Turning selective attention failure into selective attention success.

Abstract In a letter identification task in which a centrally - presented letter is to be attended and laterally - flanking letters ignored, B.A. Eriksen and C.W. Eriksen (1974) have shown effects on performance of the to - be - ignor ed letters when they contain information about the correct response. These data constitute important evidence of selection for attention on condition that the flankers are not, in fact, attended. Three experiments assess the importance of (1) practice on the task; (2) positio nal uncertainty regarding the flankers; (3) positional uncertainty regarding the target; (4) the proportion of flankers in a display which contain information about the response; and (5) the strength of the target - response mapping. The effect of flankers on performance is found to be independent of target position uncertainty and strength of target - response mapping, but to vary with practice on the task, flanker - position uncertainty, and proportion of informative flank ers. The data are interpreted as indicating that people can adjust their strategy: They are capab le of both and selection under different conditions.Resume Au cours d'un exercice d'indication d'une lettre, ou il faut faire a ttention a la lettre centrale et ne pas se preoccuper des lettres secondaires qui la co@t oient, B.A. Eriksen et C.W. Eriksen (1974) ont demontre les effets qu'ont, sur le rendement, les lettres de chaque co@te lorsqu'elles renferment des renseignements lies a la bon ne reponse. Ces donnees representent une preuve importante d'attention tardive par rapport au choix de la lettre en question, a condition qu'il ne faille pas faire attention aux lettres de co@te. Trois experiences evaluent l'importance (1) de l'habitude de l'exercice; (2) de l'incertitude quant a la position des lettres de co@te; (3) d e l'incertitude quant a la position de la lettre cible; (4) de la proportion, dans la presentati on, des lettres de co@te qui renferment des renseignements sur la reponse; et (5) de l'intensite de la mise en correspondance de la cible et de la reponse. Les resultats indiquent qu e l'effet des lettres de co@te sur le rendement ne varie ni en fonction de l'incertitude q uant a la position de la cible ni en fonction de l'intensite de la mise en correspondance de la cible et de la reponse, mais de l'habitude de l'exercice, de l'incertitude quant a la pos ition des lettres secondaires et de la proportion de lettres secondaires pertinentes. L'i nterpretation des donnees suggere que les gens peuvent modifier leur strategie: ils sont capables d'effectuer une selection precoce et une selection tardive, en fonction des cond itions.There is general acceptance of the idea that some stimulus representations can be produced by processes which operate in parallel, while ot hers can only be produced by serial processes. However, there is widespread disagreement over exactly which representations are produced by each type of process (Broadbent, 1982; Johnston & Dark, 1986; Cowan, 1988; Laberge & Brown, 1989). The basic idea that some representations a re produced only by serial processes leads to the class of models called models, th e original of which is Broadbent's (1958) theory. Early models hold that select ion on the basis of physical data is generally efficient. That is, of the stimuli which are availab le to sense receptors, most can be excluded from full analysis, allowing analysis to be focu sed on a single stimulus. Broadbent (1982) still holds the position, a view advanc ed also by Egeth (1977), Kahneman and Treisman (1984), and Johnston and Dark (1986). A variant a llowing some abstract feature detection in parallel, but only when the features are simp le, has been advanced by Treisman (Treisman & Gelade, 1980; Treisman & Souther, 1985).The view just outlined is traditionally known as early filter theory. An alternative to theory is the late filter class of models (Deutsch & Deutsch, 1963; Nor man, 1968; Allport, 1980; Duncan, 1980, 1987; Duncan & Humphreys, 1989; Allport, Tipper, & Chmiel, 1985). …