Distributional analyses in the picture–word interference paradigm: Exploring the semantic interference and the distractor frequency effects

Abstract

The present study explores the distributional features of two important effects within the picture–word interference paradigm: the semantic interference and the distractor frequency effects. These two effects display different and specific distributional profiles. Semantic interference appears greatly reduced in faster response times, while it reaches its full magnitude only in slower responses. This can be interpreted as a sign of fluctuant attentional efficiency in resolving response conflict. In contrast, the distractor frequency effect is mediated mainly by a distributional shift, with low-frequency distractors uniformly shifting reaction time distribution towards a slower range of latencies. This finding fits with the idea that distractor frequency exerts its effect by modulating the point in time in which operations required to discard the distractor can start. Taken together, these results are congruent with current theoretical accounts of both the semantic interference and distractor frequency effects. Critically, distributional analyses highlight and further describe the different cognitive dynamics underlying these two effects, suggesting that this analytical tool is able to offer important insights about lexical access during speech production.

Keywords:

• Picture–word interference
• Distributional analyses
• Semantic interference
• Distractor frequency

We thank Giovanni Nascimben, Federica Pasqualetto, and Lisa Sandri for their help in running the experiments. We are grateful to Michael Chan-Reynolds for his comments on a first version of this article.

Notes

¹We would argue that, although the intuitive appeal and the descriptive power of quantile-analyses are clear, this sort of analysis relies on larger degrees of arbitrariness than the parametrical approach. For example, the number of quantiles in which RTs are partitioned is selected by the researchers. This choice is sometimes theoretically motivated (e.g., Thomas, Neely, & O'Connor, 2012). Nonetheless, it may have an impact on the statistical results. In the present study, in line with other seminal works (e.g., Balota et al., 2008), we decided a priori to use deciles, which ensured a detailed description of the RTs distribution.

²In some trials, involuntarily produced noises activated the headset microphone ahead of time, thus terminating the presentation of the visual stimulus. Most of the times, this was unnoticeable since it occurred in a very close temporal proximity to the start of the utterance. For these trials, the onset latency was simply corrected offline using the automatic onset detection procedure implemented in the Check Vocal software. In some other instances, however, the voice-key error occurred so early that the response was disrupted, and participants produced errors or simply were unable to give any response. These instances are those reported as voice-key errors.

³In previous experiments, four words (castello–castle, granchio–crab, anatra–duck, zanzara–mosquito) were used both as semantically related/unrelated and as high-/low-frequency distractors. These materials were kept equal in Experiment 3; therefore these words appeared three times for each participant, while other distractors appeared just twice (in the case where they were used as semantically related/unrelated distractors) or once (in the case of high- and low-frequency distractors).