Automatic semantic activation of embedded words: Is there a “hat” in “that”?

Participants semantically categorized target words that contain subsets (Experiment 1; e.g., target = hatch, subset = hat) or that are parts of supersets (Experiment 2; e.g., target = bee, superset = beer). In both experiments, the targets were categorized in a congruent condition (in which the subset-superset was associated with the same response, e.g., Does hatch refer to a human body part?) and an incongruent condition (in which the subset-superset was associated with a conflicting response, e.g., Does hatch refer to a piece of clothing?). Responses were slower and less accurate in the incongruent conditions, suggesting that subsets and supersets were processed to the level of meaning. Congruency effects occurred regardless of the position of the subset or superset (e.g., hatch, drama, howl), and in Experiment 1, were obtained for subsets that maintained (e.g., card) and changed their pronunciation (e.g., crown). Congruency effects were only found when the subsets were of higher frequency than the target. The implications for theories of word identification are discussed.