Sound-meaning associations allow listeners to infer the meaning of foreign language words

Human language exhibits a seemingly arbitrary relationship between a word's form and its meaning. However, research suggests that people often share similar intuitions about the meaning of unfamiliar words, a phenomenon potentially linked to the sensorimotor qualities of speech and their connection to the external world. This "sound symbolism" is readily apparent in onomatopoeia but less understood in natural languages. This study explores whether listeners can infer the meaning of foreign words based solely on their phonological form, focusing on whether individual differences predict this ability. It investigates if cross-linguistic regularity in form-meaning associations and cognitive abilities (specifically, verbal working memory) predict the extraction of meaning from foreign language wordforms in nine different languages. Understanding this process is crucial for comprehending both the evolution of language and the cognitive mechanisms underlying language processing. This research addresses the gap in understanding how sound symbolism manifests in natural languages by investigating the role of iconicity—the resemblance between form and meaning, broader than sound symbolism, encompassing sound, visual and articulatory features.

Existing research on sound symbolism largely focuses on the objective or perceived resemblance between acoustic features of words and their referents. However, this is a limited view. The broader concept of linguistic iconicity encompasses resemblance across various modalities, including cross-modal associations (e.g., vowel length and object size, consonant voicing and object shape). Much of the past work utilizes artificial stimuli, which may not generalize well to natural language. While valuable for isolating specific associations, artificial words often maximize differences, potentially obscuring patterns in natural languages. Studies using natural language words often limit themselves to ideophones, again limiting generalizability. Recent large-scale cross-linguistic studies, however, indicate systematic form-meaning associations in natural languages. For instance, a study found a correlation between words related to roughness and the presence of a trilled /r/ across various languages. Neuroimaging studies suggest that cross-modal integration, supported by brain regions like the left superior parietal cortex and right superior temporal sulcus, might underpin form-meaning mapping. These areas are also linked to verbal working memory and cross-modal information binding. Research on non-human primates suggests that the human capacity for form-meaning mapping might stem from the advanced development of the arcuate fasciculus (AF) and superior capacity for maintaining phonological and semantic associations in verbal working memory, possibly facilitated by mirror neurons.

This study employed a forced-choice antonym task. Native monolingual English speakers (N=134) and native Spanish speakers (N=46) listened to pairs of foreign words and identified the corresponding English antonym pair. Nine languages (Japanese, Mandarin, Thai, Polish, Russian, Ukrainian, French, Romanian, Spanish) were used in the English speaker study, representing a range of language families and degrees of similarity to English. The Spanish speaker study replicated the procedure with Japanese, Polish, and English. Stimuli were created using text-to-speech services, normalized for amplitude, and checked by native speakers. Participants were screened for native language proficiency and minimal foreign language knowledge. The experiment consisted of three blocks, each containing 45 trials, with language assignments and block order counterbalanced. Each trial involved listening to a foreign antonym pair and selecting the corresponding English pair from two options. The order of words within each pair and the position of the correct answer were randomized. Participants were allowed to replay the audio and provided with feedback after each block. Following the task, participants identified any previously known foreign words, completed a verbal working memory test (digit span), and a language background survey. Data analysis involved generalized linear mixed-effects models to examine the effects of language, part of speech, phonetic distance from the native language, form-meaning regularity (measuring phonological overlap across languages for words with the same meaning), and verbal working memory on accuracy. Phonetic distances were calculated using the Sounded Cognates Algorithm (SCA). Form-meaning regularity was calculated by comparing phonetic distances between same-meaning and opposite-meaning words across different language groups. Participants with digit span scores more than 1.5 standard deviations below the mean were excluded from the interaction analysis.

Across both native English and Spanish speaker groups, participants demonstrated above-chance accuracy in identifying the meanings of foreign antonym pairs. Accuracy varied significantly across languages. For English speakers, accuracy was highest for Romance languages (French, Romanian, Spanish), followed by Slavic, and then Japonic-Sino-Tai languages. Within language families, accuracy did not vary significantly. Accuracy also varied by part of speech, with nouns often showing higher accuracy than verbs and adjectives, particularly in Romance languages. Phonetic distance from the native language positively correlated with accuracy; shorter distances led to higher accuracy. This effect varied by language group and part of speech. For English speakers, verbal working memory significantly influenced accuracy. Better working memory was associated with greater accuracy, especially for words with greater form-meaning regularity across languages. However, this correlation was not as strong in the Spanish speaker replication study. In both studies, words with the same meaning across languages showed significantly shorter phonetic distances than words with opposite meanings. Form-meaning regularity was found to enhance accuracy for participants with better verbal working memory. The Spanish replication study confirmed the main effect of language and part of speech, but the effects of verbal working memory and form-meaning regularity were less robust.

The findings challenge the traditional view of language as entirely arbitrary, providing evidence for systematic sound-meaning regularities across diverse languages. The above-chance accuracy in inferring meaning suggests that listeners leverage sublexical cues and cross-linguistic regularities. The relationship between verbal working memory and accuracy, especially for words with high form-meaning regularity, indicates that cognitive factors play a critical role in exploiting these regularities. This suggests that the ability to identify form-meaning connections is not isolated but reflects a broader cognitive function. However, the reduced effect of working memory in Spanish speakers highlights the potential influence of prior language experience, orthographic transparency, and language complexity on the reliance on working memory during auditory comprehension. The consistent influence of phonetic distance on accuracy across both studies underscores the role of cross-linguistic transfer in word meaning inference, though the ability to infer meaning from distant languages points to the broader existence of sublexical regularities. The part-of-speech effects, particularly the higher accuracy for concrete nouns, could reflect the ease of evoking perceptual features of referents, aligning with the order of word acquisition in children.

This study demonstrates consistent sound-meaning mappings across multiple languages, showing that meaning inference is possible even without prior knowledge of the language. Verbal working memory enhances this ability, particularly for words with consistent form-meaning relationships, although this effect varied by native language. Phonetic distance to the native language also played a crucial role. Future research should investigate the interplay between these factors in more detail and with larger, more diverse populations to further understand the role of linguistic iconicity and the cognitive processes behind cross-linguistic meaning inference. This will expand on our understanding of language evolution and the relationship between language and cognition.

While the findings support the prevalence of linguistic iconicity, other factors such as historical relationships between languages and communicative constraints could also explain cross-linguistic similarities in sound-meaning associations. The study did not directly assess the perceived similarity between sounds and referent qualities. Future research should directly address this to strengthen the conclusions about sound symbolism. The replication study had a smaller sample size than the primary study and this may explain some of the differing findings.