Opening a File Drawer: Surprisingly Robust Evidence for Semantic Interference from Distractor Pictures in Picture Naming

Semantic-categorically related distractor words slow down picture naming more than unrelated distractor words do (e.g., “dog” vs. “car” for the picture of a horse). This semantic interference effect in the picture-word interference task has been replicated numerous times and is an empirical cornerstone of speech production research. However, picture-picture interference studies in which the distractor words were replaced with distractor pictures have yielded highly inconsistent results, reporting semantic interference, semantic facilitation, and null effects. We present a meta-analytic summary of the published studies. We found the estimated semantic effect to be close to zero (2 ms, 95% CI [-5, 9]). This contrasts with the results of 14 previously unpublished picture-picture interference experiments from our laboratory. Across these experiments, the estimated semantic interference effect was 15 ms (95% CI [9, 21]). When all data were included in a meta-analysis, the pooled estimate was 9 ms (95% CI [4, 14]). This suggests that the magnitude of semantic interference from distractor pictures has been underestimated and that it is not as elusive as previously thought. We conclude that there is no qualitative difference in the semantic effects caused by distractor words and pictures, and we discuss the implications for modelling word production.