Mieko Ogura & William S-Y. Wang

     Project on Linguistic Analysis, University of  California at Berkeley

     ogura-m@tsurumi-u.ac.jp

     "Ambiguity and Language Evolution"

     The existence and abundance of ambiguity in languages has intrigued
linguists for a long time.  If we view language as a coding system to
encode meanings with signals, it would seem that language is not
optimal, because in an ideal code one signal should correspond to
exactly one meaning.  We assume that the number of meanings that humans
can manipulate is infinite and there may exist a cognitive constraint on
the number of forms that can be memorized.  To meet semantic need,
polysemy (forms with many [related] meanings) and homophony (two or more
[unrelated] meanings getting the same form), which are major sources of
ambiguity, are inevitable.  In this study, we would like to investigate
the evolution of homophony and its relation to the evolution of syllable
number of words, based on the quantitative analysis on the historical
data and simulation. 

     Anttila (1989) states that all languages have homophony to
different degrees, and one can never predict with complete confidence
when a community or speaker will find it inconvenient enough to be
corrected.  Even when avoidance or correction of homophony does occur,
there is no way of telling by what mechanism of change it will happen. 
Furthermore, there has been little study on the evolution of syllable
number of words except the observation that word length is inversely
related to the size of the phonological inventory (Nettle 1999). 

     Based on the CELEX lexical database of English(1991) and the LDC
Japanese Lexicon (1997), we find that 23 % of 52447 types and 17 % of
51274 types are homophones in English and Japanese respectively.  Wang
et al.  (2003), based on the simulation designed within the naming game
framework, shows that when the number of meanings and the number of
utterances are equal, the agents converge to the same vocabulary, and
their communications are successful 90% of the time, 20% of the words
having homophones.  When we compare the quantitative studies with the
simulation results, we may assume that to avoid homophones, humans try
to manifest "one meaning, one form", but homophones do occur and the
threshold is around 20% of the vocabulary. 

     Based on the CELEX database, the Brooklyn-Geneva-Amsterdam-Helsinki
Parsed Corpus of Old English, and the Penn-Helsinki Parsed Corpus of
Middle English, we find that homophones are stable and cumulate with the
times.  To avoid creating homophones, syllable number of words
increases, with more recent entry dates of words associated with more
syllables.  Furthermore, we will explain why the syllable number of
nouns is larger than that of verbs and the closed class words remain
short in Present-day English, and explore stability of homophones and
instability of synonyms in the acquisition process.  The mechanism of
the evolution of homophones and syllable length of words based on
English data works cross-linguistically from the emergence of language
and goes on at present, forming complex adaptive system (Ogura & Wang
2004). 

References:

Antilla, R.  1989.  Historical and Comparative Linguistics.  Amsterdam:
John Benjamins. 

Nettle, D.  1999.  Linguistic Diversity.  Oxford: Oxford Univ.  Press. 

Ogura, M.  & W.  S-Y.  Wang.  2004.  "Dynamic Dialectology and Complex
Adaptive System".  M.  Dossena & R.  Lass (eds.), Methods and Data in
English Historical Dialectology.  Bern: Peter Lang.  (in press)

Wang, W.  S-Y.  et al.  2003.  "Computational Studies of Language
Evolution".  Keynote lecture given at COLING-2002, Academia Sinica.  (in
press)