Rapid Statistical Learning Supporting Word Extraction From Continuous Speech

Abstract

The identification of words in continuous speech, known as speech segmentation, is a critical early step in language acquisition. This process is partially supported by statistical learning, the ability to extract patterns from the environment. Given that speech segmentation represents a potential bottleneck for language acquisition, extracting patterns in speech may occur very rapidly, without extensive exposure. This hypothesis was examined by exposing participants to continuous speech streams composed of novel, repeating nonsense words. Learning was measured online using a reaction time task. After merely one exposure to an embedded novel word, learners demonstrated significant learning effects, revealed by faster responses to predictable syllables. These results demonstrate that learners gained sensitivity to the statistical structure of unfamiliar speech on a very rapid timescale. This ability may play an essential role in early stages of language acquisition, allowing learners to rapidly identify word candidates and “break in” to an unfamiliar language.

Redefining “Learning” in Statistical Learning: What Does an Online Measure Reveal About the Assimilation of Visual Regularities?

“…Interestingly, these two accounts of learning also make contrasting predictions regarding the online formation of predictions and their temporal dynamics. In this context, it is interesting to note that in all experiments involving embedded triplets of stimuli in the current work (Experiments 1a, 1b, and 3), there was no difference in RTs between shapes in 2nd position to those in 3rd position, a result that is in line with predictions from chunk-based, but not TP-based, learning accounts (for discussion, see Batterink, 2017;Minier, Fagot, & Rey, 2016). As this is a post hoc null finding, it requires careful interpretation.…”

Do current statistical learning tasks capture stable individual differences in children? An investigation of task reliability across modality

“…In fact, using online measures to track the trajectory of visual statistical learning in adults found that learning is already evident after seven repetitions of each triplet and does not increase in magnitude with increased exposure (Siegelman et al, 2018a). Similar patterns were found in the auditory domain, in which learning plateaued after only three repetitions (Batterink, 2017; though see Batterink & Paller, 2017, for enhanced neural activation following increased exposure). That is, it is unclear whether increasing exposure length will improve learning significantly.…”

Tracking the implicit acquisition of nonadjacent transitional probabilities by ERPs

“…Although chunk learning has been found to be an important contributor to implicit statistical learning (e.g., Batterink, 2017; Perruchet & Pacton, 2006), it has recently been suggested that both rule-based statistical computation and chunk learning operate during this learning form (Christiansen, 2018; Fu, Sun, Dienes, & Fu, 2018). This would be plausible for the ASRT task, as well.…”