Broedelet, I. , Rispens, J. & Boersma, P.
University of Amsterdam
There is great need for online measures of statistical learning. To this end, we constructed a word segmentation task using four disyllabic words, with an 8-minute familiarization phase that consisted of natural co-articulated speech. Click sounds were added to the stream, to which participants were asked to respond as fast as they could by pressing a button. The clicks occurred either between words or within words. We hypothesized that knowledge of word boundaries should result in faster reaction times for clicks between words than for clicks within words. 30 adults participated in the experiment. Click position did not significantly affect detection speed. In the offline test phase, participants did not show any preference for words (TP=1) over part-words (TP=0.33). As the click detection task might have hampered learning overall, 30 more adults and 30 children were tested on a version without clicks. However, neither group showed evidence of learning word boundaries. Currently, we are conducting a third experiment with non-words (TP=0) instead of part-words (TP=0.33) in the test phase to investigate whether choosing words over part-words was problematic in our earlier experiments. Our results will be discussed in the light of methodological issues in measuring (online) statistical learning.