An onset is an onset: Evidence from abstraction of newly-learned phonotactic constraints

Abstract

Phonotactic constraints are language-specific patterns in the sequencing of speech sounds. Are these constraints represented at the syllable level (ng cannot begin syllables in English) or at the word level (ng cannot begin words)? In a continuous recognition-memory task, participants more often falsely recognized novel test items that followed than violated the training constraints, whether training and test items matched in word structure (one or two syllables) or position of restricted consonants (word-edge or word-medial position). E.g., learning that ps are onsets and fs codas, participants generalized from pef (one syllable) to putvif (two syllables), and from putvif (word-edge positions) to bufpak (word-medial positions). These results suggest that newly-learned phonotactic constraints are represented at the syllable level. The syllable is a representational unit available and spontaneously used when learning speech-sound constraints. In the current experiments, an onset is an onset and a coda a coda, regardless of word structure or word position.

Novel phonotactic learning: Tracking syllable-position and co-occurrence constraints

“…Sensitivity to phonotactic constraints is exhibited early (e.g., Jusczyk, Friederici, Wessels, Svenkerud, & Jusczyk, 1993; Jusczyk, Luce, & Charles-Luce, 1994; Zamuner & Kharlamov, 2016 for recent review) and remains flexible, allowing, in some cases, adult second-language learners to be nearly as sensitive to the phonotactics of their second language as to those of their first language, and nearly as sensitive as native speakers of that language (Weber & Cutler, 2006). Moreover, adults quickly learn novel phonotactic constraints such as P-starts and F-ends words from brief auditory (e.g., Bernard, 2015; Chambers, Onishi, & Fisher, 2010; Onishi, Chambers, & Fisher, 2002) or production experience (e.g., Dell, Reed, Adams, & Meyer, 2000; Goldrick, 2004; Goldrick & Larson, 2008; Kittredge & Dell, 2011; Warker & Dell, 2006, 2015; Warker, Dell, Whalen, & Gereg, 2008; Warker, Xu, Dell, & Fisher, 2009), enabling them to respond differentially to novel words like paf and fap that either follow or violate these constraints, as demonstrated by their rates of false recognition, repetition latencies, or production accuracies.…”

“…For instance, Bernard (2015) found support for syllable-level learning while directly manipulating syllable position by varying word structure and word position between training and test words. For example, participants trained on items like buF.Pak (where the period indicates the syllable boundary and capital letters indicate restricted consonants) distinguished between test items such as Pit.vuF (with F and P in the same syllable positions as in the training items) and Fit.vuP (with F and P in different syllable positions).…”

“…For example, participants trained on items like buF.Pak (where the period indicates the syllable boundary and capital letters indicate restricted consonants) distinguished between test items such as Pit.vuF (with F and P in the same syllable positions as in the training items) and Fit.vuP (with F and P in different syllable positions). Bernard (2015) did not, however, manipulate co-occurrences: the unordered consonant-vowel and consonant-consonant co-occurrences were the same across training and test. Thus, while it is possible that adults represented syllable-position constraints beyond the particular co-occurring contexts in which the constraints were experienced (e.g., from buF.Pak in training to ‘F is coda’ regardless of the sounds that occurs adjacent to the F)—allowing them to process F-onset and F-coda items differently (e.g., Pit.vuF vs. Fit.vuP in test)—it is also possible that the presence of familiar co-occurrences (e.g., F with U) assisted their learning and their generalization to novel word positions (e.g., ‘F is coda when it is adjacent to U’).…”

“…Some studies (mostly examining artificial-language learning) have suggested that people can track syllable-position constraints (e.g., Bernard, 2015; Chambers et al, 2003; 2010; 2011; Dell et al, 2000; Goldrick, 2004; Goldrick & Larson, 2008; Kittredge & Dell, 2011; Onishi et al, 2002; Warker & Dell, 2006; Warker et al, 2008; Warker et al, 2009) while typically not controlling for segment co-occurrence information. Other studies (mostly examining knowledge of native-language phonotactics) have suggested that people can track co-occurrence constraints (e.g., Gathercole et al, 1999; Mattys et al, 1999; Munson, 2001; Richtsmeier et al, 2009a, 2009b, 2011) while typically not controlling for syllable position information.…”

“…To assess whether phonotactics can be represented at the level of the syllable and/or at the level of local co-occurrences alone, the current experiments build on recent work that suggests that adults can learn and represent sound constraints at the syllable-position level (Bernard, 2015). In the previous work, participants heard a series of words from an artificial language, and their task was to indicate, for each word, whether they had heard it earlier in the experiment (i.e., whether it was repeated) or not (i.e., whether it was novel).…”