Sie sind auf Seite 1von 16

chapter 2

Polysynthesis as a typological feature


An attempt at a characterization from Eskimo
and Athabaskan perspectives
WillemJ.de Reuse
University of North Texas, Denton
Polysynthesis is characterized as a type of morphology, qualitatively different from
inflectional morphology and from derivational morphology and redefined as
productive noninflectional concatenation (PNC). Like syntax and unlike derivational
morphology, PNC is fully productive, potentially recursive, necessarily concatenative,
allows for ordering variability of some elements, and interacts with syntax. Unlike
inflectional morphology and like syntax and derivational morphology, PNC can
be category-changing. This postulated morphological feature is very prevalent in
polysynthetic language families such as Eskimo (illustrated by Siberian Yupik), but
not very prevalent in other language families often designated as polysynthetic,
such as Athabascan (illustrated by Western Apache). This new characterization
of polysynthesis has as an interesting consequence its existence, to a small degree,
inIndo-European languages.
Keywords: polysynthesis, morphological typology, Athabascan, Eskimo, Siberian
Yupik, Western Apache

1. Definitional problems
Characterising and defining polysynthesis has been a problem for some time.1
The concept of polysynthesis was first formulated by Duponceau (1819) for languages that can express a great variety of ideas in one single word. A modern attempt at a characterization is Fortescues statement that polysynthetic languages
are languages with complex morphologies capable of packing into a single word
many morphemes that in more analytic languages would be independent words
(1994:2600). He goes on to say that that is a very loose statement and wonders
whether it is possible to define polysynthesis more tightly. He comes to the con1. Earlier versions of this paper were presented at a seminar at the Research Center for Linguistic Typology, La Trobe University, Bundoora, Australia, 2003, and at the 15th Inuit Studies Conference, Muse du Quai Branly, Paris, 2006. Some of the ideas in this paper are also in de Reuse
(2006a). I thank the participants in this seminar and conference for their input and comments.

20

WillemJ.de Reuse

clusion that there is no unified phenomenon polysynthetic morphology that can


be defined exclusively (1994: 2601).
Evans and Sasse (2002: 3) in the introduction to their collection of edited articles Problems of Polysynthesis come to rather similar conclusions:
() a prototypical polysynthetic language is one in which it is possible, in a single
word, to use processes of morphological composition to encode information about
both the predicate and its arguments, for all major clause types (i.e. one-, two- and
three-place predicates, basic and derived), to a level of specificity allowing this word
to serve alone as a free-standing utterance without reliance on context.

Like Fortescue, they want to leave the definition intentionally vague.2


On the other hand, from a more formalist angle, an ambitious attempt has
been made by Baker (1996) in The Polysynthesis Parameter to come up with a very
restrictive definition of polysynthesis. According to Baker, polysynthesis is a macroparameter called the Morphological Visibility Condition, informally, the requirement that every argument of a head element must be related to a morpheme
in the word containing that head. That requirement can be satisfied through syntactic agreement or through noun incorporation (1996: 1415). It is interesting
to see how Baker has reached the term polysynthesis here. He first postulated a
formal characteristic of certain polysynthetic languages, called it a parameter, and
then applied the term polysynthesis to that parameter. I will not take issue with
Baker about whether such a parameter exists, but I object to the term polysynthesis being applied to the parameter. The term polysynthesis is already in use as an
(admittedly vague) typological term. A newly discovered parameter does need to
be given a name, preferably a catchy one, but it is confusing to appropriate a term
already in use and decide that that is the real meaning of the term.
The resulting situation is that Bakers too restrictive definition leaves most traditionally polysynthetic languages unaccounted for, as Evans and Sasse (2002: 4)
also point out.
Nevertheless, one can sympathize with Bakers program to define polysynthesis more tightly. But I must agree with Fortescue (1994: 26002601) that it is not
possible to give a very precise definition without doing violence to the way the
term has been used traditionally.
In this paper, I will try to circumscribe the concept of productive noninflectional concatenation, or PNC, in the belief that productive noninflectional
concatenation is a more precisely definable concept than polysynthesis. I will
not claim that the two terms are equivalent and I will not be advocating that we
2. Another interesting approach to the problem, adopted by Mattissen (2004), is to accept the
heterogeneity of morphological structure in polysynthetic languages and to engage in a detailed
classification, with several parameters, of all languages generally labelled polysynthetic.

Chapter 2. Polysynthesis as a typological feature

r eplace the term polysynthesis by PNC.3 Rather, I will argue that the existence of
noticeable amounts of PNC is what characterizes most prototypical polysynthetic languages.
I will then illustrate the concept of PNC with two languages traditionally considered polysynthetic, Central Siberian Yupik Eskimo (CSY), an EskimoAleut
language (de Reuse 1994, Nagai 2004), and Western Apache (WA), a Southern
Athabascan language (de Reuse 2006b). These two languages were chosen because
they are both representative of language families considered to be extremely polysynthetic, although, as we will see, in different ways. They were also chosen because I have extensive field experience with both.

2.Productive noninflectional concatenation (PNC): neither derivation


nor inflection?
In the search for prototypical polysynthesis, many potential typological characteristics of polysynthetic languages are listed by Fortescue (1994: 2601) and Evans
and Sasse (2002: 34), but not, apparently, the set of characteristics that I will define as PNC. In this paper, I will propose that this set is striking in the most heavily
polysynthetic languages, and therefore the set is a good candidate for a prototypical property of polysynthesis. We need to keep in mind, though, that all polysynthetic languages have characteristics that have nothing to do with PNC, and as we
will see, some polysynthetic languages have rather little PNC.
Table 1 is a chart of features distinguishing inflection, derivation, PNC, and
syntax: As seen in Table 1, PNC, even though morphologically bound, has all
six feature values in common with the syntax; it also has [6] in common with
derivation,4 and [1] and [5] in common with inflection. It is not claimed that the
features in Table 1 are independent of each other. The main definitional feature is
3. In de Reuse (1992), I used the term internal syntax for what I now prefer to call PNC. Ihad
chosen the term internal syntax (first used by Swadesh 1939 to describe the word formation
processes of Nootka, a polysynthetic Wakashan language), because PNC has many characteristics of external (or real or phrasal or sentential) syntax, but it is word-internal. It then occurred
to me that what Swadesh meant with the term internal syntax is not exactly what Imean by
PNC, and therefore I replaced internal syntax by the more descriptive but less catchy PNC,
first used in de Reuse (2006a). After completing this paper, I became aware of the concept of
quasi-inflectional morphology, as defined by Melchuk (2006), and further discussed by Beck
(2008). What Melchuk and Beck describe as quasi-inflectional morphemes appears to be quite
similar to what I call elements of PNC, but further study is needed to ascertain whether the concepts are the same.
4. At this point in the paper, derivation is assumed to mean nonproductive derivation, i.e., the
traditional concept of derivation minus what I call PNC.

21

22

WillemJ.de Reuse

Table 1.

[1] Productive?
[2] Recursivity possible?
[3] Necessarily concatenative?
[4] Variable order of elements possible in
some instances?
[5] Interaction with syntax possible?
[6] Lexical/phrasal category changing
possible?

(Nonproductive)
Inflection derivation
Pnc

Syntax

yes
no
no
no

no
no
no
no

yes
yes
yes
yes

yes
yes
yes
yes

yes
no

no
yes

yes
yes

yes
yes

probably [1], i.e., productivity. More research needs to be carried out on how to
derive features [26] from feature [1].
I will first explain features [1] through [6]. In my explanation, I will not talk
about morphemes or affixes or words, but of elements, in order to have a term covering both morphological and syntactic units.
Feature [1] means absolute productivity, i.e., there are no idiosyncratic restrictions on the use of the element. Thus, its presence is conditioned by semantic plausibility only, and not by selectional restrictions. For many polysynthetic languages,
the elements of PNC are so numerous that it is very unlikely that native speakers
would have the ability to memorize the existing sequences and store them in the
lexicon (Smith 1978, Fortescue 1980, de Reuse 1994). Inflection, of course, is also
absolutely productive, but only within its paradigm. Note that I am claiming that
derivational morphology is never absolutely productive. Since some of what is traditionally called derivational is productive, I am in effect changing the definition
of derivational morphology, so that the productive parts of derivational morphology will actually be reassigned to PNC. Note also that in syntax, productivity
does not appear to be absolute, since we have lexicalized idioms such as kick the
bucket. However, in such cases, the Amelia Bedelia style or nonidiomatic reading
of idioms is always possible.
Feature [2] means that the same element can potentially occur twice within
the same word (which is the case in PNC) or within the same sentence (which
is the case in syntax), one having scope over the other. Its presence is again conditioned by semantic plausibility. Recursion is certainly connected to productivity, since it is hard to imagine recursion without the productivity of the elements
involved.
Feature [3] means that the elements are going to be in some linear order. Ido
not expect nonconcatenative morphology such as suppletion or Semitic-style
morpheme-internal change to exist in PNC, just as it does not exist in syntax.
Feature [4] means that in some cases, the order of elements can be free. Just

Chapter 2. Polysynthesis as a typological feature

as in free word order languages in the syntax, some elements of PNC can be freely
ordered, most likely constrained by pragmatic factors only. This is impossible in
derivation.
Feature [5] has to do with relationships between the PNC element and elements of syntax. As is well known (Anderson 1982, 1988), inflection interacts with
syntax as in agreement or case marking. Derivation does not interact with syntax.
PNC can interact with syntax, especially in the case of noun incorporation.
Feature [6] means that the element can change the lexical category in the
morphology. Derivational morphology can do this as well, but inflectional morphology does not. Here, PNC lines up with derivational morphology, but in some
sense, it lines up with syntax as well. Indeed, in the syntax, the addition of an element can change the category, now called the phrasal category. For example, very
good is an Adjective Phrase, but very good quality is a Noun Phrase.

3. PNC illustrated in Central Siberian Yupik Eskimo and Western Apache


I now illustrate PNC with two polysynthetic languages from two very different
families: Central Siberian Yupik Eskimo (CSY), spoken by over a thousand people
on St. Lawrence Island, Alaska, and Chukotka, Russian Far East, of the Eskimo
Aleut family; and Western Apache (WA), spoken by about 7,000 people on and
near five reservations in central and east-central Arizona, USA, of the Athabascan
family. (1) is an example of feature [1], productivity, in CSY:

(1) neghyaghtughyugumayaghpetaa5
negh-yaghtugh-yug
-uma-yagh -pete -aa
eat -go.to.V -want.to.V-past-frustr-infrn-ind.3s.3s
It turns out s/he wanted to go eat it, but

In (1), only the stem negh- and the inflectional -aa are obligatory. The stem is
always initial, and the inflection is always final. Any or all of the other elements,
which I consider to be part of the PNC, can be left out. So using at least one PNC
element, one can make thirty different words from this (de Reuse 1992: 1667).
The word can be made longer, and this is only a slightly longer than average word.
For semantic reasons, it happens to be the case that the order of elements has to be
-yaghtugh-yug-uma-yagh-pete-. Eskimo has derivational morphology as well (de
Reuse 1992, 1994), which is not illustrated in (1).
Sentence (2) is an example of productivity in WA.
5. In the example sentences and the morphological analysis, elements to be discussed are in
bold type. The accepted St. Lawrence Island orthography is used in CSY forms (de Reuse 1994)..

23

24

WillemJ.de Reuse

(2) nanndagochilzh6
na -n -n
-da-go -chi-l
-zh
them-again-usually-pl-them-4S -them-play
1a 1b
2
3 4
5 9
10
They usually play again.

The structure of the WA verb is more complex than in CSY. In WA, one can postulate a set of position classes in which the different THEMATIC, PNC, inflectional
(and derivational, not illustrated here) prefixes have to go. The position classes are
numbered with the numbers under the glosses in (2). The order of prefixes within WA verbs is fixed (but not entirely so, as we will see later). There are about ten
or twelve positions, and one cannot really add to the verb beyond that. The three
prefixes called them have no clear meaning, they just have to be there, and together with the stem -zh, they indicate the verb to play. So the sequence nagolzh
means to to play. The elements da- and chi- are inflectional and can be replaced
by zero or other inflectional elements. The elements that are productive, but not
inflectional, are the two PNC prefixes n- again, one more time and n- usually. Each of those can be added to the verb or deleted. However, if one removes the
n- usually, the verb stem gets changed to -z. So either we have to postulate that
elements of PNC can induce changes elsewhere in the words, or maybe it is evidence that n- usually is inflectional. In the literature on Athabascan languages,
n- usually is considered to be inflectional, but I am not convinced that it is.
All in all, there is PNC in WA, but there is less of it per word than there is
in CSY. What seems to fascinate people about Athabascan languages, and what
makes their analysis difficult are all the them elements that have to be present in
various positions and interrupt the morphemes that have independent semantic
content. Hence the term interrupted synthesis, coined by Sapir in a course on
Athabascan (Kari 1989: 428).
Let me now illustrate feature [2], recursion, in CSY with (3). This is a real word
from a story. I took it out of the sentence so it would be easier to follow.

(3) iitghesqesaghiisqaa
itegh -sqe
-yaghtugh-sqe
-aa
come.in-ask.to.V-go.to.V -ask.to.V-ind.3s.3s
Hei asked himj to go ask himk to come in.

The element sqe- ask to V is used recursively. The morphophonemics explain


the differences between the analysis and the actual form. We have seen -yaghtughalready in (1). The inflection is also the same as in (1).
6. The spelling of WA is described in de Reuse (2006b).

Chapter 2. Polysynthesis as a typological feature

In WA, examples of recursion cannot be provided because recursion does not


seem to occur. Again, this is evidence that WA does not have as much PNC as CSY
has.
I do not need to illustrate feature [3], concatenation, because one can clearly
see in the CSY and WA examples that the morphology is concatenative.
Feature [4], variable order, is illustrated with (4) and (5) in CSY.

(4) aananiitkaa
aane -nanigh -utke
-aa
go.out-cease.to.V-V.on.account.of-ind.3s.3s
He ceased going out on account of it.

(5) aanutkenanighaa
aane -utke
-nanigh -aa
go.out-V.on.account.of-cease.to.V-ind.3s.3s
He ceased going out on account of it.

Even though generally in CSY the rightmost element of PNC has scope over
what is on the left, this principle does not seem to be working here. These two sentences mean exactly the same thing and were uttered within three lines from each
other in the same story (de Reuse 1994: 93).
Sentences (6) and (7) are potential WA examples. Both (6) and (7) were recorded from the same speaker, but on different occasions; (7) was recorded from
another speaker as well.

(6) yohazh
y -o -n
-H -
-ghazh
3O-dir-them-pfv-trans-bite
4 6a 6b 7 9
10
S/he took a bite out of it.

(7) hoyihazh
o -y -n
-H -
-ghazh
dir-3O-them-pfv-trans-bite
6a 4 6b 7 9
10
S/he took a bite out of it.

Some comments on the analysis of these WA forms are necessary. The expected prefix order is (6), as shown by the position class numbers underneath it, but
Ihave also heard (7) from two speakers. 3O is the third person object marker, and
the third person subject is zero here. dir is directive, marking the fact that the object is only partially affected, so this is the difference between taking a bite out of
it and completely biting it off; them is a thematic prefix, apparently conditioned

25

26

WillemJ.de Reuse

by the presence of the dir. The word-initial h- in (7) is prosthetic and phonologically predictable. So this is an example of prefix order variation in Apache. However, one might ask whether the prefixes involved are elements of PNC. The prefix
y- is inflectional, and the prefix o- is not productive and therefore has to be considered derivational. We therefore need to conclude that (6) and (7) are actually not
examples of PNC, but rather show variable order between an inflectional element
and a derivational element. As a result, we need to qualify feature [5], and postulate that variable order within elements of PNC is possible, but that variable order within inflectional elements and variable order within derivational elements is
not possible. However, variable order between an inflectional element and a derivational element is also possible, as shown in the WA examples. This sort of variable order is not relevant to my claim about PNC.
Feature [5], interaction with external syntax, is best illustrated with noun-incorporating suffixes in CSY, in (8).

(8) atan aangelghiimeng


ata -n
aange -lghii
-meng
father-poss.2s.abs.sg be.big-intr.ptcp-mod.sg
qikmilguuq
qikmigh-lgu
-uq
dog
-have.N-ind.3s
Your father has a big dog.

As Sadock (1980, 1991) has demonstrated on the basis of parallel constructions in Greenlandic Eskimo, the element -lgu have N (which I would call an
element of PNC) acts like a morphologically intransitive verb, and like other intransitive verbs, it can occur with a direct object in an oblique case (here the mod).
Since elements of PNC cannot attach to inflected words, the mod case marking
cannot occur on qikmigh- dog, but it does show up in the morphologically free
and stranded modifier aangelghiimeng big. This is expected, since CSY modifiers
agree in case with their heads. At the syntactic level then, aangelghiimeng qikmighbig dog forms a constituent to which the -lgu- is attached.
When considering WA, the situation is different again. The elements of PNC
do not seem to interact with the syntax in WA. There is no noun incorporation
in WA. This is not particularly significant for Athabascan as a whole, since some
Northern Athabascan languages have noun incorporation (Axelrod 1990).
Finally, let us look at feature [6], lexical category changing, in CSY. (9) is an
example of a word changing from a verb yughagh- to pray to a noun yughaghvigchurch to another noun yughaghvigllag- big church, back to a verb yughaghvigllange- to acquire a big church. In this word, there are actually two lexical category-changing suffixes. The first is the derivational suffix -vig- place to V, which is

Chapter 2. Polysynthesis as a typological feature

lexicalized with pray to mean church, and not the completely predictable place
to pray, any place to pray. The second lexical category-changing suffix is -nge- to
acquire N which can be productively added to any nominal form to change it to a
verbal form, and therefore is an element of PNC.

(9) yughaghvigllangyugtut
yughagh-vig
-ghllag-nge
-yug
-tut
pray
-place.to.V-big.N -acquire.N-want.to.V-ind.3p
They want to acquire a big church.

In WA, lexical category changing is not a productive process. Within the verbal word, there are no productive category-changing mechanisms for this, and
certainly no changing back and forth from one category to another. One thing
that can happen for some verbs is that the thematic prefix -- in position 9 is deleted. Alternatively, one might just as well say that for some nouns, derived verbs
are formed by adding the thematic prefix -- to them. Let us look at two examples
in (10). Note the presence of the prefix -- in the forms to the left, and its absence
in the forms to the right.
(10) yti S/he is talking. yti word, speech
tag S/he is counting. tag school
In WA, there are also some nominalizing enclitics at the right edge of the verb,
which also indicate things such as topic, specificity, and definiteness. They are not
clearly part of the morphology. There is one category-changing mechanism changing nouns to verbs, involving a lexicalized proclitic and enclitic combination, but
which is not productive, and therefore it is not PNC (de Reuse, 2008). Here again,
there are real differences between the polysynthetic morphology of CSY and that
of WA.

4.Consequences of a PNC view of polysynthesis for typology and


morphological theory
One consequence of accepting PNC as the definitional feature of polysynthesis is a
trinitarian view of morphology. Rather than distinguishing between inflectional and derivational, we now have to distinguish between inflectional, PNC, and
derivational morphology. So PNC is a type of morphology, on a par with inflectional and derivational. This is of course a controversial idea, since many linguists
are unsure about the traditional distinction between inflectional and derivational morphology, and therefore are particularly unwilling to accept a three-way distinction of morphology into inflection, derivation, and PNC.

27

28

WillemJ.de Reuse

Another consequence of this view is that there is no such thing as a polysynthetic type of languages. The languages called polysynthetic are languages with
large or notable amounts of PNC. The difference between polysynthetic languages
and non-polysynthetic languages is thus a quantitative difference, not a qualitative
one. The stereotypical polysynthetic type is represented by the languages that
happen to have many (a hundred or more) PNC elements, such as CSY. I will return to the case of WA in section 6 below.
As mentioned already, another consequence of this view is that what is traditionally called derivational is now split into derivation and productive derivation, i.e. PNC. But one should not conclude from this that only polysynthetic languages have PNC elements, and that non-polysynthetic languages have derivation
only. Indeed, PNC elements are not completely absent from non-polysynthetic
languages. Some of the affixes traditionally called derivational in Indo-European
languages are productive, and among these productive ones, some have recursion
as well. Examples of PNC affixes in non-polysynthetic languages are shown for
English in (11) and (12). They are productive and recursive as well.
(11) anti- as in antiabortion, antiantiabortion, etc
(12) re- as in rewrite, rerewrite, etc
The diminutive suffix of Dutch, -(t)je, is also productive. It can be added to any
noun with which it is semantically compatible. It cannot be used recursively, but I
think that is more a semantic and pragmatic restriction than a morphological one.
This Dutch diminutive suffix contrasts starkly with the diminutive suffixes of French
-et, (feminine) -ette,7 and the diminutive suffixes of English, such as -ette, -let, -kin,
-ling, which are unproductive. I propose to reclassify such productive affixes as anti, re-, and the Dutch diminutive as PNC suffixes, rather than as derivational ones.
Thus, elements of PNC do occur in non-polysynthetic languages, but they are not
numerous. CSY has hundreds of PNC affixes, European languages just a few.

5.A further property of PNC elements, possibly not shared by derivational


elements of the same meaning
CSY often has elements of PNC of adverbial meaning, and independent adverbial
words of exactly the same meaning. Consider the CSY word in (13). It is possible
to add a PNC suffix to it, meaning to V in vain, as in (14). It is also possible to use
7. Some varieties of Belgian French use -et, -ette productively, presumably under Germanic influence.

Chapter 2. Polysynthesis as a typological feature

the separate adverb leganiitek in vain as in (15), and it is possible to use both the
PNC suffix as well as the separate adverb, as in (16). Forms (14) through (16) have
the same meaning, although there must be pragmatic differences among the three.
(13) igaamalghiinga
igagh-uma-lghiinga
write -past-intr.ptcp.1s
I wrote.
(14) igaapigesnaamalghiinga
igagh-pigesnagh-uma-lghiinga
write -V.in.vain -past-intr.ptcp.1s
I wrote in vain.
(15) leganiitek igaamalghiinga I wrote in vain.
(16) leganiitek igaapigesnaamalghiinga I wrote in vain.
What I would like tentatively to suggest is that this co-occurrence of an item
of PNC with an independent word of the same meaning is a characteristic of PNC
and not of derivation. In other words, I am suggesting that a derivational suffix of
the same meaning cannot co-occur with an independent word of the same meaning. This co-occurrence is not a possibility with all items of PNC, however. It does
not seem possible with elements of PNC that are easily recursive; so building from
example (11), the phrase against antiabortion obviously does not mean the same
as against abortion or antiabortion.
Obviously, this suggestion needs to be checked with more data in more languages with PNC. Let us now contrast examples with a PNC suffix with examples
with a derivational suffix of the same meaning. Consider the Dutch diminutive
suffix which, as stated at the end of section 4, I consider to be a PNC suffix. From
examples (1820), it is clear that the suffix can co-occur with the independent adjective klein small. Now there are subtle pragmatic differences between (18), (19),
and (20), but it is not the case that (20) refers to a chair in some way smaller than
the chairs referred to in (18) and (19).
(17) stoel

chair

(18) stoeltje

chair-dim

(19) een kleine stoel

a little/small chair

(20) een klein stoeltje a little/small chair


Now let us compare this with the derivational diminutive suffix -ette in French.
The form (22) with the diminutive suffix does not really mean little girl (as the

29

30

WillemJ.de Reuse

form in (23) does). It is lexicalized with the meanings schoolgirl of a certain age
range or girlish boy. But regardless of what it actually means, one cannot add the
independent adjective petite small to it. My own intuitions on French, as well as
the half-dozen French speakers I consulted, reject (24) as very strange.8
(21) fille

girl

(22) fillette

girl-dim

(23) une petite fille

a little girl

(24)
*une petite fillette
The intuition of native speakers of French is that you cannot use (24) because
you are expressing smallness twice. But that is not what is happening, because in
Dutch, there are no restrictions on using the adjective and the diminutive together.
My claim is that the derivational status of the French suffix and the PNC status of
the Dutch suffix are responsible for the difference in grammaticality.

6. Conclusions
I have argued in this paper that there exists a type of productive morphology that
is neither inflection nor derivation (as traditionally defined), but is a third kind of
morphology which I call PNC. It is large numbers of PNC that best characterize
polysynthetic languages. As a result, some languages, like those belonging to the
Athabascan family, are not particularly polysynthetic, even though they have great
morphological richness and complexity.
The reader might object to my characterization of polysynthetic that by redefining the term as meaning having lots of PNC elements, I am guilty of the
same sin that I accused Baker of in section 1, since some languages traditionally
called polysynthetic, such as Athabascan, must be considered less polysynthetic
in my characterization. At the risk of engaging in some casuistry, I will argue that
I am a bit less guilty of this sin.
Eskimo, which has always been a good exemplar of polysynthesis, even in the
vaguer conceptions of polysynthesis, now becomes the best exemplar of polysynthesis. Words in other polysynthetic languages might be long, as shown by WA,
but they are never quite as long as in Eskimo languages. As Woodbury (2002)
8. Speakers of Belgian French who use -et, -ette productively can use it with the adjective petite.
An example is farde folder, une fardette a little folder, and une petite fardette a little folder (de
Reuse, 1977 Tervuren, Belgium, fieldnotes, overheard).

Chapter 2. Polysynthesis as a typological feature

points out, Eskimo languages are extreme from that point of view. In Bakers theory, it is Iroquoian that is the best exemplar of polysynthesis, and Eskimo is not
even considered polysynthetic. So I would argue that Baker does more violence to
the term polysynthesis than I do, and that my interpretation of polysynthesis as
lots of PNC is closer to the popular conception of polysynthesis.
As far as Athabascan languages are concerned, I am indeed forced to draw the
conclusion that Athabascan languages are not particularly polysynthetic. This is a
conclusion that is surprising to Athabascanists, who tend to think that Athabascan languages are rather good exemplars of polysynthetic languages. It is interesting, however, that Edward Sapir, the first linguist to study Athabascan morphology
from a typological point of view, apparently did not think of them as polysynthetic.
According to Kari (1992: 109, 203), Sapir used the term interrupted synthesis to
characterize Athabascan languages, and was using it in contradistinction to polysynthesis (emphasis Karis).9
Returning now to the issue of productivity, I agree with Fortescue, Mithun,
Sadock, and Tersis (p.c. 2006) that there are clines of productivity in affixation.
However, I would prefer to call them clines of nonproductivity. I think of productivity as I think of pregnancy, i.e., as an all-or-nothing phenomenon. And furthermore, I would like to keep the clines of nonproductivity within derivation.
Whatever is not productive can be very rare or very common. We all agree that
there is a lot of difference between an affix that only occurs with a few stems, and
an affix which, without being productive (and therefore qualifying as PNC), occurs with hundreds of stems. I claim that there is a serious qualitative difference
between productive affixation and affixation that is not productive, even though it
occurs on hundreds of stems and therefore looks productive.
This is a stronger claim than claiming that whatever is not inflection is all just
derivation (in the traditional sense), and that everything inside that derivation is
a cline. It might well be the wrong claim, but it is worth checking into more than
has been done.
One possible objection against the idea of absolute productivity in morphology
is the following. According to Mithuns research on noun incorporation in Iroquoian and her research on affixation in Central Alaskan Yupik Eskimo (Mithun
and Gorbett 1999, Mithun p.c. 2006), speakers can often tell which morphological
9. Other intriguing characterizations of polysynthesis in Athabascan languages with comparisons to Eskimo are provided in Fortescue (1992:100) where he notes that there is in Athabascan
something he calls extended derivation, () the extensive middle-ground between true derivation and inflection so characteristic of polysynthetic languages. In Fortescue (2006), he compares old polysynthesis with new polysynthesis, and concludes that both West Greenlandic
(an Eskimo language) and Koyukon (an Athabascan language) display old polysynthesis. Unfortunately, space is lacking to discuss these ideas here.

31

32

WillemJ.de Reuse

constructions are used and which ones are not in use. This might be an indication
that there is no difference between the absolute productivity I claim for PNC and
the cline of productivity in derivation. I have no fieldwork experience with Iroquoian, so I will take Mithun at her word regarding Iroquoian. Regarding Eskimo,
my experience with native speakers is rather different from that of Mithuns. Some
of my CSY consultants could indeed tell whether they had heard a particular morphological construction, but others could not tell, and others even thought that
the question was absurd. Therefore, I am not sure that we can trust Eskimo speakers with judgments about whether they have heard a word before or not. There are,
of course, elements of Eskimo morphology that have undergone lexicalization and
are derivational suffixes. An example is -vig- place to V, in (10) above. CSY speakers can generally tell which stem with a derivational suffix is in use and which one
is not. So I think that the distinction between derivational suffixes and suffixes of
PNC is genuine from the point of view of Eskimo speakers, even though it is hard
to discern from a formal point of view.
As far as WA is concerned, the reactions I obtain from native speakers are very
similar to the reactions Mithun obtained from Iroquoian and Central Alaskan
Yupik speakers. WA speakers can generally tell which morphological constructions are used and which ones are not in use, even though WA morphology is very
complex. I take this to mean that most of the affixes of WA are derivational and
not elements of PNC.

Abbreviations
3O
4S
dim
dir
frustr
H
ind.3p
ind.3s
ind.3s.3s
infrn
intr.ptcp
intr.ptcp.1s
mod.sg
N
past
pfv
pl

third person object


fourth person subject (a type of polite third person)
diminutive
directive, marking the fact that the object is only partially affected
frustrative: but.., in vain
means: put a high tone on the previous syllable
indicative mood, third-person plural subject
indicative mood, third-person singular subject
indicative mood, third person subject acting on third person object
inferential evidential, often translatable as it turns out
intransitive participial, can be a nominalizer
intransitive participial mood, first-person singular subject
modalis singular, a type of oblique case
nominal expression (always preceding)
past tense
perfective aspect
plural subject or object


poss.2s.abs.sg
them
trans
V

Chapter 2. Polysynthesis as a typological feature


absolutive singular case, second-person singular possessor
themative prefix (has no clear meaning, it just has to be there) qualifying the stem
transitivizing derivational prefix, also called classifier
verbal expression (always preceding)

References
Anderson, S. r. 1982. Wheres morphology? Linguistic Inquiry 13: 571612.
Anderson, S. r. 1988. Inflection. In Theoretical Morphology. Approaches in Modern Linguistics,
M. Hammond & M. Noonan (eds.), 2343. San Diego, CA: Academic Press.
Axelrod, M. 1990. Incorporation in Koyukon Athapaskan. International Journal of American
Linguistics 56: 17995.
Baker, M. C. 1996. The polysynthesis parameter. Oxford: OUP.
Beck, D. 2008. Variable ordering of affixes in Upper Necaxa Totonac. Paper presented at the Society for the Study of the Indigenous Languages of the Americas Meeting, Chicago, January 2008.
Duponceau, P. S. 1819. Report of the corresponding secretary to the committee, of his progress
in the investigation committed to him of the general character and forms of the languages
of the American Indians. Read 12th Jan. 1819. Transactions of the historical and literary
committee of the American Philosophical Society, held at Philadelphia, for promoting useful
knowledge I: xvii-xlvi.
Evans, N. & Sasse, H.-J. 2002. Introduction: problems of polysynthesis. In Problems of Polysynthesis, N. Evans & H.-J. Sasse (eds.), 113. Berlin: Akademie Verlag.
Fortescue, M. 1980. Affix-ordering in West Greenlandic derivational processes. International
Journal of American Linguistics 46: 25978.
Fortescue, M. 1992. Aspect and superaspect in Koyukon: An application of the Functional
Grammar model to a polysynthetic language. In Layered Structure and Reference in a Functional Perspective. Papers from the Functional Grammar Conference in Copenhagen 1990,
M. Fortescue, P. Harder & L. Kristofferson (eds.), 99141. Amsterdam: John Benjamins.
Fortescue, M. 1994. Morphology, polysynthetic. In The Encyclopedia of Language and Linguistics,
r. E. Asher (ed.), 26002602. Oxford: Pergamon.
Fortescue, M. 2006. Variations on polysynthesis: West Greenlandic, Chukchi, Koyukon and Nuuchahnulth. Paper presented at the 15th Inuit Studies Conference, Muse du Quai Branly,
Paris, 2006.
Kari, J. M. 1989. Affix positions and zones in the Athabaskan verb complex: Ahtna and Navajo.
International Journal of American Linguistics 55: 42454.
Kari, J. M. 1992. Some concepts in Ahtna Athabaskan word formation. In Morphology now,
M. Aronoff (ed.), 10731, 2036. Albany, NY: State University of New York Press.
Mattissen, J. 2004. A structural typology of polysynthesis. Word 55: 189216.
Melchuk, I. A. 2006. Aspects of the theory of morphology. Berlin: Mouton de Gruyter.
Mithun, M. & Gorbett, G. 1999. The effect of noun incorporation on argument structure. In
Boundaries of Morphology and Syntax, L. Mereu (ed.), 4971. Amsterdam: John Benjamins.
Nagai, K. 2004. A morphological study of St. Lawrence Island Yupik: Three topics on referentiality. PhD dissertation in Linguistics, Kyoto University.

33

34

WillemJ.de Reuse

de Reuse, W. J. 1992. The role of internal syntax in the historical morphology of Eskimo. In
Morphology Now, M. Aronoff (ed.), 16378, 21216. Albany, NY: State University of New
York Press.
de Reuse, W. J. 1994. Siberian Yupik Eskimo. The Language and its Contacts with Chukchi. Salt
Lake City, UT: University of Utah Press.
de Reuse, W. J. 2006a. Polysynthetic language: Central Siberian Yupik Eskimo. In Encyclopedia of
language and Linguistics, 2nd edn, Keith Brown (ed.), 7458. Oxford: Elsevier.
de Reuse, W. J., with the assistance of P. Goode. 2006b. A Practical Grammar of the San Carlos
Apache Language. [Lincom Studies in Native American Linguistics 51]. Munich: Lincom.
de Reuse, W. J. 2008. Denominal verbs in Apachean languages. International Journal of American Linguistics 74.
Sadock, J. M. 1980. Noun incorporation in Greenlandic: A case of syntactic word-formation.
Language 57: 30019.
Sadock, J. M. 1991. Autolexical Syntax A Theory of Parallel Grammatical Representations. Chicago, IL: The University of Chicago Press.
Smith, L. r. 1978. Some properties of Labrador Inuttut verbal derivation. Etudes Inuit/Inuit Studies 2: 3748.
Swadesh, M. 1939. Nootka internal syntax. International Journal of American Linguistics 9: 77
108.
Woodbury, A. C. 2002. The word in Cupik. In Word. A Cross-linguistic Typology, R. M. W. Dixon
& A. Y. Aikhenvald (eds.), 7999. Cambridge: CUP.

Das könnte Ihnen auch gefallen