BALANCE OF MEANING PROBABILITIES IN PROCESSING OF SERBIAN HOMONYMY

Dušica Filipović Đurđević

doi:10.19090/pp.2019.3.283-304

Abstract

The research deals with the set of Serbian homonymous nouns (nouns with multiple unrelated meanings) presented in the norming study and in the visual lexical decision task experiment. Native speakers listed the meanings of homonymous words and provided word familiarity and word concreteness ratings. Accordingly, the first database of Serbian homonyms was constructed containing subjective meanings of homonymous nouns along with the estimated meaning probabilities, as well as a number of meanings, redundancy and entropy of the distribution of meaning probabilities, word familiarity and word concreteness. The processing disadvantage of homonymous nouns over unambiguous nouns was replicated in the visual lexical decision task. Additionally, the processing of homonymous nouns was linked with redundancy: the information theory measure of the balance of meaning probabilities. The results revealed that homonyms with higher redundancy of the meaning probability distribution (i.e., unbalanced meaning probabilities) were processed faster. This finding was in accordance with the hypothesis derived from the Semantic Settling Dynamics account of the processing of ambiguous words, according to which the competition among the unrelated meanings derived the processing disadvantage in homonymy. However, the same pattern was not observed for the number of meanings and entropy, inviting for further research of the processing of ambiguous words.

Highlights

Processing of lexical ambiguity has long been the subject of psycholinguistic investigations
The observed asymmetry in processing effects of polysemy and homonymy was accounted for by parallel distributed model proposed by Rodd, Gaskell, and Marslen-Wilson (2004), and a similar upgraded model named Semantic Settling Dynamics (SSD) model proposed by Armstrong and Plaut (2016; and described in more detail by Armstrong, 2012)
A number of meanings and redundancy were estimated in two ways: by counting all of the meanings listed by participants, and by counting only the meanings that were listed by at least two participants

Summary

Introduction

Processing of lexical ambiguity has long been the subject of psycholinguistic investigations. The first step in resolving this inconsistency was accomplished by Rodd, Gaskell, and MarslenWilson (2002), who demonstrated that the type of lexical ambiguity was important for processing effects. They pointed to the difference between homonyms, words with unrelated meanings (e.g., river bank and financial bank), and polysemes, words with multiple related senses (e.g., daily paper and paper as a material). The delay in recognition time (i.e., a greater number of simulation cycles) for the homonyms relative to the unambiguous words was attributed to the competition among the unrelated meanings at the semantic level (due to inhibitory connections among the units which represented those meanings). In the case of polysem, competition would arise later in the processing, as predicted by the model (Armstrong, 2012; Armstrong & Plaut, 2016; Rodd et al, 2004), and observed in the empirical data (for a review see Edington & Tokowicz, 2015, and Armstrong, 2012; Armstrong & Plaut, 2016)

Objectives

Methods

Results

Conclusion