Replication of Saffran, Johnson, Aslin, & Newport (1999) Statistical learning of tone sequences by human infants and adults, Exp. 1

Abstract

We replicated Exp. 1 of Saffran, Johnson, Newport, & Aslin (1999) Statistical learning of tone sequences by human infants and adults, as part of a multi-year effort to replicate every adult statistical word segmentation study. While we were able to replicate the finding of above- chance statistical segmentation of tone sequences, many of the other findings reported in the original paper did not replicate.

Replication of Saffran, Johnson, Aslin, & Newport (1999) Statistical learning of tone sequences by human infants and adults, Exp. 2

“…Since we also replicated their Exp. 1 (Murdock et al, 2017), we were able to carry out the same analysis, with the same results: overall accuracy was lower in Exp. 2 (M=62%, SE=2%) than in Exp.…”

“…Since we also replicated their Exp. 1 (Murdock et al, 2017), we were able to carry out these analysese with a binomial mixed effects model with random intercepts of subject and file. The main effect of experiment was significant (B = 0.28, SE=0.13; Wald's z = 2.2, p=.03).…”

The Meta-Science of Adult Statistical Word Segmentation: Part 1

“…We assessed statistical significance with a binomial mixed effects regression with main effects of language and transitional probability (high/low) as well as their interaction. We included a random effects for each subject and for each word/foil pair, as well as a random slope of transitional probability (high/low) for each subject (for discussion and analysis of this random effects structure, see Murdock et al, 2017a). We found a significant intercept (B = 0.82, SE = 0.11; Wald's z = 7.8, p = 8.5*10 -15 ), indicating overall above-chance learning.…”

“…The largest difference between the replication and the original is the subject population: Although the original does not report their source of subjects, it was presumably not AMT. Because training was randomly generated according to rule, the transition probabilities were very slightly different, with a somewhat clearer delineation of words in our stimuli (see Murdock et al, 2017a). There are a number of other minor differences detailed in the preprint, such as the fact that our subjects responded by keyboard rather than with paper-and-pencil (see Murdock et al, 2017a).…”

“…The differences between the replication and original are minor and analogous to those for Exp. 2, and are detailed in the preprint (Murdock et al, 2017b).…”