Semantic Glimmers: Phonaesthemes Facilitate Access to Sentence Meaning

The association between sound and meaning is commonly thought to be symbolic and arbitrary. While this appears to be mostly correct, there is evidence that specific phonetic groupings can be broad indicators of word meaning (Bergen, 2004). Phonaesthemes are sub-morphemic clusters that are predictably related to vague semantic content; this vagueness differentiates them from morphemes. For example, the phonaestheme gl- occurs significantly more often in words related to light and vision (glimmer, gleam, glaze) than would be expected by chance, but one would not indicate that something is related to light or vision by adding gl- to a word.The current work examines the strength and significance of the relationship between form and meaning, and tests whether it affects performance on a paraphrasing task. Previous literature identifies 46 potential phonaesthemes; we added two new candidate phonaesthemes and two potential non-phonaesthemes to this list and examined their strength in a Monte-Carlo analysis of the relatedness of the words' vectors in a semantic space generated by Infomap (Schutze, 1997). Our analysis found a significant semantic relationship in 29 of the 50 word-form groups. If phonaesthemes are significantly semantically related to each other and to their context, reader/hearers may use them as clues in deciphering unfamiliar words. To test this hypothesis, we asked participants to read sentences containing phonaestheme-bearing nonsense words, and to paraphrase each sentence. Using Infomap to derive coherence values, we constructed sentences that were highly coherent with one phonaestheme-bearing target word from each of six groups. We then replaced these words with nonsense words from the following categories: (1) Target congruent with sentence; (2) Target incongruent with sentence; (3) Nonsense word with no phonaestheme (neutral).We found a significant difference in the time between the disappearance of the sentence from the screen to the beginning of typing the paraphrase, controlling for sentence length (Typing Latency). Participants were faster to begin paraphrases of Congruent sentences (M=85.03 ms/wd) than Incongruent (M=98.10 ms/wd) or Neutral (M=103.66 ms/wd). This difference is statistically significant between Congruent and Incongruent conditions (t=2.42, p=0.03), indicating that though no such effect was evident in the sentence reading/paraphrase composition times, participants had quicker access to Congruent paraphrases than in the other two conditions. In another task, participants were presented with the same sentences but with the target word replaced with a blank, and asked to complete the sentence with one of three nonsense words (one with a congruent phonaestheme, one with an incongruent phonaestheme, and one with no phonaestheme). Participants chose the congruent words at rates significantly above chance (51% of the time, t=5.0, pPrevious research has shown that when readers come across an unfamiliar word, they rely on contextual cues to help them make sense of the word. However, the current research demonstrates that word form can provide one of these cues if it is in harmony with the sentence's meaning.