ATTENTION AND FEATURE INTEGRATION 107 earlier one (Treisman et al.,1977),the results are almost identical.The requirement to search for values on two different dimensions instead of on each trial produced no ualitative and almost no quantitative change in performance:neither did the greater heterogeneity of the dis tractors.In both experiments the display was apparently searched spa- tially in parallel whenever targets could be detected on the basis of a single feature,either color or shape.Another important difference be- tween the conjunction and the feature conditions is the difference in the relation betw ositive and negati splays .The slope fo positives is about half the slope for the negatives,suggesting a seria self-ferminating search.In the reature condition.however.the slope ratio is only 1/8,and the function is linear only for the negatives.This suggests that with single feature targets,a qualitatively different process may nediate the Po itive and to n eg If the ta arget present,it is etected automaticlly:if tend to scan the display,although they may not check item by item in the strictly serial way they do in conjunction search. Practice for up to 13 sessions on the same target and distractors pro duced no alit tive cha nges in perfo ma ance in nction search. e in ty,and no se in eit or inter cept after about the seventh session.We had been interested in seeing whether practice could lead to unitization,in the sense of developing a special detector for the conjunction of green and''T,"which could allow a change to parallel search.It is of course possible that longer practice, different stimuli,or a different training method could result in a cha to parallel se rch.The present however sugges that un tion of color and shape is difficult and may be impossible to achieve There may be built-in neural constraints on which dimensions can be unitized in this way. EXPERIMENT II The next experiment explores the relation between the discriminability of the features which define a conjunction and the speed of detecting that coniunction as a target in a display.if each item must be scanned serially in order to determine how its features are conjoined,it should be possible to change the slop relating search time size,by slowi ing the decision displa L features composing each item.Thus by making the tw pes and the two colors I in a conjunction search easier or harder to distinguish,we should be able to change the rate of scanning while re- taining the characteristic serial search pattern of linear slopes and the 2/1 ratio of negative to positive slopes.We compared search for a conjunction target in distractors which were similar to each other(T in X and )and in distractors which differed ma mally from ea ch othe (0 in
ATTENTION AND FEATURE INTEGRATION 107 earlier one (Treisman et al., 1977), the results are almost identical. The requirement to search for values on two different dimensions instead of one on each trial produced no qualitative and almost no quantitative change in performance; neither did the greater heterogeneity of the distractors. In both experiments the display was apparently searched spatially in parallel whenever targets could be detected on the basis of a single feature, either color or shape. Another important difference between the conjunction and the feature conditions is the difference in the relation between positive and negative displays. The slope for conjunction positives is about half the slope for the negatives, suggesting a serial self-terminating search. In the ieature condition, however, the slope ratio is only l/8, and the function is linear only for the negatives. This suggests that with single feature targets, a qualitatively different process may mediate the responses to positive and to negative displays. If the target is present, it is detected automatically; if it is not, subjects tend to scan the display, although they may not check item by item in the strictly serial way they do in conjunction search. Practice for up to 13 sessions on the same target and distracters produced no qualitative changes in performance in conjunction search, no decrease in linearity, and no systematic decrease in either slope or intercept after about the seventh session. We had been interested in seeing whether practice could lead to unitization, in the sense of developing a special detector for the conjunction of green and “T,” which could allow a change to parallel search. It is of course possible that longer practice, different stimuli, or a different training method could result in a change to parallel search. The present experiment, however suggests that unitization of color and shape is difficult and may be impossible to achieve. There may be built-in neural constraints on which dimensions can be unitized in this way. EXPERIMENT II The next experiment explores the relation between the discriminability of the features which define a conjunction and the speed of detecting that conjunction as a target in a display. If each item must be scanned serially in order to determine how its features are conjoined, it should be possible to change the slope relating search time to display size, by slowing the decision about the features composing each item. Thus by making the two shapes and the two colors in a conjunction search easier or harder to distinguish, we should be able to change the rate of scanning while retaining the characteristic serial search pattern of linear slopes and the 2/l ratio of negative to positive slopes. We compared search for a conjunction target in distracters which were similar to each other (TBreen in X,,,,, and Tbiue) and in distracters which differed maximally from each other (Ored in
108 TREISMAN AND GELADE Oreen and Nred).The decisions whether each item had the target color and the target shape should be easier for O versus N and red versus green than for T versus X and green versus blue.(We chose green and blue inks which were very similar to each other.) A secon on w inve tigated this experiment was whethe r the previous results depended on the haphazard spatial arrangement of the items in the display.In this experiment,the letters were arranged in regular matrices of 2 x 2.4 x 4.and 6 x 6.The mean distance of the letters from the fixation point was equated,so that density again covaried with display size,but acuity was again approximately hed for ea Method Subjects.Six subjects(three females and three males)volunteered for the experiment sity of British a session for thei A two-l C ed to a ond timer was ed .4.16.or36it :The l sof2×2.4×4.or6X positions. For the displa ys of I item each of the s in the2×2mati equally often.The 6x 6 display subtended d 12.3 x 9.7:the 4 x 4 matrix subtended 9.7 x 9. and the 2 e mean distanc s from th nxation poin 0n and N and the target was O.In the difficult condition.the distractors were T and X and the target was Tr The target was presented twice in each display position plays of I and 4,in h e ns for displays of (twice Qua Results Figure 3 shows the mean R'Ts in each condition.The details of the linear regres are giv n in Table 2.None of the slo viates significantly from linearity,which accounts for more than 99. e variance display size in every case.The ratio of positive to negative slopes is 0.52 for the easy stimuli and 0.60 for the difficult ones.The slopes in the difficult discrimination are nearly three times larger than those in the easy discrimination,but the linearity and the 2/1 slope ratio is preserved across these large differences.The intercepts do not differ significantly across conditions Error rates were higher in the difficult discrimination condition.Two subjects were dropped from the experiment because they were unable to keep their false-negative errors in the large positive displays in this condi- tion below 30%.For the remaining subjects,errors averaged 5.3%for the difficult discrimination and 2.5%for the easy discrimination.They were not systematically related to display size except that the difficult positive
108 TREISMAN AND GELADE O,,,, and Nred). The decisions whether each item had the target color and the target shape should be easier for 0 versus N and red versus green than for T versus X and green versus blue. (We chose green and blue inks which were very similar to each other.) A second question we investigated in this experiment was whether the previous results depended on the haphazard spatial arrangement of the items in the display. In this experiment, the letters were arranged in regular matrices of 2 x 2, 4 x 4, and 6 x 6. The mean distance of the letters from the fixation point was equated, so that density again covaried with display size, but acuity was again approximately matched for each condition. Method Subjects. Six subjects (three females and three males) volunteered for the experiment which involved a test and re-test session. They were students and employees of the University of British Columbia ages between 16 and 45. They were paid $3.00 a session for their participation. Apparatus. A two-field Cambridge tachistoscope connected to a millisecond timer was used. The stimuli consisted, as before, of white cards with colored letters. Displays contained 1, 4, 16, or 36 items. The letters were arranged in matrices of 2 X 2, 4 x 4, or 6 x 6 positions. For the displays of 1 item each of the positions in the 2 x 2 matrix was used equally often. The 6 x 6 display subtended 12.3 x 9.7”; the 4 x 4 matrix subtended 9.7 x 9.7” and the 2 x 2 matrix subtended 7 x 7”. The mean distance of items from the fixation point was about 4.3” for all displays. Sixteen different cards, of which 8 contained a target, were made for each display size in each condition. In the easy condition, the distracters were O,,,, and Nred and the target was Ored. In the difficult condition, the distracters were Tblue and X,,,, and the target was T,,,,.. The target was presented twice in each display position for the displays of 1 and 4, in half the display positions for displays of 16 (twice in each row and twice in each column), and twice in each 3 x 3 quadrant for the displays of 36. Results Figure 3 shows the mean RTs in each condition. The details of the linear regressions are given in Table 2. None of the slopes deviates significantly from linearity, which accounts for more than 99.8% of the variance due to display size in every case. The ratio of positive to negative slopes is 0.52 for the easy stimuli and 0.60 for the difficult ones. The slopes in the difficult discrimination are nearly three times larger than those in the easy discrimination, but the linearity and the 2/l slope ratio is preserved across these large differences. The intercepts do not differ significantly across conditions. Error rates were higher in the difficult discrimination condition. Two subjects were dropped from the experiment because they were unable to keep their false-negative errors in the large positive displays in this condition below 30%. For the remaining subjects, errors averaged 5.3% for the difficult discrimination and 2.5% for the easy discrimination. They were not systematically related to display size except that the difficult positive
ATTENTION AND FEATURE INTEGRATION 109 EAS 400 4地 614 DISPLAY SIZE FG.3.Search times in Experiment Il. displays of 16 and 36 averaged 5.9 and 20.7%false-negative errors,re- spectively,compared to a mean of 2.2%errors for all other displays Discussion In both conditions we have evidence supporting serial,self-terminating search through the display for the conju ction targets.The slopes are inear and t ositives give mately half the slope of the negatives However,the rates vary dramatically:The more distinctive colors and TABLE 2 Linear Regressions of Search Times against Display Size in Experiment II Perce Slope Intercept is due to linearity Difficult Positives 55.1 453 99.8 discrimination Negatives 92.4 472 99.9 Easy Positives 20.5 437 99.8 discrimination Negatives 39.5 489 99.9
ATTENTION AND FEATURE INTEGRATION 109 DISPLAY SIZE FIG. 3. Search times in Experiment II. displays of 16 and 36 averaged 5.9 and 20.7% false-negative errors, respectively, compared to a mean of 2.2% errors for all other displays. Discussion In both conditions we have evidence supporting serial, self-terminating search through the display for the conjunction targets. The slopes are linear and the positives give approximately half the slope of the negatives. However, the rates vary dramatically: The more distinctive colors and TABLE 2 Linear Regressions of Search Times against Display Size in Experiment II Percentage variance with display size which Slope Intercept is due to linearity Difficult discrimination Easy discrimination Positives 55.1 453 99.8 Negatives 92.4 472 99.9 Positives 20.5 437 99.8 Negatives 39.5 489 99.9
110 TREISMAN AND GELADE shapes allow search to proceed nearly three times as fast as the less distinc- tive.The mean scanning rate of 62 msec per item obtained in the conjunc- tion condition of Experiment I lies between the rates obtained here with the confusable stimuli and with the highly discriminable stimuli.This wide ion in slopes,co with maintained linearity and 2/ ratios,is consistent with the theory,and puts constraints on alternative explanations.For example,we can no longer suppose that search be- comes serial only when it is difficult.The need for focused attention to each item in turn must be induced by something other than overall load. The fact that the int same for the easy and the theory. Experiment I used pseudo-random locations for the targets and dis- tractors.The present experiment extends the conclusions to displays in which the stimuli are arranged in a regular matrix.The serial scan is therefore not induced by any artifact of the locations s selected or by their haphazard arrangement. EXPERIMENT III Experiment III explores an alternative explanation for the difference 0 targets.This attributes c ndition to the centrality of the ta set of distractors:a conjunction target shares one or another feature with every distractor in the display,while each disjunctive feature target shares a feature with only half the distractors(see Fig.4).In this sense,the con- junction targets are more similar to the set of distractors than the feature iepicaadnha targ t of the similarity structure,but using uni- dimensional stimuli in which checking for conjunctions would not be necessary.We compared search times for a single unidimensional target, which was intermediate between two types of distractors on the single relevant dimension,with search times for either of two disjunctive DIAITVE AE COMAUETON ADCET IAIIVE GREEN GREEN GREEN BROWN DYe86r贤R6Y86起LY 0 0 0 0 FiG.4.Similarity relations between the stimuli in Experiments 1 and III
110 TREISMAN AND GELADE shapes allow search to proceed nearly three times as fast as the less distinctive. The mean scanning rate of 62 msec per item obtained in the conjunction condition of Experiment I lies between the rates obtained here with the confusable stimuli and with the highly discriminable stimuli. This wide variation in slopes, combined with maintained linearity and 2/l slope ratios, is consistent with the theory, and puts constraints on alternative explanations. For example, we can no longer suppose that search becomes serial only when it is difficult. The need for focused attention to each item in turn must be induced by something other than overall load. The fact that the intercepts were the same for the easy and the difficult conditions is also consistent with the theory. Experiment I used pseudo-random locations for the targets and distractors. The present experiment extends the conclusions to displays in which the stimuli are arranged in a regular matrix. The serial scan is therefore not induced by any artifact of the locations selected or by their haphazard arrangement. EXPERIMENT III Experiment III explores an alternative explanation for the difference between conjunction and feature targets. This attributes the difficulty of the conjunction condition to the centrality of the target in the set of distracters: a conjunction target shares one or another feature with every distractor in the display, while each disjunctive feature target shares a feature with only half the distracters (see Fig. 4). In this sense, the conjunction targets are more similar to the set of distracters than the feature targets. We replicated this aspect of the similarity structure, but using unidimensional stimuli in which checking for conjunctions would not be necessary. We compared search times for a single unidimensional target, which was intermediate between two types of distracters on the single relevant dimension, with search times for either of two disjunctive Dl.SJUNCTlVE NON- CONJUNCTION NON- DISJUNCTIVE TARGET TARGET TARGET TARGET TARGET GREEN-GREEN- GREEN GROWN BLUE ‘S ‘X’ ‘T ‘T ‘T -- DlSJUNCTlVE NON- MEDIUM NON- DISJUNCTIVE TARGET TARGET TARGET TARGET TARGET - - - 0 FIG. 4. Similarity relations between the stimuli in Experiments I and III