Familiar units prevail over statistical cues in word segmentation

Abstract

In language acquisition research, the prevailing position is that listeners exploit statistical cues, in particular transitional probabilities between syllables, to discover words of a language. However, other cues are also involved in word discovery. Assessing the weight learners give to these different cues leads to a better understanding of the processes underlying speech segmentation. The present study evaluated whether adult learners preferentially used known units or statistical cues for segmenting continuous speech. Before the exposure phase, participants were familiarized with part-words of a three-word artificial language. This design allowed the dissociation of the influence of statistical cues and familiar units, with statistical cues favoring word segmentation and familiar units favoring (nonoptimal) part-word segmentation. In Experiment 1, performance in a two-alternative forced choice (2AFC) task between words and part-words revealed part-word segmentation (even though part-words were less cohesive in terms of transitional probabilities and less frequent than words). By contrast, an unfamiliarized group exhibited word segmentation, as usually observed in standard conditions. Experiment 2 used a syllable-detection task to remove the likely contamination of performance by memory and strategy effects in the 2AFC task. Overall, the results suggest that familiar units overrode statistical cues, ultimately questioning the need for computation mechanisms of transitional probabilities (TPs) in natural language speech segmentation.

Linguistic entrenchment: Prior knowledge impacts statistical learning performance

“…From a theoretical perspective, the effect of prior knowledge is not uniform across all items in the verbal auditory SL task. It could facilitate performance for some items but hinder performance for others, resulting in low internal consistency without necessarily impacting overall success (see also Poulin-Charronnat et al, 2016). Thus, the only difference between the verbal and non-verbal tasks was in the amount of shared variance between items, or, in the extent to which performance in one item predicted performance in other items.…”

“…First, there is a relatively large set of studies showing that the expectations that participants bring to SL tasks can be easily manipulated, affecting task performance. For example, pre-exposing participants to isolated words or part-words before the beginning of the familiarization stream has a dramatic effect on SL performance, which can either facilitate (Cunillera, Laine, Camara, & Rodriguez-Fornells, 2010;Lew-Williams, Pelucchi, & Saffran, 2011), or hinder (Perruchet et al, 2014;Poulin-Charronnat, Perruchet, Tillmann, & Peereman, 2016) learning. In the same vein, pre-familiarizing participants with words of different length affects the size of the units they extract from the input (Lew-Williams & Saffran, 2012).…”

Statistical Learning Is Not Age‐Invariant During Childhood: Performance Improves With Age Across Modality

“…Learning these particular relations may not change much after the first year of life, when phonetic categories are established and word segmentation is performed (e.g., Kuhl, Williams, Lacerda, Stevens, & Lindblom, 1992;Werker & Tees, 1984). Linguistic auditory tasks are sensitive to prior knowledge in a way that other SL tasks are not (Perruchet, Poulin-Charronnat, Tillmann, & Peereman, 2014;Poulin-Charronnat, Perruchet, Tillmann, & Peereman, 2017;Siegelman, Bogaerts, Elazar, Arciuli, & Frost, 2018), but the effect of that knowledge may not change much after it has been established. Since knowledge about syllable co-occurrences is already present in early childhood (Storkel, 2001), our youngest age group (six-year-olds) may have been too "old" to show age effects.…”

Lexical knowledge boosts statistically-driven speech segmentation.

“…Familiarization improved subsequent segmentation of the artificial language. Finally, Poulin-Charronnat, Perruchet, Tillmann, and Peereman (2017) found that familiarizing sequences inconsistent with the statistically defined words impaired learning. Thus, short sequences familiarized immediately prior to exposure of an artificial language can constrain segmentation.…”