The Akulmiut: Territorial Dimensions of a Yup'Ik Eskimo Society
Nominalist Theory of Eskimo: A Study in Self Deception IJAL 1999
Transcript of Nominalist Theory of Eskimo: A Study in Self Deception IJAL 1999
THE NOMINALIST THEORY OF ESKIMO: A CASE STUDY
IN SCIENTIFIC SELF DECEPTION
Jerrold M. Sadock
University of Chicago
1. Introduction. Throughout this century, Eskimo languages have been
persistently claimed not to display a distinction between nouns and verbs. The
idea is seductive, but since it is false, it has led otherwise reputable scholars into
numerous errors of both fact and logic. I wish to document the idea from its
inception to the present day, not so much as a way of advancing grammatical
studies of Eskimo as of providing an object lesson in scientific self deception.
The notion that Eskimo languages fail to distinguish nouns and verbs
appears to have been the brain-child of William Thalbitzer (1873-1958), a noted
Eskimologist of the early part of this century and the first Professor of
Eskimology at the University of Copenhagen. Thalbitzer believed not only that
there was just one undifferentiated part of speech in these languages, but
moreover, that that part of speech was the noun. This idea of Eskimo as an
entirely nominal language appears in its most developed form in Thalbitzer’s
important contribution to the Handbook of American Indian Languages
(Thalbitzer 1911), where it is also invested with profound psychological
significance:
In the Eskimo mind the line of demarcation between the noun and
the verb seems to be extremely vague, as appears from the whole
structure of the language, and from the fact that the inflectional
endings are, partially at any rate, the same for both nouns and
verbs.… (1911:1057)
Judging from these considerations, we get the impression that to
the Eskimo mind the nominal concept of the phenomena of life is
predominant. The verbal idea has not emancipated itself from the
idea of things which may be owned, or which are substantial.
Anything that can be named and described in words, all real things,
actions, ideas, resting or moving, personal or impersonal, are
subject to one and the same kind of observation and expression.
We are accustomed to conceive activities or qualities as essentially
different from things in themselves, and we have a special class of
words (viz. the verbs) to express them. They seem to impress the
Eskimo mind, or to be reflected by it, as definite phenomena of the
same kind as the things, and accordingly are named and interpreted
by means of the same class of terms as are used for naming things.
The Eskimo verb merely forms a sub-class of nouns. (1911:1059)
2. Some Facts. There is considerable dispute as to whether languages of the
Salishan family (Haag 1998) or Nootkan (Jacobsen 1979) make distinctions at all
levels between nouns and verbs, but it is hard to imagine a poorer choice of a
language group to accuse of not having fundamental part-of-speech distinctions
than Eskimo. In all of the languages of this group there is a sharp formal contrast
between two classes of roots, stems, and words that is absolutely central to the
inflectional, derivational, and syntactic systems of the grammar. Furthermore,
this two-way formal distinction correlates directly with the same cognitive
complexes as characterize the noun-verb distinctions in European languages.
Thus the words for ‘house’, ‘mountain’, ‘father’, ‘milk’, and so on belong to one
class, while the words meaning ‘to walk’, ‘to see’, ‘to kill’, and ‘to give’ belong to
the other.
It is for this reason that every grammar and grammatical sketch of an
Eskimo language is organized around the distinction between noun and verb,
generally labeled as such, even those descriptions that are written under the
explicit assumption that no such distinction exists. Thalbitzer’s own description
from which the quotations above are taken divides the discussion of inflection
into large separate sections on nouns and verbs. His description of the all-
important derivational affixes of the language begins in a way that (cogently) flies
in the face of the abstract claims we saw above:
The suffixes are divided into two classes according to their use.
Some are employed to transform the nominal or verbal quality
of the independent word so that nouns are turned into verbs, and
verbs into nouns; others , merely develop the independent words
by enriching them .… Thus it may be seen, in regard both to the
suffixes and the initial stems, that a distinction may be drawn
between nouns and verbs, [and] nominal and verbal suffixes .…
(Thalbitzer 1911:1054)
A full discussion of the grammatical differences between nouns and verbs in
Eskimo would be a big undertaking – virtually exhausting the description of the
language, as a matter of fact – so I will confine my discussion here to
morphology, for it is precisely in that part of the grammar that it is commonly
claimed that little distinction is made.
3. Inflection. In West Greenlandic there are on the order of 150 inflectional forms
in the paradigm of the noun and around 700 forms in the paradigm of the verb. In
the majority of other dialects, where the dual is an active category, there are about
twice as many forms in each class.
The earmark of nominal inflection is case, of which there are eight. Seven
of these are clearly marked by overt, segmentable pieces of morphology, while
one of them – the absolutive – is unmarked. Each of the roughly 130 suffixes that
indicates case explicitly is therefore criterial for noun-hood. For illustrative
purposes we can use -tsinnut, an ending that encodes allative case (sometimes
called dative or terminalis) as well as the personal category first plural.
Hundreds of monomorphemic stems (i.e., roots) accept this inflection, and
hundreds of others do not: illu ‘house’, illutsinnut ‘to our house’: pisuk- ‘to
walk’, *pisutsinnut.
Mood is the criterial feature of verbal inflection. Neglecting the participial
mood sign, which happens to be homophonous with a deverbal derivational affix,
the other mood signs absolutely distinguish verbs from other categories in the
language. Thus the verbal mood sign of the transitive indicative is -pa-/-(v)a-,
which can be added only to certain roots and stems, and not others: pisuppa(a)
‘(s/he) walks (upon) it’ (from pisuk- ‘to walk’); but *illu(v)a(a)- (from illu
‘house’). Note that among inflectable roots, exactly those that reject nominal case
morphology, e.g. pisuk- ‘to walk’, accept verbal mood signs, and exactly those
that reject verbal mood signs, e.g. illu ‘house’, accept nominal case morphology.
The only complication is that there are a fair number of homophonous
roots and stems that (on one meaning) accept case endings and (on another
meaning) mood signs. However the alternation between noun and verb is
completely unproductive and the unpredictability of the meaning relation between
homophonous noun and verb pairs shows that this is mere etymology, as in the
first several of the following examples, or even lexical happenstance, as in the last
example: the root iga as a noun means ‘cooking pot’, and as a verb ‘to cook’; the
root sinik means ‘sleep’ when it bears nominal inflections, and ‘to sleep’ when it
bears verbal inflections; the root aak as a noun means ‘blood’, as a verb ‘to melt’;
and so on. In any case, there are hundreds of unambiguous items like illu and
pisuk- that accept case inflection and absolutely reject mood inflection, or vice
versa.
Now it is certainly true, as those who wish to deny the distinction between
nouns and verbs nearly invariably point out, that some of the personal endings
that are appended directly to noun stems and some of those that go on verb stems
plus mood inflection do resemble each other. Partly this is due to the fact that
there has been some vacillation on how to make statements in the history of
Eskimo, so that, as Bergsland (1964) put it: "What is a participle in one dialect
may be an indicative in another …". In any case, it should not be too surprising
that similar morphology is used in both nouns and verbs since, after all, the
inflections in question indicate the same thing: person and number. In a similar
way, we are not surprised to find the pronoun her in English doing double duty as
a verbal complement and a nominal modifier since the personal category that is
encoded – feminine gender, singular number – is the same in both cases. I
strongly suspect that in most if not all languages where the person and number of
possessors and the person and number of actants are inflections to nouns and
verbs respectively, there will be some resemblance between the two, since their
content, viz. person and number, is the same. This is true in Semitic, Quechua,
Hungarian, Crow, and so on.
But as a matter of fact, the personal inflections of Eskimo are not exactly
the same for nouns and verbs. A great many of the personal endings of verbs are
not to be found anywhere in the much smaller paradigms of nouns. There is, for
example, a third person singular accusative morpheme -gu/-uk that shows up in
numerous verbal forms (takumma-gu ‘when he/she saw it’, taku-uk ‘see it!’
takullu-gu ‘seeing it’, takugami-uk 'when he himself/she herself saw it', etc.), but
is never found in any nominal inflection. The first person intransitive verbal
suffixes -nga (singular) and -gut (plural) do not occur as nominal suffixes, and so
on for several other morphemes. If we consider the transitive desinences that
encode person and number information concerning both subject and object, we
find a huge range of endings unmatched by anything in the nominal paradigms,
including all combinations where the object is not a third person (e.g. takuaatigut
‘s/he sees us’, takuaanga ‘s/he sees me’, takuakkit ‘I see you (sg.)’, takuarma
‘You (sg) see me’, etc.) Thus a very large fraction of the personal endings
themselves are unique to verbs.
On the other hand, every indicator of person and number in nouns does
have an exact, or nearly exact, counterpart in some part of the much larger verbal
paradigm. This might lead one to believe that nouns are a subclass of verbs (as is
sometimes said of Hebrew, for example), but ever since Thalbitzer's original
pronouncement, the fact the nominal personal suffixes are a subset of the verbal
ones has been taken, with perverse logic, as an indication that verbs are a subset
of nouns.
4. Derivational Morphology. Just as with inflectional morphology, the highly
productive derivational system of Eskimo is organized around the distinction
between nouns and verbs. The majority of derivational affixes are quite specific
as to the class of stem to which they can be added, and quite determinate as to the
class of stem that they form. Of the approximately 500 productive derivational
suffixes of West Greenlandic, at least 300 are added unambiguously to one class
or the other. For example, the derivational suffix
-toqaq ‘old’ may be added to noun stems to create new noun stems (illutoqaq ‘old
house’) but absolutely cannot be added to verb stems like pisuk- ‘to walk’.
Conversely, the suffix -sinnaa- ‘to be able to’ is added to verb stems to form
extended verb stems (pisussinnaa- ‘to be able to walk’) and absolutely cannot be
added to noun stems: *illusinnaa-. Note that the two categories that result from
examining the derivational system coincide precisely with those that are obtained
by examining the inflectional system; illu ‘house’ belongs to one class and pisuk-
‘to walk’ belongs to the other, regardless of which subsystem of the morphology
we regard.
The two affixes mentioned in the preceding paragraph have the property of
preserving the category of the stem to which they are attached, but there are also a
great many that unambiguously attach to members of one class, converting them
to the other. The suffix -qar- ‘to have’, and at least fifty others, derives
unambiguous verb stems (according to both inflectional and derivational tests)
from unambiguous noun stems (by the same tests). Thus illoqar- ‘to have a
house’, but *pisoqar-. The suffix -v/ffik ‘place where’ unswervingly derives
unambiguous nouns from (intransitive) verbs: pisuffik ‘walking place’, but
*illuffik.
Perhaps 10% - 20% of the productive derivational affixes derive both
nouns and related verbs from the class of stems they apply to. A few affixes are
indifferent as to the class of stems to which they are added, producing nouns
when added to nouns stems and verbs when added to verb stems, but all the rest
are quite specific in their requirements and their effects. This has resulted in the
employment of a four-way classification of derivational affixes in Eskimo
something like that shown in (1) in every grammar the language following
Kleinschmidt (1851), who was the first to formulate things clearly in this fashion.
This even includes Thalbitzer's grammar and the others that deny the existence of
two classes. The classes are descriptively indispensable.
(1) INPUT | N V------------ |------------------------------OUTPUT |
| N | -toqaq -v/ffik
| ‘old’ ‘place where’|
V | -qar- -sinnaa-| ‘to have’ ‘to be able to’
These considerations show that at the level of roots (illu ‘house’; pisuk- ‘to
walk’), and at the level of derived stems (illutoqaq ‘old house’; pisussinnaa- ‘to
be able to walk’) as well as at the level of words (illorput ‘our house’; pisuppunga
‘I am walking’) the Eskimo languages (as exemplified here by West Greenlandic)
methodically distinguish two formal classes. Since this formal distinction is
correlated almost exactly with the semantic distinction between the nominal
concepts and the verbal concepts that “developed languages” represent by nouns
and verbs, there is no reason whatsoever for withholding those terms when
describing Eskimo.
5. History. The first mention that I can find of the idea that Eskimo is a verbless
language is in a little paper (Thalbitzer 1903) that was delivered in Uppsala less
than a year after the author returned from a 400 day field trip to Greenland (see
Thalbitzer 1904: 4.) The paper is pregnantly titled: “Studiet af et primitivt
sprog” (“The study of a primitive language”). In the third-person account of
Thalbitzer's presentation that has come down to us, we learn (in my translation)
that “because of his [Thalbitzer's] interest in the psychology of language, he had,
several years ago, given himself the task of studying a primitive language, and his
choice fell upon Eskimo.” (Thalbitzer 1903:51) In other words, Thalbitzer had
decided that Eskimo was a primitive language before he actually began studying
it. We learn also that the primitive quality of this language is most apparent
precisely in its failure to distinguish clearly between nouns and verbs:
“[Thalbitzer] thought he might, as a testimony to the primitive phase of language,
be able to point to the fact that for the Greenlander the boundary between noun
and verb is extremely vague, something that is demonstrated in the entire
structure of his language, and in the manner in which he uses its
forms.” (Thalbitzer 1903:58-9).
Schultz-Lorentzen (1873-1951), a Danish Lutheran minister born in
Greenland and a homespun grammarian, was also a subscriber to the view that
Eskimo (specifically West Greenlandic) did not distinguish nouns and verbs, and
he agreed that it had only the former. In his short grammar (Schultz-Lorentzen
1969), he rather poetically endorsed the idea that this supposed feature of the
language revealed a fundamental difference between the way Greenlanders view
the world and the way Europeans do. In my version of the Danish original:
In our discussion here of stems and stem words we have not drawn
any distinction between word classes as in other languages, not
even between nouns and verbs. It is true that there occur noun-like
and verb-like forms among the constructed stem words, just as the
affixes may have the sense of nouns and verbs respectively, and
impart that meaning to the stems when added to them. But the
difference between the forms is quite small, and the words are
formed with the same endings. This would seem to show clearly
that the Greenlandic mind draws no decisive distinction between
the two word classes, but expresses itself by means of a single
class, corresponding approximately to what we call nouns. The
words uvdloq, day, and nerivoq, eating, i.e., he eats; uvdlut, days,
and neríput, eating (in the plural), i.e., they eat; nunâ, his country,
and nerivâ, his eaten, i.e., he ate it, are so uniformly constructed
that they must be regarded as belonging to one word class. The
verb-like words have not then as in other languages the flowing,
streaming, active sense required to connect two nouns or noun-like
words, but they are limited, static, passive, like the nouns.
(Schultz-Lorentzen 1969:15-16)
Much of the work of Thalbitzer's student and successor in Copenhagen, L.
L. Hammerich (1892-1975), was guided by his conception of the verb as a
nominal plus “predicate of existence”. The idea is apparent in the peculiarly
strained glosses that he employs. In Hammerich 1936:214, for ernerma
panínguit asavâ we find the following gloss: “‘mein sohn liebt dein töchterchen’
eigentlich etwa ‘meines - sohnes - dein töchterchen [ist] seine liebe’ oder uns
natürlicher, ‘dein töchterchen ist seine meines sohnes liebe’”. Schultz-Lorentzen
also resorted to odd glosses in an attempt to capture what he felt was the nouny
gist of the language, as we can see in the passage cited above where, for example
nerivâ ‘he eats (or ate) it’ is rendered as hans spiste, i.e., ‘his eaten (thing)’.
Thalbitzer also seems to be the originator of the nominal glosses, as in the
following passage from his article in the Handbook (Thalbitzer 1911:1058):
Accordingly, the inflected verb [sic!] in the indicative intransitive
is properly translated in this manner:
atorpoΝa my use = I am used
...
tikippoΝa my arrival = I arrive
...
The same applies to the transitive forms of the finite verb; thus –
atorpara my its use = I use it
Note the impenetrable double possessive in this last example. Apparently, the
Eskimos not only conceive of everything as thing-like and possessable, but
simultaneously possessable in different ways by distinct individuals.
Surely the most famous linguist ever to lend his name to the nominalist
view of Eskimo was Leonard Bloomfield. In his early work, An Introduction to
the Study of Language (1914), published when he was only 27, Bloomfield
describes at length the class of "objective", or "nominal", or "attributing"
languages, as he variously calls them. In the following excerpted passage, we
hear the echo of von Humboldt and Wundt in a distinctly Bloomfieldian accent:
There are numerous languages, especially on the American
Continent, which have not gone beyond the naming of objects....
[O]ne cannot, at this stage, speak of 'white' or of 'runs', but only of
such objects as 'white rabbit' or 'running rabbit', .... Every word is
an object expression; qualities or actions are never as such
expressed by separate words .... Hence in these 'nominal' or
'attributing' languages such utterances as 'white rabbit' correspond
equally to our predication 'It is a white rabbit' and to our attributive
'white rabbit' .... (Bloomfield 1914:63-4)
"Greenlandish", according to Bloomfield 1914, is just such a language:
The explicit predication of quality or action is impossible
for languages in which every word expresses an object (p. 64). In
these languages the sentence consists of one or more object-
words... The Greenlandish [qim:eq] thus can mean 'dog' or 'It is a
dog', [saλ8:uto:qaoq] 'big liar', or 'He is a big liar', or 'He lies very
much', and so on. (Bloomfield 1914:111)
The forms are good, if somewhat archaic (West) "Greenlandish", but
their translations are only half right. Qimmeq 'dog' is only a noun form and can
amount to 'It is a dog' only in the same circumstances that the English a dog can,
e.g., as an answer to a question such as What's that? The homophonus stem is
only nominal and cannot accept verbal derivations or inflections, and so on.
Sallutooqaaq, on the other hand, is an unambiguous verb form meaning 'he/she is
a big liar.' The word can never be used in a sentence in a nominal position. The
stem /sallutuuqE-/ fails to inflect for case or unambiguous possessor and will not
support unambiguously denominal derivational affixes.
The question immediately arises: Where did Bloomfield get his
information? His 1914 book is poor on direct attribution of sources. It does,
however, contain a section toward the end enticingly entitled: "How to study
linguistics" (p. 313ff), an annotated bibliography of the then-available linguistic
literature. There are no works listed there that are entirely devoted to
Greenlandic, but there is an approving reference to a work that is hardly known
today (and which I, certainly, had never heard of before discovering it in this
connection): Franz Nikolaus Finck's (1909) Die Haupttypen des Sprachbaus, a
small volume consisting of sketches of eight languages that Finck felt represented
the eight major types linguistic systems. One of these is - you guessed it - die
grönländische Sprache. All of Bloomfield's examples are carefully copied,
including mistakes, from this work.
Finck devotes fourteen pages to a highly psychologistic discussion of
Greenlandic. At one level, his analysis is quite good: words are (with a few
important exceptions) correctly divided and the sense is (for the most part)
accurately rendered. The oddities arise in connection with the noun-verb
distinction. As an illustration of the non-distinctness of the two major classes,
Finck asks us to compare palasi-uwu-nga (palasiuvunga in the modern
orthography) 'ich bin Priester' ('I am a priest') with Kin-uwu-nga (i.e., qinuvunga)
'ich bettle' ('I beg'). The comparison, in this case, is exact because both forms are
verbs. But the morphologization is crucially incorrect. Finck, who generally
overanalyzes Greenlandic, has here missed the verb-forming element -u- 'to be a
__' in palasi-u-vu-nga. Furthermore, he misdivides what should be qinu-vu-nga,
(verb stem qinu- 'to beg, entreat') so that it seems to have the same construction as
the verbalized noun.
Now it might seem from Finck's translations that in fact, the verbal notion
is alive and well in Greenlandic, but he goes on to explain (in my translation):
The process (Vorgang) expression which forms the core of
all sentences including those which, in reality, do not deal with an
activity at all, is in no way to be interpreted as a German verb. It is
distinguished from such ... in terms of a strong objective character
which displays, instead of an indication of subject, possession-
indicating pronouns (to be sure sometimes rather veiled) and it
represents the action not as a fact ... but as an impression ... (Finck
1909:35)
His subsequent translations reflect this analysis. Thus Finck qimmip
takuaa 'the dog sees him' is ' ... (dem) Hunde Erscheinung-dessen-dessen ...', or
something like ' ... (to the) dog appearance-his-his ...' (Finck 1909:40). Except for
the Finckian peculiarity of translating what means 'see' as 'appearance', the style
of interpretation is highly reminiscent of Thalbitzer's double possessive
renderings. In Thalbitzer 1903: 60, for example, takuva-anga 'he/she sees me' is
Scandinavianized as hans min-seen or 'his my-seeing'.
Bloomfield borrowed this idea, but attempted to make some sense of it in
his glosses. Where Finck translates tukkarpaa (i.e., 'he/she tramples it'), which he
incorrectly writes with a single /k/, as 'Trampeln-dessen dessen', Bloomfield
writes (with an accurate copying of Finck's inaccuracy) '… [tukarpa:] as
'stamping-of-it-to-him' ….' But Bloomfield cheats: his rendition is an English
nominalization, which of course includes the verbal idea, or so we are inclined to
think.
Now we must also ask where Finck got his ideas about Greenlandic.
Neither of his thumbnail biographers, Lewy (1953) and Planck (1996), is able to
tell us. As far as I know, the short text that Finck presents is found nowhere else.
So it appears that he must have done his own informant work, just as we know he
did for a number of other languages. But the translations are so Thalbitzerian in
style that I cannot help but believe that Finck knew "The Study of a Primitive
Language." The date of Finck's work (1909) makes this a definite possibility, but I
cannot prove that he didn't come to this view independently of Thalbitzer.
On the modern scene, two Canadians who adhere to radically different
theoretical frameworks have recently advocated discarding the noun-verb
distinction for Eskimo. Working within the theory of Guillaumian
Psychomechanics of Language (e.g. Guillaume 1974), Ronald Lowe (1981)
minces no (French) words in lambasting the idea that nouns and verbs exist as
parts of speech in Inuktitut. To his credit he does not claim that there are only
nouns, but rather that no such classes exist at all. He writes:
Numerous errors have been committed up to now by those
linguists who have classified the radicals [of Inuktitut] ... under the
grammatical categories of noun and verb. A radical such as pisuk-,
for example, has always been considered as a verb because it
denotes a process (the idea of walking) and a radical such as nuna-
as a substantive, because it denotes a substance: the earth.
One sees here clearly that what serves as a criterion in this style
of classification is of a purely of semantic order – the opposition
process/substance – a criterion that has moreover been very badly
exploited in this case since the idea of a process is not, in itself, in
any way connected with the verbal category, as the existence in
French of words like course, marche, descente, etc. demonstrates,
which, though they denote processes, are nonetheless substantives.
(Lowe 1981:77)
But we have seen that the criteria for distinguishing nouns and verbs in
Eskimo are not of a purely semantic order. They are as formal as they can be.
Regardless of any semantic considerations, we would still arrive at the same two
classes on distributional grounds alone. It is the fact that the two formal classes
align (at least in the case of underived stems) with semantic distinctions of
process versus substance that justifies our calling them nouns and verbs (as
opposed to, say, class A and class B). It is not the case, pace Lowe, that semantic
considerations are all there is that leads the analyst of Inuktitut to postulate a
bifurcation.
Alana Johns (1987a,b), to whose work I return in more detail below, has
revived a weakened version of the Thalbitzer-Hammerich view of the nominal
character of Eskimo, without attributing any psychological ramifications to it.
The analysis I am about to present is based on the ideas of early
grammarians studying Eskimo, in particular Thalbitzer (1911) and
Hammerich (1951)... The cornerstone of this approach is that there
is no category verb at the level of word in Eskimo. This means
that all words in Eskimo are nominal in some sense to be made
more precise. (Johns 1987b)
Unlike the “early grammarians”, Johns is very clear on the issue of the
level of words at which there are only nouns in Eskimo. She argues clearly that
there are two kinds of roots and two kinds of stems, making use of the very kind
of evidence that I have used here. The “early grammarians”, however, were either
not clear on this subject or were of a different opinion. Thalbitzer’s glosses seem
to indicate that both stems and words are nouns and for Hammerich, the verbal
endings add a “predicate of existence” making the words apparently verbs, while
the stems from which they are derived, it would seem, are nominal. Schultz-
Lorentzen says in the quotation above that there are no word-class distinctions
either among stems or among “stem words”, (the smallest independently
pronounceable forms).
Johns' references to the “early grammarians studying Eskimo” is a bit
misleading. Up to Thalbitzer's time, I can find no hint of a suggestion that
Eskimo failed to make a fundamental distinction between nouns and verbs. In the
eighteenth century grammars of Top (1727), Egede (1760) and Fabricius (1791),
this is hardly surprising, since those grammars were modeled more-or-less
directly on the grammars of Classical languages, where the distinction between
nouns and verbs is, of course, unquestioned. Even in the eighteenth century,
however, the absence within Eskimo of a morphosyntactic class corresponding to
the class of adjectives in Classical languages was obvious. Thus Top 1727:121
clearly notes that there is no separate class of adjectives; verbs, he notes, are used
to predicate adjectival notions and are nominalized to form modifiers of nouns.
It is significant that there is no trace of the idea that Eskimo has only
nouns in Kleinschmidt’s stunningly original and insightful grammar of West
Greenlandic (Kleinschmidt 1851), a work in which the author explicitly set about
to liberate himself from the straitjacket of the classical grammars. Kleinschmidt’s
entire system of grammar is organized around the difference between nouns and
verbs in Greenlandic, and at every level he contrasts the properties of these two
grammatical classes. This emancipation from the bonds of tradition enabled
Kleinschmidt to follow Top in dispensing entirely with a class of adjectives, but
he felt it necessary to keep nouns and verbs apart in every respect. Rink (1887),
following Kleinschmidt, put it succinctly: "It is a peculiarity of the language that
NOUNS AND VERBS ARE ALMOST THE ONLY PARTS OF
SPEECH" [emphasis original].
In a wonderfully keen passage, Kleinschmidt points to a fundamental
formal characteristic that distinguishes “noun words” from “verbal stems”: the
former can occur as independent words without overt inflection; except for an
interjective use, the latter cannot. I translate:
Stems divide into two great groups in other respects besides
whether they are free or bound, namely into 1) noun-words, which
name a (physical or mental) object according to its characteristic
nature, e.g. igdlo house, auk blood, ujarak stone, nano bear inuk
human being; – and 2) verbal stems, which describe something
according to a general property (activity) that pertains to it, e.g.,
ajoq bad, pisuk going, tikit arrived, mike small. The former – the
noun words (nomina, i.e., names) - can be directly employed and
inflected; the latter, in contrast, are neither in and of themselves
inflectable or usable at all, except that in some dialects they can be
used as interjections (e.g. inugo ajoq! Oh, how terrible!), but
become both [i.e., usable and inflectable] only when transformed
into verb words (verba, i.e., words in the preferred sense) by means
of a further expansion, namely by means of an added mood sign
that consists of several pieces. (Kleinschmidt 1851:11)
Here Kleinschmidt not only demonstrates the formal dichotomy, he also tells us
that it correlates exactly with the psychological distinction that separates nouns
and verbs in so-called “developed” languages. It is only after Thalbitzer that this
correlation came into doubt.
Despite the clarity with which Kleinschmidt demonstrated the existence of
both nouns and verbs at the level of stems in Eskimo, Nowak in her recent book
on Kleinschmidt (Nowak 1987), is able to make the claim only two pages after
quoting the very passage that I have translated above that "both classes of stems
can be used both nominally as well as verbally (p.66)". To back up this claim, she
cites Kleinschmidt’s stem forms pisuk, tikit, and mike as if they were whole
words. But of course, as Kleinschmidt said, they are not words of the language at
all, let alone nouns. They are abstract verbal stems.
6. The anti-Whorf hypothesis. Let us consider the equation: categories of mind
= categories of language that we have found invoked explicitly or implicitly by
advocates of the nominalist view of Eskimo. There is obviously a strong
anticipation of Whorf 1956 (or reprise of von Humboldt 1836) to be found in this
identification. Some fairly direct connection between the form of language and
its cognitive value must be assumed for the conclusion to follow that, if Eskimo
has only nouns, its speakers think entirely in concrete terms. That this assumption
is tacitly made is amply demonstrated by such remarks as Thalbitzer’s: “to the
Eskimo mind the nominal concept of the phenomena of life is predominant.” For
Thalbitzer, then, as for modern scholars such as Hopper and Thompson (1984) or
Langacker (1987), such notions as noun and verb are (to a greater or lesser extent)
complexes of grammatical and mental fact.
But remarkably, Thalbitzer’s only official pronouncement on the
connection between linguistic form and thought is a radical denial of the
Whorfian hypothesis. Immediately following the passage introducing his thesis
concerning the lack of distinction between cognitive-grammatical classes in
Eskimo, we find the following illogical and methodologically perplexing
discussion, which I translate from the Dano-Norwegian:
Of course the language [Eskimo] does make a formal distinction
between nominal and verbal designations; there are overt criteria
for those words we call verbal. But that does not necessarily
indicate that the words in -poq and -voq are essentially different
from nouns.… Just as the similarity of the forms in a language
(e.g., English) does not necessarily entail identity of the concepts
of the corresponding word classes, the dissimilarity of the forms
does not necessarily mean that the concepts are sharply
distinguished in two disjoint word classes. (Thalbitzer 1903:59)
The necessity for such an extraordinary methodological hedge is clear.
Thalbitzer had decided, as we saw above, that Eskimo was a “primitive
language”. Apparently one of the things it means for a language to be primitive is
for it to fail to allow for the expression of subtle cognitive nuances (such as the
difference between entity and event) that find easy expression in more developed
languages. Now English, it goes without saying, is a developed language. Yet in
English the grammatical distinction between nouns and verbs is in fact quite
slippery. In Eskimo, which is assumed to be a primitive language, there are, as
shown above, crystal clear formal distinctions between nouns and verbs.
Therefore, the manner of expression cannot necessarily reveal the underlying
thought, else English would be the primitive language and Eskimo the developed
one! Thalbitzer is therefore forced to accept the Whorfian hypothesis in some
cases, and something like its opposite in others. The vertiginous circularity that
such a methodology makes possible needs no further comment.
7. The nominal character of the single class. Suppose, contrary to fact, that
there were only one part of speech in Eskimo. How would we know that it
corresponds to our nouns? Why couldn’t Eskimo be an entirely verbal language
(which might or might not then imply that its speakers conceive of everything as
dynamic, fluid, and active)? Or more reasonably, as Lowe 1981 suggests, why
could it not be the case that there is simply a neutral class, neither nominal nor
verbal, just as there is a neutral class of pronouns in Eskimo that fails to
distinguish masculine from feminine? (Cf. The discussion of Nootka in Sapir
1921:103-106 and 133-134.) The only argument I find in the earlier literature that
addresses this question (if one can even call it an argument) is in Thalbitzer’s
earliest paper:
One of the most remarkable things about the Eskimo inflectional
system is the almost completely consistent relationship between
the possessive suffixes for both nouns and verbs. The role that the
possessive suffixes plays goes far beyond the use to which we put
possessive suffixes in our languages. The possessive suffixes of
the verb [sic!] show themselves to be identical with the possessive
suffixes of the noun (my, your, his, his own), which can be taken as
proof of the nominal character of the verbal inflection. Even the
verb-forming -poq thereby becomes indistinguishable from a
nominal formation. (Thalbitzer 1903:60)
This argument turns on the a priori identification of the personal elements
of words as possessive suffixes. Possession is a category of the noun; therefore
all forms bearing these affixes are nouns. But where is the argument that the
personal suffixes are possessive? Why could they not be basically argument-
referencing suffixes, in which case the so-called nouns could be construed as a
sort of verb, in the fashion of generative semantics? The answer again must be
that Eskimo is prejudged to be primitive, and names for concrete things (nouns)
are more primitive than names for abstractions (verbs). Adam first named the
birds and the beasts, and only much later, we must presume, actions and events.
Alana Johns’ work referred to above (Johns 1987a,b), though purporting to
be a continuation of the ideas of Thalbitzer and Hammerich, has a much less
troubling character, partly because of its much greater sophistication and more
careful and consistent working out than what is found in the earlier work, partly
because of the fact that it specifically denies any psychological correlates of
nominalism, maintaining that there are, in fact, two lexical-semantic classes in the
language, but not least because it seems that the nominalist thesis itself is not an
absolutely necessity in Johns’ treatment of Inuktitut. She admits as much in a
footnote in her dissertation where she credits Paul Hirschbuhler with the
observation that “... if there is no difference between nouns and verbs, then the
use of the term noun is somewhat misleading.” (Johns 1987a:86-7) A lengthy
description of Johns’ intricate thesis is therefore not required here, but a short
outline of her views will, I believe, suffice to show that little in her proposal really
depends in a crucial way on its avowedly nominalist stance.
Johns argues that there are two classes of stems in Inuktitut corresponding
exactly to noun and verb stems and having exactly the semantics of noun and verb
stems, but not, she urges, the kind of syntax that we expect from noun and verb
stems. She calls these classes of stems “referential” and “predicative” and defines
them semantically. In other languages such stems are “projected in the syntax” as
noun and verb, but Eskimo, according to Johns, does not permit surface verbs.
Thus the predicative “... forms a defective category, that is, it cannot project a
syntactic category of its own, but must attach to a morpheme that can – the mood
markers.” (Johns 1987a:87) In this way Johns derives Kleinschmidt’s fact - verb
stems cannot be used without mood inflection. Now, since after a mood marker is
attached, the predicative stem can "project a syntactic category", then by her
assumption, the category must be N, not V. Predicative stems of Eskimo are then
(rather counterintuitively) syntactically nominalized by means of the mood
affixes.
This thinking might explain the inflectional and syntactic similarities
between mood-inflected verbal stems and nominally inflected referential stems,
but, as we have seen, the inflectional behavior of the two word types is not
entirely the same (as Johns herself observes) and, more importantly, the syntax of
the two word classes is, in fact, not the same. Johns provides clear evidence that
there are two classes of syntactic words that freely combine with each other – but
not with themselves – to form clauses. In particular, no two of the predicative
stems “nominalized” by mood markers can combine to form a clause. Thus one
can say in Inuktitut taku-vara sinik-tuq “I see the one who is sleeping”, where the
first word is an indicative mood form inflected for a first person singular ergative
and a third person singular absolutive, and the second word is an absolutive-case
(nominal) participle, but one cannot say *taku-vara sinik-puq, substituting an
indicative mood verb in the third person singular for the participle (Johns
1987b:ex. 10a,c).
The reason for the different syntactic behavior of the two kinds of words is
that “indicative mood markers carry some semantic feature which is necessarily
expressed in INFL or I ...”. Insofar as the mood markers are virtually obligatorily
followed directly by personal inflection, this claim is surely true in some theory-
neutral terms. But let us see where that leaves us: Eskimo has lexico-semantic
predicatives that express just what verbs do in other languages and lexico-
semantic referentials that express just what nouns do in other languages. Only the
predicatives may combine directly with mood markers and then inflections to
form one class of syntactic words. Only the referentials may combine directly
with case inflection to form a second class of syntactic words. The two classes of
syntactic words combine with each other, but not with themselves, to form
clauses. So why not just recognize these two syntactic classes as verbs and
nouns? It is not necessary to assume that there are no verb words to account for
the defectivesness of the predicative stems. It seems to me much simpler and
more straightforward to say that the defect is morphological, as Kleinschmidt did
nearly a century and a half ago: "verbal stems ... are neither in and of themselves
inflectable or usable at all ... but become both only when transformed into verb
words ... by means of ... an added mood sign." It seems to me, in other words,
that Johns’ work is nominalist in name only.
8. Derivational interconvertability. As mentioned above, Eskimo languages
contain numerous suffixes that convert nouns to verbs, and quite a few that
convert verbs to nouns. It is not unusual to find this conversion happening two or
three times inside a single word: qimmeq ‘dog’ – qimmeqar(poq) ‘to have a dog
(or dogs)’ – qimmeqartoq ‘one who has a dog (or dogs)’ – qimmeqartoqanngi(laq)
‘for there not to be anyone who has a dog (or dogs)’. Now one would think that
the possibility of converting between two categories would be extremely strong
evidence for the existence of two categories, else what sense would there be in
talking about conversion? Yet with remarkable suspension of common sense, the
very fact of interconvertability has several times been used to argue that the
categories are the same.
For example, Lowe (1981) writes:
A few examples suffice to show clearly that pisuk- [‘to walk’] and
nuna- [‘land, country’] are no more noun than verb, since in fact
one encounters them both in forms that designate a process as well
as a substance. Thus, by way of example:
pisuktuq : il marche, celui qui marche
pisungniq : la marche
pisuinnautivaa : il l’apporte en marchant..
nunannguaq : carte géographique
nunaarutivaa : il l’emporte à terre
nunatuqpuq : il mange du végétal
(Lowe 1981:77-8)
It is important to realize that in each of Lowe’s examples there is a quite
specific derivational affix that determines whether the root is encountered in a
derived stem that designates a process or in a derived stem that designates a
substance. There is no productive zero derivation in Eskimo that could obscure
the boundary between nominal and verbal roots and stems such as there is in
English (Clark and Clark 1979). The reason we encounter pisuk- ‘to walk’ as a
nominal word in pisungniq ‘the fact or act of walking’ is because it bears the
unambiguous deverbal abstract nominalizing affix -niq. The reason we encounter
nuna ‘land, country’ in a verbal form in nunatuqpuq ‘(s)he eats vegetable matter’
is because here it bears the unambiguous denominal verbalizing affix -tuq- ‘to
eat __ ’. Rather than demonstrating the identity of the classes, this sort of
interconversion by overt suffixation actually demands that we recognize their
distinctness.
Thalbitzer also used the interderivability of the two classes as evidence of
their essential unity. Furthermore, with a truly spectacular leap of illogic, he took
the convertibility to show that nouns are basic. I translate once more:
Given the nearly reckless shifting of nominal representations for
verbal ones and vice versa – and it is precisely the ease with which
the alteration takes place that bears witness to the mutual affinity
of the word classes in linguistic consciousness – it appears as if it
is the nominal conception of phenomena that has the upper hand.
What is verbal in life seems to be reflected nominally in the
Eskimo language. (Thalbitzer 1903:61)
9. Inflectional convergence. Let me return to the fact that there are parts of the
paradigms of nouns and verbs that contain similar or identical personal affixes.
Particularly striking is the fact that the possessed, absolutive-case forms of nouns,
and indicative, third-person object, transitive verbs have strikingly similar
personal inflections. Data like those in the following table are cited in nearly
every work that proposes that Eskimo is a mono-categorial language:
NOUN VERB
(absolutive, possessed) (indicative, transitive)
illuga ‘my house’ takuara ‘I see it’
illukka ‘my houses’ takuakka ‘I see them’
illut ‘your house’ takuat ‘you see it’
illutit ‘your houses’ takuatit ‘you see them’
illua ‘his house’ takuaa ‘he sees it’
illui ‘his houses’ takuai ‘he sees them’
Once again, this practice goes back to Thalbitzer, who observed the coincidence
of inflection in his 1903 article. In the same paragraph where he makes this
accurate observation, however, we find one of Thalbitzer’s very few analytical
mistakes, for there he also tried to relate a non-third-person intransitive verbal
suffix to a nominal form, when, in fact, they are quite distinct. Thalbitzer
compared the word that he transcribed: ι:υΝα ‘my house’ with the word that he
transcribed: τικιπ:υΝα ‘I arrive’, asserting that the personal suffix is the same in
both. But he was familiar with other works on Eskimo, citing Kleinschmidt 1851
and Fabricius 1791, among others, and thus he must have known that in other
dialects, there is [⊗] in the noun, but [Ν] in the verb. In the modern standard
orthography for West Greenlandic, these words are illuga and tikippunga,
respectively. In the Northern variety of West Greenlandic that Thalbitzer was
using for illustration, the contrast between intervocalic [Ν] and intervocalic [⊗]
has collapsed to [Ν], a phonological development that Thalbitzer described
accurately in his 1911 sketch. Furthermore, in most other persons and numbers,
the endings are very obviously different. For example ‘your (sg.) house’ is illut
(suffix -t), and ‘you arrive’ is tikipputit (suffix -tit). The coincidence of the
endings for the first person singular in Northern West Greenlandic is therefore just
that: a coincidence.
While it is perhaps unfair to accuse Thalbitzer of disingenuousness, at
least he fooled himself into imagining identity of nominal and verbal inflection
where it doesn’t exist and where he should have known that it doesn’t. His
delusion persisted in a way in his Handbook article (Thalbitzer 1911) as indicated
by his translation of the intransitive of verbs as possessed nouns (e.g., tikippunga
‘my arrival’) despite the fact that the form is not in general the same as the
possessive of nouns. Even in the Northern subdialect most forms of the possessed
noun are different from those of the intransitive verb; in the first person plural of
the verb we find tikippungut ‘we arrive’ (Central West Greenlandic tikippugut),
whereas the possessed noun is illorput ‘our house’, with a different desinence.
The form that would be parallel to the verb, namely *illungut, does not exist.
The only other paradigmatic backwater where there is a resemblance
between the substance of the nominal and verbal inflections is in the possessed
relative (i.e., ergative) case of nouns and the intransitive subordinate mood forms
of verbs, e.g., illuma ‘my house (relative case)’ – tikikkuma ‘when I arrive
(future)’; illutta ‘our house (relative case)’ - tikikkutta ‘when we arrive (future)’.
So far as I know, only Hammerich essays an explanation of this particular
syncretism in terms of the supposed identity of nouns and verbs – and he comes to
grief over it, as we shall see presently.
Even in those isolated inflectional regions where there is a general
resemblance between the inflection of nouns and verbs, attention to detail reveals
differences. For example, in West Greenlandic all vowel-final intransitive verb
stems geminate the /v/ of the indicative mood sign to /pp/ in the third plural :
takuvoq ‘he/she sees’, but takupput ‘they see’. While many nouns have
geminating plurals, the pattern is different: First of all, there are no nouns in
which /v/ alternates with /pp/, and secondly the gemination in nouns takes place
with any consonant-initial inflectional suffix, not just with plural /t/. If things
were really parallel in the nominal and verbal paradigms, we should expect not
only parallelism between naalagaq/naalakkat ‘master/masters’ and takuvoq/
takupput ‘(s)he sees/they see’, but also between naalagaq/naalakkatit ‘master/
your masters’ and takuvoq ‘(s)he sees with a geminating form *takupputit. But
the correct form is takuvutit ‘you see’ without gemination.
10. False facts in the name of nominalism. Another difference between the
inflection of the two great word classes in West Greenlandic is that there are
certain noun forms, namely those with a high final vowel, that make no
distinction between third singular and third plural possessors of plural nouns. The
word illui means both ‘his/her houses’ and ‘their houses’. In indicative verbs,
however, the vowel before the ending is always [a], so (as with nouns ending in
[a]) the distinction is always registered: takuai ‘he/she sees them’ - takuaat ‘they
see it/them’.
The quest for parallelism between the nominal and the verbal paradigms
led at least one researcher, Elke Nowak (1987), to forget about the peculiarity of
nouns ending in a high vowel and to invent erroneous forms in the noun that
distinguish singular and plural possessors as if they were verbs. She asserts
(Nowak 1987:69) that ‘his houses’ in West Greenlandic (in Kleinschmidt’s
orthography) is *igdluai, and ‘their houses’ is *igdluait, the correct forms being
igdlue (modern spelling illui) in both cases. If the stem had been igdlua, she
would have been right, and since this was her first foray into the morass of
Greenlandic inflectional morphology, the mistake is perhaps forgivable. The
larger point is that it was the erroneous theory that nouns and verbs are alike in
Eskimo that led Nowak to present false facts as if they were true.
The most remarkable example of pure invention in the name of the noun-
only theory of Eskimo is to be found in Hammerich’s contribution to – of all
things – the Kleinschmidt centenary volume of IJAL, (Hammerich 1951). In his
paper in that volume Hammerich sketched a highly original and in some ways
quite insightful reanalysis of Eskimo structure that was based on what he saw as a
fundamental distinction in the grammar of these languages between
superordination and subordination. One of the virtues of the theory was supposed
to be that it applies equally to what he termed “nominal nouns” (i.e., nouns) and
“verbal nouns” (i.e., verbs).
A little background on the basic morphosyntax of Eskimo is necessary to
understand Hammerich’s theory. There are two purely grammatical cases of
nouns in Eskimo, the absolutive, used for the subject of intransitive clauses and
the object of transitive clauses, and the relative (a.k.a. ergative) used for the
subject of transitive clauses and the possessor of possessed nominals. A relative-
case nominal is always associated with an inflected head that agrees with it, either
the verb of which it is the transitive subject or the noun of which it is the
possessor. Thus piniartu-p (hunter-relative/3s) illu-a (house-absolutive/3s/s) ‘the
hunter’s house’; piniartu-p (hunter-relative/3s) taku-aanga (sees-indicative/3s/1s)
‘the hunter sees me’. Hammerich saw in this the subordination of the relative
case to a superordinated possessed noun or transitive verb, thus explaining the
partial similarity in the inflectional forms in these two categories.
It was Hammerich’s idea that these two independent binary oppositions,
+/-superordinate and +/-subordinate, formed the two principal axes of Eskimo
grammar. Thus there are four “cases” corresponding to 1) neither subordination
nor superordination (which Hammerich calls “absolutive”) - 2) superordination
(his “superordinative”) - 3) subordination (called “subordinative”) - and 4) both
superordination and subordination (which he dubs the “duplex”). In Hammerich’s
words:
Case 1, the ABSOLUTIVE, is the base, the lexicological form, e.g.
igdlo hut, nanoq, bear.… If used in a verbal noun of certain moods,
the form signifies that the notion of the verb exists, has reality, is,
e.g. tikípoq does not simply signify arrival, but arrival is, there is
arrival, somebody or something has arrived.… By personal
suffixes of a non-possessive character [!] a subject of the first,
second, and – in some moods – fourth person may be expressed,
e.g. tikípunga there is I-arrival, i.e. I have arrived.(1951:20)
Since the subordinative is identified only as the relative case of nouns (nominal
nouns, in Hammerich’s terminology) the theory in question recognizes a
significant asymmetry between nouns and verbs (i.e., the “verbal nouns”.)
Hammerich makes a feeble attempt at an explanation of the difference as follows:
Case 3, the SUBORDINATIVE, is not found in purely verbal
nouns – probably because the predicate of existence which is
characteristic of verbal nouns, does not go well with the pure
subordination which is the essence of Case 3. (1951:21)
The parallelism between nouns (Hammerich’s nominal nouns) and verbs
(his verbal nouns) returns when we get to Case 4, the duplex. In nouns the
duplex will, quite logically, display the sum of the properties of the subordinative
and superordinative noun. A noun in the duplex form will be one that is both
possessed and in the relative case, that is to say, something that is itself
superordinate to a possessor and at the same time subordinate to either a transitive
verb whose subject it is, or to another possessed noun whose possessor it is. A
noun like arnaata ‘his mother (relative case)’ can be superordinate to another
subordinative (i.e., relative case) noun: piniartup arnaata ‘the hunter’s mother
(relative)’ and at the same time must be subordinate to a superordinative noun or
verb, as in piniartup arnaata illua ‘the hunter’s mother’s house’ or piniartup
arnaata takuaanga ‘the hunter’s mother saw me’.
Hammerich says that “[i]n verbal nouns [i.e., verbs] of some moods the
duplex is used to express clauses.” (1951:21) Thus a form such as tikikkutta
‘when we arrive’ is taken by Hammerich as filling the same slot in the paradigm
of verbal nouns as does the duplex arnatta ‘our mother (relative case)’, and
tikippat ‘when he/she arrives’ is the verbal correspondent, according to
Hammerich, of arnaata ‘his/her mother (relative case)’. These ideas about the
parallelism of nouns and verbs suggest to Hammerich that in clauses formed with
what he identifies as a duplex verbal noun: “What we translate as the subject
must, of course, be in the subordinative; corresponding to piniartup arnâta . . .
the hunter’s mother . . ., we have piniartup tikigpat . . . when the hunter
arrives. . .” (1951:22)
But (of course) we shouldn’t have a subordinative here. Hammerich’s
*piniartup tikigpat is wildly ungrammatical; the correct form is piniartoq tikípat
with absolutive subject, since this is an intransitive clause. The fallacious
example given by Hammerich is as ungrammatical as *when him arrives would
be in English, and always has been, so far as I can tell.
It is almost incredible that a scholar of Hammerich’s distinction and
learning could have been so blinded by his own theory as to fabricate data of as
fundamental a kind as this. But he did. Such is the seductive power of the
intriguing, romantic, but clearly false notion that Eskimo languages do not
distinguish nouns and verbs, and the racist correlate that the people speaking these
languages conceive of everything in concrete terms.
Footnotes