The distribution of motivated mappings between words and meanings both within and between the world’s languages is curious- many languages, like English, are putatively arbitrary, at least when considered as a whole. On the other hand, many languages from outside of the Indo-European family have large inventories of sound-symbolic words known as ideophones. Given that these ideophones seem easier to learn than their non-motivated counterparts, why is it that English does not take advantage of similar types of sound-meaning mappings as commonly as does, say, Japanese?
One possibility is that although the use of motivated sound-meaning mappings can enhance learning of individual meanings, it might nonetheless have an overall negative influence on learnability for non-motivated mappings (a possibility discussed in Chapters 4 and 5 of my PhD. thesis, and in a forthcoming manuscript). Effectively, the use of motivated mappings might create expectations that are deleterious to learning the ‘normal’ part of the lexicon that remains. One solution to the possibility of this type of interference effect might be to insulate the motivated and non-motivated portions of the lexicon from one another, such that they are structurally, phonologically, or temporally distinct from one another.
The pattern of acquisition of motivated tokens supports one of these possibilities- words that are sound-symbolically related to their meanings are often acquired earlier. Thus, motivated mappings are (partially) temporally bifurcated from non-motivated mappings in terms of their acquisition.
Japanese, which has a large vocabulary of ideophones, also follows this pattern, but we can additionally see evidence of structural and phonological bifurcation between ideophones and the rest of the vocabulary. Ideophones are phonotactically marked and quite different from the rest of the lexicon, and make use of structures like reduplication that are very infrequent in the remainder of the lexicon (more on this here).
Using a number of agent-based bayesian models, we are attempting to explore the influence that temporal and structural bifurcation have on the learnability of motivated associations and their maintenance in the lexicon. We propose that the bifurcation of motivatedness from the rest of the lexicon is a locally optimal solution that allows children to benefit from the learnability enhancement inherent to sound-symbolic associatons without interfering with the learning and use of the more complete lexicon required for full communication and mastery of the language. Crucially, we aim to suggest that in populations without new learners (typically children) the use of motivated mappings is not selected for, while the arbitrariness allowed for by isolation of motivated elements is crucial for the acquisition of a fully functional communication system.