Contradictory data in Travis et al., 1937 (Gahl, 2020)

Does bilingualism elevate the risk of stuttering? Whenever this question is discussed in the literature, one particular study (Travis et al., 1937) is almost invariably cited. Is this citation pattern warranted?

Travis et al. reported a stuttering prevalence of 1.8% for monolingual school children and 2.8% for bilingual ones, a difference they found to be statistically significant.

However, a difference should only be noteworthy if it is supported by the data. Here, we ask whether it was in fact the case that 1.8% of the monolingual group, versus 2.8% of the bilingual group, in Travis et al. (1937) stuttered. We then ask to what extent small discrepancies in the data affect the statistical significance of the results. To preview our results: We find that the raw counts and prevalence rates reported in Travis et al. are internally inconsistent. Among other problems, it appears that the 1.8% versus 2.8% difference is due to several stuttering children being excluded from the prevalence estimate for the monolingual group. In addition, we find that the statistical significance of the results rests on the classification of very small numbers of participants—far smaller than the uncertainties in the reported numbers.

Supplemental Material S1. (1) Travis et al. (1937)’s table of percentages as published, (2) a 264 table with the reconstructed counts of participants, and (3) a step-by-step guide to 265 reconstructing the counts and uncovering the contradictions in Travis et al.’s report.

Gahl, S. (2020). Bilingualism as a purported risk factor for stuttering: A close look at a seminal study (Travis et al., 1937). Journal of Speech, Language, and Hearing Research. Advance online publication. https://doi.org/10.1044/2020_JSLHR-20-00364