LINGUIST List 7.826

Tue Jun 4 1996

Sum: Phoneme inventory size and word length

Editor for this issue: Ljuba Veselinova <lveselinemunix.emich.edu>


Directory

  1. Martin Haspelmath, Sum: phoneme inventory size and word length

Message 1: Sum: phoneme inventory size and word length

Date: Tue, 04 Jun 1996 09:06:49 +0200
From: Martin Haspelmath <martinhazedat.fu-berlin.de>
Subject: Sum: phoneme inventory size and word length
In LINGUIST 7-739 I asked whether anyone had tested the often heard claim 
that large phoneme inventory sizes correlate with short words, and vice 
versa. Thanks to those who responded!
 I was pointed to the following recent article, where the claim is 
tested and confirmed for a sample of ten languages:

NETTLE, David (1995), "Segmental inventory size, word length, and 
communicative efficiency", Linguistics 33.2: 359-367.

Nettle explains the correlation in terms of competing motivations: 
Different languages optimize either speed of performance or simplicity of 
structure. These functional constraints are inevitably in conflict with 
each other, and different languages resolve the conflict differently by 
assigning different weights to the constraints.
 Nettle cites recent work in European "quantitative linguistics" for 
related approaches. True, such claims can only be tested by using 
quantitative methods, but are they irrelevant to phonological theory, and 
linguistic theory more generally?
 Chilin Shih objects that Mandarin Chinese has a small phoneme inventory 
and short words. In Nettle's study, Chinese is not exceptional because he 
counts compounds as single words and includes the distinctive force of 
tone in calculating the size of the phonemic inventory.
 Markus Hiller points out that it is not quite clear what should be 
counted in each case -- do we include all Japanese words of Chinese 
origin that are hardly used in the spoken language because of homonymy? 
How do we exclude them? He suggests that only the core vocaabulary should 
be taken into account. In addition, he proposes that not segments, but 
gestures (in the sense of Browman & Goldstein 1986) should be counted, 
because they are more easily classified as redundant vs. distinctive, and 
they are less abstract than features. Nettle did not use such 
sophisticated methods in his study, and the fact that he got 
statistically significant results shows that the effect is so strong that 
it is not obscured by all the "noise".
 Paul de Lacy notes that we need to distinguish between roots and words 
(although the prediction of the functional theory probably applies to 
both equally) and notes that Polynesian languages do not have 
particularly long words. Maori, for instance, has only 14 segmental 
phonemes, 80% of its roots are bimoraic, and it hasn't resorted to 
increasing word length yet. In Nettle's study, though, Hawaiian does have 
a fairly high average word length (7.08 segments, the thgird highest in 
his sample).
 I am grateful to

And Rosta (ucleaarucl.ac.uk)
Wolfgang Behr (w.behrem.uni-frankfurt.d400.de) 
Markus Hiller (markus.hillerzdv.uni-tuebingen.de)
Chilin Shih (cayenne!clsresearch.att.com)
Paul de Lacy (delacyvoyager.co.nz)


Martin Haspelmath
Department of English
Free University of Berlin
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue