Vis sammendrag
Discussions on the relationship of form and meaning in modern linguistics have been dominated by the assumption that words are arbitrary, i.e. their form does not relate to their meaning (e.g. de Saussure, 1989; Hockett, 1960). Iconicity, i.e. the perceived resemblance between the form and meaning of linguistic forms, has been considered mostly marginal. Arbitrariness of the sign is thought to result from commonly occurring phonological and semantic changes which are assumed to be blind to form-meaning correspondences (e.g. Frishberg, 1975). This view predicts that even if languages had iconic signs at some point, they should be doomed to erode over time, inevitably leading to an arbitrary lexicon (a process referred to as de-iconization, Flaksman, 2017). The evidence for de-iconization stands in stark contrast, however, to studies demonstrating widespread iconicity in basic vocabulary (e.g. Blasi et al., 2016), including sensory vocabulary for color, shape, texture, speed, and size (e.g. Johansson et al., 2020). For example, words that are associated with small size are more likely to contain a high front vowel /i/ than words associated with large size (e.g. Winter & Perlman, 2021).
Without a mechanism that reintroduces iconicity, the prevalence of iconicity across languages is unaccounted for. One prominent proposal that aims to explain the system-wide maintenance of iconicity is the repeated addition of iconic forms through lexical innovation (the “iconic treadmill account”, Flaksman, 2017). Cyclical lexical innovation, however, does not sufficiently explain the large number of sound symbolic words in basic vocabularies which are rarely prone to lexical innovation. Moreover, it does not account for evidence showing that, in certain contexts, instead of eroding, the iconicity of forms stabilizes and even increases over repeated interactions (e.g. Perlman et al., 2015; Little et al., 2017). At the very least, complementary mechanisms are needed to understand the prevalence of iconicity in spoken languages.
Here I argue that iconic prosody, i.e. the use of suprasegmental features like duration and pitch to express the resemblance of form and meaning, causes systematic segmental variation which can become phonologized. The proposed mechanism is grounded in evidence from the diachronic emergence of phono-logical patterns and requires three auxiliary assumptions, all of which are supported by independent evidence: (1) speakers modulate prosody to increase iconicity; (2) prosodic modulations lead to systematic phonetic variation of seg-ments; and (3) systematic phonetic variation of segments becomes phonologized.
(1) That speakers use prosody to express iconic relationships is well established (e.g. Perlman, 2026). For instance, speakers raise their pitch to indicate small referents and lower their pitch for large ones. Moreover, listeners seem to be sensitive to these prosodic cues when trying to interpret unknown iconic forms (Dingemanse et al., 2016; Akita & Kawahara, 2024). (2) Traditionally, prosody was thought to be constrained by the phonological segments of the words it co-occurs with (Roettger, 2017), e.g. certain lexical tones can only occur with long vowels in some tone languages. However, evidence from language descriptions and speech production studies suggest that prosody can itself constrain the segmental make-up of words it co-occurs with: For example, prosodic events such as rising tones systematically trigger segmental alternations such as changes in vowel quality and vowel length, or the insertion of non-phonological segments (Roettger & Grice, 2019). Moreover, the presence of tones can affect the perception of both vowel quality and consonant voicing (e.g. Ting & Clayards, 2024). For instance, a vowel co-occurring with high pitch is more likely to be perceived as a high vowel compared to when it co-occurs with low pitch (e.g. Steffman, 2023). Finally, (3) Sound systems are shaped by perceptual, articulatory, and acoustic aspects of speech transmission (e.g. Blevins, 2004). Reoccurring small prosodically induced biases (to e.g. express iconicity) can function as diachronic attractors for phonological segments, creating certain systematic evolutionary pathways that are more common than others.
Taking these lines of evidence into account, we can speculate that e.g. a word that systematically co-occurs with high pitch for expressive purposes, might be prone to being misperceived as a higher vowel, adding another explanatory mechanism to e.g. the prevalence of high vowels for words associated with small size. Phonetic biases grounded in low-level properties of speech transmission can lead to the misperception of segments co-occurring with these prosodic modulations and can lead to segmental iconicity through sound change in the lexicon. Such a mechanism can lead to an increase of phonological iconicity over time, thereby counteracting the general diachronic trajectory towards increasing levels of arbitrariness without lexical innovation. In the future, the proposed mechanisms should be empirically tested through controlled production tasks combined with artificial language learning paradigms.
