Senft, Gunter, ed. (2000) Systems of Nominal Classification. Cambridge
University Press, ISBN 0-521-77075-0, vii+350pp, Language, culture and
Previous review at http://linguistlist.org/issues/12/12-645.html#1
Arvi Hurskainen, Institute for Asian and African Studies,
University of Helsinki.
Although the problem of nominal classification has been a subject
of discussion and research at least since the appearance of Die
nominalen Klassifikations-Systeme in den Sprachen der Erde (Royen
1929), the subject continues to remain fragmented and we still are
far from satisfactory treatment of this vast and complicated
field. Recent attempts to illuminate the subject include
Apprehension (3 vols.) by the Cologne group (Seiler and Lehmann
1982; Seiler and Stachowiak 1982; Seiler 1986), Noun Classes and
Categorization (Craig 1986), and the more recent Gender in Grammar
and Cognition (Unterbeck and Rissanen, eds.), which the book to be
reviewed does not recognise. Because the book tries to cover all
types of languages as well as all variations of nominal
classification, without consciously excluding any area or type,
the task is formidable. It would be easy to point out issues that
are either left out or that are not treated satisfactorily, but
this is necessarily the case with a book such as this. Therefore,
it is more important to see how the authors have succeeded in what
they intended to accomplish.
Although the geographical field is global, the theoretical
viewpoint is defined more narrowly 'to clarify the interface
between anthropological and grammatical work on nominal
classification'. Whether this approach has made the work
more manageable is another issue, since the combination of two
distinct research fields, each with established concepts and
prejudices towards the other, does not make the work easy.
Although most of the chapters have what is vaguely
called 'an anthropological linguistic perspective', the
contributions are independent articles rather than integrated
chapters of a single book. This notion should not be taken as a
critic, however, because detail and faithfulness to data should
not be sacrificed for an integrated theory. In fact, although the
book has an ambitious title, it is the individual contributions,
some of which are excellent and very detailed, that are most
valuable in it. The research team of the Department of Language
and Cognition at the Max Planck Institute for Psycholinguistics in
Nijmegen had a coordinating role in compiling this book, which is
partly based on articles presented in the workshop entitled 'Back
to Basic Issues in Nominal Classification', held in Nijmegen in
More specifically, the book addresses such questions as: What
are the basic parameters, and/or the basic semantic distinctions
that are expressed in nominal classification? What is actually
classified - extra-linguistic referents or the kinds of nouns
within the language? What functions do the various types of
nominal classification fulfil and how do they function in
discourse? To what extent do they presuppose obligatory
distinctions and to what extent do they allow freedom in
allocating category? The overall fundamental question is: How is
the perceived world expressed in, and through, various systems of
nominal classification that are grammatically encoded in various
The first chapter by Gunter Senft outlines the basic
questions to be dealt with in various chapters of the book.
Drawing data from Kilivila, an Austronesian language of the
Trobriand Islanders, he describes the complex use of the so called
classificatory particles in this language and suggests that the
issues relevant in the so called 'classifier languages', which
Kilivila is a member of, are more general and are relevant also
for other systems of nominal classification. Senft lists and
describes briefly various systems of nominal classification found
among languages. Such systems include: (a) languages that classify
the nouns according to kind and degree of possession (Kilivila
being an example of such a system); (b) languages that utilise
classificatory noun incorporation, where a generic noun is
syntactically incorporated into the verb and cross-classifies a
specific noun which is syntactically governed by the verb; (c)
classification by verb, where verb roots provide a semantically
transparent classification of the intransitive subject or
transitive object; (d) the so-called 'classifier languages' with
numeral classification, where, when counting inanimate as well as
animate referents, the numerals concatenate with a certain
morpheme, which is the so-called 'classifier'; (e) noun class
systems of nominal classification with a fairly high degree of
grammaticalization, a closed system of classes and a limited
degree of semantic transparency (e.g. Bantu languages); and
(f) gender systems where the determining criterion is agreement.
As Senft points out, this list is not exhaustive, and there
is overlap between categories. Especially the terms 'noun class
system' and 'gender system' are not clear, and some researchers
(e.g. Corbett) consider noun classes as part of the wider gender
system. Finally there are languages that apply the classifier
system and the noun class system simultaneously, which gives
reason to suspect that a language may be in a process of
transition from one system to another. How this happens is still
Senft continues by discussing the inaccurate and partly
misleading terminology used in analysing classifier languages. He
points out that the term 'numeral classifier language' is partly
inaccurate, because classifier morphemes in many languages are
used also for other purposes, not only for counting animate and
inanimate referents. He joins those researchers who have proposed
'numerative' as a generic term, which then subdivides into two
distinct sub-classes, 'classifiers' and 'quantifiers'. There also
occur such terms as 'sortal classifier' (individuates whatever it
refers to in terms of the kind of entity that it is) and 'mensural
classifier' (individuates in terms of quantity).
In order to get more clarity to the use of categories, Senft
proposes that semantically based groupings should be accepted only
if they are grounded in the grammar and marked as being formally
distinct in the respective language. Because classifications are
not only formal, identifiable in linguistic forms, but basically
also semantic, this has often resulted in grouping nouns of a
language under different semantic labels, these groups
constituting 'semantic systems' or 'semantic domains'. He proposes
an analysis process, where the first and temporary semantic
categorisation, made by the researcher, is reconsidered and
amended later after having looked at their actual use in detail.
Finally, the semantic categories should not be considered static
but rather as dynamic domains which interact with the total system
The brief description and analysis of the Kilivila classifier
particles is given as a 'case study' for demonstrating in practice
what was said before on a more general level. The argumentation is
quite convincing as regards the Kilivila system and other similar
systems. It is perhaps also an important contribution to the
analysis of classifier languages, but less relevant in
understanding other types of nominal classification systems.
The second chapter also has a general scope. In it, Colette
Grinevald gives a typology of classifiers on the basis of
morphosyntactic properties. She places the typology of classifiers
into the context of the more wide system of nominal
classifications. Three major types of classifiers are
distinguished, and the main purpose of the treatise is to 'produce
a research tool that might foster the production of more thorough
and more comparable descriptions of classifier systems'. The main
motivation for working out a typology of classifiers emerges from
confusion in work on classifiers. Because the use of terms by
various researchers is not uniform, it has resulted in serious
misunderstandings, especially when data are used by other
researchers for comparative purposes.
Grinevald places the classifiers in the middle of a
continuum, where in its grammatical end are the gender and noun
class systems, and in its lexical end are measure terms and class
terms. Classifiers have a lexical origin and they are used in
specific morphosyntactic constructions. They are different from
lexical systems in that they mark categories of nouns by
independent morphemes or by affixes that are attached to other
words than the noun in question. They are also different from noun
class systems and gender systems because of their incomplete
grammaticalization. She identifies eight criteria through which
noun class-gender systems differ from classifier systems. The main
areal distribution of various systems is that gender systems are
found in Indo-European languages, noun class systems in Niger-
Congo languages (especially Bantu languages), and classifier
systems in East and Southeast Asian languages.
Within classifier languages she distinguishes such types as:
(a) numeral classifiers (e.g. Thai and Burmese in Asia); (b) noun
classifiers (e.g. Jakaltek and Akatek in Meso-America); (c)
genitive classifier systems (Micronesian family in Oceania); (d)
verbal classifiers (found in languages of North America). In
addition to these categories that she calls 'prototypes' there is
discussion on 'fuzzy edges', cases that do not fall to any of
those categories. There is blending among the classifier systems
themselves as well as blending with other nominal classification
systems. Furthermore, there are systems, such as the Amazonian
languages exemplify, that a language may have a gender system and
another type of nominal classification system (resembling a noun
class system) simultaneously. This would contradict the often-held
view that the gender system and noun class system should be
collapsed as one category.
The typology proposed by Grinevald is based on morphosyntax.
She realises the need of investigating the semantics and functions
of the classifiers, 'before it will be possible to provide a
functionally based comprehensive study of classifiers'.
While acknowledging the main types of nominal classification,
Alexandra Y. Aikhenvald gives a condensed account of such
'marginal' classifier types, which do not fall within those
categories. Among such classifier types are classifier morphemes
known by such names as 'demonstrative', 'article' and 'deictic'
classifiers, which appear only with deictic elements. Tariana, a
North Arawak language from Northern Amazonia, is an example of
such a language. It uses almost the same classifier morphemes for
such purposes as for verbal, possessive and noun classification,
as well as numeral classifiers and noun class markers. The
language also exhibits a distinction between classifiers with
demonstratives and classifiers with articles.
After a condensed (yet detailed) analysis of each of these
classifier types a concluding summary of the properties and uses
of classifiers in Tariana is given in a table. The chapter is
clearly ordered and the argumentation is convincing, and its
strength lies in that it concentrates on one language and tries to
give a full account of its classification system. Although the
book is on classification systems in general, this account on an
individual language brings us down to earth and reminds us about
the extreme complexity of systems which we have to deal with.
Roberto Zavala describes multiple classifier systems in
Akatek, which belongs to Mayan languages. Akatek has four
different paradigms of classificatory devices, which in Zavala's
terms are: classificatory suffixes, sortal numeral classifiers,
plural for humans, and noun classifiers. What is interesting in
this system is that more than one, even three, different
classificatory morphemes can co-occur in the same nominal phrase.
He claims that Akatek requires classifiers because the nouns of
the language do not make distinction between singular and plural.
Each of the four sets of noun classification has a double role.
They can function as adnominal elements or as pro-forms that
maintain the referent continuity in discourse. The use of
classificatory suffixes depends on morphosyntactic factors, but
the use of other three classificatory devices depends primarily on
discourse-pragmatic factors. What is interesting is that,
according to Zavala, none of the classificatory morphemes of those
four sets of classifiers add new semantic information to the noun
with which it combines. On the basis of phonological changes he
proposes that the originally lexical items have become purely
Chapter 5 contains a reanalysis by David Wilkins' own earlier
account of lexical noun classifiers of Mparntwe Arrernte, which
belongs to the Pama-Nyungan languages in Central Australia. Noun
classifiers are typically free morphemes standing next to the
noun. Such morphemes are often considered polysemous, because in
addition of being part of a nominal construction, they can also
stand as nouns of their own. In this chapter Wilkins takes a
different stand and, instead of treating noun classifiers as
polysemous morphemes, he emphasizes that the classificatory
morphemes should not be considered as nouns in classificatory
contexts but rather as part of a larger construction, where the
preceding morpheme expresses a semantically wider group, which the
following more specific noun is a member of. Therefore, emphasis
should not be in individual parts of the construction but rather
in its semantic content. Therefore, although formally such a
construction could be analysed as consisting of two adjacent
nouns, where the former is generic in nature and the latter
specifies it, semantically they constitute one single concept. The
construction is comparable with multi-word terms, a phenomenon
common in coining domain-specific terminology in most languages.
So far twenty-six generic terms for Mparntwe Arrernte have
been identified. And Wilkins suggests that any lexicalised
superordinate term that has identifiable lexicalised hyponyms can
occur as the generic in a generic-specific construction.
Wilkins makes explicit what he calls his theoretical biases.
He claims that generic nouns should not be treated as having one
meaning as individual nouns and another meaning as part of
generic-specific constructions - a situation common in many
dictionaries. Rather, generic lexemes are monosemous, and the
second 'meaning' is a function of the construction, not the
property of the lexeme. He further claims that constructions
themselves have a unique and non-compositional meaning. The third
'bias' concerns compositionality. It is claimed that while the
meaning of a linguistic structure is a function of its parts and
their manner of combination, this includes also that one of the
'parts' is the construction itself. The meaning of the
construction is not a concatenative process, as if the meaning of
one part would be added to the meaning of the other. Lexical,
constructional and compositional sources of meaning should not be
confused. This implies the fourth 'bias', namely the necessity of
having a method of semantic description. What method should be
employed is left open, lest one uses the method systematically.
Finally Wilkins insists on the distinction between semantics and
In building a model of the use of generic nouns in Arrernte
Wilkins includes into it also a close discourse analysis and
culturally grounded definitions. He also examines the use of such
constructions in rhetorical speech as well as in creative and
humorous contexts. When considering Arrernte data in comparison
with other Australian languages, he concludes that most of them
have merely generic-specific constructions without having
developed true noun classifiers.
Kyoko Inoue gives an account of the use of numeral
classifiers in Japanese. Attention is paid especially to the
extra-linguistic cultural factors in assigning classifiers to
constructions. Results of research with children and people with
various levels of skills in Japanese language show that classifier
choice is not constant but it depends on several non-linguistic
factors. One important factor in classifier choice seems to be
visualizing ability, i.e. how speakers can mentally view and
manipulate a noun. In fact constellations of cultural and social
knowledge override formal and semantic factors.
Jurgen Broschart argues in his contribution that the
linguistic function of classifiers follows the principles of
perception and cognition. Basing his argumentation of Gestalt
Theory he tries to find common ground for discussion between
different sciences of mind. He describes the function of
classifiers and classifier-like elements in the context of the
gestalt principles of the isolation of units and the unification
of isolates. Numeral classifiers as well as mensural and
collective classifiers define, and thus isolate, units of various
kinds. Noun classifiers are concerned with 'naturally'
independent, self-contained referents, while classifier-like
constructions (possessive and locative) specify, and thus unify,
In trying to widen the perspective to other areas of the
sciences of mind, Broschart contends that "there is no difference
of principle between the function of a classifier in linguistics
and e.g. the function of a concrete 'contour' or 'gestalt' in
visual perception". The same claim applies to morphological noun
class markers and gender markers, because they serve the purpose
of identifying manipulable entities. The difference between noun
classifiers and gender markers is the specification of domain.
While noun classifiers operate on text level, specifying e.g. the
topic of a text and are not compulsory, gender markers are
obligatory for any form of a lexical noun. The gestalt operations
(isolation of units and unification of isolates) are crucial for
informing the hearer about the manipulability and identifiability
of the parts of the domain in question.
The noun class system, one of the major nominal
classification systems, is discussed by Katherine Demuth in the
chapter on Bantu noun class systems. This group of languages is
perhaps the best-known representative of this classification
system. Noun classes are not, however, treated here from a
comparative viewpoint, for example by comparing them with other
types of classification systems. The author has chosen a group-
internal problem and she explores the nature of the semantic
productivity and competing morphophonological processes as well as
acquisition of the use of noun classes by children. The evidence
is from such languages as Sesotho, Setswana, Zulu and Siswati.
Demuth gives an outline of the Proto-Bantu noun class system
(with 25 classes) and shows how some Bantu languages and languages
distantly related to Bantu have retained noun classes in different
degrees. She is interested in the question why some noun classes
(especially 1/2 and 9/10) are more likely than others to persist
in languages that lose classes. Various attempts to provide
semantic basis for noun classes show that such a basis indeed
exists, although the precise way of forming and labelling such
groups is problematic. Some features, such as humanness and
animacy, are clear criteria for noun class allocation, but in some
other classes, such as 5/6 and 9/10, the semantic basis is much
more complex. It is not clear either whether the noun classes
should be seen to reflect semantic entities as distinct units or
whether those units form a hierarchical structure in the way Denny
and Creider (1986) have attempted.
Demuth shows how Sesotho and Setswana are examples of
different degrees of noun class productivity, especially in regard
to locative classes. On the other hand, some other classes, e.g.
the human class 1/2 seems to be productive in most Bantu
languages, as are some other classes. Perhaps the most interesting
part in this chapter is the loan word allocation, because this is
done on the basis of competing criteria. The general rule in
Sesotho seems to be that phonology and semantics are decisive, and
when neither of these two applies, the noun is assigned to the
default class, which in Sesotho is 9/10 and in Zulu 5/6. The study
of the noun class acquisition by children shows that they may make
grammatically motivated errors but not semantic
The contribution of Demuth would be a fairly well documented
piece of work on selected topics if she in the end had not
ventured to contemplate the 'necessity' of noun classification. To
compare noun classification with the classification of verbs with
respect to tense/aspect is a far-fetched idea. That 'the
classification of nouns is a semantic (and grammatical) necessity'
is simply not true, because there are languages, for example in the
Finno-Ugric family, which have no trace of nominal classification,
if we exclude personal pronouns in 3rd person where a distinction
is made between humans and non-humans.
Greville Corbett and Norman Fraser give a logically concise
account of gender assignment and demonstrate with the examples of
Russian and Arapesh (Papua New Guinea) how a formal approach to
gender typology suits for predicting gender allocation. They give
a typology of gender systems on the basis of how nouns are
assigned to genders, the basic types being semantic assignment,
predominantly semantic assignment and formal systems
(morphological assignment and phonological assignment). The
approach of using formal properties for predicting gender
assignment was tested by an inheritance formalism called DATR,
developed for encoding Network Morphology. The argumentation is
John Lucy has written a concluding chapter, where he
pinpoints three conceptual and methodological confusions in the area
and shows how they can be remedied. Especially he points out that we
need to put emphasis on the analysis of local linguistic forms,
especially on the patterns of formal meaning constructions
chracteristic of the language in question.
The book claims to give an account of the state-of-the-art in the
study of nominal classification in various languages. As far as no
competitive account is available, as the case is, it has achieved
this aim. It contains a wealth of data and scrupulous analysis
from this very wide and difficult field. It aims at terminological
clarity, and it has taken at least a step, if not a leap, into
that direction. Although the level of contributions is somewhat
uneven, I predict that it will become a standard reference work in
this field for the years to come.
My worry is that if distinction between such concepts as noun
class system and gender system is blurred we lose the
distinctiveness of those terms. Bantu languages are classical
examples of a system where nouns are classified according to
several semantic criteria, with no reference to sexual gender.
Many other languages, including the Afro-Asiatic family, tend to
classify nouns according to sexually defined gender. In some
languages, as in the Nilo-Saharan Maasai, the gender division is
very productive and nouns may switch between those two genders
because of semantic purposes (male and big vs. female and small).
It is likely that terminological distinctions will survive because
of practical reasons.
Craig, Collette (ed.), 1986. Noun classes and categorization.
Amsterdam: John Benjamins.
Denny, J.P. and C. Creider, 1986. The semantics of noun classes in
Proto-Bantu. In C. Craig (ed.), Noun classes and categorization.
Amsterdam: Benjamins. 217-39.
Royen, Gerlach, 1929. Die nominalen Klassifikations-Systeme in den
Sprachen der Erde. Historisch-kritische Studie, mit besonderer
Beruecksichtigung des Indogermanischen. Anthropos Linguistische
Bibliothek vol. 4. Vienna: Anthropos.
Seiler, Hansjakob, 1986. Apprehension. Language, object and order,
part 3: The Universal Dimension of Apprehension. Tuebingen: Narr.
Seiler, Hansjakob, and Christian Lehman (eds.), 1982.
Apprehension: das sprachliche Erfassen von Gegenstaenden, part I:
Bereich und Ordnung der Phenomene. Tuebingen: Narr.
Seiler, Hansjakob, and Franz-Joseph Stachowiak (eds.), 1982.
Apprehension: das sprachliche Erfassen von Gegenstaenden, part 2:
Die Techniken und ihr Zusammenhang in Einzelsprachen. Tuebingen:
Unterbeck, Barbara and Rissanen, Matti (eds.), 1999. Gender in
Grammar and Cognition. (Trends in Linguistics: Studies and
Monographs 124). Berlin: Mouton de Gruyter.
About the reviewer:
Arvi Hurskainen, professor of African languages at the University
of Helsinki, Finland. Main areas of interest: Bantu languages and
linguistics, computational description (morphology, syntax and
semantics) of Bantu languages (mostly Eastern and Southern),
conceptual systems (esp. Maasai).