Word superiority in signal detection: Barely a glimpse, yet reading nonetheless

In Part 1 data from a number of visual masking studies have been reanalyzed to show that, contrary to the conclusions usually drawn from this data, there is statistical evidence of semantic activation at short target-mask stimulus onset asynchrony levels (SOA) and that for levels of performance close to chance, semantic information extracted from briefly presented targets may be used to detect those targets. In Part 2 this hypothesis is tested directly in three experiments. A word superiority effect is found in a signal detection task, showing unequivocally that (i) high-level activation exists at short SOAs and (ii) such activation can affect levels of target detectability. The results undermine the use of signal detectability as a measure of awareness