Analyzing Mental Image Generation Data through Imputation: Evidence of a Visual and an Episodic Buffer for Image-Size Scaling

We reanalyzed data reported in previous mental imagery experiments (D’Angiulli, 2001; D’Angiulli & Reeves, 2002). The data consisted of self-reported image vividness ratings (1 = no image; 7 = perfectly vivid) and image latencies (image completion RTs) for “small” (i.e., subtending < 16) and “large” (i.e., > 16) images, in two image generation conditions: trial-unique and repeated. In the original experiments, the seven vividness levels were not used equally frequently. This produced “planned” missing values. Consequently, we investigated whether filling-in the missing data through regression modeling (imputation) would allow us to frame current and previous findings into a new conception of mental imagery and its architecture.