Changes in Wanka-Quechua

Studies in the Linguistic Sciences
Volume 4, Number 2, Fall, ^ol'*

MORPHOLOGICALLY CONDITIONED CHANGES IN WANKA-QUECHUA
Rodolfo Cerron-Palomino

Purpose

0.

.

The purpose of the present paper is twofold:

(a) to

provide evidence that shows the existence of genuine grammatically

conditioned sound changes; and (b) to construct a working hypothesis to
the effect that certain rules can become generalized from derived to

non-derived environments in the Kiparskyan sense.
in mind

I

V7ith

these two goals

shall present a number of instances of phonological change

affecting certain grammatical categories only, and at the end it will be
shown that some of these changes which began affecting suffixes, only
The

after eroding them propagate by generalization to the entire system.
data used come from the Wanka variety of Quechua.

For each change

I

will

try to delimit its exact distribiition interlectally as well as intra-

systematically, and references to similar changes attested in other Quechua
lects will also be made.*
0.1.

Lectal and Geographic Location

.

Wanka-Quechua is a variety

that belongs to one of the main branches (Quechua I) into which the Quechua

language family is divided.

It is spoken on the Central Highlands of

Peru, along the Mantaro Valley, covering roughly three provinces of the

Department of Junln (from South to North, Huancayo, Concepcion, and Jauja).
This variety consists of at least three lects

called (from South to North)

Waylla-Wanka, Waycha-Wanka, and Shawsha-V.'anka; these lects correspond
roughly to the political division among the three provinces mentioned
above.

Each lect, on the other heind, appears fragmented with respect

to the treatment of a niomber of phonological rules.

makes it difficult

—

if not impossible

—

This peculiar situation

to establish relatively discrete

varieties within the supralectal entity which we call Vanka (itself an

abstraction defined on the basis of certain shared miles)

imless one

embraces the rather unrealistic approach to language variation, namely

the so-called static paradigm , as opposed to the dynami c model
by the variously called sociolinguists

,

variationists

,

etc.

2

practiced

Thus, it

should be remarked at this point that the variety we are concerned with is
a highly fragmented entity.

In fact, the lectal differentiation is so

«n

strong that, to speak solely of the regional variation, not only do we
see differences from region to region hut also from town to town, from one

village to another.

As far as intelligibility is concerned, I must say that,

in spite of the surface differences, the speaker of different communolects

understand each other (although certain degrees of asymmetry can be
detected), being almost all of them, literally speaking, polylectals

This

.

knowledge on the part of the speakers is, I think, an aspect of their
linguistic competence, and we would like to capture precisely this knowledge, since that is the goal of a linguistic description.

To this effect,

one only needs to compare the varieties among themselves and thereby establish

the basic grammar from which, by means of rules of lectal differentiation,

we can derive the actual pronunciation for each variety.

Thios

in the

remainder of this paper I shaJ.1 assume the correctness of a panlectal
grammar, as opposed to an idiolectal grammar, without further discussion
1.0.

Grammatical Information and Sound Change

.

3
.

That some synchronic

phonological rules require, in order to be formulated, certain categorial

information has largely been accepted (sometimes by way of analo©r) even by
those who followed closely the strong version of the neogrammarian view
that sound change takes place only in pirely phonetic environments

.

This is

because in an historical perspective it is generally the case that most of
the grammatically conditioned rviles were originally phonetically motivated,

and as such do not contradict the "regularist" view of soxmd change.

On

the other hand, the possibility of a sound change morphologically conditioned

from its very beginning was rejected, since this goes against the very heart

of the neogrammarian doctrine.

However, there can be no doubt that

phonological change can take place in nccphonetic environments; in fact the
notion of word bovindary was largely invoked, although most of the times under
the disgaise of "phonetic paiises".

It is because of the existence of a

wide array of evidence that shows the grammatical conditioning of certain
phonological

rxiLes

that generative grammar maintains the view that soand

changes cannot always be stated in strictly phonetic terms, and therefore

this model has more empirical validity, as opposed to other theories strongly

limited by what Chomsky has called the "local determinacy" principle.

Here

it should be noted however that even if we know that the strong version of

the neogrammarian view is empirically incorrect, it "can hardly be disposed

42

of entirely, since its value

as

a heuristic principle is out of question.

This is because, as Anttila (1972:79) observes, "even if it is easy to

formulate a grammatically conditioned sound change, it need not be

historically correct".

As we have said, synchmr.ically

certain grammati-

,

cally conditioned alternations are the by-product of earlier perfectly

phonologic ally motivated rules.

This seems true, for example, of the

Iroquian sound changes reported by Postal (1968: Chap. 11 ) which were then

presented as evidence for the disconfirmation of the neogrammarian theory
of sound change.

Postal claimed that among certain Iroqaian languages

perfectly regular sound changes can be grammatically conditioned in origin.
However, Kiparsky (1973) demonstrated quite convincingly that the Iroquian
case discussed by Postal is actually "irrelevant to the neogrammarian

hypothesis, as it is not an example of sound change at all, but rather an
instance of the generalization of rules".

This of course does not mean

that there are no morphologically conditioned sound changes, but only
that we may not talk easily of grammatically conditioned changes unless

we have exhausted carefiolly all the possible phonetic conditioning factors.
Thus, as King (1969:138) remarks, "we should act as

i_f

every phonological

change were at most phonetically conditioned".

Granted the existence of nonphoneti cally conditioned sound changes, it
is however remarkable that, to our knowledge, most of the examples provided

in the literature are cases of inhibitory changes, exceptions to -recrul^r

sound changes, such as prevention of homophony, or retention of cat-^^orial
distinctions (gender, number, case, etc.).

The reason for this may be:

(a) that uncontrovertible instances of grammatical conditioning as causing

primary change are hard to find; and (b) that more often than not people
have carelessly invoked nonphonetic conditioning for phenomena they could
not explain otherwise (the Iroquian example cited .above

v.

ould be a good

case in point)

Below

I

present a number of cases of apparently genuine grammatically

conditioned changes.

If they are so,' then here we have good instances of

changes that contribute to the disconfirmation of the neogrammarian view of

sound change in its strong version.
1.1.

Glottal Absorpt ion.

U

In a relatively recent study

I

have presente

in detailed fashion the evolution of -/q/ among the Waylla and V/aycha lects

:

43

(called together henceforth Ya?a-Wanka , because here ya?a is the root for
the first person pronoun), leaving aside the treatment of the same proto-

phoneme in Shawsha-Wanka (called henceforth

Nuha-Wanka , because here nuxa

is used instead of ya?a ), since its reflex /x/ is found regularly with some

minor changes (cf. 2.1. below).

The treatment of */q/ among the Ya?a-Wanka

lects is interesting in a nvunber of respects.

Historically, the rule

involved was
(1)

v,/

>

I

%]

and the actual realization of /?/, according to the seven lects (ranging from

A to G) which

I

have identified (see 2.1. below), is subject to certain

phonologicsil rules, each of which can be considered a further step in the

change with respect ot the former, showing us a clear case of a sound change
The way this change propagates will be illustrated in the

in progress.

second section of this paper.

For our present purposes we only need to

introduce the treatment of the glottal stop in lect B, since A is the most
conservative in that it preserves the output of rule (l) unchanged.

Geographic Distribution

1.1.1.

This lect covers approximately the

.

territories of the districts of SapeuLlanga (including the locality of
Miluchaca) and Huayucachi (excepting the localities of Miraflores and

Huamanmarca)

both located at the sourthern border of the Mantaro valley,

,

in the province of Huancayo.

1.1.2.

Morphological Conditioning

that the output of rule (l

)

.

In lect B it can be observed

is maintained in careful speech, B thus being

in this sense similar to lect A.

In colloquial speech however we have the

following situation:
(2) :atu3

'fox'

Cyula:

'white'

Ccu?Iu+kaD

'the corn'

Cwa?la+na!]

'it xised to hook'

which in careful speech corresponds to (2a), respectively:
(2a) Catu?l

Cyula?:

Ccu?iu+ka?3
Cwa?la+na?3

nn

Thus, to give an account of this fact,

vre

may formulate the following

rule (leaving aside certain details which are irrelevant for our present

discussion)
(3)

/?/ ^ (0)

—

/

#

(where the use of parentheses indicate the stylistically controlled nature

of the rule).

However, in view of (k) it will be seen that rule (3) is

inadequate
Catu+miD

'(it is)

Cyula+ka+miD

(U)

'(it is) the white (one)'

fox'

Ccu?lu+kr+waQD 'with the com'
Cwa?la+na+cur)l

'did it use to hook?'

These forms correspond to the careful-speech forms of (Ua):
(i*a)

Catu?+mi:

CyulaT+kaT+miD
Ccu?iu+ka?+war)3

Cwa?la+na?+cuo3

Comparing (2) with (k) we see that the loss of the glottal stop in (h)
leaves as a reflex the lengthening of the preceding vowel, and therefore if

we do not want to miss the generalizaticn involved, we would like to set

sp

an intermediate stage for (2) whereby before the total loss of the glottal
a vowel lengthening takes palce, and only then the long vowel becomes

shortened by means of another rule, independently needed in the phonolo©r
Thus, the modified version of rule (3) would be:

of Wanka.
(5)

/V?/ - (v)

/—

^[^H

This riile will then give us an intermediate stage for (2), which we offer
as

(6):

(6) CatuD

Cyula]

Ccu?iutka]
Cwa?la+naD
and to arrive at (2) we need another rule, which shall be referred as Vowel

Shortening rule:
(7) V

-vv

/

—M

Note that this rule is obligatory once rule (5) has applied; otherwise
(6)

would constitute an incorrect output, or an ambiguous output at best.

.

45

As can be seen in the last two forms of (2), that is Ccu?iu+ka3

1.1.3.

and Cwa?la+naD, glottal stops that appear within roots are left unchanged
even in fast speech.

We must ask then what is so special in a root that

leaves the glottal stop preserved?

way of phonetic conditioning.
from Proto-Quechua,

thjs

The answer surely cannot be reached

by-

Note that the roots in question are inherited

ihe possibility of being borrowings that came into

the language after the completion of the vowel lengthening rule must be
discarded.

Note also that the data presented constitute evidence for a

—

change in its very beginning, since in the following lect

that is C

—

nole (5) is no longer a variable one, and its application is completely

obligatory; in other words, rule (5) has become categorical.

Moreover,

in lect C we find new cases of glottal elision, this time affecting the

This being so, we

roots themselves, again in the form of a variable riile.

can be sure that the change stated in (5) is morphologically conditioned

in its beginning.

The most important thing is that this change appears as

a variable rule, showing us an innovation in its embryonic stage, in this

sense, the possibility of a prior phonetic motivation has no grounds at

From these facts it seems clear that here we have a genuine morpho-

all.

logically conditioned change, since to state the rule we have to make use

of morphological information.
1.2.

The elision of /k/

following suffixes:
'accusative',

In Wanka, the phoneme /k/ drops in the

.

(a) - yka

(d) - ncik

'durative'

,

(b) - ka?

'determiner', (c) - Kta

'first person plural (inclxxsive)

'second person allocator'.

'

,

and (e) -yki

Whereas the first two changes appear to be

widespread and among different lects, th6 last three are limited to very
restricted areas.

In the following, I shall present the first two cases

separately, and then the remaining three will be treated together, since
they are found in rather small spots
1.2.1.

in two forms:

The durative -yka
-yka and -ya.

.

This modal is distributed among Wanka lects

Historically, they go back to a pre-Quechua

7
•"•-

yka

,

and the vowel was shortened

—

as it is today

—

I

in a checked syllable.

Since most suffixes in Quechua are of the pattern -C(V), it is evident
that originally this suffix was bimorphemic,

infinitive marker, and

-'-ka as

'"'-^

the verb 'to be'.

being identifiable as the
The way they become

univerbated and lost their original meaning is not our concern here; we only

46

need to point out that in the synchronic grammar of Wanka, as well as in
other Quechua lects, -yka indicates a continuous action either in the past,
present or future.

In the remaining section I assume, "without fvirther commen'

that the underlying form of this morpheme in Wanka is -yka


1.2.1.1.

.

Among the different lects of Wanka,

.

In fact, the

the innovated form covers the vast majority of territory.

-yka area forms a more or less elongated island surrounded by the territory

of -ya.

The island forms, so to speak, a bridge between the three major

Wanka lects.

Thus, going from South to North, -yka is preserved in the

communolects of Sicaya (Waylla-V/anka)

,

Orcotiina, Vixo, Huancani

,

and

a

probably Mito and Sincos (Waycha-Wanka)

;

then the isogloss continues to

the left side of the Mantaro River, beginning in the communolect of Huamali
to the North, up to the Yanamarca Valley, and to the East covering the

whole Mas ma Valley (Shawsha-Wanka)
"yaista" (that is, showing -ya)

,

.

The remaining area is entirely

except a small relic area on the mountedns

in the southwestern side of Huancayo, the communolect of Carhuacallanga,

where - yka is preserved, thus being a "kaista" area.
geographic distribution of the two variants of
1.2.1.2.

Morphological Conditioning .

th-e

This is, roughly, the

durative.

From the data

I

shall present

below it seems clear that the change that affects - yka is not due to
phonologically conditioned factors.

Thus, the only way to formulate a rule

that would account for the different realization of the durative among the

two isolects we have delimited would be by writing a morphologically
To convince ourselves of this, let us consider the

conditioned rule.

following roots, which are perfectly native words:
(8)

Cwayka-D

'to assaiilt'

Ccayku-D
Cyayku-

'to catch'
~

yayka-D 'to enter'

CaykaD

'how much/many'

Ctayka+ci-D

'to make someone to sit'

We cannot say that /k/ drops only preceded by

t>.e

semi consonant /y/,

because the sequence -yk- is regularly maintained in the roots of (8).
Furthermare, the augmentative modal - yku

yka (where the latter variant

appears when followed by certedn suffixes), which as we see has almost an

identical phonetic shape as the durative, preserves its /k/ in a very

3

'

'

U7

straightforward fashion.

9

Thus, compare (9) with (lO):

(9) -yku ~ -yka 'augmentative'

Clika+ykatma+Q3

'he/she suddenly saw me'

Clika+yku+13

'by seeing'

ChJnta+yka+?+mu+n+naI]

'

(it

suddenly became filled up'

)

(10) -yka ~ -ya 'durative'

Clika+ykatma+QD

~

Clika.+yka+13

Clika+ya+ma+nD

'he/she is seeing me'

Cliki+ya+lD

'being watching'

Chimta+yka+mu+n+naD ~ Chunta+ya+mu+n+fiaD '(it) is becoming filled
up already'
If it is the case that /k/ does not always drop after /y/, then it is
less probable that it can drop after another non-vocalic segment, as the

following situation illiistrates

.

Thus, /k/ is regularly maintained in the

durative-simviltative -cka, as well as in the dynamic - Iku

- Ika , as the

~

examples of (ll) and (12) show respectively:.
(11) Clula+cka+QkiD

'you will be doing while.

..

Cupya+cka+QkiD 'you will be drinking while.

Cawsa+cka+sa?D 'I'll be playing while ...
(12) Ctaka+lku+11

'beatihg'

Cmuyu+lku+lH

..

'

'turning'

Ctaki+lku+pti+Q]
'='-•

'once he/she sings'

Capa+ka+lka+pu+ia+y]

'please take (it) for him/her'

Thus, we have seen that it is iii5)OSsible to determine an exact phonoThis being so, the only way to write

logical environment where /k/ drops.

the rule involved would be by assigning to it morphological information,
'

by using the feature C+ DurativeD, as in (13):
(13) /y/ ->0

/

—

-;
C-r
+Durative

It follows then that a rule along these terms is part of the phonological

component of the yafsta lect; among the kaista communolects
hand, we assume that the

riile is

simply absent.

,

on the other

This means that these

latter commvinolects have the allomorph that is close to the xinderlying form.
1.2.1.3.

Phonetic Realizations

.

Whereas we have not much to say

about the phonetic realization of -ya among the yaista lects

—

it is

uniquely pronounced either as Cya3 or Cya3, according to the environment

48

we have mentioned in 1.2.1.

,

the way - yka is actualized in the kaista

lects however is interesting to note.

We shall have a look at these

phonetic variations.
Aside from the very general rule of yod elision, by which /y/ in

syllable- final position and preceded by the vowel

i_

is assimilated to

it causing the lengthening of the vowel (this, likan from /li+yka+n/

'he/she is going'), the most notorious change affecting -yka is the

palatalization of /k/.

This is true for almost all of the kaista area,

excepting the Masma Valley and some sporadic points within the Yanamarca
In these latter it is possible that the rule of palatalization

Valley.

involved is a variable one, its application being determined by the style

of speech used (casual speech favouring the innovated form).
point we may ask:

Is the process

At this

of palatalization affecting /k/ a

general process or a parti ciolar one that operates only on the durative

morpheme?
To answer the above question, let us first consider the case of the

communolect of Sicaya.

In this locality, we observe the following
(a) - yka ,

realizations of the durative:
- yea

(b) - yk a

,

(c) -yt a ,

and (d)

Although the exact distribution of these variants among the speakers

.

(taking into consideration age and style variables only) is far from clear,

we can safely say that they show a successive degree of palatalization.
To be more precise, the process involved is of the progressive type and
it does not operate when /k/ is in final syllable (thvis Clit uptikD

from /li+ku+ptitk/ 'if you go', CmitûncikD from /miku+ncik/ 'we (incl.)
eat', CtiktiD
- yka

'wart', etc.).

Thus in Sicaya this process not only affects

it is rather a general rule that applies across the board, whenever

;

its structiiral description is met:

when /n/ is mediating.
Cwasi+r)+kûnaD

*"

after /i, y/; for some speakers even

An' illxistration

Cwasi+n+tûnaD

~

of the latter case woiild be

Cwasi+n+cunal 'his Aier houses

Now let us consider other cases.

'

As I have stated earlier, the

Yanamarca Vedley also shows palatalization of

-'i^^h_.

However, here

I

have found only what appears to be the first degree of palatalization,

namely
suffix.

- yk'â

;

furthermore, this process 'only affect? this and only this

The same thing seems to happen in the cpmmunolect of Orcotuna

(Waycha-Wanka)

.

The present and other similar facts

suggest that within

U9

the kafsta territory, if the change continues, it will finfelly end up in
a true palatal, namely -yea , as in Sicaya and other areas of Quechua 1.

Thus, again, note that in Yanamarca and Orcotuna the palatalization of
/k/ seems to be morphologically conditioned, since it affects the durative
-yka only.

In view of this situation, it woioldn't be too daring to say

that in Sicaya the palatalization process began by idiosyncratically

affecting the durative morpheme and that it was only secondarily generalized
to other environments.

Another change that must be noted is that in one place, Acolla
(located in the Yanamarca Valley), I have found the variant -y^a side

by side with - yka , the former being

ased in

careless speech.

This,

I

think, is a nice illustration of the way /k/ may have disappeared among

the yaista commimolects

.

Furthermore, Torero (l96i+:i+TOA) also foimd the

same intermediate stage in some localities of Cajatambo (Lima) and

Yanacocha (Pasco).
change /k/

->

Thus I postulate that the intermediate step of the

was the yodization of the velar phoneme.

Finally, to end this section we shall note that, as has been suggested
in the preceding lines, the drop of /k/ is observed not only in Wanka but
also among other varieties.

Of special interest is the fact

tiiat

the

variety of Tarma and also that of Yauli, on the northern border of Wanka,
form an entirely yaista area (cf. note 8).

Thus it appears that the

kaista zone of Wanka can be considered a relic area with respect to the
preservation of /k/ in the durative.

Note furthermore that wherever the

velar drops we are in face of the same constraint: it occurs only in
the durative, and sometimes in the augmentative also.
The determiner - ka?

1.2.2.

.

Among the Quechua lects, Wanka is the

only one that has a determiner suffix.

Its origin is clear, since it

was a by-product of the univerbation of

-'ka-

the agentive

-?^

'to be'., nominalized by

(which goes, as any glottal stop in Ya?a-Wanka, back to

an original "/-q/), in periphrastic constructions like:

" wajpa

asi+yka+n 'he/she is looking for the (thing) that is chicken'.

ka+q+ta+m
Synchronically

it functions as a determiner, thus walpa means simply 'chicken', whereas

walpa+ka?

means

'the chicken'

(already known by the listener).

Although

its function as a determiner is beyond doubt, its syntax is far from clear.

Here

I

shall postulate - ka? as the underlying form of the determiner among

50

the Ya?a-Wanka lects

;

for Nuha-Wanka, on the other hand, - kax must be

assumed as the basic form (since here */q/

>

/x/).

This section will be

entirely devoted to the phonetic realization of the determiner among the

Ya?a-Wanka commxmolects

,

since the latter form, that is - kax , shows no

elision whatsoever of its /k/.
1.2.2.1.


As in the case of - yka , here

.

also we have a rather conservative area where /k/ never drops.

This is a

tiny area that covers approximately the localities of Sicaya (WayllaWanka)

,

Orcotxma, Vixo, Mito, Sincos and HuancanI (Waycha-Wanka)

latter on the borderline with Nuha-Wanka.
- ka? is

1.1.

,

this

The only change that affects

the loss of /?/, as it should be recalled from our discussion in

Thus, the loss of the glottal phoneme in syllable- final position

leads to compensatory vowel length (and, furthermore, the long vowel is

shortened before paise).
(li*)

Casnu+ka+ta3

This can be seen in the following examples:
'to the donkey'

Cyawar+ka+waQ3 'with the blood'
Cwalas+ka3

'the boy'

which come from imderlying forms such as those of

(l^+a)

(lUa) /asnu+ka?+ta/

/yawar+ka?+wan/
/walas+ka?/
and, as we already know, in Sicaya we may also expect forms such as
"^^
yv*^
Cwalmi+k a+taU ~ Cwalmi+t a+taD ~ Cwalmi+ca+taD 'to the woman', since

palatalization applies across the board.

The underlying form of this

expression is of course /walmitka?+ta/..
As in the case of - yka , there is also another relic area, and this
is again the communolect

of Carhuacallanga.

Here -ka? is preserved even

more intact, because /?/ is also maintained and palatalization is absent.

Thus the form observed in this area is close to the underlying representatic

we posited for the determiner.
Comparing the treatment of - yka and - ka? amonp; the conservative
communolects we note that if a lect preserves -yk " then it is likely
for the same lect to preserve also - ka?

;

the reverse situation is not

true, because - kax is regularly maintained in Nuha-Wanka, and yet we know
that in this area, there exists geographic alternation between -yka and
. 12
-ya.

51

Realization of -ka?

1.2.2.2*

.

with the exception of the two relic

areas mentioned above, in the remaining territory of Ya?a-Waiika the velar

phoneme of the determiner drops, although this time depending on the
environment.

I

offer below its various phonetic realizations.

(15) Cwamla+mi]

'the girl (is)'

Cwasi+maQl

'toward the house'

Cyaku+waf]3

'with the water'

Cwamla+kal

'the girl'

Cwalas+kaD

'the boy'

Cwalas+ka+miD 'the boy (is)'
If we were confined to use this and similar data it surely wouldn't be too

hard to arrive at a basic form -ka? (especially after having introduced
the glottal elision rule).

There is, however, an easier way of getting at

the latent form, and this is of eaurse by looking at the conservative lects.
Thus the problematic first tKree items of (15) are nicely solved by looking,
for example, at data from Orcotuna:
(l5a) Cwamla+ka+mi 3,.

Cwasi+ka+maQD
Cyaku+ka+waf)]

and if we compare (15-I5a) with the respective forms attested in Carhxiacallanga, we would be able to recover the glottal stop, otherwise realized
as vowel lengthening or as ^.

From (15) we note that:

/k/ drops when the stem to which the

(a)

determiner is added ends in a vowel, and (b) this is true only when - ka?
is not a closing morpheme.

Otherwise /k/ is preserved.

This is a categorical

rule, and a variable one only in communolects that remain in the neighbourhood

of localities which preserve

- ka?

.

This is clearly a transitional situation.

Thus, in Aco and San Jeronimo (Waycha-V/anka) I have found forms such as
Cwamla+ka+ta aya+mu+yD 'go call the girl', Cwamla+ka+waQ if+yH 'go with the
girl', together with Cwamla+ta aya+mu+yD and Cwamla+war) if+yD, respectively.

Turning now to (15), we must say that in order to derive it we have
to postulate an intermediate stage like (l5b):
(15b) Cwamla+a+mi]

Cwasi+a+maQ]

Cyaku+a+waoD

52

Cwamla+ka3
Cwalas+ka!]

(assuming that the vowel lengthening rule has operated already).
from these forms we arrive at (15 ) in the following way.

Thus,

To the first

three items of (l5h) we must apply a vowel contraction rule (needed

independently in the phonology of Ya?a-Wanka) whereby the vowel of the
determiner is eliminated in favor of the stem vowel; to the reamining two
items, on the other hand, we apply rule (T), that is, the vowel shortening

process.

Furthermore, it should be noted that the vowel contraction took

place first among homogeneous vowels, and only then among heterogeneous
ones.

I

have found communolects where only the first step was accomplished.

Thus, to give an example, in the locality of Ingenio (Waycha-Wanka) I

observed forms such as Cwalmi+a+ta aya+mutyl 'go call the woman',
Cci&u+a+ta apa+mu+yl 'bring the hat', etc.
(locality of Sincos, Waycha-Wanka).

The same is true in Chalhuas

Incidentally, note that the vowel

contraction rule is conspirative in nature, since in Quechua in general
sequences of vowels are prohibited.
1.2.2.2.2.

What seems to be a second step in the drop of /k/ is

illustrated in the following set.of items:
(l6) CweilaSTa+iniD

'the boy (is)'

Culpay+a+wEiQll

'with the dove'

Cwalas+kaH

'the boy'

Culpay+kaD

'the dove'

As we see, /k/ drops even when the stem to which the determinant is

added ends in a consonant.

Clearly we are in face of a rule generalization.

This second stage, however, is governed by a variable rule, its application

being determined depending on the style of speech used.

As usually, casual

speech tends to favor elision of /k/; in careful style however the full
form is axtomaticaHy recovered.
1.2.2.2.3.

Still a further step in the elision of /k/ is shown in

the following data:
(IT) Cwamla sa+la+mu+f5a3

'the girl had cone'

Cwalas+a sa+la'-mu+naD 'the boy had come'
(which come from underlying /wamla+ka? sa+latmutna?/ and /walas+ka?

sa+la+mu+na?/, respectively where we see that /k/ drops also when - ka?
is a closure morpheme.

Again, this is a further step in the propagation

:

'

:

53

of the change.

The generalization here involved is also a variable

rule, and furthermore it coexists with the situation described in the

preceding paragraph.

This is to say that the third stage in the rule

propagation does not presuppose the completion of the second stage.

In

other words, whereas the second step presupposes the completion of the
first one, the third stage does not imply that the second is already

accomplished.

In short, we can

s^

that

2=1,

but not that 3

=>

2.

To conclude this section, a word must be said with respect to the

sonorization of /k/ whenever it is preserved.

Thus, there exists in

Weuaka a voicing rule (see l.U. below) whereby /k/ becomes voiced, its

status being that of a variable rule, depending again on the style of

speech used.
-ka?

,

Voicing, however, is not restricted to the velar stop of

as in certain conmunolects

palatalization is limited to -yka ;

rather^ it applies across the board among the suffixes.
it is natural

This being so,

to expect that where voicing exists the allomorph

- ka is

actualized as -ga.
1.2.2.3.

Morphological Conditioninp;

As in the case of -yka , /k/

.

never drops in roots and in suffixes other than the determiner.

looking at the data given in 1.2.1.2.

pliis

Thus, by

the items provided in (l8), which

are all forms inherited from Proto-Quechua:
(18) :paka-]

'to hide'

Cpuka-D 'to blow'
Calka-] 'to wait'
CwaskaD

'

rope

and also in (19)
(19) Clumi+kuna:

'stones'

Cplasa+kamaD
ChatUQ+kaskaD

(- kuna 'plixral')

'up to (the) plaza'

(- kama *allative')

greater than'

(-kaska 'comparative')

we see that /k/ is intact, except that it becomes voiced where the vvle of

voicing is present.

Incidentally,

I

must note that voicing only would

affect the items of (l9), not those of (l8).

Still, to give a more

compelling example of how /k/ elision only affects - ka?
following forms
(20)

Chainpi+ku+?+mil 'the healer (is)'

Calkat?+naw3

'as it

were waiting'

Cmuka+?+nawD

'as it

were smoking'

,

consider the

54

(from Huacrapuquio) where the velar stop is maintained regularly.

Thus,

any attempt in trj'ing to determine a phonetic environment in which /k/

drops would immediately fail.

Here, as in the case of the durative, we have

to recognize that the only way to write the rule is by providing its

However, here

structural description with a grammatical information.
ends the parallelism we have been establishing between

- yka

and - ka?

,

for

in the former case we have no evidence whatsoever as to where and under

what conditions the velar phoneme dropped; in the latter, on the other
hand, we have good illustrations of the fact that /k/ began to drop

intervocalic ally , provided the determiner was not a closure morpheme.
Thus the data so far presented show us that /k/ began to be affected

after a vowel ending stem, and then we can also note its gradual propaga-

tion to other positions.

But even if we know that the original environ-

ment was postvocalic, we cannot avoid the conclusion that the change

For

ultimately affected idiosyncratically the determiner suffix only.

these reasons we must conclude that also here we are dealing with a
13
genuine morphologically conditioned phonological change.

Elision of /k/ in other suffixes

1.2.3.

.

Aside from the drop of

/k/ in the durative and the determiner, it also drops in the accusative

first person plural (inclusive) marker - ncik , and in the

- Kta , in the

second person allocator -yki

.

Below

I

shall consider briefly each of these

cases

1.2.3.1.

Ya?a-Wanka as

The accusative -Kta .
- kta

The accusative marker is realized in

when the stem to which it is attached ends in a short

vowel, and as -ta otherwise (including after long vowels).

This synchronic

rule is however an historical one ainong all the rest of the varieties

of Quechua, including the Nxjha-Wanka lect.
is the only

Thus, the YaTa-Wanka variety

Quechua lect that preserves - kta in its underlying form.

How-

ever, among the northeastern communolects of Waylla-Wanka the allomorph
-kta has

been leveled to -ta, its alternation with

- kta

thereby disappearing.

These commiinolects are Santo Domingo de Acdbamba, Andamarca, Comas

,

and

Cochas, all of them in the neighbourhood of Nuha-V7anka, which only has -ta.
Thus the easiest explanation for the elimination of - kta would be to say

that it is due to influence from that lect; furthermore, this hypothesis
is reinforced by the

fact that in another communolect, Chalhuas

55

(Waycha-Wanka)

,

located on the western border with Nxjha-Wanka, I have

found only -ta. However, even without recourse to extrasysteraatic

explanations, we can safely say that the change -kta

->

-ta^ is

due to the

universal principle of minimization of allomorphic alternation.

But even

if we say that the loss of /k/ in the accusative variant - kta is due to
restructuring, v^ have to admit that originally this change was also

morphologically conditioned in all of the Q;:echua lects, since, as far as
I

know, there never was a rule that otheivise affected the sequence

-kt-.

Synchronically, the fact that this sequence remains unaffected is

illustrated in the following set of native roots:
(21) ClaktaD

'thick'

Ctuktu-3 'to be asquat'
CtfktiD
1.2.3.2.

'wart', etc.

The plural marker -ncik

we have leveling of -kta

->

In the same localities where

.

-ta, I found that the velar phoneme of the

first person plural (inclxisive) marker also drops.

—

gorical one, except in one commianolect
it is found only in careless speedi.

The rule is a cate-

—

Comas (V7aylla-Wanka)

where

Furthermore, the same change was

also noticed in Roncha (locality of San Jose de Quero, W^lla-Wanka)

the western side of Waylla-Wanka.

,

Again, it cannot be sedd that /k/

on
.

regxilarly drops in final position, since froms such as:

(22) CcOsikD

'a variety of owl'

Cli+ku+pti+kD 'if you go'

—

not to speak of final /k/ preceded by a vowel other than

the velar phoneme is maintained intact.

i_

—

show that

It can be argued perhaps that

in the latter form /k/ is a morpheme that marks second person, that being

the reason why the velar does not drop, but what about the first item?
Here /k/ does not mean anything, nevertheless it is preserved.

Once more,
Ik

we are here confronted with a morphologically conditioned change.
1.2.2.3.

The second person allocator -yki

.

Among the northermidst

communolects of Shawsha-Wanka (in Ricran, for example)

I

change that consists in the syncope of the syllable

of

through an intermediate step -yki

->

-yi

->

-^.

ki_

have noticed a
- yki

,

perhaps

Thus, accordingly, we

have: Ccakla+y+taD from /cakla-yki+ta/ 'your (ace.) farm', Cmisa+y+mar)D

from /misa+yki+man/ 'on your table', etc.

The conditions under which this

56

change operates need to be investigated further, but what is beyond doubt
is that the change affects that suffix only, similar sequences being

conspicuously maintained.

Furthermore, this change seems to be more
Thus, here we

general (categorical) in the neighbouring iect of Tartoa.

have forms such as:

Cqam+qa+ta yali+sa+y+miD, Cpaga+sa+y ali+ta+mD from

/qamtqa+ta yali+sa+yki+m/ 'I'm going to win you', /paga+sa+yki ali+ta+m/
'I'm going to pay you a lot', respectively.

Velar Vcicing

1.3.

.

Except for borrowings from Spanish, Quechua in

general does not have voiced stops, although some lects

of Quechioa II

—

—

especially those

have developed them as a result of phonological changes

not only as voiced allophones, but also as underlyingly distinctively

voiced phonemes.

In the Wanka lects I have found a variable rule

affecting the velar /k/, but, as we shall see, the rxHe operates in derived
forms only, whereas the roots remain unchanged.

1.3.1.

Geographic distribution

.

The voicing rule is observed in the

Ya?a-Wanka lects, thus covering part of the provinces of Huancayo and
Concepcion.

From South to North, the change begins in the communolects of

Cocharcas (district of Sapallanga) and Huamarimarca (district of Huayucachi);
it continues covering theleft side of the Mantaro River up to the district

of Cajas Chico (Huancayo).

From here the isogloss, on the one hand

crosses the river, reaches the communolect of Huamancaca Chico, and extends

through the districts of Chupaca, Pilcomayo and Ahuac, excluding the
districts of Sicaya, Orcotuna (including Vixo), Mito, Sincos

sind

Huancani,

up to the border with Shawsha-VJanka; on the other hand, it continues

covering the remaining districts of Huancayo and Concepcion up to the border
with Shawsha-V/anka.

In this way the voicing area forms two islands of

unvoicing territories, one at the southwestern side of the Valley, and the
other at the banks of the river, in the northern territory of Ya?a-V/anka,
both islands on the right side of the river.
1.3.2.


.

To see how this rule operates,

let us consider the following data:
(23)' Cwaipa+ga wanu+gu+r)D

Caf)ka+guna+ga pali + gu+oD

'the chicken died'
'the buzzards flew'

Ccupaka+gama li+ku+ya+lga+Q] 'they are going up to Chupaca'
Ctaki+lgu+1+mi mika+pagu+Q]

'they use to eat singing'

t

57

which in carefvil speech is realized as (23a):
(23a) Cwalpa+ka wanu+ku+r)D

Carjka+kuna+ka pali+ku+Q3
C cugpaka+kama

li+ku+ya+lka+r)3

Ctaki+lku+1+mi mika+paku+r)D
As we can see, in (23) only the suffixes are affected by the rtile;
l6
the roots, on the other hand, are left unchanged.
Notice that here, as

in the other cases, the roots are pat rimoni ally inherited from Proto-Quechua.

Thus, again, there is no way of formulating this rule unless we make use

of categorial information; this can be done as follows:

where the parenthesis notation indicates that this rule is cohditioned by
the style of speech used.

As usual, fast speech tends to favor the change

Incidentally, we shovild note that this change is almost cateSorical in

certedn communolects and for certain speakers (this was noted especially in

Huarivilca and in Cochas Chico, Huancayo).

This is so, because in these

areas the speakers were aware that they pronounce

£

where others have k,

thus laughing at those who use the latter sound; this is not true for the

speakers who manifestedly make variable use of the rule, since they accept

both pronunciations while still considering the "older way" better.

From

the data we have at hand it is difficult to infer what was the primary

conditioning factor for voicing (even in that restricted environment); it
may be the case that the rule started intervocalically and then spread to

after voiced consonants, and so on; a careful study may perhaps detect this
possibility.

Whatever the inception might have, been, there is no doubt

that the change affects suffixes only; as such, this is another

exaii5)le

showing that there are morphologically conditioned changes.
l.U.

Depalatalization of /n/

.

According to Parker (1971:66-70), the

depalatalization of /n/ is a general change that covers almost the entire
territory of Quechua

I,

except the lect of Bancs Rondos and the peripheral

lects of Corongo (to the North) and Wanka (to the South).

lect however

I

have found the same

riile,

two suffixes which contain that phoneme:

-na 'inceptive'.

In this latter

but this time affecting, only the
- na?

'narrative past', and

58

l.U.l.


The area where the two mentioned

.

suffixes appear depalatalized covers the territory of Waycha-VJanka, and
also the northeastern side of VJaylla-Wanka, up to the border with ShawshaWanka.

—

Aside from these areas,

I

have also found depalatalization of /n/

this time not only restricted to the suffixes

—

villages of the district of Ricran (Shawsha-V/anka)
can'

.

among the northernmost
This latter situation

perhaps be attributed to influence from the neighboviring lect of

Tarma which shows the rule in its generalized version; for this reason

will ignore this latter area, and deal with the former areas only.

I

I

of the suffixes are - na? and -na,

assuiEe that here the linderlying forms

in view of the interlectal alternation.
l.U.2.

Grammatical Conditioning

.

That the rule of depalatalization

affects only the suffixes and not the roots can be seen in the following
examples:
'he/she was burning (it)'

(25) Ckana+ya+naD

Craana+mu+naD

'he/she had asked for (it)'

Cwayta+yatna+naD

'it was

flourishing already'

Chunu+naku+ya+lka+o+naD 'they are congregating already'
Contrast this situation to that of Huacrapuquio , for example, where we
have the fully preserved forms:
(25a)

Ckana+ya+na?3

Cmana+mu+na?D

Cwayta+ya+na?+naD

Chunu+naku+ya+lka+Q+naD
Thus, it is clear that in (25) we have another example of a nonphonetically

conditioned sound change, since there seems to be nothing special among the
roots that can prevent a rule from taking place.

Therefore, in order to

formulate the rule involved we are forced, again, to mal^e use of grammatical
Thus we can formulate the rule of depalatalization as follows:

information.
(26)

/n/

->

n /

:

+

TTT-1

suffix

It should be noted that whereas this rule is

categorical for the

narrative past, it is not so for the inceptive, since
lects where

- na

adternates with - na (for

exaii5)le

I

have found cocmuno-

in Ingenio and Andameirca)

The implication of this is that perhaps the depalatedization of - na? is

59

This is further confirmed by the fact that

earlier than that of -na.

among the localities of Masma and Molinos (Shawsha-V.'anka) the realization

of the narrative past is -nax, whereas the inceptive remains as -na.
Furthermore, an interesting situation that shows the different treatIn San

ment of /n/ within suffixes and within roots is the follov;ing.

Jeronimo

have noticed expressions such as:

I

Cna+mi li+ku+yl+na] from

/na+mi li+ku+yka+na/ 'he/she was leaving already', where na^ as an adverb

(historically related to the suffix.) maintains the palatal nasal, whereas
In the second section of this paper I will

the suffix shows the change.

have occasion of returning to this point; for the moment it may suffice to
say that again we are confronted with a grammatically conditioned sotind

change.

The Monophthong! zati on of /aw/

1.5.
I lects

,

.

In Wanlca, as in other Quechua

there are two suffixes which contain this sequence:

- caw , and the comparative - naw

.

the locative

These forms are monophthongized as -cu

(Among the Nuha-VFanka commimolects the second

and -nu, respectively.

suffix does not exist; here - nas is the equivalent (related?) morpheme).
The monophthongization of /aw/ is also found in the Ancash variety of

Quechua I, where it not only affects the suffixes mentioned, but the entire
lexicon.

Parker (1971:78-82) gives the exact distribution of this change

which he considers functionally related to the monophthongization of
/ay/

-*

e_,

and /ty/, /iy/

1.5.1.

-*

£.

In this area /aw/ is monophthongized as £.


.

In general, the diphthongal form

of the suffixes is found in all three Wanka lects.

Thus in Waylla-V/anka

we find it in the most cohservative areas, namely Huacrapuquio and
Carhuacallanga, both located at the extreme southern border of the province

of Huancayo.

We also

find another Waylla coramunolect, bordering the

Waycha variety, and this

is'

Sicaya, where however only sporadically

I

hear the locative as - caw , although - naw is the only realization of the
comparative.

Fluctuating areas contiguous to Huacrapuquio and Sicaya have

been located, and these are Viques and Vixo, respectively.

Here

I

found

variably - co and - no in nonfinal position, and - co and - no word finally
(recgill the vowel

shortening rule stated in I.I.2.).

In Waycha-Wanka

.

I

have found - caw and - naw only among the communolects of the right side of
the river, beginning in Orcotuna and ending in Huancani

,

at the

border with

60

In this latter lect

Nuha-Wanka.
- caw

find

—

ve

again only among the communolects of the left side of the
Elsewhere, we find - cu

Mantaro River,, up to the district of Paccha.
- cu

—

since - nav does not exist

and - nu

-nu, that is the monophthongized forms.

~

are - caw

that the underlying forms of the suffixes
1.5.2.


To

.

se'?

Here

and

- naw

shall assume

I
,

~

respectively.

how this change operates,

affecting the suffixes only, let us consider the following examples:
(27) Ccawpi+cuD

'in the middle'

Cnai<rpa+cu+mill

'in advance, before'

Ckawpu+nu]

'like a spin'

Cpinaw+nu+miD 'like a (variety of) herb'
As we can see, only the suffixes are affected.

Thus clearly we have again

a case of morphologically conditioned phonological chsinge.

17

The only

way to formulate the rule involved is by restricting its application to
Thus, we would have something

the locative and comparative morphemes.
like (28):
(2S) /aw/ - u

1.5.

3.

Z^r^^^^^l

/

Phonetic Realizations

.

transitional areas where we find -co

As I have said in 1.5-1., there are
~

- co

and -no_

~

-no

.

Thus obviously

an intermediate step in the change is an assimilation of the vowel a to

the following semi consonant, that is - cow and - now

monophthongization in

5_.

.

A second stage is the

Now, since among the V/anka lects we do not have

mid vowels (not even allophonically as it is the case in other varieties

of Quechua that retain /q/, a lowering phoneme par excellence )
obvious solution

—

,

the
1

as

to rise this derived orphan vowel, and thus we have the forms - cu

and - nu

~

Huayucachi
- nuy

.

-nu.
,

R

—

in the treatment of borrowings from Spanish
~

is

- cu

However, in certain communolects of Waylla-Wanka (such as

Chongos Bajo, Chacaparapa, etc.)

How do we explain these latter form.s

is an excrescent phoneme,

found the variants - cuy and

I
?

I

believe that here the yod

for similar cases are not difficult to find

elsewhere among the Wanka lects.

Thus, we have variants such as cuylu

alternating with culu from /cu?iu/ 'corn'

,

biylu side by side with bulu

from Spanish bollo 'a special type of sweet bread', etc.

These variants

with yod are less generalized, and in one and the same communolect we can

61

find both, the forms with yod and the monophthongized versions.

Finally,

as has been said elsewhere in this paper, the allomorphs with short

vowel are the product of the application of the vowel shortening rule;
thus we have here another independent motivation for this rule.

second section,

I

Nasal Absorption

1.6,

In the

shall return to this change briefly.
In all Quechua varieties there is a suffix

.

called directional, realized as -mu; this morpheme indicates the beginning

of an action in a place other than where it is spoken of, and with verbs of
motion it also indicates the direction towards the place of speaking.

This

suffix changes to -w among certain communolects of Waylla-V/anka by a
variable rule, as we shall see.
1.6.1.


The change - mu * -w is found among

.

the communolects of Sicaya and neighbouring commionities

Juan de Jarpa, in the Way 11a lect.

,

especially San

Aside from these areas, we find

constantly - mu in the remaining territories of the Wanka lect.

As far
19

as I know, this change is unique, to be found only in Waylla-Wanka.

1.6.2.


.

affecting the directional - mu only, let

To see how this change operates
vis

consider the following cases

which correspond to the colloquial style, as observed in Sicaya:
(29) Csa+w+la+ncikH

'we (incl.)

came'

Capa+ka+w+laD

'I

Casl+w+saD

'I'll look for (it) there'

Chula+la+wD

'he/she removed (it) from there'

brought (it) here for me'

Cyala+li+w3

'they got off from there'

Capa+ka+w+ki+maQD

'you should have brought (it) from there'

Ctalpu+ka+la+w+cikH 'we (incl.) have planted (it) over there'
forms which in a more careful speech correspond to those listed in (29a):
(29a) Csa+mu+la+ncikD
C

apa+ka+mii+la3

Casi+mu+sa3

Chula+la+mu+QD

•
,

Cyala+lf tmu+o3
Capa+ka+mu+Qki+maj)

]

Ctalpu+ka+la+mu+ncik3

62

As we can see, the nasal /m/ is dropped in front of the vowel /u/

which in turn is semi vocalized (or desyllabified)

,

this latter as a

result of the general principle whereby tn Quechua sequences of vowels
are rejected.

The first three examples in (29) show the change -mu

-*

-w; but, as it should have been noted, in the last four items we see not

only that the nasal /m/ is dropped, but also the same is true for the
-w;

nasal /n/ of the next suffix.

Thus we have here a change -nm+n

whereby both nasals

That it is only the following nasal and

eire

lost.

-»•

not any other consonant can be seen from the following examples:

Cca+la+mu+pti+Ol]

'if he/she arrives'

Cyayka+mu+yD

'come inl'

Cca+lka+mu+lD

(30)

'after arriving'

where the variant -w does not appear at all, not even in the most
inreflecting speech.

This is perhaps because the result of the change woul

be an intolerable cluster, completely prohibited in Quechua.
the nasal /n/ has proved to be the weakest of the resonants

segments such as

^^^

2.

incidentally, that

I

i. ^J"^

resistant to the change.

It

,

In this sense

since

should be noted,

have found no traces of a previous vowel nasalization.

Thus we do not have intermediate stages such as -mu or -w; rather, it

seems that the loss phenomenon is abrupt, since this change is governed

by a variable rule.
We must note that not only does - mu not change in checked syllables
(where the final consonant is not a nasal), it also does not operate

whenever the loss woiiLd destroy semantic information.

Thus from ca+la+mu ,

which comes from an underlying form /ca+l?a+mu+v/ 'I arrived', a possible
oxitp^it

such as "ca+la+w would be ambiguoiis since it would mean also 'he/she
Here we have a clear case of prevention of

arrived', from /ca+l?a+mu+n/.

homonyny, although there are abundant cases in Wanka which show that the
danger of homonyny is not a sufficient reason for blocking application of
It may

a rule.

well be the case also that a long vowel, which in Wanka som

times behaves as a VC sequence, is preventing the change.

V/hatever the

reason, the only way to formilate the rule involved is as follovrs
(31) /mu(n)/

-»

(w)

/

C--r:
+

-:

:

r3 |"

Directional

'

II

(where

"

i

means open syllable).

This is, as

I

have stated earlier, a

63

variable rule; and especially in Sicaya it is turning into a categorical
That the rule is grammatically conditioned is corroborated by the

one.

follovdng items:
(32) Camu+laD

'he/she used to have (it) in his/her mouth'

Chamu+ya+ma+na3 'he/she was blaming me'
Ckumu+oya+oD

'he/she starts to bow down'

Camutca+ya3

'I'm ferfcilizing (it)'

Ciamu+m3

'(it is) obtuse'

where the seq^uence mu(n) is never affected,, not even in the most careless
speech.

Thus, the obvious conclusion is that here we have another case of

genuine grammatically conditioned rule.
Glottal Metathesis

1.7.
is one

of the reflexes of

.

*/q,/

As we have seen in 1.1., the glottal stop

in Ya?a-Wanka.

Proto-Quechua suffixes ^-rgJJ 'simultative*

Thus the reflexes of the

aind ''- sqa

'participle' are,

respectively, - ?1U and - ?sa ,. showing a clear case of leftward metathesis.

However, as we shall see, the reflex of the simple past suffix "- rqa is not
•'-

tl& but - l?a , that is without metathesis.

are historically bimorphemic, where

^'-r^

Incidentally, these suffixes

and "-£ were aspect markers.

more, the' vowel of the simultative is realized normally as

vi,

but as

Furthera^

if

followed by certain suffixes (thus behaving in the same way as the augmentative -ykU (cf. 1.2.2.2.).
1.7.1.


derived forms are found only

in,

The metathesized suffixes and their

.

Ya?a-Wanka,.

In Nuha-Wanka, on the other

hand, the reflexes of the mentioned suffixes are, respectively, - lU and
-sa, showing no reflex whatsoever of the ''Vq/j in spite of the fact tiiat
in this lect, as I stated earlier, the uvular stop regularly went to /x/.

The explanation for this is that here the suffixes (including the simple
past) were subject to a cluster simplification rule which is found elsewhere

among the Quechua lects, either as an accomplished process or as a
variable rule (see 2.U. below).

To complicate matters, however, I have

noticed also in Ya?a-Wanka reflexes of the simplified form of the simiiltative;
but since

I do

not have a detailed morphological information of this variety,

it is difficult to get a clear understanSimg of the situation.

1.7.2.


.

As stated earlier, glottal metathesis

affected the simultative and the participle suffixes only; the simple past.

6I»

as well as similar sequences within the roots, are left untouched.
coinp9,re

Thus,

(33) with {3h):

(33) Cmiku+Tlu+oD

'he/she ate (it)'

Cpimu+?lu+l?aD

'he/she slept (there) for a while'

Clikali+Tla+li+mu+QkiD 'you (all) appeared suddenly'
Cpunu+l?a:

(3^*)

'I

slept'

Cawsa+l?a+nki] 'you played'
.

Cmiku+l?a]

'be/she ate'

where we see that only the simultative was metathesized.

Note particularly

the second item of (33), where we find the simple past following the

simultative.

Thus, we may question, why is that metathesis applied only

to the simultative and not to the simple past?

The same question holds

for the forms listed in (35):
(35)

Chul?u-: 'to remove'
Cul?u:

'hill'

Cai?u:

'dog'

Cpil?a:

'wall'

Csio?aIl

''nose'

where we see that the sequence

C?_,

particularly

1?_,

does not metathesize.

Now, let us consider the items of (36):
(36) Cmanca+?saD (ka+yka+n)

'he/she is scared'

Cwanu+?saD

(ka+yka+n)

'he/she is dead'

Cpakf+?sal)

(ka+yka+n)

'it is broken'

and compare them to the items of (37):
(37) Epis?u:

'bird'

Cp(ic?u:

'bitter'

Cac?aD

'rough'

Cay?iD

'to escape'

As can be seen, once more metathesis applies only to a particular category,

and similar sequences occurring within roots are left vmchanged.

Thus

,

formulation of the rule for both cases would have this form:
(38)

/CU

>

/?C/

/

<,

simultative>
I

<+ Participial>J

To be true, however,

I

must say that in Vaylla-Wanka

I

foxind the

doublet milay from /mi?iay/ 'lap' (noted in Sicaya), and mllav from

.

a

65

/mii?ay/ elsewhere (cf. Huacrapuqioio mii?ay )

Proto-Quechua 'milqay
''

thesized version.
applied to a root.

.

this form goes back to

;

As we can see, the first alternant is a meta-

This is the only example I have found where metathesis

Furthermore, the normeil development from

know, reckon' woxold have been in Ya?a-Wanka *llsi_- (or

" riqsi-

"''liTsi-

'to

among the

communolects which preserve the glottal stop, as one finds lixsi- in
most Nuha-Wanka varieties); however, what we find elsewhere in Ya?a-Wanka
is lisi- from /lis?i-/ (cf. Huacrapuquio lis?i-

communolect of Paccha, Jauja).

,

and also lisxi- in the

These two cases then seem to reduce the

validity of this example as another case of grammatically conditioned

sound change; however, in view of the overall evidence presented in this
paper for morphologically conditioned primary changes being limited to
suffixes, it is at least remarkable that the glottal metathesis is most

widespread and systematic in certain suffixes.
A language particular universal

2.0.

paper

I

.

In the first section of this

have presented several cases of morphologically conditioned

phonological changes.

We have seen also that some of these are attributable

not only to Wanka, but parallel changes taking place

cf both Quechua

I

and II can be detected.

amairtg

different lects

However, some of them are

observed in Wanka only; this can be explained partially by the fact that
the change takes place within a specific suffix that is labking among

other lects (e.g. the determiner - ka? )

,

or else by the fact that the

change affects a sound which is limited to this lect (e.g. the glottal
stop).

In any case, however, I have tried to demonstrate that it sing)ly is

impossible to detect a purely phonetic conditioning for these changes, the
natural concliasion

thjs

being that they are grammatically conditioned

changes; that is, in order to formulate the rules involved in each case
one has to state the domain of its application by providing a morphological

information to its structural description.

Moreover, it was shown that

most of these rules are not morphologized versions of prior phonologically

determined

riiles

.

Now, in view of the facts we have observed I would like to formulate

a working hypothesis according to which there is tendency in Quechua for a
chEinge to

begin affecting suffixes only, and then to spread to roots.

fact is familiar to Quechuanists

,

This

who either implicitly or explicitly have

66

assumed morphologically conditioned changes.
tendency may be remains a cystery;

thxis

21

Whatever the reason for that

in this final section I will

limit myself to illiostrating this type of rule generalization by presenting

vivid exan5)les of sound change in progress

If

.

I

succeed in demonstrating

not only the type of change we are dealing with but also its propagation

from the suffixes to the entire lexicon, then Quechua would be a clear
coxinterexample to Kiparsky's (19T3) claim that no primary' sound change can

depend on morpheme boundeiries.

What is more important, the cases presented

here would challenge Kiparsky's hypothesis according to which all phono-

logical rules that depend on morpheme boundaries arise by generalization
from non-derived to derived environments.

that a
thxas

riale

The examples I will provide show

can become generalized from derived to nonderived environments,

falsifying Kiparsky's claim.
2.1.

The generalization of glottal absorption

.

Below

offer a

I

lectal matrix that shews the generalization of the rule of glottal

absorption among the Ya?a-Wanka lects, as presented in Cerron-Palomino
(1973:79).

The interpretation of the matrix is as follows.

A-G and the rules (l) - (6) form
following way:

exi

iraplicational pattern

22

The lects
in the

G =>F =>E s>D =>C =>B =>A; that is, lect G has accomplished

the processes that took place in F; this, in turn, has the changes

observed in lect E, and so on, but not conversely; that is, lect F for
example does not have yet
rule ik)

y

rxole

(5), and lect E in turn does not have

whereas this change is categorical in F, etc.. This means,

consequently, that rule (5) implies rule (U), and this in turn implies
(3),

and so oh.

In other words, each one of the lects except the following

constitutes a recapitulation of the path followed by the immediately

following process of change.

Taking into account the temporal axis, this

means that in the implicational pattern G =>F =>E =>D =>C =>B =>A, what
is

This means that the

implied by the other is anterior to the latter.

change (l) clearly was anterior to rule (2), this in turn preceded (3),
This

and so on, until we research the most recent rule, that is (6).
can be formulated in the following way:

(l)

>

(2)

>

(3)

>

(h)

>

(5) >

(6), that is what is less than the other is posterior to the former.

68

perhaps

"be

argued that since word boundary' coincides sometimes with actual

pause, the change began originally in the prepaused (phonetic?) environmen-

and subsequently propagated by rule generalization to other environments.

The change therefore would not be a good example of genuine morphologicall:

conditioned change.

However, recall that the change takes place not only

before actual pause but also in certain suffixes where such a pause simply
does not exist.

Therefore, a natural concliisipn seems to be that the

change in qviestion

is_

grammatically conditioned from its inception.

Notice furthermore that a similar change is taking place in Kuha-Wanka,

where a variable rule governs the loss of /x/ (counterpart of /?/, both
coming from */q/).

Thus we have cases like CatuxD 'fox', CyulaxD 'white'

side by side with Csulxu+saD from /sulxu+sax/ 'I'll remove (it)', but

Csulxu+sa +mi3 'I'll remove (it)'.

We see in these examples that whereas

/x/ drops in the suffix - sax 'first person future'

(as well as in other

suffixes that contain a final /x/), the same is not true not only for the
lexiceLL items

atux a nd yulax , but also for the verb root sulxu- 'to remove

where the velar phoneme remains untouched; note also that the velar is

changing into aspiration.
2.2.

The depalatalization of /n/

change affects suffixes only.

.

As we have seen in I.U., this

Of the two suffixes involved, we noted

that the narrative - na? is the most constant in the change, whereas the

inceptive - na sometimes alternates with the anchanged form.
is, however, a general one among the northern Quechua I lects

The rule
,

beginning

from Tarma and Yauli (which borders Nuha-Wanka) up to the limits with the
dialects of Quechua II.

While in Nuha-Wanka, except for two spots

(Masma and Molinos, where we find - nax ) and the bordering villages of
Ricran, there are no traces of this change, in Waycha-Wanka however we

find it.

As there is no contact at edl between this lect and the rest

of the depalatalizing lects of QI, we must assume that the change is an
Parker (1971:66-70) studying this change among the

independent drift.

QI area of Ancash (cf. his table h) concludes that "everything seems to

indicate that the change began in the environment of a morpheme boundary";
thus in his table

k

(p.

68) the two suffixes - na and - fiaq show the most

widespread cheuige, whereas the generalization across the lexicon is
gradually less diffused, and at least in one case (the word natin 'liver')

69

the rule is a variable one in the most innovating lect.
Thus, as in the former case, here again we see how a rule spreads

leftward, from the suffixes to the roots, that is from derived to non-'

derived environments.

In the Wanka case one may ask why the change seems

to be, so to speak, fossilized in the two suffixes only.

To this we

may answer first of all that it is not true that the chemge is entirely
frozen, since we noted thet the inceptive is sometimes vised in its

unchanged form; secondly, of the three Wanka lects, the Waycha variety is
disappearing (recall note 7) rapidly due to the increasing predominance

of Spanish.

These two facts,

I

think, explain perfectly the apparent

stagnation of the change, and on the other hand do not invalidate ny claim.
2.3.

From the examples discussed in the two preceding sections, it

wouldn't be too adventurous to say that the remaining cases studied are
also examples of sound changes which began applying to suffixes, and

secondarily began to affect the roots (an exception being the glottal
metathesis rule which seems completely petrified).

What is important to

note is the fact that none of these instances show traces of a morphologi-

zation of phonological rules.
2.k.

Aside from the depalatalization of /n/ among the Quechua

I

lects of Ancash which Parker demonstrated to have begun "in the environment

of a morpheme boundary", there are other changes that have the same type

of restrictions.

Thus Parker (1971:67-68) says:

"That a change should

begin in suffixes and then spread to roots is not surprising if we know
that two other changes in Quechua are known to have the same constraint:
(1) *uy > i in parts

Bolivian (...)".

of Ancash (...); and (2) final *p

>

jc

in Cuz co-

With respect to the first change he mentioned (p. 80)

that it has "the peculiar constraint that it operates only when a morpheme

boundary (or internal word boundary) is adjacent to //y//"; thus "it does,
not affect root-internal //ijy//", as in /huytu/ 'oblong', /liycu/ 'deer',
etc.

Furthermore, there is among different varieties of Quechua a cluster

simplification process which operates in suffixes only.
(1971:98) states:

Thus Parker

"In Quechua generally there is a tendency to simplify

suffixes and combinations of the shape -CCV, almost always by dropping

the second consonant".

He then proceeds to list the following cases:

70

Bolivian - sa

— sa

<

"- cka

(also observed in Tarma,

'durative', Cuzco -yU

<

•-ykU 'augmentative'

cf. note 8), Ayacucho -ra ~- rqa

<

"- rqa 'past,

definite tense' (also observed in Wanka, cf. the discussion in l.T-l.)>

and Junln -ya

"-yka 'durative'

<

also mentions

(a change we are familiar with).

He

the fact that in the coinnunolect of Antonio Raymondi

"- rqa , "- rqu , and "- sqa (cf. with the situation in Nuha-Wanka in l.T.l.)»

all of them tense markers, "lose

'"'q

regularly by a variable rule".

All these changes listed suggest that the tendency of sound change

beginning in suffixes is perhaps a language-specific universal in all of
Quechua lects

23
.

As I have stated earlier, the reason why this is so

needs to be carefully explored, and perhaps the explanation might have

something to do with the agglutinative character of this language.

Thus,

an urgent task would be to look for similar types of changes among

languages of the Quechua type.

As Parker (1971:99) says, "changes of

this type deserve greater attention than. they have received to date".

FOOTNOTES
*
I am grateful to Professor Hans H. Hock for his criticisms and
for suggesting to me many improvements, both in form and content, for
this paper. His ideas will be apparent at many places in this paper,
nevertheless it would be unaccurate to say that he agrees with everything
I say.
I should also thank Professor Charles W. Kisseberth for his
helpful comments. All errors of fact and/or interpretation are my own.

Since the term "dialect" is rather misleading, sometimes being
defined in terms of extrelingmstic criteria, the practicioners of
the dynamic model (see note 2), especially Bailey and his followers,
These
prefer to use more appropriate terms such as lect and isolect
Thus, "a particular
concepts are definable by linguistic criteria alone.
speech form, with respect to a particular linguistic innovation, is an
(l) it lacks the innovation;
isolect in one of three possible ways:
(2) it has the innovation as a variable rule; or (3) it has the innovation
"
And lect is a speech form so defined for all the
as a categorical rule".
Cf.
innovations in the language or for some specific subset of them".
Parker 1971:^6-1+?. See also Bailey 1971a: I, 11-12 and Bailey 1973.
.

A detailed characterization of each of these paradigms, as well
evaluation of them, can be found in Bailey 1971b, also in Bailey

as an

1971a:

I,

37-39.

r>

For a Justification of this point, as well as for the theoretical
considerations concerning the naturalness of poly lect al grammars, see
Bailey 1972a.

71

k

The cases I present are also demolishing counterexamples to the
claim recently made by the adherents of the so-called "natural generative
phonology", since they maintain that soand change is statable in purely
phonetic terms. Cf. Hooper 197^.

^Cf

.

CerroD-Paloffiino 1973.

I should note incidentally that this rule does not apply to forms
such as mama 'niy mother', vasi 'ny house', awsa 'I play', miku 'I eat',
etc., since in these forms the vowel lengthening is marking the first
person singular actor and allocator. Therefore, this morpheme (which we
may represent as -v) has to be marked somehow as C- BxHe {1)1.
•7

Torero (l96U:it70A) reconstructs it as "-yka , that is without vowel
lengthening. His reason was perhaps the fact that vowel lengthening in
some of Quechua I lects is a by-product of more recent changes, especially
as a result of the drop of certain consonants in syllable final position
However, evidence from Wanka suggests that the length
(cf. 1.1.2. above).
of the vowel in the durative cannot be attributed to compensatory
phenomena. Thus, there is no doubt that the long vowel is native to ProtoQuechua I; cf. Parker 1971.
Q

In these communities as well as in others (especially on the left
side of the Mantaro River) Wanka is in the process of disappearing, even
old informants being difficult to find; some of these do recall rather
vaguely a few words. However, since those two communities are enclosed
by -yka preserving lects, we can safely infer that their /k/ was not
elided either.

However, in the communolect of Huacrapuquio (Waylla-Wanka) one
of ny informants (U3 years at the time I elicited ny data) told me that
he used to licten to his mother and elder persons saying, for example,
Cmiku+yu+yU from /miku+yku+y/ 'try to eat (it)', and also, more dramatically,
Cuyti+?lu+y3 from /uyku+71u+y/ 'give (it) to him/her/it', Cayu+?lu+yl!
from /ayku+?luty/ 'bring (it) in', etc. While this fact needs to be
checked more carefilly, we may perhaps e^lain the situation as follows. It
is possible that the change /k/ ->
after /y/ was a more general rule
in Huacrapuquio and only here, affecting not only the durative but also
the augmentative and even certain roots. Subsequently, under the pressure
of the surrounding dialects, Huacrapuqxiio undid the generalization at
least root internally and in the augmentative suffix also.
The change - ykU -> -yU 'augmentative' was also encoiantered in the
Thus, we have forms such as
communolect of Canchayllo (Shawsha-Wanka)
Csama+yu+suQD from /sama+yku+sun/ 'let us take a rest', Csama+yu+saD
from /sama+yku+sa/ 'tired', but Csayku-H 'to stand up', etc. The fact that
here /k/ drops in the augmentative can be attributed to an influence of
In the former lects the
the Tarma-Yauli lect on which Canchayllo borders.
augmentative regularly loses its /k/ (and, of course, here -yka is -ya,
as expected)
Another variety where the same drop takes place is Cuzco
Quechua (here - yka does not exist). Note, however, that in both cases the
change operates on these suffixes only.
,

.

.

72

The only iniportant variation I have found is what would appear
to he a further developinent in the change of -yka and this is the
Thus, in the communolect
drop of the intervocalically derived yod in - ya
of Muqui , which borders Huancani, the last kaista area) I noticed in the
speech of an 8^4 year old woman forns such as Cpuli+ku+wa+QkiD from
/pulitku+yka+nki/ 'you are weLLking', Casi+ku+wa+QkiD from /asi+ku+yka+nki/
'you are looking for', etc. (where the nonpermissible vowel sequence is
eliminated by the epenthetic w )
In view of the scant data I have for that
communolect, I cannot say whether this change is regvLLar or idiosyncratic;
however, aside from Muqui, I did not find any other community showing that
type of change. It is dubious then that it can be taken seriously as a
regular change in progress, except possibly for that community.
,

.

.

Torero (196U:U51B-A, itJOA) mentions similar changes in some
cocmunolects of the provinces of Ceijatambo, Chancay (Lima), and Daniel
Carri6n (Pasco), although in some of them palatalization seems to apply
Creider
across the board also (in the same environment as in Sicaya).
(1967) also observes the intermediate stage -yt a for Chancay, as a variable
rule.

12

The only grammar written we have for the Wanka lects , that of the
Franciscan friar Francisco Jose Raez (1917), has - yka 'durative' (and in
dicen
a foot note he says, "algunos suprimen la £ de este interf. por ej
caian en lugar de caican" , cf. p. 1^+9), and also - ka 'determiner' (where
he fails to "reconstruct" the final /?/, which he represents by h otherwise, but notes that in some places the c_ is "syncopated", as in hualass-a
From this it woiiLd seem that the
for hualass-ca 'the boy'; cf. p. 5^).
variety he describes is that of Siceya, where it is known he was a priest
for a time; but what casts doubt on this is the fact that he gives - cu
for the locative and - nu for the comparative (here as elsewhere he does
not pay much attention with respect to the transcription of long vowels);
however, as will be seen in 1.7. below, Sicaya has - naw for the comparative.
For this reason I think that Raez's grammar cannot be identified with any
of the varieties of Wanka; in this sense it is better to consider it as
an attempt at a panlectal grammar.
:

13

As I have said, historically it seems transparent that In both
especially in the latter, we are confronted with a by,
product of the suffixation of the verb -ka- 'to be'. This being so, to
what extent caii it be said that the elision of /k/ was due to the fact
that this segment was next to a word boundary? To this, one may answer
- yka and - ka?

by saying that even if we correlate word boundaries with actual physiological pauses, why is it that the same change did not affect other suffixes
which historically seem to contain the seme root "ka-? I am referring
to suffixes such as the durative-simultative -£Ka, the pluralizer -Ika,
In view of this, clearly the elision of /k/ hes affected idiosynetc.
cratically r..- the peLLatalization of -yka . since the quasi-homophonous
-yku - -yka 'augmentative' also seems to have been bireorphemic.
This change is also common in other varieties of Quechua, including
Thus in the neighbouring lect of Tarma-Yaxili this
the Ecuadorian variety.
morpheme drops regilarly its /k/, except in word final position; thus we

•

.

,

73

have situations such as Cmiku+na+nci+baql from /miku+na+ncik+paq/ for
us to eat', Cyaca+ya+nci+mD from /yaca+ya+nciktini/ 'we (incl.) are
knowing', but Cmama+ncikD 'our mother' , Cnuqa+ncik3 'we (incl.)', etc.
In view of this situation one may perhaps argue that the drop of /k/ in
Wanka is a by-product of dialectal borrowing. However, even if this were
true, it still remains to be explained why is it that the change observed
in Tarma-Yauli affects to that suffix only.
'

This is especially true for the Ecuadorian group of Quechua.
Parker 1969:15^ff.

Cf

16
The only exception to this rule is li-t-ku- 'to go (ethical dative)'
which synchronically should be analyzed as bimorphemic (cf. li+n 'he/she
goes').
The fact that here -ku does not voice can be perhaps attributed
to univerbation whereby the whole sequence liku- is interpreted as a single
Incidentally, that
root (cf. miku- 'to eat', taku- 'to blend', etc.).
this is not an ad hoc explanation can be proved by the fact that Quechua
shows an astonishing process of univerbation of suffixes; thus cf. verbs
such as Vatyku- 'to enter', %a+l?u- 'to get off, *ya+lku- 'to climb
v[p "ya+lpu- 'to get down', etc. where the second element in each case
was clearly a suffix which historically became completely amalgamated with
the root, which per se is now devoid of sense.
'

17
One may ask whether the monophthongization process observed in
Wanka is merely a propagation of the more generalized process found in
(a) the Wanka lects do
Ancash. To this I must answer by saying that:
not border on the Ancash variety; (b) the Tarma-Yauli variety, on which
Wanka borders, has uniformly - caw (although it presents monophthongization
of the comparative, realized as - ny ); and (c) the immediately neighbouring
variety of Wanka (i.e. Nuha-Wanka) lacks the Coii5)arative, and therefore
the potential linkage between the Tarma form - n-gy and the - nu ~ - nu
alternation found in Ya?a-Wanka is broken. In view of this, it seems safe
to conclude that the monophthongization of /aw/ in Wanka is an independent
development (= convergence), and, as such, another good example of a
morphologically conditioned sound change.

For a length study of the treatment of Spanish mid vowels /e, 0/
in borrowings, see Cerron-Palomino 197^'
19

A change from '"'-ma -» -wa 'directional' can however be postulated for
Proto-Quechua II; again, here we have a case where only the speaker object
suffix is affected by this "non-regular" change. Cf. Parker 1969:151
20

That /n/ is easily elided can be seen again ±n the sequence -n?which disappears altogether in lect G. Thus, we have /sin?a/ -* Csiyal
'nose', /xm?a-/ -> Cuwa-D 'to forget', /tan?a/
Cta-D 'to push', etc.
It seems that in certain lects the nasal disappeared first (thus in
Carhuacallanga we have forms such as Csi?a3, Cu?a-D, Cta?a-Il," res-'.iectively)
but in others it is the glottal that went first (thus Chongos Bajo CsiQall,
CuQa-D, and CtaQa-D). But again in the former case I was unable to find
intermediate steps such as Csi?aD, Cu?a3 or Cta?a-D. This, in away, would
seem to contradict the claim made in the sense that nasal dropping is always

—

74

preceded by a prior nasalization of the preceding vowel; however, it may
well be the case that a kind of rule "telescoping" is involved here.
21
Parker (1969:150), following Torero (196U), sets up the change
V > ^^ in the evolution of Quechua II (= A in Parker's notation); this
change applies only to the first person affix to produce Proto-Quechua II
"-jr_ 'first person singular (actor and allocator)'.
The other alternative,
which would be to reconstruct *-^ for Proto-Quechua and then assume a
change V;/ > V in Proto-Quechua I, runs, according to Parker, into serious
difficulties since "we must explain why the first person indergoes such
a change while infinitive '''-^, imperative "-i[_, second person *-yki and
many roots are unaffected". Having seen cases of sound change that affects
only specific morphemes, I think that Parker's reticence can no longer be
taken seriously. The examples provided in this paper make viable either
of the alternatives; thus the problem still remains as to which one vis the
correct.
,

"^For

this and the following discussion I follow Bailey 1972b.

Tlowever, interestingly enough. Dressier (1973:135) points, although
tentatively, the following for Latin: "On peu etablir que le developpement
phonique de desinences et suffixes anticipe, parfois , des changements
generaux posterieurs. Ainsi les monophthongaisons de ae^ et ei_ flexionnels
semblent preceder les changements correspondants dans les lexemes quoique
les temoignages ne soient pas abondants ou faciles a interpreter (a cause
d'influences dialectales)".
,

REFERENCES

Anttila, Raimo.
1972. An Introduction to Historiceil and Comparative
Linguistics
New York:
The Macmillan Coinpany.
A preliminary
Bailey, Charles-James. 1971a. Variation and Language Theory
mimeographed version. To appear as one of the volumes published by
the Center for Applied Linguistics.
1971b.
"Trying to talk in the New Paradigm". Working Papers in
Dept. of Linguistics
Linguistics k:2, pp. 312-38. Honolulu, Hawaii:
University of Hawaii.
Internal
"The Integration of Lingmstic Theory:
1972a.
Reconstruction and the Comparative Method in Descriptive Linguistic
Analysis", in Linguistic Change and Generative Theory , ed. by Robert
Stockwell and Ronald K.S. Macaulay, pp. 22-31. Bloomington, Indiana:
Indiana University Press.
Mimeographed
"How the Wave Model explains what it explains".
1972b.
version.
"The Patterning of Language Variation", in Varieties in Present
1973.
English , ed. by Richard W. Bailey and Jay L. Robinson, pp. 156-86.
New York:
The Macmillan Con^jany.
Cerron-Palomino, Rudolfo.
La evolucion del fonema */q/ en Ya?a1973.
Wanka.
Lima:
Universidad Nacional Mayor de San Marcos, C.I.L.A.
Doc. de Trabajo IIo. 15197'*
"Hispanismos en el Wanka". Mimeographed.
.

.

.

,

.

.

•

75

Creider, Chester. 196?.
"Fonologia del Quechua de Picoy", in Cuatro
fonologias quechuas , ed, by Alberto Escobar, Gary Parker, Chester
Creider, and Rodolfo Cerron Palomino, pp. i;l-52. Lima: UNMSM, PFL.
Dressier, Wolfgang. 1973.
"Poir une stylistique phonologique Su Latin; a propos
des. styles nepligents d'lone langue morte"'.
Bulletin de la
Societe de Linguistiqvte de. -Paris , Topie LXVIII:!, pp. 129-ii5.
Hooper, Joan Bybee. 197^. Aspects "of Natural Generative Phonology
Unpublished Doctoral dissertation, reproduced by the Indiana. University
Linguistics Club.
King, Robert D.
Historical Linguistics and Generative GraTMiar
19^9
Englewood Cliffs, NeV Jersey: Prentice-Hall, Inc.
Kiparsky, Paul. 1973.
"Abstractness , Opacity and Global Rules".
Mimeographed.
Parker, Gary J. 1969.
"Comparative Quechua Phonology and Grammar IV:
The Evolution of Quechua A". WPinL , 1:9, pp. li+9-20U.
"Comparative Quechua Phonology and Grammar V: The Evolution
1971.
of Quechua B". WPinL , 3:3, pp. 1+5-109.
New York:
Postal, Paul M.
1968.
Aspects of Phonological Theoiy
Harper & Row, Publishers.
Raez, Fray Jose Francisco Maria.
Gramaticas en el Quichua1917.
Huanca y en el de Ayachucho
Lima: Sanmarti y Cia.
Anales Cientificos
Torero, Alfredo. I96U.
"Los dialectos quechuas".
de la Universidad Agraria , 2:k, pp. Ul+6-78.
Lima, Peru.
.

'

.

.

.

•

.

.

.

Changes in Wanka-Quechua

More Related Content

Viewers also liked (13)

Similar to Changes in Wanka-Quechua (20)

More from Marcos Luk'aña (20)

Recently uploaded (20)

Changes in Wanka-Quechua