The traditional account of the relationship between phonetics and meaning holds that there is a “duality of patterning” (Bloomfield 1933). Phonemes are individually meaningless and are separated by an iron wall from morphemes
The tools to resolve this difficulty have been sitting in front of us almost as long, but have been blocked by a slavish adherence to surface forms. It is well known that a single phoneme may have several different realizations. For example the English /p/ is aspirated word initially but unaspirated after a sibilant. In German some think that the same phoneme may be realized as [h] in the syllable onset but [ŋ] in the coda. In some American dialects /t/ and /d/ have the same phonetic surface form of [ɾ]. The assumption of a simple one-
I propose that sound symbolism is the default and unmarked case. Instances where the phoneme meaning is harder to divine can be seen as examples of homophony where different phonemes have the same surface realization. You can imagine the difficulty a theory of lexical semantics would have trying to account for the meanings of both wear and where in the same word. Furthermore, once we properly distinguish between the fundamental phoneme and mere accidents of production, we can see that morphemes are actually compositionally produced from both the sound and the meanings of their component phonemes. For example, /gl/ is traditionally thought of as a two-
A critic may charge us with spurious duplication of meaning. Why should a word have so many phonemes if just one or two would be enough to specify its meaning? The reason for this redundancy is that there are a limited number of distinguishable sounds, so in the noisy production environment different phonemes can be confused with each other. Mandarin Chinese, faced with a similar problem of massive homophony, solved it by compounding words with extra morphemes to disambiguate; for example, hǔ (“tiger”) was expanded to lǎohǔ (“old tiger”). The same process could have occurred to give us our multi-
Someone might complain that a particular phoneme may appear so rarely that a child would not have enough time to learn its meaning before learning the meaning of the whole word. This objection can be trivially answered by positing a Universal Phonology. The set of semantic phonemes is fixed in the human blueprint. A child merely has to learn the language-
In conclusion, Socrates was fundamentally right. There is a direct, non-