The paper by Hengeveld et al. (2012) in Functions of Language on “Semantic categories in the indigenous languages of Brazil” was probably written by Kees Hengeveld (with his characteristically clear style), but there are many “al.”: eleven additional coauthors! So it beats Bickel et al. (2007), which with its nine coauthors seems to be the record for papers in Language so far. In the natural sciences, of course, papers with many authors are now commonplace (sometimes even with hundreds of authors), so linguistics seems to be catching up with a wider trend. With so many authors, one may wonder what the contributions of the individual authors were, and the present paper doesn’t say, but from various indications it is obvious that Hengeveld was responsible the theoretical background, while the eleven coauthors (mostly from the Universidade Federal do Rio de Janeiro) provided the data on the sample of 24 Brazilian languages from 23 different families. This may well be a model for much future comparative work, with a comparative linguist being responsible for the overall project, and full coauthor credit given to the experts of the individual languages. (What is missing in this paper, however, is further explanation on where the data come from – did all these authors work directly with speakers of the languages? Or did they get the data from reference materials? The paper doesn’t say.)
This paper reports on different ways of expressing the ontological categories of Individual (x), Location (l), Manner (m), Time (t), Reason (r), and Quantity (q), especially in interrogatives and demonstratives, but also in nominalizations. English, for instance, has basic interrogatives for all categories (what, where, when, how, why) except quantity (where it uses the non-basic how much). It turns out that the categories can be arranged on a hierarchy such that languages are more likely to have basic interrogatives for the categories higher on the hierarchy:
(1) Individual > Location > Manner > Time, Reason, Quantity
Thus, English is in line with this in that how much (Quantity) is low on the hierarchy, and Spanish is in line with it in that por qué (‘why’, Reason) is low on the hierarchy.
Hengeveld et al. call the entities in (1) “semantic categories”, which is the standard Functional Discourse Grammar (FDG) term for them; I prefer the term “ontological categories” (from Jackendoff 1983: 51), because there are many different kinds of categories that one might want to distinguish semantically. FDG (Hengeveld & Mackenzie 2008) is probably the only grammatical theory where these categories play a major role, so it is not surprising that not only Hengeveld, but also another FDG proponent, Lachlan Mackenzie, has worked on them from a typological perspective (Mackenzie 2009). Mackenzie focuses on interrogative pronouns, but asks not only which categories are expressed by basic pronouns, but also what the segmental length of the interrogative pronouns is. That length should go along with non-basicness is of course to be expected (so por qué is longer than why). The diversity within Mackenzie’s world-wide sample is greater than in Hengeveld’s Brazilian sample, but overall the sequence in (1) seems to be obeyed, at least as a strong trend (Mackenzie says in the last paragraph that “semantic categories cannot be ordered in a single cross-linguistically valid implicational hierarchy”, but given the clear statistical skewing, this is too negative a conclusion).
Hengeveld and his coauthors go beyond interrogatives and also study demonstratives (‘this’, ‘there’, ‘now’, ‘like this’, ‘therefore’) as well as nominalizations (agent/instrument nouns, location nouns, time nouns), and find broadly similar patterns, always with individuals and locations at the top.
I found this work very interesting, though the authors could have referred to Heine et al. (1991: 55-59), who also discuss interrogative pronouns cross-linguistically and whose hierarchy “person/thing > space > time” is fairly similar (though based on quite different considerations). But I wondered about the explanation of the hierarchy. Hengeveld et al. are content with concluding that the hierarchy provides evidence for FDG semantic categories, but Mackenzie (2009) advances a functional explanation in terms of cognitive complexity (reminiscent of Heine et al.): “there is an iconic relation between the hierarchy of cognitive complexity and the signal complexity [i.e. segment length] of items” (2009: 1135). I have argued (Haspelmath 2008) that such iconic explanations are better replaced by explanations in terms of frequency of use, and indeed we find that the items higher on the hierarchy tend to be more frequently used (e.g. the following frequencies in the BNC of English: what 2401 per million words, where 1059, when 2076, how 981, why 486; but note that this includes relative and conjunction uses).