Abstract image of a woman sitting on books and studying on a laptop

The Limits Of Syntactic Variation Theresa Biberauer

K
keleshamler

The Limits Of Syntactic Variation Theresa Biberauer The Limits Of Syntactic Variation Theresa Biberauer The Limits Of Syntactic Variation Theresa Biberauer

Education
1 of 81
Download to read offline
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
The Limits Of Syntactic Variation Theresa Biberauer download https://ebookbell.com/product/the-limits-of-syntactic-variation- theresa-biberauer-1645896 Explore and download more ebooks at ebookbell.com
Here are some recommended products that we believe you will be interested in. You can click the link to download. Syntactic Features And The Limits Of Syntactic Change Jhannes Gsli Jnsson And Thrhallur Eythrsson https://ebookbell.com/product/syntactic-features-and-the-limits-of- syntactic-change-jhannes-gsli-jnsson-and-thrhallur-eythrsson-30396290 The Poverty Of Conceptual Truth Kants Analyticsynthetic Distinction And The Limits Of Metaphysics 1st Edition R Lanier Anderson https://ebookbell.com/product/the-poverty-of-conceptual-truth-kants- analyticsynthetic-distinction-and-the-limits-of-metaphysics-1st- edition-r-lanier-anderson-5852594 The Limits Of Ancient Biography Brian Mcging Judith Mossman https://ebookbell.com/product/the-limits-of-ancient-biography-brian- mcging-judith-mossman-46469680 The Limits Of The World Andrew Raymond Drennan https://ebookbell.com/product/the-limits-of-the-world-andrew-raymond- drennan-46571530
The Limits Of Scientific Reason Habermas Foucault And Science As A Social Institution John D Mcintyre https://ebookbell.com/product/the-limits-of-scientific-reason- habermas-foucault-and-science-as-a-social-institution-john-d- mcintyre-46707938 The Limits Of Exactitude In Greek Roman And Byzantine Literature And Textual Transmission 1st Edition Franco Montanari And Antonios Rengakos https://ebookbell.com/product/the-limits-of-exactitude-in-greek-roman- and-byzantine-literature-and-textual-transmission-1st-edition-franco- montanari-and-antonios-rengakos-47238282 The Limits Of Regulation A Critical Analysis Of Capitalist Development Stavros Mavroudeas https://ebookbell.com/product/the-limits-of-regulation-a-critical- analysis-of-capitalist-development-stavros-mavroudeas-47252632 The Limits Of The British Empire John Dee Ken Macmillan Editor https://ebookbell.com/product/the-limits-of-the-british-empire-john- dee-ken-macmillan-editor-47288444 The Limits Of Europe Daniel C Thomas https://ebookbell.com/product/the-limits-of-europe-daniel-c- thomas-47440488
The Limits Of Syntactic Variation Theresa Biberauer
The Limits Of Syntactic Variation Theresa Biberauer
The Limits of Syntactic Variation
Volume 132 The Limits of Syntactic Variation Edited by Theresa Biberauer General Editors Werner Abraham University of Vienna / Rijksuniversiteit Groningen Elly van Gelderen Arizona State University Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective. Linguistik Aktuell/Linguistics Today (LA) Advisory Editorial Board Cedric Boeckx Harvard University Guglielmo Cinque University of Venice Günther Grewendorf J.W. Goethe-University, Frankfurt Liliane Haegeman University of Lille, France Hubert Haider University of Salzburg Christer Platzack University of Lund Ian Roberts Cambridge University Lisa deMena Travis McGill University Sten Vikner University of Aarhus C. Jan-Wouter Zwart University of Groningen
The Limits of Syntactic Variation Edited by Theresa Biberauer University of Cambridge and Stellenbosch University John Benjamins Publishing Company Amsterdam/Philadelphia
Library of Congress Cataloging-in-Publication Data The limits of syntactic variation / edited by Theresa Biberauer.        p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166-0829 ; v. 132) Includes bibliographical references and index. 1. Grammar, Comparative and general--Syntax. 2. Principles and parameters (Linguistics) 3. Language and languages--Variation. 4. Language acquisition. I. Biberauer, Theresa. P295.L56    2008 415--dc22                                                            2008024028 isbn 978 90 272 5515 0 (Hb; alk. paper) © 2008 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984. 8 TM
Preface and Acknowledgements vii Introduction 1 Part I. The locus of (parametric) variation Parametric versus functional explanations of syntactic universals 75 Martin Haspelmath Three fundamental issues in parametric linguistics 109 Chiara Gianollo, Cristina Guardiano Giuseppe Longobardi On the syntactic flexibility of formal features 143 Hedde Zeijlstra Expletives, datives, and the tension between morphology and syntax 175 Richard S. Kayne Mapping a Parochial Lexicon onto a universal semantics 219 Gillian Ramchand Peter Svenonius Aspect matters in the middle 247 Marika Lekakou Part II. A classic parameter revisited: The null subject parameter The null subject parameter and correlating properties: The case of Creole languages 271 Marco Nicolis The Case-F valuation parameter in Romance 295 Gerardo Fernández-Salgueiro Silent arguments without pro: The case of Basque 311 Maia Duguine Case morphology and radical pro-drop 331 Ad Neeleman Kriszta Szendrői Table of contents
Part III. Parametric clustering The macroparameter in a microparametric world 351 Mark C. Baker Topic prominence and null subjects 375 Marcello Modesto Non-configurationality: Free word order and argument drop in Turkish 411 Balkiz Öztürk Diachronic stability and feature interpretability 441 E. Phoevos Panagiotidis Part IV. The acquisition of parameters Can children tell us anything we did not know about parameter clustering? 459 Larisa Avram Martine Coene Parameter setting and input reduction 483 Arnold Evers Jacqueline van Kampen Index 517  Table of contents
Preface and Acknowledgements The contributions in this volume represent a selection of the papers that were originally accepted for presentation at the conference on The Structure of Para- metric Variation which was held at the University of Newcastle Upon Tyne in September 2005 under the auspices of the Arts and Humanities Research Coun- cil (AHRC)-funded project on “Null Subjects and the structure of parametric variation” (AR14458). That conference also included three papers which could unfortunately not be included in the present volume owing to prior publication commitments. These are: • David Adger – Variation without parameters; • Sjef Barbiers – Word-order alternations in three-verb clusters as non-parametric variation; and • Geoff Horrocks Melita Stavrou – Parametric variation in the lexicalisation of semantic properties: The role of grammatical aspect in shifting a verb’s basic Aktionsart My sincerest thanks to all the contributors for the focused, co-operative and ex- tremely patient manner in which they approached the various stages of the publi­ cation process. Grateful thanks also to the reviewers for their valuable input – in alphabetical order: Karlos Arregi, Anders Holmberg, John A. Hawkins, Kasia Jaszczolt, Chris Johns, Jaklin Kornfilt, Ian Roberts, Michelle Sheehan, Ianthi Tsimpli and David Willis. And to Vicki Hart and Glenda Newton for their careful proof- reading at different stages of the process. Additional thanks to Glenda for her help with the Index. Further thanks to the Cambridge students who showed such a lively interest in many of the papers in this volume, particularly – once again in alphabetical order – Bob Davies, Luke Donnan, Emma Game, Nadine Hitchon, Iain Mobbs and Neil Myler. And, last but not least, my thanks to Kees Vaes and Martine van Marsbergen at Benjamins for their patience, support and guidance.
The Limits Of Syntactic Variation Theresa Biberauer
Introduction 1. Introduction Martin Joos’s famous statement that “(l)anguages can differ from each other without limit and in unpredictable ways” (Joos 1957: 96) is often misrepresented as an assertion reflecting the thinking current during the first half of the 20th century.1 In reality, even the American descriptivist Franz Boas, to whom this view is ascribed, did not hold this view. “[I]n all the languages of the world, the number of processes which are utilized to express the relations of terms is limited”, Boas wrote in the introduction to the Handbook of American Indian Languages (1911: 67). More generally, there is also no evidence that other American de- scriptivists (Sapir, Bloomfield, etc.) held the extreme view expressed by Joos: these linguists appear to have been far less committed to the notion of boundless variation in human languages than to the belief that it was simply too early to speculate about the possible limits on this variation, given how little was known at the time about non-Indo-European languages. Thus Bloomfield (1933: 20), for example, observes that: [t]he fact that some features are … widespread is worthy of notice and calls for explanation; when we have adequate data about many languages, we shall have to return to the problem of general grammar and to explain these similarities and divergences and he also writes of: [l]ooking forward to a larger synthesis, a General Grammar, which will register similarities between languages (Bloomfield 1934: 46).2 In the 50 years since Joos’s statement, “general grammar” has, in one form or another, been very much at the forefront of linguistic research, with researchers in various frameworks having become interested in language universals and also the associated question of what limits human languages are subject to. Thus the work . Cf. Thomas (2002) for extensive discussion of this misinterpretation. . Cf. Roberts (2007b) for further discussion of the extent to which linguists of the late 19th and early 20th centuries were interested in comparative issues of this sort.
 The limits of syntactic variation of Greenberg and Chomsky and their followers has always been guided by the fundamental idea that the very striking and seemingly limitless superficial varia- tion between languages in fact conceals “uniformities of universal scope” (Green- berg, Osgood Jenkins 1961/1963: xv). Furthermore, while the manner in which typologists and generativists have approached the search for these “uniformities”, or “universals” as they are more commonly known, has been very different, there is today agreement between those in the two camps that it is indeed the case that the variation exhibited by the world’s languages is not random. In particular, there is agreement that the syntactic variation across languages is neither “without limit” nor “unpredictable”, and it is this aspect of linguistic variation that is the focus of this volume. 2. The non-randomness of syntactic variation Arguably the most oft-cited illustration of the non-arbitrariness of syntactic varia- tion has its origins in Greenberg’s pioneering work on word-order universals (cf. Greenberg 1963). This work brought to light the surprising fact that seemingly unrelated word-order patternings tend to correlate3 – S(ubject)-O(bject)-V(erb), postverbal auxiliaries and postpositions, for example; VSO, preverbal auxiliaries and prepositions; and so on. Crucially, it also introduced the idea that one might want to think of language types defined in syntactic terms. Thus Greenberg himself distinguished SVO, SOV and VSO languages and proposed that these might be viewed as the “basic word-order types”, with a language’s “type” determining the patterning of various pairs of elements, e.g., V-Auxiliary and N-Adposition order, as noted above, and also Adjective-Noun, Genitive-Noun, and so on. Sub- sequent work, notably that by Lehmann (1973); Venneman (1974) and Hawkins (1983), refined not only Greenberg’s word-order typology, but also the notion of “harmony” that emerged from it. Thus the well-known generalizations in (1) became established, along with the idea that so-called “harmonic” languages, which conform to these correlations, could be distinguished from so-called “disharmonic” languages, which do not: (1) VO: Aux-V, Prepositions, Noun-Adjective, Noun-Genitive, Noun-Relative OV: V-Aux, Postpositions, Adjective-Noun, Genitive-Noun, Relative-Noun . In itself, this was not an entirely new observation: as Greenberg (1963: 104, note 4) himself observes, researchers such as Lepsius (1880) and in particular Schmidt (1926) had previously drawn attention to a subset of the word-order correlations identified by Greenberg.
Introduction  Later work, i.a. by Tomlin (1986); Dryer (1992) and by the contributors to the World Atlas of Language Structures (WALS, cf. Haspelmath et al. 2005), has served to reveal the extent to which the Greenbergian language types are in fact instanti- ated in the world’s languages. Certain correlations are shown to constitute very strong correlations, the OV/VO and adposition one being a case in point, as (2) below shows: (2) Correlation between OV/VO ordering and adpositional structure in the 892 languages surveyed in WALS (cf. Baker, this volume) OV Postposition 427 (47.87%) VO Preposition 417 (46.75%) OV Prepostion 10 (1.12%) VO Postposition 38 (4.26%) Thus 94.5% of the languages surveyed in the WALS are “harmonic” in respect of their OV/VO and adposition structure. As Dryer (1992, note 17) notes, however: The general conclusion is that completely consistent [i.e., harmonic – TB] types represent a minority, but that the majority of inconsistencies among inconsistent languages can be attributed to a small number of pairs of elements for which there is a skewed distribution, such as the general preferences for NRel (ative) order. Viewed across the board, then, it is not realistic to think about the Greenber- gian word-order-based types in absolute terms – they do not entail absolute universals, but rather statistical universals or universal tendencies (cf. i.a. Green- berg 1963; Hawkins 1983).Worth noting, however, is Dryer’s observation that the limited number of fully harmonic languages is largely due to the existence of a fixed set of harmony-disrupting element-pairs which tend to surface in only one of the two possible orderings relative to each other. What we are left with, then, is the impression that, even though harmonic languages are in the minority, there is a significant amount of non-randomness in the domain of word order, non-randomness which even extends to the domain of exceptions to the observed regularities. And the same also appears to be true in a range of other syntactic and non-syntactic domains (cf. the Universals Archive at http:// typo.uni-konstanz.de/archive/intro/index.php for an inventory of the currently more than 2000 typological universals recorded to date). The question, of course, is why?
 The limits of syntactic variation 3. Accounting for limits on syntactic variation 3.1 Two different approaches: Functionalism vs. formalism4 As Haspelmath (this volume) notes, typologists have tended to follow a function- alist line in attempting to answer the question of why languages should display the structural parallels that they do. More specifically, researchers working within the typological framework generally appeal to communicative and/or processing con- siderations in order to account for the structural skewings that have been observed in the world’s languages (cf. also Hawkins 1988, 2004 and Newmeyer 1998, 2005a for extensive discussion). Thus, for example, Hawkins (2004: 103) proposes that the following parsing efficiency principle plays a central role in accounting for the word-order correlations in (1) and (2): (3) Minimize Domains (MiD) The human processor prefers to minimize the connected sequences of linguistic forms and their conventionally associated syntactic and semantic properties in which relations of combination and/or dependency are processed. The degree of this preference is proportional to the number of relations whose domains can be minimized in competing sequences, and to the extent of the minimization difference in each domain. In terms of (3), then, the adjacency of V and P and P and N that is ensured in con- sistently head-initial and consistently head-final languages (head-initial: [VP V [PP P [NP N …]]] and head-final: [VP [PP [NP … N] P] V]]) minimizes the domain within which the elements making up VP and PP (V and PP and P and NP respectively) will be recognized. The fact that “disharmonic” patterns generally are not as common as “harmonic” ones is thus ascribed to the added processing burden that they entail. In contrast to typologists, generativists in general mostly seek to understand crosslinguistic regularities and gaps by relating these to the influence of an in- nately specified and formally oriented linguistic capacity, Universal Grammar (UG), on which, more below. It is worth noting, however, that it is not the case that . As Newmeyer (1998: 7–8) notes, formal is an unhelpfully ambiguous term, meaning both “pertaining to grammatical form (as opposed to meaning and/or use)” and “stated in math- ematically precise terms as in formalized”. Clearly, there are formal approaches to language that are only formal in the former sense (cf. Saussurean structuralism), but these will not concern us here. For the purposes of the present discussion, then, formal is to be understood as implying both of the above meanings in the manner generally assumed by generativists. Worth noting in connection with formal in the second sense is that it does not generally in syntactic theory mean anything as strong as “axiomatised” or even “fully formalized”, as it does in, for example, logic, computational work or formal semantics.
Introduction  generativists uniformly reject the possibility that functional and, particularly, pro- cessing considerations may determine the distribution of syntactic patterns, al- though this has undoubtedly been the most common trend to date (see below). Thus, for example, Ackema Neeleman (2002) argue that the very striking asymmetry between leftward and rightward movement observed in the world’s languages – in terms of which leftward movement is vastly more common than rightward move- ment and additionally appears to be unbounded in a way that rightward movement is not – can in fact best be understood in parsing terms.5 More specifically, they, like non-generativist psycholinguists (cf. i.a. Just Carpenter 1992; Gibson 1998), propose that processing of syntactic representations involves a left-to-right parsing procedure which involves both processing and storage: since it is presumably more economical to assign already-parsed structure to short-term memory as soon as this becomes possible rather than to wait until the entire structure has been parsed, it is assumed that already-parsed structure is “closed off” at appropriate points during the parsing process. This, then, renders its internal structure “invisible” to the parser. If, as is usually thought, the parser employs a filler-driven strategy in dealing with movement dependencies (cf. Clifton Frazier 1989; Frazier Flores D’Arcais 1989), it will only postulate a trace after it has encountered a moved element. Clearly, this is always the case when we are dealing with leftward movement (cf. (4a)), but the opposite is true wherever rightward movement has taken place (cf. (4b)): (4) a. wh-element/topicalized dp, etc. ….. twh-element/tDP, etc. b. twh-element/tDP, etc ……… wh-element/topicalized dp, etc. As (4b) illustrates, rightward movement is problematic within filler-driven, interim storage-oriented approaches to parsing in that it requires post hoc trace insertion in a partially analysed string: left-to-right parsing must be suspended so that a trace can be inserted at an appropriate place in an already-parsed piece of struc- ture located to the left of the point the parser had reached. If it is indeed the case that such structure has been “closed off” and therefore lacks internal structure, we might expect this kind of trace-insertion to be impossible or at least costly. And the same mode of explanation is available to account for the locality restric- tions on rightward movement. As Ross (1967) first observed, this is always clause- bounded (cf. the so-called Right Roof Constraint). On the natural assumption that distinct clauses constitute distinct parsing domains, we would expect it to be . Other well-known examples of generativists appealing to parsing to account for formal patterns include: Bever Langendoen (1971) in their work on the factors determining the availability of relative pronoun omission in English, Fodor (1978, 1984) in her work on nested dependencies, Berwick Weinberg (1984) in their work on Subjacency, and the work of Janet Dean Fodor Lyn Frazier more generally.
 The limits of syntactic variation impossible for rightward movement to be unbounded in the manner that leftward movement appears to be. This is because unbounded leftward movement is unprob- lematic if long-distance movement is indeed, as has generally been assumed since Chomsky (1973), successive-cyclic: if elements move leftward successive-cyclically, each clause will be headed by a trace, facilitating easy tracking of the filler-gap depen- dency,evenwhereearlierclauseshavealreadybeen“closedoff”–thetraceatthestart of every parsing domain alerts the parser to the fact that its search is not yet complete. As far as the parser is concerned, then, filler-gap dependencies can be established in the usual local manner wherever movement is successive-cyclic and leftward. The same is not, however, true for rightward movement beyond the clause to which the moved element belongs. Even if this movement is successive-cyclic, it will still entail the requirement that previously “closed off” material needs to be “reopened”. This is because traces in clauses preceding that to which the element has moved cannot be postulated until the parser has encountered the gap located at the rightmost edge of the final clause. Long-distance rightward movement therefore entails multiple “reopenings” of the type required in monoclausal rightward-movement structures. Ackema Neeleman (2002) explicitly indicate why a purely formal account of these phenomena – for example, one appealing to Kayne’s (1994) Antisymmetry proposal, in terms of which all movement is necessarily leftwards because of structural con- straints (specifiers all being left-adjoined, with the result that upward movement can only be to the left) – fares less well than a parsing-oriented approach (cf. also Abels Neeleman 2007). The details of this argument fall beyond the scope of this introduction; the point here is simply that it is not always the case that formal- ists exclusively seek to understand syntactic distributions and the factors deter- mining them in purely formal terms. And, more generally, there is, as Newmeyer (1998: 21) rightly notes, “nothing in the generative program that demands that all typological facts be attributed to the setting of innately specified parameters” (cf. also Newmeyer 1998, chapter 3). Furthermore, there is also nothing in the generative re- search programme that excludes the possibility that innately prespecified principles may ultimately be functionally grounded (cf. Müller 2007a, numerous contribu- tions in Good (2008), in particular, Kiparsky 2008, who proposes criteria designed to distinguish UG-based universals from those which have historical – and thus potentially ultimately functional – explanations, and Mobbs 2008, who suggests a “third factor” reinterpretation of Hawkins’s performance principles6). We return to these matters below. . Kiparsky (ibid.3) explicitly refers to functional principles which “might have become biologized within UG itself”. Cf. Section 3.4 for discussion of “third factor” considerations within current Chomskyan theory which may also include functional considerations.
Introduction  First, however, let us consider the UG-based approach and, in particular, that which generativists have espoused since the introduction of the Principles and Parameters model in the late 1970s/early 1980s. 3.2 Introducing the Principles and Parameters Framework Universal Grammar (UG) is the innately specified mental endowment which is assumed to be uniquely shared by all humans. In terms of the current general model of UG, the Principles and Parameters (PP) model first outlined in Chom- sky’s 1979 Pisa Lectures (published in revised form as Chomsky 1981), UG con- sists of invariant principles and unset parameters whose values must be set on the basis of the linguistic input available to language-acquiring children (the primary linguistic data or PLD) – cf. (5): (5) UG (principles unset parameters) + PLD → Adult grammar S0 Ss (the initial state) (the steady state) Together, the immutable principles and the initially unset parameters (i.e., S0 in (5)) are taken to determine the form of a possible human grammar (i.e., Ss in (5)), where “grammar” is understood as the abstract mental system, including specific parametric settings, – the I-language in Chomsky’s terms (cf. Chomsky 1986 et seq.) – on the basis of which actual utterances are produced. For Chom- skyans, then, UG plays a role not only in language acquisition, but also in defining the notion “possible language”, both synchronically and diachronically. And, crucially, the limits it is assumed to impose on “possible languages” and, in particular, the syntactic variation that they may exhibit, are conceived of in formal terms. In PP’s first incarnation during the Government and Binding (GB) period, a range of principles and parameters were proposed, with the X-bar schema, the Theta Criterion, the Extended Projection Principle (EPP) and Subjacency/Rela- tivised Minimality numbering among the most frequently cited principles and the Head Parameter (HP) and the Null Subject Parameter (NSP) undoubtedly being the most well-known parameters. In terms of the former, there are universal con- straints on the way languages construct phrases, the way in which thematic roles like Agent, Patient and Goal are assigned, the positioning of subjects, and the dis- tance over which elements can undergo movement: no languages should violate these principles, with the result that no variation in these domains is expected (but see below). Parameters, on the other hand, do permit variation, although it is variation of a highly restricted kind. Consider, by way of illustration, the HP and the NSP in (6–7).
 The limits of syntactic variation (6) Head Parameter (cf. Travis 1984) Heads X precede/follow their complements X' a. b. X Comp  → - X' X Comp  → - (7) Null Subject Parameter (cf. Rizzi 1982, 1986) Subjects can be pro, where pro is a [+pronominal, -anaphoric] entity which is licensed by X0 y where X0 y is a subset of the heads in a given language. The gram- matical specification of pro is recovered from the features of the licensing head. e.g. I' ... I [+pronominal] → I licenses pro Spec pro IP a. b. I' ... I [–pronominal] → I does not license pro Spec overt pronoun IP As shown above, both the HP and the NSP define an innately specified choice- point, the HP in respect of head-complement directionality and the NSP in respect of the availability of pronominal null subjects. Just like principles, then, parameters impose universal restrictions – they determine the set of structural choices avail- able to languages in various domains, thereby ensuring that all languages make choices about the same set of choice-points. Because languages do not make the same choices in each case, variation results. Thus the PP approach opens up the possibility of having a model of grammar which entails a finite and potentially relatively simple UG, but which nevertheless leads us to expect what might on the surface seem like considerable crosslinguistic variation – consider the striking dif- ferences between languages that opt for different settings of the HP, for example. That is, the PP approach, unlike those that went before it, facilitates a resolution of the tension between the search for both descriptive and explanatory adequacy, being able to account not only for the (apparent) diversity of syntactic structures, but also for how these could have been successfully acquired given the constraints under which language acquisition takes place (cf. Chomsky 2004b for overview discussion, and also Section 7 below). In accordance with the schema in (5), the values assigned to the various parametric choice-points are assumed to be determined on the basis of specific
Introduction  PLD, signalling the appropriate setting. With the HP, at least in the case of the consistently head-initial/head-final languages (i.e., the ones exhibiting harmonic order in respect of V and its complement, Aux and V(P), Adpositions and their complements, Nouns and their complements, so on7), it is not hard to see what would be an appropriate cue (cf. Lightfoot 1999 and Dresher 1999, 2003 for discussion of the notion “cue”): any head + complement-containing XP will provide the necessary cue. Thus, for example, any one of the following Japanese structures would enable the learner to set the HP to head-final: (8) a. Sensei-wa [Taro-o] sikata teacher-top Taro-acc scolded ‘The teacher scolded Taro.’ → Objects precedes V b. [Nihon] kara Japan from ‘from Japan’ → Complement precedes adposition c. John-ga [Mary- to renaisite] iru John-nom Mary-within.love is ‘John is in love with Mary.’ → V precedes Auxiliary d. [John-ga Mary-o nikundeiro] to John-nom Mary-acc be-hating that ‘that John hated Mary’ → Clause precedes Complementizer8 As noted above, however, fully harmonic languages represent the minority of the world’s languages and many languages, including very well studied ones like German, exhibit mixed headedness, with certain phrases seeming to be . It is worth noting that the HP does not, as is often thought, determine the relative ordering of Noun-Adjective as adjectives are not nominal complements. Similarly, the HP has nothing to say about those Noun-Genitive and Noun-Relative pairings which do not involve a comple- mentation relation. The fact that, for example, English exhibits Adjective-Noun rather than the Noun-Adjective order expected in a harmonic head-initial language (cf. (1) above) does not, therefore, reflect a shortcoming of the HP. . If, as Lightfoot (1989, 1991) has suggested, learners do indeed only pay attention to matrix clauses at the stage at which parameter setting is underway, (8d) may play no role in the setting process. As it is only necessary for the child to establish the directionality of a single XP (say, the VP) in order to set the HP, this does not, however, have to pose particular problems. The same is obviously not true for “mixed” languages, though (see main text).
 The limits of syntactic variation consistently head-initial (e.g., CP and DP in German), while others appear to be consistently head-final (VP and TP in German). This is illustrated in (9): (9) a. … daß er das Buch von Chomsky unter dem Tisch gefunden hat that he the book from Chomsky under the table found has ‘… that he found the book by Chomsky under the table’ b. [CP daß [TP eri [VP ti [DP das Buch von Chomsky] [PP unter dem Tisch] gefunden] hat]] i.e., C, D, N and P precede their complements; V and T follow them. There also appear to be quite a few languages which exhibit head-directionality of one kind in the clausal domain and the opposite in nominals (cf. Thai, Khmer and Burmese as discussed in Simpson 2005, Bangla as discussed in Bhattacharya 1999, Marathi as discussed in Nayudu 2008, and Gungbe as discussed in Aboh 2004, etc.). This is shown for Gungbe in (10): (10) dàwé l� x� kὲkέ man det bought.perf bicycle ‘The man bought a bicycle.’ (from Aboh 2004: 32) Worse still, there are further cases where individual lexical items appear to determine the headedness of the phrases in which they surface. Thus German adpositions are generally head-initial (cf. von and unter in (9)), but there are a few exceptions like zufolge (“according to”) and halber (“for the sake of, because of”) and the “shadow prepositions” (cf. Noonan 2003) herauf (“up”) and drin (“in”) which only surface finally. The “mixed” complementiser class in languages like Bengali, Marathi and Assamese, in which certain complementisers are head-initial while others are head-final, represents another problematic case (cf. Bayer 1999, 2001). Furthermore, there are lexical items like German entlang (“along(side)/beside”) and Afrikaans in (“in”/“into”) which occur both initially and finally, but where positioning determines other morphosyntactic and/or semantic properties – in the case of entlang, which case its complement takes (genitive where entlang is initial and accusative where it is final) and in the case of in, precisely how it is interpreted (as locative “in” when it is initial and as directional “into” when it is final).9 The examples in (11–12) illustrate: (11) a. Sie ist entlang des Flusses gefahren [German] she is along the-gen river driven ‘She drove along the river.’ . We leave aside here the potential further complication represented by so-called circumposi- tions (cf. Dutch naar … toe – “to … to”) which, superficially at least, appear to consist of nested adpositional phrases, with a head-initial phrase embedded under a (superficially) head-final one (cf. i.a. van Riemsdijk 1978, 1990; Koopman 1997; Zeller 2001; and den Dikken forthcoming).
Introduction  b. Sie ist den Fluss entlang gefahren she is the-acc river along driven ‘She drove along the river.’ (12) a. Hy loop in die bos [Afrikaans] he walk in the forest ‘He is walking in the forest.’ b. Hy loop die bos in he walk the forest in ‘He is walking into the forest.’ In at least some cases, then, head-directionality appears to play a lexical disam- biguation role.10 Finally, there are elements like the prepositional complement-taking adjective stolz (“proud”) in German and its counterpart in Dutch (trots) and adpositions like ilman (“without”) and kohti (“towards”) in Finnish which can surface either before or after their complements without either morphological or semantic con- sequences. Consider (13–14) in this connection: (13) a. Die Frau ist stolz auf sich [German] the woman is proud of self ‘The woman is proud of herself.’ b. Die Frau ist auf sich stolz the woman is of self proud ‘The woman is proud of herself.’ (14) a. kohti Stalinin kuvaa [Finnish] towards Stalin-gen picture ‘towards a picture of Stalin’ b. Stalinin kuvaa kohti Stalin-gen picture towards ‘towards a picture of Stalin’ . If it is indeed correct to think of lexically-specified head-directionality, one might want to appeal to Elsewhere considerations (cf. Kiparsky 1973) in accounting for the word-order properties of structures containing these elements. More specifically, it could conceivably be the case that the HP is appropriately set for a specific set of syntactic categories, but that lexical items associated with independent (i.e., irregular) linearisation instructions are linearised in accordance with their more specific specification. If this approach is extended to cases permit- ting both linearisation options (e.g., the German and Dutch adjectives and Finnish adpositions mentioned in the main text), we would need to postulate two distinct lexical entries which are phonologically and semantically identical, but which differ in terms of their linearisation in- structions (cf. Adger 2006 for an optionality proposal along these lines).
 The limits of syntactic variation Taken together, these facts raise serious questions about how children success- fully identify the appropriate cues to set the HP (cf. Dresher 1999, 2003; Fodor 1998a, 1998b, 2001; Fodor Sakas 2005; Sakas Fodor 2001, and Lightfoot 1999 for recent discussion of the general problem of cue identification, and also the contributions by Gianollo, Guardiano Longobardi (henceforth: GGL), Zeijlstra, Avram Coene and Evers van Kampen in this volume). And they also suggest that the HP may be a rather different type of parameter to the NSP, which is usually (though see Sections 3.4 and 5 below) thought to involve a binary choice which is made once and for all for the language as a whole. Specifically, one might view the HP as a parameter which can operate in a more or less “myopic” fashion as outlined below: a. “blindly”, being set once for all categories, thereby delivering harmonic languages; b. category-sensitively, being consistently set in one way for a fixed set of syntactic categories and another way for another fixed set; or c. lexical item-sensitively, being set individually for specific lexical items (although see note 10). Despite this fact, the general assumption about the HP during the GB era was that it was, universally, a parameter specifically associated with the phrase-structure (X-bar) module of the grammar. Turning to the NSP: this parameter clearly involves two components, namely (a) that a null-subject language (henceforth: NSL) include in its pronoun inventory the phonologically null, lexically unspecified pronoun pro and (b) that it include in its inventory of functional categories subject-related ones (e.g., I in the clausal domain and/or D in the nominal domain) which are capable of assigning pro the appropriate φ-features (here: person and number). Non-NSLs, by contrast, lack one or both of these properties. Traditionally, the assumption was that the NSP was set to its positive value – in terms of which children postulate the availability of pro and the appropriate licensing head(s) – on the basis of PLD such as (15) featuring unrealised referential pronominal subjects as in (15a) and unrealised expletive pronominal subjects as in (15b) and (15c): (15) a. Habl- a español [Spanish] speak-3sg Spanish ‘(S)he speaks Spanish.’ b. Piove [Italian] rain-3sg ‘It is raining.’
Introduction  c. Sono venute delle ragazze [Italian] be-3pl come-f.pl some girls ‘There came some girls.’ Just as was the case with the HP, the NSP was assumed to determine the (un)avail- ability of a range of structures not limited to those which, like those in (15), were generally assumed to constitute the actual cue required to trigger the appropriate setting of the parameter. Thus the NSP sanctioned not only null-subject structures such as those in (15), but also structures like the following: (16) a. Ha telefonato Gianni [Italian] have-3sg telephoned Gianni ‘Gianni has called.’ b. Pjos ipes oti egrapse afto to vivlio? [Greek] who said-2sg that wrote this the book ‘Who did you say wrote this book?’ (16a) represents a so-called Free Inversion structure, in which the overt subject follows the lexical verb rather than occupying the canonical subject position, Spec-IP. (16b), in turn, illustrates the absence of so-called that-trace effects in NSLs: unlike in non-NSLs like English, wh-subjects can be extracted across overtly realised complementisers without resulting in an ungrammatical structure (con- trast the English counterpart of (16b), *Who did you say that wrote this book?). Neither of the structures in (16) at first sight seems to relate to the availability of null subjects, but Rizzi’s (1982, 1986) characterization of the NSP links all of these phenomena and additionally also the fact that Romance NSLs exhibit “rich” agree- ment morphology. The reasoning is as follows: the availability of pro facilitates the existence of null-subject structures such as those in (15), with the “rich” agreement morphology here and elsewhere reflecting the fact that the language has the kind of I required to supply pro with the φ-features it intrinsically lacks. The availability of pro also enables a subject to occur in inverted position as in (16a) since pro is available to satisfy the Extended Projection Principle (EPP), i.e., the requirement that every clause contain a subject located in Spec-IP (cf. Chomsky 1981, 1982). Finally, pro also facilitates the circumvention of that-trace violations, as in (16b), by allowing wh-subjects to be extracted from a lower position, while pro itself once again satisfies the EPP. Clearly, then, the NSP as formulated by Rizzi leads us to expect superfi- cially unrelated phenomena to cluster together.11 And if it is indeed correctly . Cf. Roberts (2007c) for discussion of three further phenomena that have been proposed to be connected to the NSP: clitic climbing, infinitive-clitic ordering and realisation of an arbitrary subject as a clitic.
 The limits of syntactic variation formulated, we would expect it to impose a universal limit on the types of struc- tures available both in NSLs and in non-NSLs: the former would all be expected to exhibit all of the structures facilitated by pro and a suitable I, while the latter would all be expected to systematically lack these structures. Looking beyond the NSP, this was also the expectation for GB parameters more generally since parameters with a rich deductive structure simultaneously facilitate an account of systematic syntactic correspondences across languages and an account of how it might be possible for children to acquire these systematic correspondences with relative ease. If we consider the NSP, for example, it is arguably the case that unrealised pronominal subjects like those in (15) represent a salient and easily detectable component of the PLD. By contrast, that-trace circumvention structures such as those in (16b) are likely to feature far less frequently in the input and, superficially at least, these structures seem a great deal more complex than unrealised subject- containing ones like those in (15). Even more striking in this connection is the scenario faced by children acquiring non-NSLs: in this case, the input does not contain any positive evidence that that-trace structures are unavailable, with the result that one would expect that children would have to be explicitly told that subject wh-extraction is not possible in the presence of an overt complementiser, unlike extraction of other wh-elements. Such direct negative evidence is, however, not usually thought to play a significant role in first language acquisition (cf. i.a. Bowerman 1988; Marcus 1993) and, in the case at hand, it is furthermore not at all clear that adult native- speakers are actually consciously aware of the existence of this structural gap – it is certainly not one that is noted in prescriptive grammars or explicitly taught in English grammar classes. In both the NSL and the non-NSL case, then, a question arises as to how native speakers are able to acquire structures that are infrequent or even unattested in the input. If we can, however, assume in the NSL case that unrealised pronominal sub- jects alert children to the other NSP-related properties, the salience and frequency discrepancies noted above need not pose an insurmountable obstacle. Similarly, if we can assume that the obligatoriness of overtly realised subjects (and particularly, overt expletives) serves the same function for children acquiring a non-NSL, we can understand the successful acquisition of that-trace phenomena as the consequence of the fact that the unavailability of wh-subject extraction is simply part of the cluster of NSP-related properties associated with the negative setting of this parameter. More generally, if innately specified parameters do indeed determine the way in which certain types of syntactic phenomena cluster, we would expect, firstly, that the world’s languages (past, present and future) will exhibit certain clusters of properties and not others and, secondly, that language acquisition will proceed in a way which reflects the (incremental) acquisition of relevant parametric settings
Introduction  and that will therefore never feature stages at which children correctly use some of the structures associated with a given parameter, but also others that are associated with the opposite setting of that same parameter. That is, parameters as they are conceived of in the PP framework aim to contribute to an explanation not only of the limits of syntactic variation, but also of the limits on language acquisition and, by extension, language change. 3.3 Problems with GB principles and parameters As is now well known, many of the principles and parameters proposed during the first phase of the PP era did not ultimately prove either descriptively or explana- torily adequate. Consider, for example, the fact, already noted above, that, while the HP facilitates an elegant account of languages exhibiting “harmonic” word orders, it is evidently too simplistic to assume that head-directionality is universally fixed once-and-for-all on the basis of the simple binary choice in (6): the empirical evi- dence clearly points to a much more complex situation. Similarly, the NSP as formulated above leads to various expectations: that NSLs will be morphologically rich,12 that morphologically poor languages will not be NSLs, that the properties governed by the NSP (cf. (15–16) above) will always cluster together, and so on. As work as early as Gilligan (1987) has shown, however, these predictions are not borne out: numerous morphologically rich languages are not NSLs (e.g., Icelandic and Russian), there are various morphologically poor languages that are NSLs (e.g., Chinese and Cheke Holo), and there are also other respects in which the proposed correlations do not always hold (cf. Bantu lan- guages exhibiting referential null subjects, but not Free Inversion and non-NSLs like Icelandic exhibiting the no that-trace effect property of NSLs). Some of these failed predictions have been ascribed to parameter interac- tion, with the idea being that the non-occurrence of an expected parametric reflex follows from the obscuring effects of the setting of another parameter which inter- acts with the first. Thus, for example, the non-occurrence of referential null subjects in inflectionally rich Icelandic is commonly ascribed to the positive setting of the V2 Parameter in this language – that is, since the V2 Parameter always requires an XP to occupy the preverbal position in main clauses, preverbal null subjects are . It is worth being precise here: what is specifically expected is that NSLs will have rich agreement (i.e., person/number) morphology; rich tense, mood or other verbal morphology should not be relevant to the setting of the NSP, as the crucial consideration is that the richness of the verbal agreement morphology should facilitate recovery of pro’s content (cf. Biberauer Roberts, forthcoming).
 The limits of syntactic variation ruled out in these contexts and consequently the effects of the NSP are masked. Clearly, though, this proposal only accounts for the absence of preverbal null sub- jects; it tells us nothing about why referential null subjects are also unavailable else- where and why expletive null subjects, but not their referential counterparts, may surface in postverbal position. More recently, researchers have adopted a different tack, opting instead to re-examine the morphological paradigms of languages which at first sight appear to qualify as “morphologically rich” and proposing that languages like Icelandic do, after all, systematically differ from NSLs (cf. Müller 2005, 2007b Tamburelli 2006 for two recent proposals of this type). More generally, however, the notion of “morphological richness” has proven very hard to quantify, not only in the NSP context, but also more generally (cf. Jaeggli Safir 1989; Vikner 1997; Rohrbacher 1999 for different, but ultimately unsuccessful proposals). Clearly, this is a highly problematic state of affairs for a parameter whose setting was initially thought to be at least partially cued by inflectional wealth. Looking beyond the HP and the NSP, we see that other GB parameters also encountered difficulties of various kinds. Thus, for example, in the V2 context, the discovery that V2 languages are not all “well behaved” and that languages like Icelandic and Yiddish systematically permit V2 in the presence of overtly realised complementisers first raised the question whether one can in fact think of a unitary V2 parameter (cf. Schwartz Vikner 1996 and Vikner 1995 for relevant discus- sion). The rise since the late 1980s of so-called “cartographic” approaches to clause structure, in terms of which functional structure is assumed to be a great deal more articulated than initially thought, has only underscored this question (cf. Poletto 2000; Mohr 2005 and Roberts 2005 for illustrations of articulated CP analyses of V2 phenomena). Furthermore, independently of their precise formulation, this and other head-movement-related parameters (e.g., the V-to-I Parameter, the N-to-D Parameter, etc.) would also seem to face a further conceptual question, namely whether they really meet the requirement that parameters should govern clusters of empirical phenomena: head-movement parameters are not generally thought to give rise to the sorts of superficially unrelated phenomena proposed for the NSP.13 . Viewed from a Germanic perspective, the V-to-I parameter could be argued to govern a cluster of unrelated phenomena – consider, for example, Bobaljik Thráinsson’s (1998) Split- INFL parameter, in terms of which languages exhibiting V-to-I also feature a split IP, full DP Object Shift (OS), and Transitive Expletive Constructions (TECs). This (full) cluster does not, however, seem to be more generally associated with languages which have been analysed as fea- turing V-to-I movement, e.g., the Celtic languages, the Semitic languages, which at most permit pronominal OS (cf. Shlonsky 1997, chapter 10), and the Romance languages, which also lack non-pronominal OS and TECs of the Germanic type (cf. Cornilescu 2000 for a proposal that Romanian features TECs). Strikingly, Avram and Coene, this volume, propose a very different
Introduction  Baker (this volume) dubs this the extent of variation question, and we return to it in Sections 3.4 and 3.5 below. Related to it is the question that we have already raised in connection with the HP, namely whether binary parameters can in fact hope to account for the attested range of variation within a given domain. Since this question in fact raises a number of important issues, let us consider it in a little more detail on the basis of a previously undiscussed parameter. Both during the GB era and beyond, it has frequently been proposed that the presence vs. absence of wh-movement in a language is determined by the positive vs. negative setting of a Wh-movement Parameter (cf. Fukui 1995; Cheng 1997 and Simpson 2000 for discussion). If this parameter simply offers a binary choice, we would expect to find only two types of languages: wh-movement and wh-in-situ languages, illustrated in (17): (17) a. Whati did Steve eat ti? b. Hufei chi- le shenme ne [Chinese] Hufei eat-asp thing Qwh ‘What did Hufei eat?’ In actual fact, however, some languages have been shown to exhibit a mixture of these properties, permitting both movement and in situ structures (cf. Cheng 1997). Crucially, closer investigation has revealed that some of the “mixed” lan- guages do not in fact pose an empirical challenge to the postulation of a binary Wh-movement Parameter. Some, like Egyptian Arabic, Irish, Malagasy and Zulu, can, for example, be shown to involve very different underlying struc- tures: a (pseudo-)cleft (i.e., a non-movement structure) in the (apparent) wh- “movement” case and an in situ structure in the non-movement case. This is illustrated in (18): (18) a. U- bona ubani?14 [Zulu] 2sg- see who.1a ‘Who did you see?’ b. Ng- ubani o- m- bona-yo? cop-who.1a rc.2sg-oc.1a- see -rs ‘Who is it that you see?’ (data from Sabel Zeller 2006) cluster of Split-INFL-related effects to that proposed by Bobaljik Thráinsson. At the very least, then, it does not seem that V-to-I can be viewed as correlating with a single cluster of effects. . Sabel Zeller’s abbreviations are as follows: 1a = noun class 1a, i.e., that mostly reserved for singular human entities; cop = copula; rc = relative concord/agreement; oc = object clitic; rs = relative suffix.
 The limits of syntactic variation As (18) clearly shows, the usual SVO order is maintained in monoclausal wh- interrogatives, with the wh-element surfacing in situ; it is only in the presence of a cleft-marking copula that wh-elements may appear in a moved position, and then it is clearly a relativised one, rather than the clause-initial Spec-CP position associ- ated with wh-movement languages. Similarly, wh-“movement” in Malay and Iraqi Arabic seems to be possible only in focus structures and not in unmarked interrogatives, i.e., wh-movement, where it occurs, takes place as a consequence of focus- rather than specifically wh- driven movement (cf. Ouhalla 1996). From the perspective of the Wh-movement Parameter, then, these languages are negatively specified for wh-movement and the superficial “mix” of positioning possibilities can be explained despite the avail- ability of only two mutually exclusive parametric options. Importantly, however, there do appear to be systems within which a putative Wh-movement Parameter could not be set once and for all. Duala, a Bantu language spoken in Cameroon, is a case in point: it permits both wh-movement and wh-in situ in matrix clauses, but necessarily requires wh-movement in embedded contexts (cf. Sabel 1998 for detailed discussion). Languages of this sort clearly cannot be specified as [+movement] as wh-elements are evidently not required to move (there is no (nec- essary) interpretive difference between movement and non-movement structures). Similarly,theycannotbespecifiedas[–movement]sincethereisacontext(embedded clauses) where movement is systematically required. Here, then, the Wh-Parameter does appear to face a genuine empirical challenge from optionality phenomena. Furthermore, a simple binary parameter also will not be able to account for the op- tionality exhibited by languages like Bangla, Basque, Imbabura Quechua and others which permit so-called clausal pied-piping to alternate with wh-DP raising, once again without necessary interpretive consequences (cf. Richards 2001 for overview discus- sion); and the same is true of that found in Slavic and other languages which permit Left Branch Condition/LBC-violating wh-extractions to alternate – often without interpretive consequences – with their non-LBC-violating counterparts (contrast English *Whose did he read book? [LBC violation] which is not available alongside Whose book did he read?). Additionally, a simple binary Wh-movement Parameter cannot explain why some languages (e.g., those belonging to the Slavic family) either obligatorily or optionally front all the wh-words present in multiple wh-questions, whereasothersmayonlyfrontoneofthem(cf.Bošković2002fordetaileddiscussion). Like the “size”-related optionality mentioned above, these differences have the flavour of “sub-parameters” associated with the Wh-movement Parameter initially proposed: more precisely, they would seem to be choices which only come into play following positive setting of the Wh-movement Parameter (cf. also Uriagereka 2007 for recent discussion of parametric structure in terms of parameters which differ in their “deep- ness”, delivering, in his words (p.1), “progressively deeper forms of variation”).
Introduction  From a parametric perspective, the notion that a parameter that was ini- tially viewed as a unitary Yes/No choice-point in fact turns out to initiate or be part of a sequenced network or hierarchy of related parameters need, of course, not be problematic. We return to this matter in Section 3.3 below. For the moment, it worth noting, however, that the more articulated view of pa- rameters suggested by closer examination of the wh-data also brings with it various complications which should not be overlooked. There is, for example, the matter of how many (sub-)parameters we will ultimately need to postu- late, i.e., how many innately specified choice-points UG will need to contain. Related to this is the question of whether each choice-point must be associated with its own cue (see Section 3.2 above and Section 7 below) and, if so, how one can ensure that the cues are “noticed” in the correct sequence by the language acquirer. Specific acquisitional concerns aside, there is also more generally a clear sense in which an articulated Wh-movement parameter is at odds with the original GB notion of parameters as innate entities whose setting “gives” the acquirer a range of language-specific information – some of it “deep” in the sense of not necessarily being superficially evident from the input – “for free”. This is once again a matter to which we will return in Section 7. Here we simply wish to highlight the fact that closer consideration of GB parameters like the Wh-Parameter has led to the discovery that initially proposed parameters do in some cases hold up under closer scrutiny, but that this is not always the case and, furthermore, that it sometimes appears to be necessary to adopt a more articulated view of the structure of a parameter. To conclude, let us briefly consider the relation between GB principles and parameters, and also the fate of principles proposed during the GB era. Whereas principles and parameters are presently often thought of as distinct and sepa- rate components of UG – cf. the Wh- or Head-movement parameters discussed above, which are not typically viewed as choice-points (directly) connected to any particular UG principle – the earliest conception of “parameter” entailed a close connection between the two entities: parameters were necessarily “parameterised principles” (cf. Newmeyer 2005a). Consider (19–20) in this connection: (19) A parameterised principle in the phrase structure module: a. Principle: X’ → X; Complement (where; signifies an unordered pair) b. Parameter: the Head Parameter (HP; cf. (6) above) Heads X precede/follow their complements (20) A parameterised principle in the bounding module: a. Principle: Subjacency In a structure α … [β … [γ … δ…] …] …, movement cannot apply if β and γ both belong to the class of bounding nodes.
 The limits of syntactic variation b. Parameter: the class of bounding nodes = DP and/or TP and/or CP (where bounding nodes in English = DP and TP, in Italian = DP and CP and in Russian = DP, TP and CP) As shown above, phrase structure and movement are both subject to inviolable principles (a), but the precise manner in which these principles manifest in dif- ferent languages is subject to specific points of parametric variation (b). As GB parameters do not all appear to be so directly connected to principles, it seems necessary to acknowledge that the parameters assumed during the GB era were not of a single general type: at the very least, one needs to distinguish parameters associated with principles (“parameterised principles”) from parameters whose choice-points are less directly connected to specific principles, a state of affairs which immediately raises questions about the relation between these two, the “locus” of the latter, and so on. We return to these questions in Section 3.4. The preceding discussion has shown that principles may be directly associated with parametric choice-points. This need, of course, not be the case – consider, for example, the Projection Principle and the Theta Criterion (cf. Section 3.2 above), neither of which entails a (b)-component of the type illustrated in (19–20). Subja- cency, too, was initially thought to be an invariant principle (cf. Chomsky 1973), but Rizzi’s (1982) work revealed this assumption to be incorrect. Similarly, work on the wide range of languages that Chomskyans have investigated since the early PP days has shown that certain further properties initially ascribed universally invariant status also cannot be regarded as such. Let us consider this matter on the basis of the Ex- tended Projection Principle (EPP) in terms of which every clause must have a subject located in Spec-IP (i.e., the canonical subject position; cf. Chomsky 1981, 1982). As is well known, this principle facilitates a unified account of subject and ex- pletive distribution in modern English, but necessitates sometimes quite complex assumptions in the context of the many languages that do not exhibit the same patterns of subject behaviour. Thus, for example, in languages which set the NSP to its positive setting, two assumptions need to be made, namely (a) that these lan- guages feature a particular type of I and (b) that they include in their lexical stock an element, pro, which is able to satisfy the EPP-imposed requirement that Spec-IP be projected and filled (henceforth: the EPP-requirement). In the absence of an EPP- requirement, just the (a)-assumption would be sufficient, and numerous researchers have in fact proposed that it may not be necessary to project and fill Spec-IP in all NSLs (cf. Borer 1986; Barbosa 1995, to appear, Alexiadou Anagnostopoulou 1998 on so-called “I-subject” approaches to NSLs, and see also the discussion in Section 5 below).15 In other NSLs, postulating pro and therefore also the projection . The “I-subjects” approach necessarily has various consequences for Theta Theory, which
Introduction  of Spec-IP may nevertheless be justified (cf. Sheehan 2006 and Barbosa, to appear for recent discussion and see Section 5 below for further discussion). In the Germanic context too, there are a range of subject-related phenomena which cast doubt on the universal operation of an EPP-requirement. Consider, for example, the fact that subjects in Icelandic, German and Dutch may surface in a posi- tion which is clearly lower than the V2-adjacent canonical subject position (Spec-IP). The Icelandic examples in (21) illustrate (subjects in bold; Spec-IP underlined; % in- dicates that the structure in question is ill-formed in the context in question):16 (21) a. Við settum allar bækurnar fyrir málstofuna í lestrarsalinn … We put all books-the for seminar-the in reading-room-the i. … en það hafa margar af þessum bókum verið lesnar áður … but there have many of these books been read before ‘… but many of these books have been read before’ ii. %… en það hafa ___ verið lesnar margar af þessum bókum áður … but there have been read many of these books before b. Við settum færeyskar bækur fyrir málstofuna í lestrarsalinn … We put Faroese books for seminar-the in reading-room-the i. %… en það höfðu íslenskar bækur verið lesnar í fyrra … but there had Icelandic books been read last year ii. … en það höfðu ___ verið lesnar íslenskar bækur í fyrra … but there had been read Icelandic books last year ‘… but Icelandic books had been read last year’ (from Jonas 1996: 63) As shown above, it is not always necessary for Spec-IP to be filled: where a subject constitutes new information, as in (21bii), it does not undergo raising; it is only where the subject encodes old information, as in (21ai), that raising to Spec-TP is required. Similar effects emerge in German and Dutch, where “low” and raised (i.e., Spec-TP-occupying) subjects are also said to differ in respect of information- structure-related properties (e.g., specificity; cf. Diesing 1992 and in particular Rosengren2002).IfoneistomaintainthattheselanguageshaveanEPP-requirement, it is necessary to postulate the presence of expletive pro in the case of the structures we will not go into here, but see Baker (1996, 1997) and Alexiadou Anagnostopoulou (1998) for discussion. . Note that Icelandic is a symmetric V2 language, exhibiting V2 orders in matrix and em- bedded contexts. Following Schwartz Vikner (1996) and many others, I assume that all V2 clauses are CPs in which the finite verb is located in C.
 The limits of syntactic variation featuring “low” subjects. Given the presence of an overt expletive in cases like (21bii), it is, however, not obvious that this is a reasonable postulation, particularly if LF considerations are borne in mind. Further, it is striking that raising vs. non- raising of the subject is necessarily associated with specific interpretive effects, a state of affairs which suggests an optional rather than an obligatory operation (cf. Chomsky 2001: 34 on the so-called Fox-Reinhart intuition on optionality, and also Biberauer Richards 2006; Biberauer 2004, 2006; Richards Biberauer 2005 for detailed discussion of the status of the EPP in Germanic). In at least some Ger- manic varieties, then, it is not so clear that the EPP as originally formulated holds. Looking beyond languages that have traditionally been labelled subject- prominent (cf. Li Thompson 1976), it is not so clear that the subject-centric EPP accurately captures the grammatical focus of so-called topic-prominent languages (cf. Modesto, this volume for overview discussion). More recently, it has also been proposed that it may be productive to think of languages as being either D-prominent or V-prominent (cf. Davies Dubinsky 2002), with the latter privileging a verbal rather than a nominal category. On this view, it could then be the case that I – or, following Pollock (1989) and others, the relevant I-related head in a more ar- ticulated structure17 – always requires a specifier, but it need not be the case that this specifier is always reserved for the subject or expletive. As with (the relevant subset of) topic-prominent languages, the absence of expletives in V-prominent languages could therefore be viewed as a consequence of the non-pivotal nature of subject/nominal categories in these systems, rather than as an indication of their semi-null-subject nature (cf. Section 5 below). Finally, researchers working on a wide spectrum of languages ranging from Germanic to Austronesian have also considered the possibility that what one might think of as D- and V-prominent languages may in fact differ in respect of the “size” of the category that undergoes raising to Spec-TP (cf. various contributions in Carnie Guilfoyle 2000 and Carnie, Dooley Harley 2005, and also Richards . It is worth noting that the genuinely universal status of the EPP becomes extremely difficult to maintain in the context of so-called “cartographic” approaches to functional structure within which I may be decomposed into several projections. What analyses couched in these terms have tended to highlight is that languages in fact appear to have available to them a range of IP-internal subject positions (cf. Bobaljik Jonas 1996; Cardinaletti 1997, 2004; in some cases, subject positions that have been proposed may in fact be located in the lower CP-domain, but we leave this matter aside here). If this is correct, the implication is that one cannot think of a single, EPP-imposed, obligatorily filled subject position in these languages. To “save” the EPP, one could potentially appeal to a “spreading” device of the kind mooted in Chomsky (2008) and in terms of which the phase-head C “spreads” its EPP-feature to one and only one of the I-heads in its domain. In pre-2008 Oh Phases terms, though, it is not clear how this would be achieved.
Introduction  Biberauer 2005 and Travis 2006). In terms of this type of approach, then, the idea is that Spec-TP must be projected and filled, but that is not specifically filled exclusively by a subject; even where a language is D-prominent, it may permit or even require a larger category containing the subject to raise to Spec-TP. Biberauer (2004) argues this to be the case in Afrikaans, where the subject standardly piedpipes its containing vP to Spec-TP, thereby delivering Afrikaans’s characteristic SOVAux ordering, and Biberauer Roberts (2005, 2008) make a similar case for earlier English. Among the advantages of proposals of this sort is that they can account not just for unexpected subject-related phenomena – e.g., the absence in certain structures of an overtly rea- lised expletive where a language does not generally license expletive pro (expletive pro is not expected to be present if Spec-TP is occupied by a larger category) – but also for unrelated phenomena such as word order and word-order variation. Given the above, then, it would seem that there are a range of empirical and theoretical reasons why it is not feasible to maintain a universally instantiated priv- ileged subject position. One could, of course, maintain the projection-of-Spec-IP component of the EPP independently of its subject connection, but this represents a departure from the principle as originally formulated. What we have seen in this section, therefore, is that a number of prominently cited GB principles and parameters do not appear to be empirically supported in the manner originally predicted. As we will see below, this has led some researchers to reject the usefulness of the parametric enterprise as a whole and to question the extent to which it is in fact feasible to identify universally instantiated limits on variation. In the following section, we will, however, consider an alternative per- spective on this matter, one which was, in part, necessarily brought about by the shift to the current phase of the PP enterprise, namely the Minimalist Program. 3.4 Principles and Parameters in the Minimalist Program The Minimalist Program (MP) has its origins in the theoretical “stock take” that Chomsky Lasnik initiated at the start of the 1990s, after roughly 10 years of work within the PP model. Rather than rejecting GB and its predecessors and “starting afresh”, as critics sometimes suggest, it aims to critically evaluate the significant body of theoretical work produced previously and to subject it to the kind of Occam’s Razor assessment that characterises the normal course of scientific research more generally. Additionally, however, the MP, at least as pursued by researchers with a “biolinguistic” commitment (cf. Chomsky 2005, 2007, 2008 for discussion and references), also has at its core a new hypothesis, namely the Strong Minimalist Thesis (SMT) given in (22): (22) The Strong Minimalist Thesis (SMT) Language is an optimal solution to legibility conditions (Chomsky 2000: 97)
 The limits of syntactic variation In terms of the SMT, the starting point for any MP model of the language faculty is that its design is such that it satisfies the requirements that it clearly must meet18 in respect of feeding linguistic information to and from the systems with which it interfaces – the conceptual-intentional (“meaning”) system and the ar- ticulatory-perceptual (“sound”) systems – in the most efficient manner possible. The assumption is not that this hypothesis, which in effect postulates a maxi- mally “well-designed” or “perfect” language faculty, is likely to be correct; instead, Chomsky’s rationale is that research within the PP programme had, by the early 1990s, advanced to the point where enough was understood for a hypothesis like the SMT to serve the kind of heuristic function that is likely to drive the UG enterprise forward (cf. Chomsky 2005, 2007 for discussion). Instead of simply continuing to add to the machinery already proposed in the haphazard way that often becomes inevitable in the context of a widely adopted research enterprise, the suggestion, then, was that the SMT should serve as a yardstick to determine whether particular theoretical notions can and should be maintained or not: the SMT requires that each theoretical assumption be subjected to the question of whether a principled explanation can be found for it, with principled being un- derstood as meaning “imposed by one or both interfaces” and/or “following from general considerations of computational efficiency” (cf. Chomsky 2000, 2001, 2004a, 2004b, 2005, 2007, 2008). A direct consequence of the adoption of the SMT is, of course, the need to minimise the number of purely language-internal entities one postulates. On the architectural front, this led to the elimination of syntax-internal levels (D- and S- structure) and also, crucially for our purposes, of the modules associated with the GB model of the grammar (the Phrase-Structure, Theta, Case, Bounding, Binding and Control modules). As noted in Section 3.3, at least some GB parameters were assumed to be associated with specific modules of the grammar. Thus the HP, for example, was clearly a phrase-structure module-related parameter, while the Wh- movement parameter was associated with the Bounding module. In the absence of the modules with which they were associated, it is clear that these parameters must be “located” elsewhere if they are to be retained within an MP grammar. In principle, the simplified MP architecture, comprising just a Lexicon, the Com- putational System (which “does” Narrow Syntax) and (possibly) two interface components, PF and LF, allows for two general possibilities: either parameters are located in the Lexicon or they are located at one of the interfaces. The notion that parameters might in fact be associated with lexical items is a long-standing one, first proposed during the GB era as the Lexical Parametrisation . Chomsky (1995 et seq.) refers to these as “virtually conceptually necessary” requirements.
Introduction  Hypothesis (cf. Borer 1984; cf. also Manzini Wexler 1987 for further seminal discussion). As noted by Baker (this volume: 253), the idea that much, if not all, parametric variation is located in the Lexicon constitutes a core assumption in minimalist theorising. Baker (ibid) labels it the Borer-Chomsky Conjecture (BCC) and states it as follows: (23) The Borer-Chomsky Conjecture (BCC) All parameters of variation are attributable to differences in the features of particular items (e.g., the functional heads) in the lexicon. The appeal of the BCC is clear: if lexical items constitute the sole locus of para- metric variation, acquiring these will, in effect, amount to acquiring the language as a whole. In other words, on this view, “learning vocabulary” does not just entail learning the idiosyncratic component of language (i.e., that traditionally associ- ated with Saussurean arbitrariness) and, presumably at the same time, also what the phonotactic and morphological properties of the language in question are (i.e., two types of system-defining knowledge); it also entails acquiring the parametric profile of the language, fixing the unvalued parameters of S0 (cf. (5) above). Reconsidering the parameters discussed above, it is clear that a number of them are plausible candidates for reformulation in terms of the BCC. The NSP in (7), for example, in effect specifies two lexical requirements, namely that a lan- guage must have both a particular type of functional category (“richly” specified I) and a particular type of pronominal (pro). The HP in (6), by contrast, initially looks like a far less likely candidate, but this parameter is also amenable to reformulation in lexical terms and numerous researchers have in recent years done so, either explicitly or tacitly. Thus one might, for example, assume Kayne’s (1994) Antisymmetry proposal in terms of which phrase structure is exceptionlessly head-initial and propose that surface head-final orderings are the consequence of leftward movement of complements. In practice, what this entails is that the heads which surface head-finally bear a movement- triggering feature (an EPP-feature in current minimalist parlance), whereas those which surface initially do not. On this model, then, at least all the functional heads in harmonic head-final languages, and possibly heads generally, are lexically speci- fied as bearing movement-triggering EPP-features (precisely what determines the “size” of the moved constituent – i.e., whether we see DP-movement or “massive pied-piping” of the sort proposed in so-called roll-up analyses – seems to require independent specification; cf. discussion below).19 By contrast, the (functional) . Note that, in the context of a theory that assumes the distribution of EPP-features to con- stitute a central point of crosslinguistic variation (cf. also GGL this volume), with functional categories (and possibly lexical items) either bearing or not bearing this movement diacritic, the
 The limits of syntactic variation heads in harmonic head-initial languages systematically lack EPP-features (except where these are optionally present in order to guarantee a non-neutral, move- ment-derived structure20). In disharmonic languages like German and Gungbe, in which phrases do not all exhibit the same head-directionality, the same system would obtain, with the result that the Lexicon in the case of languages of this type would contain both EPP-bearing (head-final) and EPP-lacking (head-initial) heads. Finally, in languages in which linearisation appears to play a lexical disam- biguation role (cf. the examples mentioned in Section 3.2), EPP-features would only be associated with the specific lexical items exhibiting final ordering.21 This last case suggests that not just functional categories, but individual lexical items may bear independent linearisation properties. In the context of the BCC, this does not sound problematic, but it is worth considering the implications of this possibility. One very obvious implication is that it would seem to entail massive redun- dancy in the context of all languages in which the HP is not operating “myopi- cally” (i.e., lexical item-sensitively; cf. Section 3.3 above) – that is, in all languages where head-directionality is more systematically determined on the basis of either syntactic category membership (the German/Gungbe case) or in an across-the- board fashion (the harmonic cases). In other words, ascribing head-directionality specifications (via diacritics or otherwise) to each and every lexical item would seem, in the vast majority of cases, to be complicating a system which could be significantly simpler if this information were located elsewhere (cf. also Baker, this volume, for the same argument). As such, this kind of proposal would not seem to be in keeping with the SMT in (22).22 GB Extended Projection Principle (discussed in Section 3.3) emerges as a stipulation that would require principled explanation: why would it be necessary for T (=I) in every language to bear an EPP-feature, whereas all other functional categories either may or may not do so? . Cf. Biberauer Richards (2006) on the difference between obligatory and optional EPP- features and the interpretive effects they bring about, and Biberauer, Holmberg Roberts (2007) for discussion of the necessary distinction between linearisation-ensuring features and those responsible for A’- and possibly also A-movement. . Precisely how this system would account for heads that may appear both initially and finally is not clear. Possibly, the relevant lexical entries would be underspecified for linearization information or perhaps one could, as suggested in note 10, follow Adger (2006) in proposing that certain types of lexically-related optionality may in fact involve two distinct lexical items with the same phonological and semantic characteristics, but different linearisation specifica- tions. Given how strongly languages in general seem to resist this latter “homonymy” scenario, however (cf. i.a. Kroch 1994), the latter may not be a plausible hypothesis. . Worth noting, however, is an alternative perspective on this matter, in terms of which the “redundancy” entailed by assigning an EPP-feature to all the categories in a given system in fact
Introduction  Anotherseriousproblem,alsodiscussedinBaker(thisvolume),withthenotion that linearisation information is stored on a lexical-item-by-lexical-item basis is that it leads us to expect “a relatively smooth continuum of languages” (Baker, this volume: 360), in which there are many mixed (i.e., disharmonic) languages of dif- ferent kinds and relatively few harmonic languages. In reality, however, the facts seem to be rather different. Consider in this connection Greenberg’s original ob- servations, the obvious skewing in favour of harmonic languages registered in (2) above, and Dryer’s (1992, note 17) observation, which points to “disharmonies” having predictable origins not only within languages, but also crosslinguistically. Looking beyond the domain of the HP, Dryer (1998: 123–124) has observed that empirical work strongly suggests that languages “tend to cluster together around a certain type and that they become increasingly infrequent as we move away from the core type”. As Baker (this volume) notes, facts of this kind raise serious ques- tions about the currently quite generally accepted and largely unexamined idea that parametric variation is necessarily lexically specified. Significantly, linearisa- tion was precisely one of the domains that Fukui (1988) argued should fall outside the domain of lexical parameterisation (cf. his Ordering Restriction Hypothesis). Additionally, he also argued that parametric variation should be restricted to func- tional categories, with lexical categories like N and V, for example, not playing any role in parametric variation (cf. his Functional Parameterisation Hypothesis). This latter idea is one that is currently widely held, with minimalists frequently inter- preting the BCC as applying specifically to functional categories (FCs).23 As the contributions by GGL and Kayne in particular highlight, however, it is not nec- essary to adopt this approach and the presence versus absence of specific lexical items may in fact also play a central role in determining the more general syntactic options available to a given system. We return to this matter in Section 4 below. emerges as highly economical. If we adopt the perspective of a simplicity-driven acquirer who is trying to establish the directionality of the syntactic categories in their language (i.e., whether they are associated with an EPP-feature or not), one could argue that the simplest systems are those in which the acquirer simply postulates C[+EPP] or C[–EPP], i.e., the same EPP-specifica- tion for all categories (C) in the system (see i.a. Clark Roberts 1993; Roberts Roussou 2003 and Roberts 2007a for discussion of simplicity-driven acquisition). This does, of course, raise the complex question of “markedness” – specifically, whether “harmonic” systems should not actually be easier to acquire than “disharmonic” ones – but we leave this matter aside here. Thanks to Ian Roberts for discussion of this point . Under the view that lexical categories like N and V do not constitute primitive entities, but that “nominalisation” and “verbalisation” are in fact achieved via merger of acategorial roots with functional categories like n and v (cf. Marantz 1997; Chomsky 2001: 9), it is arguably im- possible for lexical categories to play a role in parametric variation independently of functional categories (for a very different view on the noun/verb distinction, see Kayne 2008).
 The limits of syntactic variation For the moment, we conclude, firstly, that the BCC is a hypothesis which requires more critical examination than that to which it has been subjected to date and, secondly, that linearisation in particular may well not be the kind of parametric property that is best located in the Lexicon. Leaving aside the objections raised above, there is also the in recent years oft-discussed question of whether linear ordering should play any role in Narrow Syntax (NS), as it would presumably have to if it is already lexically specified. As has been pointed out many times, the computational system does not appear to be sensitive to linear ordering; only to hierarchical structure. As such, one might expect that linear ordering – even if this is ultimately determined via Kayne’s (1994) Linear Correspondence Axiom (LCA) – is only imposed at the stage at which it is self-evidently required, namely at PF (cf. Chomsky 1995: 335, 340). We pursue this matter below. One aspect of linearisation that is, however, usually thought to lend itself to lexical parameterisation is that which was previously encoded by the various movement parameters – V-to-I movement, N-to-D movement, V2 and Wh- movement, to name but a few. In each case, the relevant movement requirement can be lexically encoded as a movement diacritic (EPP-feature, etc.), with lan- guages lacking movement in the relevant domains correspondingly lacking this diacritic.24 There is little doubt that the easy “translateability” of the movement pa- rameters into BCC terms has played a significant role within minimalist research in foregrounding the role of the Lexicon in determining parametric variation. It is, however, worth noting that the matter may be significantly more complicated than “recastings” of this kind might suggest. Consider, firstly, the fact that movement parameters, if they are lexically encoded and implemented in the syntax, once again introduce linear ordering into Narrow Syntax. Furthermore, if Chomsky (2001: 37–38) and others are in fact correct in consigning (at least some) head- but not phrasal movement to the PF component (i.e., to an interface), it is not clear how the computational system can effectively “ignore” the movement dia- critics that will ultimately result in head-movement effects, while satisfying those which will ultimately deliver phrasal movement. It is, of course, conceivable that head-movement triggers are invisible (i.e., uninterpretable) to the computational system, while XP-triggers are not and that we are thus effectively dealing with two different types of movement diacritic. Given that certain head-movements do appear to take place in the syntax, however (cf. Lechner 2006 and Roberts 2006 on LF effects associated with head-movement), it would seem that there also needs to . Alternatively, one might, like Brody (1995), Bobaljik (2002), Landau (2006) and others, think in terms of P-features which determine which copy of a chain is ultimately pronounced.
Introduction  be a Narrow Syntax-internally interpretable head-movement trigger – and, addi- tionally, since syntactic relations always involve head-head relations (cf. Chomsky 2004a et seq.), some means of signalling when head- rather than XP-movement is to take place (cf. Watanabe 2006 on “pied-piper” features). If some head-movement is PF-movement, then, quite an extensive inventory of movement-triggers would seem to be required. If it is not, there are still many currently unanswered questions as to how the computational system is able to determine the “size” of the category it is to move (cf. Biberauer Richards 2006; Biberauer Roberts 2006 and Watanabe 2006 for some recent discussion). Let us now turn to the alternative potential loci for Minimalist parameters: the interfaces. Since the possibility of LF parameters has not received much at- tention to date, we will focus on the PF interface here (cf., however, Ramchand Svenonius, this volume, for discussion of the role that LF may potentially play in determining parametric variation). As noted above, the PF interface would seem the most plausible locus for lineari- sationgenerallyandthereforealsoforparametricallyvariantaspectsoflinearization.25 Chomsky himself has always held that “[t]here is no linear order in the N[umeration] → LF computation” (Chomsky 1995: 335; cf. also Chomsky 2001: 7) and that linear ordering be imposed at PF, where the LCA may, but need not necessarily, determine linear precedence. Actually formulating a plausible PF linearisation system, let alone a credible PF version of the HP has, however, proven quite difficult. As regards linearisation more generally, the problem is ensuring that the structures sent to PF are in fact unambiguously linearisable. If one assumes that the very restrictive (Spec-Head-Comp) phrase structure assumed by Kayne (1994) is the only kind permitted in NS (i.e., that a very strict version of the X-bar tem- plate is in fact encoded as a UG principle), the problem of course becomes trivial: such structure is always linearisable via the LCA. On this view, phrase structure is, however, strictly constrained precisely to ensure a linearisable output, with a “simple PF” coming at the expense of a more complicated computation (including, for example, unrealised heads and specifiers, and “extra” leftward movement wherever head-final structures result – see discussion above and note 22).26 As such, a Kaynian approach to linearisation must be regarded as a departure from . For a minimalist analysis which proposes relocating the HP to Narrow Syntax, however, see Saito Fukui’s (1998) Parameterised Merge proposal. . As a result of the latter, particularly if leftward movements may be of the remnant variety often assumed in Kaynian analyses, there is, of course, also one aspect of the PF “computation” that is not simplified, namely the copy spellout process, i.e., determining which copy/copies to spell out and which to suppress (cf. Corver Nunes 2007 for discussion).
 The limits of syntactic variation the SMT and alternatives should at least be investigated to determine whether a strict LCA-based system in fact constitutes a genuine case where there is a “gap between SMT and the true nature of FL [the Language Faculty – TB]” (Chomsky 2007: 3). An obvious way to relax the Kaynian constraints on phrase structure is to propose, as many researchers following Moro (2000) have done, that syntactic structureneednotbelinearisableateverystageofthederivation,butthatstructures must be linearisable at the stage at which they are sent to PF. Within proposals of this kind, linearisation is therefore only partially a PF matter, with initially unlin- earisable structures being “rescuable” via subsequent movement operations which may – as in Moro’s Dynamic Antisymmetry approach – or may not be directly aimed at neutralising troublesome “points of symmetry” (i.e., symmetrical and thus unlinearisable structures). Languages may then differ as to the points of sym- metry that are generated during the course of the computation27 and possibly also as to when and how they are neutralised (cf. Hsieh Sybesma 2007 for a phase- based implementation of DA and Takita 2007 for a proposed symmetry-breaking parameter). Approaches of this type can, however, clearly be criticised for “look- ahead”, i.e., NS operating in such a way that subsequent requirements – here, PF legibility – can be met.28 We leave this and other matters aside here since our main concern is to consider how feasible it in fact is to completely banish linearisation information from NS. AverydifferentapproachisthatofRichards(2004).Heproposesparameterising the LCA such that PF systematically ignores one set of symmetric c-command relations (cf. also Epstein et al. 1998), thereby capturing the effects of the HP in the manner outlined for VP below: 29 (24) The Parameterised LCA (a PF HP) a. VO = Ignore all O V b. OV = Ignore all V O . Points of symmetry may in principle be either head-head or XP-XP adjunction pairs. . Additionally, it is worth noting that DA-based approaches once again necessitate the postulation of a range of movement-triggering factors since movement in natural languages is clearly not limited to the symmetry-breaking variety. . Richards’s proposal takes as its point of departure the observation that the first two ele- ments Merged in any structure will be heads, i.e., elements which necessarily symmetrically c-command each other and which will therefore also necessarily be unlinearisable in LCA terms without some additional instruction. The Parameterised LCA is intended to supply this instruction and, simultaneously, to state a point of parametric variation. For a very different solution to the head-head merger problem, see Kayne (2008).
Introduction  In terms of (24), then, head-initial languages consistently ignore PF linearisation instructions (i.e., symmetric c-command structures) ordering complements before heads (a), whereas head-final languages consistently ignore the opposite set of linearisation instructions. This system does therefore remove HP-related linearisa- tion information from NS, allowing the HP to be stated as an interface parameter. As with the original HP, however, it is not so clear how an interface HP operates in the context of disharmonic languages, i.e., those which do not set (24) so that it applies across-the-board. One might propose that (24) is sensitive to syntactic- category information and that lexically specified linearisation instructions can over- ride its across-the-board application (cf. note 10), or possibly that the application of (24) is determined on a phase-by-phase basis, with some languages opting for a subset of phases which are linearised in accordance with (a), while the remainder are linearised in accordance with (b). The implications of these possibilities remain to be worked out in further detail, with linearisation patterns that systematically fail to surface in the world’s languages (i.e., impossible linearisations) potentially rep- resenting a crucial empirical testbed (cf. Biberauer, Holmberg Roberts 2007 for discussion, and also Kural 2005 and Cinque 2005, 2008). For the moment, however, what is worth noting about the HP and its potential MP reformulations is that an exclusively lexically based approach would seem to entail considerable redundancy for harmonic languages in particular, effectively making the most idiosyncratic of the disharmonic exceptions seem the most “natural”, whereas a purely PF-oriented approach seems best suited to the harmonic languages while raising questions as to how the disharmonic orders are to be accounted for. If UG is to contain a linearisation parameter, it clearly needs to be one which could in principle be set so as to generate both the harmonic and the disharmonic word orders and it is not presently clear either how this can be achieved or whether (a) suitable linearisation parameter(s) will require a departure from the SMT. We have already noted above that movement parameters would, at first sight, seem trivially recodable as lexical parameters involving the presence versus absence of movement diacritics. Another possibility is that these parameters may in fact be located at PF. Bobaljik (2002) offers a proposal along these lines, sug- gesting that the difference between overt and covert movement may in fact not lie in whether an Agree operation between two elements is or is not associated with a movement diacritic (presence of EPP-feature = overt movement; absence of EPP- feature = covert movement), but in which copy of a moved element PF chooses to spell out. In Bobaljik’s terms, then, overt movement entails spelling out the highest copy of a moved element whereas covert movement falls out from spelling out the lowest copy; movement therefore always takes place in NS, but this plays no role in linearisation (cf. Brody 2000 for an in many ways similar approach focused on a different empirical domain). This proposal is particularly appealing in the context
 The limits of syntactic variation of covert movements that were during the GB era assumed to take place in the der- ivation from S-structure to LF (QR, wh-movement in wh-in situ languages, etc.). If Bobaljik’s proposals are on the right track, it may therefore be the case that GB parameters like the Wh-movement Parameter have both a lexical and an interface component (cf. the oft-discussed correlations between (non-)movement and, inter alia, the availability of what Cheng 1997 designates clause-typing particles and the morphosyntactic structure of wh-elements). To conclude on the question of how GB parameters may be reformulated in the current framework: we have seen that, while the available loci would seem to be clear (the Lexicon and one or more of the interfaces), GB parameters are not equally easily captured in minimalist terms. This may, of course, indicate various things. One possibility – and, given the course of normal scientific progress, it is likely to be a very strong possibility – is that some or all of the GB parameters were misfocused and that they should therefore be viewed, in the context of the PP enterprise, as “first pass” attempts at pinpointing recurring points of syntactic variation (see Section 3.5 below). Another possibility, which arose in the course of our discussion of the Wh-movement Parameter, is that parameters may in fact be the reflex of potentially quite complex interactions between various aspects of language structure (lexical, morphological and syntactic), with the result that for- mulating and “localising” them might involve consideration of a greater number of factors than has traditionally been done (but see Haspelmath, this volume, for a different perspective on the likelihood of cross-domain relations). At the most general level, though, what the minimalist heuristic strongly suggests about some of the GB parameters is the fact that they may not ultimately be formulable in a manner which respects the SMT. Let us consider this point in a little more detail. As Chomsky (2007: 1) notes, considerations of language acquisition had, until the advent of the PP programme, always led Chomskyans to assume that UG must be “rich, highly structured and substantially unique”: if the acquirer was to be able to abduce (i.e., correctly infer on the basis of innately specified knowledge, cf. Peirce 1955) the grammar underlying the PLD to which they are exposed (cf. (5) above), a suitably specific UG seemed a prerequisite. The PP programme led to the revision of this notion since it opened up the possibility of an underlying system in which parameter settings might differ only slightly, while the surface effects of those slight differences might be extensive (consider, for example, the considerable surface difference between consistently head-initial and consistently head-final languages, where the relevant underlying difference, in GB terms, is simply the way in which the HP is set). Within the MP, consideration of the role of a “third factor”, whose potential relevance had previously only been mentioned in passing, but which was never explicitly included in the acquisitional scenario alongside UG (innate endowment) and PLD (experience), has become central.
Introduction  This “third factor” entails principles of mental/neural organisation that are not specific to the faculty of language, but are common to all cognitive faculties (cf. Chomsky 2005, 2007, 2008, Hauser, Chomsky Fitch 2002 and the ensuing debate for further discussion). Furthermore, in terms of the hypothesis driving minimalist research – the SMT in (22) – the null hypothesis is that UG, or the specifically linguistic infor- mation attributed to the human genome, is minimal. In other words, minimalist research, unlike that which went before, explicitly proceeds on the assumption that, in the absence of evidence to the contrary, as little as possible should be at- tributed to UG – specifically, only those properties which do not appear to be re- ducible to more general, possibly partially functionally motivated principles. The question is, as Chomsky (2005: 11) puts it: “How far can we progress in showing that all … language-specific technology is reducible to principled explanation, thus isolating core processes that are essential to the language faculty?”. In the context of SMT-oriented theorising, then, a rich parametric inventory of the type assumed during the GB era would constitute a considerable departure from the SMT and, as such, should only be assumed if the evidence points to the need for it. In the first instance, however, the goal should, as in minimalist theorising more generally, be to consider how much of what was previously proposed can be accounted for by appealing to “non-domain-specific” principles (e.g., those governing efficient computation). The greater the number of language-specific parameters that ultimately needs to be proposed, the more extensive the depar- ture from the SMT will be. As we have seen above, first attempts suggest that it is possible to “locate” some GB parameters in the Lexicon and also that it might be possible to view others as the consequence of particular options which arise at the interfaces. As such, these parameters would represent content that needs to be ascribed spe- cifically to UG, but at least one aspect of their nature – specifically, the fact that they can be formulated in lexical and/or interface terms – is such that one could offer a principled answer to the question: “Why those parameters and not others that one could possibly formulate?”. Answer: the parameters in question are di- rectly associated with the “virtually conceptually necessary” components of the language faculty and do not entail postulating additional modules or otherwise complicating the overall architecture of the language faculty. In the context of the current phase of the PP programme, then, the question is not just what sorts of innately specified knowledge might facilitate language acquisition in the manner that it occurs, but additionally also whether knowledge that seems to require innate prespecification can in fact be seen to take the form it does by virtue of this having been “imposed” by considerations of “good design”. Insofar as the content of UG and the design of the language faculty more generally can be understood
 The limits of syntactic variation as having the form that they do in virtue of “third factor” considerations, it is, in Chomsky’s terms, possible to go “beyond explanatory adequacy”; and the MP aims to probe the extent to which it is possible to do so, approaching the ques- tion both “from above” (i.e., continuing the search for explanatory adequacy) and “from below” (i.e., attempting to discover the extent to which “third factor” considerations have shaped UG). So far, we have speculated that GB parameters which can be viewed as being “located” in one of the “virtually conceptually necessary” components of the lan- guage faculty might be thought of as meeting a “good design” requirement, namely that they do not require the postulation of additional, purely language faculty- internal architecture. A further “good design” question that the SMT forces us to ask is whether it is in fact necessary to view each of the specific parameters that were previously identified as independent entities within UG. Consider, for example, the individual head-raising parameters (e.g., V-to-T, N-to-D, V-to-C, etc.): these clearly share the same format – X raises to Y – and the question the SMT raises is whether UG actually requires each of these parameters to be speci- fied individually. One might, for example, alternatively think of a general template which serves, in much the same way as the X-bar template during the GB era, to determine aspects of syntactic structure (cf. GGL, this volume); or, more radically, that even the template may arise as a result of the interplay between PLD and “third factor” considerations (cf. Zeijlstra and Evers van Kampen, this volume and also Roberts 2007d for different versions of this idea). Evidently, then, there is a very general question within the MP as to the forms that minimalist parameters might take. A further question concerns their status within UG: are they present in “unset” form at S0 or can they arise during the process of acquisition? These are questions which have not featured prominently on the MP research agenda to date, but which clearly need to be addressed if we are to be able to offer a coherent and up-to-date minimalist answer to the question of how principles and param- eters constrain syntactic variation. In the absence of a coherent minimalist theory of this type, it is not in fact possible to meaningfully evaluate the success of the PP enterprise to date. The remainder of Section 3 focuses on a recent evalua- tion by Newmeyer (2005a, 2005b and 2006) which starkly underlines the need for more systematic work on notions like “minimalist parameter”. 3.5 Recent criticism of the parametric enterprise Newmeyer (2005b: 185) states seven desiderata that a parametric approach to syntactic variation should meet if it is to be superior to a potential rule-based alternative. These include descriptive simplicity, binarity, smallness of number, implicational organisation and a rich deductive structure. He then considers the
Introduction  extent to which GB parameters, notably the NSP, satisfy these desiderata, ul- timately concluding that they fall short in various ways, and that the limits of syntactic variation should be sought in “the interplay between unparameterized UG principles, language-particular rules, and processing principles sensitive to structural relations holding among grammatical elements” (2005b: 225). In other words, Newmeyer proposes that parameters should play no role in our model of UG and that the parametric enterprise can be viewed as having failed. It is, however, important to note that his conclusions are based exclusively on consid- eration of GB parameters.30 As such, they can at most be interpreted as potential conclusions as to the feasibility of the “first pass” parameters of Stage I of the PP enterprise, and not as conclusions on the enterprise as a whole; as was discussed in Section 3.4, minimalist parameters will necessarily differ from their GB prede- cessors in various ways and work on precisely what the inventory of minimalist parameters might ultimately look like has scarcely begun. Many of the contribu- tions in this volume are centrally concerned with this question, which was also a major focus of the conference at which they were first presented. What we will briefly consider here is the extent to which some of Newmeyer’s criticisms can also be levelled against the parametric enterprise as it can now be and is now being approached. Our discussion here will focus on the descriptive simplicity, richness of deductive structure and smallness of number desiderata (cf. discus- sion of implicational organisation in Section 4 below, and Roberts Holmberg 2005 for discussion of the other desiderata). Firstly,letusbrieflyconsiderthematterofdescriptivesimplicityandtherelated matter of smallness of number. As discussed in Section 3.4 above, the hypothesis driving minimalist research (the SMT in (22)) takes as its point of departure that UG does not contain domain-specific information. As we have seen, it may well be possible that it need not contain specific parameters of the type assumed during the GB era and that any parametric options that are innately specified may take the form of extremely general and thus also descriptively extremely simple tem- plates (cf. GGL, Zeijlstra and Evers van Kampen for suggestions as to the format that such templates might take). If this thinking proves to be correct, it may then well be the case that a very small number of (or possibly even no) innately speci- fied templates gives rise to a diverse range of steady-state (Ss) parameters. This line of thinking has an important consequence in the context of the point New- meyer (2005b: 185) raises regarding the complexity of the variation observed in the world’s languages. According to Newmeyer, this is such that it simply is not . Cf. Roberts Holmberg (2005) for a critical evaluation of some of the conclusions Newmeyer draws.
 The limits of syntactic variation feasible to postulate only a small inventory of innately specified parametric options, with the result that one might as well assume that children do not draw on innately specified parametric choices, but instead learn language-specific rules which capture the relevant dimensions of variation. If it can, however, be shown that it is possible for the Ss parameters to be “constructed” during the course of acquisition, that these do not have to be individually prespecified and that their effects might nevertheless be wide-ranging, Newmeyer’s conclusion no longer follows. Let us turn now to the matter of the “rich deductive structure” traditionally ascribed to parameters. As we have seen in Section 3.4, lexically specified parame- ters associated with specific lexical items and/or functional categories have played a prominent role in minimalist theorising to date. At first sight, these might seem to be parameters which may not always meet the “rich deductive structure” desid- eratum since their effects are naturally limited to the domains in which the relevant lexical item is present, with the result that lexically specified parameters are often (erroneously–seethediscussionofpossibleencodingsoftheHPabove,andbelow) equated with microparameters, i.e., parameters that define the kinds of small-scale differences in terms of which closely related dialects may differ (cf. Kayne 2005b for discussion). Newmeyer (2005b, 2006), like i.a. Safir (1987: 78) and Bouchard (2003) before him, is of the view that parameters of this type entirely undermine the rationale for parameters, essentially reducing the term parameter to a synonym for rule. In the words of Newmeyer (2006: 9), “if every difference or even most dif- ferences are attributed to a difference in “parameter setting”, then saying that two languages have different parameter settings merely becomes an obfuscatory way of saying that they have different rules”. For Newmeyer, then, only parameters of the type Baker (this volume) designates macro- and medioparameters have a suf- ficiently rich deductive structure to justify speaking of “parameters” as opposed to “rules”. Like Pica (2001), however, he believes that the past quarter century of research within the parametric framework has shown that parameters of this kind are not empirically motivated, leaving us, at best, with lexically specified micropa- rameters, which, for him, are in fact most realistically viewed as language-specific rules. For Newmeyer, then, one of the central failings of the parametric enterprise is its failure to bring to light plausible, empirically substantiated parameters which have wide-ranging and at first sight seemingly unrelated effects. Worth noting about Newmeyer’s conclusion are the assumptions on which it is based. Firstly, it takes for granted that lexically specified (micro) parameters necessarily have very limited surface effects of the kind that can be very straight- forwardly captured in generalised rule format. That microparameters could also entail “deep” effects, with the presence of a lexical item or particular feature speci- fication, for example, determining the (un)availability and/or the form of further, often seemingly quite unrelated structures in the language is not entertained as a
Introduction  possibility. As Kayne’s contribution to this volume and his and other researchers’ microparametric work more generally has strikingly shown, however, points of microparametric variation do not have to differ from medio- or macroparametric ones in this regard (cf. also Rizzi 2004: 332 on this point). Furthermore, as Baker (this volume: 355) points out, “[i]t is perfectly possible that a lexical parameter … could have a substantial impact on the language generated, particularly if it concerned some very prominent item (such as the finite Tense node)”. And, con- versely, a macroparameter like the Polysynthesis Parameter might only have a very limited impact on a language lacking extensive agreement (cf. also Baker 1996: note 35). Kayne (2005b: 5–6) likewise registers the opacity surrounding the notion of parametric “size”, pointing out that “what seems “dramatic” [in terms of number of empirical reflexes or range of effects – TB] depends on expectations that may themselves be somewhat arbitrary”. Given all of this, it seems fair to say that it is not at all clear that the discovery and existence of “large” parameters, assuming the meaningfulness of a notion of this type, is crucial to the success of the parametric enterprise. Furthermore, given the potentially extensive effects of “smaller” pa- rameters, it is also cannot be said to follow that the possible non-existence of “large” parameters entails that one might as well conceptualise crosslinguistic variation in terms of language-specific rules. Another point worth noting in this connection is that Newmeyer’s conclu- sions rest on the assumption that macroparametric research is no longer being seriously entertained by researchers engaged in the parametric enterprise. As Baker himself indicates in his contribution to this volume, it is certainly true that genuinely macroparametric research is not receiving nearly the attention that Kayne-style microparametric research is. This is, however, different from saying that researchers are no longer engaged in this type of research at all, and various contributions to this volume show that this is not in fact the case – cf. Modesto on a potential Topic Parameter, Zeijlstra on a Negative Concord Parameter, Öztürk on a Non-configurationality Parameter and Avram Coene on a Split INFL Pa- rameter which, unlike Bobaljik Thráinsson’s (1998) original, is hypothesized to have large-scale effects both in the clausal and the nominal domain. And beyond the articles collected here, Huang’s (2006) work on the synchronic and diachronic consequences of a so-called Analyticity Parameter, in some senses the diametric opposite of Baker’s (1996) Polysynthesis Parameter, constitutes further evidence that this line of research is also receiving attention. Further, this volume also contains papers which consider parameters that might, in Baker’s terms, be thought of a medioparameters – cf. Lekakou’s Aspect Parameter, Fernandez-Salgueiro’s Case-F Valuation Parameter and Neeleman Szendröi’s Radical Pro-drop Param- eter. In sum, then, the contributions in this volume do not support the view that present-day parametric research is exclusively microparametric or that “larger”
 The limits of syntactic variation parameters are no longer on the research agenda. Furthermore, as already noted above and as is also specifically highlighted in Baker’s paper, there is, quite indepen- dently of the amount of research being done on non-microparametric syntax, unde- niable empirical evidence – uncovered by both typologists and generativists – that the world’s languages robustly exhibit skewings of various kinds; and, moreover, skewings that often cannot be economically stated in rule form (cf. the discus- sion in Section 3.4 of lexical reformulations of the HP). These empirical facts clearly require explanation, and it is not obvious how a purely (language-specific) rule-based account of the kind Newmeyer espouses gets us any closer to under- standing them. To conclude, then: consideration of just three of Newmeyer’s desiderata in relation to the current stage of the parametric enterprise would seem to lead to very different conclusions as to its potential value to those drawn by Newmeyer himself. There is no denying the fact that core questions relating to the number, format, size, interrelationship and, in fact, given Chomsky’s “three factors” specu- lations, the very existence of innately prespecified parameters do not have clearcut answers at this stage of the PP enterprise. These are matters which need to be at the forefront of the MP research agenda to a much greater extent than they currently are if a meaningful evaluation of the parametric aspects of the PP en- terprise as it stands now is to become a realistic prospect. That the GB parameters which are the focus of Newmeyer’s evaluation do not constitute plausible innately specified constraints on syntactic variation is beyond dispute; what remains to be seen now is whether or not the same will be true of the current phase of the PP programme. 3.6 Conclusion This section has shown that the factors determining the limits of syntactic varia- tion are in many respects still poorly understood. Accordingly, questions as funda- mental as whether innately specified knowledge plays a role or not and if so, what form such knowledge should take are still under debate. Against this background, what the present volume aims to do, then, is to give a flavour of current thinking on the nature and value of parameters and parametric theory in accounting for syntactic variation. The remainder of this introduction will now focus more specifically on the contents of its sixteen papers and the issues that they address. The volume is or- ganised into four sections which respectively focus on the following topics: • the locus of (parametric) variation (cf. Section 4) • the NSP (cf. Section 5)
Introduction  • parametric clustering (cf. Section 6); and • the acquisition of parameters (cf. Section 7) 4. The locus of parametric variation This section considers the under-discussed question of “where” in terms of the overall architecture of the grammar we might expect points of syntactic variation to be specified. It opens with a paper by Haspelmath which aims to fill a gap in the currently existing literature on (the limits of) syntactic variation by providing (i) a system- atic comparison of the different ways in which parametrically and functionally oriented linguists traditionally approach the limits question and (ii) a critical evaluation of the success the two approaches have achieved. Haspelmath’s main contention is that the recurrent patterns in the world’s languages – so-called phe- nomenological universals in his (2004) terminology – require an observationally adequate account, i.e., one that distinguishes possible from impossible expressions in a language (p. 2) and that does not make any claims about speakers’ knowledge; attempts at achieving any “higher” level of adequacy and, particularly, at trying to gain insight into the workings of the human mind via the study of syntactic limits are not, in Haspelmath’s view, feasible nor, given the availability of functional and processing explanations, even necessary. The proposal is that languages should (contra Meillet 1903: 407) not be viewed as “systems in which everything hangs together”. Consequently, so-called cross- domain implications where, for example, semantic organisation correlates with aspects of phonological organisation, should be expected to be rather rare, whereas so-called intra-domain implications where, for example, one aspect of phonology correlates with one or more other aspects of phonological organisation, need not be so. In keeping with these expectations, typological research has revealed a great many intra-domain implications, many of which are, according to Haspelmath, amenable to a functional explanation (cf. pp. 96–100). Cross-domain implications, on the other hand, have proved rare, producing only weak correlations which are in many cases quite difficult to distinguish from areal effects (cf. p. 19). The conclusion to be drawn from this, then, is argued to be that a research programme which seeks to understand apparently unrelated surface phenomena, particularly where these do not belong to the same domain, as the reflexes of a far smaller number of parameters is misfocused: the data do not warrant such a programme and it is doomed to fail in the same way in which previous attempts to propose “holistic” lan- guage types have failed (cf. the fate of the 19th century morphological typologies). Instead, the question that should be being asked about recurrent patterns, given
Introduction  restrictions of computational efficiency and conditions of the interfaces” (p. 118). On this basis, they identify five “core” parameter schemata, leaving open the possibility of further schemata relating to phonetic nullness and lexico-semantic parameterisation. In the context of GGL’s so-called Parameters and Schemata (PS) model, then, the contents of UG is restricted to a limited number of func- tional features, and lexical categories, together with “a tiny class of parameter schemata” (p. 122); the parameters with which we are familiar are not therefore linguistic primitives; instead, they are constructed on the basis of innately speci- fied features, lexical categories and schemata, which are the actual primitives. Aside from the new possibilities that this type of approach opens up in the domain of language acquisition (cf. Section 3.4 and 3.5 above), it also makes a very interesting new prediction about the extent to which grammatical variation is expected to emerge over time. As GGL note, current minimalist-inspired theories might lead one to conclude that the amount of variation should de- crease over time, whereas the linguistic reality suggests that the opposite is in fact true – consider, for example, the immense grammatical diversity found in Italo-Romance, a language family which has a 1500–2000 year history, com- pared to the restricted grammatical diversity found in the various varieties of Afrikaans, a language which has only existed as a variety speakers think of as an independent language, distinct from Dutch since the late 19th century (cf. Deumert 2004).32 GGL argue that their PS model predicts that grammatical diversity will increase over time. In their words: “once the introduction of a pa- rameter schema into the language faculty is justified for a single case, it will … bring about every more potential variation for all possible combinations of relevant entities in the lexicon (features and categories)” (p. 123). Finally, GGL also consider the role that parameters – viewed in the de- rivative, schemata-determined manner outlined above – might potentially be able to play in facilitating a better understanding of the phylogenetic relations between languages. They argue that the correctness of the phylogenetic predic- tions made on the basis of MGP, firstly, corroborates the validity of postulating parameters and, secondly and more importantly, opens up the possibility of using parametric values as comparanda in seeking to establish linguistic relationships. . As Ian Roberts (p.c.) observes, it should, however, be noted that it is too simplistic to view the emergence of grammatical variation as a process which proceeds uniformly: change can sometimes occur in sudden spurts, as seems to have been the case in Middle English (cf. Fischer et al. 2000) and 9th century Icelandic (cf. Hróarsdóttir 2000); conversely, languages may remain fairly static for relatively long periods of time, at least as far as certain properties are concerned – consider, for example, the lack of change in most of the Romance languages with respect to their NSL status.
 The limits of syntactic variation For GGL, then, parametric theory, rigorously pursued, may facilitate insights into a wide range of issues, going well beyond the points at which languages may differ. Zeijlstra’s paper, like GGL’s, pursues the question of what a truly minimalist, plausibly specified UG would contain and the proposal he makes resonates with theirs in various interesting ways. Zeijlstra’s specific focus is the question of the identity of the set of formal features available to UG and his central proposal is that this set is in fact empty. Instead, the inventory of formal features that can head a functional projection in a given language is argued to be fixed on the basis of PLD, during first-language acquisition (cf. also Evers van Kampen for a similar pro- posal in another domain). Strikingly, Zeijlstra and GGL’s proposals converge on one of the choice-points requiring fixing during language acquisition – Is feature F formalized as a formal (grammatical) feature? (cf. the first of GGL’s schemata). Where they differ is in respect of the ontological status that they assign to this choice-point: for GGL, it is one of the innately specified schemata in UG, whereas for Zeijlstra, it simply constitutes a linguists’ description of one of the major types of PLD-driven choice that children make during acquisition. In terms of Zeijl- stra’s Flexible Formal Feature Hypothesis (FFFH), children must determine for the features encoded in their language whether a given feature is a purely semantic feature which has C-I-relevant content, but none which is relevant to the compu- tational system (i.e., NS) or whether it has both. In the latter case, a formal (i.e., grammatical) feature will be postulated. The cue for deciding the status of a feature is the presence vs. absence in the PLD of semantically vacuous doubling: where it is present, the child “notices” and a formal feature is postulated; where it is absent, there is nothing to “notice” and therefore nothing in effect happens as the FFFH does not assume the prior existence of an entity requiring “negative setting” (cf. also Roberts 2007d: 17–18). Zeijlstra presents the crosslinguistic distribution of Negative Concord phenomena as a case study to illustrate the workings of the proposal, demonstrating that [negation] must be analysed as a formal feature in Negative Concord languages, where negations do not cancel each other out. In so- called Double Negation languages, by contrast, each negative element does in fact bear negative meaning, with the result that children acquiring these languages will conclude that [negation] is always and only a semantic feature. In general, then, the proposal is that instances of syntactic agreement (“concord”) will signal the presence of a formal feature in the language being acquired. As Zeijlstra notes, the proposal has clear implications not only for the de- scription of synchronically attested varieties, but also for the manner in which diachronic changes will take place. In the latter context, it is clear that the loss and/ or introduction of sound-meaning forms (words, clitics, inflections, etc.) bearing features which consequently are or are not doubled would be expected to result
Random documents with unrelated content Scribd suggests to you:
Eskimo, and the so-called “shamans” of southern Alaska, but, as far as we could see, do not possess the power and influence usually elsewhere ascribed to this class. It was exceedingly difficult to obtain any definite information concerning these people, and we only discovered casually that such and such a person was a “doctor” by hearing that he had been employed in a certain case of sickness, or to perform some ceremony of incantation. We did not even succeed in learning the name of this class of people, who, in talking with us, would call themselves “tûktĕ,” as they did our surgeon. On one occasion some of the party happened to visit the house of a sick man where one of these “doctors” was at work. He sat facing the entrance of the house, beating his drum at intervals, and making a babbling noise with his lips, followed by long speeches addressed to something down the trapdoor, bidding it “go!” We were given to understand that these speeches were addressed to a tuɐña or supernatural being. 576 Their only idea of direct treatment of disease is apparently to apply a counterirritant by scarification of the surface of the part affected. We know of one case where a sufferer from some liver complaint had inflicted on himself, or had had inflicted upon him, quite a considerable cut on the right side with a view of relieving the pain. We also know of several cases where the patients had themselves cut on the scalp or back to relieve headache or rheumatism, and one case where the latter disorder, I believe, had been treated by a severe cut on the side of the knee. A similar practice has been observed at Plover Bay, Siberia, by Hooper, 577 who also mentions the use of a kind of seton for the relief of headache. They also practice a sort of rough-and-ready surgery, as in the case of the man already mentioned, whose feet had both been amputated. One of the men who lost the tip of his forefinger by the explosion of a cartridge was left with a stump of bone protruding at the end of the finger. Our surgeon attempted to treat this, but after two unsuccessful trials to etherize the patient he was obliged to give
it up. When, however, the young man’s father-in-law, who was a noted “doctor,” came home he said at once that the stump must come off, and the patient had to submit to the operation without ether. The “doctor” tried to borrow Dr. Oldmixon’s bone forceps, and when these were refused him cut the bone off, I believe, with a chisel. They appear to have no cure for blindness. We heard nothing of the curious process of “couching” described by Egede in Greenland, p. 121. We had no opportunity of observing their methods of treating wounds or other external injuries. Sufferers were very glad to be treated by our surgeon, and eagerly accepted his medicines, though he had considerable difficulty in making them obey his directions about taking care of themselves. After they had been in the habit of receiving the surgeon’s medicine for some time, one of the Utkiavwĭñ natives gave Capt. Herendeen what he said was their own medicine. It is a tiny bit of turf which they called nuna kĭñmölq, and which, therefore, probably came from the highland of the upper Meade River, which region bears the name of Kĭñmölq. We were able to get very little information about this substance, but my impression is that it was said to be administered internally, and I believe was specially recommended for bleeding at the lungs. Possibly this is the same as “the black moss that grows on the mountain,” which, according to Crantz 578 was eaten by the Greenlanders to stop blood-spitting. CUSTOMS CONCERNING THE DEAD. Abstentions.— From the fact that we did not hear of any of the deaths until after their occurrence, we were able to learn very few of their customs concerning the dead. The few observations we were able to make agree in the main with those made elsewhere. For instance, we learned with tolerable certainty that the relatives of the dead, at least, must abstain from working on wood with an ax or hammer for
a certain period—I believe, four or five days. According to Dall, 579 in the region about Norton Sound the men can not cut wood with an ax for five days after a death has occurred. In Greenland the household of the deceased were obliged to abstain for a while from certain kinds of food and work. 580 A woman from Utkiavwĭñ, who came over to the station one day in the autumn of 1881, declined to sew on clothing, even at our house, because, as she told Lieut. Ray, there was a dead man in the village who had not yet been carried out to the cemetery and “he would see her.” After consulting with her husband, however, she concluded she could protect herself from him by tracing a circle about her on the floor with a snow-knife. In this circle she did the sewing required, and was careful to keep all her work inside of it. One of the natives informed me that when a man died his labrets were taken out and thrown away. I remember, however, seeing a young man wearing a plug labret of syenite, which he said had belonged to an old man who died early in the winter of 1881-’82. It was perhaps removed before he actually died. Manner of disposing of the dead.— The corpse is wrapped up in a piece of sailcloth (deerskin was formerly used), laid upon a flat sled, and dragged out by a small party of people—perhaps the immediate relatives of the deceased, though we never happened to see one of these funeral processions except from a distance—to the cemetery, the place where “they sleep on the ground.” This place at Utkiavwĭñ is a rising ground about a mile and a half east of the village, near the head of the southwest branch of the Isûtkwa lagoon. At Nuwŭk the main cemetery is at “Nexeurá,” between the village and Pernyû. The bodies are laid out upon the ground without any regular arrangement apparently, though it is difficult to be sure of this, as most of the remains have been broken up and scattered by dogs and foxes. With a freshly wrapped body it is almost impossible to tell which is the head and which the feet. We unfortunately never
noticed whether the heads were laid toward any particular point of the compass, as has been observed in other localities. Dr. Simpson says that the head is laid to the east at Point Barrow. Various implements belonging to the deceased are broken and laid beside the corpse, and the sled is sometimes broken and laid over it. Sometimes, however, the latter is withdrawn a short distance from the cemetery and left on the tundra for one moon, after which it is brought back to the village. Most people do not seem to be troubled at having the bodies of their relatives disturbed by the dogs or other animals, 581 but we know of one case where the parents of two children who died very nearly at the same time, finding that the dogs were getting at the bodies, raised them on stages of driftwood about 4 or 5 feet high. Similar stages were observed by Hooper at Plover Bay; 582 but this method of disposing of the dead appears to have gone out of use at the present day, since Dall 583 describes the ordinary Siberian method of laying out the dead in ovals of stone as in use at Plover Bay at the time of his visit. The cemetery at Utkiavwĭñ is not confined to the spot I have mentioned, though most of the bodies are exposed there. A few bodies are also exposed on the other side of the lagoon, and one body, that of a man, was laid out at the edge of the higher tundra, about a mile due east from the station. The body was covered with canvas, staked down all round with broken paddles, and over it was laid a flat sledge with one runner broken. 584 At one end of the body lay a wooden dish, and under the edge of the canvas were broken seal-darts and other spears. The body lay in an east and west line, but we could not tell which end was the head. All sorts of objects were scattered round the cemetery—tools, dishes, and even a few guns—though we saw none that appeared to have been serviceable when exposed, except one Snider rifle. If, as is the case among Eskimo in a good many other places, all the personal property of the deceased is supposed to become unclean and must be exposed with him, it is probable that his friends manage to remove the more valuable articles before he is actually dead. 585
The method of disposing of the dead varies slightly among the Eskimo in different localities, but the weapons or other implements belonging to the deceased are always laid beside the corpse. The custom at Smith Sound, as described by Bessels, 586 is remarkably like that at Point Barrow. The corpse was wrapped in furs, placed on a sledge, and dragged out and buried in the snow with the face to the west. The sledge was laid over the body and the weapons of the deceased were deposited beside it. Unlike the Point Barrow natives, however, they usually cover the body with stones. In the same passage Dr. Bessels describes a peculiar symbol of mourning, not employed, so far as I can learn, elsewhere. The male mourners plugged up the right nostril with hay and the females the left, and these plugs were worn for several days. The custom of covering the body with stones appears to be universally prevalent east of the Mackenzie region. 587 The bodies seen by Dr. Richardson in the delta of the Mackenzie were wrapped in skins and loosely covered with driftwood, 588 and a similar arrangement was noticed at Kotzebue Sound by Beechey, who figures 589 a sort of little wigwam of driftwood built over the dead man. At Port Clarence Nordenskiöld 590 saw two corpses “laid on the ground, fully clothed, without protection of any coffin, but surrounded by a close fence consisting of a number of tent-poles driven crosswise into the ground. Alongside one of the corpses lay a kayak with oars, a loaded double-barreled gun with locks at half- cock and caps on, various other weapons, clothes, tinder-box, snowshoes, drinking-vessels, two masks, * * * and strangely shaped animal figures.” On the Siberian coast the dead are sometimes burned. 591 Nordenskiöld believes that the coast Chukches have perhaps begun to abandon the custom of burning the dead, but I am rather inclined to think that is a custom of the “deermen,” which the people of the coast of pure or mixed Eskimo blood never fully adopted. Dall, indeed, was explicitly informed that the custom was only used with the bodies of “good” men, and at the time of Nordenskiöld’s visit he
found it “at least certain that the people of Pitlekaj exclusively bury their dead by laying them out on the tundra.” The body is surrounded by an oval of stones, but apparently not covered with them as in the east. 592 The Krause brothers observed by the bodies, besides “die erwähnten Geräthschaften” [Lanzen, Bogen und Pfeile für die Männer, Koch- und Hausgeräthe für die Weiber], “unter einen kleinen Steinhaufen ein Hunde-, Renthier-, Bären- oder Walross- Schädel.” This custom shows a curious resemblance to that described by Egede 593 in Greenland: “When little Children die and are buried, they put the Head of a Dog near the Grave, fancying that Children, having no Understanding, they can not by themselves find the Way, but the Dog must guide them to the Land of the Souls.” The body is usually laid out at full length upon the ground. Among the ancient Greenlanders, 594 however, and in the Yukon region the body was doubled up. In the latter region the body was laid on its side in a box of planks four feet long and raised on four supports 595 or wrapped up in mats and covered with rocks or driftwood. 596 The custom of inclosing the dead in a short coffin, to judge from the figures given by the latter writer in P1. vi. of his report, appears also to prevail at the mouth of the Kuskokwim. In the island of Kadiak, according to Dall and Lisiansky, 597 the dead were buried.
Two lines near the bottom of the page were reversed in print: GOVERNMENT. In the family.— I can hardly do better than quote Dr. Simpson’s words, already referred to (op. cit. page 252), on this subject: “A man seems to have unlimited authority in his own hut.” Nevertheless, his rule seems to be founded on respect and mutual agreement, rather than on despotic authority. The wife appears to be consulted, as already stated, on all important occasions, and, to quote Dr. Simpson again (ibid.): “Seniority gives precedence when there are several women in one hut, and the sway of the elder in the direction of everything connected with her duties seems never disputed.” When more than one family inhabit the same house the head of each family appears to have authority over his own relatives, while the relations between the two are governed solely by mutual agreement. In the village.— These people have no established form of government nor any chiefs in the ordinary sense of the word, but appear to be ruled by a strong public opinion, combined with a certain amount of respect for the opinions of the elder people, both men and women, and by a large number of traditional observances like those concerning the whale fishery, the deceased, etc., already described. In the ordinary relations of life a person, as a rule, avoids doing anything to his neighbor which he would not wish to have done to himself, and affairs which concern the community as a whole, as for instance
their relations with us at the station, are settled by a general and apparently informal discussion, when the opinion of the majority carries the day. The majority appears to have no means, short of individual violence, of enforcing obedience to its decisions, but, as far as we could see, the matter is left to the good sense of the parties concerned. Respect for the opinions of elders is so great that the people may be said to be practically under what is called “simple elder rule.” 598 Public opinion has formulated certain rules in regard to some kinds of property and the division of game, which are remarkably like those noticed among Eskimo elsewhere, and which may be supposed to have grown up among the ancestors of the Eskimo, before their separation. For instance, in Greenland, 599 “Anyone picking up pieces of driftwood or goods lost at sea or on land was considered the rightful owner of them; and to make good his possession he had only to carry them up above high-water mark and put stones upon them, no matter where his homestead might be.” Now, at Point Barrow we often saw the natives dragging driftwood up to the high-water mark, and the owner seemed perfectly able to prove his claim. Lieut. Ray informs me that he has seen men mark such sticks of timber by cutting them with their adzes and that sticks so marked were respected by the other natives. On one occasion, when he was about to have a large piece of drift-timber dragged up to the station, a woman came up and proved that the timber belonged to her by pointing out the freshly cut mark. I have myself seen a native claim a barrel which had been washed ashore, by setting it up on end. As far as we could learn, the smaller animals, as for instance, birds, the smaller seals, reindeer, etc., are the property of the hunter, instead of being divided as in some other localities, for example at Smith Sound, 600. The larger seals and walruses appeared to be divided among the boat’s crew, the owner of the boat apparently keeping the tusks of the walrus and perhaps the skin. A bear, however, both flesh and skin, is equally divided among all who in any way had a hand in the killing. We learned this with certainty from
having to purchase the skin of a bear killed at the village, where a number of men had been engaged in the hunt. When a whale is taken, as I have already said, the whalebone is equally divided among the crews of all the boats in sight at the time of killing. All comers, however, have a right to all the flesh, blubber, and blackskin that they can cut off. 601 Dr. Rink, in describing the social order of the ancient Greenlanders, 602 says: “Looking at what has been said regarding the rights of property and the division of the people into certain communities, in connection with the division of property into the classes just given, we are led to the conclusion that the right of any individual to hold more than a certain amount of property was, if not regulated by law, at least jealously watched by the rest of the community, and that virtually the surplus of any individual or community, fixed by the arbitrary rate which tradition or custom had assigned, was made over to those who had less.” At Point Barrow, however, the idea of individual ownership appears to be much more strongly developed. As far as we could learn, there is no limit to the amount of property which an individual, at least the head of a family, may accumulate. Even though the whalebone be, as already described, divided among all the boats’ crews “in at the death,” no objection is made to one man buying it all up, if he has the means, for his own private use. This has given rise to a regular wealthy and aristocratic class, who, however, are not yet sufficiently differentiated from the poorer people to refuse to associate on any terms but those of social equality. The men of this class are the umialiks, a word which appears in many corrupted forms on the coast of Western America and is often supposed to mean “chief.” Dr. Simpson 603 says: “The chief men are called O-mé-liks (wealthy),” but “wealthy” is an explanation of the position of these men, and not a translation of the title, which, as we obtained it, is precisely the same as the Greenland word for owner of a boat, umialik (from umia(k), and the
termination lĭk or lĭ-ñ. This is one of the few cases in which the final k is sounded at Point Barrow as in Greenland). Dr. Rink has already observed 604 that the word used by Simpson “no doubt must be the same as the Greenlandish umialik, signifying owner of a boat,” and as I heard the title more than once carefully pronounced at Point Barrow it was the identical word. The umialiks, as Simpson says, 605 “have acquired their position by being more thrifty and intelligent, better traders, and usually better hunters, as well as physically stronger and more daring.” 606 They have acquired a certain amount of influence and respect from these reasons, as well as from their wealth, which enables them to purchase the services of others to man their boats, but appear to have absolutely no authority outside of their own families. 607 Petroff 608 considers them as a sort of “middlemen or spokesmen,” who make themselves “prominent by superintending all intercourse and traffic with visitors.” This sort of prominence, however, appears to have been conferred upon them by the traders, who, ignorant of the very democratic state of Eskimo society, naturally look for “chiefs” to deal with. They pick out the best looking and best dressed man in the village and endeavor to win his favor by giving him presents, receiving him into the cabin, and conducting all their dealings with the natives through him. The chief, thus selected, is generally shrewd enough to make the most of the greatness thrust upon him, and no doubt often pretends to more influence and power than he actually possesses. 609 As to the story of the whalemen, that the “chieftainship” is the reward of the best fighter, who holds it like a “challenge cup,” subject to being called out at any time to defend his rank in a duel, as far as concerns Point Barrow, this is a sheer fable, perhaps invented by the Eskimo to impose upon the strangers, but more likely the result of misunderstanding and a vivid imagination on the part of the whites. Among umialiks, one or two appear to have more wealth and influence than the rest. Tcuñaura in Utkiavwĭñ and the
late Katiga at Nuwŭk were said, according to Captain Herendeen, to be “great umialiks” and Tcuñaura was always spoken of as the foremost man in Utkiavwĭñ. We knew of one party coming up from Sidaru with presents for Tcuñaura, and were informed that the other Eskimo never sold to him, but only gave him presents. It was also said that Katiga’s infant son would one day be a “great umialik.” All these men are or have been captains of whaling umiaks, and the title umialiks appears to be applied to them in this capacity, since many of the poorer men, who, as far as we could learn, were not considered umialiks, own umiaks which they do not fit out for whaling, but use only to transport their families from place to place in the summer. RELIGION. General ideas.— It was exceedingly difficult to get any idea of the religious belief of the people, partly from our inability to make ourselves understood in regard to abstract ideas and partly from ignorance on our part of the proper method of conducting such inquiries. For instance, in trying to get at their ideas of a future life, we could only ask “Where does a man go when he dies?” to which we, of course, received the obvious answer, “To the cemetery!” Moreover, such a multitude of other and easier lines of investigation presented themselves for our attention that we were naturally inclined to neglect the difficult field of religion, and besides under the circumstances of our intercourse it was almost impossible to get the attention of the natives when their minds were not full of other subjects. Nevertheless, many of the fragments of superstition and tradition that we were able to collect agree remarkably with what has been observed among the Eskimo elsewhere, so that it is highly probable that their religion is of the same general character as that of the Greenlanders, namely, a belief in a multitude of supernatural beings,
who are to be exorcised or propitiated by various observances, especially by the performances of certain specially gifted people, who are something of the nature of wizards. So much has been written by many authors about these wizards or “doctors,” the angekut of the eastern Eskimo, the so-called “shamans” of Alaska and Siberia, that I need make no special reference to their writings except where they happen to throw light on our own observations. Dr. Simpson succeeded in obtaining more information concerning the religious belief of these people than our party was able to do, and his observations, 610 to which ours are in some degree supplementary, tend to corroborate the conclusion at which I have arrived. Our information in regard to the special class of wizards was rather vague. We learned that many men in the village, distinguishable from the rest by no visible characteristics, were able to heal the sick, procure good weather, favorable winds, plenty of game, and do other things by “talking” and beating the drum. We did not learn the number of these men in either village, but we heard of very many different men doing one or the other of these things, while others of our acquaintance never attempted them. Neither did we learn that any one of these men was considered superior to the rest, as appears to be the case in some regions, nor how a man could attain this power. Some of these men, who appeared to give particular attention to curing the sick, called themselves “tû´ktĕ” (“doctor”), but, probably for want of properly directed inquiries, we did not learn the Eskimo name of these people. We were definitely informed, however, that their “talk,” when treating disease or trying to obtain fair weather, etc., was addressed to “tu´ɐña,” or a supernatural being. This name, of course, differs only in dialectic form from that applied in other places to the universal familiar spirits of Eskimo superstition. We at first supposed that “tuɐña” meant some particular individual demon, but Dr. Simpson is probably right in saying that the Point Barrow natives, like the rest of the Eskimo, recognize a host of tuɐñain, since “tuɐña” was described to us under a variety of forms.
Most of the natives whom we asked if they had seen tuɐña, said that they had not, but that other men, mentioning certain “doctors,” had seen him. One man, however, said that he had seen tuɐña in the kûdyĭgĭ, when the people “talked” sitting in the dark, with their heads bowed and faces covered, and tuɐña came with a noise like a great bird. 611 He had raised his head and saw tuɐña, like a man with bloodless cheeks. 612 Tuɐña again was called “a bad man, dead” (apparently a ghost), sometimes as large as a man and sometimes dwarfish, sometimes a fleshless skeleton, while one man, to describe him, made the same grimace that a white man would use to indicate a hobgoblin, with staring eyes, gaping mouth, and hands outstretched like claws. Apparently “tuɐña” in conversation with us was used to designate all the various supernatural objects of their belief, ghosts as well as familiar spirits. For instance, in Greenland, according to Rink, 613 a ghost “manifests himself by whistling or singing in the ears.” Now, Lieut. Ray was walking rapidly one day in the winter with an Eskimo and his wife, and the woman suddenly stopped and said she “heard tuɐña”—that he made a noise like singing in the ears. The people generally have a great dread of “tuɐña,” who they say would kill them, and are very averse to going out alone in the dark. One of each party that came over from the village in the evening usually carried a drawn knife, preferably one of the large double- edged knives, supposed to be Siberian and already described, in his hand as defense against tuɐña, and a drawn knife was sometimes even carried in the daylight “nanumunlu tuɐñamunlu,” “for bear and demon.” Notwithstanding their apparently genuine dread of “tuɐña,” they are by no means averse to talking or even joking about him. The knife also serves as a protection against the aurora, which most of them agree is bad, and when bright likely to kill a person by striking him in the back of the neck. However, brandishing the knife at it will keep it off. Besides, as a woman told me one night, you can drive off a “bad” aurora by throwing at it dog’s excrement and urine. 614
Lieut. Ray saw in one of the houses in Utkaiwiñ, a contrivance for frightening away a “tuɐña” from the entrance to a house should he try to get in. The man had hung in the trapdoor the handle of a seal- drag by means of a thong spiked to the wall with a large knife, and told Lieut. Ray that if “tuɐña” tried to get into the house he would undoubtedly catch hold of the handle to help himself up, which would pull down the knife upon his head and frighten him off. We never had an opportunity of witnessing the ceremony of summoning “tuɐña,” nor did we ever hear of the ceremony taking place during our stay at the station, but we were fortunate enough to observe several other performances, though they do not appear to be frequent. The ceremony of healing the sick and the ceremonies connected with the whale-fishery have already been described. On the 21st of February, 1883, Lieut. Ray and Capt. Herendeen happened to be at the village on time to see the tuɐña, who had been causing the bad weather, expelled from the village. Some of the natives said the next day that they had killed the tuɐña, but they said at the same time he had gone “a long way off.” When Lieut. Ray reached the village, women were standing at the doors of the houses armed with snow-knives and clubs with which they made passes over the entrance when the people inside called out. He entered one house and found a woman vigorously driving the tuɐña out of every corner with a knife. They then repaired to the kûdyĭgi, where there were ten or twelve people, each of whom, to quote from Lieut. Ray’s note book, “made a charge against the evil spirit, telling what injuries they had received from it.” Then they went into the open air, where a fire had been built in front of the entrance, and formed a half circle around the fire. Each then went up and made a speech, bending over the fire (according to Simpson, who describes a similar ceremony at Nuwŭk on p. 274 of his paper, coaxing the tuɐña to come under the fire to warm himself). Then they brought out a large tub full of urine, to which, Simpson says, each man present had contributed, and held it ready near the fire, while two men stood with their rifles in readiness, and a boy stood near the fire with a large stone in his hands, bracing himself firmly
with his feet spread apart for a vigorous throw. Then they chanted as follows (the words of this chant were obtained afterward by the writer): Tâk tâk tâk tohâ! Nìju´a hâ! He! he! he! Haiyahe! Yaiyahe! Hwi! And instantly the contents of the tub were dashed on the fire, the stone thrown into the embers, and both men discharged their rifles, one into the embers, and one into the cloud of steam as it rose. Then all brushed their clothes violently and shouted, and the tuɐña was killed. By a fortunate coincidence, the next day was the finest we had had for a long time. Sacrifices are also occasionally made to these supernatural beings as in Greenland “gifts were offered to the inue of certain rocks, capes and ice firths, principally when traveling and passing those places.” 615 Capt. Herendeen, in the fall of 1882, went to the rivers in company with one of the “doctors.” When they arrived at the river Kuaru, where the latter intended to stay for the fishing, he got out his drum and “talked” for a long time, and breaking off very small pieces of tobacco threw them into the air, crying out, “Tuɐña, tuɐña, I give you tobacco! give me plenty of fish.” When they passed the dead men at the cemetery, he gave them tobacco in the same way, asking them also for fish. 616 We noticed but few other superstitious observances which have not been already described. As in Greenland and elsewhere, superstition requires certain persons to abstain from certain kinds of food. For instance, Mûñialu, and apparently many others, were not permitted to eat the burbot, another man was denied ptarmigan, and a woman 617 at Nuwŭk was not allowed to
eat “earth food,” that is, anything which grew upon the ground. Lieut. Ray also mentions a man who was forbidden bear’s flesh. 618 We observed some traces of the superstition concerning the heads of seals and other marine animals taken in the chase, which has been noticed elsewhere. Crantz says: 619 “The heads of seals must not be fractured, nor must they be thrown into the sea, but be piled in a heap before the door, 620 that the souls of the seals may not be enraged and scare their brethren from the coast.” And Capt. Parry found that at Winter Island they carefully preserved the heads of all the animals killed during the winter, except two or three of the walrus which he obtained with great difficulty. The natives told him that they were to be thrown into the sea in the summer, but at Iglulik they readily sold them before the summer arrived. 621 I tried very hard to get a full series of skulls from the seals taken at Utkiavwĭñ in the winter of 1882-’83, but though I frequently asked the natives to bring them over for sale, they never did so, till at last one young woman promised to bring me all I wanted at the price of half a pound of gunpowder a skull. Nevertheless, she brought over only two or three at that price. We did not observe what was done with the skulls, but frequently observed quantities of the smaller bones of the seals carefully tucked away in the crevices of the ice at some distance from the shore. We had comparatively little difficulty in obtaining skulls of the walrus, but I observed that the bottom of Tûseráru, the little pond at the edge of the village, was covered with old walrus skulls, as if they had been deposited there for years. The superstition appears to be in full force among the Chukches, who live near the place where the Vega wintered. Nordenskiöld was unable to purchase a pair of fresh walrus heads at the first village he visited, though the tusks were offered for sale the next day 622 and at Pitlekaj. 623 “Some prejudice * * * prevented the Chukches from parting with the heads of the seal, though * * * we offered a high price for them. ‘Irgatti’ (to-morrow) was the usual answer. But the promise was never kept.”
Fig. 421.—Whale flaked from glass. Amulets.— Like the Greenlanders 624 and other Eskimos, they place great reliance on amulets or talismans, which are carried on the person, in the boat, or even inserted in weapons, each apparently with some specific purpose, which indeed we learned in the case of some of those in the collection. Like the amulets of the Greenlanders, they appear to be 625 “certain animals or things which had belonged to or been in contact with certain persons (e.g., the people of ancient times, or fortunate hunters) or supernatural beings,” and “objects which merely by their appearance recalled the effect expected from the amulet, such as figures of various objects.” To the latter class belong the rudely flaked flint images of whales, already mentioned, and probably many of the other small images of men and animals already described, especially those fitted with holes for strings to hang them up by. The flint whale is a very common amulet, intended, as we understood, to give good luck, in whaling, and is worn habitually by many of the men and boys under the clothes, suspended around the neck by a string. The captain and harpooner of a whaling crew also wear them as pendants on the fillets already described, and on the breast of the jacket. We obtained five of these objects, all of very nearly the same shape, but of different materials and varying somewhat in size. Fig. 421 represents one of these (No. 56703 [208] from Utkiavwĭñ) made of a piece of hard colorless glass, probably a fragment of a ship’s “deadlight.” It is rather roughly flaked into a figure of a “bowhead” whale, 3.4 inches long, as seen from above and very much flattened with exaggerated flukes. The flippers were rudely indicated in the outline, but the left one is broken off.
Fig. 422.—Whale flaked from red jasper. No. 89613 [771] from Utkiavwĭñ is a very similar image, 2.4 inches long, which perhaps is of the same material, though it may be made of rock crystal. No. 56707 [159] from Utkiavwĭñ is a very small whale (1.4 inches long), chipped in large flakes out of a water-worn pebble of smoky quartz, while No. 89577 [939] Fig. 422, from the same village, which is a trifle larger (2 inches long), is made of dark crimson jasper. The large black flint whale, No. 56683 [61], also from Utkiavwĭñ, which is 3.9 inches long, is the rudest of all the figures of the whales. It is precisely the shape of the blade of a skin scraper, except for the roughly indicated flukes. Fig. 423 (No. 89524 [1299] from Utkiavwĭñ) is a rude wooden image of the same animal, 3½ inches long, very broad and flat- bellied. It is smoothly carved and has a fragment of sky-blue glass inlaid to represent the left eye and a bit of iron pyrites for the right. The flukes have been split wholly off and fastened on with a lashing of narrow whalebone passing through a vertical hole in the “small” and round the edge of the flukes. The flukes themselves have been split across and appear to have been doweled together. This shows that the owner attached considerable value to the object, or he would not have taken the trouble to mend it when another could have been so easily whittled out. In the middle of the belly is an oblong cavity, containing something which probably adds greater power to the charm. What this is can not be seen, as a band of sealskin with the hair shaved off has been shrunk on round the hinder half of the body and secured by a seam on the right side. A double turn of sinew braid is knotted round the middle of the body, leaving two ends which are tied together in a loop, showing that this object was meant to be attached somewhere about the person.
Fig. 423.—Ancient whale amulet, of wood. To this class also probably belong the skins or pieces of animals worn as amulets, probably with a view of obtaining the powers of the particular animal, as in so many cases in the stories related in Rink’s Tales and Traditions. We frequently saw men wearing at the belt bunches of the claws of the bear or wolverine, or the metacarpal bones of the wolf. 626 The head or beak of the gull or raven 627 is also a common personal amulet, and one man wore a small dried flounder. 628 We collected a number of these animal amulets to be worn on the person, but only succeeded in learning the special purpose of one of them, No. 89532 [1307], from Utkiavwĭñ, which was said to be intended to give good luck in deer hunting. It is a young unbranched antler of a reindeer, 6 inches long, and apparently separated from the skull at the “bur,” with the “velvet” skin still adhering, though most of the hair is worn off except at the tip. A bit of sinew is tied round the base. No. 89522 [1573], from Utkiavwĭñ, is an amulet consisting of the last three joints of the foot of a reindeer fawn, with the skin and hoof and about 1½ inches of tendon attached behind, through a hole in the end of which is knotted about 3 inches of seal thong. No. 89525 [1314] from the same village, is a precisely similar charm. No. 89699 [779] from Utkiavwĭñ , is the subfossil incisor tooth of some ruminant with a hole drilled through the root for a string to hang it up by. It was said to be the tooth of the “ug’ru´nû,” a large animal, long extinct. As the natives said, “Here on the land are none, only
the bones remain.” No. 89743 [1110], from Utkiavwĭñ, is a molar tooth of the same animal, probably, weathered and old, with a hole freshly drilled through one root and a long piece of sinew braid with the ends knotted together looped into it. There are also in the collection two very old teeth which probably were inclosed in little sacks of skin and worn as amulets. No. 89698 [1580], from Utkiavwĭñ, is the tusk of a very young walrus, only 2½ inches long, and No. 89452 [1148] from Utkiavwĭñ, is the canine tooth of a polar bear. No. 56547 [656], from the same village, is a similar tooth. 629 The only amulet attached to a weapon, which we collected, is the tern’s bill, already alluded to, placed under the whalebone lashing on the seal-spear, No. 89910 [1694]. Perhaps the idea of this charm is that the spear should plunge down upon the seal with as sure an aim as the tern does upon its prey. 630 A number of amulets of this class are always carried in the whaling- umiak. I have already mentioned the wolf-skulls, stuffed ravens and eagles, fox-tails 631 and bunches of feathers used for this purpose. Most of these charms are parts of some rapacious animal or bird, but parts of other animals seem to have some virtue on these occasions. For instance, I noticed the axis vertebra of a seal in one whaling- umiak, and we collected a rudely stuffed skin of a godwit (Limosa lapponica baueri), which, we were informed, was “for whales.” This specimen (No. 89526 [1328], Fig. 424, from Utkiavwĭñ) is soiled and ragged, and has a stick thrust through the neck to hold it out. The neck is wrapped around with a narrow strip of whalebone and some coarse thread, part of which serves to lash on a slip of wood, apparently to splice the stick inside. A bit of white man’s string is passed around the body and tied in a loop to hang it up by. This charm is perhaps to keep the boat from capsizing, since Crantz says that the Greenlanders “like to fasten to their kajak a model of it * * * or only a dead sparrow or snipe, or a bit of wood, stone, some feathers or hair, that they may not overset” (vol. 1, p. 216), and
perhaps the bone of a marine animal, like the seal, is to protect the crew from drowning should the boat upset, after all. No. 89529 [1150] from Utkiavwĭñ is a bunch of feathers to be carried in the boat. It consists of nine wing feathers of the golden eagle, four tied in a bunch with a bit of sinew round the quills, four tied up with one end of the short bit of seal thong which serves to tie the whole bundle together, one of which has all the light-colored parts of the feather stained with red ocher, and a single feather shaft carefully wrapped up in a piece of entrail and wound spirally with a piece of sinew braid. Fig. 424.—Amulet of whaling; stuffed godwit. No. 89527-8 [1327] from Utkiavwĭñ is the charm which will secure good success in deerhunting if it is hung up outside of the snow house in which the family is encamped. It consists of two roughly stuffed skins of the black bellied plover (Charadrius squatarola), each with a stick run through the body so that one end supports the neck and the other the tail, and the necks wound with sinew. One has no head. A string of sinew braid is tied around the body of each, so as to leave a free end at the back, to which is fastened a little cross piece of bone, by which it may be secured to a becket. Like the rest of the amulets in the collection this has evidently seen service, being very old, worn, and faded. The other class of amulets, namely objects which have belonged to or been in contact with certain persons or supernatural beings, or I may add apparently certain localities, is represented by a number of specimens. To the custom of using such things as amulets, we
Fig. 425.—Amulet consisting of ancient jade adz. Fig. 426.—Little box containing amulet for whaling. undoubtedly owe the preservation of most of the ancient weapons and other implements, especially those made of wood, bone, or other perishable substances, like the ancient harpoon heads already described, one of which, No. 89544 [1419], is still attached to the belt on which it was worn. Fig. 425, No. 56668 [308], from Utkiavwĭñ is one of the ancient black jade adzes 5.1 inches, slung with thong and whalebone, making a becket by which it can be hung up. We did not learn the history of this amulet, which at the time of collecting it was supposed to be a net sinker. There would, however, be no reason for using so valuable an object for such a purpose, when a common beach pebble would do just as well, unless it was intended as a charm to insure success in fishing. It may even have been carried as a charm on the person, since we afterwards saw a still more bulky object used for such a purpose. Such an object seems rather heavy to be carried on the person, but a well known man in Utkiavwĭñ always carried with him when he went sealing a large pear-shaped stone, which must have weighed upwards of two pounds, suspended somewhere about his person. It is not unlikely that this stone acquired its virtue as an amulet from having been a sinker used by some lucky fisherman in former time or in a distant country. Mr. H. W. Henshaw has already referred to the resemblance of this amulet to the plummet-like “medicine stones” of some of our Indians. 632 Fig. 426, (No. 89534 [1306] from Utkiavwĭñ) is an amulet for success in whaling. It consists of three little irregular water-worn fragments of amber carefully wrapped in a bit of parchment and inclosed in a little wooden box 1½ inches long, made of two semicylindrical bits of cottonwood, with the flat faces hollowed out and put together and fastened up by three turns of
Fig. 427.— Amulet for catching fowl with bolas. sinew braid round the middle, tied in a loose knot. The box is old and brown from age and handling. We heard of other pieces of amber and earth (“nuna”) worn as amulets, wrapped up in bits of leather and hung on the belt. No. 89533 [1247], from Utkiavwĭñ, is simply a nearly square pebble, 1.4 inches long, of dark red jasper, slung in a bit of sinew braid so that it can be hung on the belt. Fig. 427 (No. 89525 [1308] from Utkiavwĭñ) is some small object, placed in the center of the grain side of a square bit of white sealskin, the edges of which are folded up around it and tied tightly round with deer sinew, so as to make a little round knob. I collected this amulet, and was particularly informed how it was to be used. If it be fastened on the right shoulder it will insure success in taking ducks with the “bolas.” Fig. 428 (No. 89535 [1244] from Utkiavwĭñ) is an amulet whose history we did not learn. It is a little oblong box 3.3 inches long, carved from a block of cottonwood, with a flat cover tied on with nine turns of sinew braid, and contains twenty-one dried humble-bees, which it was said came from the river Kulugrua. The natives have a great dread, apparently superstitious, of these bees and the large gadflies (Œstens tarandi), one of which I have seen scatter half a dozen people. A man one day caught one of these, and whittled out a little box of wood, in which he shut the insect up and tied it up with a shred of sinew, telling Capt. Herendeen that it was “tuɐñamun,” for “tuɐña.” Fig. 428.—Box of dried bees—amulet.
A small lump of indurated gravel (No. 56725) [273] was one day brought over from Utkiavwĭñ, with the story that it was a “medicine” for driving away the ice. The man who uses this charm stands on the high bank at the village, and breaking off grains of the gravel throws them seaward. This will cause the ice to move off from the shore. The essential identity of the amulets of the Point Barrow natives with those used by the Eskimo elsewhere is shown by the following passages from other writers. Egede says: 633 A Superstition very common among them is to load themselves with Amulets or Pomanders, dangling about their Necks and Arms, which consist in some Pieces of old Wood, Stones or Bones, Bills and Claws of Birds, or Anything else which their Fancy suggests to them. Crantz says: 634 They are so different in the amulets or charms they hang on people, that one laughs at another’s. These powerful preventives consist in a bit of old wood hung around their necks, or a stone, or a bone, or a beak or claw of a bird, or else a leather strap tied round their forehead, breast, or arm. Parry speaks 635 of what he supposes were amulets at Iglulik, consisting of teeth of the fox, wolf, and musk-ox, bones of the “kablĕĕarioo” (supposed to be the wolverine), and foxes’ noses. Kumlien says 636 that at Cumberland Gulf, “among the many superstitious notions, the wearing of charms about the person is one of the most curious. These are called angoouk or amusit, and may be nothing but pieces of bone or wood, birds’ bills or claws, or an animal’s teeth or skin.” A little girl “had a small envelope of sealskin that was worn on the back of her inside jacket” containing two small stones. Such little pockets of skin sewed to the inner jacket are very common at Point Barrow, but we did not succeed in any case in learning their contents. At Kotzebue Sound, Beechey saw ravens’ skins on which the natives set a high value, while the beaks and claws of these birds were attached to their belts and headbands. 637 Petitot describes 638 the amulets used in the Mackenzie district, in
the passage already quoted, as “défroques empaillées de corbeau, de faucon ou d’hermine.” It is not likely that the use of these is confined to the women, as his words, “Elles y portent,” would seem to imply. Among the sedentary Chukches of Siberia amulets were seen consisting of wooden forks and wood or ivory carvings. 639 A wolf’s skull, hung up by a thong; the skin, together with the whole cartilaginous portion of a wolf’s nose, and a flat stone, are also mentioned. 640 Capt. Holm also found wonderfully similar customs among the East Greenlanders. He says, 641 “bære alle Folk Amuletter af de mest forskjelligartede Ting” to guard against sickness and to insure long life, and also for specific purposes. The men wear them slung round the neck or tied round the upper arm, the women in their knot of hair or “i Snippen foran paa Pelsen.” Footnotes 411-641 411. Nordenskiöld, Vega, vol. 2, pp. 122, and Fig. 1, p. 117. 412. Crantz describes the process of preparing boat covers as follows: “The boat skins are selected out of the stoutest seals’ hides, from which the fat is not quite taken off; they roll them up, and sit on them, or let them lie in the sun covered with grass several weeks, ’till the hair will come off.” History of Greenland, vol. 1, p. 167. 413. Gilder describes a similar process of manufacturing these lines at Hudson’s Bay. (Schwatka’s Search, p. 176.) 414. W. J. Sollas, in Jour. Anthrop. Inst. of Great Britain and Ireland, vol. 9, pp. 329-336. 415. Nordenskiöld’s figures, Vega, vol. 2, p. 123. 416. Parry’s Second Voy., pl. opposite p. 548, Fig. 5. 417. Vega, vol. 1, p. 493.
Welcome to our website – the perfect destination for book lovers and knowledge seekers. We believe that every book holds a new world, offering opportunities for learning, discovery, and personal growth. That’s why we are dedicated to bringing you a diverse collection of books, ranging from classic literature and specialized publications to self-development guides and children's books. More than just a book-buying platform, we strive to be a bridge connecting you with timeless cultural and intellectual values. With an elegant, user-friendly interface and a smart search system, you can quickly find the books that best suit your interests. Additionally, our special promotions and home delivery services help you save time and fully enjoy the joy of reading. Join us on a journey of knowledge exploration, passion nurturing, and personal growth every day! ebookbell.com

More Related Content

PDF
Determinants Of Grammatical Variation In English Gnter Rohdenburg Editor Brit...
mayouhertlyp
 
PDF
Determinants Of Grammatical Variation In English Britta Mondorf
tcxdzoed223
 
PDF
Describing And Modeling Variation In Grammar Andreas Dufter Editor Jrg Fleisc...
ghydayandja30
 
PDF
Variation And Change In Morphology Selected Papers From The 13th Internationa...
outinsibala
 
PDF
Syntax And Variation Reconciling The Biological And The Social 1st Leonie Mea...
nirschlaasma
 
PDF
Syntactic Variation And Genre 1st Edition Heidrun Dorgeloh Ed
ojqntkacox3296
 
PDF
Syntactic Variation And Genre Heidrun Dorgeloh Editor Anja Wanner Editor
inlowschub7r
 
PDF
Measuring Grammatical Complexity Frederick J Newmeyer Laurel B Preston
persondhass
 
Determinants Of Grammatical Variation In English Gnter Rohdenburg Editor Brit...
mayouhertlyp
 
Determinants Of Grammatical Variation In English Britta Mondorf
tcxdzoed223
 
Describing And Modeling Variation In Grammar Andreas Dufter Editor Jrg Fleisc...
ghydayandja30
 
Variation And Change In Morphology Selected Papers From The 13th Internationa...
outinsibala
 
Syntax And Variation Reconciling The Biological And The Social 1st Leonie Mea...
nirschlaasma
 
Syntactic Variation And Genre 1st Edition Heidrun Dorgeloh Ed
ojqntkacox3296
 
Syntactic Variation And Genre Heidrun Dorgeloh Editor Anja Wanner Editor
inlowschub7r
 
Measuring Grammatical Complexity Frederick J Newmeyer Laurel B Preston
persondhass
 

Similar to The Limits Of Syntactic Variation Theresa Biberauer (20)

PDF
marathi grammar by Rajeshwari Pandharipande .pdf
CantodaSereia
 
PDF
Variation And Change In Spanish 1st Edition Ralph Penny
solasbaney4d
 
PDF
Grammaticalization And Pragmatics Facts Approaches Theoretical Issues Corinne...
gnaftimlaker
 
PDF
Standard Negation The Negation Of Declarative Verbal Main Clauses In A Typolo...
suncakakuda
 
PDF
On Diversity And Complexity Of Languages Spoken In Europe And North And Centr...
jxrxyrqm951
 
PDF
The Variability Of Current World Englishes Eugene Green Editor Charles F Meye...
lukacovitzr4
 
PPTX
Anthrtopological linguistics1
Diego Morales
 
PDF
Scopes of linguistic description 1
Bel Abbes Neddar
 
PDF
The Lexicon Ebook James Pustejovsky Olga Batiukova
carpesvellen
 
PDF
I01066062
IOSR Journals
 
DOC
Language structure is partly by social structure
SpanishinBuenosAires
 
PDF
Variation In Datives A Microcomparative Perspective Beatriz Fernndez
marathokala90
 
PDF
The Structure Of Discoursepragmatic Variation Heike Pichler
buyensthike
 
PDF
The Structure Of Discoursepragmatic Variation Heike Pichler
buyensthike
 
PDF
The Syntactic Variation Of Spanish Dialects Illustrated Ngel J Gallego
tesinzimvo43
 
PDF
Perspectives on Variation Sociolinguistic Historical Comparative Nicole Delbe...
brodapatatu4
 
PDF
The Diachrony Of Verb Meaning Aspect And Argument Structure Elly Van Gelderen
hvoxizmp464
 
PDF
Newest Trends In The Study Of Grammaticalization And Lexicalization In Chines...
shottdulla25
 
PDF
Information Structure Theoretical Typological And Experimental Perspectives M...
conranulz15
 
PPTX
language variation and change Presentation1.pptx
DESTAWWAGNEW
 
marathi grammar by Rajeshwari Pandharipande .pdf
CantodaSereia
 
Variation And Change In Spanish 1st Edition Ralph Penny
solasbaney4d
 
Grammaticalization And Pragmatics Facts Approaches Theoretical Issues Corinne...
gnaftimlaker
 
Standard Negation The Negation Of Declarative Verbal Main Clauses In A Typolo...
suncakakuda
 
On Diversity And Complexity Of Languages Spoken In Europe And North And Centr...
jxrxyrqm951
 
The Variability Of Current World Englishes Eugene Green Editor Charles F Meye...
lukacovitzr4
 
Anthrtopological linguistics1
Diego Morales
 
Scopes of linguistic description 1
Bel Abbes Neddar
 
The Lexicon Ebook James Pustejovsky Olga Batiukova
carpesvellen
 
I01066062
IOSR Journals
 
Language structure is partly by social structure
SpanishinBuenosAires
 
Variation In Datives A Microcomparative Perspective Beatriz Fernndez
marathokala90
 
The Structure Of Discoursepragmatic Variation Heike Pichler
buyensthike
 
The Structure Of Discoursepragmatic Variation Heike Pichler
buyensthike
 
The Syntactic Variation Of Spanish Dialects Illustrated Ngel J Gallego
tesinzimvo43
 
Perspectives on Variation Sociolinguistic Historical Comparative Nicole Delbe...
brodapatatu4
 
The Diachrony Of Verb Meaning Aspect And Argument Structure Elly Van Gelderen
hvoxizmp464
 
Newest Trends In The Study Of Grammaticalization And Lexicalization In Chines...
shottdulla25
 
Information Structure Theoretical Typological And Experimental Perspectives M...
conranulz15
 
language variation and change Presentation1.pptx
DESTAWWAGNEW
 
Ad

Recently uploaded (20)

PPTX
Virtual and Augmented Reality in Current Scenario
Shruti Dalela
 
PDF
Trump Administration's workforce development strategy
Mebane Rash
 
PDF
My India Quiz Book_20210205121199924.pdf
AtandraDey2
 
PDF
CISA (Certified Information Systems Auditor) Domain-Wise Summary.pdf
infosecTrain
 
PDF
Empowerment Technology for Senior High School Guide
solthereseamericandr
 
PDF
advance database management system book.pdf
hinamannan364
 
PDF
medical_surgical_nursing_10th_edition_ignatavicius_TEST_BANK_pdf.pdf
victor kamau
 
PDF
FORM 1 BIOLOGY MIND MAPS and their schemes
arcade5
 
PDF
Hazard Identification & Risk Assessment .pdf
estagiarioprotegesms
 
PPTX
Computer Architecture Input Output Memory.pptx
Gunasundari Selvaraj
 
PPTX
B.Sc. DS Unit 2 Software Engineering.pptx
Dr. Pallawi Bulakh
 
PDF
Complications of Minimal Access-Surgery.pdf
Laparoscopy Hospital
 
PDF
Practical Manual AGRO-233 Principles and Practices of Natural Farming
MilindKhedekar2
 
PDF
BP 704 T. NOVEL DRUG DELIVERY SYSTEMS (UNIT 2).pdf
CMEmpire
 
PDF
Vision Prelims GS PYQ Analysis 2011-2022 www.upscpdf.com.pdf
navneetgupta711851
 
PPTX
Unit 4 Computer Architecture Multicore Processor.pptx
Gunasundari Selvaraj
 
PDF
1.3 FINAL REVISED K-10 PE and Health CG 2023 Grades 4-10 (1).pdf
JohnJerryDalawampu
 
PPTX
Introduction to pro and eukaryotes and differences.pptx
AvaniKarumathil
 
PDF
HVAC Specification 2024 according to central public works department
amitrobanipl
 
PPTX
20th Century Theater, Methods, History.pptx
cpadilla9
 
Virtual and Augmented Reality in Current Scenario
Shruti Dalela
 
Trump Administration's workforce development strategy
Mebane Rash
 
My India Quiz Book_20210205121199924.pdf
AtandraDey2
 
CISA (Certified Information Systems Auditor) Domain-Wise Summary.pdf
infosecTrain
 
Empowerment Technology for Senior High School Guide
solthereseamericandr
 
advance database management system book.pdf
hinamannan364
 
medical_surgical_nursing_10th_edition_ignatavicius_TEST_BANK_pdf.pdf
victor kamau
 
FORM 1 BIOLOGY MIND MAPS and their schemes
arcade5
 
Hazard Identification & Risk Assessment .pdf
estagiarioprotegesms
 
Computer Architecture Input Output Memory.pptx
Gunasundari Selvaraj
 
B.Sc. DS Unit 2 Software Engineering.pptx
Dr. Pallawi Bulakh
 
Complications of Minimal Access-Surgery.pdf
Laparoscopy Hospital
 
Practical Manual AGRO-233 Principles and Practices of Natural Farming
MilindKhedekar2
 
BP 704 T. NOVEL DRUG DELIVERY SYSTEMS (UNIT 2).pdf
CMEmpire
 
Vision Prelims GS PYQ Analysis 2011-2022 www.upscpdf.com.pdf
navneetgupta711851
 
Unit 4 Computer Architecture Multicore Processor.pptx
Gunasundari Selvaraj
 
1.3 FINAL REVISED K-10 PE and Health CG 2023 Grades 4-10 (1).pdf
JohnJerryDalawampu
 
Introduction to pro and eukaryotes and differences.pptx
AvaniKarumathil
 
HVAC Specification 2024 according to central public works department
amitrobanipl
 
20th Century Theater, Methods, History.pptx
cpadilla9
 
Ad

The Limits Of Syntactic Variation Theresa Biberauer

  • 1. The Limits Of Syntactic Variation Theresa Biberauer download https://ebookbell.com/product/the-limits-of-syntactic-variation- theresa-biberauer-1645896 Explore and download more ebooks at ebookbell.com
  • 2. Here are some recommended products that we believe you will be interested in. You can click the link to download. Syntactic Features And The Limits Of Syntactic Change Jhannes Gsli Jnsson And Thrhallur Eythrsson https://ebookbell.com/product/syntactic-features-and-the-limits-of- syntactic-change-jhannes-gsli-jnsson-and-thrhallur-eythrsson-30396290 The Poverty Of Conceptual Truth Kants Analyticsynthetic Distinction And The Limits Of Metaphysics 1st Edition R Lanier Anderson https://ebookbell.com/product/the-poverty-of-conceptual-truth-kants- analyticsynthetic-distinction-and-the-limits-of-metaphysics-1st- edition-r-lanier-anderson-5852594 The Limits Of Ancient Biography Brian Mcging Judith Mossman https://ebookbell.com/product/the-limits-of-ancient-biography-brian- mcging-judith-mossman-46469680 The Limits Of The World Andrew Raymond Drennan https://ebookbell.com/product/the-limits-of-the-world-andrew-raymond- drennan-46571530
  • 3. The Limits Of Scientific Reason Habermas Foucault And Science As A Social Institution John D Mcintyre https://ebookbell.com/product/the-limits-of-scientific-reason- habermas-foucault-and-science-as-a-social-institution-john-d- mcintyre-46707938 The Limits Of Exactitude In Greek Roman And Byzantine Literature And Textual Transmission 1st Edition Franco Montanari And Antonios Rengakos https://ebookbell.com/product/the-limits-of-exactitude-in-greek-roman- and-byzantine-literature-and-textual-transmission-1st-edition-franco- montanari-and-antonios-rengakos-47238282 The Limits Of Regulation A Critical Analysis Of Capitalist Development Stavros Mavroudeas https://ebookbell.com/product/the-limits-of-regulation-a-critical- analysis-of-capitalist-development-stavros-mavroudeas-47252632 The Limits Of The British Empire John Dee Ken Macmillan Editor https://ebookbell.com/product/the-limits-of-the-british-empire-john- dee-ken-macmillan-editor-47288444 The Limits Of Europe Daniel C Thomas https://ebookbell.com/product/the-limits-of-europe-daniel-c- thomas-47440488
  • 6. The Limits of Syntactic Variation
  • 7. Volume 132 The Limits of Syntactic Variation Edited by Theresa Biberauer General Editors Werner Abraham University of Vienna / Rijksuniversiteit Groningen Elly van Gelderen Arizona State University Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective. Linguistik Aktuell/Linguistics Today (LA) Advisory Editorial Board Cedric Boeckx Harvard University Guglielmo Cinque University of Venice Günther Grewendorf J.W. Goethe-University, Frankfurt Liliane Haegeman University of Lille, France Hubert Haider University of Salzburg Christer Platzack University of Lund Ian Roberts Cambridge University Lisa deMena Travis McGill University Sten Vikner University of Aarhus C. Jan-Wouter Zwart University of Groningen
  • 8. The Limits of Syntactic Variation Edited by Theresa Biberauer University of Cambridge and Stellenbosch University John Benjamins Publishing Company Amsterdam/Philadelphia
  • 9. Library of Congress Cataloging-in-Publication Data The limits of syntactic variation / edited by Theresa Biberauer.        p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166-0829 ; v. 132) Includes bibliographical references and index. 1. Grammar, Comparative and general--Syntax. 2. Principles and parameters (Linguistics) 3. Language and languages--Variation. 4. Language acquisition. I. Biberauer, Theresa. P295.L56    2008 415--dc22                                                            2008024028 isbn 978 90 272 5515 0 (Hb; alk. paper) © 2008 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984. 8 TM
  • 10. Preface and Acknowledgements vii Introduction 1 Part I. The locus of (parametric) variation Parametric versus functional explanations of syntactic universals 75 Martin Haspelmath Three fundamental issues in parametric linguistics 109 Chiara Gianollo, Cristina Guardiano Giuseppe Longobardi On the syntactic flexibility of formal features 143 Hedde Zeijlstra Expletives, datives, and the tension between morphology and syntax 175 Richard S. Kayne Mapping a Parochial Lexicon onto a universal semantics 219 Gillian Ramchand Peter Svenonius Aspect matters in the middle 247 Marika Lekakou Part II. A classic parameter revisited: The null subject parameter The null subject parameter and correlating properties: The case of Creole languages 271 Marco Nicolis The Case-F valuation parameter in Romance 295 Gerardo Fernández-Salgueiro Silent arguments without pro: The case of Basque 311 Maia Duguine Case morphology and radical pro-drop 331 Ad Neeleman Kriszta Szendrői Table of contents
  • 11. Part III. Parametric clustering The macroparameter in a microparametric world 351 Mark C. Baker Topic prominence and null subjects 375 Marcello Modesto Non-configurationality: Free word order and argument drop in Turkish 411 Balkiz Öztürk Diachronic stability and feature interpretability 441 E. Phoevos Panagiotidis Part IV. The acquisition of parameters Can children tell us anything we did not know about parameter clustering? 459 Larisa Avram Martine Coene Parameter setting and input reduction 483 Arnold Evers Jacqueline van Kampen Index 517  Table of contents
  • 12. Preface and Acknowledgements The contributions in this volume represent a selection of the papers that were originally accepted for presentation at the conference on The Structure of Para- metric Variation which was held at the University of Newcastle Upon Tyne in September 2005 under the auspices of the Arts and Humanities Research Coun- cil (AHRC)-funded project on “Null Subjects and the structure of parametric variation” (AR14458). That conference also included three papers which could unfortunately not be included in the present volume owing to prior publication commitments. These are: • David Adger – Variation without parameters; • Sjef Barbiers – Word-order alternations in three-verb clusters as non-parametric variation; and • Geoff Horrocks Melita Stavrou – Parametric variation in the lexicalisation of semantic properties: The role of grammatical aspect in shifting a verb’s basic Aktionsart My sincerest thanks to all the contributors for the focused, co-operative and ex- tremely patient manner in which they approached the various stages of the publi­ cation process. Grateful thanks also to the reviewers for their valuable input – in alphabetical order: Karlos Arregi, Anders Holmberg, John A. Hawkins, Kasia Jaszczolt, Chris Johns, Jaklin Kornfilt, Ian Roberts, Michelle Sheehan, Ianthi Tsimpli and David Willis. And to Vicki Hart and Glenda Newton for their careful proof- reading at different stages of the process. Additional thanks to Glenda for her help with the Index. Further thanks to the Cambridge students who showed such a lively interest in many of the papers in this volume, particularly – once again in alphabetical order – Bob Davies, Luke Donnan, Emma Game, Nadine Hitchon, Iain Mobbs and Neil Myler. And, last but not least, my thanks to Kees Vaes and Martine van Marsbergen at Benjamins for their patience, support and guidance.
  • 14. Introduction 1. Introduction Martin Joos’s famous statement that “(l)anguages can differ from each other without limit and in unpredictable ways” (Joos 1957: 96) is often misrepresented as an assertion reflecting the thinking current during the first half of the 20th century.1 In reality, even the American descriptivist Franz Boas, to whom this view is ascribed, did not hold this view. “[I]n all the languages of the world, the number of processes which are utilized to express the relations of terms is limited”, Boas wrote in the introduction to the Handbook of American Indian Languages (1911: 67). More generally, there is also no evidence that other American de- scriptivists (Sapir, Bloomfield, etc.) held the extreme view expressed by Joos: these linguists appear to have been far less committed to the notion of boundless variation in human languages than to the belief that it was simply too early to speculate about the possible limits on this variation, given how little was known at the time about non-Indo-European languages. Thus Bloomfield (1933: 20), for example, observes that: [t]he fact that some features are … widespread is worthy of notice and calls for explanation; when we have adequate data about many languages, we shall have to return to the problem of general grammar and to explain these similarities and divergences and he also writes of: [l]ooking forward to a larger synthesis, a General Grammar, which will register similarities between languages (Bloomfield 1934: 46).2 In the 50 years since Joos’s statement, “general grammar” has, in one form or another, been very much at the forefront of linguistic research, with researchers in various frameworks having become interested in language universals and also the associated question of what limits human languages are subject to. Thus the work . Cf. Thomas (2002) for extensive discussion of this misinterpretation. . Cf. Roberts (2007b) for further discussion of the extent to which linguists of the late 19th and early 20th centuries were interested in comparative issues of this sort.
  • 15.  The limits of syntactic variation of Greenberg and Chomsky and their followers has always been guided by the fundamental idea that the very striking and seemingly limitless superficial varia- tion between languages in fact conceals “uniformities of universal scope” (Green- berg, Osgood Jenkins 1961/1963: xv). Furthermore, while the manner in which typologists and generativists have approached the search for these “uniformities”, or “universals” as they are more commonly known, has been very different, there is today agreement between those in the two camps that it is indeed the case that the variation exhibited by the world’s languages is not random. In particular, there is agreement that the syntactic variation across languages is neither “without limit” nor “unpredictable”, and it is this aspect of linguistic variation that is the focus of this volume. 2. The non-randomness of syntactic variation Arguably the most oft-cited illustration of the non-arbitrariness of syntactic varia- tion has its origins in Greenberg’s pioneering work on word-order universals (cf. Greenberg 1963). This work brought to light the surprising fact that seemingly unrelated word-order patternings tend to correlate3 – S(ubject)-O(bject)-V(erb), postverbal auxiliaries and postpositions, for example; VSO, preverbal auxiliaries and prepositions; and so on. Crucially, it also introduced the idea that one might want to think of language types defined in syntactic terms. Thus Greenberg himself distinguished SVO, SOV and VSO languages and proposed that these might be viewed as the “basic word-order types”, with a language’s “type” determining the patterning of various pairs of elements, e.g., V-Auxiliary and N-Adposition order, as noted above, and also Adjective-Noun, Genitive-Noun, and so on. Sub- sequent work, notably that by Lehmann (1973); Venneman (1974) and Hawkins (1983), refined not only Greenberg’s word-order typology, but also the notion of “harmony” that emerged from it. Thus the well-known generalizations in (1) became established, along with the idea that so-called “harmonic” languages, which conform to these correlations, could be distinguished from so-called “disharmonic” languages, which do not: (1) VO: Aux-V, Prepositions, Noun-Adjective, Noun-Genitive, Noun-Relative OV: V-Aux, Postpositions, Adjective-Noun, Genitive-Noun, Relative-Noun . In itself, this was not an entirely new observation: as Greenberg (1963: 104, note 4) himself observes, researchers such as Lepsius (1880) and in particular Schmidt (1926) had previously drawn attention to a subset of the word-order correlations identified by Greenberg.
  • 16. Introduction  Later work, i.a. by Tomlin (1986); Dryer (1992) and by the contributors to the World Atlas of Language Structures (WALS, cf. Haspelmath et al. 2005), has served to reveal the extent to which the Greenbergian language types are in fact instanti- ated in the world’s languages. Certain correlations are shown to constitute very strong correlations, the OV/VO and adposition one being a case in point, as (2) below shows: (2) Correlation between OV/VO ordering and adpositional structure in the 892 languages surveyed in WALS (cf. Baker, this volume) OV Postposition 427 (47.87%) VO Preposition 417 (46.75%) OV Prepostion 10 (1.12%) VO Postposition 38 (4.26%) Thus 94.5% of the languages surveyed in the WALS are “harmonic” in respect of their OV/VO and adposition structure. As Dryer (1992, note 17) notes, however: The general conclusion is that completely consistent [i.e., harmonic – TB] types represent a minority, but that the majority of inconsistencies among inconsistent languages can be attributed to a small number of pairs of elements for which there is a skewed distribution, such as the general preferences for NRel (ative) order. Viewed across the board, then, it is not realistic to think about the Greenber- gian word-order-based types in absolute terms – they do not entail absolute universals, but rather statistical universals or universal tendencies (cf. i.a. Green- berg 1963; Hawkins 1983).Worth noting, however, is Dryer’s observation that the limited number of fully harmonic languages is largely due to the existence of a fixed set of harmony-disrupting element-pairs which tend to surface in only one of the two possible orderings relative to each other. What we are left with, then, is the impression that, even though harmonic languages are in the minority, there is a significant amount of non-randomness in the domain of word order, non-randomness which even extends to the domain of exceptions to the observed regularities. And the same also appears to be true in a range of other syntactic and non-syntactic domains (cf. the Universals Archive at http:// typo.uni-konstanz.de/archive/intro/index.php for an inventory of the currently more than 2000 typological universals recorded to date). The question, of course, is why?
  • 17.  The limits of syntactic variation 3. Accounting for limits on syntactic variation 3.1 Two different approaches: Functionalism vs. formalism4 As Haspelmath (this volume) notes, typologists have tended to follow a function- alist line in attempting to answer the question of why languages should display the structural parallels that they do. More specifically, researchers working within the typological framework generally appeal to communicative and/or processing con- siderations in order to account for the structural skewings that have been observed in the world’s languages (cf. also Hawkins 1988, 2004 and Newmeyer 1998, 2005a for extensive discussion). Thus, for example, Hawkins (2004: 103) proposes that the following parsing efficiency principle plays a central role in accounting for the word-order correlations in (1) and (2): (3) Minimize Domains (MiD) The human processor prefers to minimize the connected sequences of linguistic forms and their conventionally associated syntactic and semantic properties in which relations of combination and/or dependency are processed. The degree of this preference is proportional to the number of relations whose domains can be minimized in competing sequences, and to the extent of the minimization difference in each domain. In terms of (3), then, the adjacency of V and P and P and N that is ensured in con- sistently head-initial and consistently head-final languages (head-initial: [VP V [PP P [NP N …]]] and head-final: [VP [PP [NP … N] P] V]]) minimizes the domain within which the elements making up VP and PP (V and PP and P and NP respectively) will be recognized. The fact that “disharmonic” patterns generally are not as common as “harmonic” ones is thus ascribed to the added processing burden that they entail. In contrast to typologists, generativists in general mostly seek to understand crosslinguistic regularities and gaps by relating these to the influence of an in- nately specified and formally oriented linguistic capacity, Universal Grammar (UG), on which, more below. It is worth noting, however, that it is not the case that . As Newmeyer (1998: 7–8) notes, formal is an unhelpfully ambiguous term, meaning both “pertaining to grammatical form (as opposed to meaning and/or use)” and “stated in math- ematically precise terms as in formalized”. Clearly, there are formal approaches to language that are only formal in the former sense (cf. Saussurean structuralism), but these will not concern us here. For the purposes of the present discussion, then, formal is to be understood as implying both of the above meanings in the manner generally assumed by generativists. Worth noting in connection with formal in the second sense is that it does not generally in syntactic theory mean anything as strong as “axiomatised” or even “fully formalized”, as it does in, for example, logic, computational work or formal semantics.
  • 18. Introduction  generativists uniformly reject the possibility that functional and, particularly, pro- cessing considerations may determine the distribution of syntactic patterns, al- though this has undoubtedly been the most common trend to date (see below). Thus, for example, Ackema Neeleman (2002) argue that the very striking asymmetry between leftward and rightward movement observed in the world’s languages – in terms of which leftward movement is vastly more common than rightward move- ment and additionally appears to be unbounded in a way that rightward movement is not – can in fact best be understood in parsing terms.5 More specifically, they, like non-generativist psycholinguists (cf. i.a. Just Carpenter 1992; Gibson 1998), propose that processing of syntactic representations involves a left-to-right parsing procedure which involves both processing and storage: since it is presumably more economical to assign already-parsed structure to short-term memory as soon as this becomes possible rather than to wait until the entire structure has been parsed, it is assumed that already-parsed structure is “closed off” at appropriate points during the parsing process. This, then, renders its internal structure “invisible” to the parser. If, as is usually thought, the parser employs a filler-driven strategy in dealing with movement dependencies (cf. Clifton Frazier 1989; Frazier Flores D’Arcais 1989), it will only postulate a trace after it has encountered a moved element. Clearly, this is always the case when we are dealing with leftward movement (cf. (4a)), but the opposite is true wherever rightward movement has taken place (cf. (4b)): (4) a. wh-element/topicalized dp, etc. ….. twh-element/tDP, etc. b. twh-element/tDP, etc ……… wh-element/topicalized dp, etc. As (4b) illustrates, rightward movement is problematic within filler-driven, interim storage-oriented approaches to parsing in that it requires post hoc trace insertion in a partially analysed string: left-to-right parsing must be suspended so that a trace can be inserted at an appropriate place in an already-parsed piece of struc- ture located to the left of the point the parser had reached. If it is indeed the case that such structure has been “closed off” and therefore lacks internal structure, we might expect this kind of trace-insertion to be impossible or at least costly. And the same mode of explanation is available to account for the locality restric- tions on rightward movement. As Ross (1967) first observed, this is always clause- bounded (cf. the so-called Right Roof Constraint). On the natural assumption that distinct clauses constitute distinct parsing domains, we would expect it to be . Other well-known examples of generativists appealing to parsing to account for formal patterns include: Bever Langendoen (1971) in their work on the factors determining the availability of relative pronoun omission in English, Fodor (1978, 1984) in her work on nested dependencies, Berwick Weinberg (1984) in their work on Subjacency, and the work of Janet Dean Fodor Lyn Frazier more generally.
  • 19.  The limits of syntactic variation impossible for rightward movement to be unbounded in the manner that leftward movement appears to be. This is because unbounded leftward movement is unprob- lematic if long-distance movement is indeed, as has generally been assumed since Chomsky (1973), successive-cyclic: if elements move leftward successive-cyclically, each clause will be headed by a trace, facilitating easy tracking of the filler-gap depen- dency,evenwhereearlierclauseshavealreadybeen“closedoff”–thetraceatthestart of every parsing domain alerts the parser to the fact that its search is not yet complete. As far as the parser is concerned, then, filler-gap dependencies can be established in the usual local manner wherever movement is successive-cyclic and leftward. The same is not, however, true for rightward movement beyond the clause to which the moved element belongs. Even if this movement is successive-cyclic, it will still entail the requirement that previously “closed off” material needs to be “reopened”. This is because traces in clauses preceding that to which the element has moved cannot be postulated until the parser has encountered the gap located at the rightmost edge of the final clause. Long-distance rightward movement therefore entails multiple “reopenings” of the type required in monoclausal rightward-movement structures. Ackema Neeleman (2002) explicitly indicate why a purely formal account of these phenomena – for example, one appealing to Kayne’s (1994) Antisymmetry proposal, in terms of which all movement is necessarily leftwards because of structural con- straints (specifiers all being left-adjoined, with the result that upward movement can only be to the left) – fares less well than a parsing-oriented approach (cf. also Abels Neeleman 2007). The details of this argument fall beyond the scope of this introduction; the point here is simply that it is not always the case that formal- ists exclusively seek to understand syntactic distributions and the factors deter- mining them in purely formal terms. And, more generally, there is, as Newmeyer (1998: 21) rightly notes, “nothing in the generative program that demands that all typological facts be attributed to the setting of innately specified parameters” (cf. also Newmeyer 1998, chapter 3). Furthermore, there is also nothing in the generative re- search programme that excludes the possibility that innately prespecified principles may ultimately be functionally grounded (cf. Müller 2007a, numerous contribu- tions in Good (2008), in particular, Kiparsky 2008, who proposes criteria designed to distinguish UG-based universals from those which have historical – and thus potentially ultimately functional – explanations, and Mobbs 2008, who suggests a “third factor” reinterpretation of Hawkins’s performance principles6). We return to these matters below. . Kiparsky (ibid.3) explicitly refers to functional principles which “might have become biologized within UG itself”. Cf. Section 3.4 for discussion of “third factor” considerations within current Chomskyan theory which may also include functional considerations.
  • 20. Introduction  First, however, let us consider the UG-based approach and, in particular, that which generativists have espoused since the introduction of the Principles and Parameters model in the late 1970s/early 1980s. 3.2 Introducing the Principles and Parameters Framework Universal Grammar (UG) is the innately specified mental endowment which is assumed to be uniquely shared by all humans. In terms of the current general model of UG, the Principles and Parameters (PP) model first outlined in Chom- sky’s 1979 Pisa Lectures (published in revised form as Chomsky 1981), UG con- sists of invariant principles and unset parameters whose values must be set on the basis of the linguistic input available to language-acquiring children (the primary linguistic data or PLD) – cf. (5): (5) UG (principles unset parameters) + PLD → Adult grammar S0 Ss (the initial state) (the steady state) Together, the immutable principles and the initially unset parameters (i.e., S0 in (5)) are taken to determine the form of a possible human grammar (i.e., Ss in (5)), where “grammar” is understood as the abstract mental system, including specific parametric settings, – the I-language in Chomsky’s terms (cf. Chomsky 1986 et seq.) – on the basis of which actual utterances are produced. For Chom- skyans, then, UG plays a role not only in language acquisition, but also in defining the notion “possible language”, both synchronically and diachronically. And, crucially, the limits it is assumed to impose on “possible languages” and, in particular, the syntactic variation that they may exhibit, are conceived of in formal terms. In PP’s first incarnation during the Government and Binding (GB) period, a range of principles and parameters were proposed, with the X-bar schema, the Theta Criterion, the Extended Projection Principle (EPP) and Subjacency/Rela- tivised Minimality numbering among the most frequently cited principles and the Head Parameter (HP) and the Null Subject Parameter (NSP) undoubtedly being the most well-known parameters. In terms of the former, there are universal con- straints on the way languages construct phrases, the way in which thematic roles like Agent, Patient and Goal are assigned, the positioning of subjects, and the dis- tance over which elements can undergo movement: no languages should violate these principles, with the result that no variation in these domains is expected (but see below). Parameters, on the other hand, do permit variation, although it is variation of a highly restricted kind. Consider, by way of illustration, the HP and the NSP in (6–7).
  • 21.  The limits of syntactic variation (6) Head Parameter (cf. Travis 1984) Heads X precede/follow their complements X' a. b. X Comp  → - X' X Comp  → - (7) Null Subject Parameter (cf. Rizzi 1982, 1986) Subjects can be pro, where pro is a [+pronominal, -anaphoric] entity which is licensed by X0 y where X0 y is a subset of the heads in a given language. The gram- matical specification of pro is recovered from the features of the licensing head. e.g. I' ... I [+pronominal] → I licenses pro Spec pro IP a. b. I' ... I [–pronominal] → I does not license pro Spec overt pronoun IP As shown above, both the HP and the NSP define an innately specified choice- point, the HP in respect of head-complement directionality and the NSP in respect of the availability of pronominal null subjects. Just like principles, then, parameters impose universal restrictions – they determine the set of structural choices avail- able to languages in various domains, thereby ensuring that all languages make choices about the same set of choice-points. Because languages do not make the same choices in each case, variation results. Thus the PP approach opens up the possibility of having a model of grammar which entails a finite and potentially relatively simple UG, but which nevertheless leads us to expect what might on the surface seem like considerable crosslinguistic variation – consider the striking dif- ferences between languages that opt for different settings of the HP, for example. That is, the PP approach, unlike those that went before it, facilitates a resolution of the tension between the search for both descriptive and explanatory adequacy, being able to account not only for the (apparent) diversity of syntactic structures, but also for how these could have been successfully acquired given the constraints under which language acquisition takes place (cf. Chomsky 2004b for overview discussion, and also Section 7 below). In accordance with the schema in (5), the values assigned to the various parametric choice-points are assumed to be determined on the basis of specific
  • 22. Introduction  PLD, signalling the appropriate setting. With the HP, at least in the case of the consistently head-initial/head-final languages (i.e., the ones exhibiting harmonic order in respect of V and its complement, Aux and V(P), Adpositions and their complements, Nouns and their complements, so on7), it is not hard to see what would be an appropriate cue (cf. Lightfoot 1999 and Dresher 1999, 2003 for discussion of the notion “cue”): any head + complement-containing XP will provide the necessary cue. Thus, for example, any one of the following Japanese structures would enable the learner to set the HP to head-final: (8) a. Sensei-wa [Taro-o] sikata teacher-top Taro-acc scolded ‘The teacher scolded Taro.’ → Objects precedes V b. [Nihon] kara Japan from ‘from Japan’ → Complement precedes adposition c. John-ga [Mary- to renaisite] iru John-nom Mary-within.love is ‘John is in love with Mary.’ → V precedes Auxiliary d. [John-ga Mary-o nikundeiro] to John-nom Mary-acc be-hating that ‘that John hated Mary’ → Clause precedes Complementizer8 As noted above, however, fully harmonic languages represent the minority of the world’s languages and many languages, including very well studied ones like German, exhibit mixed headedness, with certain phrases seeming to be . It is worth noting that the HP does not, as is often thought, determine the relative ordering of Noun-Adjective as adjectives are not nominal complements. Similarly, the HP has nothing to say about those Noun-Genitive and Noun-Relative pairings which do not involve a comple- mentation relation. The fact that, for example, English exhibits Adjective-Noun rather than the Noun-Adjective order expected in a harmonic head-initial language (cf. (1) above) does not, therefore, reflect a shortcoming of the HP. . If, as Lightfoot (1989, 1991) has suggested, learners do indeed only pay attention to matrix clauses at the stage at which parameter setting is underway, (8d) may play no role in the setting process. As it is only necessary for the child to establish the directionality of a single XP (say, the VP) in order to set the HP, this does not, however, have to pose particular problems. The same is obviously not true for “mixed” languages, though (see main text).
  • 23.  The limits of syntactic variation consistently head-initial (e.g., CP and DP in German), while others appear to be consistently head-final (VP and TP in German). This is illustrated in (9): (9) a. … daß er das Buch von Chomsky unter dem Tisch gefunden hat that he the book from Chomsky under the table found has ‘… that he found the book by Chomsky under the table’ b. [CP daß [TP eri [VP ti [DP das Buch von Chomsky] [PP unter dem Tisch] gefunden] hat]] i.e., C, D, N and P precede their complements; V and T follow them. There also appear to be quite a few languages which exhibit head-directionality of one kind in the clausal domain and the opposite in nominals (cf. Thai, Khmer and Burmese as discussed in Simpson 2005, Bangla as discussed in Bhattacharya 1999, Marathi as discussed in Nayudu 2008, and Gungbe as discussed in Aboh 2004, etc.). This is shown for Gungbe in (10): (10) dàwé l� x� kὲkέ man det bought.perf bicycle ‘The man bought a bicycle.’ (from Aboh 2004: 32) Worse still, there are further cases where individual lexical items appear to determine the headedness of the phrases in which they surface. Thus German adpositions are generally head-initial (cf. von and unter in (9)), but there are a few exceptions like zufolge (“according to”) and halber (“for the sake of, because of”) and the “shadow prepositions” (cf. Noonan 2003) herauf (“up”) and drin (“in”) which only surface finally. The “mixed” complementiser class in languages like Bengali, Marathi and Assamese, in which certain complementisers are head-initial while others are head-final, represents another problematic case (cf. Bayer 1999, 2001). Furthermore, there are lexical items like German entlang (“along(side)/beside”) and Afrikaans in (“in”/“into”) which occur both initially and finally, but where positioning determines other morphosyntactic and/or semantic properties – in the case of entlang, which case its complement takes (genitive where entlang is initial and accusative where it is final) and in the case of in, precisely how it is interpreted (as locative “in” when it is initial and as directional “into” when it is final).9 The examples in (11–12) illustrate: (11) a. Sie ist entlang des Flusses gefahren [German] she is along the-gen river driven ‘She drove along the river.’ . We leave aside here the potential further complication represented by so-called circumposi- tions (cf. Dutch naar … toe – “to … to”) which, superficially at least, appear to consist of nested adpositional phrases, with a head-initial phrase embedded under a (superficially) head-final one (cf. i.a. van Riemsdijk 1978, 1990; Koopman 1997; Zeller 2001; and den Dikken forthcoming).
  • 24. Introduction  b. Sie ist den Fluss entlang gefahren she is the-acc river along driven ‘She drove along the river.’ (12) a. Hy loop in die bos [Afrikaans] he walk in the forest ‘He is walking in the forest.’ b. Hy loop die bos in he walk the forest in ‘He is walking into the forest.’ In at least some cases, then, head-directionality appears to play a lexical disam- biguation role.10 Finally, there are elements like the prepositional complement-taking adjective stolz (“proud”) in German and its counterpart in Dutch (trots) and adpositions like ilman (“without”) and kohti (“towards”) in Finnish which can surface either before or after their complements without either morphological or semantic con- sequences. Consider (13–14) in this connection: (13) a. Die Frau ist stolz auf sich [German] the woman is proud of self ‘The woman is proud of herself.’ b. Die Frau ist auf sich stolz the woman is of self proud ‘The woman is proud of herself.’ (14) a. kohti Stalinin kuvaa [Finnish] towards Stalin-gen picture ‘towards a picture of Stalin’ b. Stalinin kuvaa kohti Stalin-gen picture towards ‘towards a picture of Stalin’ . If it is indeed correct to think of lexically-specified head-directionality, one might want to appeal to Elsewhere considerations (cf. Kiparsky 1973) in accounting for the word-order properties of structures containing these elements. More specifically, it could conceivably be the case that the HP is appropriately set for a specific set of syntactic categories, but that lexical items associated with independent (i.e., irregular) linearisation instructions are linearised in accordance with their more specific specification. If this approach is extended to cases permit- ting both linearisation options (e.g., the German and Dutch adjectives and Finnish adpositions mentioned in the main text), we would need to postulate two distinct lexical entries which are phonologically and semantically identical, but which differ in terms of their linearisation in- structions (cf. Adger 2006 for an optionality proposal along these lines).
  • 25.  The limits of syntactic variation Taken together, these facts raise serious questions about how children success- fully identify the appropriate cues to set the HP (cf. Dresher 1999, 2003; Fodor 1998a, 1998b, 2001; Fodor Sakas 2005; Sakas Fodor 2001, and Lightfoot 1999 for recent discussion of the general problem of cue identification, and also the contributions by Gianollo, Guardiano Longobardi (henceforth: GGL), Zeijlstra, Avram Coene and Evers van Kampen in this volume). And they also suggest that the HP may be a rather different type of parameter to the NSP, which is usually (though see Sections 3.4 and 5 below) thought to involve a binary choice which is made once and for all for the language as a whole. Specifically, one might view the HP as a parameter which can operate in a more or less “myopic” fashion as outlined below: a. “blindly”, being set once for all categories, thereby delivering harmonic languages; b. category-sensitively, being consistently set in one way for a fixed set of syntactic categories and another way for another fixed set; or c. lexical item-sensitively, being set individually for specific lexical items (although see note 10). Despite this fact, the general assumption about the HP during the GB era was that it was, universally, a parameter specifically associated with the phrase-structure (X-bar) module of the grammar. Turning to the NSP: this parameter clearly involves two components, namely (a) that a null-subject language (henceforth: NSL) include in its pronoun inventory the phonologically null, lexically unspecified pronoun pro and (b) that it include in its inventory of functional categories subject-related ones (e.g., I in the clausal domain and/or D in the nominal domain) which are capable of assigning pro the appropriate φ-features (here: person and number). Non-NSLs, by contrast, lack one or both of these properties. Traditionally, the assumption was that the NSP was set to its positive value – in terms of which children postulate the availability of pro and the appropriate licensing head(s) – on the basis of PLD such as (15) featuring unrealised referential pronominal subjects as in (15a) and unrealised expletive pronominal subjects as in (15b) and (15c): (15) a. Habl- a español [Spanish] speak-3sg Spanish ‘(S)he speaks Spanish.’ b. Piove [Italian] rain-3sg ‘It is raining.’
  • 26. Introduction  c. Sono venute delle ragazze [Italian] be-3pl come-f.pl some girls ‘There came some girls.’ Just as was the case with the HP, the NSP was assumed to determine the (un)avail- ability of a range of structures not limited to those which, like those in (15), were generally assumed to constitute the actual cue required to trigger the appropriate setting of the parameter. Thus the NSP sanctioned not only null-subject structures such as those in (15), but also structures like the following: (16) a. Ha telefonato Gianni [Italian] have-3sg telephoned Gianni ‘Gianni has called.’ b. Pjos ipes oti egrapse afto to vivlio? [Greek] who said-2sg that wrote this the book ‘Who did you say wrote this book?’ (16a) represents a so-called Free Inversion structure, in which the overt subject follows the lexical verb rather than occupying the canonical subject position, Spec-IP. (16b), in turn, illustrates the absence of so-called that-trace effects in NSLs: unlike in non-NSLs like English, wh-subjects can be extracted across overtly realised complementisers without resulting in an ungrammatical structure (con- trast the English counterpart of (16b), *Who did you say that wrote this book?). Neither of the structures in (16) at first sight seems to relate to the availability of null subjects, but Rizzi’s (1982, 1986) characterization of the NSP links all of these phenomena and additionally also the fact that Romance NSLs exhibit “rich” agree- ment morphology. The reasoning is as follows: the availability of pro facilitates the existence of null-subject structures such as those in (15), with the “rich” agreement morphology here and elsewhere reflecting the fact that the language has the kind of I required to supply pro with the φ-features it intrinsically lacks. The availability of pro also enables a subject to occur in inverted position as in (16a) since pro is available to satisfy the Extended Projection Principle (EPP), i.e., the requirement that every clause contain a subject located in Spec-IP (cf. Chomsky 1981, 1982). Finally, pro also facilitates the circumvention of that-trace violations, as in (16b), by allowing wh-subjects to be extracted from a lower position, while pro itself once again satisfies the EPP. Clearly, then, the NSP as formulated by Rizzi leads us to expect superfi- cially unrelated phenomena to cluster together.11 And if it is indeed correctly . Cf. Roberts (2007c) for discussion of three further phenomena that have been proposed to be connected to the NSP: clitic climbing, infinitive-clitic ordering and realisation of an arbitrary subject as a clitic.
  • 27.  The limits of syntactic variation formulated, we would expect it to impose a universal limit on the types of struc- tures available both in NSLs and in non-NSLs: the former would all be expected to exhibit all of the structures facilitated by pro and a suitable I, while the latter would all be expected to systematically lack these structures. Looking beyond the NSP, this was also the expectation for GB parameters more generally since parameters with a rich deductive structure simultaneously facilitate an account of systematic syntactic correspondences across languages and an account of how it might be possible for children to acquire these systematic correspondences with relative ease. If we consider the NSP, for example, it is arguably the case that unrealised pronominal subjects like those in (15) represent a salient and easily detectable component of the PLD. By contrast, that-trace circumvention structures such as those in (16b) are likely to feature far less frequently in the input and, superficially at least, these structures seem a great deal more complex than unrealised subject- containing ones like those in (15). Even more striking in this connection is the scenario faced by children acquiring non-NSLs: in this case, the input does not contain any positive evidence that that-trace structures are unavailable, with the result that one would expect that children would have to be explicitly told that subject wh-extraction is not possible in the presence of an overt complementiser, unlike extraction of other wh-elements. Such direct negative evidence is, however, not usually thought to play a significant role in first language acquisition (cf. i.a. Bowerman 1988; Marcus 1993) and, in the case at hand, it is furthermore not at all clear that adult native- speakers are actually consciously aware of the existence of this structural gap – it is certainly not one that is noted in prescriptive grammars or explicitly taught in English grammar classes. In both the NSL and the non-NSL case, then, a question arises as to how native speakers are able to acquire structures that are infrequent or even unattested in the input. If we can, however, assume in the NSL case that unrealised pronominal sub- jects alert children to the other NSP-related properties, the salience and frequency discrepancies noted above need not pose an insurmountable obstacle. Similarly, if we can assume that the obligatoriness of overtly realised subjects (and particularly, overt expletives) serves the same function for children acquiring a non-NSL, we can understand the successful acquisition of that-trace phenomena as the consequence of the fact that the unavailability of wh-subject extraction is simply part of the cluster of NSP-related properties associated with the negative setting of this parameter. More generally, if innately specified parameters do indeed determine the way in which certain types of syntactic phenomena cluster, we would expect, firstly, that the world’s languages (past, present and future) will exhibit certain clusters of properties and not others and, secondly, that language acquisition will proceed in a way which reflects the (incremental) acquisition of relevant parametric settings
  • 28. Introduction  and that will therefore never feature stages at which children correctly use some of the structures associated with a given parameter, but also others that are associated with the opposite setting of that same parameter. That is, parameters as they are conceived of in the PP framework aim to contribute to an explanation not only of the limits of syntactic variation, but also of the limits on language acquisition and, by extension, language change. 3.3 Problems with GB principles and parameters As is now well known, many of the principles and parameters proposed during the first phase of the PP era did not ultimately prove either descriptively or explana- torily adequate. Consider, for example, the fact, already noted above, that, while the HP facilitates an elegant account of languages exhibiting “harmonic” word orders, it is evidently too simplistic to assume that head-directionality is universally fixed once-and-for-all on the basis of the simple binary choice in (6): the empirical evi- dence clearly points to a much more complex situation. Similarly, the NSP as formulated above leads to various expectations: that NSLs will be morphologically rich,12 that morphologically poor languages will not be NSLs, that the properties governed by the NSP (cf. (15–16) above) will always cluster together, and so on. As work as early as Gilligan (1987) has shown, however, these predictions are not borne out: numerous morphologically rich languages are not NSLs (e.g., Icelandic and Russian), there are various morphologically poor languages that are NSLs (e.g., Chinese and Cheke Holo), and there are also other respects in which the proposed correlations do not always hold (cf. Bantu lan- guages exhibiting referential null subjects, but not Free Inversion and non-NSLs like Icelandic exhibiting the no that-trace effect property of NSLs). Some of these failed predictions have been ascribed to parameter interac- tion, with the idea being that the non-occurrence of an expected parametric reflex follows from the obscuring effects of the setting of another parameter which inter- acts with the first. Thus, for example, the non-occurrence of referential null subjects in inflectionally rich Icelandic is commonly ascribed to the positive setting of the V2 Parameter in this language – that is, since the V2 Parameter always requires an XP to occupy the preverbal position in main clauses, preverbal null subjects are . It is worth being precise here: what is specifically expected is that NSLs will have rich agreement (i.e., person/number) morphology; rich tense, mood or other verbal morphology should not be relevant to the setting of the NSP, as the crucial consideration is that the richness of the verbal agreement morphology should facilitate recovery of pro’s content (cf. Biberauer Roberts, forthcoming).
  • 29.  The limits of syntactic variation ruled out in these contexts and consequently the effects of the NSP are masked. Clearly, though, this proposal only accounts for the absence of preverbal null sub- jects; it tells us nothing about why referential null subjects are also unavailable else- where and why expletive null subjects, but not their referential counterparts, may surface in postverbal position. More recently, researchers have adopted a different tack, opting instead to re-examine the morphological paradigms of languages which at first sight appear to qualify as “morphologically rich” and proposing that languages like Icelandic do, after all, systematically differ from NSLs (cf. Müller 2005, 2007b Tamburelli 2006 for two recent proposals of this type). More generally, however, the notion of “morphological richness” has proven very hard to quantify, not only in the NSP context, but also more generally (cf. Jaeggli Safir 1989; Vikner 1997; Rohrbacher 1999 for different, but ultimately unsuccessful proposals). Clearly, this is a highly problematic state of affairs for a parameter whose setting was initially thought to be at least partially cued by inflectional wealth. Looking beyond the HP and the NSP, we see that other GB parameters also encountered difficulties of various kinds. Thus, for example, in the V2 context, the discovery that V2 languages are not all “well behaved” and that languages like Icelandic and Yiddish systematically permit V2 in the presence of overtly realised complementisers first raised the question whether one can in fact think of a unitary V2 parameter (cf. Schwartz Vikner 1996 and Vikner 1995 for relevant discus- sion). The rise since the late 1980s of so-called “cartographic” approaches to clause structure, in terms of which functional structure is assumed to be a great deal more articulated than initially thought, has only underscored this question (cf. Poletto 2000; Mohr 2005 and Roberts 2005 for illustrations of articulated CP analyses of V2 phenomena). Furthermore, independently of their precise formulation, this and other head-movement-related parameters (e.g., the V-to-I Parameter, the N-to-D Parameter, etc.) would also seem to face a further conceptual question, namely whether they really meet the requirement that parameters should govern clusters of empirical phenomena: head-movement parameters are not generally thought to give rise to the sorts of superficially unrelated phenomena proposed for the NSP.13 . Viewed from a Germanic perspective, the V-to-I parameter could be argued to govern a cluster of unrelated phenomena – consider, for example, Bobaljik Thráinsson’s (1998) Split- INFL parameter, in terms of which languages exhibiting V-to-I also feature a split IP, full DP Object Shift (OS), and Transitive Expletive Constructions (TECs). This (full) cluster does not, however, seem to be more generally associated with languages which have been analysed as fea- turing V-to-I movement, e.g., the Celtic languages, the Semitic languages, which at most permit pronominal OS (cf. Shlonsky 1997, chapter 10), and the Romance languages, which also lack non-pronominal OS and TECs of the Germanic type (cf. Cornilescu 2000 for a proposal that Romanian features TECs). Strikingly, Avram and Coene, this volume, propose a very different
  • 30. Introduction  Baker (this volume) dubs this the extent of variation question, and we return to it in Sections 3.4 and 3.5 below. Related to it is the question that we have already raised in connection with the HP, namely whether binary parameters can in fact hope to account for the attested range of variation within a given domain. Since this question in fact raises a number of important issues, let us consider it in a little more detail on the basis of a previously undiscussed parameter. Both during the GB era and beyond, it has frequently been proposed that the presence vs. absence of wh-movement in a language is determined by the positive vs. negative setting of a Wh-movement Parameter (cf. Fukui 1995; Cheng 1997 and Simpson 2000 for discussion). If this parameter simply offers a binary choice, we would expect to find only two types of languages: wh-movement and wh-in-situ languages, illustrated in (17): (17) a. Whati did Steve eat ti? b. Hufei chi- le shenme ne [Chinese] Hufei eat-asp thing Qwh ‘What did Hufei eat?’ In actual fact, however, some languages have been shown to exhibit a mixture of these properties, permitting both movement and in situ structures (cf. Cheng 1997). Crucially, closer investigation has revealed that some of the “mixed” lan- guages do not in fact pose an empirical challenge to the postulation of a binary Wh-movement Parameter. Some, like Egyptian Arabic, Irish, Malagasy and Zulu, can, for example, be shown to involve very different underlying struc- tures: a (pseudo-)cleft (i.e., a non-movement structure) in the (apparent) wh- “movement” case and an in situ structure in the non-movement case. This is illustrated in (18): (18) a. U- bona ubani?14 [Zulu] 2sg- see who.1a ‘Who did you see?’ b. Ng- ubani o- m- bona-yo? cop-who.1a rc.2sg-oc.1a- see -rs ‘Who is it that you see?’ (data from Sabel Zeller 2006) cluster of Split-INFL-related effects to that proposed by Bobaljik Thráinsson. At the very least, then, it does not seem that V-to-I can be viewed as correlating with a single cluster of effects. . Sabel Zeller’s abbreviations are as follows: 1a = noun class 1a, i.e., that mostly reserved for singular human entities; cop = copula; rc = relative concord/agreement; oc = object clitic; rs = relative suffix.
  • 31.  The limits of syntactic variation As (18) clearly shows, the usual SVO order is maintained in monoclausal wh- interrogatives, with the wh-element surfacing in situ; it is only in the presence of a cleft-marking copula that wh-elements may appear in a moved position, and then it is clearly a relativised one, rather than the clause-initial Spec-CP position associ- ated with wh-movement languages. Similarly, wh-“movement” in Malay and Iraqi Arabic seems to be possible only in focus structures and not in unmarked interrogatives, i.e., wh-movement, where it occurs, takes place as a consequence of focus- rather than specifically wh- driven movement (cf. Ouhalla 1996). From the perspective of the Wh-movement Parameter, then, these languages are negatively specified for wh-movement and the superficial “mix” of positioning possibilities can be explained despite the avail- ability of only two mutually exclusive parametric options. Importantly, however, there do appear to be systems within which a putative Wh-movement Parameter could not be set once and for all. Duala, a Bantu language spoken in Cameroon, is a case in point: it permits both wh-movement and wh-in situ in matrix clauses, but necessarily requires wh-movement in embedded contexts (cf. Sabel 1998 for detailed discussion). Languages of this sort clearly cannot be specified as [+movement] as wh-elements are evidently not required to move (there is no (nec- essary) interpretive difference between movement and non-movement structures). Similarly,theycannotbespecifiedas[–movement]sincethereisacontext(embedded clauses) where movement is systematically required. Here, then, the Wh-Parameter does appear to face a genuine empirical challenge from optionality phenomena. Furthermore, a simple binary parameter also will not be able to account for the op- tionality exhibited by languages like Bangla, Basque, Imbabura Quechua and others which permit so-called clausal pied-piping to alternate with wh-DP raising, once again without necessary interpretive consequences (cf. Richards 2001 for overview discus- sion); and the same is true of that found in Slavic and other languages which permit Left Branch Condition/LBC-violating wh-extractions to alternate – often without interpretive consequences – with their non-LBC-violating counterparts (contrast English *Whose did he read book? [LBC violation] which is not available alongside Whose book did he read?). Additionally, a simple binary Wh-movement Parameter cannot explain why some languages (e.g., those belonging to the Slavic family) either obligatorily or optionally front all the wh-words present in multiple wh-questions, whereasothersmayonlyfrontoneofthem(cf.Bošković2002fordetaileddiscussion). Like the “size”-related optionality mentioned above, these differences have the flavour of “sub-parameters” associated with the Wh-movement Parameter initially proposed: more precisely, they would seem to be choices which only come into play following positive setting of the Wh-movement Parameter (cf. also Uriagereka 2007 for recent discussion of parametric structure in terms of parameters which differ in their “deep- ness”, delivering, in his words (p.1), “progressively deeper forms of variation”).
  • 32. Introduction  From a parametric perspective, the notion that a parameter that was ini- tially viewed as a unitary Yes/No choice-point in fact turns out to initiate or be part of a sequenced network or hierarchy of related parameters need, of course, not be problematic. We return to this matter in Section 3.3 below. For the moment, it worth noting, however, that the more articulated view of pa- rameters suggested by closer examination of the wh-data also brings with it various complications which should not be overlooked. There is, for example, the matter of how many (sub-)parameters we will ultimately need to postu- late, i.e., how many innately specified choice-points UG will need to contain. Related to this is the question of whether each choice-point must be associated with its own cue (see Section 3.2 above and Section 7 below) and, if so, how one can ensure that the cues are “noticed” in the correct sequence by the language acquirer. Specific acquisitional concerns aside, there is also more generally a clear sense in which an articulated Wh-movement parameter is at odds with the original GB notion of parameters as innate entities whose setting “gives” the acquirer a range of language-specific information – some of it “deep” in the sense of not necessarily being superficially evident from the input – “for free”. This is once again a matter to which we will return in Section 7. Here we simply wish to highlight the fact that closer consideration of GB parameters like the Wh-Parameter has led to the discovery that initially proposed parameters do in some cases hold up under closer scrutiny, but that this is not always the case and, furthermore, that it sometimes appears to be necessary to adopt a more articulated view of the structure of a parameter. To conclude, let us briefly consider the relation between GB principles and parameters, and also the fate of principles proposed during the GB era. Whereas principles and parameters are presently often thought of as distinct and sepa- rate components of UG – cf. the Wh- or Head-movement parameters discussed above, which are not typically viewed as choice-points (directly) connected to any particular UG principle – the earliest conception of “parameter” entailed a close connection between the two entities: parameters were necessarily “parameterised principles” (cf. Newmeyer 2005a). Consider (19–20) in this connection: (19) A parameterised principle in the phrase structure module: a. Principle: X’ → X; Complement (where; signifies an unordered pair) b. Parameter: the Head Parameter (HP; cf. (6) above) Heads X precede/follow their complements (20) A parameterised principle in the bounding module: a. Principle: Subjacency In a structure α … [β … [γ … δ…] …] …, movement cannot apply if β and γ both belong to the class of bounding nodes.
  • 33.  The limits of syntactic variation b. Parameter: the class of bounding nodes = DP and/or TP and/or CP (where bounding nodes in English = DP and TP, in Italian = DP and CP and in Russian = DP, TP and CP) As shown above, phrase structure and movement are both subject to inviolable principles (a), but the precise manner in which these principles manifest in dif- ferent languages is subject to specific points of parametric variation (b). As GB parameters do not all appear to be so directly connected to principles, it seems necessary to acknowledge that the parameters assumed during the GB era were not of a single general type: at the very least, one needs to distinguish parameters associated with principles (“parameterised principles”) from parameters whose choice-points are less directly connected to specific principles, a state of affairs which immediately raises questions about the relation between these two, the “locus” of the latter, and so on. We return to these questions in Section 3.4. The preceding discussion has shown that principles may be directly associated with parametric choice-points. This need, of course, not be the case – consider, for example, the Projection Principle and the Theta Criterion (cf. Section 3.2 above), neither of which entails a (b)-component of the type illustrated in (19–20). Subja- cency, too, was initially thought to be an invariant principle (cf. Chomsky 1973), but Rizzi’s (1982) work revealed this assumption to be incorrect. Similarly, work on the wide range of languages that Chomskyans have investigated since the early PP days has shown that certain further properties initially ascribed universally invariant status also cannot be regarded as such. Let us consider this matter on the basis of the Ex- tended Projection Principle (EPP) in terms of which every clause must have a subject located in Spec-IP (i.e., the canonical subject position; cf. Chomsky 1981, 1982). As is well known, this principle facilitates a unified account of subject and ex- pletive distribution in modern English, but necessitates sometimes quite complex assumptions in the context of the many languages that do not exhibit the same patterns of subject behaviour. Thus, for example, in languages which set the NSP to its positive setting, two assumptions need to be made, namely (a) that these lan- guages feature a particular type of I and (b) that they include in their lexical stock an element, pro, which is able to satisfy the EPP-imposed requirement that Spec-IP be projected and filled (henceforth: the EPP-requirement). In the absence of an EPP- requirement, just the (a)-assumption would be sufficient, and numerous researchers have in fact proposed that it may not be necessary to project and fill Spec-IP in all NSLs (cf. Borer 1986; Barbosa 1995, to appear, Alexiadou Anagnostopoulou 1998 on so-called “I-subject” approaches to NSLs, and see also the discussion in Section 5 below).15 In other NSLs, postulating pro and therefore also the projection . The “I-subjects” approach necessarily has various consequences for Theta Theory, which
  • 34. Introduction  of Spec-IP may nevertheless be justified (cf. Sheehan 2006 and Barbosa, to appear for recent discussion and see Section 5 below for further discussion). In the Germanic context too, there are a range of subject-related phenomena which cast doubt on the universal operation of an EPP-requirement. Consider, for example, the fact that subjects in Icelandic, German and Dutch may surface in a posi- tion which is clearly lower than the V2-adjacent canonical subject position (Spec-IP). The Icelandic examples in (21) illustrate (subjects in bold; Spec-IP underlined; % in- dicates that the structure in question is ill-formed in the context in question):16 (21) a. Við settum allar bækurnar fyrir málstofuna í lestrarsalinn … We put all books-the for seminar-the in reading-room-the i. … en það hafa margar af þessum bókum verið lesnar áður … but there have many of these books been read before ‘… but many of these books have been read before’ ii. %… en það hafa ___ verið lesnar margar af þessum bókum áður … but there have been read many of these books before b. Við settum færeyskar bækur fyrir málstofuna í lestrarsalinn … We put Faroese books for seminar-the in reading-room-the i. %… en það höfðu íslenskar bækur verið lesnar í fyrra … but there had Icelandic books been read last year ii. … en það höfðu ___ verið lesnar íslenskar bækur í fyrra … but there had been read Icelandic books last year ‘… but Icelandic books had been read last year’ (from Jonas 1996: 63) As shown above, it is not always necessary for Spec-IP to be filled: where a subject constitutes new information, as in (21bii), it does not undergo raising; it is only where the subject encodes old information, as in (21ai), that raising to Spec-TP is required. Similar effects emerge in German and Dutch, where “low” and raised (i.e., Spec-TP-occupying) subjects are also said to differ in respect of information- structure-related properties (e.g., specificity; cf. Diesing 1992 and in particular Rosengren2002).IfoneistomaintainthattheselanguageshaveanEPP-requirement, it is necessary to postulate the presence of expletive pro in the case of the structures we will not go into here, but see Baker (1996, 1997) and Alexiadou Anagnostopoulou (1998) for discussion. . Note that Icelandic is a symmetric V2 language, exhibiting V2 orders in matrix and em- bedded contexts. Following Schwartz Vikner (1996) and many others, I assume that all V2 clauses are CPs in which the finite verb is located in C.
  • 35.  The limits of syntactic variation featuring “low” subjects. Given the presence of an overt expletive in cases like (21bii), it is, however, not obvious that this is a reasonable postulation, particularly if LF considerations are borne in mind. Further, it is striking that raising vs. non- raising of the subject is necessarily associated with specific interpretive effects, a state of affairs which suggests an optional rather than an obligatory operation (cf. Chomsky 2001: 34 on the so-called Fox-Reinhart intuition on optionality, and also Biberauer Richards 2006; Biberauer 2004, 2006; Richards Biberauer 2005 for detailed discussion of the status of the EPP in Germanic). In at least some Ger- manic varieties, then, it is not so clear that the EPP as originally formulated holds. Looking beyond languages that have traditionally been labelled subject- prominent (cf. Li Thompson 1976), it is not so clear that the subject-centric EPP accurately captures the grammatical focus of so-called topic-prominent languages (cf. Modesto, this volume for overview discussion). More recently, it has also been proposed that it may be productive to think of languages as being either D-prominent or V-prominent (cf. Davies Dubinsky 2002), with the latter privileging a verbal rather than a nominal category. On this view, it could then be the case that I – or, following Pollock (1989) and others, the relevant I-related head in a more ar- ticulated structure17 – always requires a specifier, but it need not be the case that this specifier is always reserved for the subject or expletive. As with (the relevant subset of) topic-prominent languages, the absence of expletives in V-prominent languages could therefore be viewed as a consequence of the non-pivotal nature of subject/nominal categories in these systems, rather than as an indication of their semi-null-subject nature (cf. Section 5 below). Finally, researchers working on a wide spectrum of languages ranging from Germanic to Austronesian have also considered the possibility that what one might think of as D- and V-prominent languages may in fact differ in respect of the “size” of the category that undergoes raising to Spec-TP (cf. various contributions in Carnie Guilfoyle 2000 and Carnie, Dooley Harley 2005, and also Richards . It is worth noting that the genuinely universal status of the EPP becomes extremely difficult to maintain in the context of so-called “cartographic” approaches to functional structure within which I may be decomposed into several projections. What analyses couched in these terms have tended to highlight is that languages in fact appear to have available to them a range of IP-internal subject positions (cf. Bobaljik Jonas 1996; Cardinaletti 1997, 2004; in some cases, subject positions that have been proposed may in fact be located in the lower CP-domain, but we leave this matter aside here). If this is correct, the implication is that one cannot think of a single, EPP-imposed, obligatorily filled subject position in these languages. To “save” the EPP, one could potentially appeal to a “spreading” device of the kind mooted in Chomsky (2008) and in terms of which the phase-head C “spreads” its EPP-feature to one and only one of the I-heads in its domain. In pre-2008 Oh Phases terms, though, it is not clear how this would be achieved.
  • 36. Introduction  Biberauer 2005 and Travis 2006). In terms of this type of approach, then, the idea is that Spec-TP must be projected and filled, but that is not specifically filled exclusively by a subject; even where a language is D-prominent, it may permit or even require a larger category containing the subject to raise to Spec-TP. Biberauer (2004) argues this to be the case in Afrikaans, where the subject standardly piedpipes its containing vP to Spec-TP, thereby delivering Afrikaans’s characteristic SOVAux ordering, and Biberauer Roberts (2005, 2008) make a similar case for earlier English. Among the advantages of proposals of this sort is that they can account not just for unexpected subject-related phenomena – e.g., the absence in certain structures of an overtly rea- lised expletive where a language does not generally license expletive pro (expletive pro is not expected to be present if Spec-TP is occupied by a larger category) – but also for unrelated phenomena such as word order and word-order variation. Given the above, then, it would seem that there are a range of empirical and theoretical reasons why it is not feasible to maintain a universally instantiated priv- ileged subject position. One could, of course, maintain the projection-of-Spec-IP component of the EPP independently of its subject connection, but this represents a departure from the principle as originally formulated. What we have seen in this section, therefore, is that a number of prominently cited GB principles and parameters do not appear to be empirically supported in the manner originally predicted. As we will see below, this has led some researchers to reject the usefulness of the parametric enterprise as a whole and to question the extent to which it is in fact feasible to identify universally instantiated limits on variation. In the following section, we will, however, consider an alternative per- spective on this matter, one which was, in part, necessarily brought about by the shift to the current phase of the PP enterprise, namely the Minimalist Program. 3.4 Principles and Parameters in the Minimalist Program The Minimalist Program (MP) has its origins in the theoretical “stock take” that Chomsky Lasnik initiated at the start of the 1990s, after roughly 10 years of work within the PP model. Rather than rejecting GB and its predecessors and “starting afresh”, as critics sometimes suggest, it aims to critically evaluate the significant body of theoretical work produced previously and to subject it to the kind of Occam’s Razor assessment that characterises the normal course of scientific research more generally. Additionally, however, the MP, at least as pursued by researchers with a “biolinguistic” commitment (cf. Chomsky 2005, 2007, 2008 for discussion and references), also has at its core a new hypothesis, namely the Strong Minimalist Thesis (SMT) given in (22): (22) The Strong Minimalist Thesis (SMT) Language is an optimal solution to legibility conditions (Chomsky 2000: 97)
  • 37.  The limits of syntactic variation In terms of the SMT, the starting point for any MP model of the language faculty is that its design is such that it satisfies the requirements that it clearly must meet18 in respect of feeding linguistic information to and from the systems with which it interfaces – the conceptual-intentional (“meaning”) system and the ar- ticulatory-perceptual (“sound”) systems – in the most efficient manner possible. The assumption is not that this hypothesis, which in effect postulates a maxi- mally “well-designed” or “perfect” language faculty, is likely to be correct; instead, Chomsky’s rationale is that research within the PP programme had, by the early 1990s, advanced to the point where enough was understood for a hypothesis like the SMT to serve the kind of heuristic function that is likely to drive the UG enterprise forward (cf. Chomsky 2005, 2007 for discussion). Instead of simply continuing to add to the machinery already proposed in the haphazard way that often becomes inevitable in the context of a widely adopted research enterprise, the suggestion, then, was that the SMT should serve as a yardstick to determine whether particular theoretical notions can and should be maintained or not: the SMT requires that each theoretical assumption be subjected to the question of whether a principled explanation can be found for it, with principled being un- derstood as meaning “imposed by one or both interfaces” and/or “following from general considerations of computational efficiency” (cf. Chomsky 2000, 2001, 2004a, 2004b, 2005, 2007, 2008). A direct consequence of the adoption of the SMT is, of course, the need to minimise the number of purely language-internal entities one postulates. On the architectural front, this led to the elimination of syntax-internal levels (D- and S- structure) and also, crucially for our purposes, of the modules associated with the GB model of the grammar (the Phrase-Structure, Theta, Case, Bounding, Binding and Control modules). As noted in Section 3.3, at least some GB parameters were assumed to be associated with specific modules of the grammar. Thus the HP, for example, was clearly a phrase-structure module-related parameter, while the Wh- movement parameter was associated with the Bounding module. In the absence of the modules with which they were associated, it is clear that these parameters must be “located” elsewhere if they are to be retained within an MP grammar. In principle, the simplified MP architecture, comprising just a Lexicon, the Com- putational System (which “does” Narrow Syntax) and (possibly) two interface components, PF and LF, allows for two general possibilities: either parameters are located in the Lexicon or they are located at one of the interfaces. The notion that parameters might in fact be associated with lexical items is a long-standing one, first proposed during the GB era as the Lexical Parametrisation . Chomsky (1995 et seq.) refers to these as “virtually conceptually necessary” requirements.
  • 38. Introduction  Hypothesis (cf. Borer 1984; cf. also Manzini Wexler 1987 for further seminal discussion). As noted by Baker (this volume: 253), the idea that much, if not all, parametric variation is located in the Lexicon constitutes a core assumption in minimalist theorising. Baker (ibid) labels it the Borer-Chomsky Conjecture (BCC) and states it as follows: (23) The Borer-Chomsky Conjecture (BCC) All parameters of variation are attributable to differences in the features of particular items (e.g., the functional heads) in the lexicon. The appeal of the BCC is clear: if lexical items constitute the sole locus of para- metric variation, acquiring these will, in effect, amount to acquiring the language as a whole. In other words, on this view, “learning vocabulary” does not just entail learning the idiosyncratic component of language (i.e., that traditionally associ- ated with Saussurean arbitrariness) and, presumably at the same time, also what the phonotactic and morphological properties of the language in question are (i.e., two types of system-defining knowledge); it also entails acquiring the parametric profile of the language, fixing the unvalued parameters of S0 (cf. (5) above). Reconsidering the parameters discussed above, it is clear that a number of them are plausible candidates for reformulation in terms of the BCC. The NSP in (7), for example, in effect specifies two lexical requirements, namely that a lan- guage must have both a particular type of functional category (“richly” specified I) and a particular type of pronominal (pro). The HP in (6), by contrast, initially looks like a far less likely candidate, but this parameter is also amenable to reformulation in lexical terms and numerous researchers have in recent years done so, either explicitly or tacitly. Thus one might, for example, assume Kayne’s (1994) Antisymmetry proposal in terms of which phrase structure is exceptionlessly head-initial and propose that surface head-final orderings are the consequence of leftward movement of complements. In practice, what this entails is that the heads which surface head-finally bear a movement- triggering feature (an EPP-feature in current minimalist parlance), whereas those which surface initially do not. On this model, then, at least all the functional heads in harmonic head-final languages, and possibly heads generally, are lexically speci- fied as bearing movement-triggering EPP-features (precisely what determines the “size” of the moved constituent – i.e., whether we see DP-movement or “massive pied-piping” of the sort proposed in so-called roll-up analyses – seems to require independent specification; cf. discussion below).19 By contrast, the (functional) . Note that, in the context of a theory that assumes the distribution of EPP-features to con- stitute a central point of crosslinguistic variation (cf. also GGL this volume), with functional categories (and possibly lexical items) either bearing or not bearing this movement diacritic, the
  • 39.  The limits of syntactic variation heads in harmonic head-initial languages systematically lack EPP-features (except where these are optionally present in order to guarantee a non-neutral, move- ment-derived structure20). In disharmonic languages like German and Gungbe, in which phrases do not all exhibit the same head-directionality, the same system would obtain, with the result that the Lexicon in the case of languages of this type would contain both EPP-bearing (head-final) and EPP-lacking (head-initial) heads. Finally, in languages in which linearisation appears to play a lexical disam- biguation role (cf. the examples mentioned in Section 3.2), EPP-features would only be associated with the specific lexical items exhibiting final ordering.21 This last case suggests that not just functional categories, but individual lexical items may bear independent linearisation properties. In the context of the BCC, this does not sound problematic, but it is worth considering the implications of this possibility. One very obvious implication is that it would seem to entail massive redun- dancy in the context of all languages in which the HP is not operating “myopi- cally” (i.e., lexical item-sensitively; cf. Section 3.3 above) – that is, in all languages where head-directionality is more systematically determined on the basis of either syntactic category membership (the German/Gungbe case) or in an across-the- board fashion (the harmonic cases). In other words, ascribing head-directionality specifications (via diacritics or otherwise) to each and every lexical item would seem, in the vast majority of cases, to be complicating a system which could be significantly simpler if this information were located elsewhere (cf. also Baker, this volume, for the same argument). As such, this kind of proposal would not seem to be in keeping with the SMT in (22).22 GB Extended Projection Principle (discussed in Section 3.3) emerges as a stipulation that would require principled explanation: why would it be necessary for T (=I) in every language to bear an EPP-feature, whereas all other functional categories either may or may not do so? . Cf. Biberauer Richards (2006) on the difference between obligatory and optional EPP- features and the interpretive effects they bring about, and Biberauer, Holmberg Roberts (2007) for discussion of the necessary distinction between linearisation-ensuring features and those responsible for A’- and possibly also A-movement. . Precisely how this system would account for heads that may appear both initially and finally is not clear. Possibly, the relevant lexical entries would be underspecified for linearization information or perhaps one could, as suggested in note 10, follow Adger (2006) in proposing that certain types of lexically-related optionality may in fact involve two distinct lexical items with the same phonological and semantic characteristics, but different linearisation specifica- tions. Given how strongly languages in general seem to resist this latter “homonymy” scenario, however (cf. i.a. Kroch 1994), the latter may not be a plausible hypothesis. . Worth noting, however, is an alternative perspective on this matter, in terms of which the “redundancy” entailed by assigning an EPP-feature to all the categories in a given system in fact
  • 40. Introduction  Anotherseriousproblem,alsodiscussedinBaker(thisvolume),withthenotion that linearisation information is stored on a lexical-item-by-lexical-item basis is that it leads us to expect “a relatively smooth continuum of languages” (Baker, this volume: 360), in which there are many mixed (i.e., disharmonic) languages of dif- ferent kinds and relatively few harmonic languages. In reality, however, the facts seem to be rather different. Consider in this connection Greenberg’s original ob- servations, the obvious skewing in favour of harmonic languages registered in (2) above, and Dryer’s (1992, note 17) observation, which points to “disharmonies” having predictable origins not only within languages, but also crosslinguistically. Looking beyond the domain of the HP, Dryer (1998: 123–124) has observed that empirical work strongly suggests that languages “tend to cluster together around a certain type and that they become increasingly infrequent as we move away from the core type”. As Baker (this volume) notes, facts of this kind raise serious ques- tions about the currently quite generally accepted and largely unexamined idea that parametric variation is necessarily lexically specified. Significantly, linearisa- tion was precisely one of the domains that Fukui (1988) argued should fall outside the domain of lexical parameterisation (cf. his Ordering Restriction Hypothesis). Additionally, he also argued that parametric variation should be restricted to func- tional categories, with lexical categories like N and V, for example, not playing any role in parametric variation (cf. his Functional Parameterisation Hypothesis). This latter idea is one that is currently widely held, with minimalists frequently inter- preting the BCC as applying specifically to functional categories (FCs).23 As the contributions by GGL and Kayne in particular highlight, however, it is not nec- essary to adopt this approach and the presence versus absence of specific lexical items may in fact also play a central role in determining the more general syntactic options available to a given system. We return to this matter in Section 4 below. emerges as highly economical. If we adopt the perspective of a simplicity-driven acquirer who is trying to establish the directionality of the syntactic categories in their language (i.e., whether they are associated with an EPP-feature or not), one could argue that the simplest systems are those in which the acquirer simply postulates C[+EPP] or C[–EPP], i.e., the same EPP-specifica- tion for all categories (C) in the system (see i.a. Clark Roberts 1993; Roberts Roussou 2003 and Roberts 2007a for discussion of simplicity-driven acquisition). This does, of course, raise the complex question of “markedness” – specifically, whether “harmonic” systems should not actually be easier to acquire than “disharmonic” ones – but we leave this matter aside here. Thanks to Ian Roberts for discussion of this point . Under the view that lexical categories like N and V do not constitute primitive entities, but that “nominalisation” and “verbalisation” are in fact achieved via merger of acategorial roots with functional categories like n and v (cf. Marantz 1997; Chomsky 2001: 9), it is arguably im- possible for lexical categories to play a role in parametric variation independently of functional categories (for a very different view on the noun/verb distinction, see Kayne 2008).
  • 41.  The limits of syntactic variation For the moment, we conclude, firstly, that the BCC is a hypothesis which requires more critical examination than that to which it has been subjected to date and, secondly, that linearisation in particular may well not be the kind of parametric property that is best located in the Lexicon. Leaving aside the objections raised above, there is also the in recent years oft-discussed question of whether linear ordering should play any role in Narrow Syntax (NS), as it would presumably have to if it is already lexically specified. As has been pointed out many times, the computational system does not appear to be sensitive to linear ordering; only to hierarchical structure. As such, one might expect that linear ordering – even if this is ultimately determined via Kayne’s (1994) Linear Correspondence Axiom (LCA) – is only imposed at the stage at which it is self-evidently required, namely at PF (cf. Chomsky 1995: 335, 340). We pursue this matter below. One aspect of linearisation that is, however, usually thought to lend itself to lexical parameterisation is that which was previously encoded by the various movement parameters – V-to-I movement, N-to-D movement, V2 and Wh- movement, to name but a few. In each case, the relevant movement requirement can be lexically encoded as a movement diacritic (EPP-feature, etc.), with lan- guages lacking movement in the relevant domains correspondingly lacking this diacritic.24 There is little doubt that the easy “translateability” of the movement pa- rameters into BCC terms has played a significant role within minimalist research in foregrounding the role of the Lexicon in determining parametric variation. It is, however, worth noting that the matter may be significantly more complicated than “recastings” of this kind might suggest. Consider, firstly, the fact that movement parameters, if they are lexically encoded and implemented in the syntax, once again introduce linear ordering into Narrow Syntax. Furthermore, if Chomsky (2001: 37–38) and others are in fact correct in consigning (at least some) head- but not phrasal movement to the PF component (i.e., to an interface), it is not clear how the computational system can effectively “ignore” the movement dia- critics that will ultimately result in head-movement effects, while satisfying those which will ultimately deliver phrasal movement. It is, of course, conceivable that head-movement triggers are invisible (i.e., uninterpretable) to the computational system, while XP-triggers are not and that we are thus effectively dealing with two different types of movement diacritic. Given that certain head-movements do appear to take place in the syntax, however (cf. Lechner 2006 and Roberts 2006 on LF effects associated with head-movement), it would seem that there also needs to . Alternatively, one might, like Brody (1995), Bobaljik (2002), Landau (2006) and others, think in terms of P-features which determine which copy of a chain is ultimately pronounced.
  • 42. Introduction  be a Narrow Syntax-internally interpretable head-movement trigger – and, addi- tionally, since syntactic relations always involve head-head relations (cf. Chomsky 2004a et seq.), some means of signalling when head- rather than XP-movement is to take place (cf. Watanabe 2006 on “pied-piper” features). If some head-movement is PF-movement, then, quite an extensive inventory of movement-triggers would seem to be required. If it is not, there are still many currently unanswered questions as to how the computational system is able to determine the “size” of the category it is to move (cf. Biberauer Richards 2006; Biberauer Roberts 2006 and Watanabe 2006 for some recent discussion). Let us now turn to the alternative potential loci for Minimalist parameters: the interfaces. Since the possibility of LF parameters has not received much at- tention to date, we will focus on the PF interface here (cf., however, Ramchand Svenonius, this volume, for discussion of the role that LF may potentially play in determining parametric variation). As noted above, the PF interface would seem the most plausible locus for lineari- sationgenerallyandthereforealsoforparametricallyvariantaspectsoflinearization.25 Chomsky himself has always held that “[t]here is no linear order in the N[umeration] → LF computation” (Chomsky 1995: 335; cf. also Chomsky 2001: 7) and that linear ordering be imposed at PF, where the LCA may, but need not necessarily, determine linear precedence. Actually formulating a plausible PF linearisation system, let alone a credible PF version of the HP has, however, proven quite difficult. As regards linearisation more generally, the problem is ensuring that the structures sent to PF are in fact unambiguously linearisable. If one assumes that the very restrictive (Spec-Head-Comp) phrase structure assumed by Kayne (1994) is the only kind permitted in NS (i.e., that a very strict version of the X-bar tem- plate is in fact encoded as a UG principle), the problem of course becomes trivial: such structure is always linearisable via the LCA. On this view, phrase structure is, however, strictly constrained precisely to ensure a linearisable output, with a “simple PF” coming at the expense of a more complicated computation (including, for example, unrealised heads and specifiers, and “extra” leftward movement wherever head-final structures result – see discussion above and note 22).26 As such, a Kaynian approach to linearisation must be regarded as a departure from . For a minimalist analysis which proposes relocating the HP to Narrow Syntax, however, see Saito Fukui’s (1998) Parameterised Merge proposal. . As a result of the latter, particularly if leftward movements may be of the remnant variety often assumed in Kaynian analyses, there is, of course, also one aspect of the PF “computation” that is not simplified, namely the copy spellout process, i.e., determining which copy/copies to spell out and which to suppress (cf. Corver Nunes 2007 for discussion).
  • 43.  The limits of syntactic variation the SMT and alternatives should at least be investigated to determine whether a strict LCA-based system in fact constitutes a genuine case where there is a “gap between SMT and the true nature of FL [the Language Faculty – TB]” (Chomsky 2007: 3). An obvious way to relax the Kaynian constraints on phrase structure is to propose, as many researchers following Moro (2000) have done, that syntactic structureneednotbelinearisableateverystageofthederivation,butthatstructures must be linearisable at the stage at which they are sent to PF. Within proposals of this kind, linearisation is therefore only partially a PF matter, with initially unlin- earisable structures being “rescuable” via subsequent movement operations which may – as in Moro’s Dynamic Antisymmetry approach – or may not be directly aimed at neutralising troublesome “points of symmetry” (i.e., symmetrical and thus unlinearisable structures). Languages may then differ as to the points of sym- metry that are generated during the course of the computation27 and possibly also as to when and how they are neutralised (cf. Hsieh Sybesma 2007 for a phase- based implementation of DA and Takita 2007 for a proposed symmetry-breaking parameter). Approaches of this type can, however, clearly be criticised for “look- ahead”, i.e., NS operating in such a way that subsequent requirements – here, PF legibility – can be met.28 We leave this and other matters aside here since our main concern is to consider how feasible it in fact is to completely banish linearisation information from NS. AverydifferentapproachisthatofRichards(2004).Heproposesparameterising the LCA such that PF systematically ignores one set of symmetric c-command relations (cf. also Epstein et al. 1998), thereby capturing the effects of the HP in the manner outlined for VP below: 29 (24) The Parameterised LCA (a PF HP) a. VO = Ignore all O V b. OV = Ignore all V O . Points of symmetry may in principle be either head-head or XP-XP adjunction pairs. . Additionally, it is worth noting that DA-based approaches once again necessitate the postulation of a range of movement-triggering factors since movement in natural languages is clearly not limited to the symmetry-breaking variety. . Richards’s proposal takes as its point of departure the observation that the first two ele- ments Merged in any structure will be heads, i.e., elements which necessarily symmetrically c-command each other and which will therefore also necessarily be unlinearisable in LCA terms without some additional instruction. The Parameterised LCA is intended to supply this instruction and, simultaneously, to state a point of parametric variation. For a very different solution to the head-head merger problem, see Kayne (2008).
  • 44. Introduction  In terms of (24), then, head-initial languages consistently ignore PF linearisation instructions (i.e., symmetric c-command structures) ordering complements before heads (a), whereas head-final languages consistently ignore the opposite set of linearisation instructions. This system does therefore remove HP-related linearisa- tion information from NS, allowing the HP to be stated as an interface parameter. As with the original HP, however, it is not so clear how an interface HP operates in the context of disharmonic languages, i.e., those which do not set (24) so that it applies across-the-board. One might propose that (24) is sensitive to syntactic- category information and that lexically specified linearisation instructions can over- ride its across-the-board application (cf. note 10), or possibly that the application of (24) is determined on a phase-by-phase basis, with some languages opting for a subset of phases which are linearised in accordance with (a), while the remainder are linearised in accordance with (b). The implications of these possibilities remain to be worked out in further detail, with linearisation patterns that systematically fail to surface in the world’s languages (i.e., impossible linearisations) potentially rep- resenting a crucial empirical testbed (cf. Biberauer, Holmberg Roberts 2007 for discussion, and also Kural 2005 and Cinque 2005, 2008). For the moment, however, what is worth noting about the HP and its potential MP reformulations is that an exclusively lexically based approach would seem to entail considerable redundancy for harmonic languages in particular, effectively making the most idiosyncratic of the disharmonic exceptions seem the most “natural”, whereas a purely PF-oriented approach seems best suited to the harmonic languages while raising questions as to how the disharmonic orders are to be accounted for. If UG is to contain a linearisation parameter, it clearly needs to be one which could in principle be set so as to generate both the harmonic and the disharmonic word orders and it is not presently clear either how this can be achieved or whether (a) suitable linearisation parameter(s) will require a departure from the SMT. We have already noted above that movement parameters would, at first sight, seem trivially recodable as lexical parameters involving the presence versus absence of movement diacritics. Another possibility is that these parameters may in fact be located at PF. Bobaljik (2002) offers a proposal along these lines, sug- gesting that the difference between overt and covert movement may in fact not lie in whether an Agree operation between two elements is or is not associated with a movement diacritic (presence of EPP-feature = overt movement; absence of EPP- feature = covert movement), but in which copy of a moved element PF chooses to spell out. In Bobaljik’s terms, then, overt movement entails spelling out the highest copy of a moved element whereas covert movement falls out from spelling out the lowest copy; movement therefore always takes place in NS, but this plays no role in linearisation (cf. Brody 2000 for an in many ways similar approach focused on a different empirical domain). This proposal is particularly appealing in the context
  • 45.  The limits of syntactic variation of covert movements that were during the GB era assumed to take place in the der- ivation from S-structure to LF (QR, wh-movement in wh-in situ languages, etc.). If Bobaljik’s proposals are on the right track, it may therefore be the case that GB parameters like the Wh-movement Parameter have both a lexical and an interface component (cf. the oft-discussed correlations between (non-)movement and, inter alia, the availability of what Cheng 1997 designates clause-typing particles and the morphosyntactic structure of wh-elements). To conclude on the question of how GB parameters may be reformulated in the current framework: we have seen that, while the available loci would seem to be clear (the Lexicon and one or more of the interfaces), GB parameters are not equally easily captured in minimalist terms. This may, of course, indicate various things. One possibility – and, given the course of normal scientific progress, it is likely to be a very strong possibility – is that some or all of the GB parameters were misfocused and that they should therefore be viewed, in the context of the PP enterprise, as “first pass” attempts at pinpointing recurring points of syntactic variation (see Section 3.5 below). Another possibility, which arose in the course of our discussion of the Wh-movement Parameter, is that parameters may in fact be the reflex of potentially quite complex interactions between various aspects of language structure (lexical, morphological and syntactic), with the result that for- mulating and “localising” them might involve consideration of a greater number of factors than has traditionally been done (but see Haspelmath, this volume, for a different perspective on the likelihood of cross-domain relations). At the most general level, though, what the minimalist heuristic strongly suggests about some of the GB parameters is the fact that they may not ultimately be formulable in a manner which respects the SMT. Let us consider this point in a little more detail. As Chomsky (2007: 1) notes, considerations of language acquisition had, until the advent of the PP programme, always led Chomskyans to assume that UG must be “rich, highly structured and substantially unique”: if the acquirer was to be able to abduce (i.e., correctly infer on the basis of innately specified knowledge, cf. Peirce 1955) the grammar underlying the PLD to which they are exposed (cf. (5) above), a suitably specific UG seemed a prerequisite. The PP programme led to the revision of this notion since it opened up the possibility of an underlying system in which parameter settings might differ only slightly, while the surface effects of those slight differences might be extensive (consider, for example, the considerable surface difference between consistently head-initial and consistently head-final languages, where the relevant underlying difference, in GB terms, is simply the way in which the HP is set). Within the MP, consideration of the role of a “third factor”, whose potential relevance had previously only been mentioned in passing, but which was never explicitly included in the acquisitional scenario alongside UG (innate endowment) and PLD (experience), has become central.
  • 46. Introduction  This “third factor” entails principles of mental/neural organisation that are not specific to the faculty of language, but are common to all cognitive faculties (cf. Chomsky 2005, 2007, 2008, Hauser, Chomsky Fitch 2002 and the ensuing debate for further discussion). Furthermore, in terms of the hypothesis driving minimalist research – the SMT in (22) – the null hypothesis is that UG, or the specifically linguistic infor- mation attributed to the human genome, is minimal. In other words, minimalist research, unlike that which went before, explicitly proceeds on the assumption that, in the absence of evidence to the contrary, as little as possible should be at- tributed to UG – specifically, only those properties which do not appear to be re- ducible to more general, possibly partially functionally motivated principles. The question is, as Chomsky (2005: 11) puts it: “How far can we progress in showing that all … language-specific technology is reducible to principled explanation, thus isolating core processes that are essential to the language faculty?”. In the context of SMT-oriented theorising, then, a rich parametric inventory of the type assumed during the GB era would constitute a considerable departure from the SMT and, as such, should only be assumed if the evidence points to the need for it. In the first instance, however, the goal should, as in minimalist theorising more generally, be to consider how much of what was previously proposed can be accounted for by appealing to “non-domain-specific” principles (e.g., those governing efficient computation). The greater the number of language-specific parameters that ultimately needs to be proposed, the more extensive the depar- ture from the SMT will be. As we have seen above, first attempts suggest that it is possible to “locate” some GB parameters in the Lexicon and also that it might be possible to view others as the consequence of particular options which arise at the interfaces. As such, these parameters would represent content that needs to be ascribed spe- cifically to UG, but at least one aspect of their nature – specifically, the fact that they can be formulated in lexical and/or interface terms – is such that one could offer a principled answer to the question: “Why those parameters and not others that one could possibly formulate?”. Answer: the parameters in question are di- rectly associated with the “virtually conceptually necessary” components of the language faculty and do not entail postulating additional modules or otherwise complicating the overall architecture of the language faculty. In the context of the current phase of the PP programme, then, the question is not just what sorts of innately specified knowledge might facilitate language acquisition in the manner that it occurs, but additionally also whether knowledge that seems to require innate prespecification can in fact be seen to take the form it does by virtue of this having been “imposed” by considerations of “good design”. Insofar as the content of UG and the design of the language faculty more generally can be understood
  • 47.  The limits of syntactic variation as having the form that they do in virtue of “third factor” considerations, it is, in Chomsky’s terms, possible to go “beyond explanatory adequacy”; and the MP aims to probe the extent to which it is possible to do so, approaching the ques- tion both “from above” (i.e., continuing the search for explanatory adequacy) and “from below” (i.e., attempting to discover the extent to which “third factor” considerations have shaped UG). So far, we have speculated that GB parameters which can be viewed as being “located” in one of the “virtually conceptually necessary” components of the lan- guage faculty might be thought of as meeting a “good design” requirement, namely that they do not require the postulation of additional, purely language faculty- internal architecture. A further “good design” question that the SMT forces us to ask is whether it is in fact necessary to view each of the specific parameters that were previously identified as independent entities within UG. Consider, for example, the individual head-raising parameters (e.g., V-to-T, N-to-D, V-to-C, etc.): these clearly share the same format – X raises to Y – and the question the SMT raises is whether UG actually requires each of these parameters to be speci- fied individually. One might, for example, alternatively think of a general template which serves, in much the same way as the X-bar template during the GB era, to determine aspects of syntactic structure (cf. GGL, this volume); or, more radically, that even the template may arise as a result of the interplay between PLD and “third factor” considerations (cf. Zeijlstra and Evers van Kampen, this volume and also Roberts 2007d for different versions of this idea). Evidently, then, there is a very general question within the MP as to the forms that minimalist parameters might take. A further question concerns their status within UG: are they present in “unset” form at S0 or can they arise during the process of acquisition? These are questions which have not featured prominently on the MP research agenda to date, but which clearly need to be addressed if we are to be able to offer a coherent and up-to-date minimalist answer to the question of how principles and param- eters constrain syntactic variation. In the absence of a coherent minimalist theory of this type, it is not in fact possible to meaningfully evaluate the success of the PP enterprise to date. The remainder of Section 3 focuses on a recent evalua- tion by Newmeyer (2005a, 2005b and 2006) which starkly underlines the need for more systematic work on notions like “minimalist parameter”. 3.5 Recent criticism of the parametric enterprise Newmeyer (2005b: 185) states seven desiderata that a parametric approach to syntactic variation should meet if it is to be superior to a potential rule-based alternative. These include descriptive simplicity, binarity, smallness of number, implicational organisation and a rich deductive structure. He then considers the
  • 48. Introduction  extent to which GB parameters, notably the NSP, satisfy these desiderata, ul- timately concluding that they fall short in various ways, and that the limits of syntactic variation should be sought in “the interplay between unparameterized UG principles, language-particular rules, and processing principles sensitive to structural relations holding among grammatical elements” (2005b: 225). In other words, Newmeyer proposes that parameters should play no role in our model of UG and that the parametric enterprise can be viewed as having failed. It is, however, important to note that his conclusions are based exclusively on consid- eration of GB parameters.30 As such, they can at most be interpreted as potential conclusions as to the feasibility of the “first pass” parameters of Stage I of the PP enterprise, and not as conclusions on the enterprise as a whole; as was discussed in Section 3.4, minimalist parameters will necessarily differ from their GB prede- cessors in various ways and work on precisely what the inventory of minimalist parameters might ultimately look like has scarcely begun. Many of the contribu- tions in this volume are centrally concerned with this question, which was also a major focus of the conference at which they were first presented. What we will briefly consider here is the extent to which some of Newmeyer’s criticisms can also be levelled against the parametric enterprise as it can now be and is now being approached. Our discussion here will focus on the descriptive simplicity, richness of deductive structure and smallness of number desiderata (cf. discus- sion of implicational organisation in Section 4 below, and Roberts Holmberg 2005 for discussion of the other desiderata). Firstly,letusbrieflyconsiderthematterofdescriptivesimplicityandtherelated matter of smallness of number. As discussed in Section 3.4 above, the hypothesis driving minimalist research (the SMT in (22)) takes as its point of departure that UG does not contain domain-specific information. As we have seen, it may well be possible that it need not contain specific parameters of the type assumed during the GB era and that any parametric options that are innately specified may take the form of extremely general and thus also descriptively extremely simple tem- plates (cf. GGL, Zeijlstra and Evers van Kampen for suggestions as to the format that such templates might take). If this thinking proves to be correct, it may then well be the case that a very small number of (or possibly even no) innately speci- fied templates gives rise to a diverse range of steady-state (Ss) parameters. This line of thinking has an important consequence in the context of the point New- meyer (2005b: 185) raises regarding the complexity of the variation observed in the world’s languages. According to Newmeyer, this is such that it simply is not . Cf. Roberts Holmberg (2005) for a critical evaluation of some of the conclusions Newmeyer draws.
  • 49.  The limits of syntactic variation feasible to postulate only a small inventory of innately specified parametric options, with the result that one might as well assume that children do not draw on innately specified parametric choices, but instead learn language-specific rules which capture the relevant dimensions of variation. If it can, however, be shown that it is possible for the Ss parameters to be “constructed” during the course of acquisition, that these do not have to be individually prespecified and that their effects might nevertheless be wide-ranging, Newmeyer’s conclusion no longer follows. Let us turn now to the matter of the “rich deductive structure” traditionally ascribed to parameters. As we have seen in Section 3.4, lexically specified parame- ters associated with specific lexical items and/or functional categories have played a prominent role in minimalist theorising to date. At first sight, these might seem to be parameters which may not always meet the “rich deductive structure” desid- eratum since their effects are naturally limited to the domains in which the relevant lexical item is present, with the result that lexically specified parameters are often (erroneously–seethediscussionofpossibleencodingsoftheHPabove,andbelow) equated with microparameters, i.e., parameters that define the kinds of small-scale differences in terms of which closely related dialects may differ (cf. Kayne 2005b for discussion). Newmeyer (2005b, 2006), like i.a. Safir (1987: 78) and Bouchard (2003) before him, is of the view that parameters of this type entirely undermine the rationale for parameters, essentially reducing the term parameter to a synonym for rule. In the words of Newmeyer (2006: 9), “if every difference or even most dif- ferences are attributed to a difference in “parameter setting”, then saying that two languages have different parameter settings merely becomes an obfuscatory way of saying that they have different rules”. For Newmeyer, then, only parameters of the type Baker (this volume) designates macro- and medioparameters have a suf- ficiently rich deductive structure to justify speaking of “parameters” as opposed to “rules”. Like Pica (2001), however, he believes that the past quarter century of research within the parametric framework has shown that parameters of this kind are not empirically motivated, leaving us, at best, with lexically specified micropa- rameters, which, for him, are in fact most realistically viewed as language-specific rules. For Newmeyer, then, one of the central failings of the parametric enterprise is its failure to bring to light plausible, empirically substantiated parameters which have wide-ranging and at first sight seemingly unrelated effects. Worth noting about Newmeyer’s conclusion are the assumptions on which it is based. Firstly, it takes for granted that lexically specified (micro) parameters necessarily have very limited surface effects of the kind that can be very straight- forwardly captured in generalised rule format. That microparameters could also entail “deep” effects, with the presence of a lexical item or particular feature speci- fication, for example, determining the (un)availability and/or the form of further, often seemingly quite unrelated structures in the language is not entertained as a
  • 50. Introduction  possibility. As Kayne’s contribution to this volume and his and other researchers’ microparametric work more generally has strikingly shown, however, points of microparametric variation do not have to differ from medio- or macroparametric ones in this regard (cf. also Rizzi 2004: 332 on this point). Furthermore, as Baker (this volume: 355) points out, “[i]t is perfectly possible that a lexical parameter … could have a substantial impact on the language generated, particularly if it concerned some very prominent item (such as the finite Tense node)”. And, con- versely, a macroparameter like the Polysynthesis Parameter might only have a very limited impact on a language lacking extensive agreement (cf. also Baker 1996: note 35). Kayne (2005b: 5–6) likewise registers the opacity surrounding the notion of parametric “size”, pointing out that “what seems “dramatic” [in terms of number of empirical reflexes or range of effects – TB] depends on expectations that may themselves be somewhat arbitrary”. Given all of this, it seems fair to say that it is not at all clear that the discovery and existence of “large” parameters, assuming the meaningfulness of a notion of this type, is crucial to the success of the parametric enterprise. Furthermore, given the potentially extensive effects of “smaller” pa- rameters, it is also cannot be said to follow that the possible non-existence of “large” parameters entails that one might as well conceptualise crosslinguistic variation in terms of language-specific rules. Another point worth noting in this connection is that Newmeyer’s conclu- sions rest on the assumption that macroparametric research is no longer being seriously entertained by researchers engaged in the parametric enterprise. As Baker himself indicates in his contribution to this volume, it is certainly true that genuinely macroparametric research is not receiving nearly the attention that Kayne-style microparametric research is. This is, however, different from saying that researchers are no longer engaged in this type of research at all, and various contributions to this volume show that this is not in fact the case – cf. Modesto on a potential Topic Parameter, Zeijlstra on a Negative Concord Parameter, Öztürk on a Non-configurationality Parameter and Avram Coene on a Split INFL Pa- rameter which, unlike Bobaljik Thráinsson’s (1998) original, is hypothesized to have large-scale effects both in the clausal and the nominal domain. And beyond the articles collected here, Huang’s (2006) work on the synchronic and diachronic consequences of a so-called Analyticity Parameter, in some senses the diametric opposite of Baker’s (1996) Polysynthesis Parameter, constitutes further evidence that this line of research is also receiving attention. Further, this volume also contains papers which consider parameters that might, in Baker’s terms, be thought of a medioparameters – cf. Lekakou’s Aspect Parameter, Fernandez-Salgueiro’s Case-F Valuation Parameter and Neeleman Szendröi’s Radical Pro-drop Param- eter. In sum, then, the contributions in this volume do not support the view that present-day parametric research is exclusively microparametric or that “larger”
  • 51.  The limits of syntactic variation parameters are no longer on the research agenda. Furthermore, as already noted above and as is also specifically highlighted in Baker’s paper, there is, quite indepen- dently of the amount of research being done on non-microparametric syntax, unde- niable empirical evidence – uncovered by both typologists and generativists – that the world’s languages robustly exhibit skewings of various kinds; and, moreover, skewings that often cannot be economically stated in rule form (cf. the discus- sion in Section 3.4 of lexical reformulations of the HP). These empirical facts clearly require explanation, and it is not obvious how a purely (language-specific) rule-based account of the kind Newmeyer espouses gets us any closer to under- standing them. To conclude, then: consideration of just three of Newmeyer’s desiderata in relation to the current stage of the parametric enterprise would seem to lead to very different conclusions as to its potential value to those drawn by Newmeyer himself. There is no denying the fact that core questions relating to the number, format, size, interrelationship and, in fact, given Chomsky’s “three factors” specu- lations, the very existence of innately prespecified parameters do not have clearcut answers at this stage of the PP enterprise. These are matters which need to be at the forefront of the MP research agenda to a much greater extent than they currently are if a meaningful evaluation of the parametric aspects of the PP en- terprise as it stands now is to become a realistic prospect. That the GB parameters which are the focus of Newmeyer’s evaluation do not constitute plausible innately specified constraints on syntactic variation is beyond dispute; what remains to be seen now is whether or not the same will be true of the current phase of the PP programme. 3.6 Conclusion This section has shown that the factors determining the limits of syntactic varia- tion are in many respects still poorly understood. Accordingly, questions as funda- mental as whether innately specified knowledge plays a role or not and if so, what form such knowledge should take are still under debate. Against this background, what the present volume aims to do, then, is to give a flavour of current thinking on the nature and value of parameters and parametric theory in accounting for syntactic variation. The remainder of this introduction will now focus more specifically on the contents of its sixteen papers and the issues that they address. The volume is or- ganised into four sections which respectively focus on the following topics: • the locus of (parametric) variation (cf. Section 4) • the NSP (cf. Section 5)
  • 52. Introduction  • parametric clustering (cf. Section 6); and • the acquisition of parameters (cf. Section 7) 4. The locus of parametric variation This section considers the under-discussed question of “where” in terms of the overall architecture of the grammar we might expect points of syntactic variation to be specified. It opens with a paper by Haspelmath which aims to fill a gap in the currently existing literature on (the limits of) syntactic variation by providing (i) a system- atic comparison of the different ways in which parametrically and functionally oriented linguists traditionally approach the limits question and (ii) a critical evaluation of the success the two approaches have achieved. Haspelmath’s main contention is that the recurrent patterns in the world’s languages – so-called phe- nomenological universals in his (2004) terminology – require an observationally adequate account, i.e., one that distinguishes possible from impossible expressions in a language (p. 2) and that does not make any claims about speakers’ knowledge; attempts at achieving any “higher” level of adequacy and, particularly, at trying to gain insight into the workings of the human mind via the study of syntactic limits are not, in Haspelmath’s view, feasible nor, given the availability of functional and processing explanations, even necessary. The proposal is that languages should (contra Meillet 1903: 407) not be viewed as “systems in which everything hangs together”. Consequently, so-called cross- domain implications where, for example, semantic organisation correlates with aspects of phonological organisation, should be expected to be rather rare, whereas so-called intra-domain implications where, for example, one aspect of phonology correlates with one or more other aspects of phonological organisation, need not be so. In keeping with these expectations, typological research has revealed a great many intra-domain implications, many of which are, according to Haspelmath, amenable to a functional explanation (cf. pp. 96–100). Cross-domain implications, on the other hand, have proved rare, producing only weak correlations which are in many cases quite difficult to distinguish from areal effects (cf. p. 19). The conclusion to be drawn from this, then, is argued to be that a research programme which seeks to understand apparently unrelated surface phenomena, particularly where these do not belong to the same domain, as the reflexes of a far smaller number of parameters is misfocused: the data do not warrant such a programme and it is doomed to fail in the same way in which previous attempts to propose “holistic” lan- guage types have failed (cf. the fate of the 19th century morphological typologies). Instead, the question that should be being asked about recurrent patterns, given
  • 53. Introduction  restrictions of computational efficiency and conditions of the interfaces” (p. 118). On this basis, they identify five “core” parameter schemata, leaving open the possibility of further schemata relating to phonetic nullness and lexico-semantic parameterisation. In the context of GGL’s so-called Parameters and Schemata (PS) model, then, the contents of UG is restricted to a limited number of func- tional features, and lexical categories, together with “a tiny class of parameter schemata” (p. 122); the parameters with which we are familiar are not therefore linguistic primitives; instead, they are constructed on the basis of innately speci- fied features, lexical categories and schemata, which are the actual primitives. Aside from the new possibilities that this type of approach opens up in the domain of language acquisition (cf. Section 3.4 and 3.5 above), it also makes a very interesting new prediction about the extent to which grammatical variation is expected to emerge over time. As GGL note, current minimalist-inspired theories might lead one to conclude that the amount of variation should de- crease over time, whereas the linguistic reality suggests that the opposite is in fact true – consider, for example, the immense grammatical diversity found in Italo-Romance, a language family which has a 1500–2000 year history, com- pared to the restricted grammatical diversity found in the various varieties of Afrikaans, a language which has only existed as a variety speakers think of as an independent language, distinct from Dutch since the late 19th century (cf. Deumert 2004).32 GGL argue that their PS model predicts that grammatical diversity will increase over time. In their words: “once the introduction of a pa- rameter schema into the language faculty is justified for a single case, it will … bring about every more potential variation for all possible combinations of relevant entities in the lexicon (features and categories)” (p. 123). Finally, GGL also consider the role that parameters – viewed in the de- rivative, schemata-determined manner outlined above – might potentially be able to play in facilitating a better understanding of the phylogenetic relations between languages. They argue that the correctness of the phylogenetic predic- tions made on the basis of MGP, firstly, corroborates the validity of postulating parameters and, secondly and more importantly, opens up the possibility of using parametric values as comparanda in seeking to establish linguistic relationships. . As Ian Roberts (p.c.) observes, it should, however, be noted that it is too simplistic to view the emergence of grammatical variation as a process which proceeds uniformly: change can sometimes occur in sudden spurts, as seems to have been the case in Middle English (cf. Fischer et al. 2000) and 9th century Icelandic (cf. Hróarsdóttir 2000); conversely, languages may remain fairly static for relatively long periods of time, at least as far as certain properties are concerned – consider, for example, the lack of change in most of the Romance languages with respect to their NSL status.
  • 54.  The limits of syntactic variation For GGL, then, parametric theory, rigorously pursued, may facilitate insights into a wide range of issues, going well beyond the points at which languages may differ. Zeijlstra’s paper, like GGL’s, pursues the question of what a truly minimalist, plausibly specified UG would contain and the proposal he makes resonates with theirs in various interesting ways. Zeijlstra’s specific focus is the question of the identity of the set of formal features available to UG and his central proposal is that this set is in fact empty. Instead, the inventory of formal features that can head a functional projection in a given language is argued to be fixed on the basis of PLD, during first-language acquisition (cf. also Evers van Kampen for a similar pro- posal in another domain). Strikingly, Zeijlstra and GGL’s proposals converge on one of the choice-points requiring fixing during language acquisition – Is feature F formalized as a formal (grammatical) feature? (cf. the first of GGL’s schemata). Where they differ is in respect of the ontological status that they assign to this choice-point: for GGL, it is one of the innately specified schemata in UG, whereas for Zeijlstra, it simply constitutes a linguists’ description of one of the major types of PLD-driven choice that children make during acquisition. In terms of Zeijl- stra’s Flexible Formal Feature Hypothesis (FFFH), children must determine for the features encoded in their language whether a given feature is a purely semantic feature which has C-I-relevant content, but none which is relevant to the compu- tational system (i.e., NS) or whether it has both. In the latter case, a formal (i.e., grammatical) feature will be postulated. The cue for deciding the status of a feature is the presence vs. absence in the PLD of semantically vacuous doubling: where it is present, the child “notices” and a formal feature is postulated; where it is absent, there is nothing to “notice” and therefore nothing in effect happens as the FFFH does not assume the prior existence of an entity requiring “negative setting” (cf. also Roberts 2007d: 17–18). Zeijlstra presents the crosslinguistic distribution of Negative Concord phenomena as a case study to illustrate the workings of the proposal, demonstrating that [negation] must be analysed as a formal feature in Negative Concord languages, where negations do not cancel each other out. In so- called Double Negation languages, by contrast, each negative element does in fact bear negative meaning, with the result that children acquiring these languages will conclude that [negation] is always and only a semantic feature. In general, then, the proposal is that instances of syntactic agreement (“concord”) will signal the presence of a formal feature in the language being acquired. As Zeijlstra notes, the proposal has clear implications not only for the de- scription of synchronically attested varieties, but also for the manner in which diachronic changes will take place. In the latter context, it is clear that the loss and/ or introduction of sound-meaning forms (words, clitics, inflections, etc.) bearing features which consequently are or are not doubled would be expected to result
  • 55. Random documents with unrelated content Scribd suggests to you:
  • 56. Eskimo, and the so-called “shamans” of southern Alaska, but, as far as we could see, do not possess the power and influence usually elsewhere ascribed to this class. It was exceedingly difficult to obtain any definite information concerning these people, and we only discovered casually that such and such a person was a “doctor” by hearing that he had been employed in a certain case of sickness, or to perform some ceremony of incantation. We did not even succeed in learning the name of this class of people, who, in talking with us, would call themselves “tûktĕ,” as they did our surgeon. On one occasion some of the party happened to visit the house of a sick man where one of these “doctors” was at work. He sat facing the entrance of the house, beating his drum at intervals, and making a babbling noise with his lips, followed by long speeches addressed to something down the trapdoor, bidding it “go!” We were given to understand that these speeches were addressed to a tuɐña or supernatural being. 576 Their only idea of direct treatment of disease is apparently to apply a counterirritant by scarification of the surface of the part affected. We know of one case where a sufferer from some liver complaint had inflicted on himself, or had had inflicted upon him, quite a considerable cut on the right side with a view of relieving the pain. We also know of several cases where the patients had themselves cut on the scalp or back to relieve headache or rheumatism, and one case where the latter disorder, I believe, had been treated by a severe cut on the side of the knee. A similar practice has been observed at Plover Bay, Siberia, by Hooper, 577 who also mentions the use of a kind of seton for the relief of headache. They also practice a sort of rough-and-ready surgery, as in the case of the man already mentioned, whose feet had both been amputated. One of the men who lost the tip of his forefinger by the explosion of a cartridge was left with a stump of bone protruding at the end of the finger. Our surgeon attempted to treat this, but after two unsuccessful trials to etherize the patient he was obliged to give
  • 57. it up. When, however, the young man’s father-in-law, who was a noted “doctor,” came home he said at once that the stump must come off, and the patient had to submit to the operation without ether. The “doctor” tried to borrow Dr. Oldmixon’s bone forceps, and when these were refused him cut the bone off, I believe, with a chisel. They appear to have no cure for blindness. We heard nothing of the curious process of “couching” described by Egede in Greenland, p. 121. We had no opportunity of observing their methods of treating wounds or other external injuries. Sufferers were very glad to be treated by our surgeon, and eagerly accepted his medicines, though he had considerable difficulty in making them obey his directions about taking care of themselves. After they had been in the habit of receiving the surgeon’s medicine for some time, one of the Utkiavwĭñ natives gave Capt. Herendeen what he said was their own medicine. It is a tiny bit of turf which they called nuna kĭñmölq, and which, therefore, probably came from the highland of the upper Meade River, which region bears the name of Kĭñmölq. We were able to get very little information about this substance, but my impression is that it was said to be administered internally, and I believe was specially recommended for bleeding at the lungs. Possibly this is the same as “the black moss that grows on the mountain,” which, according to Crantz 578 was eaten by the Greenlanders to stop blood-spitting. CUSTOMS CONCERNING THE DEAD. Abstentions.— From the fact that we did not hear of any of the deaths until after their occurrence, we were able to learn very few of their customs concerning the dead. The few observations we were able to make agree in the main with those made elsewhere. For instance, we learned with tolerable certainty that the relatives of the dead, at least, must abstain from working on wood with an ax or hammer for
  • 58. a certain period—I believe, four or five days. According to Dall, 579 in the region about Norton Sound the men can not cut wood with an ax for five days after a death has occurred. In Greenland the household of the deceased were obliged to abstain for a while from certain kinds of food and work. 580 A woman from Utkiavwĭñ, who came over to the station one day in the autumn of 1881, declined to sew on clothing, even at our house, because, as she told Lieut. Ray, there was a dead man in the village who had not yet been carried out to the cemetery and “he would see her.” After consulting with her husband, however, she concluded she could protect herself from him by tracing a circle about her on the floor with a snow-knife. In this circle she did the sewing required, and was careful to keep all her work inside of it. One of the natives informed me that when a man died his labrets were taken out and thrown away. I remember, however, seeing a young man wearing a plug labret of syenite, which he said had belonged to an old man who died early in the winter of 1881-’82. It was perhaps removed before he actually died. Manner of disposing of the dead.— The corpse is wrapped up in a piece of sailcloth (deerskin was formerly used), laid upon a flat sled, and dragged out by a small party of people—perhaps the immediate relatives of the deceased, though we never happened to see one of these funeral processions except from a distance—to the cemetery, the place where “they sleep on the ground.” This place at Utkiavwĭñ is a rising ground about a mile and a half east of the village, near the head of the southwest branch of the Isûtkwa lagoon. At Nuwŭk the main cemetery is at “Nexeurá,” between the village and Pernyû. The bodies are laid out upon the ground without any regular arrangement apparently, though it is difficult to be sure of this, as most of the remains have been broken up and scattered by dogs and foxes. With a freshly wrapped body it is almost impossible to tell which is the head and which the feet. We unfortunately never
  • 59. noticed whether the heads were laid toward any particular point of the compass, as has been observed in other localities. Dr. Simpson says that the head is laid to the east at Point Barrow. Various implements belonging to the deceased are broken and laid beside the corpse, and the sled is sometimes broken and laid over it. Sometimes, however, the latter is withdrawn a short distance from the cemetery and left on the tundra for one moon, after which it is brought back to the village. Most people do not seem to be troubled at having the bodies of their relatives disturbed by the dogs or other animals, 581 but we know of one case where the parents of two children who died very nearly at the same time, finding that the dogs were getting at the bodies, raised them on stages of driftwood about 4 or 5 feet high. Similar stages were observed by Hooper at Plover Bay; 582 but this method of disposing of the dead appears to have gone out of use at the present day, since Dall 583 describes the ordinary Siberian method of laying out the dead in ovals of stone as in use at Plover Bay at the time of his visit. The cemetery at Utkiavwĭñ is not confined to the spot I have mentioned, though most of the bodies are exposed there. A few bodies are also exposed on the other side of the lagoon, and one body, that of a man, was laid out at the edge of the higher tundra, about a mile due east from the station. The body was covered with canvas, staked down all round with broken paddles, and over it was laid a flat sledge with one runner broken. 584 At one end of the body lay a wooden dish, and under the edge of the canvas were broken seal-darts and other spears. The body lay in an east and west line, but we could not tell which end was the head. All sorts of objects were scattered round the cemetery—tools, dishes, and even a few guns—though we saw none that appeared to have been serviceable when exposed, except one Snider rifle. If, as is the case among Eskimo in a good many other places, all the personal property of the deceased is supposed to become unclean and must be exposed with him, it is probable that his friends manage to remove the more valuable articles before he is actually dead. 585
  • 60. The method of disposing of the dead varies slightly among the Eskimo in different localities, but the weapons or other implements belonging to the deceased are always laid beside the corpse. The custom at Smith Sound, as described by Bessels, 586 is remarkably like that at Point Barrow. The corpse was wrapped in furs, placed on a sledge, and dragged out and buried in the snow with the face to the west. The sledge was laid over the body and the weapons of the deceased were deposited beside it. Unlike the Point Barrow natives, however, they usually cover the body with stones. In the same passage Dr. Bessels describes a peculiar symbol of mourning, not employed, so far as I can learn, elsewhere. The male mourners plugged up the right nostril with hay and the females the left, and these plugs were worn for several days. The custom of covering the body with stones appears to be universally prevalent east of the Mackenzie region. 587 The bodies seen by Dr. Richardson in the delta of the Mackenzie were wrapped in skins and loosely covered with driftwood, 588 and a similar arrangement was noticed at Kotzebue Sound by Beechey, who figures 589 a sort of little wigwam of driftwood built over the dead man. At Port Clarence Nordenskiöld 590 saw two corpses “laid on the ground, fully clothed, without protection of any coffin, but surrounded by a close fence consisting of a number of tent-poles driven crosswise into the ground. Alongside one of the corpses lay a kayak with oars, a loaded double-barreled gun with locks at half- cock and caps on, various other weapons, clothes, tinder-box, snowshoes, drinking-vessels, two masks, * * * and strangely shaped animal figures.” On the Siberian coast the dead are sometimes burned. 591 Nordenskiöld believes that the coast Chukches have perhaps begun to abandon the custom of burning the dead, but I am rather inclined to think that is a custom of the “deermen,” which the people of the coast of pure or mixed Eskimo blood never fully adopted. Dall, indeed, was explicitly informed that the custom was only used with the bodies of “good” men, and at the time of Nordenskiöld’s visit he
  • 61. found it “at least certain that the people of Pitlekaj exclusively bury their dead by laying them out on the tundra.” The body is surrounded by an oval of stones, but apparently not covered with them as in the east. 592 The Krause brothers observed by the bodies, besides “die erwähnten Geräthschaften” [Lanzen, Bogen und Pfeile für die Männer, Koch- und Hausgeräthe für die Weiber], “unter einen kleinen Steinhaufen ein Hunde-, Renthier-, Bären- oder Walross- Schädel.” This custom shows a curious resemblance to that described by Egede 593 in Greenland: “When little Children die and are buried, they put the Head of a Dog near the Grave, fancying that Children, having no Understanding, they can not by themselves find the Way, but the Dog must guide them to the Land of the Souls.” The body is usually laid out at full length upon the ground. Among the ancient Greenlanders, 594 however, and in the Yukon region the body was doubled up. In the latter region the body was laid on its side in a box of planks four feet long and raised on four supports 595 or wrapped up in mats and covered with rocks or driftwood. 596 The custom of inclosing the dead in a short coffin, to judge from the figures given by the latter writer in P1. vi. of his report, appears also to prevail at the mouth of the Kuskokwim. In the island of Kadiak, according to Dall and Lisiansky, 597 the dead were buried.
  • 62. Two lines near the bottom of the page were reversed in print: GOVERNMENT. In the family.— I can hardly do better than quote Dr. Simpson’s words, already referred to (op. cit. page 252), on this subject: “A man seems to have unlimited authority in his own hut.” Nevertheless, his rule seems to be founded on respect and mutual agreement, rather than on despotic authority. The wife appears to be consulted, as already stated, on all important occasions, and, to quote Dr. Simpson again (ibid.): “Seniority gives precedence when there are several women in one hut, and the sway of the elder in the direction of everything connected with her duties seems never disputed.” When more than one family inhabit the same house the head of each family appears to have authority over his own relatives, while the relations between the two are governed solely by mutual agreement. In the village.— These people have no established form of government nor any chiefs in the ordinary sense of the word, but appear to be ruled by a strong public opinion, combined with a certain amount of respect for the opinions of the elder people, both men and women, and by a large number of traditional observances like those concerning the whale fishery, the deceased, etc., already described. In the ordinary relations of life a person, as a rule, avoids doing anything to his neighbor which he would not wish to have done to himself, and affairs which concern the community as a whole, as for instance
  • 63. their relations with us at the station, are settled by a general and apparently informal discussion, when the opinion of the majority carries the day. The majority appears to have no means, short of individual violence, of enforcing obedience to its decisions, but, as far as we could see, the matter is left to the good sense of the parties concerned. Respect for the opinions of elders is so great that the people may be said to be practically under what is called “simple elder rule.” 598 Public opinion has formulated certain rules in regard to some kinds of property and the division of game, which are remarkably like those noticed among Eskimo elsewhere, and which may be supposed to have grown up among the ancestors of the Eskimo, before their separation. For instance, in Greenland, 599 “Anyone picking up pieces of driftwood or goods lost at sea or on land was considered the rightful owner of them; and to make good his possession he had only to carry them up above high-water mark and put stones upon them, no matter where his homestead might be.” Now, at Point Barrow we often saw the natives dragging driftwood up to the high-water mark, and the owner seemed perfectly able to prove his claim. Lieut. Ray informs me that he has seen men mark such sticks of timber by cutting them with their adzes and that sticks so marked were respected by the other natives. On one occasion, when he was about to have a large piece of drift-timber dragged up to the station, a woman came up and proved that the timber belonged to her by pointing out the freshly cut mark. I have myself seen a native claim a barrel which had been washed ashore, by setting it up on end. As far as we could learn, the smaller animals, as for instance, birds, the smaller seals, reindeer, etc., are the property of the hunter, instead of being divided as in some other localities, for example at Smith Sound, 600. The larger seals and walruses appeared to be divided among the boat’s crew, the owner of the boat apparently keeping the tusks of the walrus and perhaps the skin. A bear, however, both flesh and skin, is equally divided among all who in any way had a hand in the killing. We learned this with certainty from
  • 64. having to purchase the skin of a bear killed at the village, where a number of men had been engaged in the hunt. When a whale is taken, as I have already said, the whalebone is equally divided among the crews of all the boats in sight at the time of killing. All comers, however, have a right to all the flesh, blubber, and blackskin that they can cut off. 601 Dr. Rink, in describing the social order of the ancient Greenlanders, 602 says: “Looking at what has been said regarding the rights of property and the division of the people into certain communities, in connection with the division of property into the classes just given, we are led to the conclusion that the right of any individual to hold more than a certain amount of property was, if not regulated by law, at least jealously watched by the rest of the community, and that virtually the surplus of any individual or community, fixed by the arbitrary rate which tradition or custom had assigned, was made over to those who had less.” At Point Barrow, however, the idea of individual ownership appears to be much more strongly developed. As far as we could learn, there is no limit to the amount of property which an individual, at least the head of a family, may accumulate. Even though the whalebone be, as already described, divided among all the boats’ crews “in at the death,” no objection is made to one man buying it all up, if he has the means, for his own private use. This has given rise to a regular wealthy and aristocratic class, who, however, are not yet sufficiently differentiated from the poorer people to refuse to associate on any terms but those of social equality. The men of this class are the umialiks, a word which appears in many corrupted forms on the coast of Western America and is often supposed to mean “chief.” Dr. Simpson 603 says: “The chief men are called O-mé-liks (wealthy),” but “wealthy” is an explanation of the position of these men, and not a translation of the title, which, as we obtained it, is precisely the same as the Greenland word for owner of a boat, umialik (from umia(k), and the
  • 65. termination lĭk or lĭ-ñ. This is one of the few cases in which the final k is sounded at Point Barrow as in Greenland). Dr. Rink has already observed 604 that the word used by Simpson “no doubt must be the same as the Greenlandish umialik, signifying owner of a boat,” and as I heard the title more than once carefully pronounced at Point Barrow it was the identical word. The umialiks, as Simpson says, 605 “have acquired their position by being more thrifty and intelligent, better traders, and usually better hunters, as well as physically stronger and more daring.” 606 They have acquired a certain amount of influence and respect from these reasons, as well as from their wealth, which enables them to purchase the services of others to man their boats, but appear to have absolutely no authority outside of their own families. 607 Petroff 608 considers them as a sort of “middlemen or spokesmen,” who make themselves “prominent by superintending all intercourse and traffic with visitors.” This sort of prominence, however, appears to have been conferred upon them by the traders, who, ignorant of the very democratic state of Eskimo society, naturally look for “chiefs” to deal with. They pick out the best looking and best dressed man in the village and endeavor to win his favor by giving him presents, receiving him into the cabin, and conducting all their dealings with the natives through him. The chief, thus selected, is generally shrewd enough to make the most of the greatness thrust upon him, and no doubt often pretends to more influence and power than he actually possesses. 609 As to the story of the whalemen, that the “chieftainship” is the reward of the best fighter, who holds it like a “challenge cup,” subject to being called out at any time to defend his rank in a duel, as far as concerns Point Barrow, this is a sheer fable, perhaps invented by the Eskimo to impose upon the strangers, but more likely the result of misunderstanding and a vivid imagination on the part of the whites. Among umialiks, one or two appear to have more wealth and influence than the rest. Tcuñaura in Utkiavwĭñ and the
  • 66. late Katiga at Nuwŭk were said, according to Captain Herendeen, to be “great umialiks” and Tcuñaura was always spoken of as the foremost man in Utkiavwĭñ. We knew of one party coming up from Sidaru with presents for Tcuñaura, and were informed that the other Eskimo never sold to him, but only gave him presents. It was also said that Katiga’s infant son would one day be a “great umialik.” All these men are or have been captains of whaling umiaks, and the title umialiks appears to be applied to them in this capacity, since many of the poorer men, who, as far as we could learn, were not considered umialiks, own umiaks which they do not fit out for whaling, but use only to transport their families from place to place in the summer. RELIGION. General ideas.— It was exceedingly difficult to get any idea of the religious belief of the people, partly from our inability to make ourselves understood in regard to abstract ideas and partly from ignorance on our part of the proper method of conducting such inquiries. For instance, in trying to get at their ideas of a future life, we could only ask “Where does a man go when he dies?” to which we, of course, received the obvious answer, “To the cemetery!” Moreover, such a multitude of other and easier lines of investigation presented themselves for our attention that we were naturally inclined to neglect the difficult field of religion, and besides under the circumstances of our intercourse it was almost impossible to get the attention of the natives when their minds were not full of other subjects. Nevertheless, many of the fragments of superstition and tradition that we were able to collect agree remarkably with what has been observed among the Eskimo elsewhere, so that it is highly probable that their religion is of the same general character as that of the Greenlanders, namely, a belief in a multitude of supernatural beings,
  • 67. who are to be exorcised or propitiated by various observances, especially by the performances of certain specially gifted people, who are something of the nature of wizards. So much has been written by many authors about these wizards or “doctors,” the angekut of the eastern Eskimo, the so-called “shamans” of Alaska and Siberia, that I need make no special reference to their writings except where they happen to throw light on our own observations. Dr. Simpson succeeded in obtaining more information concerning the religious belief of these people than our party was able to do, and his observations, 610 to which ours are in some degree supplementary, tend to corroborate the conclusion at which I have arrived. Our information in regard to the special class of wizards was rather vague. We learned that many men in the village, distinguishable from the rest by no visible characteristics, were able to heal the sick, procure good weather, favorable winds, plenty of game, and do other things by “talking” and beating the drum. We did not learn the number of these men in either village, but we heard of very many different men doing one or the other of these things, while others of our acquaintance never attempted them. Neither did we learn that any one of these men was considered superior to the rest, as appears to be the case in some regions, nor how a man could attain this power. Some of these men, who appeared to give particular attention to curing the sick, called themselves “tû´ktĕ” (“doctor”), but, probably for want of properly directed inquiries, we did not learn the Eskimo name of these people. We were definitely informed, however, that their “talk,” when treating disease or trying to obtain fair weather, etc., was addressed to “tu´ɐña,” or a supernatural being. This name, of course, differs only in dialectic form from that applied in other places to the universal familiar spirits of Eskimo superstition. We at first supposed that “tuɐña” meant some particular individual demon, but Dr. Simpson is probably right in saying that the Point Barrow natives, like the rest of the Eskimo, recognize a host of tuɐñain, since “tuɐña” was described to us under a variety of forms.
  • 68. Most of the natives whom we asked if they had seen tuɐña, said that they had not, but that other men, mentioning certain “doctors,” had seen him. One man, however, said that he had seen tuɐña in the kûdyĭgĭ, when the people “talked” sitting in the dark, with their heads bowed and faces covered, and tuɐña came with a noise like a great bird. 611 He had raised his head and saw tuɐña, like a man with bloodless cheeks. 612 Tuɐña again was called “a bad man, dead” (apparently a ghost), sometimes as large as a man and sometimes dwarfish, sometimes a fleshless skeleton, while one man, to describe him, made the same grimace that a white man would use to indicate a hobgoblin, with staring eyes, gaping mouth, and hands outstretched like claws. Apparently “tuɐña” in conversation with us was used to designate all the various supernatural objects of their belief, ghosts as well as familiar spirits. For instance, in Greenland, according to Rink, 613 a ghost “manifests himself by whistling or singing in the ears.” Now, Lieut. Ray was walking rapidly one day in the winter with an Eskimo and his wife, and the woman suddenly stopped and said she “heard tuɐña”—that he made a noise like singing in the ears. The people generally have a great dread of “tuɐña,” who they say would kill them, and are very averse to going out alone in the dark. One of each party that came over from the village in the evening usually carried a drawn knife, preferably one of the large double- edged knives, supposed to be Siberian and already described, in his hand as defense against tuɐña, and a drawn knife was sometimes even carried in the daylight “nanumunlu tuɐñamunlu,” “for bear and demon.” Notwithstanding their apparently genuine dread of “tuɐña,” they are by no means averse to talking or even joking about him. The knife also serves as a protection against the aurora, which most of them agree is bad, and when bright likely to kill a person by striking him in the back of the neck. However, brandishing the knife at it will keep it off. Besides, as a woman told me one night, you can drive off a “bad” aurora by throwing at it dog’s excrement and urine. 614
  • 69. Lieut. Ray saw in one of the houses in Utkaiwiñ, a contrivance for frightening away a “tuɐña” from the entrance to a house should he try to get in. The man had hung in the trapdoor the handle of a seal- drag by means of a thong spiked to the wall with a large knife, and told Lieut. Ray that if “tuɐña” tried to get into the house he would undoubtedly catch hold of the handle to help himself up, which would pull down the knife upon his head and frighten him off. We never had an opportunity of witnessing the ceremony of summoning “tuɐña,” nor did we ever hear of the ceremony taking place during our stay at the station, but we were fortunate enough to observe several other performances, though they do not appear to be frequent. The ceremony of healing the sick and the ceremonies connected with the whale-fishery have already been described. On the 21st of February, 1883, Lieut. Ray and Capt. Herendeen happened to be at the village on time to see the tuɐña, who had been causing the bad weather, expelled from the village. Some of the natives said the next day that they had killed the tuɐña, but they said at the same time he had gone “a long way off.” When Lieut. Ray reached the village, women were standing at the doors of the houses armed with snow-knives and clubs with which they made passes over the entrance when the people inside called out. He entered one house and found a woman vigorously driving the tuɐña out of every corner with a knife. They then repaired to the kûdyĭgi, where there were ten or twelve people, each of whom, to quote from Lieut. Ray’s note book, “made a charge against the evil spirit, telling what injuries they had received from it.” Then they went into the open air, where a fire had been built in front of the entrance, and formed a half circle around the fire. Each then went up and made a speech, bending over the fire (according to Simpson, who describes a similar ceremony at Nuwŭk on p. 274 of his paper, coaxing the tuɐña to come under the fire to warm himself). Then they brought out a large tub full of urine, to which, Simpson says, each man present had contributed, and held it ready near the fire, while two men stood with their rifles in readiness, and a boy stood near the fire with a large stone in his hands, bracing himself firmly
  • 70. with his feet spread apart for a vigorous throw. Then they chanted as follows (the words of this chant were obtained afterward by the writer): Tâk tâk tâk tohâ! Nìju´a hâ! He! he! he! Haiyahe! Yaiyahe! Hwi! And instantly the contents of the tub were dashed on the fire, the stone thrown into the embers, and both men discharged their rifles, one into the embers, and one into the cloud of steam as it rose. Then all brushed their clothes violently and shouted, and the tuɐña was killed. By a fortunate coincidence, the next day was the finest we had had for a long time. Sacrifices are also occasionally made to these supernatural beings as in Greenland “gifts were offered to the inue of certain rocks, capes and ice firths, principally when traveling and passing those places.” 615 Capt. Herendeen, in the fall of 1882, went to the rivers in company with one of the “doctors.” When they arrived at the river Kuaru, where the latter intended to stay for the fishing, he got out his drum and “talked” for a long time, and breaking off very small pieces of tobacco threw them into the air, crying out, “Tuɐña, tuɐña, I give you tobacco! give me plenty of fish.” When they passed the dead men at the cemetery, he gave them tobacco in the same way, asking them also for fish. 616 We noticed but few other superstitious observances which have not been already described. As in Greenland and elsewhere, superstition requires certain persons to abstain from certain kinds of food. For instance, Mûñialu, and apparently many others, were not permitted to eat the burbot, another man was denied ptarmigan, and a woman 617 at Nuwŭk was not allowed to
  • 71. eat “earth food,” that is, anything which grew upon the ground. Lieut. Ray also mentions a man who was forbidden bear’s flesh. 618 We observed some traces of the superstition concerning the heads of seals and other marine animals taken in the chase, which has been noticed elsewhere. Crantz says: 619 “The heads of seals must not be fractured, nor must they be thrown into the sea, but be piled in a heap before the door, 620 that the souls of the seals may not be enraged and scare their brethren from the coast.” And Capt. Parry found that at Winter Island they carefully preserved the heads of all the animals killed during the winter, except two or three of the walrus which he obtained with great difficulty. The natives told him that they were to be thrown into the sea in the summer, but at Iglulik they readily sold them before the summer arrived. 621 I tried very hard to get a full series of skulls from the seals taken at Utkiavwĭñ in the winter of 1882-’83, but though I frequently asked the natives to bring them over for sale, they never did so, till at last one young woman promised to bring me all I wanted at the price of half a pound of gunpowder a skull. Nevertheless, she brought over only two or three at that price. We did not observe what was done with the skulls, but frequently observed quantities of the smaller bones of the seals carefully tucked away in the crevices of the ice at some distance from the shore. We had comparatively little difficulty in obtaining skulls of the walrus, but I observed that the bottom of Tûseráru, the little pond at the edge of the village, was covered with old walrus skulls, as if they had been deposited there for years. The superstition appears to be in full force among the Chukches, who live near the place where the Vega wintered. Nordenskiöld was unable to purchase a pair of fresh walrus heads at the first village he visited, though the tusks were offered for sale the next day 622 and at Pitlekaj. 623 “Some prejudice * * * prevented the Chukches from parting with the heads of the seal, though * * * we offered a high price for them. ‘Irgatti’ (to-morrow) was the usual answer. But the promise was never kept.”
  • 72. Fig. 421.—Whale flaked from glass. Amulets.— Like the Greenlanders 624 and other Eskimos, they place great reliance on amulets or talismans, which are carried on the person, in the boat, or even inserted in weapons, each apparently with some specific purpose, which indeed we learned in the case of some of those in the collection. Like the amulets of the Greenlanders, they appear to be 625 “certain animals or things which had belonged to or been in contact with certain persons (e.g., the people of ancient times, or fortunate hunters) or supernatural beings,” and “objects which merely by their appearance recalled the effect expected from the amulet, such as figures of various objects.” To the latter class belong the rudely flaked flint images of whales, already mentioned, and probably many of the other small images of men and animals already described, especially those fitted with holes for strings to hang them up by. The flint whale is a very common amulet, intended, as we understood, to give good luck, in whaling, and is worn habitually by many of the men and boys under the clothes, suspended around the neck by a string. The captain and harpooner of a whaling crew also wear them as pendants on the fillets already described, and on the breast of the jacket. We obtained five of these objects, all of very nearly the same shape, but of different materials and varying somewhat in size. Fig. 421 represents one of these (No. 56703 [208] from Utkiavwĭñ) made of a piece of hard colorless glass, probably a fragment of a ship’s “deadlight.” It is rather roughly flaked into a figure of a “bowhead” whale, 3.4 inches long, as seen from above and very much flattened with exaggerated flukes. The flippers were rudely indicated in the outline, but the left one is broken off.
  • 73. Fig. 422.—Whale flaked from red jasper. No. 89613 [771] from Utkiavwĭñ is a very similar image, 2.4 inches long, which perhaps is of the same material, though it may be made of rock crystal. No. 56707 [159] from Utkiavwĭñ is a very small whale (1.4 inches long), chipped in large flakes out of a water-worn pebble of smoky quartz, while No. 89577 [939] Fig. 422, from the same village, which is a trifle larger (2 inches long), is made of dark crimson jasper. The large black flint whale, No. 56683 [61], also from Utkiavwĭñ, which is 3.9 inches long, is the rudest of all the figures of the whales. It is precisely the shape of the blade of a skin scraper, except for the roughly indicated flukes. Fig. 423 (No. 89524 [1299] from Utkiavwĭñ) is a rude wooden image of the same animal, 3½ inches long, very broad and flat- bellied. It is smoothly carved and has a fragment of sky-blue glass inlaid to represent the left eye and a bit of iron pyrites for the right. The flukes have been split wholly off and fastened on with a lashing of narrow whalebone passing through a vertical hole in the “small” and round the edge of the flukes. The flukes themselves have been split across and appear to have been doweled together. This shows that the owner attached considerable value to the object, or he would not have taken the trouble to mend it when another could have been so easily whittled out. In the middle of the belly is an oblong cavity, containing something which probably adds greater power to the charm. What this is can not be seen, as a band of sealskin with the hair shaved off has been shrunk on round the hinder half of the body and secured by a seam on the right side. A double turn of sinew braid is knotted round the middle of the body, leaving two ends which are tied together in a loop, showing that this object was meant to be attached somewhere about the person.
  • 74. Fig. 423.—Ancient whale amulet, of wood. To this class also probably belong the skins or pieces of animals worn as amulets, probably with a view of obtaining the powers of the particular animal, as in so many cases in the stories related in Rink’s Tales and Traditions. We frequently saw men wearing at the belt bunches of the claws of the bear or wolverine, or the metacarpal bones of the wolf. 626 The head or beak of the gull or raven 627 is also a common personal amulet, and one man wore a small dried flounder. 628 We collected a number of these animal amulets to be worn on the person, but only succeeded in learning the special purpose of one of them, No. 89532 [1307], from Utkiavwĭñ, which was said to be intended to give good luck in deer hunting. It is a young unbranched antler of a reindeer, 6 inches long, and apparently separated from the skull at the “bur,” with the “velvet” skin still adhering, though most of the hair is worn off except at the tip. A bit of sinew is tied round the base. No. 89522 [1573], from Utkiavwĭñ, is an amulet consisting of the last three joints of the foot of a reindeer fawn, with the skin and hoof and about 1½ inches of tendon attached behind, through a hole in the end of which is knotted about 3 inches of seal thong. No. 89525 [1314] from the same village, is a precisely similar charm. No. 89699 [779] from Utkiavwĭñ , is the subfossil incisor tooth of some ruminant with a hole drilled through the root for a string to hang it up by. It was said to be the tooth of the “ug’ru´nû,” a large animal, long extinct. As the natives said, “Here on the land are none, only
  • 75. the bones remain.” No. 89743 [1110], from Utkiavwĭñ, is a molar tooth of the same animal, probably, weathered and old, with a hole freshly drilled through one root and a long piece of sinew braid with the ends knotted together looped into it. There are also in the collection two very old teeth which probably were inclosed in little sacks of skin and worn as amulets. No. 89698 [1580], from Utkiavwĭñ, is the tusk of a very young walrus, only 2½ inches long, and No. 89452 [1148] from Utkiavwĭñ, is the canine tooth of a polar bear. No. 56547 [656], from the same village, is a similar tooth. 629 The only amulet attached to a weapon, which we collected, is the tern’s bill, already alluded to, placed under the whalebone lashing on the seal-spear, No. 89910 [1694]. Perhaps the idea of this charm is that the spear should plunge down upon the seal with as sure an aim as the tern does upon its prey. 630 A number of amulets of this class are always carried in the whaling- umiak. I have already mentioned the wolf-skulls, stuffed ravens and eagles, fox-tails 631 and bunches of feathers used for this purpose. Most of these charms are parts of some rapacious animal or bird, but parts of other animals seem to have some virtue on these occasions. For instance, I noticed the axis vertebra of a seal in one whaling- umiak, and we collected a rudely stuffed skin of a godwit (Limosa lapponica baueri), which, we were informed, was “for whales.” This specimen (No. 89526 [1328], Fig. 424, from Utkiavwĭñ) is soiled and ragged, and has a stick thrust through the neck to hold it out. The neck is wrapped around with a narrow strip of whalebone and some coarse thread, part of which serves to lash on a slip of wood, apparently to splice the stick inside. A bit of white man’s string is passed around the body and tied in a loop to hang it up by. This charm is perhaps to keep the boat from capsizing, since Crantz says that the Greenlanders “like to fasten to their kajak a model of it * * * or only a dead sparrow or snipe, or a bit of wood, stone, some feathers or hair, that they may not overset” (vol. 1, p. 216), and
  • 76. perhaps the bone of a marine animal, like the seal, is to protect the crew from drowning should the boat upset, after all. No. 89529 [1150] from Utkiavwĭñ is a bunch of feathers to be carried in the boat. It consists of nine wing feathers of the golden eagle, four tied in a bunch with a bit of sinew round the quills, four tied up with one end of the short bit of seal thong which serves to tie the whole bundle together, one of which has all the light-colored parts of the feather stained with red ocher, and a single feather shaft carefully wrapped up in a piece of entrail and wound spirally with a piece of sinew braid. Fig. 424.—Amulet of whaling; stuffed godwit. No. 89527-8 [1327] from Utkiavwĭñ is the charm which will secure good success in deerhunting if it is hung up outside of the snow house in which the family is encamped. It consists of two roughly stuffed skins of the black bellied plover (Charadrius squatarola), each with a stick run through the body so that one end supports the neck and the other the tail, and the necks wound with sinew. One has no head. A string of sinew braid is tied around the body of each, so as to leave a free end at the back, to which is fastened a little cross piece of bone, by which it may be secured to a becket. Like the rest of the amulets in the collection this has evidently seen service, being very old, worn, and faded. The other class of amulets, namely objects which have belonged to or been in contact with certain persons or supernatural beings, or I may add apparently certain localities, is represented by a number of specimens. To the custom of using such things as amulets, we
  • 77. Fig. 425.—Amulet consisting of ancient jade adz. Fig. 426.—Little box containing amulet for whaling. undoubtedly owe the preservation of most of the ancient weapons and other implements, especially those made of wood, bone, or other perishable substances, like the ancient harpoon heads already described, one of which, No. 89544 [1419], is still attached to the belt on which it was worn. Fig. 425, No. 56668 [308], from Utkiavwĭñ is one of the ancient black jade adzes 5.1 inches, slung with thong and whalebone, making a becket by which it can be hung up. We did not learn the history of this amulet, which at the time of collecting it was supposed to be a net sinker. There would, however, be no reason for using so valuable an object for such a purpose, when a common beach pebble would do just as well, unless it was intended as a charm to insure success in fishing. It may even have been carried as a charm on the person, since we afterwards saw a still more bulky object used for such a purpose. Such an object seems rather heavy to be carried on the person, but a well known man in Utkiavwĭñ always carried with him when he went sealing a large pear-shaped stone, which must have weighed upwards of two pounds, suspended somewhere about his person. It is not unlikely that this stone acquired its virtue as an amulet from having been a sinker used by some lucky fisherman in former time or in a distant country. Mr. H. W. Henshaw has already referred to the resemblance of this amulet to the plummet-like “medicine stones” of some of our Indians. 632 Fig. 426, (No. 89534 [1306] from Utkiavwĭñ) is an amulet for success in whaling. It consists of three little irregular water-worn fragments of amber carefully wrapped in a bit of parchment and inclosed in a little wooden box 1½ inches long, made of two semicylindrical bits of cottonwood, with the flat faces hollowed out and put together and fastened up by three turns of
  • 78. Fig. 427.— Amulet for catching fowl with bolas. sinew braid round the middle, tied in a loose knot. The box is old and brown from age and handling. We heard of other pieces of amber and earth (“nuna”) worn as amulets, wrapped up in bits of leather and hung on the belt. No. 89533 [1247], from Utkiavwĭñ, is simply a nearly square pebble, 1.4 inches long, of dark red jasper, slung in a bit of sinew braid so that it can be hung on the belt. Fig. 427 (No. 89525 [1308] from Utkiavwĭñ) is some small object, placed in the center of the grain side of a square bit of white sealskin, the edges of which are folded up around it and tied tightly round with deer sinew, so as to make a little round knob. I collected this amulet, and was particularly informed how it was to be used. If it be fastened on the right shoulder it will insure success in taking ducks with the “bolas.” Fig. 428 (No. 89535 [1244] from Utkiavwĭñ) is an amulet whose history we did not learn. It is a little oblong box 3.3 inches long, carved from a block of cottonwood, with a flat cover tied on with nine turns of sinew braid, and contains twenty-one dried humble-bees, which it was said came from the river Kulugrua. The natives have a great dread, apparently superstitious, of these bees and the large gadflies (Œstens tarandi), one of which I have seen scatter half a dozen people. A man one day caught one of these, and whittled out a little box of wood, in which he shut the insect up and tied it up with a shred of sinew, telling Capt. Herendeen that it was “tuɐñamun,” for “tuɐña.” Fig. 428.—Box of dried bees—amulet.
  • 79. A small lump of indurated gravel (No. 56725) [273] was one day brought over from Utkiavwĭñ, with the story that it was a “medicine” for driving away the ice. The man who uses this charm stands on the high bank at the village, and breaking off grains of the gravel throws them seaward. This will cause the ice to move off from the shore. The essential identity of the amulets of the Point Barrow natives with those used by the Eskimo elsewhere is shown by the following passages from other writers. Egede says: 633 A Superstition very common among them is to load themselves with Amulets or Pomanders, dangling about their Necks and Arms, which consist in some Pieces of old Wood, Stones or Bones, Bills and Claws of Birds, or Anything else which their Fancy suggests to them. Crantz says: 634 They are so different in the amulets or charms they hang on people, that one laughs at another’s. These powerful preventives consist in a bit of old wood hung around their necks, or a stone, or a bone, or a beak or claw of a bird, or else a leather strap tied round their forehead, breast, or arm. Parry speaks 635 of what he supposes were amulets at Iglulik, consisting of teeth of the fox, wolf, and musk-ox, bones of the “kablĕĕarioo” (supposed to be the wolverine), and foxes’ noses. Kumlien says 636 that at Cumberland Gulf, “among the many superstitious notions, the wearing of charms about the person is one of the most curious. These are called angoouk or amusit, and may be nothing but pieces of bone or wood, birds’ bills or claws, or an animal’s teeth or skin.” A little girl “had a small envelope of sealskin that was worn on the back of her inside jacket” containing two small stones. Such little pockets of skin sewed to the inner jacket are very common at Point Barrow, but we did not succeed in any case in learning their contents. At Kotzebue Sound, Beechey saw ravens’ skins on which the natives set a high value, while the beaks and claws of these birds were attached to their belts and headbands. 637 Petitot describes 638 the amulets used in the Mackenzie district, in
  • 80. the passage already quoted, as “défroques empaillées de corbeau, de faucon ou d’hermine.” It is not likely that the use of these is confined to the women, as his words, “Elles y portent,” would seem to imply. Among the sedentary Chukches of Siberia amulets were seen consisting of wooden forks and wood or ivory carvings. 639 A wolf’s skull, hung up by a thong; the skin, together with the whole cartilaginous portion of a wolf’s nose, and a flat stone, are also mentioned. 640 Capt. Holm also found wonderfully similar customs among the East Greenlanders. He says, 641 “bære alle Folk Amuletter af de mest forskjelligartede Ting” to guard against sickness and to insure long life, and also for specific purposes. The men wear them slung round the neck or tied round the upper arm, the women in their knot of hair or “i Snippen foran paa Pelsen.” Footnotes 411-641 411. Nordenskiöld, Vega, vol. 2, pp. 122, and Fig. 1, p. 117. 412. Crantz describes the process of preparing boat covers as follows: “The boat skins are selected out of the stoutest seals’ hides, from which the fat is not quite taken off; they roll them up, and sit on them, or let them lie in the sun covered with grass several weeks, ’till the hair will come off.” History of Greenland, vol. 1, p. 167. 413. Gilder describes a similar process of manufacturing these lines at Hudson’s Bay. (Schwatka’s Search, p. 176.) 414. W. J. Sollas, in Jour. Anthrop. Inst. of Great Britain and Ireland, vol. 9, pp. 329-336. 415. Nordenskiöld’s figures, Vega, vol. 2, p. 123. 416. Parry’s Second Voy., pl. opposite p. 548, Fig. 5. 417. Vega, vol. 1, p. 493.
  • 81. Welcome to our website – the perfect destination for book lovers and knowledge seekers. We believe that every book holds a new world, offering opportunities for learning, discovery, and personal growth. That’s why we are dedicated to bringing you a diverse collection of books, ranging from classic literature and specialized publications to self-development guides and children's books. More than just a book-buying platform, we strive to be a bridge connecting you with timeless cultural and intellectual values. With an elegant, user-friendly interface and a smart search system, you can quickly find the books that best suit your interests. Additionally, our special promotions and home delivery services help you save time and fully enjoy the joy of reading. Join us on a journey of knowledge exploration, passion nurturing, and personal growth every day! ebookbell.com