Volume 8 Issue 1 (2010)
Note: Linguistic Discovery uses Unicode characters
to represent phonetic symbols. Please see Optimizing Display
for requirements to accurately reproduce this page.
Polysemous Qualities, Continuity, and Graduation
Author’s reply to ‘Drawing Networks from Recurrent
Polysemies’ by Michael Cysouw (2010)
Cysouw’s (2010) remarks concern three issues: the
presentation of the concepts studied, the arbitrariness of the statistical
approach, and the visualization of the networks. I will briefly comment upon
each of those in turn.
As for the presentation of the cross-linguistic concepts, Cysouw is
right: exemplifying by a sentence or above all by a small paragraph would have
been very useful. The pilot study, conducted for African languages, was aimed at
naming the concepts and assumed that these prototypical qualities were easily
understandable; it made sense as a first step. But the approach Cysouw suggests
is indeed preferable. In some instances, it would be even better to use a short
paragraph rather that a single sentence, as some concepts, e.g. ‘dried
fruit’ or ‘dry weather’, may be not cross-culturally
Concerning the issue of statistical assessment of the frequencies
attested, I will try to clarify my point of view. Cysouw points out that
“arbitrary divisions of a scale [are] not very suitable for interpretative
practice”. I fully agree with him: some qualities are very federative and
some others are less federative, just as some polysemous patterns recur
frequently while others are less recurrent. But it is clear that the algorithm
used by Cysouw avoids the problem of quantifying a threshold because his
representations give an account of the attestation frequencies (by means of the
length of the line), whereas my
qualitative notional maps forced me to
quantitative viewpoint. For instance, I defined the federative
notions as qualities involved in a minimum of five polysemous patterns and
across a minimum of six languages. The reason behind specifying a threshold was
that the particular functioning of these polysemous qualities becomes
statistically more significant. Nevertheless, although a statistical approach
enables me to distinguish regularities, it does not commit me to final
inferences. As I tried to explain in the conclusion, a strictly quantitative
conception of typology is not very productive because it goes beyond binary
definitions and all-or-nothing oppositions.
We need to maintain a constructivist approach of language, which implies
“a rejection of substantive universals, a definition of dynamic invariant
features […], and a dynamic construction of salient patterns” (Fuchs
1999:17-18). And, once again, I will refer to Lazard (1992:432) who states that
“semantic substance is continuous and unstable. Differences within it are
gradual.” So, federative notions as well as recurring phenomena must be
considered as continuous and gradual facts.
As for the networks drawn by Cysouw, there is no doubt that his model
significantly improves the analysis of the polysemous patterns (whether
recurring or not) because, as I just mentioned, his representations capture the
principles of continuity/graduation better than mere reliance on attestation
frequencies. As a consequence, we can clearly observe that Cysouw’s
“clusters” are organized around conceptual cores of federative
fat/big/thick, etc. And I think
that, with such a method of representation, the expression “semantic
field” as employed by Lazard (1992) really becomes meaningful.
Cysouw, Michael. 2010. Drawing networks from recurrent polysemies.
Comment on Perrin 2010. Linguistic Discovery, this issue. doi:10.1349/ps1.1537-0852.a.377
Fuchs, Catherine. 1999. Diversity in linguistic representations: A
challenge for cognitive science. Language diversity and cognitive
representations, ed. by Catherine Fuchs and Stéphane Robert, 3-19.
Lazard, Gilbert. 1992. Y a-t-il des catégories
interlangagières? Texte, Sätze, Wörter und Moneme, ed. by S.
Anschütz, 427-434. Heidelberg: Heidelberger Orientverlag.
Perrin, Loïc-Michel. 2010. Polysemous qualities and universal
networks. Linguistic Discovery, this issue. doi:10.1349/ps1.1537-0852.a.353
Author’s contact information:
Langage, Langues et Cultures d'Afrique Noire (LLACAN, CNRS -
7, rue Guy Môquet
94 801 Villejuif