LINGUIST List 26.3113

Wed Jul 01 2015

Review: Ling Theories; Morphology; Semantics; Syntax; Typology: Masini, Simone (2014)

Editor for this issue: Sara Couture <>

Date: 11-Mar-2015
From: Volker Struckmeier <>
Subject: Word Classes
E-mail this message to a friend

Discuss this message

Book announced at

EDITOR: Raffaele Simone
EDITOR: Francesca Masini
TITLE: Word Classes
SUBTITLE: Nature, typology and representations
SERIES TITLE: Current Issues in Linguistic Theory 332
PUBLISHER: John Benjamins
YEAR: 2014

REVIEWER: Volker Struckmeier, Universität zu Köln

Review's Editor: Helen Aristar-Dry


As the editors to this volume point out, the question of word classes goes back to Greek antiquity. However, basic questions regarding the nature, typology and representation of word classes remain unsolved:

- What are the definitions of word classes in specific languages?
- Can the same categorizations be applied to different languages (or even universally)?

The editors specifically consider two approaches to categorization: ''a typologically-oriented approach'' and a more ''theoretically-oriented one'' (p. 4), structuring their book accordingly:

Typologists attempt to describe and compare parts-of-speech systems of unrelated languages. Therefore, Part I of the book gathers articles that address properties of major word classes (verbs, nouns, adjectives, etc.). Part II includes articles on lesser-studied word classes (e.g., classifying nouns, cardinal numerals, preverbs, etc.)

Theoretical linguists attempt to derive categories from grammatical properties needed for a grammatical model to operate, e.g. semanto-pragmatic, or distributional ones.
 As a result, Part III of the book discusses cases where categoriality seems to fail (e.g., category/function mismatches, or decategorization). Part IV questions the need for the very concept of word classes (e.g, to be replaced by features or abandoned for words which cannot be assigned to any category in the first place).

The volume thus runs the gamut from empirical descriptions to theoretical reflections. In the following, some short summaries are given for the individual articles contained in this edited volume.

Allesandro Lenci discusses problems of various approaches to the categorization of verbs: His article “Carving verb classes from corpora” argues that ontological classifications base solely on ''our intuitions or presumptions“ (p. 19). Lenci therefore goes on to develop a typology of distributional subcategorization frames (SCF): What number and types of arguments does a verb appear with? Corpus analyses then allow us to compute the frequency for each SCF, reflecting selectional preferences of the verb. However, the polysemy of certain verbs may be overlooked in a computerized corpus analysis (p. 27). Therefore, Lenci analyzes certain SFCs in-depth.

In “Classes of creation verbs”, Elisabetta Jezek proposes to differentiate two classes of creation verbs: ''create verbs“ and ''verbs of derived creation''.
- Create verbs (p. 42ff.) such as ''coin'', ''invent'' have 'creation semantics' as their primary, and often only, meaning. They have a human agent and an artefact object and show change-of-state properties (resulting in the existence of the object).

- Verbs of derived creation (p.44ff.) can have readings where no argument comes into existence (e.g. ''paint a dozen portraits'' vs. ''paint the walls white''). Subject arguments can be humans, other concrete entities, or even have abstract reference.

In “On Light Nouns”, Raffaele Simone and Francesca Masini present graded categorizations: Subclasses of verbs are well-investigated in this regard, but the authors investigate 'nouniness' scales for light nouns, which serve as modifiers of other nouns in constructions like ''N1 of N2“ (''sort of a mouse'', ''bit of bread'', ''fit of temper''). Simone and Masini propose a scaled referential force (RF): high RF points towards the 'nouny' end of the nouniness scale, whereas lower RF is found with 'less nouny' elements: Taxonomic nouns, e.g., have low RF (''a kind of conclusion“ is a ''conclusion“). However, support nouns are more nouny in this regard (e.g. ''un colpo di pistola“ is indeed a ''hit“ or ''shot“, p.64).

Denis Creissels establishes two types of adjectives in Tswana “The ‘new adjectives’ of Tswana”: ''New adjectives“ differ from elements traditionally considered adjectives in class agreement morphology and have other, nominal uses (p. 75ff.). These elements constitute an emergent class of adjective which, the author argues, cannot be captured in analyses that represent them as relative clauses.

In “The Chinese adjective as a word class”, Giorgio Francesco Arcodia also discusses adjectives. Even though some authors claim that Chinese lacks adjectives, Arcodia argues that adjectives occupy intermediate positions on a scale between nouns and verbs. In Chinese, the adjectives simply turn out to be rather ''verb-like“ (p. 97), but are not verbs:

- Some adjectives constitute predicates without a copula (p. 100)
- their negation is similar to verbal, rather than nominal negation (p. 100), and

- both verbs and some adjectives require a particle ''de“ to function as modifiers (p. 102).

However, ''de'' is not obligatory for all attributive uses of adjectives (p. 104)
- adjectives have derivational affixes from verbs (p. 106), and

- semantically, adjectival and verbal reduplication differ in forms and meaning (p. 105f.).

Luca Alfieri, in “Qualifying modifier encoding and adjectival typology”, argues that categories are not primitives but should derive from the ''constructions defining them“ (p. 119). For example, languages might differ in that:

- they separate object modifiers, quality modifiers and action modifiers

- or lump together the action/quality kinds

- or they lumped together the object/quality kinds (p. 121).

The author goes on to portray different languages (Latin, Quechua, Hausa, Lavukaleve, Vedic, Lao, etc.) and demonstrates their empirical differences. From these findings, Alfieri concludes that their common function is the basis for their categoriality, so that elements only display categoriality in their relation to functional notions.

Aniko Csirmaz and Éva Dékány discuss light nouns in Hungarian in “Hungarian is a classifier language”: Hungarian bare nouns denote an undifferentiated mass (p. 142), so classifying elements are added to a nominal construction in order to denote individuals. The authors point out the following subtypes:

- Syntactically, sortal classifiers come between lexical nouns and numerals or quantifiers. Group classifiers, on the other hand, must come in ''structurally higher“ (p. 153). Many adjectives must precede sortal classifiers, whereas group classifiers can be preceded or followed by all adjectives (p.153f.).
- The meaning of group classifiers is not affected much under NP ellipsis (p. 154f.), whereas sortal classifiers will shift their meaning to the N that they are homophonous with.

Rossella Pannain and Anna Riccio show that, cross-linguistically, both 'adjectival' or 'nouny' types of numerals can be found (p. 165f.), in “Cardinal numerals: A syntax-semantics interface analysis”. Numerals show a semantically uniform function across languages, but the formal implementation of this function differs cross-linguistically (p. 168f.). The authors present analyses for nominal (p.169-72), as well as adverbial uses (p. 173ff.). Pannain and Riccio conclude that a unified analyses of cardinal numerals is constituted by their common functional core (expression of cardinality) – rather than by specific morphosyntactic implementations of this function.

Anna Sörés proposes a functional word class called ''path satellites'', based on an analysis of Hungarian preverbs, in her “On the borders of neglected word classes”: The article claims that these elements cannot be subsumed under the usual adverbial categories, despite the fact that this category is the least uniform cross-linguistically (p. 184). Hungarian preverbs

- denote the direction or orientation of a movement or other actions (p. 186)

- interact with aspectual interpretations (p.187)

- can co-occur with verbal structures (p. 191f.);
but furthermore they

- can occur independently from modified structures in some discourse settings (p. 188).

The article presents cross-linguistic comparisons with a variety of related elements (e.g., in German, Italian, Rama, Polish, Ancient Greek, and Jakaltek). The cross-linguistically common function of various elements establishes for them the category ''path satellite''.

Peter Lauwers investigates non-canonical uses of members of word classes in French, e.g. the use of adjectives as nouns and vice versa (e.g., ''des costumes très théâtre''). Null derivations are rejected as analyses, since the elements in question never quite fit the canonical properties of the 'other' word class:

Adjectives used as nouns
- are modified by adverbs (''le facilement accessible'', p. 206)

- cannot be modified by adjectives, except for some specific exceptions (p. 207), and

- cannot be picked up referentially by anaphoric pronouns (p. 208).

Lauwers proposes constructions to represent the resulting structures, as well as the properties of the adjectives that fit into the larger construct in the first place (p. 217). Also, the article proposes a construction to represent nouns that are used like adjectives (p.221):

- Not all can be modified by adverbs (p. 218)

- The ones that do are preceded only by very specific adverbs (''très'', ''assez'', etc., p. 220),

Livio Gaeta discusses decategorization phenomena in German: The right sentence bracket, according to the author, displays ''low categoriality“ (p. 227). 'Decategorization' here refers to the (synchronic) loss of categorial properties of an element (without shift to a different category):

- Inside compounds, the first morphological element loses its inflection (e.g, to test-drive is ''testfahren'', not ''*testefahren“, p. 235).
- In complex predicates, adjectives are used without their inflection, too.

Haritini Kallergi investigates reduplication effects: The paper argues that these differ across languages with regard to their meaning effects: Nominal reduplication often intensifies meanings (e.g., reduplication of ''woman'' means 'very feminine woman' and reduplication of ''white'' yield 'very white' in Greek, p.248). Verbal elements, however, show aspectual meaning effects (''say say'' means 'to say constantly' in Greek). However, these effects are not replicated across languages (p. 250):

- Rennelese reduplicates verbs for intensification (''want want'' = 'want very much')

- Malagasy reduplicates nouns for continuation (''speech speech'' = 'repeated speech').

Kallergi therefore proposes to predict reduplication effects on the basis of individual semantic features (countability, distributivity of event arguments, gradability, etc) rather than categories.

Maarten Janssen discusses ''unique words'': English ''half“, e.g., can be used as an adjective, determiner, noun or quantifier – but operates quite unlike other members of those categories. To establish the properties of elements, Janssen presents a Corpus Pattern Analysis (CPA), to investigate ''semantic [...] behavior of words from a corpus in terms of corpus patterns“ (p. 269). Interestingly, the multiple patterns an element like ''half'' can occur in have a striking effect on the structure of the lexicon: Categories cease to be a taxonomy of categories with subcategories (and sub-subcategories, and so on). Rather, the established patterns arrange in ''a lattice, in which nodes can have more than one parent. This means that in CPA, it is possible that fire and toss are both ''throw“ verbs, while toss is also a ''give“ verb, whereas fire is not'' (p. 280).

In sum, the book presents many analyses (some of which are quite detailed) for various members of various word classes: The typological variation within what is considered 'a single word class' is pointed out very nicely in many of the articles. However, the book contains very few 'deep' theoretical analyses of the syntacto-semantic kind that would define categories, or help elucidate the notion of categoriality in the first place (as, e.g., Baker 2003, Sasse 1993, or Wunderlich 1992 do). Readers who wish to learn more about the theoretical background of word classes will most likely be disappointed by this book.


The book is highly successful in the empirical presentation of categorial scales, variation within categories across languages, and in presenting vexing cases of uncategorizable elements. However, the data often point at larger questions that the authors sadly do not attempt to address: Why would, e.g., the subdistinctions of categories presented in some papers matter, while further subdistinctions can somehow be ignored? What level of granularity, or detail resolution, do categorizations need? Given that any answer to this question will be at best 'relativistic' vis-a-vis the intended application of the categorization, it may be wiser to avoid decisions on the granularity of categorizations in the first place. Features, however, readily serve as predictors for the behavior that an element will ultimately show in morphological, syntactic, and larger contexts. Combinations of these features will define ever smaller sets of elements, ending in singleton sets (where every element is in 'a class of its own'). These feature sets would thus cut across established (but often euro-centric) word classes and provide to every purpose exactly the level of granularity it requires. Kallergi‘s article, e.g., points out very clear-cut properties of structures (options for reduplication, and their semantic effects) which can be tied to the semantic features of the reduplicated elements. Therefore, these features can serve as non-circular explanations for items' behavior, explaining (not restating) the facts.

On the other hand, circularity occurs in other analyses: We simply cannot ever explain the behavior of elements by assigning them to some word class, as long as the assignment of elements to word classes is handled on the basis of the very same properties. The circularity is obvious: Why do elements a, b, and c belong to category X? Because they have the properties associated with X. And why do a, b, and c have these properties? Because they belong to category X! The same problem befalls construction-based approaches, too: By generalizing the fact that some elements a and b can occur in some construction Y, the construction, in turn, cannot be used to 'explain' the properties of a/b. What is needed, instead, is an independent explanation as to why a/b operate in the context of Y in the first place (while another element c may not).

Some theoretical claims may also seem a little too grandiose: Why, e.g., would functional or semantic notions be prior to formal distinctions, as some of the authors argue? The fact that functions can be tracked across languages even if formal implementations differ, in fact demonstrates very little here: Formal theories can always argue that categories arise on a language-specific basis precisely because these individual languages only allow for certain feature combinations to be used in their grammars, which (by assumption) differ mostly in formal, rather than functional aspects. The question of function- versus form-oriented grammars will quite simply not be solved by pointing out lexical elements with peculiar properties in some language, since grammars of all persuasions can easily accommodate these facts.

In many articles in the book, there is also a palpable disregard for 'rigid' categories, which is understandable in view of the data discussed. However, some authors associate these rigid categories with generative syntax – incorrectly, I think: Indeed, nothing at all is gained when some authors demonstrate quite sardonically that 'classic' categories like N, V, A, etc. cannot be used in re-write rules to yield the observable data. Re-write rules have not been in use in many generative grammars for at least the last 35 years! The shift from ''top-down'' structure generation to ''bottom-up'' projection of lexical properties that dramatically changed, e.g., Chomsky'an proposals from the late 1970es onwards has obviously failed to make much of an impression on some authors included here. No article even only mentions the fact that the very of notion of 'syntactic category' has been abandoned in generative grammars from the mid-1990es onwards (cf., e.g., Chomsky 1995).

In sum, the theoretical conclusions drawn in this book may fail to influence future theoretical research. However, the book will be welcome to all who are looking for empirical testing grounds for their categorization theories: Many of the gradual (and not so gradual) linguistic distinctions underlying the categories presented here will hopefully lay one simplistic assumption to rest: A handful of 'word classes' handed down from antiquity will not help us in defining satisfactory linguistic theories. Given the lack of attention this important fact often receives, this is a very welcome result indeed.


Baker, Mark C. 2003. Lexical Categories: Verbs, Nouns, and Adjectives. Cambridge: Cambridge University Press.

Chomsky, Noam. 1995. The Minimalist Program. Cambridge: MIT Press.

Sasse, Hans-Jürgen. 1993. Das Nomen - eine universale Kategorie? Sprachtypologie und Universalienforschung 46-3. 187-221.

Wunderlich, Dieter. 1992. Lexical Categories. Theoretical Linguistics 22. 1-48.


Volker Struckmeier is a senior lecturer at the University of Cologne, Germany. He has worked on word classes and questions of categoriality in language, elements of questionable categorial status (e.g. modal particles) and hybrid categories (e.g. participles), as well as different phenomena (e.g., information structure, scrambling, attribution structures, and the syntax-prosody interface more generally). He is currently carrying out acoustic experiments that investigate the interaction of word order, prosody, and semantic effects in German scrambling phenomena. Future goals include the empirical validation of his scrambling theory in visually presented stimuli, and resulting effects for preferred and dispreferred word orders in written texts.

Page Updated: 01-Jul-2015