The masked semantic priming effect is task dependent: Reconsidering the automatic spreading activation process.

Abstract

Semantic priming effects are popularly explained in terms of an automatic spreading activation process, according to which the activation of a node in a semantic network spreads automatically to interconnected nodes, preactivating a semantically related word. It is expected from this account that semantic priming effects should be routinely observed when the prime identity is veiled from conscious awareness, but the extant literature on masked semantic priming effects is notoriously mixed. The authors use the same prime-target pairs in the lexical decision task and the semantic categorization task and show that although masking the prime eliminates the semantic priming effect in lexical decision, reliable semantic priming effects are observed with both masked and unmasked primes in the semantic categorization task. The authors explain this task dependence in terms of their account of semantic priming effects based on notions of evidence accumulation and source confusion and support their account by means of reaction time distribution analyses.

Evidence accumulation in the integrated and primed Stroop tasks

“…Evidence subsequently accumulated from the target therefore starts from a different baseline than it would in the absence of a prime, but the rate of evidence accumulation remains unchanged. In effect, the prime gives the target a head start, and this was the explanation suggested by de Wit and Kinoshita (2014, 2015a, 2015b) for the flat-delta-plot pattern seen with the semantic-priming effects in semantic categori-zation. (As we will see shortly, a head start and lowering of the threshold can produce quantile plots that are indistinguishable.…”

“…The discrepancy is also unexpected according to the view that both the Stroop effect and the semantic-priming effect found with a short prime–target stimulus onset asynchrony (SOA), as in the studies described above (de Wit & Kinoshita, 2014, 2015a, 2015b; Voss et al, 2013), reflect an automatic semantic process. Since most previous studies using the Stroop task and the semantic categorization task have not examined the congruence effect at the level of RT distributions, this question has not arisen.…”

“…In addition, de Wit and Kinoshita (2014) showed that increasing the proportion of semantically related trials magnified the size of the semantic-priming effect and that this increase was constant across the range of response latencies (affecting only the μ parameter); that is, it increased the amount of distributional shift. De Wit and Kinoshita (2015a) then showed that the distributional shift pattern of semantic-priming effects is also found when the primes are presented very briefly and (forward- and) backward-masked, so that the participants are unaware of its identity. Kinoshita and Hunt (2008) had earlier reported this distributional-shift pattern with a number classification (“bigger than 5?”) task.…”

“…The interpretation offered by de Wit and Kinoshita (2014, 2015a, 2015b) of the distributional-shift pattern is that the semantically related, category-congruent prime provides a head start in the categorization process. In contrast to the tra-ditional “activation” view of semantic priming, in which the activation from the prime spreads and preactivates the target, priming effects here are explained in terms of an evidence accumulation process in which the information accumulated from the prime is combined with that accumulated from the target.…”

The onset and time course of semantic priming during rapid recognition of visual words.

“…De Wit & Kinoshita, 2015a b). Instead, these results are consistent with a limited set of previous findings showing greater priming for slower responses in both LDT and naming tasks (Thomas et al 2012; Balota et al 2008, for visually-degraded targets only).…”