Abstract
Sanguinetti & Peterson (VSS, 2013) found that silhouettes semantically related to preceding masked words can “unmask” the words. Participants viewed masked words (50ms). 60ms later, a masked silhouette (Familiar or Novel) was displayed (175ms). Familiar silhouettes depicted a recognizable, nameable object. Novel silhouettes were unnamable objects. Familiar silhouettes were named by the masked words on half the trials (Match) and not on the rest (Mismatch). Participants indicated, on each trial, whether they saw a word. Participants saw more words on Familiar Match trials than other trials (40%vs.20%, p< .05). Participants also categorized the silhouettes (novel/familiar); accuracy = 77%. Sanguinetti & Peterson hypothesized that word presentation triggers a hypothesis of the word’s presence that is disconfirmed by the mask, preventing its conscious awareness. The semantically-related silhouette unmasks the word because the semantic relation serves as confirmation of the word. An alternative explanation is that the word is perceived when presented but immediately forgotten or degraded due to the mask; conscious recognition of the Familiar silhouettes simply facilitates memory of the word. We present evidence against this alternative explanation. The memory explanation depends on conscious recognition of the Familiar silhouettes. We shortened the silhouette exposure to 70ms, reducing iterative processing of the silhouette, but not recognition. Participants saw fewer words on Familiar Match trials (22%, p< .01), suggesting that unmasking strength depends on silhouette duration; recognition is insufficient. A separate study showed the silhouettes were indeed recognizable with a 70ms exposure. The same display sequence was used except participants categorized the silhouettes (novel/familiar). Accuracy was 81%, equivalent to Sanguinetti & Peterson p>.05). Our results show that the memory explanation is incorrect, and demonstrate that the unmasking effect requires a longer silhouette exposure, perhaps because longer exposures allow more iterative processing and therefore more access to the silhouette semantics necessary to unmask the word.
Meeting abstract presented at VSS 2015