Cross-linguistic categories - what are they?

[Bill Croft]

Esa Itkonen's comment (and paper) does not 
consider another alternative, namely that the 
notion of crosslinguistic formal categories is a 
counterproductive fiction (the only options he 
offers are "psychologically real entity" and 
"useful fiction"). There are a number of 
misinterpretations of my position and that of 
Haspelmath in Itkonen's paper.

I do not subscribe to a "conventionalist" view of 
crosslinguistic formal categories; I argue that 
they do not exist at all, not just in Radical 
Construction Grammar but in a number of follow-on 
papers (Croft 2005, 2007, 2009, 2010). This is 
the same position taken by Martin Haspelmath in a 
passage from an unpublished 2008 paper cited by 
Itkonen: "the adoption of categorial universalism 
has actually impeded, not facilitated, 
crosslinguistic research".

Itkonen presents some quotations from 
Haspelmath's paper and claims they are 
contradictory. In fact they present a coherent 
and consistent position, one which I also 
advocate and is part of the typological method: 
crosslinguistic formal linguistic categories are 
invalid, but crosslinguistic comparison can be 
based on semantic categories which are 
crosslinguistically valid - albeit in terms of 
fine-grained definitions of situation types, not 
broad conceptual categories like "IN 
[containment]" (see Croft 2001, chapter 3; Croft 
to appear a, b; Croft and Poole 2008:31-33).

I do not reject the distributional method (Croft 
2001:45-46; Croft 2010:344-45). It is the only 
valid method of formal linguistic analysis, if 
done carefully and thoroughly - that is, not 
ignoring distributional facts that don't match up 
or don't match expectations. I do reject the 
opportunistic use of selective distributional 
facts to support categories assumed to exist a 
priori.

Finally, and most importantly, abandoning 
crosslinguistic formal categories allows 
typologists to develop valild methods and 
concepts to understand crosslinguistic diversity 
and universals. The chief method is the semantic 
map model, which is basically a multidimensional 
generalization of implicational hierarchies. It 
has been used productively by many typologists 
including Lloyd Anderson, Suzanne Kemmer, Martin 
Haspelmath, Leon Stassen, Johan van der Auwera, 
Andrej Malchukov, Nikolaus Himmelmann & Eva 
Schultze-Berndt, and others to whom I apologize 
for not remembering to name here. 
Multidimensional scaling can be used in order to 
extend the applicability of the semantic map 
model to larger and more complex datasets (Croft 
and Poole 2008, Croft to appear a, b). MDS has 
been used in this way by Stephen Levinson & 
Sergio Meira, Michelle Feist and Steven Clancy; 
Melissa Bowerman & Asifa Majid have used related 
multivariate techniques for the same purposes. 
The semantic map model is an empirical inductive 
method, but it does not presuppose 
crosslinguistic formal categories.

Bill Croft


Croft, William. 2001. Radical Construction 
Gammar: syntactic theory in typological 
perspective. Oxford: Oxford University Press.

------. 2005. Word classes, parts of speech and 
syntactic argumentation [Commentary on Evans and 
Osada, Mundari: the myth of a language without 
word classes]. Linguistic Typology 9.431-41.

------. 2007. Beyond Aristotle and gradience: a 
reply to Aarts. Studies in Language 31.409-30.

------. 2009. Methods for finding language 
universals in syntax. Universals of language 
today, ed. Sergio Scalise, Elisabetta Magni and 
Antonietta Bisetto, 145-64. Berlin: Springer.

------. 2010. Ten unwarranted assumptions in 
syntactic argumentation. Language usage and 
language structure, ed. Kasper Bøye and Elisabeth 
Engberg-Pedersen, 313-50. Berlin: Mouton de 
Gruyter.

------. To appear a. Relativity, linguistic 
variation and language universals. CogniTextes.

------. To appear b. Exemplar semantics. To 
appear in a volume ed. Seana Coulson. Stanford: 
Center for the Study of Language and Information.

------ and Keith T. Poole. 2008. Inferring 
universals from grammatical variation: 
multidimensional scaling for typological 
analysis. Theoretical Linguistics 34.1-37.