Abstract This experiment tests a prediction of a model of priming proposed by McLeod and Walley (1989) in which interference is caused by inhibitory processes between nodes in semantic memory. This model predicts that interference normally found in a high cue validity condition at long SOAS will be considerably increased when a brief target is masked. This prediction was confirmed.In a priming task it has typically been found that interference associated with a semantically unrelated prime occurs only when the target follows the prime at a long stimulus onset asynchrony (SOA) and subjects pay attention to the prime, using it as a cue to the target (e.g., Neely, 1977). This pattern of priming interference is well accounted for by Posner and Snyder's (1975) two process model of attention.It has sometimes been found, however, that priming interference occurs with short SOA's: McLeod and Walley (1989), for example, found large interference effects with a short SOA in both a high and a low cue validity condition when brief targets were immediately masked, effects not easily explained by the two process model. To explain these effects, we have proposed a model of priming in which nodes corresponding to concepts in semantic memory may be connected by both excitatory and inhibitory links. Inhibitory links may be an essential mechanism of selective attention, filtering out irrelevant activation (Walley & Weiden, 1973), and may also account for priming and other interference effects.When a node in memory is activated, activation tends to spread to semantically related nodes through excitatory links. At the same time, activation of other nodes is inhibited due to inhibitory links. The more strongly a node is activated, the more strongly it will tend to inhibit other nodes. If more than one node is activated, therefore, the more strongly activated node will win out in the competition for response mechanisms because it will inhibit other nodes more strongly than it is itself being inhibited.In a priming task with a short SOA, nodes corresponding to both the prime and the target are activated and compete with each other through these inhibitory processes. Normally, activation of an unrelated prime node does not present serious competition to the target node because, by the time the target is presented, activation of the prime node has already begun to decay. However, when a very brief target is immediately masked, erasing or overwriting the visual image, processing is interrupted before the target node is fully activated, making it susceptible to inhibition generated by the prime node.With a long SOA in a high cue validity condition, if subjects are using the prime as a cue to develop an expectation regarding the target, then the node in memory corresponding to the expected target would be activated. If the target is unrelated, and therefore unexpected, there will be activation of two competing nodes: the node corresponding to the expected target and the node corresponding to the target actually presented. Because these nodes are both strongly activated at the same time, inhibition generated by the expected target node results in the interference effects typically found.We predicted that interference would be considerably increased in this condition if, as in our earlier studies, targets were brief and immediately masked. In this case, our model suggests that activation of the target node would be considerably reduced and it would therefore be affected even more strongly by inhibition generated by the node corresponding to the expected target.In the following experiment, this prediction was tested. The task was a primed lexical decision with an SOA of 800 ms. Cue validity was either high (75% of prime target pairs related), or low (25% of prime target pairs related). Targets were brief, and masked in either block 1 or block 2 of trials. Masking was blocked because in a previous experiment we had found that subjects had extreme difficulty recognizing brief masked targets mixed with brief unmasked targets. …
Read full abstract