Analysis of distributions reveals real differences on dichotic listening scores between left- and right-handers

Abstract

About 95% of right-handers and 70% of left-handers have a left-hemispheric specialization for language. Dichotic listening is often used as an indirect measure of this language asymmetry. However, while it reliably produces a right-ear advantage (REA), corresponding to the left-hemispheric specialization of language, it paradoxically often fails to obtain statistical evidence of mean differences between left- and right-handers. We hypothesized that non-normality of the underlying distributions might be in part responsible for the similarities in means. Here, we compare the mean ear advantage scores, and also contrast the distributions at multiple quantiles, in two large independent samples (Ns = 1,358 and 1,042) of right-handers and left-handers. Right-handers had an increased mean REA, and a larger proportion had an REA than in the left-handers. We also found that more left-handers are represented in the left-eared end of the distribution. These data suggest that subtle shifts in the distributions of DL scores for right- and left-handers may be at least partially responsible for the unreliability of significantly reduced mean REA in left-handers.

Keywords: cerebral lateralization; dichotic listening; handedness; hemispheric asymmetry; language.

Fig. 1

Overlapping density curves of left-handers (LH; green) and right-handers (RH; blue) dichotic listening LI scores for (A) Bergen and (B) Bangor. The overlap between the two handedness group distributions can be seen as turquoise in both graphs.

Fig. 2

Scatterplots and shift function for Bergen DL LI scores. Panel a shows the distribution of LH and RH scores. Panel B illustrates the same distributions. The dark vertical lines mark the deciles in each distribution, with the median line of each distribution slightly thicker. Between distributions, the matching deciles are joined by purple lines, indicating a negative decile difference between the LH and RH groups. The values of the differences for deciles 1 and 9 are indicated in the superimposed labels. In panel C, the x-axis shows the deciles of LH scores, and the y-axis the differences between deciles (how much LI score deciles from the RH distribution needs to be shifted to match those of the LH distribution). The vertical lines indicate the 95% bootstrap confidence intervals. These negative LI quantile difference scores indicate that the left-handed group have smaller LIs at all points of the distribution.

Fig. 3

Scatterplot and shift function for Bangor DL LI scores. Panel a shows the distribution of LH and RH scores. Panel B illustrates the same distributions. The dark vertical lines mark the deciles in each distribution, with the median line of each distribution slightly thicker. Between distributions, the matching deciles are joined by purple lines, indicating a negative decile difference between the LH and RH groups. The values of the differences for deciles 1 and 9 are indicated in the superimposed labels. In panel C, the x-axis shows the deciles of LH, and the y-axis the differences between deciles (how much deciles from the RH distribution need to be shifted to match those of the LH distribution). The first and fourth deciles significantly differed in the two groups.