A neuropsychological theory of attention is described in which the encoding of one stimulus interferes with the encoding of other stimuli. This interference effect is termed cognitive masking and is attributed to recurrent lateral inhibition between neurons in association cortex. Evidence is reviewed which indicates that there is a facilitation of cortical recurrent inhibition during arousal, and it is suggested that cognitive masking is related to the level of arousal. A mathematical model is presented which describes the activity of sets of cortical neurons having reciprocal inhibitory interconnections. The results of some computer simulations of this model indicate that the model can account for cognitive masking and several other attentional effects. Implications of the theory for a variety of problems are discussed.

In a spatial version of the Stroop task where subjects must attend to the position of a word and ignore the word itself, we argue that interference results because selective attention is only partially successful. If so, then increasing attention to the word should increase interference. Experiment 1 showed increased interference with position naming when subjects must be prepared to attend to either the word or its position. Experiment 2 required subjects to first name the position of the word and then later to recall which word had been presented. This resulted in increased interference with position naming when the word was followed at a short SOA by a pattern mask. Experiment 3 employed both word reading and position naming and showed that a large reverse Stroop effect occurred with both vocal and manual responses when subjects had to be prepared to attend to either dimension of the stimulus.

No abstract

Rabbit eyeblinks were conditioned to either the visual or the nonvisual element of a compound conditioned stimulus. Evoked potentials (EPs) were simultaneously recorded from the visual cortex. Changes in the late negative component of the EP were related to nonspecific arousal effects. The initial surface positive component, however, was significantly enhanced during visual stimulus control. This enhancement was observed whether the visual stimulus was presented alone or in compound and with either photic or optic chiasma stimulation as the visual stimulus. The effect could not be attributed to arousal, sensitization, or to peripheral mechanisms. Some implications of the data were assessed through a subsequent investigation of geniculostriate recovery cycles under the same conditions.

Priming effects were examined in two experiments using either a pronunciation or lexical decision task. The prime, cither a strong associate of the target, an unrelated word, or a neutral prime, was presented for 200 ms. After an SOA ol 200, 400, or 8(K) ms, a masked target was presented for 33.3, 50, or 66.7 ms. Attention was manipulated by varying the probability that prime and target would be strongly associated. Both experiments showed significant interference in the low attention condition and at the 200-ms SOA, presumably before the onset of consciously directed processing. Two subsequent experiments using a short SOA and the low attention condition attempted to determine the conditions under which this interference will occur by varying the interstimulus interval, target duration, and the mask. It was found that interference occurred only when targets were brief and masked. These results are discussed in terms of a model involving lateral inhibition between nodes in semantic memory. It is suggested that when the target is brief and masked, the node in memory corresponding to the target is much less strongly activated and therefore more susceptible to the inhibitory effects of other activated nodes.RKSUMK On a examine dans deux experiences les effets d'amorcage a l'aide d'unc tache de decision lexicale ou de pronunciation. L'amorce, cjui elail soil une forte association de la cible, soil un mot non relie, ou une amorce neutre etait presentee pendant 200 ins. Apres un SOA de 200, 400, ou 800 ms, une cible masqucc ctait presentee pendant 33.3, 50, ou 66.7 ms. L'atiention etait manipulec en variant la probability que l'amorce et la cible soil fortement associees. Les deux experiences ont montrc une interference significative dans la condition ou I'attcntion ctait basse ct au SOA de 200 ms, prcsumement avant le declenehement d'un processing dirige consciemment. Deux experiences subscquentcs utilisant un SOA court el la condition d'aiiemion basse lemaient de determiner les conditions dans lesquellcs survicndraint cette interference en variant I'intcrvallc cntre les stimuli, la durcc dc la cible ct Ic masque. I/interference survenait seulemcnt quand les cibles elaienl presentees brievement ct avee Ic masque. Les rcsultats sont diseutcs en fonction d'un modelc impliquant 1'inhibition latcralc cntrc les nocuds dans la memoire semanlique. Nous suggcrons que lorsque la cible est breve ct masquees, le noeud dans la memoire corrcspondant a la cible est active beaucoup moins fortement et est ainsi plus sueeptiblc aux effcts inhibitcurs d'autres nocuds actives.