Relating emotional variables to recognition memory performance: a large-scale re-analysis of megastudy data

The megastudy paradigm has become an important tool for cognitive science. One advantage to the megastudy is that existing data can be reanalysed in light of novel hypotheses. In the current study, recognition memory data for 4819 words were obtained. Multiple regression analyses assessed the influence of emotional variables on recognition memory performance (i.e., hits minus false alarm rates H-FAs) for the words. The predictor variables included valence, arousal, extremity of valence (the degree of negative or positive meaning), context valence (the degree to which a word typically appears in positive or negative contexts), context arousal (how emotionally reactive are contexts in which the word appears), and context extremity of valence (the degree of this typical emotional context). This study extended earlier work by implementing more thorough controls, maximising the number of words, assessing a more comprehensive set of emotional variables, and introducing the context extremity of valence variable. We found extremity of valence, context extremity of valence, context valence, and context arousal all were significant predictors of H-FAs. We interpret the results in terms of the dual-coding theory and hub and spoke model. We also explain how single-process models could accommodate the results in terms of context diversity.