Only after the publication of the paper “Foxes, deer, and hedgehogs: The recall of focus alternatives in Vietnamese” did we become aware of the existence of an erratum to Baron Cohen et al.’s (2001) autism spectrum questionnaire. This erratum states the necessity to reverse-code item 1 of the questionnaire. Thus, items that had been coded as 1 by us needed to be re-coded as 0 and vice versa. While this did not alter the overall findings, all numbers from calculations involving the autism scores need to be adjusted a little:

p. 19: Adding participants’ scaled and centered autism scores did improve model fit significantly (𝝌²(1) = 4.85, p = .03). However, adding the interaction between condition and autism scores did not lead to any further improvement (𝝌²(1) < 1, p = .82).

p. 19: Both variables were scaled and centered and neither the interaction of condition by age of acquisition (𝝌²(2) = 1.06, p = .59) nor the interaction of condition by length of residence (𝝌²(2) = 2.60, p = .27) improved model fit.

p. 19: The following adjustments are necessary for Table 6.

Fixed effect estimates and variance estimates for GLMER of recall for list items (recall~condition * gender + autism_score + word type + poly(NumInExpC,2) + (1+gender|item) + (1|word) + (1+poly(NumInExpC,2)|participants), n = 9648, log-likelihood = –5337.6). Coding scheme for gender, condition, and word type = sum coding.

The only difference with the previous model is that the main effect of participant gender is no longer significant but only a trend with a p-value of .07. Thus, with the correct autism scores, slightly more variance seems to be explained by the autism scores and slightly less by gender. Still, the critical focus condition by gender interaction is still significant and adding a focus condition by autism score interaction still did not improve model fit.

p. 20: The following adjustments are necessary for Table 7.

Fixed effect estimates and variance estimates for GLMER of recall for list items for men (recall ~ condition + autism_score + word type + poly(NumInExpC, 2) + (1 | item) + (1 | word) + (1 + poly(NumInExpC, 2) | participants), n = 3312, log-likelihood = –1856.7). Coding scheme for condition and word type = sum coding.

p. 21: The following adjustments are necessary for Table 8.

Fixed effect estimates and variance estimates for GLMER of recall for list items for women (recall~ condition + autism_score + word type + poly(NumInExpC, 2) + (1|item) + (1|word) + (1+poly(NumInExpC, 2)|participants), n = 6336, log-likelihood = –3534.4). Coding scheme for condition and word type = sum coding.

p. 22: First, we ran the statistical model on this reduced data set and still observed the relevant effects: an interaction of condition with participant gender (B = 0.25, |z| = 2.21, p = .03), and main effect of gender (B = 0.45, |z| = 2.25, p = .02), word type (B = – 1.23, |z| = 9.64, p < .001), and autism score (B = 0.25, |z| = 2.77, p < .01). The linear component of the predictor for trial number was also significant (B = 31.55, |z| = 3.05, p < .01). We tested for an influence of German proficiency by adding an interaction term for German proficiency and focus condition. This did not improve model fit (𝝌²(2) < 1, p = .94). We tested for an effect of Vietnamese attrition by adding Vietnamese proficiency to the model, which, again, did not improve model fit (𝝌²(1) = 2.16, p = .14).