Linguistic Typology 2016; 20(2): 377393

William Croft
Comparative concepts and language-specific
categories: Theory and practice
DOI 10.1515/lingty-2016-0012
Received March 7, 2016

Abstract: What are comparative concepts and how are they related to language-
specific categories used in language description? Three general categories of
comparative concepts are defined here: purely functional comparative concepts
and two types of hybrid formal-functional concepts, constructions and strate-
gies. The two hybrid types provide more explicit and precise definitions of
common typological practice. However a terminological issue is that Western
grammatical terms are frequently used to describe strategies which are not
universal rather than constructions which are. Language-specific categories
appear to be radically different from comparative concepts because the former
are defined distributionally whereas the latter are defined in universal func-
tional and formal terms. But language-specific constructions have functions,
that is, they are instances of constructions in the comparative sense and their
form is an instantiation of a strategy. Typology forms generalizations across
language-specific constructions in both their form and their function. Finally, a
major issue is the confusion of terminological choices for language-specific
categories. Four rules of thumb for useful labeling of language-specific cate-
gories, largely following best descriptive practice, are offered.

Keywords: categories, comparison, construction, methodology

1 Introduction
A recent debate on the Lingtyp e-mail discussion list (see Supplementary Online
Materials) once again raises the issue of comparative concepts in typology and
their relationship to the description and analysis of the grammars of particular
languages. There has been much debate on this topic by many scholars in the
typological literature recently, including myself. Here I offer definitions of both
comparative concept and specific types of comparative concepts, and compare
them to language-specific categories. Much typological practice informally

William Croft, Department of Linguistics, University of New Mexico, MSC03 2130, Albuquerque,
NM 87131-0001, U.S.A., E-mail: [email protected]

Unauthenticated
Download Date | 11/18/16 7:17 PM
378 William Croft

conforms to the theoretical notions advocated by typologists such as Matthew

Dryer, Martin Haspelmath, and myself. Other practice does not, but in fairly
consistent ways that usually do not obscure interpretation by a sympathetic
reader.
The question arose in the Lingtyp discussion (and earlier as well) whether
there is any relation at all between the comparative concepts used by typologists
to formulate universals and the language-specific categories used by descrip-
tive/field linguists to produce grammatical analyses of the language in question.
I believe that there is a strong relationship, despite major differences between
comparative concepts and language-specific categories. Practical suggestions
can be made to strengthen the connection between typological theory and
descriptive practice.

2 Comparative concepts: Functional and hybrid

Haspelmath (2010) defines comparative concepts as theoretical concepts used
for crosslinguistic comparison, in typology or other approaches to grammatical
theory for that matter. He describes two broad categories of comparative con-
cepts. The first type are purely semantic concepts; I use the broader term
FUNCTIONAL, so as to include what I will call (following Chafe 1976 and others)
information packaging functions (also called information structure, discourse
structure, or discourse function).
The use of purely functional concepts as comparative concepts is relatively
uncontroversial, but by no means entirely so. The main issue here is that those
who support semantic relativity in a fairly strong form would argue that seman-
tic concepts are language- or culture-specific and hence not usable as compara-
tive concepts. There are significant issues with language-based semantic
relativity arguments (Croft 2001: 108131, 2010). I will not survey those argu-
ments here, but I will bring out one issue arising in the semantic relativity
debate that requires an important qualification in using semantic or functional
concepts for crosslinguistic comparison.
In general, extensionally large semantic categories that are given mono-
semous definitions (that is, Aristotelian categories defined by necessary and
sufficient conditions) do in fact often fail as comparative concepts. For example,
property concepts taken as a broad category (stative, unary valency, gradable,
inherent) do not serve well as a comparative concept for understanding adjec-
tives, because semantic subclasses of property concepts age, dimension,
color, value, etc. exhibit different grammatical behavior in one and the same

Unauthenticated
Download Date | 11/18/16 7:17 PM
 Comparative concepts 379

language (Dixon 1977). The usual solution to this problem is to use finer-grained
categories, such as the property subclasses, or for core participant roles, the
division into A, S, and P.
This solution has a problem as well, namely that one may need to make
further divisions, for example, the division into Sa and Sp, in order to capture
crosslinguistic variation. Semantic categories conceived of as an exhaustive
partition of conceptual space are probably not going to serve well as compara-
tive concepts because there will be some language that is grammatically variable
within even those fine-grained categories. Instead, what typologists do is use
points or very small regions in conceptual space for crosslinguistic comparison.
Classic examples of this approach are questionnaire surveys such as Dahls
(1985) tense-aspect questionnaire, or the stimulus sets created at the Max
Planck Institute for Psycholinguistics at Nijmegen, e.g., for spatial relations
(the Bowerman-Pederson set; Levinson et al. 2003) or for cutting/breaking
events (Majid et al. 2008). These stimuli provide a denser distribution of com-
parative concepts in particular regions of conceptual space. Those studies reveal
that linguistic categorization is even more variable than we believed, and that
the conceptual space is fine-grained and probably continuous in many dimen-
sions (Croft & Poole 2008; Croft 2010).
In theory, one cannot extrapolate beyond the point in conceptual space for
which one has elicited an example. In practice, there appears to be substantial
regularity such that robust crosslinguistic generalizations can be formed.
The second type of comparative concept that Haspelmath describes com-
bines both functional concepts and formal concepts. Haspelmath cautions that
formal concepts must be defined in a crosslinguistically valid way, and be
applied consistently across the languages to be compared. To distinguish
these comparative concepts from the purely functional ones, I will call them
HYBRID comparative concepts. This second type of comparative concept is the
more contentious category, in part because the names for such concepts tend to
be traditional grammatical category terms such as adjective, relative clause,
etc., and so they invite confusion with language-specific categories to which
those terms are applied. For this reason, typologists have proposed the conven-
tion of capitalizing the terms when used as language-specific categories and
using lower case when the terms are used as comparative concepts (Lazard 1975;
Comrie 1976: 10; Bybee 1985: 141; Dahl 1985: 34; Croft 2001: 12; Haspelmath
2010: 674; see Section 8).
It is actually quite difficult to define hybrid comparative concepts in a
crosslinguistically valid way; it is much easier to criticize other scholars
attempts to do so. One must not make reference to any formal properties or
categories defined by language-specific constructions. Also, the truth is that

Unauthenticated
Download Date | 11/18/16 7:17 PM
380 William Croft

even the most innocent-looking crosslinguistically valid formal concepts pre-

suppose some functionally defined properties (Croft 2009: 161162). For exam-
ple, the order of elements such as AN vs. NA requires definitions of the
categories being ordered (adjective and noun in this example); such defini-
tions will make reference to some functionally defined properties. Since these
are hybrid comparative concepts, this is not a serious objection.

3 Comparative concepts: Constructions

and strategies
In Croft (2014, in preparation) I describe two types of hybrid comparative
concepts that are useful for crosslinguistic comparison. The first is what I call
a CONSTRUCTION (Croft 2014: 537):

a construction (or any construction) in a language (or any language) used to express a
particular combination of semantic structure and information packaging function.

This definition fits with the construction grammar definition of a construction as

any pairing of form and meaning (Fillmore et al. 1988; Goldberg 1995; Croft
2001). It is hence a hybrid comparative concept, including both constructional
form and its meaning; but the comparative concept is individuated by the
meaning used to define the construction. As noted in Croft (2014), this definition
simply follows the widely held view in typology that crosslinguistic comparison
is based on function; the grammatical form is the dependent variable in typolo-
gical analysis.
The functional side of the definition of a construction refers to both seman-
tics and information packaging. For example, an adjective construction can be
defined in terms of both property concepts (semantics) and modification (infor-
mation packaging). One could define adjectives solely in terms of semantics. But
in many languages, the grammatical behavior of property concept words differs
in modification, predication, and/or reference. Hence from a practical point of
view, the purely semantic definition may not be useful for typological analysis.
The second type of hybrid comparative concept I describe is a STRATEGY (Croft
2014: 537):

a construction in a language (or any language), used to express a particular combination

of semantic structure and information packaging function, that is further distinguished by
certain characteristics of grammatical form that can be defined in a crosslinguistically
consistent fashion.

Unauthenticated
Download Date | 11/18/16 7:17 PM
 Comparative concepts 381

The term strategy also goes back to the early days of typology (Keenan &
Comrie 1977; Givn 1979), and I am simply making explicit, and perhaps more
rigorous, general typological practice here, as with constructions.
Strategies can be defined in different ways. One can use crosslinguistically
valid properties of grammatical structure, such as the order of elements or the
occurrence of overt coding such as a copula in predication. Another type of
strategy is how categories are defined in a language. A well-known example of
this type of strategy are alignment strategies: an ergative language categorizes S
and P together against A, while an accusative language categorizes S and A
together against P. A third type of strategy is to define a construction in terms of
the form also used for another construction. For example, the locative strategy
for predicative possession (Stassen 2009) defines the construction used for
predicative possession in a language as a formal structure which is identical
in the relevant respects to the construction used for the predication of location in
that language.

4 Comparative concepts in typological practice

Comparative concepts are a means to an end, the end being the discovery of
crosslinguistic generalizations, that is, language universals. Purely func-
tional comparative concepts and hybrid comparative concepts have been
used in practice in modern (Greenbergian) typology from its inception. The
same is true of constructions and strategies, though the formal properties
used for defining strategies have not always been consistently defined in a
crosslinguistic fashion. There are probably other types of comparative con-
cepts that are useful for typological comparison, from which typological
generalizations can be made. In other words, the comparative concepts
described in Sections 2 and 3 are more or less congruent with typological
practice.
Where is the disagreement, then? I believe that it is in large part in the terms
used. A corollary of the definition of a construction as a comparative concept is
that constructions are found in all languages. Since human languages are
general-purpose communication systems, they express all functional concepts,
at least the basic concepts that tend to figure in most typological research.
(There are culture-specific concepts that may not have translation equivalents,
though if such concepts became salient enough, a language would borrow or
otherwise form a construction to express them. One should not assume a
language is a fixed object.)

Unauthenticated
Download Date | 11/18/16 7:17 PM
382 William Croft

If a grammatical term is used for a construction as defined in Section 3,

then it will by definition be universal. For example, if the term adjective
is defined as the construction expressing a property concept used in mod-
ification, then adjectives are universal by definition. But many linguists,
including many typologists, state that adjective, or some other category,
is not universal. They cannot mean Adjective as a language-specific cate-
gory. The category of Adjective in Language X is not universal, also by
definition: as a language-specific category, it is not found in any other
language.
Instead, what is meant is that a comparative concept like adjective is
not assumed to be a construction as defined in Section 3. Instead, it is
assumed to be a strategy, namely the strategy used for what I would call
the adjective construction in English or some western European language: a
strategy using a copula (overt morpheme) in predication, or a zero-coded
strategy for modification distinct from the relative clause construction in that
language. In that sense, adjectives are indeed not universal: strategies vary
from language to language, and most languages use a subset of the known
strategies for a construction, or just one strategy. In Croft (2014), a review
article on Malchukov et al. (eds.) (2010), I noted that some authors who
argued that a language lacked ditransitives actually meant the language
lacked a particular strategy (the double-object strategy) for what I would
call the ditransitive construction, namely, any construction expressing the
predication of a transfer event.
So the disagreement appears to be basically about terminology: some
authors use a traditional grammatical term for a construction, while others
use it for a particular strategy found for that construction. One can usually
figure that out by reading the supporting arguments. But I would argue that
it is better to use the basic traditional grammatical terms for a construction,
not a strategy. The western European grammatical tradition often did not
make a distinction between construction and strategy, because the languages
were typologically similar and used the same or similar strategies for the
same constructions. Typology looks across all languages, of course, and finds
other strategies used for various constructions. But the western European
grammatical terms are so deeply ingrained in usage that they are better
used for the universal comparative concept, not a strategy that is restricted
to a subset of languages. The common use of notional, that is, semantic,
definitions for traditional western European grammatical terms also suggests
that there is some continuity in using traditional grammatical terms for
constructions rather than just for the strategies found in western European
languages.

Unauthenticated
Download Date | 11/18/16 7:17 PM
 Comparative concepts 383

5 Language-specific categories vs. comparative

concepts
Language-specific categories are very different from comparative concepts.
Language-specific categories are classes of words, morphemes, or larger
grammatical units that are defined distributionally, that is, by their occur-
rence in roles in constructions of the language. A construction is defined here
broadly as in construction grammar: a morphosyntactic unit of any size
paired with its meaning, so, for example, English Plural is a morphological
construction that can be used to define a language-specific category such as
English Count Noun.
It may appear that the term construction is being used differently here
than in Section 3. It is not, because the definition of language-specific
categories refers to constructions OF THE LANGUAGE. Constructions (in the com-
parative concept sense) of the particular language are language-specific
entities, which typologists have suggested should be referred to with capita-
lized terms, such as Plural (see Section 2). The term should also ideally be
compounded with the language name, i.e., English Plural. But in many
contexts, such as a reference grammar of a particular language, that part of
the label for the language-specific construction, or category, can be safely
dropped without confusion.
Language-specific categories are language-specific because they are defined
by their distribution in constructions, and the constructions that define them are
also language-specific. Lango Adjectives are not the same as English Adjectives
because Lango Adjectives are defined by Lango constructions. English does not
have Lango Adjectives, and Lango does not have English Adjectives. Here, using
the language name in the term, as well as capitalization, makes it clearer that
these categories are not comparative concepts.
Comparative concepts are different from language-specific categories in
another way. In Section 2, I argued that large conceptual categories are in
fact not good comparative concepts, and typologists use narrower concep-
tual categories or even individual tokens (as in elicitation from a stimulus
like a cutting/break video clip). Language-specific categories are often
large, especially if they are defined by occurrence in a role in just one
construction, and are defined as all elements that occur in that construc-
tional role. Comparative concepts cannot be all elements occurring in a
constructional role, because they are not defined distributionally.
Distributional patterns are language-specific, because constructions are lan-
guage-specific.

Unauthenticated
Download Date | 11/18/16 7:17 PM
384 William Croft

6 The relationship between language-specific

categories and typology and universals
research
The comparison of language-specific categories and comparative concepts
in Section 5 makes it look like language-specific categories have nothing to
do with comparative concepts. A linguist writing a grammatical description
using language-specific categories appears to be doing something comple-
tely unrelated to a typologist using comparative concepts to discover lan-
guage universals. This would be an unfortunate state of affairs, not unlike
the disconnect between generative syntactic theory and almost all current
descriptive grammatical practice. There are reasons to believe, however,
that this is not the case. Just as descriptive grammatical research is essen-
tial to typology and universals research, the results of typology and uni-
versals research is valuable if not essential to descriptive grammatical work.
The empirical observations that led to the need to identify comparative
concepts that are valid across languages also apply to distributional analysis
within languages. Not only do distributionally-defined categories differ across
languages, if one compares distributional patterns of functionally equivalent
constructions. Distributionally-defined categories differ across the constructions
of a single language that are used to define language-specific categories. In
other words, grammatical categories are not just language-specific, but con-
struction-specific (Croft 2001).
This point, that the constructions used to define grammatical categories do not
match in their distribution, is well known. However, it clashes with the assumption
that grammatical categories, particularly word classes, exist independently of the
constructions that define them. The assumption of the independence of gramma-
tical categories, and the assumption that they are the building blocks of gram-
matical structure, is a legacy of structuralist and generative syntactic analysis (if not
also traditional grammar before them). Ironically, careful and complete distribu-
tional analysis is also a legacy of at least American structuralism (Croft 2001: 36, 41).
However, these two theoretical constructs are incompatible (Croft 2001: 4547).
Distributional analysis, broadened to include meaning as well as form, is the basic
empirical method of grammatical description and analysis. Grammatical categories
are not independent of the constructions that define them, because they are defined
by those very constructions.
This circularity is obscured by the fact that the term construction is
avoided in describing distributional analysis. The constructions used to define

Unauthenticated
Download Date | 11/18/16 7:17 PM
 Comparative concepts 385

word classes are called many different things: criteria (Givn 2001: 49; Dixon
2010: 38), tests (McCawley 1998; Carnie 2013: 47, 98100), evidence, phe-
nomena, operation, and process (Mulder 1994: 114), instead of simply
being called constructions defining a word class. The words in a word class
are said to have a particular grammatical or syntactic distribution (Harris 1951:
5; Carnie 2013: 47), behavior (McCawley 1998: 186), properties (McCawley
1998: 18; Evans & Osada 2005: 452; Schachter & Shopen 2007: 2), features
(Amha 2001: 89), use (Jagersma 2010: 268), or function (Palmer 2009: 94),
instead of simply saying that they occur in certain constructions and not in
others.
Applying the same reasoning for comparative concepts and language-spe-
cific categories to those language-specific categories leads to the conclusion that
language-specific categories are in fact construction-specific. Language-specific
categories overlap in many complex ways, thanks to the different constructions
that define them, and in some sense are even epiphenomenal. Just as recogniz-
ing that categories are language-specific frees us to describe languages as they
are without imposing universal categories, recognizing that categories are
construction-specific frees us to describe grammatical categories of languages
as they are for each individual word, morpheme, or larger unit, what
constructions it occurs in and what ones it doesnt; and why, to the extent
that we can motivate their distribution in functional terms.
Comparative concepts are not distributionally defined, while language- or
construction-specific concepts are so defined. But language- or construction-
specific categories play a major role in typology and universals research. The
point of comparative concepts is not to impose yet another set of grammatical
categories that partitions the relevant space of possibilities. It is to provide a
basis for comparing languages that differ in the categories that they have, in
terms of the functions they perform, or the strategies they employ in performing
those functions. Once that is done, then much typology and universals research
is devoted to placing those categories in a crosslinguistic comparative context.
Language universals in typological theory are constraints, often probabilis-
tic, on crosslinguistic variation. Many if not most or even all such universals are
constraints on crosslinguistic variation in language-specific categories. The
clearest example of this is the semantic map model (Croft 2001, 2003;
Haspelmath 2003; and references cited therein; for a computational model, see
Regier et al. 2013) and the use of multidimensional scaling for semantic map
modeling (Croft & Poole 2008; Croft 2010; Rogers 2015; Garca Macas 2016). The
data used to construct semantic maps are language-specific (or more precisely,
construction-specific) categories. The extensions of the categories are calibrated
via comparative concepts. A semantic map analysis uses the variation of

Unauthenticated
Download Date | 11/18/16 7:17 PM
386 William Croft

categories across constructions and across languages to infer the structure of the
conceptual space of comparative concepts, and to identify constraints on possi-
ble (or probable) language-specific and construction-specific categories on that
conceptual space.
Language-specific categories and comparative concepts are very different
entities that serve different analytical purposes. But typological research uses
language-specific and construction-specific categories as well as comparative
concepts in order to discover language universals.

7 Language- and construction-specific categories

in descriptive practice: What typology
and universals research can contribute
to language description
I believe that the results of typology and universals research can be fruitfully
married to descriptive practice in several ways, as described in this and the
following section.
A grammatical description should provide as detailed a characterization of
distributional patterns of the language as time and other resources allow. The
distributional patterns are a many-to-many mapping between complex gramma-
tical units multimorphemic words, phrases, clauses, multiclausal sentences
and the grammatical units that fill the roles in those complex grammatical units:
morphemes, words, phrases, clauses, etc. Many of the best current grammatical
descriptions do this to some degree. But the distributional patterns are often
obscured by the widespread organization of reference grammars by word class
(often defined by morphological constructions, particularly inflectional con-
structions), with relatively few (if any) or relatively brief chapters explicitly
organized by phrasal, clausal, and sentential constructions.
Of course, there are many constructions in a particular language.
Typological research can provide an overview of the constructions in the
comparative sense, that is, forms expressing a particular function that allows
a grammar to cover the broad scope of a language, even if a grammatical
description can only do so in a skeletal fashion due to resource limitations. In
particular, typological research can identify the constructions which are com-
mon in discourse, and the functional distinctions that frequently are reflected in
major formal constructional distinctions. This allows a descriptive linguist to

Unauthenticated
Download Date | 11/18/16 7:17 PM
 Comparative concepts 387

capture much of the constructional diversity of the language. Typological

research also identifies the range of major constructional strategies found in
the worlds languages. Recognizing those strategies allows a descriptive linguist
to situate the language being described in its typological context, including
what is unusual or unique about the strategy or strategies employed for that
function in the language. An overview of constructions and strategies will be
found in Morphosyntax: Constructions of the worlds languages (Croft, in
preparation).

8 Labeling language-specific, and construction-

specific, categories
A final issue that frequently arises, including in the Lingtyp discussion, is how
to name language-specific and construction-specific categories and construc-
tions in language description. Superficially, one might think that since distribu-
tionally defined categories in a language are language-specific and construction-
specific, then unique names should be used for such categories in order to
minimize confusion. In practice, however, having unique names for every lan-
guage-specific and construction-specific category, as well as unique names for
the constructions that define them, would make a grammar completely unusa-
ble. The current practice of using common grammatical terms is largely con-
venient for use, which seems surprising given the issues that led to the
identification of comparative concepts in typological theory.
Why is this so? The main difficulties are not descriptive but theoretical. The
problems tend to arise when language-specific categories are identified with
theoretical, putatively universal, categories, or vice versa. When this happens,
then the lengthy but ultimately pointless discussions of whether some language-
specific category is an instance of theoretical universal category X or not, or
whether some language has theoretical universal category X at all, arise (Croft
2001: 3032; Haspelmath 2012). One can easily avoid such debates by avoiding
making this false identification, and simply read the grammatical description as
a description of language-specific categories and constructions, their functions,
and their distribution.
The reason that language descriptions using terms like verb, subject,
and relative clause are mostly easy to read for comparative and theoretical
purposes is that most descriptive linguists implicitly follow some common rules
of thumb for labeling language-specific and construction-specific categories.

Unauthenticated
Download Date | 11/18/16 7:17 PM
388 William Croft

There are some common exceptions, for example, the use (or avoidance) of the
term adjective, and there are some less common cases in which the descriptive
linguist has made an odd choice from a comparative or typological perspective.
Part of the reason for this is that the rules of thumb are not explicit, and
sometimes the descriptive linguist is not familiar with either typological or
comparative research that has led to commonly accepted terms for the construc-
tions and strategies being described. Here I will briefly summarize rules of
thumb, some more widely followed than others, which would mutually benefit
descriptive linguists and typologists.

(1) Rule of Thumb 1:

Use semantic class terms to denote semantic classes/functions, and gram-
matical terms to denote word classes/constructions, if distinct terms are in
general use.

For example, use property word and not adjective if talking about the
semantic class of property concept words. I have chosen an example here
where this rule of thumb is not commonly followed, but in other cases, e.g.,
noun and object word (or more commonly, semantic subclasses such as
humans, animals, and artifacts), this rule of thumb is usually applied.
Nevertheless there are numerous cases where a careful reading of a grammar
indicates that a traditional grammatical category term is being used for a
semantic category.

(2) Rule of Thumb 2:

If the same term is used for both a semantic class/function and a (lan-
guage-specific) word class/construction, capitalize the term when used for
the word class/construction and leave the term in lower case for the
semantic class/function.

For example, use numeral for the semantic class and (English) Numeral for
the English word class that includes this semantic class. This rule of thumb is
not widely followed. Here the risk of misinterpretation is greatest in using the
same term for a language-specific construction such as (English) Relative Clause
and the hybrid comparative concept of the relative clause construction. This is
because most proposed terms for hybrid comparative concepts are grammatical
terms from the Western grammatical tradition, for good reasons (see Section 4).
But again, in a grammatical description, one can usually assume that the
grammatical term is being used for the language-specific category or construc-
tion, so the absence of capitalization is often not a problem.

Unauthenticated
Download Date | 11/18/16 7:17 PM
 Comparative concepts 389

(3) Rule of Thumb 3:

Use as the label for a word class/construction the prototype function of
the construction used to define the member(s) of the word class/construc-
tion category, even if the words in question occur in other constructions,
or the construction has other functions.

To put it in comparative concept terms: label the language-specific word class or

construction with the name used for the construction or role in the construction
qua comparative concept, albeit capitalized following Rule of Thumb 2. For
example, name the word class (English) Adjective if it includes the Dixon core
property concept words in the modification construction. Recall that a construc-
tion is defined as the form used for a particular meaning or function, the latter
usually quite narrowly defined. Typological research has identified or distin-
guished meanings/functions that are useful in analysis of crosslinguistic varia-
tion, because of the universals that they enter into. Hence it is likely that the
construction in a language that expresses that function will be relevant to
grammatical distinctions or distributional patterns found in the language, and
so it will be useful to label that construction with a capitalized version of the
comparative concept. In practice most descriptive linguists follow this rule of
thumb, though careful reading of a grammar sometimes reveals unexpected
labeling of constructions or categories.
There is a problem with Rule of Thumb 3. In some languages, there is no
distinct form or word class for the function that the label implies. For example,
Old English lacks a special Reflexive Pronoun form, and uses the same form for
noncoreferential objects as well as coreferential objects (Comrie 2003: 203):

(4) hi slh hinei/j

Hei hit himselfi/himj.

The term typically used for noncoreferential object pronoun forms is Anaphoric
Pronoun. Old English hine is used for both coreferential and noncoreferential
object pronouns. Should hine be labeled an Anaphoric Pronoun or a Reflexive
Pronoun? The usual labeling strategy is described in the following rule of
thumb:

(5) Rule of Thumb 4:

There is an implicit implicational hierarchy of labeling, so that when the
same form occurs in two distinct constructions which are used to define a
category, one construction trumps the other in labeling the category.

Unauthenticated
Download Date | 11/18/16 7:17 PM
390 William Croft

In the case of object pronouns, use in the subject-object noncoreference con-

struction trumps use in the subject-object coreference construction: we call hine
an Object Anaphoric Pronoun, not a Reflexive Pronoun.
Unfortunately, these generally used implicational hierarchies of labeling are
not codified anywhere. The labeling choices implicitly refer to universal typolo-
gical patterns: e.g., when a language has a distinct Reflexive Pronoun form in
the subject-object coreference construction, it represents an extension of a form
from certain other domains such as intensification, or it has greater structural
coding (herself vs. her) which often represents a lower degree of grammaticaliza-
tion, while an Anaphoric Pronoun form has other sources (demonstrative pro-
noun, etc.).
Yet another problem in word class labeling is found when multiple con-
structions are used for a word class, but each construction defines different
(though overlapping) sets of words. This problem is a manifestation of the fact
that grammatical categories are construction-specific, not just language-specific
(see Section 6). For example, the English Adjective class is often defined by
occurrence in several constructions, including the Attributive (modifying)
Adjective construction, the Predicate Adjective construction, the Degree
Inflection construction, and the Degree Modifier construction. However,
English alive does not occur as a modifier, English intelligent does not inflect
for degree, English even does not take degree modifiers, and English entire
cannot be predicated.
Most linguistic descriptions avoid the problem of labeling such distinct but
overlapping word classes; at best they list exceptions, set up subclasses, or treat
the category as a prototype without justifying the prototype analysis. In a fully
construction-based approach to grammatical description, one can use the fol-
lowing rule of thumb to name these overlapping classes (Croft 2001: 50):

(6) Rule of Thumb 5:

Name the word class using the name of the construction that defines the
word class.

In the case of English Adjectives, we should therefore define four word classes
corresponding to the four constructions that define them: Attributive Adjectives,
Predicate Adjectives, Inflectional Adjectives, and Gradable Adjectives. Although
the word class labels suggested here all include the word Adjective, these
should not be thought of as subclasses of a word class Adjective. Instead, they
are a set of largely overlapping word classes that have in common a substantial
proportion of property concept words. Again, we would not want to obscure this
fact by coining unique, novel labels for each of these overlapping classes of

Unauthenticated
Download Date | 11/18/16 7:17 PM
 Comparative concepts 391

property concept words. Since Adjective is the term often used for the word
class that includes mostly property concept words, it is most convenient to use
the term Adjective compounded with a modifying label as I have done here.
It must not be forgotten that the labels for word classes are simply con-
venient ways to talk about patterns of occurrence of words with particular
meanings in constructions with particular information packaging functions.
Any label can be chosen by a descriptive grammarian. But some labels are
more convenient than others, in particular labels that clearly distinguish gram-
matical form from word meaning and constructional function (most easily by
capitalizing the former and not the latter), and labels that evoke the prototypical
meaning of members of the word class and the prototypical function of the
construction, as well as likely sources and extensions of the construction. The
labels may be evocative, but what they evoke are crosslinguistic patterns, and
the universals that constrain them, that have been confirmed or discovered by
typological research.

References
Amha, Azeb. 2001. The Maale language. Leiden: CNWS Publications.
Bybee, Joan L. 1985. Morphology: A study into the relation between meaning and form.
Amsterdam: Benjamins.
Carnie, Andrew. 2013. Syntax: A generative introduction. 3rd edn. Chichester: Wiley-Blackwell.
Chafe, Wallace. 1976. Givenness, contrastiveness, definiteness, subjects, topics and points of
view. In Charles Li (ed.), Subject and topic, 2556. New York: Academic Press.
Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press.
Comrie, Bernard. 2003. On explaining language universals. In Michael Tomasello (ed.), The new
psychology of language: Cognitive and functional approaches to language structure,
Vol. 2, 195209. Mahwah, NJ: Erlbaum.
Croft, William. 2001. Radical Construction Grammar: Syntactic theory in typological perspective.
Oxford: Oxford University Press.
Croft, William. 2003. Typology and universals. 2nd edn. Cambridge: Cambridge University
Press.
Croft, William. 2009. Methods for finding language universals in syntax. In Sergio Scalise,
Elisabetta Magni & Antonietta Bisetto (eds.), Universals of language today, 145164.
Berlin: Springer.
Croft, William. 2010. Relativity, linguistic variation and language universals. CogniTextes 4.303
http://cognitextes.revues.org/303/
Croft, William. 2014. Comparing categories and constructions crosslinguistically (again): The
diversity of ditransitives (review article on Malchukov et al. (eds.) 2010). Linguistic
Typology 18. 533551.
Croft, William (in preparation). Morphosyntax: Constructions of the worlds languages.
Cambridge: Cambridge University Press.

Unauthenticated
Download Date | 11/18/16 7:17 PM
392 William Croft

Croft, William & Keith T. Poole. 2008. Inferring universals from grammatical variation: multi-
dimensional scaling for typological analysis. Theoretical Linguistics 34. 137.
Dahl, sten. 1985. Tense and aspect systems. Oxford: Blackwell.
Dixon, R. M. W. 1977. Where have all the adjectives gone? Studies in Language 1. 1980.
Dixon, R. M. W. 2010. Basic linguistic theory, Vol. 2: Grammatical topics. Oxford: Oxford
University Press.
Evans, Nicholas & Toshiki Osada. 2005. Mundari and argumentation in word-class analysis.
Linguistic Typology 9. 442457.
Fillmore, Charles J., Paul Kay & Mary Catherine OConnor. 1988. Regularity and idiomaticity in
grammatical constructions: The case of let alone. Language 64). 501538.
Garca Macas, Jos Hugo. 2016. From the unexpected to the unbelievable: Thetics, miratives
and exclamatives in conceptual space. Albuquerque, NM: University of New Mexico doc-
toral dissertation.
Givn, Talmy. 1979. On understanding grammar. New York: Academic Press.
Givn, Talmy. 2001. Syntax, Volume 1. Amsterdam: Benjamins.
Goldberg, Adele E. 1995. Constructions: A construction grammar approach to argument
structure. Chicago: University of Chicago Press.
Harris, Zellig S. 1951. Methods in structural linguistics. Chicago: University of Chicago Press.
Haspelmath, Martin. 2003. The geometry of grammatical meaning: Semantic maps and cross-
linguistic comparison. In Michael Tomasello (ed.), The new psychology of language:
Cognitive and functional approaches to language structure, Vol. 2, 211242. Mahwah,
NJ: Erlbaum.
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in crosslinguistic
studies. Language 86. 663687.
Haspelmath, Martin. 2012. How to compare major word-classes across the worlds languages.
In Thomas Graf, Denis Paperno, Anna Szabolcsi & Jos Tellings (eds.), Theories of every-
thing: In honor of Edward Keenan (UCLA Working Papers in Linguistics 17), 109130. Los
Angeles: UCLA. http://www.linguistics.ucla.edu/faciliti/wpl/issues/wpl17/wpl17.html
Jagersma, Abraham Hendrik. 2010. A descriptive grammar of Sumerian. Leiden: Universiteit
Leiden doctoral dissertation.
Keenan, Edward L. & Bernard Comrie. 1977. Noun phrase accessibility and universal grammar.
Linguistic Inquiry 8. 6399.
Lazard, Gilbert. 1975. La catgorie de lventuel. In Mlanges linguistiques offers mile
Benveniste, 347358. Leuven: Peeters.
Levinson, Stephen C., Srgio Meira, and the Language and Cognition Group. 2003. Natural
concepts in the spatial topological domain adpositional meanings in crosslinguistic
perspective: An exercise in semantic typology. Language 79. 485516.
Majid, Asifa, James S. Boster, Melissa Bowerman. 2008. The cross-linguistic categorization of
everyday events: A study of cutting and breaking. Cognition 109. 235250.
Malchukov, Andrej, Martin Haspelmath & Bernard Comrie (eds.). 2010. Studies in ditransitive
constructions: A comparative handbook. Berlin: De Gruyter Mouton.
McCawley, James D. 1998. The syntactic phenomena of English. 2nd edn. Chicago: University of
Chicago Press.
Mulder, Jean Gail. 1994. Ergativity in Coast Tsimshian (Smalgyax). Berkeley, CA: University of
California Press.
Palmer, Bill. 2009. Kokota grammar (Oceanic Linguistics Special Publications 35). Honolulu:
University of Hawaii Press.

Unauthenticated
Download Date | 11/18/16 7:17 PM
 Comparative concepts 393

Regier, Terry, Naveen Khetarpal & Asifa Majid. 2013. Inferring semantic maps. Linguistic
Typology 17. 89105.
Rogers, Phillip. 2015. Illustrating the prototype structures of parts of speech: A multidimen-
sional scaling analysis. Albuquerque, NM: University of New Mexico MA thesis.
Schachter, Paul & Timothy Shopen. 2007. Parts of speech systems. In Timothy Shopen (ed.),
Language typology and syntactic description (2nd edn.), Vol. 1: Clause structure, 160.
Cambridge: Cambridge University Press.
Stassen, Leon. 2009. Predicative possession. Oxford: Oxford University Press.

Unauthenticated
Download Date | 11/18/16 7:17 PM