Corrigendum to “Parallels between self-monitoring for speech errors and identification of the misspoken segments” [J. Mem. Lang. 69(3) (2013) 417-428]

Abstract

In Table 4, a coefficient of +0.4091 (s.e. 0.1553) was reported for the categorical predictor “Undetected vs Detected”. The contrasts of this predictor were coded as −1 for undetected speech errors, +0.5 for errors detected early, and +0.5 for errors detected late (p.422). On p.422 it was correctly stated that “[the] significant second contrast confirms that misidentification occurs more often in segmental errors detected by the speaker than in undetected errors (β = 0.4091, odds ratio 1.51, p = .0085)”. This statement matches the misidentification rates reported in Table 3 (undetected errors 3%, early detected errors 4%, late detected errors 6%). Later in the article, however, this effect was discussed as if it were in the opposite direction, and it was stated erroneously that “. undetected errors suffer significantly more from misidentification than detected errors” (p.425). The authors apologize for this error and for any inconvenience caused. As argued on p.419 (prediction 2), the observed effect, with higher odds of consonant misidentification of detected speech errors than of undetected speech errors, supports a conflict-based account of self-monitoring during speech production. The authors thank Holger Mitterer for noticing this error.