![INTRODUCTION 9
paradox seemed worth pointing out for the implications it has for the understanding
the mechanisms involved in these two types of change (see this volume, chapter 2).
Admittedly, at the time I did not realize the potential for both of these observations
to totally change the way I looked at phonology and grammar.
When I undertook a cross-linguistic study of morphology, the importance of
frequency was observable both within and across paradigms in all the languages I
investigated, leading to a novel conception of morphological structure (see Bybee
1985 and in this volume part II). A view of morphological structure as affected by
language use kept producing more and more interesting hypotheses and findings.
Similarly, the investigation of frequency effects in phonology has led to many inter-
esting questions and solutions to some old problems (see Bybee 2001b and in this
volume part III). I feel the issues in morphosyntax may be more complex, but many
interesting findings have already arisen in this usage-based conception of grammar
(see part IV). This volume reports on these investigations, which together paint a
picture of language as having a variegated and mutable shape. While there are many
general patterns of great interest, the topology of the cognitive organization of lan-
guage is neither flat, regular, nor permanent. High-frequency words and phrases grow
strong with repetition and loom large, forming looser connections with other items,
while low-frequency words and expressions are less prominent but gain stability by
conforming to patterns used by other items. General patterns dominate networks where
more specific patterns can be overpowered unless represented by high-frequency
items. Words that have phonological similarities cluster together; constructions are
connected if they have properties in common. Instances of constructions that grow
to high frequency slowly disengage from the more general pattern to become inde-
pendent constructions. Thus the phonetic and semantic substance of language is ever
being shaped by the effects of usage.
In the next section I provide an overview of the frequency effects that I have
investigated. At the beginning of each part of the book, I have provided a few para-
graphs that put the ideas in these articles in perspective with respect to the environ-
ment in which they were written, as well as subsequent research.
2. What are frequency effects?
There are various ways of counting frequency, various units upon which to base the
count, and various ways in which frequency has cognitive effects. First, I distinguish
token and type frequency. Token frequency counts the number of times a unit ap-
pears in running text. Any specific unit, such as particular consonant [s], a syllable
[ba], a word dog or the, a phrase take a break, or even a sentence such as Your toast
popped up can have a token frequency. Type frequency is a very different sort of
count. Only patterns of language have type frequency, because this refers to how
many distinct items are represented by the pattern. Type frequency may apply to
phonotactic sequences; it would be the count of how many words of the language
begin with [sp] versus how many begin with [sf]. It may apply to morphological
patterns, such as stem + affix combinations. For instance, the English past tense pat-
tern exemplified by know, knew; blow, blew has a lower type frequency than the regu-](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-22-320.jpg)
![10 BACKGROUND AND CURRENT CONTEXT
lar pattern of adding the –ed suffix. Syntactic patterns or constructions also have type
frequencies: the ditransitive pattern in English, exemplified by He gave me the change,
is used with only a small set of verbs, while the alternate pattern He gave the change
to me is possible with a large class of verbs (Goldberg 1995). In this section I treat
the three known effects of token frequency and the effect of type frequency, as well
as their interaction.
2.1. Token frequency: The Conserving Effect
The first effect of token frequency to be discussed, which we can call the Conserv-
ing Effect, depends upon the fact that repetition strengthens memory representations
for linguistic forms and makes them more accessible. Accessibility in this sense re-
fers to the fact that in experiments where subjects are asked to say whether a string
of letters or sounds is a word of their language, they respond much more quickly to
high-frequency words than to low-frequency words. This greater accessibility sug-
gests that each token of use strengthens the memory representation for a word or
phrase (Bybee 1985; this volume, chapters 2, 3, 4, 5, and 8). The strength of repre-
sentation of higher frequency forms explains why they resist reformation on the basis
of analogy with other forms. For instance, for English irregular verbs, there is a gen-
eral trend diachronically toward regularization, a trend also witnessed in child lan-
guage development. However, the higher frequency verbs resist this trend; thus keep,
sleep, weep, leap, and creep and other verbs of this shape acquired irregular past forms
when the vowel was shortened in early Middle English, giving kept, slept, wept, leapt,
and crept. Only the lower frequency verbs of this class have subsequently developed
regularized pasts weeped, leaped, and creeped (still used alongside the irregulars).
The mechanism behind this type of change (analogical reformation) is that a new
past form is created by accessing the base/present form and adding the suffix –ed (in
this case its allomorph [t]) to it. For those verbs of high frequency, the greater acces-
sibility of the irregular past makes such a reformation unlikely. For this reason, the
lower frequency paradigms tend to regularize before the higher frequency paradigms.
Another aspect of the Conserving Effect is the fact that within a paradigm it is
the higher frequency form that serves as the basis for the reformation just illustrated
(Ma¤czak 1980; Bybee 1985; this volume, chapters 3 and 4). Thus in our example, the
base/present form weep, which is more frequent than wept, survives and serves as the
base upon which the new past is created, in contrast to the hypothetical situation in
which a new base is formed from wept, creating *wep as the new base. The latter result
would also be a regular paradigm, but this type of change is rare (see Tiersma 1982).
Thus Ma¤czak (1980) has written that the most frequent form(s) of a paradigm are
the most likely to resist change and to serve as the basis for change in other forms.
The Conserving Effect is also observable in morphosyntactic constructions (this
volume, chapters 12 and 16). For instance, of the two forms of negation exemplified
by I know nothing about it (negative incorporation) and I don’t know anything about
it (not negation), the former is older and more conservative. Tottie (1991) shows that
where the two constructions are interchangeable, the older construction (negative
incorporation) is used in certain high-frequency constructions, such as existential
constructions, constructions with possessive have and copular be. That frequency is](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-23-320.jpg)
![12 BACKGROUND AND CURRENT CONTEXT
As mentioned before, the lexical regularity of sound change is the cornerstone
of the comparative method. It was argued for vociferously by the Neogrammarians
of the late nineteenth century. Over the years, there have been many detractors from
this strict position. Contemporaneously with the Neogrammarians, Schuchardt (1885;
see Vennemann and Wilbur 1972) documented his observations that sound changes
in progress saw frequent words scurrying ahead, while the less frequent ones dragged
behind. Many philologists maintained open minds about regularity; the famous his-
torian of the Spanish language Ramón Menéndez-Pidal (1950) observed that in sound
change words were like leaves floating on running water: some were caught by the
current and rushed ahead, while others circled in eddies and lagged behind, yet the
general flow was unmistakably directional. In my first investigation of frequency
effects, I attempted to document the role of high-token frequency in sound change
(see chapter 2 of this collection). Taking off from this point, Phillips (1984, 2001,
and other works) has studied many sound changes of various types in the light of
frequency of use. The effect of frequency can be seen in many cases of more or less
stable variation, as well as in changes that end up affecting all the words of a lan-
guage, and turn out regular. In the latter case, the course of development of the change
involves the gradual spread through the lexicon known as lexical diffusion.
The idea that high-frequency words grow shorter brings to mind the work of
George Kingsley Zipf, which, however, proposes a very different explanation for
the correlation of high frequency with shorter expression. In The Psychobiology of
Language, published in 1935, Zipf documents across a number of languages the ten-
dency for high-frequency words to be shorter than lower frequency words. Zipf takes
the frequency distribution of short and long words to represent an equilibrium that
languages strive to maintain. The mechanism he sees as responsible for this equilib-
rium is the process of “clipping” or abbreviation by which laboratory becomes lab,
and so on. Ma¤czak’s theory of irregular sound change due to frequency is similar to
Zipf’s:
There is a synchronic law according to which the linguistic elements which are more
often used are smaller than those which are less often used . . . If a word or mor-
pheme becomes too short in relation to its frequency, it is replaced by a longer one.
But if a linguistic element (i.e., a morpheme, word or group of words) becomes too
long in relation to its frequency, it must be shortened, and then there are two pos-
sibilities: either mechanical shortening (Universität > Uni) or an irregular sound
change due to frequency. (1978: 309–310)
My own view of Zipf’s finding, as represented in the chapters of this volume,
is that high-frequency words undergo reductive changes at a faster rate than low-
frequency words. While “clipping” certainly occurs and produces this result as well,
the major mechanism is gradual phonetic reduction brought about by the reduction
and overlapping of articulatory gestures. In phrases of extreme high frequency this
reduction may appear “irregular,” as Ma¤czak has observed, but my view is that such
reduction follows the regular phonetic patterns of the language but carried to a more
extreme point than in words of lesser frequency. Thus when I don’t know reduces,
the [d] undergoes flapping and even deletion, the nasal consonant deletes, and the
[o] vowel may be reduced to schwa. All of these phonetic changes affect other words](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-25-320.jpg)
![INTRODUCTION 13
of English as well. They are not “irregular”; they are just extreme instances of other-
wise regular phonetic processes.
A related effect of the repetition of sequences of units is chunking, or the repre-
sentation of the sequence as a single unit at a higher level (Haiman 1994; Boyland
1997; this volume, chapters 14 and 15). As I don’t know is repeated frequently, it
ceases to be produced by the concatenation of three items (as don’t has long ago
coalesced) and comes to be accessed as a single unit (this volume, chapter 14). This
packaging facilitates both access and further phonetic reduction in the sequence.
Chunking is particularly obvious in high-frequency word sequences, but Jurafsky et al.
(2001) and other studies (Krug 1998; Brown 2004; Alba 2005) have shown that
phonetic reduction between words is highly dependent upon how frequently the words
co-occur, suggesting the probabilistic relations among words at all levels of frequency.
Phonetic reduction of high-frequency combinations of words is also one of the
processes of change that make up the phenomenon of grammaticization (Bybee,
Perkins, and Pagliuca 1994; this volume, chapter 16). As the words or phrases that
enter into the grammaticization process increase in frequency, they also undergo
extreme versions of phonetic reduction, as, for instance, when be going to reduces to
[gənə] or I’m gonna reduces to [aimənə]. Other concomitant processes that lead to
this reduction are the loss of stress on grammaticizing items and their increasing use
in redundant contexts.
The Reducing Effect and the Conserving Effect at first blush seem contradictory
(see this volume, chapter 2) since in one case high frequency encourages change and
in the other high frequency discourages change. The contradiction is only apparent,
however, because very different types of change are involved. On the one hand, the
strengthening of memories makes complex units resist change by reformation on anal-
ogy with productive patterns, while on the other hand, the greater fluency and reduc-
tion of repeated units is a factor in phonetic change and semantic change. The two types
of change involved here are very different. This is the point of chapter 2. (See also
chapter 16 for these effects in grammaticization of morphosyntax.)
2.3. Token frequency: Autonomy
In previous work, I have identified two effects of token frequency, the two just dis-
cussed. However, I think it is worthwhile to take “autonomy” to be a third effect of
token frequency even though it may be also considered an extreme instance of the
Conserving Effect. In Bybee and Brewer 1980 (this volume, chapter 3) we introduce
the notion of autonomy, as it applies to words (although it can also be extended to
phrases), by saying it is “the extent to which a word is likely to be represented in the
speaker’s lexicon as a whole and separate unit” (see page 50 of this volume). Au-
tonomy is thus probabilistic, varying for each word or phrase. Three factors are said
to influence the degree of autonomy: semantic simplicity, morphophonemic regu-
larity, and word frequency. Autonomy is based on MacWhinney’s (1978) notion of
lexical strength, which is a reflection of frequency in experience and discussed more
thoroughly in Bybee (1985). Highly autonomous words have weaker connections to
other related words—either words of the same paradigm or words of the same lexi-
cal class. The idea behind autonomy is that when words (and phrases) are highly](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-26-320.jpg)
![INTRODUCTION 15
be taken into account: often the member items that occur with a construction must
also belong to certain phonological or semantic categories. The verbs of the string,
strung class must end in a nasal or a velar (chapters 5 and 6); the adjectives that can
be used in the construction [X drives me (or someone) ADJ], (as in it drives me mad,
it drives me crazy) must suggest some degree of insanity, either literally or figura-
tively (Boas 2003). As pointed out in the previous section, not all members contrib-
ute equally to productivity: extreme high-frequency members of classes may not
contribute at all, due to their autonomy.
The contribution of type frequency to productivity is the fact that when a con-
struction is experienced with different items occupying a position, it enables the
parsing of the construction. If happiness is learned by someone who knows no re-
lated words, there is no way to infer that it has two morphemes. If happy is also learned,
then the learner could hypothesize that –ness is a suffix, but only if it occurs on other
adjectives would its status as a suffix become established. Thus a certain degree of
type frequency is needed to uncover the structure of words and phrases. In addition,
a higher type frequency also gives a construction a stronger representation, making
it more available or accessible for novel uses.
In addition, other factors important for productivity have been identified by
Jennifer Hay and colleagues. In particular, Hay (2002) and Hay and Baayen (2002)
present evidence that it is the type frequency of words within which the affix can
be parsed out that determines productivity, not just the type frequency of the affix.
The “parsibility” of a word depends on both phonological and semantic factors.
2.5. Frequency and category formation
The formation of grammar is in large part an issue of categorization. Every construc-
tion is characterized by the categories that fill its various slots. Every phoneme con-
stitutes a category, often made complex by the existence of allophones. Semantic
notions such as past or future tense also constitute categories, again made complex
by meaning changes in specific contexts. Even the contexts of use of particular con-
structions constitute categories. Both type and token frequency are important to cate-
gory formation.
The properties of the types included in a category establish its boundaries while
the number of types relates to the degree of productivity of the construction refer-
ring to the category. In research into exemplar models (in which the category con-
sists of the experienced exemplars), token frequency can be seen to influence
the perception of the center of the category, as well as its boundaries (Nosofsky
1988). In phonetic categorization, high-frequency exemplars tend to be maintained
while low-frequency ones are marginalized and lost (this volume, chapter 10). In
semantic categorization, a similar phenomenon occurs; in a corpus and experimental
study of the pairing of verbs that mean ‘become’ with adjectives in Spanish, it was
found that the high-frequency pairs served as the center of some of the most pro-
ductive categories (Bybee and Eddington 2006). Similarly, Casenhiser and Gold-
berg (2005) show that children and adults learn a new construction faster if they
are exposed to one higher frequency token as well as several types that exemplify
the construction.](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-28-320.jpg)
![16 BACKGROUND AND CURRENT CONTEXT
3. How high, how low?
Throughout this introduction, I have used the terms high and low frequency in a fairly
vague way. In the chapters collected here frequency is counted with a variety of cutoff
points, depending upon the goals of the particular study. In addition, different sources
for frequency information are consulted. There is no one method for doing frequency
research. When I began to study frequency effects, the available counts were almost
all based on written language. Fortunately, now there are large databases of spoken
language as well as software that makes extracting frequency information from such
databases fairly straightforward. Thus now it is possible to choose a database that is
relevant to the particular hypotheses being tested.
When one is studying token frequency, there is an inherent problem in deter-
mining the point at which high should be distinguished from low. The problem is
that there will be many tokens of high-frequency types and many types represented
among the low-frequency tokens. Thus if one draws the line so that half the tokens
are high frequency and half are low, the high-frequency group may have very few
types in it. Conversely, if one puts half the types in the high group and half in the low
group, the number of tokens in the high group will vastly outnumber those in the
low group. Most studies take a compromise between these two position by looking
for a natural gap in the frequency ranks that puts about 30 to 50% of the tokens in
one group and 50 to 70% of the tokens in the other.
For the research I have done, the strategy of including more than two frequency
groups—say high-, mid-, and low-frequency groups—has not improved the descrip-
tion of the data. But the possibility certainly remains that items in the medium range
have their own properties and effects. Certainly it has been useful to refer to items of
extreme high frequency, as these are most likely grammaticizing or exhibiting autonomy
in one way or another. However, it has also been found that for types of phonetic re-
duction that are quite general (e.g., final [t] or [d] deletion in English) a continuous
effect throughout the frequency levels can be observed (Gregory et al. 1999).
It is important not to overlook the fact that even a small number of repetitions in
the speaker’s experience has a cognitive effect. As pointed out in chapter 13, certain
English words that occur only in fixed expressions, such as dint, bated, and hale in
the expressions by dint of, with bated breath, and hale and hearty, are actually of
rather low frequency (not listed at all among the 1 million words of Francis and Kucera
1982). Yet these expressions are part of the knowledge of English speakers. The
pervasive use of prefabricated word sequences, which form a major part of any dis-
course, points to the same conclusion. Prefabs are word sequences such as promi-
nent role, beyond repair, to need help, which are transparent semantically but used
conventionally. Some counts show that more than half of the word choices in a spo-
ken or written discourse are determined by membership in a prefab (Pawley and Syder
1983; Erman and Warren 2000; Wray 2002). Prefabs are not especially high in fre-
quency, and yet it is clearly their repetition that has given them their social status of
conventionalized and their cognitive status of easily accessed routines. In addition,
certain phrases that are not particularly frequent, though conventionalized for par-
ticular situations, such as who goes there? and how goes it?, are conservative in using
main verb inversion rather than do. We also find conservative behavior in phrases](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-29-320.jpg)
![Chapter 2
Word Frequency in Lexical
Diffusion and the Source of
Morphophonological Change
23
Two ideas that have been around for almost a century need to be reexamined in
the light of our current theories of linguistic change. These ideas involve the rela-
tionship between word frequency and morphophonological change. I will divide
morphophonological change in the traditional way into sound change, which I view
as being largely, perhaps entirely, phonetically motivated, and analogy, which I con-
sider to be primarily conceptually motivated. The particular point I want to discuss
is that the relation between sound change and word frequency is just the reverse of
the relation between analogy and word frequency. That is, as Schuchardt observed
in 1885, sound changes affect the most frequent lexical items first (1885 [1972]).
And Paul, around 1886, observed that analogical leveling tends to affect infrequent
items first (1890 [1972]). Or, stated differently, infrequent items are the most resis-
tant to phonetically motivated change, while frequent items are the most resistant to
conceptually motivated change. I have gathered some data that document these two
tendencies, and I will present these data before speculating on the meaning of this
difference between the types of change for theories of the source of morphophono-
logical change. In particular I will speculate on the role of the child versus the adult
speaker in the initiation of linguistic change.
1. Phonetically conditioned change
First consider phonetically motivated change, sound change. It has often been ob-
served that reductive changes affect frequent words and phrases, thus God be with](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-36-320.jpg)
![WORD FREQUENCY IN LEXICAL DIFFUSION 25
deletion. When words that do not have schwa deletion for phonological reasons are
taken out of the Rarely category, as we might expect, the average frequency of words
in this category falls to the level that we see in the fourth column.
There are three phonological environments that block schwa deletion before r.
One is the presence of a preceding flap, derived historically from a t (and possible
from a d, but there are so few words in the corpus with -dVry that no conclusions can
be drawn).5 The flap appears where a /t/ was preceded by a vowel or /r/, for example,
artery, flattery, but not where a consonant precedes /t/, for example, factory, adul-
tery, elementary. In the latter forms, deletion may occur, but in the former forms, the
eight words with flaps, artery, battery, buttery, flattery, lottery, pottery, rotary, and
watery, deletion does not usually occur.
Deletion is also constrained if a syllable-initial consonant cluster precedes the
schwa. In burglary a deletion would leave a totally unacceptable consonant cluster
of glr for syllable-initial position. Similarly, deletion is constrained in library, con-
trary, mercury, and penury, although [laybri] and [mkri] are possible pronuncia-
tions. Finally, a preceding affricate, as in forgery or butchery, constrains deletion for
many speakers.
The individual speakers displayed different phonological constraints, so the
fourth column was computed separately for each subject. I considered the subject to
be exercising a phonological constraint if all or all but one of the words of a phono-
logical type appeared in the Rarely category. When this occurred, all the words of
that type were removed and the average was recomputed. However, if a subject put
two or more words of a phonological type in the Sometimes or Usually category, I
did not take the words of this type out of the Rarely category. The lowercase initials
following the fourth column indicate which constraints were considered applicable
for that subject: f stands for flap, c for cluster, a for affricate. Thus subjects A, B, and
2.1. Average Frequency of Words Categorized for the
Application of Schwa-deletion
Rarely without
Phonologically
Usually Sometimes Rarely Constrained Words
A 38.75 (16) 13.95 (38) 14.17 (53) 6.40 (f, c, a)
B a 23.87 (46) 12.33 (66) 5.81 (f, c, a)
C 22.00 (67) 15.54 (24) 15.00 (20) 8.70 (c)
D 24.03 (37) 20.44 (33) 18.76 (42) 5.81 (c)
E 29.57 (35) 13.90 (11) 12.67 (66) 5.48 (f, c, a)
F 10.38 (8) 31.44 (25) 13.11 (79) 11.77 (f, c)
20.75 (4)
G 27.24 (45) 21.27(11) 19.04 (49) 7.28 (f, c)
H 35.04 (28) 17.5 (10) 11.57 (74) 9.93 (f, c)
Average of
all subjects 26.72 15.99 6.18
Average frequency of all 112 words: 17.07
aSubject B did not put any words in the Usually category.](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-38-320.jpg)
![WORD FREQUENCY IN LEXICAL DIFFUSION 27
primary difference in word frequency between written and spoken language will be
that frequent words in written language will be even more frequent in spoken lan-
guage, while infrequent words in written language will be even more infrequent or
totally non-existent in spoken language. Thus the difference in frequency found in
written texts will be increased, probably in geometric proportions, in spoken language.
This means that the frequency differences displayed in tables 2.1 and 2.2 would be
greatly increased if I had used a frequency count of spoken language.
Another problem may be the fact that only eight subjects were used. However, the
findings are quite consistent, and it seems to me in the case used here, of schwa dele-
tion, the hypothesis is really non-controversial. This is because it is a reductive type
change and I do not think anyone would deny that reduction in frequent items is to be
expected. It would be more interesting (although more difficult) to run more extensive
tests on sound changes in progress that are not so obviously reductive, for example,
assimilations and vowel shifts. I would venture to guess that all sound changes, even
those that we do not consider to be reductive, take place first in frequent words.
Let me make clear my interpretation of the status of schwa deletion in English.
I claim that schwa deletion, especially poststress schwa deletion, represents a “sound
change” in progress. It is potentially totally general, but at present it has not yet reached
all the lexical items of English. At present schwa deletion may appear sporadic. Simi-
larly, vowel reduction in English appeared sporadic to Fidelholtz, who referred to it
as “an exceptional rule” and concluded that “frequent words can do exceptional
things.” I do not think that frequent words do “exceptional things”; rather, they are
merely ahead of other words in undergoing phonetic change. My view is more like
that of Schuchardt, who says “every sound change in some phase is sporadic” (1885
[1972]: 63), except that I would say that every sound change in some phase appears
to be sporadic until we know what all the variables are. I will return in section 3 to a
discussion of the implications of word frequency as a variable affecting sound change.
2. Analogical leveling
Analogical leveling appears to affect less frequent forms; frequent paradigms stub-
bornly cling to their suppletion. (I am leaving analogical extension out of the dis-
cussion.) While our general experience supports this idea, I wanted to check the
hypothesis a little more rigorously before basing any claims on it. In particular I looked
at classes of Old English strong verbs that have some members that survive as strong
verbs in Modern English and some members that have become weak verbs, that is,
that have undergone analogical leveling in Modern English. Using Sweet’s Anglo-
Saxon Primer (Davis 1965), I found three appropriate classes of verbs.6 The frequency
counts for these verbs (again from Kucera and Francis) are listed in table 2.3. To
obtain these frequency counts, I totaled all entries that could be verb forms, for ex-
ample, drive, drives, driven, drove, and driving. In some cases, however, the frequency
score will include non-verbs that are spelled identically, since the frequency count is
done totally on the basis of spelling, disregarding the words’ function in the sentence.
The verbs whose frequency score may include non-verbs and thus overrepresent the
frequency of the verb, are marked with asterisks.](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-40-320.jpg)
![30 BACKGROUND AND CURRENT CONTEXT
surely be the more frequent paradigms; for frequent forms there is a greater avail-
ability of the model, more opportunity to practice, and greater pressure to conform.
(Notice that I am not claiming that children acquire the most frequent words first.)
Actually, the imperfect learning hypothesis need not be restricted to young chil-
dren. Adult speakers may very well play a role in analogical leveling. For infre-
quent paradigms adults may not be exactly sure of all the forms, or the mere
infrequency of a suppletive paradigm makes an analogical formation more accept-
able. For instance, as I mentioned earlier, creeped is not standard, but I would not
flinch if I heard it, and I might even produce it myself, although I know crept is
“correct.” However, keeped would definitely cause a negative reaction, because
the form kept is so solidly established, due to its frequency. Thus I conclude that
the imperfect learning hypothesis is quite a reasonable hypothesis about analogi-
cal change, but it need not be applied only to children.
The imperfect learning hypothesis has also been proposed to explain phonetic
changes, chiefly by David Stampe (1969, 1973). Again we find certain correspon-
dences between the substitutions that children make and known historical sound
changes, although the correspondence is by no means perfect (Dressler 1974). Stampe
claims that children possess a set of innate phonological processes that they must
learn to suppress in order to acquire a language. The failure to suppress a natural
process may result in a sound change.
If the reasoning I have followed in explaining how imperfect learning allows
analogical change is correct, that is, if it is correct that the forms that are most likely
to be imperfectly learned are the infrequent forms, then phonetic changes such as
schwa deletion and any other change that follows the same pattern of lexical diffu-
sion cannot be attributed to imperfect learning. A child’s command of the phonetic
shape for a frequent word will certainly be closer to the adult model than for an in-
frequent word, since as I mentioned earlier, for frequent words the model is more
available and the pressure to conform is greater. Thus we cannot postulate that chil-
dren initiate phonetic changes such as vowel reduction or schwa deletion.
What, then, is the source of a change such as schwa deletion? I suggest that the
source and cause of such changes should be sought in adult speech, especially the
casual style of speech. It is no accident that “new rules” appear to start out as op-
tional or variable rules in the most casual styles of speech. These phonetic changes
(or rules) are naturally occurring in casual or fast speech and are due to a tendency to
minimize effort; thus vowels move closer to the neutral position, as in Fidelholtz’s vowel
reduction, and schwas grow shorter and shorter until they are simply skipped. Or these
phonetic changes come about as changes in the timing of articulatory gestures—thus
we get assimilations as in didya, wouldya, and so on.
Zipf (1935 [1965]) would claim that these phonetic changes originate because
certain sounds or sequences of sounds are frequent. I would suggest, however, that
these phonetic changes are motivated solely by phonetics, not frequency, but that
they affect frequent items first because it is the frequent words that make up the sub-
stance of casual speech. That is, the more casual the style, the greater the chance that
the lexical items used will be the more frequent ones. If schwa deletion in its initial
stages is restricted to casual speech, then only the words used in casual speech will
have variable forms.](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-43-320.jpg)
![32 BACKGROUND AND CURRENT CONTEXT
Calvary Introductory Savory
Cannery Ivory Scenery
Celery Leathery Sedimentary
Century Library Sensory
Chancery Lottery Shivery
Chicanery Luxury Shrubbery
Complementary Machinery Silvery
Complimentary Mastery Slavery
Compulsory Memory Slippery
Contradictory Mercury Snobbery
Contrary Misery Sorcery
Creamery Mockery Spidery
Cursory Mystery Splintery
Debauchery Nursery Summary
Delivery Papery Supervisory
Desultory Parliamentary Supplementary
Directory Penury Surgery
Discovery Peppery Treachery
Dispensary Peremptory Treasury
Documentary Perfumery Unsatisfactory
Drapery Perfunctory Vagary
Drudgery Periphery Victory
Elementary Perjury Watery
Embroidery
Notes
I am very grateful to Richard Mowrey for many stimulating discussions on the topic of this
paper.
1. Kucera and Francis (1967) give a frequency of 4 for celery and 11 for artillery; this
is due to the type of written texts they use. I am certain that for speakers other than militarists
celery is the more frequent of the pair in spoken language. This pair illustrates clearly a gen-
eral problem with relying on frequency counts of written material. Since I have computed the
results of the test in terms of average frequencies for groups of words, this problem does not
interfere greatly with the validity of my conclusions.
2. The “double-standard” reverse word list was used. It is a list of all left-justified, bold-
face words, with standard meanings occurring in both the Shorter Oxford English Dictionary
and the Merriam Webster New International Dictionary, third edition.
3. For nouns total the frequency counts for singulars and plurals to obtain the figure I
used.
4. One of the problems that made the task difficult is that particular words may have
deletion or not, depending on how they are used. Some examples: (i) [mkri] is a possible
shortening of mercury only when it refers to the make of an automobile, but never for the
element or planet; (ii) Richard Mowrey (personal communication) observed the following:
parliamentary reduces in a parliamentary procedure, but not in Their form of government is
parliamentary; rosary for this speaker usually has a schwa, but in the name Rosary Hill (a
college in Buffalo) the schwa is deleted. These examples suggest that the degree of stress
may condition the presence or absence of the schwa in words that have variable form.
5. The phonological constraints on schwa deletion are due to syllable structure con-
straints. These are explained in detail in Hooper (1976). The deletion under discussion here](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-45-320.jpg)
![WORD FREQUENCY IN LEXICAL DIFFUSION 33
creates a new syllable-initial cluster; e.g., hickory gives [hIkri]. Under the theory of syllable
structure developed in Vennemann and Wilbur (1972) and Hooper (1976), a consonant in
initial position must be stronger than the consonant in second position. The flap in English is
an extremely weak consonant; I suggest that it is too weak to cluster with an /r/. In order to
have a cluster with an /r/, the alveolar consonant must be pronounced as a stop, either [th] or
[d]. The data show, however, that speakers prefer to keep the schwa, rather than pronouncing
a full stop here.
Observe further that the phonological constraints operating here are not language-specific
constraints, for many violations of English syllable structure are allowed; e.g., nasal plus /r/
is allowed, memory, scenery; /s/ plus /r/, nursery. Neither of these clusters occurs ordinarily
as a syllable-initial cluster in English.
6. Sweet’s other classes of strong verbs were not appropriate, either because they had
too few surviving members or because leveling occurred according to phonological subtype.
7. Labov (1966) finds that the degree of vowel shift is greater for some speakers in ca-
sual styles than in more formal styles, but this fact in itself does not prove that the vowel
shifts originate in casual speech.
References
Davis, Norman. 1965. Sweet’s Anglo-Saxon primer. Oxford, Clarendon Press.
Dolby, J. L., and H. L. Resnikoff. 1967. The English word speculum. The Hague: Mouton.
Dressler, Wolfgang. 1974. Diachronic puzzles for natural phonology, Papers from the
Parasession on Natural Phonology, 95–102. Chicago: Chicago Linguistic Society.
Fidelholtz, James L. 1975. Word frequency and vowel reduction in English. In annual Papers
from the eleventh meeting of the Chicago Linguistic Society, 200–213. Chicago: Chi-
cago Linguistic Society.
Halle, Morris. 1964. Phonology in generative grammar. In The structure of language, ed.
J. A. Fodor and J. J. Katz, 334–352. Englewood Cliffs, N.J.: Prentice-Hall.
Hooper, Joan B. 1976. An introduction to natural generative phonology. New York: Aca-
demic Press.
King, Robert D. 1969. Historical linguistics and generative grammar. Englewood Cliffs, N.J.:
Prentice-Hall.
Kiparsky,Paul.1968.Linguisticuniversalsandlinguisticchange.InUniversalsinlinguistictheory,
ed. Emmond Bach and Robert Harms, 170–202. New York: Holt, Rinehart and Winston.
———. 1970. Historical linguistics. In New horizons in linguistics, ed. John Lyons, 302–
315. London: Penguin.
Kucera, Henry, and Nelson Francis. 1967. Computational analysis of present-day American
English. Providence, R.I.: Brown University Press.
Labov, William. 1966. The social stratification of English in New York City. Arlington: Cen-
ter for Applied Linguistics.
Leslau, Wolf. 1969. Frequency as a determinant of linguistic changes in the Ethiopian lan-
guages,” Word 25, 180–189.
Moore, Samuel. 1968. Historical outlines of English sounds and inflections. Revised by Albert
H. Marckwardt. Ann Arbor, Mich.: George Wahr.
Paul, Hermann. 1970 [1890]. Principles of the history of language. Trans. by H. A. Strong.
College Park, Md.: McGrath.
Schuchardt, Hugo. 1885 [1972]. On sound laws: Against the Neogrammarians, in Schuchardt,
the Neogrammarians, and the transformational theory of phonological change(Linguistische
Forschungen 26), ed. Theo Venneman and Terence H. Wilbur, 29–72. Frankfurt am Main.
Athenaum.](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-46-320.jpg)
![34 BACKGROUND AND CURRENT CONTEXT
Stampe, David. 1969. On the acquisition of phonetic representation, Papers from the fifth
regional meeting of the Chicago Linguistic Society, 443–454.
———. 1973. A dissertation on natural phonology. Diss., University of Chicago.
Zipf, George K. 1935 [1965]. The psycho-biology of language. Cambridge, Mass.: MIT
Press.
Zwicky, Arnold. 1972. Note on a phonological hierarchy in English. In Linguistic change
and generative theory, ed. R. P. Stockwell and R. K. S. Macaulay, 275–301. Bloomington:
Indiana University Press.](https://image.slidesharecdn.com/2216798-250610153547-bf6f1557/85/Frequency-Of-Use-And-The-Organization-Of-Language-Joan-Bybee-47-320.jpg)
Frequency Of Use And The Organization Of Language Joan Bybee
- 1. Frequency Of Use And The Organization Of Language Joan Bybee download https://ebookbell.com/product/frequency-of-use-and-the- organization-of-language-joan-bybee-4433596 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. The Frequency Of Life A Collection Of Word Art And Artwork Kari Beyer https://ebookbell.com/product/the-frequency-of-life-a-collection-of- word-art-and-artwork-kari-beyer-48735598 The Collapse Frequency Of Structures Bridges Dams Tunnels Retaining Structures Buildings Dirk Proske https://ebookbell.com/product/the-collapse-frequency-of-structures- bridges-dams-tunnels-retaining-structures-buildings-dirk- proske-48751488 The Red Rover Frequency Of Distress The Rover Series Universe Book 5 Ce Whitaker Iii https://ebookbell.com/product/the-red-rover-frequency-of-distress-the- rover-series-universe-book-5-ce-whitaker-iii-46419648 Frequency Conversion Of Ultrashort Pulses In Extended Laserproduced Plasmas 1st Edition Rashid A Ganeev Auth https://ebookbell.com/product/frequency-conversion-of-ultrashort- pulses-in-extended-laserproduced-plasmas-1st-edition-rashid-a-ganeev- auth-5359188
- 3. Frequency Analysis Of Vibration Energy Harvesting Systems 1st Edition Xu Wang https://ebookbell.com/product/frequency-analysis-of-vibration-energy- harvesting-systems-1st-edition-xu-wang-5601462 Frequency Dictonary Of Spanish Davies M https://ebookbell.com/product/frequency-dictonary-of-spanish- davies-m-1345088 Frequency Conversion Of Uitrashort Pulses In Extended Laserproduced Plasmas 2016th Edition Rashid Aganeev https://ebookbell.com/product/frequency-conversion-of-uitrashort- pulses-in-extended-laserproduced-plasmas-2016th-edition-rashid- aganeev-59735234 Handbook Of Frequency Allocations And Spectrum Protection For Scientific Uses Panel On Frequency Allocations And Spectrum Protection For Scientific Uses https://ebookbell.com/product/handbook-of-frequency-allocations-and- spectrum-protection-for-scientific-uses-panel-on-frequency- allocations-and-spectrum-protection-for-scientific-uses-2006802 Handbook Of Frequency Allocations And Spectrum Protection For Scientific Uses Second Edition 2nd Edition And Medicine Engineering National Academies Of Sciences Division On Engineering And Physical Sciences Board On Physics And Astronomy Committee On Radio Frequencies Panel On Frequency Allocations And Spectrum Protection For Scientific Uses https://ebookbell.com/product/handbook-of-frequency-allocations-and- spectrum-protection-for-scientific-uses-second-edition-2nd-edition- and-medicine-engineering-national-academies-of-sciences-division-on- engineering-and-physical-sciences-board-on-physics-and-astronomy- committee-on-radio-frequencies-panel-on-frequency-allocations-and- spectrum-protection-for-scientific-uses-51782004
- 6. Frequency of Use and the Organization of Language
- 7. This page intentionally left blank
- 8. Frequency of Use and the Organization of Language Joan Bybee 1 2007
- 9. 3 Oxford University Press, Inc., publishes works that further Oxford University’s objective of excellence in research, scholarship, and education. Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With offices in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Copyright © 2007 by Oxford University Press, Inc. Published by Oxford University Press, Inc. 198 Madison Avenue, New York, New York 10016 www.oup.com Oxford is a registered trademark of Oxford University Press All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press. Library of Congress Cataloging-in-Publication Data Bybee, Joan L. Frequency of use and the organization of language / Joan Bybee. p. cm. Includes bibliographical references and index. ISBN-13 978-0-19-530156-4; 978-0-19-530157-1 (pbk.) ISBN 0-19-530156–0; 0-19-530157-9 (pbk.) 1. Frequency (Linguistics) 2. Linguistic change. I. Title. P128.F73B935 2006 410.1'51—dc22 2006040009 1 3 5 7 9 8 6 4 2 Printed in the United States of America on acid-free paper
- 10. This volume reprints articles written over a period of almost three decades. Needless to say, I am grateful to a large community of scholars who have reacted to the hy- potheses and findings of these articles and eventually to the theory that has emerged from them. Of course, I am most indebted to my co-authors, Mary Alexandra Brewer, Elly Pardo, Dan I. Slobin, Carol Lynn Moder, Jean E. Newman, Sandra Thompson, and Joanne Scheibman. Collaboration with these researchers allowed me to move beyond my own limitations and produce results that I could never have achieved alone. For helping with the preparation of the manuscript, I thank Vsevolod Kapatsinski and Clayton Beckner. I am also grateful to Peter Ohlin of Oxford University Press for suggesting this volume and to my family for their continued support. I gratefully acknowledge permission from the publishers to reprint the following articles and book chapters, which constitute this volume. Hooper, Joan B. 1976. Word frequency in lexical diffusion and the source of morphophonological change. In Current progress in historical linguistics, ed. William Christie, 95–105. Amsterdam: North Holland. Reprinted with kind permission from Elsevier. Bybee, Joan L., and Mary Alexandra Brewer. 1980. Explanation in morphophonemics: Changes in Provençal and Spanish preterite forms. Lingua 52, 201–242. Reprinted with kind permission from Elsevier. Bybee, Joan L., and Elly Pardo. 1981. On lexical and morphological conditioning of alternations: A nonce-probe experiment with Spanish verbs. Linguistics 19, 937–968. Reprinted with kind permission from Mouton de Gruyter. Bybee, Joan L., and Dan I. Slobin. 1982. Rules and schemas in the development and use of the English past tense. Language 58, 265–289. Reprinted with kind permission from the Linguistic Society of America. ACKNOWLEDGMENTS
- 11. Bybee, Joan L., and Carol Lynn Moder. 1983. Morphological classes as natural categories. Language 59, 25l–270. Reprinted with kind permis- sion from the Linguistic Society of America. Bybee, Joan L., and Jean E. Newman. 1995. Are stem changes as natural as affixes? Linguistics 33, 633–654. Reprinted with kind permission from Mouton de Gruyter. Bybee, Joan L. 1995. Regular morphology and the lexicon. Language and Cognitive Processes 10, 425–455. Reprinted with kind permission from Lawrence Erlbaum Associates. Bybee, Joan. 2000. The phonology of the lexicon: Evidence from lexical diffusion. In Usage-based models of language, ed. Michael Barlow and Suzanne Kemmer, 65–85. Stanford, Calif.: CSLI. Reprinted with kind permission from CSLI Publications. Bybee, Joan L. 2000. Lexicalization of sound change and alternating environments. In Papers in Laboratory Phonology, vol. 5: Acquisition and the Lexicon, ed. Michael Broe and Janet Pierrehumbert, 250–268. Cambridge: Cambridge University Press. Reprinted with kind permission from Cambridge University Press. Bybee, Joan L. 2002. Word frequency and context of use in the lexical diffusion of phonetically conditioned sound change. Language variation and change 14, 261–290. Reprinted with kind permission from Cam- bridge University Press. Bybee, Joan, and Sandra Thompson. 1997. Three frequency effects in syntax. Berkeley Linguistics Society 23, 65–85. Reprinted with kind permission from the Berkeley Linguistics Society. Bybee, Joan L. 1998. The emergent lexicon. In CLS 34: The panels, ed. M. Catherine Gruber, Derrick Higgins, Kenneth S. Olson, and Tamra Wysocki, 421–435. University of Chicago: Chicago Linguistic Society. Reprinted with kind permission from the Chicago Linguistic Society. Bybee, Joan L., and Joanne Scheibman. 1999. The effect of usage on degrees of constituency: The reduction of don’t in American English. Linguistics 37(4), 575–596. Reprinted with kind permission from Mouton de Gruyter. Bybee, Joan L. 2002. Sequentiality as the basis of constituent structure. In The evolution of language out of pre-language, ed. T. Givón and Bertram F. Malle, 109–132. Amsterdam: John Benjamins. Reprinted with kind permission from John Benjamins Publishing. Bybee, Joan L. 2003. Mechanisms of change in grammaticization: The role of frequency. In Handbook of historical linguistics, ed. R. Janda and B. Joseph, 602–623. Oxford: Blackwell. Reprinted with kind permission from Blackwell Publishers Limited. vi ACKNOWLEDGMENTS
- 12. CONTENTS Part I Background and Current Context Introduction to Part I, 3 1. Introduction, 5 2. Word Frequency in Lexical Diffusion and the Source of Morphophonological Change, 23 Part II Frequency as a Determinant of Morphological Structure Introduction to Part II, 37 3. Explanation in Morphophonemics: Changes in Provençal and Spanish Preterite Forms, 41 with Mary Alexandra Brewer 4. On Lexical and Morphological Conditioning of Alternations: A Nonce-Probe Experiment with Spanish Verbs, 74 with Elly Pardo 5. Rules and Schemas in the Development and Use of the English Past Tense, 101 with Dan I. Slobin
- 13. 6. Morphological Classes as Natural Categories, 127 with Carol Lynn Moder 7. Are Stem Changes as Natural as Affixes?, 148 with Jean E. Newman 8. Regular Morphology and the Lexicon, 167 Part III Phonetic Change: Frequency in Context Introduction to Part III, 197 9. The Phonology of the Lexicon: Evidence from Lexical Diffusion, 199 10. Lexicalization of Sound Change and Alternating Environments, 216 11. Word Frequency and Context of Use in the Lexical Diffusion of Phonetically Conditioned Sound Change, 235 Part IV Frequency Effects in Mophosyntax Introduction to Part IV, 267 12. Three Frequency Effects in Syntax, 269 with Sandra Thompson 13. The Emergent Lexicon, 279 14. The Effect of Usage on Degrees of Constituency: The Reduction of Don’t in English, 294 with Joanne Scheibman 15. Sequentiality as the Basis of Constituent Structure, 313 16. Mechanisms of Change in Grammaticization: The Role of Frequency, 336 Index, 359 viii CONTENTS
- 14. PART I BACKGROUND AND CURRENT CONTEXT
- 15. This page intentionally left blank
- 16. Introduction to Part I 3 Chapter 1 provides background on the effects of frequency of use on linguistic elements, explaining how such effects have been regarded in recent linguistic theory. This introduction lays out briefly the topics further developed in the chapters of this volume, which together demonstrate a major role for frequency of use in determin- ing the nature of language. The second chapter is my first paper on the topic of frequency of use; it was presented at a conference on historical linguistics in 1975. My interest in and in- vestigations of frequency began rather tentatively, given the climate in which I was educated and in which I worked. The first research in which I investigated frequency effects was in the context of language change, and it involved two frequency ef- fects that most historical linguists knew about and perhaps considered plausible, though of little interest given the prevailing view that language change is change in the grammar. The strength of the paper is not in its methodology; at that time, linguists were unaccustomed to evaluating quantitative or experimental data and I was no excep- tion. The evidence presented, however, seems to be suggestive enough to render the theoretical points valid. The paper addresses mechanisms of change that involve usage and the interaction of usage with cognitive representations. The hypotheses of this paper are further developed in the chapters presented here and in the work on lexical diffusion by Phillips (1984, 2001, and elsewhere). The phonological aspects of the schwa-deletion process discussed in this chapter are further explored in Hooper (1978).
- 17. References Hooper, Joan B. 1978. Constraints on schwa deletion in American English. In Recent devel- opments in historical phonology, ed. Jacek Fisiak, 183–207. The Hague: Mouton. Phillips, Betty S. 1984. Word frequency and the actuation of sound change. Language 60, 320–342. ———. 2001. Lexical diffusion, lexical frequency, and lexical analysis. In Frequency and the emergence of linguistic structure, ed. Joan L. Bybee and Paul Hopper, 123–126. Amsterdam: John Benjamins. 4 BACKGROUND AND CURRENT CONTEXT
- 18. 1. Background A newcomer to the field of linguistics might be surprised to learn that for most of the twentieth century facts about the frequency of use of particular words, phrases, or constructions were considered irrelevant to the study of linguistic structure. To the uninitiated, it does not seem unreasonable at all to suppose that high-frequency words and expressions might have one set of properties and low-frequency words and ex- pressions another. So how is it that so many professional linguists for so many decades, maybe even centuries, have missed (or perhaps avoided) this basic point? One factor is that frequency effects tend to be observable at the level of the in- dividual word or expression, while linguists have tended to focus their interest on the broader patterns of structure and the more abstract and generalized categories. While language is full of both broad generalizations and item-specific properties, linguists have been dazzled by the quest for general patterns. Of course, the abstract structures and categories of language are fascinating. But I would submit that what is even more fascinating is the way that these general structures arise from and inter- act with the more specific items of language use, producing a highly conventional set of general and specific structures that allow the expression of both conventional and novel ideas. In terms of the history of the field, one can see not just the glamour but also the utility of generalization as the basis of the Neogrammarian doctrine. Many sound changes turn out to be completely regular in the sense that they affect all the words of a language with the relevant phonetic environment. This fact is interesting in its Chapter 1 Introduction 5
- 19. 6 BACKGROUND AND CURRENT CONTEXT own right, but it is also the indispensable basis of the comparative method—the fact that systematic relations can be found between the sounds of the words of two lan- guages establishes their genetic relationship. It has also become the basis of the gen- erative position that sound change is grammar change. This latter position in particular also rules out irregularity in the spread of a sound change. Similarly, on the level of grammar for both descriptive and theoretical purposes, a focus on general patterns is absolutely essential. In cross-linguistic studies, the discovery of typologies of word order, case marking, or morphological form would be impossible without a focus on the general patterns of a language. However, it is now time to examine the role that specific expressions and structures play in gram- mar and here frequency becomes important. The other major theoretical factor working against an interest in frequency of use in language is the distinction, traditionally traced back to Ferdinand de Saussure (1916), between the knowledge that speakers have of the signs and structures of their language and the way language is used by actual speakers communicating with one another. American structuralists, including those of the generativist tradition, accept this distinction and assert furthermore that the only worthwhile object of study is the underlying knowledge of language (Chomsky 1965 and subsequent works). In this view, any focus on the frequency of use of the patterns or items of language is con- sidered irrelevant. In fact, given Chomsky’s disdainful review of Skinner’s Verbal Behavior, any mention of frequency, repetition, or imitation has been dismissed by Chomsky’s followers as irrelevant to the true nature of language. A major challenge to this doctrine, arising in the 1970s and continuing with growing strength to the present, is a view of grammar as arising from the patterns of language use in actual discourse. This functionalist view fostered studies of gram- mar in the context of discourse and interaction, as well as the dynamic study of gram- mar as found in works on grammaticization. In this view, it is proposed that grammar arises diachronically because of the commonly used discourse patterns that humans need to communicate (Li 1975, 1977; Givón 1979; Hopper and Thompson 1980, 1984). The empirical support for this hypothesis is found in diachronic, typological, and discourse studies (e.g., DuBois 1987). The importance of the work of Joseph Greenberg in inspiring this movement cannot be overestimated. His detailed cross- linguistic investigations revealed common diachronic patterns across related and unrelated languages, allowing him to formulate both synchronic universals and uni- versals of change. This typological work inspired others to seek to explain the implicational hierarchies he discovered by considering discourse patterns (e.g., Greenberg 1978; Givón 1979). Perhaps more germane to this work, however, is Greenberg’s (1966) explicit investigations of the role of frequency in deriving the effects underlying the theory of markedness. While not proposing a causal mecha- nism, he demonstrates that unmarked members of categories throughout the gram- mar are more frequent than marked members. The argument that is gaining strength is that separating language from the way it is used removes a valuable source of explanation for why language has grammar and what form that grammar takes. Despite the empirical bent of the functionalist movement and the acceptance of the notion that conventionalization through repetition creates grammar, there was still very little investigation into the nature of the effects of repetition or frequency
- 20. INTRODUCTION 7 on the cognitive representation of language. One reason for this gap has to do with the unfortunate separation in both method and theory between psycholinguists and linguists, such that linguists often have trouble making use of psycholinguistic find- ings. The other reason is that the more empirically minded functionalists, reacting negatively to the arrogance of generativists’ proposing “mental representations” on very scant evidence, shied away from proposals about what might be going on in people’s heads. More recently the field of linguistics has begun to diversify in the range of evi- dence considered relevant and the range of views of grammar and how it relates to the rest of cognition. In particular there is a new appreciation of item-specific knowl- edge and how it interacts with and gives rise to more general knowledge. The role of experience in the formation of linguistic categories and representations is being in- vestigated in what have come to be called usage-based models (Bybee 1985, this volume, chapter 8; Langacker 1987, 2000; Barlow and Kemmer 2000) and probabi- listic models (Bod, Hay, and Jannedy 2003) An important force in raising the vis- ibility of item-specific factors is the continuing work on connectionist modeling of language, where it is shown that it is possible to model the buildup of generaliza- tions from specific instances without throwing away the knowledge of the specific instances (Rumelhart and McClelland 1986; Elman 1990; Bybee and McClelland 2005). This suggests a model of language in which the predictable patterns are not separate from representations, where lexicon and grammar are interwoven as are the specific and the general. Both type and token frequency in the input have been found to be factors important to the modeling of linguistic systems, such as the English past tense (see section 2 of this chapter). A new movement in language acquisition focuses on specific word combina- tions used early by children (Lieven, Pine, and Baldwin 1997; Tomasello 2003), and how these specific combinations begin to pattern into constructions (Savage et al. 2003; Casenhiser and Goldberg 2005; Daçbrowska and Lieven 2005). In these works, both token and type frequency in the child’s experience are shown to be important for acquiring the specific combinations and the generalized pattern. Rather than view- ing acquisition as guided by innate categories and rules, these researchers demon- strate that children build up categories and sequential patterns using the utterances they have experienced in their own and in their caregivers’ speech. An important methodological breakthrough comes with the availability of large corpora recording natural language use, and the concordance software that allows for quantitative studies. Corpus studies often reveal quantitative patterns that are not available to introspection but that are likely to be important to the understanding of how speakers store and access units of language. Along with fairly traditional posi- tional analysis, which for instance reveals that English of does not behave like other prepositions (Sinclair 1991), quantitative studies of phonetic reduction using large corpora reveal that the units of speech are grouped in production according to their specific histories of co-occurrence (Gregory et al. 1999; Jurafsky et al. 2001; this volume, chapters 11, 14, and 15). Thus more recent theories are approaching a common ground in which it is hy- pothesized that specific instances of experience give rise to generalizations, and they can do so without being swallowed up themselves by the general pattern. This inter-
- 21. 8 BACKGROUND AND CURRENT CONTEXT action of the specific and the general is a general property of human cognition that has been known in psychology from the late 1960s. Experiments in which subjects are exposed to patterns of dots, colored geometrical objects, and even line drawings of facial features demonstrate that subjects form categories based on similarity to a prototype that may never have appeared in the experiment, but they do so without losing the specific information about the exemplars they have experienced (Posner and Keele 1968; Medin and Schaffer 1978). These studies have shown that both simi- larity and frequency in experience determine the nature of categories. When the stimuli are very different and each is shown repeatedly, sensitivity to the individual patterns is relatively high; when the stimuli are very similar and each is shown only once, sensitivity to the individual patterns is relatively low and sensitivity to the prototype or central tendency is higher. The essential points are that performance with each item reflects both specific and general information and, in addition, the number of repetitions influences the nature of the category. This type of system, often referred to as an emergent system, has the properties we find described as complex systems in nature. In complex systems, a small num- ber of mechanisms operate in real time and with repetition lead to the emergence of what appears to be an organized structure, such as a sand dune. However, we know that a sand dune is not fixed in time and space but is ever altering and becoming. So we see that language is also always in a process of becoming—creating, losing, and re-creating structures that are never absolutely fixed, allowing for continued varia- tion and change (Lindblom, MacNeilage, and Studdert-Kennedy 1984; Hopper 1987; Holland 1998). In such theories repetition of actions brings about the formation of structures; thus in language, too, we see that repetition is a necessary component of grammar formation (Haiman 1994). The reason frequency or repetition plays a role in grammar formation is that the mind is sensitive to repetition. This is a domain- general principle; that is, it does not apply just to language but to other cognitive domains as well. What are the cognitive responses to repetition? There are multiple responses and their effects can change according to the extent of the repetition. The chapters of this volume represent my investigations into these effects over the last thirty years, from the first recognition of the potential explanatory power of frequency effects through a continuing effort to understand them and their interaction with processing factors, as well as phonetics, semantics, and grammar. One might reasonably ask how it was that in the early 1970s I became interested in a topic that was so unpopular, one that indeed might brand me as a behaviorist. The source of my interest was, and remains, language change. When I was a student at UCLA, Theo Vennemann had briefly introduced the historical linguistic class I attended to the ideas of Hugo Schuchardt, who had observed that high-frequency words tend to undergo sound change before low-frequency words. Of course, this phenomenon is readily observable in the extreme reduction of high-frequency phrases. I, in turn, passed on this observation to my students at SUNY at Buffalo. Two of my students, Richard Mowrey and William Pagliuca, became very interested in phonetic reduction and frequency, which led me to consider the topic further. In teaching his- torical linguistics, I also noted that it was generally observed that analogy in the form of regularization affected low-frequency paradigms before high-frequency ones. This
- 22. INTRODUCTION 9 paradox seemed worth pointing out for the implications it has for the understanding the mechanisms involved in these two types of change (see this volume, chapter 2). Admittedly, at the time I did not realize the potential for both of these observations to totally change the way I looked at phonology and grammar. When I undertook a cross-linguistic study of morphology, the importance of frequency was observable both within and across paradigms in all the languages I investigated, leading to a novel conception of morphological structure (see Bybee 1985 and in this volume part II). A view of morphological structure as affected by language use kept producing more and more interesting hypotheses and findings. Similarly, the investigation of frequency effects in phonology has led to many inter- esting questions and solutions to some old problems (see Bybee 2001b and in this volume part III). I feel the issues in morphosyntax may be more complex, but many interesting findings have already arisen in this usage-based conception of grammar (see part IV). This volume reports on these investigations, which together paint a picture of language as having a variegated and mutable shape. While there are many general patterns of great interest, the topology of the cognitive organization of lan- guage is neither flat, regular, nor permanent. High-frequency words and phrases grow strong with repetition and loom large, forming looser connections with other items, while low-frequency words and expressions are less prominent but gain stability by conforming to patterns used by other items. General patterns dominate networks where more specific patterns can be overpowered unless represented by high-frequency items. Words that have phonological similarities cluster together; constructions are connected if they have properties in common. Instances of constructions that grow to high frequency slowly disengage from the more general pattern to become inde- pendent constructions. Thus the phonetic and semantic substance of language is ever being shaped by the effects of usage. In the next section I provide an overview of the frequency effects that I have investigated. At the beginning of each part of the book, I have provided a few para- graphs that put the ideas in these articles in perspective with respect to the environ- ment in which they were written, as well as subsequent research. 2. What are frequency effects? There are various ways of counting frequency, various units upon which to base the count, and various ways in which frequency has cognitive effects. First, I distinguish token and type frequency. Token frequency counts the number of times a unit ap- pears in running text. Any specific unit, such as particular consonant [s], a syllable [ba], a word dog or the, a phrase take a break, or even a sentence such as Your toast popped up can have a token frequency. Type frequency is a very different sort of count. Only patterns of language have type frequency, because this refers to how many distinct items are represented by the pattern. Type frequency may apply to phonotactic sequences; it would be the count of how many words of the language begin with [sp] versus how many begin with [sf]. It may apply to morphological patterns, such as stem + affix combinations. For instance, the English past tense pat- tern exemplified by know, knew; blow, blew has a lower type frequency than the regu-
- 23. 10 BACKGROUND AND CURRENT CONTEXT lar pattern of adding the –ed suffix. Syntactic patterns or constructions also have type frequencies: the ditransitive pattern in English, exemplified by He gave me the change, is used with only a small set of verbs, while the alternate pattern He gave the change to me is possible with a large class of verbs (Goldberg 1995). In this section I treat the three known effects of token frequency and the effect of type frequency, as well as their interaction. 2.1. Token frequency: The Conserving Effect The first effect of token frequency to be discussed, which we can call the Conserv- ing Effect, depends upon the fact that repetition strengthens memory representations for linguistic forms and makes them more accessible. Accessibility in this sense re- fers to the fact that in experiments where subjects are asked to say whether a string of letters or sounds is a word of their language, they respond much more quickly to high-frequency words than to low-frequency words. This greater accessibility sug- gests that each token of use strengthens the memory representation for a word or phrase (Bybee 1985; this volume, chapters 2, 3, 4, 5, and 8). The strength of repre- sentation of higher frequency forms explains why they resist reformation on the basis of analogy with other forms. For instance, for English irregular verbs, there is a gen- eral trend diachronically toward regularization, a trend also witnessed in child lan- guage development. However, the higher frequency verbs resist this trend; thus keep, sleep, weep, leap, and creep and other verbs of this shape acquired irregular past forms when the vowel was shortened in early Middle English, giving kept, slept, wept, leapt, and crept. Only the lower frequency verbs of this class have subsequently developed regularized pasts weeped, leaped, and creeped (still used alongside the irregulars). The mechanism behind this type of change (analogical reformation) is that a new past form is created by accessing the base/present form and adding the suffix –ed (in this case its allomorph [t]) to it. For those verbs of high frequency, the greater acces- sibility of the irregular past makes such a reformation unlikely. For this reason, the lower frequency paradigms tend to regularize before the higher frequency paradigms. Another aspect of the Conserving Effect is the fact that within a paradigm it is the higher frequency form that serves as the basis for the reformation just illustrated (Ma¤czak 1980; Bybee 1985; this volume, chapters 3 and 4). Thus in our example, the base/present form weep, which is more frequent than wept, survives and serves as the base upon which the new past is created, in contrast to the hypothetical situation in which a new base is formed from wept, creating *wep as the new base. The latter result would also be a regular paradigm, but this type of change is rare (see Tiersma 1982). Thus Ma¤czak (1980) has written that the most frequent form(s) of a paradigm are the most likely to resist change and to serve as the basis for change in other forms. The Conserving Effect is also observable in morphosyntactic constructions (this volume, chapters 12 and 16). For instance, of the two forms of negation exemplified by I know nothing about it (negative incorporation) and I don’t know anything about it (not negation), the former is older and more conservative. Tottie (1991) shows that where the two constructions are interchangeable, the older construction (negative incorporation) is used in certain high-frequency constructions, such as existential constructions, constructions with possessive have and copular be. That frequency is
- 24. INTRODUCTION 11 important to the retention of this construction is shown by the fact that it also occurs commonly with certain high-frequency verbs, such as know, do, give, and make. The Conserving Effect also operates in the formation of new constructions, such as those involving the English auxiliary: certain properties of the auxiliaries, such as inverting with the subject, being followed by negation, and taking an infinitive with- out to are actually ancient grammatical properties that applied to all verbs. After some changes that occurred in the fifteenth through the seventeenth century, only the most frequent verbs (the modals, progressive be, perfect have, copular be, and possessive have) retained these characteristics (see this volume, chapters 12 and 16, and Bybee forthcoming). Another interesting phenomenon is the interaction of phonological alternations with morphosyntactic constructions such as consonant liaison in French (Bybee 2001b; this volume, chapter 13). Obligatory liaison occurs primarily in grammatical morphemes (determiners, clitic pronouns, verb endings, plural on adjectives), all of which are very frequent. A corpus study by Ågren (1973) demonstrates that optional liaison is highly influenced by frequency of use, such that even forms of the same verb maintain liaison to a greater or lesser extent depending upon the frequency with which they occur in the liaison environment. While this volume is only suggestive of the ways that morphosyntactic construc- tions might be affected by frequency, the articles contained here point to important areas for future research in usage-based theory. As we will see below, all of the fre- quency effects discussed here are highly relevant for the understanding of the cogni- tive representation of constructions. 2.2. Token frequency: The Reducing Effect It is a common observation that oft-repeated phrases, such as greetings (God be with you > goodbye, how are you > hi) and titles, tend to reduce phonetically. The same observation applies to much-used grammatical items, such as auxiliaries, modals, negatives, and pronouns. As mentioned above, it has recently been fully documented that reductive sound change applies probabilistically across all frequency levels, affecting high-frequency items more quickly and radically than low-frequency items (this volume, chapters 2, 9, and 11). The reason for this trend is that repetition of neuromotor sequences leads to greater overlap and reduction of the component ar- ticulatory gestures. As articulation becomes more efficient, the output appears more and more to have been affected by assimilation and reduction. What is the cause of such reduction? My view, as presented in the chapters of this book, is that phonetic reduction is directly tied to neuromotor processing: re- peated sequences of neuromotor commands and actions tend to be processed as single units; at the same time, repeated sequences tend to become more efficient by the increased overlap and reduction of the gestures involved (Browman and Goldstein 1992; Haiman 1994; Mowrey and Pagliuca 1995; Boyland 1997). This domain-general process is responsible both for the fact that general reductive sound change occurs earlier in high-frequency words and that special reduction occurs in very high-frequency words and phrases. Thus frequency of use is one factor in ex- plaining sound change.
- 25. 12 BACKGROUND AND CURRENT CONTEXT As mentioned before, the lexical regularity of sound change is the cornerstone of the comparative method. It was argued for vociferously by the Neogrammarians of the late nineteenth century. Over the years, there have been many detractors from this strict position. Contemporaneously with the Neogrammarians, Schuchardt (1885; see Vennemann and Wilbur 1972) documented his observations that sound changes in progress saw frequent words scurrying ahead, while the less frequent ones dragged behind. Many philologists maintained open minds about regularity; the famous his- torian of the Spanish language Ramón Menéndez-Pidal (1950) observed that in sound change words were like leaves floating on running water: some were caught by the current and rushed ahead, while others circled in eddies and lagged behind, yet the general flow was unmistakably directional. In my first investigation of frequency effects, I attempted to document the role of high-token frequency in sound change (see chapter 2 of this collection). Taking off from this point, Phillips (1984, 2001, and other works) has studied many sound changes of various types in the light of frequency of use. The effect of frequency can be seen in many cases of more or less stable variation, as well as in changes that end up affecting all the words of a lan- guage, and turn out regular. In the latter case, the course of development of the change involves the gradual spread through the lexicon known as lexical diffusion. The idea that high-frequency words grow shorter brings to mind the work of George Kingsley Zipf, which, however, proposes a very different explanation for the correlation of high frequency with shorter expression. In The Psychobiology of Language, published in 1935, Zipf documents across a number of languages the ten- dency for high-frequency words to be shorter than lower frequency words. Zipf takes the frequency distribution of short and long words to represent an equilibrium that languages strive to maintain. The mechanism he sees as responsible for this equilib- rium is the process of “clipping” or abbreviation by which laboratory becomes lab, and so on. Ma¤czak’s theory of irregular sound change due to frequency is similar to Zipf’s: There is a synchronic law according to which the linguistic elements which are more often used are smaller than those which are less often used . . . If a word or mor- pheme becomes too short in relation to its frequency, it is replaced by a longer one. But if a linguistic element (i.e., a morpheme, word or group of words) becomes too long in relation to its frequency, it must be shortened, and then there are two pos- sibilities: either mechanical shortening (Universität > Uni) or an irregular sound change due to frequency. (1978: 309–310) My own view of Zipf’s finding, as represented in the chapters of this volume, is that high-frequency words undergo reductive changes at a faster rate than low- frequency words. While “clipping” certainly occurs and produces this result as well, the major mechanism is gradual phonetic reduction brought about by the reduction and overlapping of articulatory gestures. In phrases of extreme high frequency this reduction may appear “irregular,” as Ma¤czak has observed, but my view is that such reduction follows the regular phonetic patterns of the language but carried to a more extreme point than in words of lesser frequency. Thus when I don’t know reduces, the [d] undergoes flapping and even deletion, the nasal consonant deletes, and the [o] vowel may be reduced to schwa. All of these phonetic changes affect other words
- 26. INTRODUCTION 13 of English as well. They are not “irregular”; they are just extreme instances of other- wise regular phonetic processes. A related effect of the repetition of sequences of units is chunking, or the repre- sentation of the sequence as a single unit at a higher level (Haiman 1994; Boyland 1997; this volume, chapters 14 and 15). As I don’t know is repeated frequently, it ceases to be produced by the concatenation of three items (as don’t has long ago coalesced) and comes to be accessed as a single unit (this volume, chapter 14). This packaging facilitates both access and further phonetic reduction in the sequence. Chunking is particularly obvious in high-frequency word sequences, but Jurafsky et al. (2001) and other studies (Krug 1998; Brown 2004; Alba 2005) have shown that phonetic reduction between words is highly dependent upon how frequently the words co-occur, suggesting the probabilistic relations among words at all levels of frequency. Phonetic reduction of high-frequency combinations of words is also one of the processes of change that make up the phenomenon of grammaticization (Bybee, Perkins, and Pagliuca 1994; this volume, chapter 16). As the words or phrases that enter into the grammaticization process increase in frequency, they also undergo extreme versions of phonetic reduction, as, for instance, when be going to reduces to [gənə] or I’m gonna reduces to [aimənə]. Other concomitant processes that lead to this reduction are the loss of stress on grammaticizing items and their increasing use in redundant contexts. The Reducing Effect and the Conserving Effect at first blush seem contradictory (see this volume, chapter 2) since in one case high frequency encourages change and in the other high frequency discourages change. The contradiction is only apparent, however, because very different types of change are involved. On the one hand, the strengthening of memories makes complex units resist change by reformation on anal- ogy with productive patterns, while on the other hand, the greater fluency and reduc- tion of repeated units is a factor in phonetic change and semantic change. The two types of change involved here are very different. This is the point of chapter 2. (See also chapter 16 for these effects in grammaticization of morphosyntax.) 2.3. Token frequency: Autonomy In previous work, I have identified two effects of token frequency, the two just dis- cussed. However, I think it is worthwhile to take “autonomy” to be a third effect of token frequency even though it may be also considered an extreme instance of the Conserving Effect. In Bybee and Brewer 1980 (this volume, chapter 3) we introduce the notion of autonomy, as it applies to words (although it can also be extended to phrases), by saying it is “the extent to which a word is likely to be represented in the speaker’s lexicon as a whole and separate unit” (see page 50 of this volume). Au- tonomy is thus probabilistic, varying for each word or phrase. Three factors are said to influence the degree of autonomy: semantic simplicity, morphophonemic regu- larity, and word frequency. Autonomy is based on MacWhinney’s (1978) notion of lexical strength, which is a reflection of frequency in experience and discussed more thoroughly in Bybee (1985). Highly autonomous words have weaker connections to other related words—either words of the same paradigm or words of the same lexi- cal class. The idea behind autonomy is that when words (and phrases) are highly
- 27. 14 BACKGROUND AND CURRENT CONTEXT frequent they can be accessed independently of related items and are thus not as in- terconnected in the network. Autonomy can be observed to be operating in grammaticization. A common result of grammaticization is that the lexical item within a grammaticizing construction splits off from its origins: thus possessive have and perfect have in English are no longer the “same” item (Heine and Reh 1984; Hopper 1991). Besides the semantics, evi- dence for this is the fact that in American English possessive have takes the auxil- iary do in questions and negatives and perfect have does not (Do you have a driver’s license? vs. Have you seen Betty?). Thus the construction exemplified by have seen, have gone, have eaten is independent of other uses of have. Similarly, be going to, discussed in chapter 16, is autonomous from other uses of go as well as from other instances of the purpose construction that gave rise to it. Bybee and Brewer (1980; this volume, chapter 3) argue that autonomy in inflec- tional morphology leads to resistance to change and it also allows forms to serve as the basis of reformation within a paradigm. The high frequency of autonomous forms weakens their connections to other forms, leading in extreme cases to suppletion, as when the Past form went split off from wend and became the past of go (Bybee 1985). Such cases of suppletion occur only in the highest frequency paradigms of a language. In derivational morphology a related tendency is for high-frequency derived forms to split off semantically from their bases (Bybee 1985). Hay (2001) has re- fined this hypothesis by postulating that derived forms move away from their bases semantically if they are more frequent than their bases. Bybee and Moder (1983; this volume, chapter 6) also argue that high-frequency inflected forms do not contribute to the productivity of morphological patterns. Thus an irregular verb such as begin, began, begun is of extreme high frequency and could well be autonomous from the semiproductive class to which it originally belonged. This could explain why the class it belonged to extended by adding verbs that end in velars (such as string, strung; strike, struck) rather than adding more members end- ing in alveolars. Moder (1992) tests this hypothesis experimentally, demonstrating that high-frequency class members are less successful at priming nonce responses than lower frequency class members. The same principle has been shown to apply in phonotactic studies. Bailey and Hahn (2001) show that the type frequency (see section 2.4) of phonotactic patterns partially determines how acceptable native speakers will find them, but that patterns with very high token frequency and very low token frequency have less of an effect on acceptability judgments. 2.4. Type frequency Type frequency is a property of patterns or constructions and refers to the number of distinct items that can occur in the open slot of a construction or the number of items that exemplify a pattern, such as a phonotactic sequence. Type frequency is a major factor determining the degree of productivity of a construction (Guillaume 1927/1973; MacWhinney 1978; Bybee 1985; this volume, chapters 4–8). Constructions that ap- ply to a high number of distinct items also tend to be highly applicable to new items. In determining productivity, however, factors other than type frequency must also
- 28. INTRODUCTION 15 be taken into account: often the member items that occur with a construction must also belong to certain phonological or semantic categories. The verbs of the string, strung class must end in a nasal or a velar (chapters 5 and 6); the adjectives that can be used in the construction [X drives me (or someone) ADJ], (as in it drives me mad, it drives me crazy) must suggest some degree of insanity, either literally or figura- tively (Boas 2003). As pointed out in the previous section, not all members contrib- ute equally to productivity: extreme high-frequency members of classes may not contribute at all, due to their autonomy. The contribution of type frequency to productivity is the fact that when a con- struction is experienced with different items occupying a position, it enables the parsing of the construction. If happiness is learned by someone who knows no re- lated words, there is no way to infer that it has two morphemes. If happy is also learned, then the learner could hypothesize that –ness is a suffix, but only if it occurs on other adjectives would its status as a suffix become established. Thus a certain degree of type frequency is needed to uncover the structure of words and phrases. In addition, a higher type frequency also gives a construction a stronger representation, making it more available or accessible for novel uses. In addition, other factors important for productivity have been identified by Jennifer Hay and colleagues. In particular, Hay (2002) and Hay and Baayen (2002) present evidence that it is the type frequency of words within which the affix can be parsed out that determines productivity, not just the type frequency of the affix. The “parsibility” of a word depends on both phonological and semantic factors. 2.5. Frequency and category formation The formation of grammar is in large part an issue of categorization. Every construc- tion is characterized by the categories that fill its various slots. Every phoneme con- stitutes a category, often made complex by the existence of allophones. Semantic notions such as past or future tense also constitute categories, again made complex by meaning changes in specific contexts. Even the contexts of use of particular con- structions constitute categories. Both type and token frequency are important to cate- gory formation. The properties of the types included in a category establish its boundaries while the number of types relates to the degree of productivity of the construction refer- ring to the category. In research into exemplar models (in which the category con- sists of the experienced exemplars), token frequency can be seen to influence the perception of the center of the category, as well as its boundaries (Nosofsky 1988). In phonetic categorization, high-frequency exemplars tend to be maintained while low-frequency ones are marginalized and lost (this volume, chapter 10). In semantic categorization, a similar phenomenon occurs; in a corpus and experimental study of the pairing of verbs that mean ‘become’ with adjectives in Spanish, it was found that the high-frequency pairs served as the center of some of the most pro- ductive categories (Bybee and Eddington 2006). Similarly, Casenhiser and Gold- berg (2005) show that children and adults learn a new construction faster if they are exposed to one higher frequency token as well as several types that exemplify the construction.
- 29. 16 BACKGROUND AND CURRENT CONTEXT 3. How high, how low? Throughout this introduction, I have used the terms high and low frequency in a fairly vague way. In the chapters collected here frequency is counted with a variety of cutoff points, depending upon the goals of the particular study. In addition, different sources for frequency information are consulted. There is no one method for doing frequency research. When I began to study frequency effects, the available counts were almost all based on written language. Fortunately, now there are large databases of spoken language as well as software that makes extracting frequency information from such databases fairly straightforward. Thus now it is possible to choose a database that is relevant to the particular hypotheses being tested. When one is studying token frequency, there is an inherent problem in deter- mining the point at which high should be distinguished from low. The problem is that there will be many tokens of high-frequency types and many types represented among the low-frequency tokens. Thus if one draws the line so that half the tokens are high frequency and half are low, the high-frequency group may have very few types in it. Conversely, if one puts half the types in the high group and half in the low group, the number of tokens in the high group will vastly outnumber those in the low group. Most studies take a compromise between these two position by looking for a natural gap in the frequency ranks that puts about 30 to 50% of the tokens in one group and 50 to 70% of the tokens in the other. For the research I have done, the strategy of including more than two frequency groups—say high-, mid-, and low-frequency groups—has not improved the descrip- tion of the data. But the possibility certainly remains that items in the medium range have their own properties and effects. Certainly it has been useful to refer to items of extreme high frequency, as these are most likely grammaticizing or exhibiting autonomy in one way or another. However, it has also been found that for types of phonetic re- duction that are quite general (e.g., final [t] or [d] deletion in English) a continuous effect throughout the frequency levels can be observed (Gregory et al. 1999). It is important not to overlook the fact that even a small number of repetitions in the speaker’s experience has a cognitive effect. As pointed out in chapter 13, certain English words that occur only in fixed expressions, such as dint, bated, and hale in the expressions by dint of, with bated breath, and hale and hearty, are actually of rather low frequency (not listed at all among the 1 million words of Francis and Kucera 1982). Yet these expressions are part of the knowledge of English speakers. The pervasive use of prefabricated word sequences, which form a major part of any dis- course, points to the same conclusion. Prefabs are word sequences such as promi- nent role, beyond repair, to need help, which are transparent semantically but used conventionally. Some counts show that more than half of the word choices in a spo- ken or written discourse are determined by membership in a prefab (Pawley and Syder 1983; Erman and Warren 2000; Wray 2002). Prefabs are not especially high in fre- quency, and yet it is clearly their repetition that has given them their social status of conventionalized and their cognitive status of easily accessed routines. In addition, certain phrases that are not particularly frequent, though conventionalized for par- ticular situations, such as who goes there? and how goes it?, are conservative in using main verb inversion rather than do. We also find conservative behavior in phrases
- 30. INTRODUCTION 17 such as far be it from me, which includes a subjunctive form and an archaic word order. Perhaps related to these lower frequency repetitions, subordinate clauses often exhibit conservative behavior (Bybee 2001a). In addition, constructions can take on special pragmatic and semantic values through repetition despite not being of especially high frequency. The special con- notations of the construction discussed by Kay and Fillmore (1999), “what is X doing Y” of incongruity or disapproval, as in What’s that box doing up there?, have to be acquired by repetition in the appropriate contexts. A few repetitions in the relevant context allow a new construction to be formed (Johnson 1997). I conclude, then, that repetition is necessary for the formation of grammar (Haiman 1994), but that differ- ent levels of repetition have different effects (Bybee forthcoming). It is important to bear in mind, as always, that frequency interacts with other factors, such as phonological and semantic similarity, categorization, and semantic/ pragmatic change. It is often difficult to discern which factors are the most impor- tant in determining linguistic behavior. 4. Frequency of what? In order for frequency counts to be useful the researcher must know what to count. Often it is a more specific and longer string that must be counted rather than a more general and shorter one. For instance, the frequency of don’t in itself does not ac- count for its reduction, because it only reduces in certain phrases. Thus it is the fre- quency of phrases such as I don’t know, I don’t think, I don’t have, and so on, that figures in its most extreme reduction. Whether the reduction can spread from these contexts into new ones is a question that remains to be answered. Similarly, in using type frequency to study productivity, the question arises as to how similar or how specific the pattern is that counts toward productivity. For instance, in Spanish the alternations in the verbs tener/tengo, ‘to have, 1s have’, poner/ pongo ‘to put, 1s put’, do not extend to nonce verbs such as *roner (this volume, chapter 4). However, the derivatives of tener and poner all use the same alternations in their conjugations: detener/detengo ‘to detain’, proponer/propongo ‘to propose’. Thus while the –n/ngo alternation does not seem productive, a longer, more specific string tener/tengo, poner/pongo appears to be. Thus the linguist’s goal of always finding the most general patterns may not serve to reveal the significant patterns in the speaker’s mental grammar. However, sometimes the relevant comparison class is broader than the phenom- enon under study. Diessel and Tomasello (2005) argue that the ease of acquisition of relative clauses does not always depend upon the frequency of the type of relative clause but also on the similarity of the word order of that clause to a simple clause. 5. Is frequency a cause or an effect? Over the years many linguists have expressed the opinion that what is really inter- esting is the question of why certain words, expressions, or sounds are frequent.
- 31. 18 BACKGROUND AND CURRENT CONTEXT My claim is that no matter the source of the frequency, the cognitive effects shape grammar. However, it is worthwhile to investigate the question of why certain items are frequent, even though we will find that items become frequent for many dif- ferent reasons. First, there is the question of what people want to talk about. The answer is, “Themselves!” One of the most used words in English is I, and other first person pronouns (object and possessive) are also very frequent (Scheibman 2002). Second, in addition to the content, the way speakers structure their discourse leads some elements to be more frequent than others. These discourse patterns are conventionalized into grammar, as summed up in DuBois’s slogan: “grammars code best what speakers do most.” One could question the use of the word best here, but presumably DuBois means that through repetition, reduction, and conventionaliza- tion grammars arise that provide economical means of expressing frequent discourse patterns. His example is the development of ergative cross-referencing out of the tendency to pronominalize or omit the agent and use full noun phrases for the absolutives (the subject of the intransitive and the object of the transitive verb). There is also a relationship between frequency and generality or flexibility of meaning. Grammaticized elements, such as past or future markers (will, be going to), have very general and abstract meaning that can be applied to almost any verb type. Therefore, these markers tend to be of very high frequency. Polysemous items also can have a higher frequency. For instance, a phrase such as I don’t know, which has its literal meaning in addition to a pragmatic use as a conversational hedge or turn organizer, will have a higher frequency than a phrase with only literal meaning (Scheibman 2000). Gentner (1981) demonstrates that verbs are more flexible in their interpretation and thus more polysemous than nouns are. This relative flexibility of meaning gives rise to the different frequency distributions of nouns versus verbs in English: there are more high-frequency verbs than nouns and more low-frequency nouns than verbs. Thus the answer to the question of whether frequency is a cause or an effect is complex. On the one hand, frequency is just a tally, a pattern observable in texts, which is of course an effect. On the other hand, frequency or repetition of experi- ences has an impact on cognitive representations and in this way becomes a cause for the effects discussed in this book. 6. The parts of this volume This book is divided into four parts. The second chapter of part I consists of an article that discusses two effects of token frequency; it represents my first foray into this territory and lays out the premises for much of the later work. Part II is made up of six chapters that discuss the way in which token and type frequency modulate morphological structure. Part III has three chapters that explore further how token frequency affects the spread of sound change. The five chapters of part IV apply the findings for phonetics and morphology to the level of morphosyntax. The organization of the book partly reflects chronology: part I’s chapter 2 was originally published in 1976, the articles of part II appeared between 1980 and 1995,
- 32. INTRODUCTION 19 the articles of part III between 2000 and 2002, the articles of part IV between 1997 and 2003. Each part of the book begins with commentary on the chapters that follow. References Ågren, John. 1973. Etude sur quelques liasons facultatives dans le français de conversation radiophonique: Fréquence et facteurs. Uppsala: Acta Universitatis Upsaliensis. Studia Romanica Upsaliensa 10. Alba, Matt. 2005. Hiatus resolution between words in New Mexican Spanish: A usage-based account. Ph.D. diss., University of New Mexico. Bailey, Todd M., and Ulrike Hahn. 2001. Determinants of wordlikeness: Phonotactics or lexical neighborhoods? Journal of Memory and Language 44, 568–591. Barlow, Michael, and Suzanne Kemmer. 2000. Usage-based models of language. Stanford, Calif.: CSLI. Boas, Hans C. 2003. A constructional approach to resultatives. Stanford, Calif.: CSLI. Bod, Rens, Jennifer Hay, and Stefanie Jannedy (eds.). 2003. Probabilistic linguistics. Cam- bridge, Mass.: MIT Press. Boyland, Joyce Tang. 1997. Morphosyntactic change in progress: A psycholinguistic approach. Ph.D. diss., University of California at Berkeley. Browman, Catherine P., and Louis M. Goldstein. 1992. Articulatory phonology: An over- view. Phonetica 49, 155–180. Brown, Esther. 2004. Reduction of syllable-initial /s/ in the Spanish of New Mexico and Southern Colorado: A usage-based approach. Ph.D. diss., University of New Mexico. Bybee, Joan L. 1985. Morphology: A study of the relation between meaning and form. Amsterdam: John Benjamins. ———. 1995. Regular morphology and the lexicon. Language and Cognitive Processes 10, 425–455. ———. 2001a. Main clauses are innovative, subordinate clauses are conservative: Conse- quences for the nature of constructions. In Complex sentences in grammar and discourse: Essays in honor of Sandra A. Thompson, ed. Joan L. Bybee and Michael Noonan, 1–17. Amsterdam: John Benjamins. ———. 2001b. Phonology and language use. Cambridge: Cambridge University Press. ———. forthcoming. From usage to grammar: The mind’s response to repetition. Language 82. Bybee, Joan L., and Mary Alexander Brewer. 1980. Explanation in morphophonemics: Changes in Provençal and Spanish preterite forms. Lingua 52, 201–242. Bybee, Joan L., and David Eddington. 2006. A usage-based exemplar model approach to Spanish verbs of “becoming.” Language 82, 323–355. Bybee, Joan L., and James McClelland. 2005. Alternatives to the combinatorial paradigm of linguistic theory based on domain general principles of human cognition. In The role of linguistics in cognitive science, ed. Nancy A. Ritter. Special Issue of The Linguistic Re- view 22(2–4), 381–410. Bybee, Joan, Revere D. Perkins, and William Pagliuca. 1994. The evolution of grammar: Tense, aspect and modality in the languages of the world. Chicago: University of Chicago Press. Casenhiser, Devin, and Adele Goldberg. 2005. Fast mapping between a phrasal form and meaning. Paper presented at the International Association for the Study of Child Lan- guage Development. Berlin. Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, Mass.: MIT Press. Daçbrowska, Ewa, and Elena Lieven. 2005. Towards a lexically specific grammar of children’s
- 33. 20 BACKGROUND AND CURRENT CONTEXT question constructions. Paper presented at the International Association for the Study of Child Language Development. Berlin. Diessel, Holger, and Michael Tomasello. 2005. A new look at the acquisition of relative clauses. Language 81.4, 882–906. DuBois, John. 1987. The discourse basis of ergativity. Language 63, 805–855. Elman, Jeffrey L. 1990. Finding structure in time. Cognitive Science 14, 179–211. Erman, Britt, and Beatrice Warren. 2000. The idiom principle and the open choice principle. Text 20, 29–62. Francis, W. Nelson, and Henry Kucera. 1982. Frequency analysis of English usage. Boston: Houghton Mifflin. Gentner, Dedre. 1981. Some interesting differences between nouns and verbs. Cognition and Brain Theory 4, 161–178. Givón, Talmy. 1979. On understanding grammar. New York: Academic Press. Goldberg, Adele. 1995. Constructions: A construction grammar approach to argument struc- ture. Chicago: University of Chicago Press. Greenberg, Joseph. 1966. Language universals: With special reference to feature hierarchies. The Hague: Mouton. ———. 1978. How does a language acquire gender markers? In Universals of Human Lan- guage, vol. 3: Word Structure, ed. Joseph H. Greenberg, Charles A. Ferguson, and Edith A. Moravcsik, 47–82. Stanford, Calif.: Stanford University Press. Gregory, Michelle L., William D. Raymond, Alan Bell, Eric Fosler-Lussier, and Daniel Jurafsky. 1999. The effects of collocational strength and contextual predictability in lexical production. CLS 99, 151–166. Guillaume, P. 1927/1973. The development of formal elements in the child’s speech. In Studies of child language development, ed. C. A. Ferguson and Dan I. Slobin, 240–251. New York: Holt, Rinehart and Winston. Haiman, John. 1994. Ritualization and the development of language. In Perspectives on grammaticalization, ed. William Pagliuca, 3–28. Amsterdam: John Benjamins. Hay, Jennifer. 2001. Lexical frequency in morphology: Is everything relative? Linguistics 39(6), 1041–1070. ———. 2002. From speech perception to morphology: Affix ordering revisited. Language 78, 527–555. Hay, Jennifer, and Harald Baayen. 2002. Parsing and productivity. In Yearbook of morphol- ogy 2001, ed. Geert Booij and Jaap van Marle, 203–235. Dordrecht: Klewer Academic. Heine, Bernd, and Mechthild Reh. 1984. Grammaticalization and reanalysis in African lan- guages. Hamburg: Helmut Buske Verlag. Holland, John H. 1998. Emergence: From chaos to order. New York: Basic Books. Hopper, Paul J. 1987. Emergent grammar. BLS 13, 139–157. ———. 1991. On some principles of grammaticization. In Approaches to grammaticali- zation, vol. 1, ed. Elizabeth C. Traugott and Bernd Heine, 17–35. Amsterdam: John Benjamins. Hopper, Paul, and Sandra Thompson. 1980. Transitivity in grammar and discourse. Language 56, 251–299. ———. 1984. The discourse basis for lexical categories in universal grammar. Language 60, 703–752. Johnson, Christopher. 1997. The acquisition of the “What’s X doing Y?” construction. Pro- ceedings of the 21st annual Boston University Conference on Language Development 2, 343–353. Somerville, Mass.: Cascadilla Press. Jurafsky, Daniel, Alan Bell, Michelle Gregory, and William D. Raymond. 2001. Probabilis- tic relations between words: Evidence from reduction in lexical production. In Frequency
- 34. INTRODUCTION 21 and the emergence of linguistic structure, ed. Joan L. Bybee and Paul Hopper, 229–254. Amsterdam: John Benjamins. Kay, Paul, and Charles Fillmore. 1999. Grammatical constructions and linguistic generaliza- tions: The What’s X doing Y? construction. Language 75, 1–33. Krug, Manfred. 1998. String frequency: A cognitive motivating factor in coalescence, language processing and linguistic change. Journal of English Linguistics 26, 286–320. Langacker, Ronald. 1987. Foundations of cognitive grammar, vol. 1: Theoretical prerequi- sites. Stanford, Calif.: Stanford University Press. ———. 2000. A dynamic usage-based model. In Usage-based models of language, ed. Michael Barlow and Suzanne Kemmer, 1–63. Stanford, Calif.: CSLI. Li, Charles N., ed. 1975. Word order and word order change. Austin: University of Texas Press. ———, ed. 1977. Mechanisms of syntactic change. Austin: University of Texas Press. Lieven, Elena V. M., Julian M. Pine, and Gillian Baldwin. 1997. Lexically-based learning and early grammatical development. Journal of Child Language 24, 187–220. Lindblom, Björn, Peter MacNeilage, and Michael Studdert-Kennedy. 1984. Self-organizing processes and the explanation of phonological universals. In Explanations for language universals, ed. Brian Butterworth, Bernard Comrie, and Östen Dahl, 181–203. New York: Mouton. MacWhinney, Brian. 1978. The acquisition of morphophonology. Monographs of the Soci- ety for Research in Child Development, no. 174, vol. 43. Ma¤czak, Witold. 1978. Irregular sound change due to frequency in German. In Recent devel- opments in historical phonology, ed. Jacek Fisiak, 309–319. The Hague: Mouton. ———. 1980. Laws of analogy. Historical morphology, ed. Jacek Fisiak, 283–288. Berlin: Mouton. Medin, Douglas L., and M. M. Schaffer. 1978. Context theory of classification learning. Psychological Review 85, 207–238. Menéndez-Pidal, Ramón. (1950). Orígenes del español: Estado lingüístico de la península ibérica hasta el siglo XI. Madrid: Espasa-Calpe. Moder, Carol Lynn. 1992. Productivity and categorization in morphological classes. Ph.D. diss., SUNY at Buffalo. Mowrey, Richard, and William Pagliuca. 1995. The reductive character of articulatory evo- lution. Rivista di Linguistica 7(1), 37–124. Nosofsky, R. M. 1988. Similarity, frequency and category representation. Journal of Experi- mental Psychology: Learning, Memory and Cognition 14:54–65. Pawley, Andrew, and Frances Hodgetts Syder. 1983. Two puzzles for linguistic theory: Nativelike selection and nativelike fluency. In Language and communication, ed. Jack C. Richards and Richad W. Schmidt, 191–226. Longman: London. Phillips, Betty S. 1984. Word frequency and the actuation of sound change. Language 60, 320–342. ———. 2001. Lexical diffusion, lexical frequency, and lexical analysis. In Frequency and the emergence of linguistic structure, ed. Joan L. Bybee and Paul Hopper, 123–126. Amsterdam: John Benjamins. Posner, Michael I., and Steven W. Keele. 1968. On the genesis of abstract ideas. Journal of Experimental Psychology 77, 353–363. Rumelhart, David E., and James L. McClelland. 1986. On learning the past tenses of English verbs: Implicit rules or parallel distributed processing? In Parallel distributed process- ing: Explorations in the microstructure of cognition, ed. James L. McClelland, David E. Rumelhart, and the PDP Research Group, 216–271. Cambridge, Mass.: MIT Press. Saussure, Ferdinand de. 1916. Cours de linguistique générale. Paris: Payot. Savage, Ceri, Elena V. M. Lieven, Anna Theakston, and Michael Tomasello. 2003. Testing
- 35. 22 BACKGROUND AND CURRENT CONTEXT the abstractness of young children’s linguistic representations: Lexical and structural priming of syntactic constructions? Developmental Science 6, 557–567. Scheibman, Joanne. 2000. I dunno but: A usage-based account of the phonological reduction of don’t in American English conversation. Journal of Pragmatics 32(1), 105–124. ———. 2002. Point of view and grammar: Structural patterns of subjectivity in American- English conversation. Amsterdam: John Benjamins. Schuchardt, Hugo. 1885. Über die Lautgesetze: Gegen die Junggrammatiker. Berlin: Robert Oppenheimer. Sinclair, John. 1991. Corpus, concordance, collocation. Oxford: Oxford University Press. Tiersma, Peter. 1982. Local and general markedness. Language 58, 832–849. Tomasello, Michael. 2003. Constructing a language. Cambridge, Mass.: Harvard University Press. Tottie, Gunnel. 1991. Lexical diffusion in syntactic change: Frequency as a determinant of linguistic conservatism in the development of negation in English. In Historical English syntax, ed. Dieter Kastovsky, 439–467. Berlin: Mouton de Gruyter. Vennemann, Theo, and Terence Wilbur. 1972. Schuchardt, the Neogrammarians and the transformational theory of phonological change. Frankfurt: Athenäum. Wray, Alison. 2002. Formulaic language and the lexicon. Cambridge: Cambridge Univer- sity Press. Zipf, George K. 1935. The psychobiology of language. Boston: Houghton Mifflin.
- 36. Chapter 2 Word Frequency in Lexical Diffusion and the Source of Morphophonological Change 23 Two ideas that have been around for almost a century need to be reexamined in the light of our current theories of linguistic change. These ideas involve the rela- tionship between word frequency and morphophonological change. I will divide morphophonological change in the traditional way into sound change, which I view as being largely, perhaps entirely, phonetically motivated, and analogy, which I con- sider to be primarily conceptually motivated. The particular point I want to discuss is that the relation between sound change and word frequency is just the reverse of the relation between analogy and word frequency. That is, as Schuchardt observed in 1885, sound changes affect the most frequent lexical items first (1885 [1972]). And Paul, around 1886, observed that analogical leveling tends to affect infrequent items first (1890 [1972]). Or, stated differently, infrequent items are the most resis- tant to phonetically motivated change, while frequent items are the most resistant to conceptually motivated change. I have gathered some data that document these two tendencies, and I will present these data before speculating on the meaning of this difference between the types of change for theories of the source of morphophono- logical change. In particular I will speculate on the role of the child versus the adult speaker in the initiation of linguistic change. 1. Phonetically conditioned change First consider phonetically motivated change, sound change. It has often been ob- served that reductive changes affect frequent words and phrases, thus God be with
- 37. 24 BACKGROUND AND CURRENT CONTEXT you becomes good-bye, vuestra merced becomes usted. In English, auxiliaries and object pronouns contract. Leslau (1969) has observed similar reductions of very fre- quent verbs in Ethiopian languages. Furthermore, Schuchardt observed that general sound changes affect very frequent items before they affect infrequent items. Some recent investigations support Schuchardt’s hypothesis. For example, Fidelholtz (1975) has shown that vowel reduction in initial syllables in English correlates significantly with frequency: the more frequent the word is, the more likely it is to have a reduced vowel. I have gathered some data on schwa deletion in English that also support Schuchardt’s hypothesis. Schwa deletion is in an extremely variable state in Ameri- can English. Unstressed schwas tend to be deleted generally except for a few pho- nologically constrained environments that I will mention later. Yet there are some words for which a schwa-less pronunciation is clearly the norm and others for which a schwa-ful pronunciation is preferred; words of this sort troubled Zwicky (1972): compare nursery and cursory, scenery and chicanery, celery and artillery,1 memory and armory. I was interested in finding out if the frequency of a word had any correlation to the progress of schwa deletion with regard to that word, so I devised a test, which I tried out on a few subjects. To get a corpus of words with phonologically similar environments, I extracted from Dolby and Resnikoff, The English Word Speculum, (1967) all words ending in stressed V Co V r y.2 I checked the frequency of these words in Kucera and Francis (1967).3 I took out the words not listed in the Kucera and Francis frequency count and was left with a list of 112 words. To find out which of these words undergo schwa deletion, I could not consult a dictionary as Fidelholtz did for his vowel reduction data, because dictionary pronunciations are extremely conservative and usually give a pronunciation with the schwa. The ideal way to find out which words undergo schwa deletion would be to listen to speakers’ pronunciations in spontaneous conversation, but this is not practical, for the amount of data one can gather in this way is far out of proportion to the amount of time it would take. Instead, I asked speakers rather di- rectly how these words are normally pronounced: I gave eight speakers a list of the 112 words and asked them to classify the words into one of three categories: Usually delete, Sometimes delete, and Rarely delete. All of the subjects but one were gradu- ate students in linguistics, so they were able to understand the directions. None of them knew my hypothesis. This was a difficult task, and in some cases the decisions were rather arbitrary. A speaker’s judgment on any one word may be different today from what it was yesterday.4 Since the conclusions are all stated in rather general terms, that is, in terms of the average frequencies of the words in each category, I make no claims about any particular word. It should also be remembered that my test shows only the speaker’s conception of how these words are pronounced; it does not reveal any phonetic reality. Table 2.1 gives the results for each individual speaker. There are four columns; the first three represent the three categories the subjects used to categorize the words. The first number in each column is the average frequency of the words the subject put into this category; the number in parentheses is the number of words in the cate- gory. The fourth column represents adjustments made on the Rarely category. As I suspected, there are certain phonological conditions that systematically block schwa
- 38. WORD FREQUENCY IN LEXICAL DIFFUSION 25 deletion. When words that do not have schwa deletion for phonological reasons are taken out of the Rarely category, as we might expect, the average frequency of words in this category falls to the level that we see in the fourth column. There are three phonological environments that block schwa deletion before r. One is the presence of a preceding flap, derived historically from a t (and possible from a d, but there are so few words in the corpus with -dVry that no conclusions can be drawn).5 The flap appears where a /t/ was preceded by a vowel or /r/, for example, artery, flattery, but not where a consonant precedes /t/, for example, factory, adul- tery, elementary. In the latter forms, deletion may occur, but in the former forms, the eight words with flaps, artery, battery, buttery, flattery, lottery, pottery, rotary, and watery, deletion does not usually occur. Deletion is also constrained if a syllable-initial consonant cluster precedes the schwa. In burglary a deletion would leave a totally unacceptable consonant cluster of glr for syllable-initial position. Similarly, deletion is constrained in library, con- trary, mercury, and penury, although [laybri] and [mkri] are possible pronuncia- tions. Finally, a preceding affricate, as in forgery or butchery, constrains deletion for many speakers. The individual speakers displayed different phonological constraints, so the fourth column was computed separately for each subject. I considered the subject to be exercising a phonological constraint if all or all but one of the words of a phono- logical type appeared in the Rarely category. When this occurred, all the words of that type were removed and the average was recomputed. However, if a subject put two or more words of a phonological type in the Sometimes or Usually category, I did not take the words of this type out of the Rarely category. The lowercase initials following the fourth column indicate which constraints were considered applicable for that subject: f stands for flap, c for cluster, a for affricate. Thus subjects A, B, and 2.1. Average Frequency of Words Categorized for the Application of Schwa-deletion Rarely without Phonologically Usually Sometimes Rarely Constrained Words A 38.75 (16) 13.95 (38) 14.17 (53) 6.40 (f, c, a) B a 23.87 (46) 12.33 (66) 5.81 (f, c, a) C 22.00 (67) 15.54 (24) 15.00 (20) 8.70 (c) D 24.03 (37) 20.44 (33) 18.76 (42) 5.81 (c) E 29.57 (35) 13.90 (11) 12.67 (66) 5.48 (f, c, a) F 10.38 (8) 31.44 (25) 13.11 (79) 11.77 (f, c) 20.75 (4) G 27.24 (45) 21.27(11) 19.04 (49) 7.28 (f, c) H 35.04 (28) 17.5 (10) 11.57 (74) 9.93 (f, c) Average of all subjects 26.72 15.99 6.18 Average frequency of all 112 words: 17.07 aSubject B did not put any words in the Usually category.
- 39. 26 BACKGROUND AND CURRENT CONTEXT E put all flap words, affricate words, and cluster words in the Rarely category; for these subjects all words of these categories were removed to obtain the fourth col- umn average. However, subjects C and D had two or more flap words and two or more affricate words in the Sometimes and Usually categories, so flap and affricate words were averaged in to get the fourth column score. Now if we examine the figures we see that the overall pattern is what Schuchardt’s hypothesis predicts, although there are a few problems of detail, which I will return to. To make the general pattern more apparent, compare the highest and lowest cat- egories for each speaker. The highest category has at worst an average frequency three times greater than the lowest category and at best an average frequency six times greater than the lowest (cf. subject A versus subject F). The major problems in table 2.1 show up in the middle category, where for three subjects the average frequency of the Sometimes category is higher than the average frequency of the Rarely category. I attribute this largely to the difficulty of assigning words to three arbitrary categories. In fact, given the difficulty of the task, it is amaz- ing that the results are so consistent. Notice also that for subjects E and H the Some- times category has so few members that their average is hardly comparable to the other averages. The only other incongruity in table 2.1 occurs in the Usually column for subject F. The average frequency for his Usually category is much lower than his Sometimes category. However, his Usually category contained only eight words, and of these five were words ending in -mentary. It appears that while he is a very conservative deletor in general, for him certain words or formatives are restructured, and -mentary is one of them. If the five -mentary words are counted as a single item the frequency average for this category doubles to 20.75. Still the average for this category is lower than for his Sometimes category, but with the five -mentary words counted as one item there are only four items in this category, which makes the average consider- ably less reliable and not comparable to the other averages. For this reason, in table 2.2 I used subject F’s Sometimes category, where for the other subjects I used the Usu- ally category. I will only mention two other obvious problems with the test. First, the frequency count used is a frequency count based on written texts. Though a variety of types of texts were used, written texts will never reflect spoken language accurately. But the 2.2. High and Low Categories High Low A 38.75 6.40 B 23.87 5.81 C 22.00 8.70 D 24.03 5.81 E 29.57 5.48 F 31.44 11.77 G 27.24 7.28 H 35.04 9.93 Total average 28.99 6.18
- 40. WORD FREQUENCY IN LEXICAL DIFFUSION 27 primary difference in word frequency between written and spoken language will be that frequent words in written language will be even more frequent in spoken lan- guage, while infrequent words in written language will be even more infrequent or totally non-existent in spoken language. Thus the difference in frequency found in written texts will be increased, probably in geometric proportions, in spoken language. This means that the frequency differences displayed in tables 2.1 and 2.2 would be greatly increased if I had used a frequency count of spoken language. Another problem may be the fact that only eight subjects were used. However, the findings are quite consistent, and it seems to me in the case used here, of schwa dele- tion, the hypothesis is really non-controversial. This is because it is a reductive type change and I do not think anyone would deny that reduction in frequent items is to be expected. It would be more interesting (although more difficult) to run more extensive tests on sound changes in progress that are not so obviously reductive, for example, assimilations and vowel shifts. I would venture to guess that all sound changes, even those that we do not consider to be reductive, take place first in frequent words. Let me make clear my interpretation of the status of schwa deletion in English. I claim that schwa deletion, especially poststress schwa deletion, represents a “sound change” in progress. It is potentially totally general, but at present it has not yet reached all the lexical items of English. At present schwa deletion may appear sporadic. Simi- larly, vowel reduction in English appeared sporadic to Fidelholtz, who referred to it as “an exceptional rule” and concluded that “frequent words can do exceptional things.” I do not think that frequent words do “exceptional things”; rather, they are merely ahead of other words in undergoing phonetic change. My view is more like that of Schuchardt, who says “every sound change in some phase is sporadic” (1885 [1972]: 63), except that I would say that every sound change in some phase appears to be sporadic until we know what all the variables are. I will return in section 3 to a discussion of the implications of word frequency as a variable affecting sound change. 2. Analogical leveling Analogical leveling appears to affect less frequent forms; frequent paradigms stub- bornly cling to their suppletion. (I am leaving analogical extension out of the dis- cussion.) While our general experience supports this idea, I wanted to check the hypothesis a little more rigorously before basing any claims on it. In particular I looked at classes of Old English strong verbs that have some members that survive as strong verbs in Modern English and some members that have become weak verbs, that is, that have undergone analogical leveling in Modern English. Using Sweet’s Anglo- Saxon Primer (Davis 1965), I found three appropriate classes of verbs.6 The frequency counts for these verbs (again from Kucera and Francis) are listed in table 2.3. To obtain these frequency counts, I totaled all entries that could be verb forms, for ex- ample, drive, drives, driven, drove, and driving. In some cases, however, the frequency score will include non-verbs that are spelled identically, since the frequency count is done totally on the basis of spelling, disregarding the words’ function in the sentence. The verbs whose frequency score may include non-verbs and thus overrepresent the frequency of the verb, are marked with asterisks.
- 41. 28 BACKGROUND AND CURRENT CONTEXT In the left-hand column are strong verbs that remained strong; in the right-hand column are strong verbs that lost the vocalic alternation. I used all the verbs that Sweet listed under each class, except, of course, the verbs that do not survive in Modern English. The results overwhelmingly bear out the hypothesis; in fact, the frequency differences are so great that I am certain an expanded corpus would produce the same results. Let me comment briefly on a few details: (1) In Class I, shine is not classified, because as an intransitive verb it has, presumably, the past tense and past participle shone; but as a transitive verb, it is weak and has the past tense and past participle form shined. (2) In Class VII, weep and leap are considered to be leveled, despite the lax vowels occurring in wept and leapt, because this lax vowel is a result of an early Middle English vowel shortening (Moore 1968, 67) and is not the same as the ablaut found in the Old English strong verbs. A problem with the results displayed in table 2.3 is that the frequency count used was based on Modern English, but the analogical leveling took place sometime dur- ing the last ten centuries. However, since the results show such a striking difference TABLE 2.3. Frequency of Leveled versus Unleveled Old English Strong Verbs Strong Verbs Strong Verbs That Have Become Weak Class I *drive 208 *bide 1 *rise 280 *reap 5 *ride 150 *slit 8 *write 599 *sneak 11 *bite 128 Partially leveled *shine 35 Average frequency 273.00 Average frequency 6.25 Class II *choose 177 *rue 6 *fly 119 *seethe 0 *shoot 187 *smoke 59 *lose 274 *float 23 *flee 40 *shove 16 Average frequency 159.40 Average frequency 32.50 Class VII *fall 338 *wax 19 *hold 498 *weep 31 *know 1227 *beat 96 *grow 257 *hew 1 *blow 81 *leap 42 *mow 1 Average frequency 473.80 sow 3 *flow 95 *row 53 Average frequency 37.89
- 42. WORD FREQUENCY IN LEXICAL DIFFUSION 29 in frequency between leveled and nonleveled forms, I do not think a more accurate frequency count would alter the general picture. A way to avoid this problem would be to study modern leveling. One case I have investigated involves the six verbs creep, keep, leap, leave, sleep, and weep, all of which have a past form with a lax vowel (due to the Middle English laxing mentioned earlier). Of these verbs, three, creep, leap, and weep, all may have, at least marginally, a past forms with a tense vowel, creeped, leaped, and weeped. The other three verbs are in no way threatened by lev- eling; past forms *keeped, *leaved, *sleeped are clearly out of the question. Now consider the frequency differences among these verbs, in table 2.4. Again the hy- pothesis that less frequent forms are leveled first is supported. 3. Implications for the source of change Of course I have not examined enough data to produce strong statistical support for the two tendencies upon which I will base my theoretical speculations, but this small body of data satisfies me that the intuitions of linguists over the years have been essentially correct: phonetic change tends to affect frequent words first, while ana- logical leveling tends to affect infrequent words first. The implication of this differ- ence is that phonetic change and analogical change must be treated differently in both diachronic theory and synchronic theory. With regard to theories of diachronic change, I would like to explore the im- plications of the reverse tendencies for theories concerning the source of linguistic change. I will argue that since the two types of change diffuse through the lexicon in opposite manners, we must posit two different sources for these two types of change. One hypothesis about the source of linguistic change, which I will call the im- perfect learning hypothesis, holds that a language is changed in the process of being transmitted from one generation to the next. This view has been proposed by Halle (1964), Kiparsky (1968), King (1969), Stampe (1969, 1973), and others. One point of view, stated in Kiparsky (1970), is that analogical change comes about as the result of overgeneralization on the part of children acquiring their language. This claim is quite plausible: the similarities between children’s overgeneralization and analogical levelings are quite striking and familiar. The tendency I have discussed here concerning the lexical diffusion of analogical change fits in quite well with the imperfect learning hypothesis. If children are going to get away with any overgeneralization, it is more likely to be in infrequent forms, where it will not be noticed so much. If they are going to learn any suppletive paradigms, these will TABLE 2.4. Modern English Leveling Not Subject to Leveling Subject to Leveling *keep 531 *creep 37 *leave 792 *leap 42 *sleep 132 *weep 31
- 43. 30 BACKGROUND AND CURRENT CONTEXT surely be the more frequent paradigms; for frequent forms there is a greater avail- ability of the model, more opportunity to practice, and greater pressure to conform. (Notice that I am not claiming that children acquire the most frequent words first.) Actually, the imperfect learning hypothesis need not be restricted to young chil- dren. Adult speakers may very well play a role in analogical leveling. For infre- quent paradigms adults may not be exactly sure of all the forms, or the mere infrequency of a suppletive paradigm makes an analogical formation more accept- able. For instance, as I mentioned earlier, creeped is not standard, but I would not flinch if I heard it, and I might even produce it myself, although I know crept is “correct.” However, keeped would definitely cause a negative reaction, because the form kept is so solidly established, due to its frequency. Thus I conclude that the imperfect learning hypothesis is quite a reasonable hypothesis about analogi- cal change, but it need not be applied only to children. The imperfect learning hypothesis has also been proposed to explain phonetic changes, chiefly by David Stampe (1969, 1973). Again we find certain correspon- dences between the substitutions that children make and known historical sound changes, although the correspondence is by no means perfect (Dressler 1974). Stampe claims that children possess a set of innate phonological processes that they must learn to suppress in order to acquire a language. The failure to suppress a natural process may result in a sound change. If the reasoning I have followed in explaining how imperfect learning allows analogical change is correct, that is, if it is correct that the forms that are most likely to be imperfectly learned are the infrequent forms, then phonetic changes such as schwa deletion and any other change that follows the same pattern of lexical diffu- sion cannot be attributed to imperfect learning. A child’s command of the phonetic shape for a frequent word will certainly be closer to the adult model than for an in- frequent word, since as I mentioned earlier, for frequent words the model is more available and the pressure to conform is greater. Thus we cannot postulate that chil- dren initiate phonetic changes such as vowel reduction or schwa deletion. What, then, is the source of a change such as schwa deletion? I suggest that the source and cause of such changes should be sought in adult speech, especially the casual style of speech. It is no accident that “new rules” appear to start out as op- tional or variable rules in the most casual styles of speech. These phonetic changes (or rules) are naturally occurring in casual or fast speech and are due to a tendency to minimize effort; thus vowels move closer to the neutral position, as in Fidelholtz’s vowel reduction, and schwas grow shorter and shorter until they are simply skipped. Or these phonetic changes come about as changes in the timing of articulatory gestures—thus we get assimilations as in didya, wouldya, and so on. Zipf (1935 [1965]) would claim that these phonetic changes originate because certain sounds or sequences of sounds are frequent. I would suggest, however, that these phonetic changes are motivated solely by phonetics, not frequency, but that they affect frequent items first because it is the frequent words that make up the sub- stance of casual speech. That is, the more casual the style, the greater the chance that the lexical items used will be the more frequent ones. If schwa deletion in its initial stages is restricted to casual speech, then only the words used in casual speech will have variable forms.
- 44. WORD FREQUENCY IN LEXICAL DIFFUSION 31 After the initial stages, the language acquisition process begins to play a role, for it is probably in the transmission of the language that restructuring takes place. A very frequent word such as every may be variable for an adult speaker, but a new learner may take the schwa-less pronunciation to be the norm or base form. The younger speaker will have a schwa-less underlying form, and the process will be complete for that word. The change in underlying forms will progress similarly over several generations. This is only a very general outline of how changes originating in casual speech may become permanent changes in lexical items. A number of problems remain. These concern in particular a synchronic grammar, such as that of present-day American English, which must represent an ongoing change such as schwa deletion. For ex- ample, if schwa deletion is a “rule,” as we normally think of rules, how can it be made to affect different words in different ways? I do not think the relative frequency of words is a part of native speaker competence, so I would not propose to make the rule sensitive to word frequency. We could propose that schwa deletion is not a rule in the ordinary sense of “rule,” but then we must discover what sort of phenomenon it is. Let me now draw some conclusions for diachronic linguistics. I have suggested earlier that I expect that sound changes that we do not ordinarily think of as reduc- tions, for example, vowel shifts, would also affect frequent items before infrequent items.7 If it turns out that they do, then we can make the general claim that all sound change initiates in language use or casual speech rather than in the acquisition pro- cess. However, if it turned out that vowel shifts and some other phonetic changes affect infrequent forms before frequent forms, then we would have an interesting indication that phonetic changes arise from different sources, and furthermore, if my hypotheses are correct, a way of determining which types of changes are traceable to which source. Thus it appears that lexical diffusion, studied in terms of word fre- quency, may turn up some interesting evidence concerning the source of morpho- phonological change. Appendix: Word list Accessory Factory Plenary Adultery Feathery Plenipotentiary Anniversary Fernery Pottery Armory Fishery Powdery Artery Flattery Quackery Artillery Forgery Raillery Battery Gallery Recovery Beggary Glossary Refectory Boundary Granary Refractory Bravery Haberdashery Robbery Brewery Hickory Rosary Burglary History Rotary Butchery Illusory Salary Buttery Injury Satisfactory
- 45. 32 BACKGROUND AND CURRENT CONTEXT Calvary Introductory Savory Cannery Ivory Scenery Celery Leathery Sedimentary Century Library Sensory Chancery Lottery Shivery Chicanery Luxury Shrubbery Complementary Machinery Silvery Complimentary Mastery Slavery Compulsory Memory Slippery Contradictory Mercury Snobbery Contrary Misery Sorcery Creamery Mockery Spidery Cursory Mystery Splintery Debauchery Nursery Summary Delivery Papery Supervisory Desultory Parliamentary Supplementary Directory Penury Surgery Discovery Peppery Treachery Dispensary Peremptory Treasury Documentary Perfumery Unsatisfactory Drapery Perfunctory Vagary Drudgery Periphery Victory Elementary Perjury Watery Embroidery Notes I am very grateful to Richard Mowrey for many stimulating discussions on the topic of this paper. 1. Kucera and Francis (1967) give a frequency of 4 for celery and 11 for artillery; this is due to the type of written texts they use. I am certain that for speakers other than militarists celery is the more frequent of the pair in spoken language. This pair illustrates clearly a gen- eral problem with relying on frequency counts of written material. Since I have computed the results of the test in terms of average frequencies for groups of words, this problem does not interfere greatly with the validity of my conclusions. 2. The “double-standard” reverse word list was used. It is a list of all left-justified, bold- face words, with standard meanings occurring in both the Shorter Oxford English Dictionary and the Merriam Webster New International Dictionary, third edition. 3. For nouns total the frequency counts for singulars and plurals to obtain the figure I used. 4. One of the problems that made the task difficult is that particular words may have deletion or not, depending on how they are used. Some examples: (i) [mkri] is a possible shortening of mercury only when it refers to the make of an automobile, but never for the element or planet; (ii) Richard Mowrey (personal communication) observed the following: parliamentary reduces in a parliamentary procedure, but not in Their form of government is parliamentary; rosary for this speaker usually has a schwa, but in the name Rosary Hill (a college in Buffalo) the schwa is deleted. These examples suggest that the degree of stress may condition the presence or absence of the schwa in words that have variable form. 5. The phonological constraints on schwa deletion are due to syllable structure con- straints. These are explained in detail in Hooper (1976). The deletion under discussion here
- 46. WORD FREQUENCY IN LEXICAL DIFFUSION 33 creates a new syllable-initial cluster; e.g., hickory gives [hIkri]. Under the theory of syllable structure developed in Vennemann and Wilbur (1972) and Hooper (1976), a consonant in initial position must be stronger than the consonant in second position. The flap in English is an extremely weak consonant; I suggest that it is too weak to cluster with an /r/. In order to have a cluster with an /r/, the alveolar consonant must be pronounced as a stop, either [th] or [d]. The data show, however, that speakers prefer to keep the schwa, rather than pronouncing a full stop here. Observe further that the phonological constraints operating here are not language-specific constraints, for many violations of English syllable structure are allowed; e.g., nasal plus /r/ is allowed, memory, scenery; /s/ plus /r/, nursery. Neither of these clusters occurs ordinarily as a syllable-initial cluster in English. 6. Sweet’s other classes of strong verbs were not appropriate, either because they had too few surviving members or because leveling occurred according to phonological subtype. 7. Labov (1966) finds that the degree of vowel shift is greater for some speakers in ca- sual styles than in more formal styles, but this fact in itself does not prove that the vowel shifts originate in casual speech. References Davis, Norman. 1965. Sweet’s Anglo-Saxon primer. Oxford, Clarendon Press. Dolby, J. L., and H. L. Resnikoff. 1967. The English word speculum. The Hague: Mouton. Dressler, Wolfgang. 1974. Diachronic puzzles for natural phonology, Papers from the Parasession on Natural Phonology, 95–102. Chicago: Chicago Linguistic Society. Fidelholtz, James L. 1975. Word frequency and vowel reduction in English. In annual Papers from the eleventh meeting of the Chicago Linguistic Society, 200–213. Chicago: Chi- cago Linguistic Society. Halle, Morris. 1964. Phonology in generative grammar. In The structure of language, ed. J. A. Fodor and J. J. Katz, 334–352. Englewood Cliffs, N.J.: Prentice-Hall. Hooper, Joan B. 1976. An introduction to natural generative phonology. New York: Aca- demic Press. King, Robert D. 1969. Historical linguistics and generative grammar. Englewood Cliffs, N.J.: Prentice-Hall. Kiparsky,Paul.1968.Linguisticuniversalsandlinguisticchange.InUniversalsinlinguistictheory, ed. Emmond Bach and Robert Harms, 170–202. New York: Holt, Rinehart and Winston. ———. 1970. Historical linguistics. In New horizons in linguistics, ed. John Lyons, 302– 315. London: Penguin. Kucera, Henry, and Nelson Francis. 1967. Computational analysis of present-day American English. Providence, R.I.: Brown University Press. Labov, William. 1966. The social stratification of English in New York City. Arlington: Cen- ter for Applied Linguistics. Leslau, Wolf. 1969. Frequency as a determinant of linguistic changes in the Ethiopian lan- guages,” Word 25, 180–189. Moore, Samuel. 1968. Historical outlines of English sounds and inflections. Revised by Albert H. Marckwardt. Ann Arbor, Mich.: George Wahr. Paul, Hermann. 1970 [1890]. Principles of the history of language. Trans. by H. A. Strong. College Park, Md.: McGrath. Schuchardt, Hugo. 1885 [1972]. On sound laws: Against the Neogrammarians, in Schuchardt, the Neogrammarians, and the transformational theory of phonological change(Linguistische Forschungen 26), ed. Theo Venneman and Terence H. Wilbur, 29–72. Frankfurt am Main. Athenaum.
- 47. 34 BACKGROUND AND CURRENT CONTEXT Stampe, David. 1969. On the acquisition of phonetic representation, Papers from the fifth regional meeting of the Chicago Linguistic Society, 443–454. ———. 1973. A dissertation on natural phonology. Diss., University of Chicago. Zipf, George K. 1935 [1965]. The psycho-biology of language. Cambridge, Mass.: MIT Press. Zwicky, Arnold. 1972. Note on a phonological hierarchy in English. In Linguistic change and generative theory, ed. R. P. Stockwell and R. K. S. Macaulay, 275–301. Bloomington: Indiana University Press.
- 48. PART II FREQUENCY AS A DETERMINANT OF MORPHOLOGICAL STRUCTURE
- 49. This page intentionally left blank
- 50. The chapters of part II resulted from a period of intense focus on the impact of meaning and use on morphological structure. Data from language change, child lan- guage, and experimentation are considered in trying to determine the nature of the cognitive organization of morphology, especially inflectional morphology. While the effects of both type and token frequency are explored in these works, frequency ef- fects are by no means the only factors considered in these chapters. Phonological similarity and degree of semantic relatedness are also considered to have an effect on the organization of paradigms. Chapter 3 reprints Bybee and Brewer (1980), which focuses on analogical changes in the preterite verbal paradigms of dialects of Castilian and Provençal. Though the changes work with different phonological material, indicating their in- dependence from one another, the principles underlying the changes are the same. These changes point to the relative autonomy of certain forms, particularly the third singular and the first singular, which both resist change and serve as the basis of new analogical formations. Autonomy is said to depend upon semantic simplicity, morphophonemic irregularity, and word frequency. Because of the accepted dogma of the day—that frequency has nothing to do with linguistic structure—we were somewhat hesitant to hypothesize frequency as the major factor, a position I would be more likely to choose today. While chapter 3 deals with token frequency, chapter 4 (Bybee and Pardo 1981) reports on a nonce-probe experiment with Spanish-speaking adults that investigates a variety of issues, including the internal organization of paradigms, as in chapter 3, but also the role of type and token frequency in determining productivity in Introduction to Part II 37
- 51. morphophonological alternations. It is found that (1) certain widespread vowel al- ternations, especially diphthongization, are not readily extended to nonce forms; (2) when alternations are extended, it is under conditions in which both alternates are made available in the experiment and one is in a form that is closely related se- mantically to the target form; (3) morphophonological classes do not necessarily make use of the broadest phonological generalizations but often have more specific and local phonological reference (see also Bybee 2001:130ff); (4) type frequency and phonological similarity relate to the degree of productivity; and (5) paradigms with very high token frequency do not serve as analogical models, thus detracting from productivity. Subsequent studies of Spanish diphthongization have confirmed the lexical specificity of diphthongization, identifying both morphological and phono- logical contexts that favor or disfavor diphthongs (Eddington 1996, 1998; Albright, Andrade, and Hayes 2001). Chapter 5 (Bybee and Slobin 1982) deals with the English past tense in children and adults, taking into account the type frequency of the irregular patterns, the token frequency of members of the irregular classes, the phonological similarity between the base and the past, and the phonological similarity among past forms themselves. Because of its focus on the differences in type and token frequency between regular and irregular past forms, Rumelhart and McClelland (1986) chose to simulate the results of this study with a connectionist model, setting off a long-standing debate about the role of structure versus frequency in the processing of regular versus ir- regular forms (Pinker 1991; McClelland and Patterson 2002 and references there). One particular class of irregular verbs in English, exemplified by string/strung, shows some productivity, offering an opportunity to understand further the factors that govern productivity. This class is studied in chapter 6, which reprints Bybee and Moder (1983). Here it is shown that the phonological shape of the verb, including not just the final consonants, but also the initial ones, defines the class. This chapter argues for product-oriented schemas on the basis of the fact that the vowel of the base is the least important factor in predicting class membership, for example, verbs such as strike/struck are recent additions to the class. The results reported in this paper have been replicated by Prasada and Pinker (1993). In chapter 7, Bybee and Newman (1995) contrast explanations traced to grammati- cization with explanations based on synchronic processing preferences. We argue that the rarity of stem changes versus affixes in the languages of the world is due to the way each mode of expression develops diachronically and not to a postulated extra pro- cessing cost for stem changes. The acquisition difficulty associated with stem changes is due to their low type frequency in the languages in which they have been studied. We report here on an experiment in which subjects were taught an artificial language in which stem changes have type frequency comparable to affixes. In this experiment, subjects learned the stem changes as easily as the affixes. Chapter 8 is a response to some claims that arose in the Great Past Tense De- bate, which, as mentioned before, was triggered by Rumelhart and McClelland’s connectionist simulation of the acquisition of the English past tense. An important topic in the debate centers on the role of type frequency. The dual-mechanism camp (Pinker 1991; Marcus et al. 1992 and subsequent papers) argue that productivity is structurally determined; somehow the child learns which is the default rule, possibly 38 FREQUENCY AS A DETERMINANT OF MORPHOLOGICAL STRUCTURE
- 52. by having heard it in a default environment. In this theory, type frequency is of no consequence. In the connectionist account and my own network model, type frequency is a strong determining factor for productivity. In this chapter, I argue that there are degrees of productivity that correspond to the interaction of phonological similarity with type frequency. The chapter also argues that high-frequency regular past tense forms are represented in the mental lexicon, for which further evidence is presented in chapter 9. References Albright, Adam, Argelia Andrade, and Bruce Hayes. 2001. Segmental environments of Span- ish diphthongization. In Papers in phonology 5. UCLA Working Papers in Linguistics 7, ed. A. Albright and T. Cho, 117–151. Los Angeles: UCLA Department of Linguistics. Bybee, Joan L. 2001. Phonology and language use. Cambridge: Cambridge University Press. Eddington, David. 1996. Diphthongization in Spanish derivational morphology: An empiri- cal investigation. Hispanic Linguistics 8(1), 1–13. ———. 1998. Spanish diphthongization as a non-derivational phenomenon. Rivista di Linguistica 10(2), 335–354. Marcus, G. F., S. Pinker, M. Ullman, M. Hollander, T. J. Rosen, F. Xu, and H. Clahsen. 1992. Overregularization in language acquisition. Monographs of the Society for Research in Child Development 57. Chicago: University of Chicago Press. McClelland, James L., and Karalyn Patterson. 2002. Rules or connections in past-tense inflec- tions: What does the evidence rule out? Trends in Cognitive Sciences 6, 465–472. Pinker, Steven. 1991. Rules of language. Science 253, 530–535. Prasada, Sandeep, and Pinker, Steven. 1993. Generalisation of regular and irregular morpho- logical patterns. Language and Cognitive Processes 8, 1–15. Rumelhart, David E., and James L. McClelland. 1986. On learning the past tenses of English verbs: Implicit rules or parallel distributed processing? In Parallel distributed processing: Explorations in the microstructure of cognition, ed. James L. McClelland, David E. Rumelhart, and the PDP Research Group, 216–271. Cambridge, Mass.: MIT Press. FREQUENCY AS A DETERMINANT OF MORPHOLOGICAL STRUCTURE 39
- 53. This page intentionally left blank
- 54. Chapter 3 WITH MARY ALEXANDRA BREWER Explanation in Morphophonemics Changes in Provençal and Spanish Preterite Forms 41 Morphophonemic alternations (alternations with some morphosyntactic or lexi- cal conditioning) would seem to be among the most arbitrary and conventional as- pects of natural language. Their treatment in most theories characterizes them as synchronically arbitrary—Bloomfield (1933) and later Chomsky and Halle (1968) describe morphophonemics with historical reconstructions, while others are satis- fied with long lists of alternants and statements about distribution. More recently, morphophonemic alternations have been excluded from phonology, as not being explainable by phonological principles but belonging to some other domain (Andersen 1969; Vennemann 1972; Stampe 1973). The question now arises, what is this other domain, and are morphophonemic alternations synchronically arbitrary, or are they somehow motivated by the communication process and its participants? In this essay we take the point of view that morphophonemic alternations are not arbitrary but rather tend to form diagrammatic relations as described in Andersen (1980). Andersen’s proposal (based on some brief suggestions in Jakobson 1965) is that the relations among signantia (the phonological forms of morphemes) or the relations among variants of signantia reflect parallel or diagram relations among sig- nata (the semantic values of morphemes). In particular, we present evidence, as others have, that there is a basic-derived relationship among the forms of a stem paradigm that diagrams the basic-derived relationship in the semantic domain (section 1). More important, we investigate the question of what determines particular basic-derived relationships and the question of why diagrams exist. Our proposal is that words may be classified as to degree of autonomy—the extent to which a word is likely to be represented in the speaker’s lexicon as a whole
- 55. Another Random Document on Scribd Without Any Related Topics
- 56. Rosamond's face. From all her other features the expression seemed to be gone. She said nothing, she noticed nothing. She did not start, she did not move aside an inch, when the landlady returned, and led Mr. Frankland to his wife. Lenny! don't let the new nurse stop here to-night—pray, pray don't! whispered Rosamond, eagerly catching her husband by the arm. Warned by the trembling of her hand, Mr. Frankland laid his fingers lightly on her temples and on her heart. Good Heavens, Rosamond! what has happened? I left you quiet and comfortable, and now— I've been frightened, dear—dreadfully frightened, by the new nurse. Don't be hard on her, poor creature; she is not in her right senses—I am certain she is not. Only get her away quietly—only send her back at once to where she came from. I shall die of the fright, if she stops here. She has been behaving so strangely—she has spoken such words to me—Lenny! Lenny! don't let go of my hand. She came stealing up to me so horribly, just where you are now; she knelt down at my ear, and whispered—oh, such words! Hush, hush, love! said Mr. Frankland, getting seriously alarmed by the violence of Rosamond's agitation. Never mind repeating the words now; wait till you are calmer—I beg and entreat of you, wait till you are calmer. I will do every thing you wish, if you will only lie down and be quiet, and try to compose yourself before you say another word. It is quite enough for me to know that this woman has frightened you, and that you wish her to be sent away with as little harshness as possible. We will put off all further explanations till to-morrow morning. I deeply regret now that I did not persist in carrying out my own idea of sending for a proper nurse from London. Where is the landlady? The landlady placed herself by Mr. Frankland's side. Is it late? asked Leonard. Oh no, Sir; not ten o'clock yet.
- 57. Order a fly to be brought to the door, then, as soon as possible, if you please. Where is the nurse? Standing behind you, Sir, near the wall, said the maid. As Mr. Frankland turned in that direction, Rosamond whispered to him: Don't be hard on her, Lenny. The maid, looking with contemptuous curiosity at Mrs. Jazeph, saw the whole expression of her countenance alter, as those words were spoken. The tears rose thick in her eyes, and flowed down her cheeks. The deathly spell of stillness that had lain on her face was broken in an instant. She drew back again, close to the wall, and leaned against it as before. Don't be hard on her! the maid heard her repeat to herself, in a low sobbing voice. Don't be hard on her! Oh, my God! she said that kindly—she said that kindly, at least! I have no desire to speak to you, or to use you unkindly, said Mr. Frankland, imperfectly hearing what she said. I know nothing of what has happened, and I make no accusations. I find Mrs. Frankland violently agitated and frightened; I hear her connect that agitation with you—not angrily, but compassionately—and, instead of speaking harshly, I prefer leaving it to your own sense of what is right, to decide whether your attendance here ought not to cease at once. I have provided the proper means for your conveyance from this place; and I would suggest that you should make our apologies to your mistress, and say nothing more than that circumstances have happened which oblige us to dispense with your services. You have been considerate toward me, Sir, said Mrs. Jazeph, speaking quietly, and with a certain gentle dignity in her manner, and I will not prove myself unworthy of your forbearance by saying what I might say in my own defense. She advanced into the middle of the room, and stopped where she could see Rosamond plainly. Twice she attempted to speak, and twice her voice failed her. At the third effort she succeeded in controlling herself. Before I go, ma'am, she said, I hope you will believe that I have no bitter feeling against you for sending me away. I am not
- 58. angry—pray remember always that I was not angry, and that I never complained. There was such a forlornness in her face, such a sweet, sorrowful resignation in every tone of her voice during the utterance of these few words, that Rosamond's heart smote her. Why did you frighten me? she asked, half relenting. Frighten you? How could I frighten you? Oh me! of all the people in the world, how could I frighten you? Mournfully saying those words, the nurse went to the chair on which she had placed her bonnet and shawl, and put them on. The landlady and the maid, watching her with curious eyes, detected that she was again weeping bitterly, and noticed with astonishment, at the same time, how neatly she put on her bonnet and shawl. The wasted hands were moving mechanically, and were trembling while they moved,—and yet, slight thing though it was, the inexorable instinct of propriety guided their most trifling actions still. On her way to the door, she stopped again at passing the bedside, looked through her tears at Rosamond and the child, struggled a little with herself, and then spoke her farewell words— God bless you, and keep you and your child happy and prosperous, she said. I am not angry at being sent away. If you ever think of me again, after to-night, please to remember that I was not angry, and that I never complained. She stood for a moment longer, still weeping, and still looking through her tears at the mother and child—then turned away and walked to the door. Something in the last tones of her voice caused a silence in the room. Of the four persons in it not one could utter a word, as the nurse closed the door gently, and went out from them alone.
- 59. CHAPTER V. A COUNCIL OF THREE. On the morning after the departure of Mrs. Jazeph, the news that she had been sent away from the Tiger's Head by Mr. Frankland's directions, reached the doctor's residence from the inn just as he was sitting down to breakfast. Finding that the report of the nurse's dismissal was not accompanied by any satisfactory explanation of the cause of it, Mr. Orridge refused to believe that her attendance on Mrs. Frankland had really ceased. However, although he declined to credit the news, he was so far disturbed by it that he finished his breakfast in a hurry, and went to pay his morning visit at the Tiger's Head nearly two hours before the time at which he usually attended on his patient. On his way to the inn, he was met and stopped by the one waiter attached to the establishment. I was just bringing you a message from Mr. Frankland, Sir, said the man. He wants to see you as soon as possible. Is it true that Mrs. Frankland's nurse was sent away last night by Mr. Frankland's order? asked Mr. Orridge. Quite true, Sir, answered the waiter. The doctor colored, and looked seriously discomposed. One of the most precious things we have about us—especially if we happen to belong to the medical profession—is our dignity. It struck Mr. Orridge that he ought to have been consulted before a nurse of his recommending was dismissed from her situation at a moment's notice. Was Mr. Frankland presuming upon his position as a gentleman of fortune? The power of wealth may do much with impunity, but it is not privileged to offer any practical contradictions to a man's good opinion of himself. Never had the doctor thought more disrespectfully of rank and riches; never had he been conscious of reflecting on republican principles with such absolute
- 60. impartiality, as when he now followed the waiter in sullen silence to Mr. Frankland's room. Who is that? asked Leonard, when he heard the door open. Mr. Orridge, Sir, said the waiter. Good-morning, said Mr. Orridge, with self-asserting abruptness and familiarity. Mr. Frankland was sitting in an arm-chair, with his legs crossed. Mr. Orridge carefully selected another arm-chair, and crossed his legs on the model of Mr. Frankland's the moment he sat down. Mr. Frankland's hands were in the pockets of his dressing-gown. Mr. Orridge had no pockets, except in his coat-tails, which he could not conveniently get at; but he put his thumbs into the arm-holes of his waistcoat, and asserted himself against the easy insolence of wealth in that way. It made no difference to him—so curiously narrow is the range of a man's perceptions when he is insisting on his own importance—that Mr. Frankland was blind, and consequently incapable of being impressed by the independence of his bearing. Mr. Orridge's own dignity was vindicated in Mr. Orridge's own presence, and that was enough. I am glad you have come so early, doctor, said Mr. Frankland. A very unpleasant thing happened here last night. I was obliged to send the new nurse away at a moment's notice. Were you, indeed! said Mr. Orridge, defensively matching Mr. Frankland's composure by an assumption of the completest indifference. Aha! were you indeed? If there had been time to send and consult you, of course I should have been only too glad to have done so, continued Leonard; but it was impossible to hesitate. We were all alarmed by a loud ringing of my wife's bell; I was taken up to her room, and found her in a condition of the most violent agitation and alarm. She told me she had been dreadfully frightened by the new nurse; declared her conviction that the woman was not in her right senses; and entreated that I would get her out of the house with as little delay and as little harshness as possible. Under these circumstances,
- 61. what could I do? I may seem to have been wanting in consideration toward you, in proceeding on my own sole responsibility; but Mrs. Frankland was in such a state of excitement that I could not tell what might be the consequence of opposing her, or of venturing on any delays; and after the difficulty had been got over, she would not hear of your being disturbed by a summons to the inn. I am sure you will understand this explanation, doctor, in the spirit in which I offer it. Mr. Orridge began to look a little confused. His solid substructure of independence was softening and sinking from under him. He suddenly found himself thinking of the cultivated manners of the wealthy classes; his thumbs slipped mechanically out of the arm- holes of his waistcoat; and, before he well knew what he was about, he was stammering his way through all the choicest intricacies of a complimentary and respectful reply. You will naturally be anxious to know what the new nurse said or did to frighten my wife so, pursued Mr. Frankland. I can tell you nothing in detail; for Mrs. Frankland was in such a state of nervous dread last night that I was really afraid of asking for any explanations; and I have purposely waited to make inquiries this morning until you could come here and accompany me up stairs. You kindly took so much trouble to secure this unlucky woman's attendance, that you have a right to hear all that can be alleged against her, now she has been sent away. Considering all things, Mrs. Frankland is not so ill this morning as I was afraid she would be. She expects to see you with me; and, if you will kindly give me your arm, we will go up to her immediately. On entering Mrs. Frankland's room, the doctor saw at a glance that she had been altered for the worse by the events of the past evening. He remarked that the smile with which she greeted her husband was the faintest and saddest he had seen on her face. Her eyes looked dim and weary, her skin was dry, her pulse was irregular. It was plain that she had passed a wakeful night, and that her mind was not at ease. She dismissed the inquiries of her medical
- 62. attendant as briefly as possible, and led the conversation immediately, of her own accord, to the subject of Mrs. Jazeph. I suppose you have heard what has happened, she said, addressing Mr. Orridge. I can't tell you how grieved I am about it. My conduct must look in your eyes, as well as in the eyes of the poor unfortunate nurse, the conduct of a capricious, unfeeling woman. I am ready to cry with sorrow and vexation when I remember how thoughtless I was, and how little courage I showed. Oh, Lenny, it is dreadful to hurt the feelings of any body, but to have pained that unhappy, helpless woman as we pained her, to have made her cry so bitterly, to have caused her such humiliation and wretchedness— My dear Rosamond, interposed Mr. Frankland, you are lamenting effects, and forgetting causes altogether. Remember what a state of terror I found you in—there must have been some reason for that. Remember, too, how strong your conviction was that the nurse was out of her senses. Surely you have not altered your opinion on that point already? It is that very opinion, love, that has been perplexing and worrying me all night. I can't alter it; I feel more certain than ever that there must be something wrong with the poor creature's intellect—and yet, when I remember how good-naturedly she came here to help me, and how anxious she seemed to make herself useful, I can't help feeling ashamed of my suspicions; I can't help reproaching myself for having been the cause of her dismissal last night. Mr. Orridge, did you notice any thing in Mrs. Jazeph's face or manner which might lead you to doubt whether her intellects were quite as sound as they ought to be? Certainly not, Mrs. Frankland, or I should never have brought her here. I should not have been astonished to hear that she was suddenly taken ill, or that she had been seized with a fit, or that some slight accident, which would have frightened nobody else, had seriously frightened her; but to be told that there is any thing approaching to derangement in her faculties, does, I own, fairly surprise me.
- 63. Can I have been mistaken! exclaimed Rosamond, looking confusedly and self-distrustfully from Mr. Orridge to her husband. Lenny! Lenny! if I have been mistaken, I shall never forgive myself. Suppose you tell us, my dear, what led you to suspect that she was mad? suggested Mr. Frankland. Rosamond hesitated. Things that are great in one's own mind, she said, seem to get so little when they are put into words. I almost despair of making you understand what good reason I had to be frightened—and then, I am afraid, in trying to do justice to myself, that I may not do justice to the nurse. Tell your own story, my love, in your own way, and you will be sure to tell it properly, said Mr. Frankland. And pray remember, added Mr. Orridge, that I attach no real importance to my opinion of Mrs. Jazeph. I have not had time enough to form it. Your opportunities of observing her have been far more numerous than mine. Thus encouraged, Rosamond plainly and simply related all that had happened in her room on the previous evening, up to the time when she had closed her eyes and had heard the nurse approaching her bedside. Before repeating the extraordinary words that Mrs. Jazeph had whispered in her ear, she made a pause, and looked earnestly in her husband's face. Why do you stop? asked Mr. Frankland. I feel nervous and flurried still, Lenny, when I think of the words the nurse said to me, just before I rang the bell. What did she say? Was it something you would rather not repeat? No! no! I am most anxious to repeat it, and to hear what you think it means. As I have just told you, Lenny, we had been talking of Porthgenna, and of my project of exploring the north rooms as soon as I got there; and she had been asking many questions about the old house; appearing, I must say, to be unaccountably interested in it, considering she was a stranger.
- 64. Yes? Well, when she came to the bedside, she knelt down close at my ear, and whispered all on a sudden—'When you go to Porthgenna, keep out of the Myrtle Room!' Mr. Frankland started. Is there such a room at Porthgenna? he asked, eagerly. I never heard of it, said Rosamond. Are you sure of that? inquired Mr. Orridge. Up to this moment the doctor had privately suspected that Mrs. Frankland must have fallen asleep soon after he left her the evening before; and that the narrative which she was now relating, with the sincerest conviction of its reality, was actually derived from nothing but a series of vivid impressions produced by a dream. I am certain I never heard of such a room, said Rosamond. I left Porthgenna at five years old; and I had never heard of it then. My father often talked of the house in after-years; but I am certain that he never spoke of any of the rooms by any particular names; and I can say the same of your father, Lenny, whenever I was in his company after he had bought the place. Besides, don't you remember, when the builder we sent down to survey the house wrote you that letter, he complained that there were no names of the rooms on the different keys to guide him in opening the doors, and that he could get no information from any body at Porthgenna on the subject? How could I ever have heard of the Myrtle Room? Who was there to tell me? Mr. Orridge began to look perplexed; it seemed by no means so certain that Mrs. Frankland had been dreaming, after all. I have thought of nothing else, said Rosamond to her husband, in low, whispering tones. I can't get those mysterious words off my mind. Feel my heart, Lenny—it is beating quicker than usual only with saying them over to you. They are such very strange, startling words. What do you think they mean? Who is the woman who spoke them?—that is the most important question, said Mr. Frankland.
- 65. But why did she say the words to me? That is what I want to know—that is what I must know, if I am ever to feel easy in my mind again! Gently, Mrs. Frankland, gently! said Mr. Orridge. For your child's sake, as well as for your own, pray try to be calm, and to look at this very mysterious event as composedly as you can. If any exertions of mine can throw light upon this strange woman and her still stranger conduct, I will not spare them. I am going to-day to her mistress's house to see one of the children; and, depend upon it, I will manage in some way to make Mrs. Jazeph explain herself. Her mistress shall hear every word that you have told me; and I can assure you she is just the sort of downright, resolute woman who will insist on having the whole mystery instantly cleared up. Rosamond's weary eyes brightened at the doctor's proposal. Oh, go at once, Mr. Orridge! she exclaimed—go at once! I have a great deal of medical work to do in the town first, said the doctor, smiling at Mrs. Frankland's impatience. Begin it, then, without losing another instant, said Rosamond. The baby is quite well, and I am quite well—we need not detain you a moment. And, Mr. Orridge, pray be as gentle and considerate as possible with the poor woman; and tell her that I never should have thought of sending her away if I had not been too frightened to know what I was about. And say how sorry I am this morning, and say— My dear, if Mrs. Jazeph is really not in her right senses, what would be the use of overwhelming her with all these excuses? interposed Mr. Frankland. It will be more to the purpose if Mr. Orridge will kindly explain and apologize for us to her mistress. Go! Don't stop to talk—pray go at once! cried Rosamond, as the doctor attempted to reply to Mr. Frankland. Don't be afraid; no time shall be lost, said Mr. Orridge, opening the door. But remember, Mrs. Frankland, I shall expect you to reward your embassador, when he returns from his mission, by showing him that you are a little more quiet and composed than I
- 66. find you this morning. With that parting hint, the doctor took his leave. 'When you go to Porthgenna, keep out of the Myrtle Room,' repeated Mr. Frankland, thoughtfully. Those are very strange words, Rosamond. Who can this woman really be? She is a perfect stranger to both of us; we are brought into contact with her by the merest accident; and we find that she knows something about our own house of which we were both perfectly ignorant until she chose to speak! But the warning, Lenny—the warning, so pointedly and mysteriously addressed to me? Oh, if I could only go to sleep at once, and not wake again till the doctor comes back! My love, try not to count too certainly on our being enlightened, even then. The woman may refuse to explain herself to any body. Don't even hint at such a disappointment as that, Lenny—or I shall be wanting to get up, and go and question her myself! Even if you could get up and question her, Rosamond, you might find it impossible to make her answer. She may be afraid of certain consequences which we can not foresee; and, in that case, I can only repeat that it is more than probable she will explain nothing— or, perhaps, still more likely that she will coolly deny her own words altogether. Then, Lenny, we will put them to the proof for ourselves. And how can we do that? By continuing our journey to Porthgenna the moment I am allowed to travel, and by leaving no stone unturned when we get there until we have discovered whether there is or is not any room in the old house that ever was known, at any time of its existence, by the name of the Myrtle Room. And suppose it should turn out that there is such a room? asked Mr. Frankland, beginning to feel the influence of his wife's enthusiasm.
- 67. If it does turn out so, said Rosamond, her voice rising, and her face lighting up with its accustomed vivacity, how can you doubt what will happen next? Am I not a woman? And have I not been forbidden to enter the Myrtle Room? Lenny! Lenny! Do you know so little of my half of humanity as to doubt what I should do the moment the room was discovered? My darling, as a matter of course, I should walk into it immediately. CHAPTER VI. ANOTHER SURPRISE. With all the haste he could make, it was one o'clock in the afternoon before Mr. Orridge's professional avocations allowed him to set forth in his gig for Mrs. Norbury's house. He drove there with such good-will that he accomplished the half-hour's journey in twenty minutes. The footman having heard the rapid approach of the gig, opened the hall door the instant the horse was pulled up before it, and confronted the doctor with a smile of malicious satisfaction. Well, said Mr. Orridge, bustling into the hall, you were all rather surprised last night when the housekeeper came back, I suppose? Yes, Sir, we certainly were surprised when she came back last night, answered the footman; but we were still more surprised when she went away again this morning. Went away! You don't mean to say she is gone? Yes, I do, Sir—she has lost her place, and gone for good. The footman smiled again, as he made that reply; and the housemaid, who happened to be on her way down stairs while he was speaking,
- 68. and to hear what he said, smiled too. Mrs. Jazeph had evidently been no favorite in the servants' hall. Amazement prevented Mr. Orridge from uttering another word. Hearing no more questions asked, the footman threw open the door of the breakfast-parlor, and the doctor followed him into the room. Mrs. Norbury was sitting near the window in a rigidly upright attitude, inflexibly watching the proceedings of her invalid child over a basin of beef-tea. I know what you are going to talk about before you open your lips, said the outspoken lady. But just look to the child first, and say what you have to say on that subject, if you please, before you enter on any other. The child was examined, was pronounced to be improving rapidly, and was carried away by the nurse to lie down and rest a little. As soon as the door of the room had closed, Mrs. Norbury abruptly addressed the doctor, interrupting him, for the second time, just as he was about to speak. Now, Mr. Orridge, she said, I want to tell you something at the outset. I am a remarkably just woman, and I have no quarrel with you. You are the cause of my having been treated with the most audacious insolence by three people—but you are the innocent cause, and, therefore, I don't blame you. I am really at a loss, Mr. Orridge began—quite at a loss, I assure you— To know what I mean? said Mrs. Norbury. I will soon tell you. Were you not the original cause of my sending my housekeeper to nurse Mrs. Frankland? Yes. Mr. Orridge could not hesitate to acknowledge that. Well, pursued Mrs. Norbury, and the consequence of my sending her is, as I said before, that I am treated with unparalleled insolence by no less than three people. Mrs. Frankland takes an insolent whim into her head, and affects to be frightened by my housekeeper. Mr. Frankland shows an insolent readiness to humor that whim, and hands me back my housekeeper as if she was a bad
- 69. shilling; and last, and worst of all, my housekeeper herself insults me to my face as soon as she comes back—insults me, Mr. Orridge, to that degree that I give her twelve hours' notice to leave the place. Don't begin to defend yourself! I know all about it; I know you had nothing to do with sending her back; I never said you had. All the mischief you have done is innocent mischief. I don't blame you, remember that—whatever you do, Mr. Orridge, remember that! I had no idea of defending myself, said the doctor, for I have no reason to do so. But you surprise me beyond all power of expression when you tell me that Mrs. Jazeph treated you with incivility. Incivility! exclaimed Mrs. Norbury. Don't talk about incivility—it's not the word. Impudence is the word—brazen impudence. The only charitable thing to say of Mrs. Jazeph is that she is not right in her head. I never noticed any thing odd about her myself; but the servants used to laugh at her for being as timid in the dark as a child, and for often running away to her candle in her own room when they declined to light the lamps before the night had fairly set in. I never troubled my head about this before; but I thought of it last night, I can tell you, when I found her looking me fiercely in the face, and contradicting me flatly the moment I spoke to her. I should have thought she was the very last woman in the world to misbehave herself in that way, answered the doctor. Very well. Now hear what happened when she came back last night, said Mrs. Norbury. She got here just as we were going up stairs to bed. Of course, I was astonished; and, of course, I called her into the drawing-room for an explanation. There was nothing very unnatural in that course of proceeding, I suppose? Well, I noticed that her eyes were swollen and red, and that her looks were remarkably wild and queer; but I said nothing, and waited for the explanation. All that she had to tell me was that something she had unintentionally said or done had frightened Mrs. Frankland, and that Mrs. Frankland's husband had sent her away on the spot. I disbelieved this at first—and very naturally, I think—but she persisted in the story, and answered all my questions by declaring that she could tell me nothing more. 'So then,' I said, 'I am to
- 70. believe that, after I have inconvenienced myself by sparing you, and after you have inconvenienced yourself by undertaking the business of nurse, I am to be insulted, and you are to be insulted, by your being sent away from Mrs. Frankland on the very day when you get to her, because she chooses to take a whim into her head?' 'I never accused Mrs. Frankland of taking a whim into her head,' said Mrs. Jazeph, and stares me straight in the face, with such a look as I never saw in her eyes before, after all my five years' experience of her. 'What do you mean?' I asked, giving her back her look, I can promise you. 'Are you base enough to take the treatment you have received in the light of a favor?' 'I am just enough,' said Mrs. Jazeph, as sharp as lightning, and still with that same stare straight at me—'I am just enough not to blame Mrs. Frankland.' 'Oh, you are, are you?' I said. 'Then all I can tell you is, that I feel this insult, if you don't; and that I consider Mrs. Frankland's conduct to be the conduct of an ill-bred, impudent, capricious, unfeeling woman.' Mrs. Jazeph takes a step up to me—takes a step, I give you my word of honor—and says distinctly, in so many words, 'Mrs. Frankland is neither ill-bred, impudent, capricious, nor unfeeling.' 'Do you mean to contradict me, Mrs. Jazeph?' I asked. 'I mean to defend Mrs. Frankland from unjust imputations,' says she. Those were her words, Mr. Orridge—on my honor, as a gentlewoman, those were exactly her words. The doctor's face expressed the blankest astonishment. Mrs. Norbury went on— I was in a towering passion—I don't mind confessing that, Mr. Orridge—but I kept it down. 'Mrs. Jazeph,' I said, 'this is language that I am not accustomed to, and that I certainly never expected to hear from your lips. Why you should take it on yourself to defend Mrs. Frankland for treating us both with contempt, and to contradict me for resenting it, I neither know nor care to know. But I must tell you, in plain words, that I will be spoken to by every person in my employment, from my housekeeper to my scullery-maid, with respect. I would have given warning on the spot to any other servant in this house who had behaved to me as you have behaved.' She tried to interrupt me there, but I would not allow her. 'No,' I
- 71. said, 'you are not to speak to me just yet; you are to hear me out. Any other servant, I tell you again, should have left this place to- morrow morning; but I will be more than just to you. I will give you the benefit of your five years' good conduct in my service. I will leave you the rest of the night to get cool, and to reflect on what has passed between us; and I will not expect you to make the proper apologies to me until the morning.' You see, Mr. Orridge, I was determined to act justly and kindly; I was ready to make allowances—and what do you think she said in return? 'I am willing to make any apologies, ma'am, for offending you,' she said, 'without the delay of a single minute; but, whether it is to-night, or whether it is to-morrow morning, I can not stand by silent when I hear Mrs. Frankland charged with acting unkindly, uncivilly, or improperly toward me or toward any one.' 'Do you tell me that deliberately, Mrs. Jazeph?' I asked. 'I tell it you sincerely, ma'am,' she answered; 'and I am very sorry to be obliged to do so.' 'Pray don't trouble yourself to be sorry,' I said, 'for you may consider yourself no longer in my service. I will order the steward to pay you the usual month's wages instead of the month's warning the first thing to-morrow; and I beg that you will leave the house as soon as you conveniently can afterward.' 'I will leave to-morrow, ma'am,' says she, 'but without troubling the steward. I beg respectfully, and with many thanks for your past kindness, to decline taking a month's money which I have not earned by a month's service.' And thereupon she courtesies and goes out. That is, word for word, what passed between us, Mr. Orridge. Explain the woman's conduct in your own way, if you can. I say that it is utterly incomprehensible, unless you agree with me that she was not in her right senses when she came back to this house last night. The doctor began to think, after what he had just heard, that Mrs. Frankland's suspicions in relation to the new nurse were not quite so unfounded as he had been at first disposed to consider them. He wisely refrained, however, from complicating matters by giving utterance to what he thought; and, after answering Mrs. Norbury in a few vaguely polite words, endeavored to soothe her irritation
- 72. against Mr. and Mrs. Frankland by assuring her that he came as the bearer of apologies from both husband and wife, for the apparent want of courtesy and consideration in their conduct which circumstances had made inevitable. The offended lady, however, absolutely refused to be propitiated. She rose up, and waved her hand with an air of great dignity. I can not hear a word more from you, Mr. Orridge, she said; I can not receive any apologies which are made indirectly. If Mr. Frankland chooses to call, and if Mrs. Frankland condescends to write to me, I am willing to think no more of the matter. Under any other circumstances, I must be allowed to keep my present opinions both of the lady and the gentleman. Don't say another word, and be so kind as to excuse me if I leave you, and go up to the nursery to see how the child is getting on. I am delighted to hear that you think her so much better. Pray call again to-morrow or next day, if you conveniently can. Good-morning! Half amused at Mrs. Norbury, half displeased at the curt tone she adopted toward him, Mr. Orridge remained for a minute or two alone in the breakfast-parlor, feeling rather undecided about what he should do next. He was, by this time, almost as much interested in solving the mystery of Mrs. Jazeph's extraordinary conduct as Mrs. Frankland herself; and he felt unwilling, on all accounts, to go back to the Tiger's Head, and merely repeat what Mrs. Norbury had told him, without being able to complete the narrative by informing Mr. and Mrs. Frankland of the direction that the housekeeper had taken on leaving her situation. After some pondering, he determined to question the footman, under the pretense of desiring to know if his gig was at the door. The man having answered the bell, and having reported the gig to be ready, Mr. Orridge, while crossing the hall, asked him carelessly if he knew at what time in the morning Mrs. Jazeph had left her place. About ten o'clock, Sir, answered the footman. When the carrier came by from the village, on his way to the station for the eleven o'clock train. Oh! I suppose he took her boxes? said Mr. Orridge.
- 73. And he took her, too, Sir, said the man, with a grin. She had to ride, for once in her life, at any rate, in a carrier's cart. On getting back to West Winston, the doctor stopped at the station to collect further particulars, before he returned to the Tiger's Head. No trains, either up or down, happened to be due just at that time. The station-master was reading the newspaper, and the porter was gardening on the slope of the embankment. Is the train at eleven in the morning an up-train or a down-train? asked Mr. Orridge, addressing the porter. A down-train. Did many people go by it? The porter repeated the names of some of the inhabitants of West Winston. Were there no passengers but passengers from the town? inquired the doctor. Yes, Sir. I think there was one stranger—a lady. Did the station-master issue the tickets for that train? Yes, Sir. Mr. Orridge went on to the station-master. Do you remember giving a ticket this morning, by the eleven o'clock down-train, to a lady traveling alone? The station-master pondered. I have issued tickets, up and down, to half-a-dozen ladies to-day, he answered, doubtfully. Yes, but I am speaking only of the eleven o'clock train, said Mr. Orridge. Try if you can't remember? Remember? Stop! I do remember; I know who you mean. A lady who seemed rather flurried, and who put a question to me that I am not often asked at this station. She had her veil down, I recollect, and she got here for the eleven o'clock train. Crouch, the carrier, brought her trunk into the office. That is the woman. Where did she take her ticket for? For Exeter.
- 74. You said she asked you a question? Yes: a question about what coaches met the rail at Exeter to take travelers into Cornwall. I told her we were rather too far off here to have the correct time-table, and recommended her to apply for information to the Devonshire people when she got to the end of her journey. She seemed a timid, helpless kind of woman to travel alone. Any thing wrong in connection with her, Sir? Oh, no! nothing, said Mr. Orridge, leaving the station-master and hastening back to his gig again. When he drew up, a few minutes afterward, at the door of the Tiger's Head, he jumped out of his vehicle with the confident air of a man who has done all that could be expected of him. It was easy to face Mrs. Frankland with the unsatisfactory news of Mrs. Jazeph's departure, now that he could add, on the best authority, the important supplementary information that she had gone to Cornwall.
- 75. BOOK IV. CHAPTER I. A PLOT AGAINST THE SECRET. Toward the close of the evening, on the day after Mr. Orridge's interview with Mrs. Norbury, the Druid fast coach, running through Cornwall as far as Truro, set down three inside passengers at the door of the booking-office on arriving at its destination. Two of these passengers were an old gentleman and his daughter; the third was Mrs. Jazeph. The father and daughter collected their luggage and entered the hotel; the outside passengers branched off in different directions with as little delay as possible; Mrs. Jazeph alone stood irresolute on the pavement, and seemed uncertain what she should do next. When the coachman good-naturedly endeavored to assist her in arriving at a decision of some kind, by asking whether he could do any thing to help her, she started, and looked at him suspiciously; then, appearing to recollect herself, thanked him for his kindness, and inquired, with a confusion of words and a hesitation of manner which appeared very extraordinary in the coachman's eyes, whether she might be allowed to leave her trunk at the booking-office for a little while, until she could return and call for it again. Receiving permission to leave her trunk as long as she pleased, she crossed over the principal street of the town, ascended the pavement on the opposite side, and walked down the first turning she came to. On entering the by-street to which the turning led, she glanced back, satisfied herself that nobody was following or
- 76. watching her, hastened on a few yards, and stopped again at a small shop devoted to the sale of book-cases, cabinets, work-boxes, and writing-desks. After first looking up at the letters painted over the door—Buschmann, Cabinet-maker, c.—she peered in at the shop window. A middle-aged man, with a cheerful face, sat behind the counter, polishing a rosewood bracket, and nodding briskly at regular intervals, as if he were humming a tune and keeping time to it with his head. Seeing no customers in the shop, Mrs. Jazeph opened the door and walked in. As soon as she was inside, she became aware that the cheerful man behind the counter was keeping time, not to a tune of his own humming, but to a tune played by a musical box. The clear ringing notes came from a parlor behind the shop, and the air the box was playing was the lovely Batti, Batti, of Mozart. Is Mr. Buschmann at home? asked Mrs. Jazeph. Yes, ma'am, said the cheerful man, pointing with a smile toward the door that led into the parlor. The music answers for him. Whenever Mr. Buschmann's box is playing, Mr. Buschmann himself is not far off from it. Did you wish to see him, ma'am? If there is nobody with him. Oh, no, he is quite alone. Shall I give any name? Mrs. Jazeph opened her lips to answer, hesitated, and said nothing. The shopman, with a quicker delicacy of perception than might have been expected from him, judging by outward appearances, did not repeat the question, but opened the door at once, and admitted the visitor to the presence of Mr. Buschmann. The shop parlor was a very small room, with an old three- cornered look about it, with a bright green paper on the walls, with a large dried fish in a glass case over the fire-place, with two meerschaum pipes hanging together on the wall opposite, and a neat round table placed as accurately as possible in the middle of the floor. On the table were tea-things, bread, butter, a pot of jam, and a musical box in a quaint, old-fashioned case; and by the side of the table sat a little, rosy-faced, white-haired, simple-looking old
- 77. man, who started up, when the door was opened, with an appearance of extreme confusion, and touched the top of the musical box so that it might cease playing when it came to the end of the air. A lady to speak with you, Sir, said the cheerful shopman. That is Mr. Buschmann, ma'am, he added in a lower tone, seeing Mrs. Jazeph stop in apparent uncertainty on entering the parlor. Will you please to take a seat, ma'am? said Mr. Buschmann, when the shopman had closed the door and gone back to his counter. Excuse the music; it will stop directly. He spoke these words in a foreign accent, but with perfect fluency. Mrs. Jazeph looked at him earnestly while he was addressing her, and advanced a step or two before she said any thing. Am I so changed? she asked softly. So sadly, sadly changed, Uncle Joseph? Gott im Himmel! it's her voice—it's Sarah Leeson! cried the old man, running up to his visitor as nimbly as if he was a boy again, taking both her hands, and kissing her with an odd, brisk tenderness on the cheek. Although his niece was not at all above the average height of women, Uncle Joseph was so short that he had to raise himself on tiptoe to perform the ceremony of embracing her. To think of Sarah coming at last! he said, pressing her into a chair. After all these years and years, to think of Sarah Leeson coming to see Uncle Joseph again! Sarah still, but not Sarah Leeson, said Mrs. Jazeph, pressing her thin, trembling hands firmly together, and looking down on the floor while she spoke. Ah! married? said Mr. Buschmann, gayly. Married, of course. Tell me all about your husband, Sarah. He is dead. Dead and forgiven. She murmured the last three words in a whisper to herself. Ah! I am so sorry for you! I spoke too suddenly, did I not, my child? said the old man. Never mind! No, no; I don't mean that—I mean let us talk of something else. You will have a bit of bread and
- 78. jam, won't you, Sarah?—ravishing raspberry jam that melts in your mouth. Some tea, then? So, so, she will have some tea, to be sure. And we won't talk of our troubles—at least, not just yet. You look very pale, Sarah—very much older than you ought to look—no, I don't mean that either; I don't mean to be rude. It was your voice I knew you by, my child—your voice that your poor Uncle Max always said would have made your fortune if you would only have learned to sing. Here's his pretty music box going still. Don't look so downhearted—don't, pray. Do listen a little to the music: you remember the box?—my brother Max's box? Why, how you look! Have you forgotten the box that the divine Mozart gave to my brother with his own hand, when Max was a boy in the music school at Vienna? Listen! I have set it going again. It's a song they call 'Batti, Batti;' it's a song in an opera of Mozart's. Ah! beautiful! beautiful! Your Uncle Max said that all music was comprehended in that one song. I know nothing about music, but I have my heart and my ears, and they tell me that Max was right. Speaking these words with abundant gesticulation and amazing volubility, Mr. Buschmann poured out a cup of tea for his niece, stirred it carefully, and, patting her on the shoulder, begged that she would make him happy by drinking it all up directly. As he came close to her to press this request, he discovered that the tears were in her eyes, and that she was trying to take her handkerchief from her pocket without being observed. Don't mind me, she said, seeing the old man's face sadden as he looked at her; and don't think me forgetful or ungrateful, Uncle Joseph. I remember the box—I remember every thing that you used to take an interest in, when I was younger and happier than I am now. When I last saw you, I came to you in trouble; and I come to you in trouble once more. It seems neglectful in me never to have written to you for so many years past; but my life has been a very sad one, and I thought I had no right to lay the burden of my sorrow on other shoulders than my own. Uncle Joseph shook his head at these last words, and touched the stop of the musical box. Mozart shall wait a little, he said, gravely,
- 79. till I have told you something. Sarah, hear what I say, and drink your tea, and own to me whether I speak the truth or not. What did I, Joseph Buschmann, tell you, when you first came to me in trouble, fourteen, fifteen, ah more! sixteen years ago, in this town, and in this same house? I said then, what I say again now: 'Sarah's sorrow is my sorrow, and Sarah's joy is my joy;' and if any man asks me reasons for that, I have three to give him. He stopped to stir up his niece's tea for the second time, and to draw her attention to it by tapping with the spoon on the edge of the cup. Three reasons, he resumed. First, you are my sister's child— some of her flesh and blood, and some of mine, therefore, also. Second, my sister, my brother, and lastly me myself, we owe to your good English father—all. A little word that means much, and may be said again and again—all. Your father's friends cry, Fie! Agatha Buschmann is poor! Agatha Buschmann is foreign! But your father loves the poor German girl, and he marries her in spite of their Fie, Fie. Your father's friends cry Fie! again; Agatha Buschmann has a musician brother, who gabbles to us about Mozart, and who can not make to his porridge salt. Your father says, Good! I like his gabble; I like his playing; I shall get him people to teach; and while I have pinches of salt in my kitchen, he to his porridge shall have pinches of salt too. Your father's friends cry Fie! for the third time. Agatha Buschmann has another brother, a little Stupid-Head, who to the other's gabble can only listen and say Amen. Send him trotting; for the love of Heaven, shut up all the doors and send Stupid-Head trotting, at least. Your father says, No! Stupid-Head has his wits in his hands; he can cut and carve and polish; help him a little at the starting, and after he shall help himself. They are all gone now but me! Your father, your mother, and Uncle Max—they are all gone. Stupid-Head alone remains to remember and to be grateful—to take Sarah's sorrow for his sorrow, and Sarah's joy for his joy. He stopped again to blow a speck of dust off the musical box. His niece endeavored to speak, but he held up his hand, and shook his forefinger at her warningly.
- 80. No, he said. It is yet my business to talk, and your business to drink tea. Have I not my third reason still? Ah! you look away from me; you know my third reason before I say a word. When I, in my turn, marry, and my wife dies, and leaves me alone with little Joseph, and when the boy falls sick, who comes then, so quiet, so pretty, so neat, with the bright young eyes, and the hands so tender and light? Who helps me with little Joseph by night and by day? Who makes a pillow for him on her arm when his head is weary? Who holds this box patiently at his ear?—yes! this box, that the hand of Mozart has touched—who holds it closer, closer always, when little Joseph's sense grows dull, and he moans for the friendly music that he has known from a baby, the friendly music that he can now so hardly, hardly hear? Who kneels down by Uncle Joseph when his heart is breaking, and says, 'Oh, hush! hush! The boy is gone where the better music plays, where the sickness shall never waste or the sorrow touch him more?' Who? Ah, Sarah! you can not forget those days; you can not forget the Long Ago! When the trouble is bitter, and the burden is heavy, it is cruelty to Uncle Joseph to keep away; it is kindness to him to come here. The recollections that the old man had called up found their way tenderly to Sarah's heart. She could not answer him; she could only hold out her hand. Uncle Joseph bent down, with a quaint, affectionate gallantry, and kissed it; then stepped back again to his place by the musical box. Come! he said, patting it cheerfully, we will say no more for a while. Mozart's box, Max's box, little Joseph's box, you shall talk to us again! Having put the tiny machinery in motion, he sat down by the table, and remained silent until the air had been played over twice. Then observing that his niece seemed calmer, he spoke to her once more. You are in trouble, Sarah, he said, quietly. You tell me that, and I see it is true in your face. Are you grieving for your husband? I grieve that I ever met him, she answered. I grieve that I ever married him. Now that he is dead, I can not grieve—I can only forgive him.
- 81. Forgive him? How you look, Sarah, when you say that! Tell me— Uncle Joseph! I have told you that my husband is dead, and that I have forgiven him. You have forgiven him? He was hard and cruel with you, then? I see; I see. That is the end, Sarah—but the beginning? Is the beginning that you loved him? Her pale cheeks flushed; and she turned her head aside. It is hard and humbling to confess it, she murmured, without raising her eyes; but you force the truth from me, uncle. I had no love to give to my husband—no love to give to any man. And yet you married him! Wait! it is not for me to blame. It is for me to find out, not the bad, but the good. Yes, yes; I shall say to myself, she married him when she was poor and helpless; she married him when she should have come to Uncle Joseph instead. I shall say that to myself, and I shall pity, but I shall ask no more. Sarah half reached her hand out to the old man again—then suddenly pushed her chair back, and changed the position in which she was sitting. It is true that I was poor, she said, looking about her in confusion, and speaking with difficulty. But you are so kind and so good, I can not accept the excuse that your forbearance makes for me. I did not marry him because I was poor, but— She stopped, clasped her hands together, and pushed her chair back still farther from the table. So! so! said the old man, noticing her confusion. We will talk about it no more. I had no excuse of love; I had no excuse of poverty, she said, with a sudden burst of bitterness and despair. Uncle Joseph, I married him because I was too weak to persist in saying No! The curse of weakness and fear has followed me all the days of my life! I said No to him once. I said No to him twice. Oh, uncle, if I could only have said it for the third time! But he followed me, he frightened me, he took away from me all the little will of my own that I had. He made me speak as he wished me to speak, and go where he wished me to go. No, no, no—don't come to me, uncle;
- 82. don't say any thing. He is gone; he is dead—I have got my release; I have given my pardon! Oh, if I could only go away and hide somewhere! All people's eyes seem to look through me; all people's words seem to threaten me. My heart has been weary ever since I was a young woman; and all these long, long years it has never got any rest. Hush! the man in the shop—I forgot the man in the shop. He will hear us; let us talk in a whisper. What made me break out so? I'm always wrong. Oh me! I'm wrong when I speak; I'm wrong when I say nothing; wherever I go and whatever I do, I'm not like other people. I seem never to have grown up in my mind since I was a little child. Hark! the man in the shop is moving—has he heard me? Oh, Uncle Joseph! do you think he has heard me? Looking hardly less startled than his niece, Uncle Joseph assured her that the door was solid, that the man's place in the shop was at some distance from it, and that it was impossible, even if he heard voices in the parlor, that he could also distinguish any words that were spoken in it. You are sure of that? she whispered, hurriedly. Yes, yes, you are sure of that, or you would not have told me so, would you? We may go on talking now. Not about my married life: that is buried and past. Say that I had some years of sorrow and suffering, which I deserved—say that I had other years of quiet, when I was living in service with masters and mistresses who were often kind to me when my fellow-servants were not—say just that much about my life, and it is saying enough. The trouble that I am in now, the trouble that brings me to you, goes back further than the years we have been talking about—goes back, back, back, Uncle Joseph, to the distant day when we last met. Goes back all through the sixteen years! exclaimed the old man, incredulously. Goes back, Sarah, even to the Long Ago! Even to that time. Uncle, you remember where I was living, and what had happened to me, when— When you came here in secret? When you asked me to hide you? That was the same week, Sarah, when your mistress died; your
- 83. 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