Accounting for the stochastic nature of sound symbolism using Maximum Entropy model

Abstract

Sound symbolism refers to stochastic and systematic associations between sounds and meanings. Sound symbolism has not received much serious attention in the generative phonology literature, perhaps because most if not all sound symbolic patterns are probabilistic. Building on the recent proposal to analyze sound symbolic patterns within a formal phonological framework (Alderete and Kochetov 2017), this paper shows that MaxEnt grammars allow us to model stochastic sound symbolic patterns in a very natural way. The analyses presented in the paper show that sound symbolic relationships can be modeled in the same way that we model phonological patterns. We suggest that there is nothing fundamental that prohibits formal phonologists from analyzing sound symbolic patterns, and that studying sound symbolism using a formal framework may open up a new, interesting research domain. The current study also reports two hitherto unnoticed cases of sound symbolism, thereby expanding the empirical scope of sound symbolic patterns in natural languages.

Teaching phonetics through sound symbolism

“…It was therefore expected that Takarazuka players may try to convey their "Takarazuka-gender" through their names. As predicted, Katsuda's analysis found that as shown in Figure 3, the more sonorants are contained in their name, the more likely that these names are used for female roles [14]. This result is a clear case in which the sound symbolic relationship between sonorants and females is playing a crucial role in naming.…”

“…On the other hand, only 7 out of 643 obstruents (1%) were targeted by reduplication. This difference is statistically significant, and shows that Japanese AKB idols are actively choosing to increase the number of sonorants in their nicknames [14].…”

Sound symbolism and theoretical phonology

“…A similar question arises when studying sound symbolism. For instance, some studies report that two instances of the same segment can evoke a stronger image than one instance (Hamano, 2013; Kawahara et al, 2019; Kawahara and Kumagai, to appear; Martin, 1962), which is akin to what Jäger and Rosenbach (2006) refer to as counting cumulativity . Moreover, D'Onofrio (2014) shows that in the bouba ‐ kiki effect, several phonetic/phonological features matter in determining the perceived roundness/angularity of visual images (e.g., vowel backness, voicing, and place of articulation), and that these effects are cumulative, which would remind phonologists of ganging‐up cumulativity (Jäger & Rosenbach, 2006).…”

“…They propose an analysis of expressive palatalization using Optimality Theory (Prince & Smolensky, 2004), with a set of violable constraints (E xpress[x] ) specifying which sounds should be realized to express which meanings. Kawahara, Katsuda, and Kumagai (2019) analyze sound symbolic connections themselves (rather than alternations triggered by sound symbolic considerations), and argue that as long as generative phonology is a function that maps one representation (e.g., underlying forms) to another (e.g., surface forms)—as it in fact has been—there is nothing that prevents us from using the same formalism to model the mapping from representation in one modality (i.e., sound) to representation in another modality (i.e., meaning). In other words, the grammatical device that phonologists have been using for decades can be applied to formalize sound symbolic connections at no additional costs.…”

“…Likewise, recent studies have demonstrated that phonological knowledge can be stochastic; for example, some structures tend to be preferred over others, and some alternations are more likely to occur in one environment than in other environments (Boersma & Hayes, 2001; Hayes & Londe, 2006; Pierrehumbert, 2001; Zuraw & Hayes, 2017). Focusing on this parallel between phonological patterns and sound symbolic connections, Kawahara et al (2019) propose to use the MaxEnt model with Optimality Theoretic constraints (Goldwater & Johnson, 2003) to account for stochastic aspects of sound symbolic patterns. For example, in Japanese Takarazuka actress names, the more sonorants are contained in their name, the more likely that name is used for a female name.…”