One form, many meanings: iconicity in phonological and semantic development

Abstract

This thesis investigates the consequences of iconicity—a motivated mapping between form and meaning—for the phonological and semantic development of iconic words, through the lens of dialectal variation in the forms and meanings of ideophones in Japonic (Japanese, Ryukyuan). I use the term ideophone to refer to a lexical class of iconic words dedicated to the depiction of sensory phenomena. In Japanese, this class numbers well into the thousands, and includes ideophones mimicking sound (onomatopoeia), e.g. kokekkokoo 'cock-a-doodle-doo', as well as ideophones depicting movement (buruburu ‘wobbling’), shape (kunyakunya ‘twisting’), texture (nebaneba 'sticky'), and even internal states (mukamuka ‘nauseous’, wakuwaku 'excited') (Akita, 2017). The phonological and semantic development of ideophones is presumed to be shaped and constrained by their iconicity. For example, it has been noted that sound changes occurring regularly in the arbitrary lexicon are often absent in ideophones, presumably to conserve iconic mappings between form and meaning (e.g. Diffloth, 1979). Similarly, the semantic development of ideophones is assumed to follow an implicational hierarchy with sound—the easiest domain to represent iconically in speech—as the first domain in which ideophones appear, followed by domains that are more difficult to represent iconically. Thus, we find languages whose ideophone systems are limited to the depiction of sound, and those whose ideophones depict a broader range of sensory perceptions in addition to sound, but we do not find languages that depict broader sensory perceptions without also depicting sound (Dingemanse, 2012). This thesis investigates these two ideas by using dialectal variation to explore the phonological and semantic development of ideophones in Japanese. The first part of the thesis compares the degree to which sound change has occurred between Japanese and two varieties of Ryukyuan (Miyako and Ishigaki) in the iconic versus the arbitrary lexicon (e.g. the iconic gunyagunya 'mushy' vs. the arbitrary juunan 'pliable'). Taking Japanese as a baseline, I adopt a double-pronged approach to quantifying sound change. First, for Ryukyuan words with a Japanese cognate (N=330), string edit distances between cognates provide a measure of how much sound change has occurred in the cognate set. Second, for all words (N=404), I developed a phonotactic deviation score, which is a measure of how much a Ryukyuan word differs from Japanese words in general, based on the number of deviations it makes from Japanese phonotactics. Both measures confirmed that the iconic words studied had undergone significantly less sound change than arbitrary words from the same semantic domains. However, onomatopoeia were more likely to incorporate new sounds and sound configurations than ideophones depicting inaudible phenomena. The directness of the iconic mapping in onomatopoeia means that when new sounds or sound configurations enter the language, these can be used in onomatopoeia with a transparent interpretation. On the other hand, less directly iconic mappings between sound and other sensory domains must be established through large sets of similar systematic correspondences, and thus it takes longer for sound changes to make their way into this part of the lexicon. This confirms that, compared to other vocabulary, the forms of ideophones are diachronically very stable, but with the insight that this is less likely to be the case for onomatopoeia. The second part of the thesis addresses the idea of an implicational hierarchy for the semantic development of ideophone systems, with a broader cross-linguistic sample than has been used in previous studies. Using this broader cross-linguistic sample, I formulate an implicational hierarchy SOUND < MOVEMENT < SHAPE < TEXTURE < OTHER SENSORY PERCEPTIONS, which says that languages develop ideophones first in the domain of sound, next in the domain of movement, third in the domain of shape, and fourth in the domain of texture, but that after this point their developmental trajectories may vary. This is illustrated by highlighting asymmetries in the cross-linguistic distribution of ideophones. Next, I investigate why the hierarchy should emerge in this way with a detailed analysis of Japanese and Ryukyuan ideophones elicited in response to sensory stimuli of varying modalities (including for e.g. sounds, movements, shapes, colours, textures, tastes, smells and stimuli representing internal perceptions such as pain and emotion). To introduce a diachronic perspective, the task was conducted in locations representing both the centre (Tokyo, Kyoto) and the furthest edges (Tohoku, Okinawa) of Japan, in order to contrast older forms and usages preserved in peripheral varieties with the newest forms and usages appearing in central varieties (Onishi, 2010). In total, data was collected from 80 participants representing seven major linguistic varieties: Eastern Japanese, Western Japanese, Tohoku Japanese, Amami Ryukyuan, Okinawa Ryukyuan, Miyako Ryukyuan and Yaeyama Ryukyuan. More data was collected from speakers of Ryukyuan varieties, as these are highly endangered and their ideophone systems are not well described. Thus, this thesis also makes a small contribution to understanding differences between the ideophone systems of Japanese and Ryukyuan, and to the documentation of ideophones in these languages. The results of the stimuli task support the idea that domains appearing earlier in the hierarchy are easier to represent iconically, in that ideophones were elicited most frequently in these domains. This was further supported by the multisensory meanings of ideophones elicited, as ideophones in domains appearing later in the hierarchy often had related associations in domains appearing earlier in the hierarchy (e.g. ideophones elicited in response to textural stimuli were often onomatopoeic, mimicking the sound produced by the stimulus when touched). In contrast, ideophones from domains early in the hierarchy were more likely to be monosensory, meaning that they did not have any related associations in domains other than that of the stimulus which elicited them (e.g. ideophones elicited in response to auditory stimuli tended to depict sound without reference to any other sensory domain). This reinforces the idea that iconic mappings are formed more easily in domains from early in the hierarchy, since these are frequently recruited to depict domains appearing later in the hierarchy, whereas the reverse situation is much rarer. To explain this, I consider the relative iconicity of mappings from speech to domains appearing earlier or later in the hierarchy. Specifically, I argue that mappings to domains appearing early in the hierarchy are likely to exhibit a greater degree of iconicity than mappings to domains appearing later in the hierarchy. This is attributed to the iconic affordances of the acoustic and articulatory properties of speech, as well as the multimodal integration of our sensory processing systems. In sum, this thesis develops a detailed account of how the phonological and semantic development of ideophones is intimately related to their iconicity. In terms of phonology, a close analysis of the influence of different types and degrees of iconicity on sound change revealed that while the iconicity of ideophones generally results in diachronically stable forms, paradoxically, the more transparent iconicity of onomatopoeia means that these more iconic words can exhibit less diachronic stability. This is a much more nuanced account of iconicity and sound change than has previously been developed. Secondly, in terms of semantics, the study demonstrates how the iconicity of ideophones leads to an implicational hierarchy, with some domains (e.g. sound and movement) appearing earlier than others (e.g. shape and texture) due to the ease with which they can be represented iconically in speech. However, the study also reveals that this implicational hierarchy is not as pervasive as previously assumed, and that there is room for cross-linguistic variation as well. Finally, the last chapter of the thesis considers the broader implications of these findings for future directions in the study of iconicity across different languages and modalities.