Perceptual d n d Motor Skills, 1975,41, 755-760. @ Perceptual and Motor Skills 1975
EFFECT OF TARGET SEPARATION O N SELECTIVE ATTENTION
Arlstralian National Ufiiversity Summasy.-A visual search task, in which subjects searched circular stimulus displays for two instances of a prespecified target, was used to investigate the effects of target-separation on accuracy. When a comparison was made of the total number of targets correctly located at each separation, no significant differences were found, and this suggested that, within the range of separations (maximum of 2.53") examined, the selective processing of the relevant items was not influenced by the distance separating them. Also, assuming that the redundancy of cargec items increased the probability of a target being perceived, the differences between the number of first targets correctly located and the number correct in a single target condition, were in the predicted direction. However, they failed to reach significance.
Over the past decade or so, some considerable interest has been displayed in the spatial factors affecting visual information processing. For example, ( 1 ) Volkmann, Corbin, Eddy, and Coonley (1964) and Vollunann and Corbin (1965) using visual search, obtained results indicating an ovaloid area around the fixation point as an area of fast search. ( 2 ) Mackworth (1965) spoke of a useful field of view, as the area around the retina where information was processed exhaustively. ( 3 ) Flom, Weymouth, and Kahneman ( 1963), Eriksen and Rohrbaugh (1970) and Strangert and Brannsuom (1973) obtained results indicating that contour interference occurred in foveal vision, if letters were separated by less than approximately 0.33' of visual angle. ( 4 ) Eriksen and Hoffman (1972) suggested that the ability of the attention mechanism to focus has a lower limit of approximately 1' of visual angle. ( 5 ) The effect of distance from the fovea has been fairly well explored (cf. Strangert & Brannstrom, 1973; Collins, 1973). However, whether the distance separating two relevant pieces of information (or items) has an effect on the selection process has not been systematically investigated. If a task requires that, out of a number of items only two are to be processed and if the brevity of the stimulus display negates the possibility of more than one fixation, then how does the attentive process accomplish this selection? Given that there are clear physical differences between the relevant and irrelevant items, the preattentive mechanism (Neisser, 1967) should direct focal processing resources onto the relevant items. However, since focal processing is carried out in serial (Neisser, 1967; Eriksen & Colegate, 1971), one of the items will be processed before the other and it is of some interest to decide if 'Requests for reprints should be addressed to R. T. Solman C/-Department of Psychology, School of General Studies, The Australian National University, Box 4, P.O. Canberra, A.C.T. 2600.
the distance separating the targets has any influence on the manner in which they are processed. The study by Eriksen and Colegate ( 1971), where subjects reported two indicated items separated by varying distances, could have been relevant to this issue. However, no attempt was made to extract such information. Also, with the possible exception of the simultaneous indicator condition, a probe technique might not be expected to encourage preattentive selection. In an effort to facilitate initial selection, the present study employed a visual search task where the identity of the target instances did not change. As well as examining the effect of target separation on selective attention, the design of the present study makes it possible to examine the effect of target redundancy on accuracy. The results of previous studies concerned with the effects of redundant critical items have been equivocal, e.g., earlier studies by Wolford, Wessel, and Estes (1968) and Bjork and Estes ( 1971) showed that an increase in the number of critical elements failed to reduce reaction times significantly, and later studies by Kinsbourne and Innis (1972), Estes ( 1972), and Van Der Heijden and Menckenberg (1974) showed that increases in the number of target instances reduced reaction times and increased the percentage of correct detections. In the study reported here, the presence of two target instances should increase the probability that a target will be perceived. It would therefore be predicted that, when corrected for chance, the percentage of correct responses to the first target processed would be greater than the percentage of correct responses in a single-target case.
METHOD The task was visual search, and throughout the experiment a trial consisted of the following steps. After placing a stimulus card in a tachistoscope, E gave a verbal ready signal upon which S fixated a cross and pressed a hand switch. The switch simultaneously removed the cross and initiated a display, which remained present for a pre-set period (20, 40, or 80 msec. for experimental sessions), before being replaced by a 50-msec. mask. The fixation cross returned upon termination of the mask. At the completion of each trial S recorded his result. This was done by marking the positions of the targets, placing a confidence rating beside each, and indicating which target was located first or i f they were both located at the same time. Design The design was three-way ( 2 X 3 X 6 ) with repeated measures on the last rwo factors. The independent variables were, firstly, the order of presenting the display exposures (either 20, 40, 80 or 80, 40, 2 0 ) , secondly display exposure, and thirdly the distance separating the two targets (0.67, 1.27, 1.80, 2.20, 2.43, and 2.53' of visual angle). c- were 4 undergraduates ( 2 males, 2 females) with normal 6/6 vision,
TARGET SEPARATION AND SELECTIVE ATTENTION
757
who received $1 per hour as payment for taking part in the experiment. TWO Ss searched displays where the exposures were ordered from shortest to longest, while the remaining two received the exposures in the reverse order. All SS searched for 2 practice and 9 experimental sessions, each of d ~ r a t i o napproximately 1 hr. At the completion of each trial they recorded the positions of the targets, indicated which was first to be located or if they were both located together, and rated each on a four-point confidence scale ( 4 = sure, 3 = almost sure, 2 = might have been, and 1 = guessing). A target was correctly located if the correct position or one either side was marked; the reasons for adopting this scoring method were outlined by Solman (1974).
Materials and Apparatus The letter F was the target item and G , 0, Q, U, and W were selected as irrelevant items. The stimuli were displayed in a Model G.B. Scientific Prototype Three-channel Tachistoscope. The luminance of the fixation, stimulus and mask channels was 10.8, 17.2, and 38.7 mL respectively. Stimulus displays were constructed by mounting black Letraset Upper Case letters ( 18 pt Grotesque 216) on square white cards of side length 203 mm. Each display contained 12 items positioned at the 12 clock positions on an imaginary circle, the diameter of which subtended a visual angle of 2.53". There were 36 displays constructed containing two Fs, i.e., given the restriction that over the 36 displays each of the 12 positions was equally likely to contain an F, a random selection of 6 of the 12 possible target arrangements was made for each separation ( 6 X 6 = 36). Twenty-four single-target displays were constructed with the target appearing twice at each position. When positioned each display item was 0.27' high with a stroke width of 0.06". The fixation card had a black cross mounted at the centre (arms and stroke width 0.43' and 0.06' respectively) and the mask consisted of a collection of broken, distorted, and whole Letraset letters of the same kind used in the displays. T o allow Ss to record their results sheets were constructed containing six circles with the 12 clock positions numbered 1 to 12. Procedure Upon arrival at the first practice session Ss were told that the task was visual search, the target appeared once on some cards and twice on others, the fixation cross should be in clear focus before the initiation of the display, and the confidence ratings referred to target presence. They were also informed that prior knowledge would be given concerning the number of targets in a display, and that a response, including a confidence rating, must be made to all targets. They were then familiarized with the displays and given 54 practice trials with those containing two targets ( 1 8 at each exposure) and 9 practice trials with the single-target displays ( 3 at each exposure), using exposures of 40, 80, and
160 msec. Two Ss received the exposues in ascending and two in descending order, and this order was maintained throughout the study. The order of presenting the single- and double-target displays was counterbalanced across subject groups. The same procedure was followed for the second practice session, using the experimental exposures (20, 40, and 80 msec.). After completion of practice Ss attended 9 experimental sessions. During the first 8 they searched 6 practice trials (2 at each exposure) before both the double- and single-target conditions and 126 experimental trials. For the doubletarget condition 36 displays were given at each exposure, and for the single-target condition 6. During Session 9 only 42 experimental trials were given (12 at each exposure for the double-target condition and 2 for the single target), which made up a total of 50 responses for each experimental condition.
RESULTSAND DISCUSSION A three-way analysis of variance for stimulus exposure presentation order, stimulus exposure, and the distance separating the two targets, showed that only the main effect of exposure and the interaction between exposure and separation reached statistical significance (F2,4 = 40.72, P < 0.01; PI0,20 = 6.49, P < 0.001). Since order of presentation of the stimulus exposures was not a significant variable, it has been excluded from the results. Table 1 shows the mean number of correctly located targets for the 18 experimental conditions given by the 3 exposures and the 6 distances separating the targets. TABLE 1 MEANNUMBER OF CORRECTLY LOCATED TARGET INSTANCES FOR 18 CONDITIONS GIVEN BY THREEEXPOSURESAND SIX DISTANCES SEPARATTNGTWO OF TARGET INSTANCES Separation (degrees of visual angle) 0.67 1.27 1.80 2.20 2.43 2.53
20
Stimulus Exposure (in msec.) 40
80
62.00 54.75 56.50 54.25 51.00 56.75
70.25 75.00 73.00 64.25 74.50 72.25
76.50 80.50 74.25 75.50 74.50 76.00
The failure of the distance separating the targets to have a significant effect on the total numher located ( F 5 . , ~= 2.35, P > 0.05) suggested that their selection for processing was independent of this variable. In terms of the analogy about a torch beam (Eriksen & Hoffman, 1972), the distance which must be travelled by the beam of processing capacicy as it moves from one target to the other, does not influence the rate and detail of the processing. At first this seems surprising, since it might be expected that moving processing capacity - -
TARGET SEPARATION AND SELECTIVE ATTENTION
759
from one target to the next would require time and that accuracy levels would fall as the distance moved and the time required increased. However, it is not necessary to conclude that the spatial separation of targets in the stimulus environment is represented the same way during processing. An explanation of the significant interaction between target separation and exposure (F10,20= 6.49, P < 0.001) was not obvious from the data. Observation of Table 1 showed that performance at separation 0.67' for 20 msec. and 2.20" for 40 msec. did not conform to the general pattern of the results. Accepting the 1' limit on focussing (Eriksen & Hoffman, 1972), it might be expected that performance would be superior at a separation of 0.67' for all exposures. Why this superiority was only found for an exposure of 20 msec. is not clear. The low level of performance at separation 2.20° for the 40-msec. exposure eludes theoretical interpretation. The significant increase in accuracy with increasing exposure was consistent with the results obtained by Solman (1974), i.e., in the earlier study the results showed a rapid increase in accuracy as exposure increased from 15 to 50 msec., with a tendency toward stable performance for further increases up to 180 msec. In the present study the accuracy changes were greater from 20 to 40 msec. than from 40 to 80 msec. In order to assess the effect of a redundancy of target instances, the data for the first target located were averaged over separation distances and compared with data for a single-target condition. To make the comparison meaningful both sets of results were corrected for chance performance.* The mean number of correct responses for two-target and one-target conditions (original and data corrected for chance), for each of the three levels of stimulus exposure, are shown in Table 2. Although the differences observed in the corrected data ( a higher level of performance for the two-target condition) suggested that target
TABLE 2 MEAN NUMBEROF
CORRECTLYLOCATED FIRSTTARGETS FORm0-TARGET CONDITION A N D SINGLE-TARGET CONDITION -
Stimulus Exposure (msec.)
Note.-Both
Corrected Data T w o targets One target
20 13.20 11.67 40 31.96 28.00 80 36.63 35.00 original and data corrected for chance are shown.
-
original Data T w o targets One target 30.83 42.08 43.50
21.25 33.50 38.75
T h e data for the single-target condition were corrected by subtracting one third of the incorrect responses from the correct responses for each subject, and then calculating the mean for all subjects (see Solman, 1974, Footnote 5 ) . The correction for the data o n the first target located in the two-target condition was carried out by subtracting the single-target factor multiplied by the value of the ratio of chance performance in the two conditions.
760
R. T.SOLMAN
redundancy increased the probability of a target being perceived, an analysis of variance showed that they did not reach significance (F1,2 = 5.57, P 0.15). Why should redundancy of targets increase the probability that a target will be perceived? The salience of the target features in the icon might be expected to determine the probability with which focal processing is directed to these features, and if there are two instances of the one target, this probability should be greater than in the single-target case. In the present study the number of instances of the target item only varied from one to two in a 12-item display, and this increase may not have been sufficient to raise the probability of perceiving a target significantly. Also, since subjects were required to report both targets, there may have been some trade-off in processing efficiency which increased performance for the second target, while slightly decreasing performance for the first.
EFFECT OF TARGET SEPARATION O N SELECTIVE ATTENTION
Arlstralian National Ufiiversity Summasy.-A visual search task, in which subjects searched circular stimulus displays for two instances of a prespecified target, was used to investigate the effects of target-separation on accuracy. When a comparison was made of the total number of targets correctly located at each separation, no significant differences were found, and this suggested that, within the range of separations (maximum of 2.53") examined, the selective processing of the relevant items was not influenced by the distance separating them. Also, assuming that the redundancy of cargec items increased the probability of a target being perceived, the differences between the number of first targets correctly located and the number correct in a single target condition, were in the predicted direction. However, they failed to reach significance.
Over the past decade or so, some considerable interest has been displayed in the spatial factors affecting visual information processing. For example, ( 1 ) Volkmann, Corbin, Eddy, and Coonley (1964) and Vollunann and Corbin (1965) using visual search, obtained results indicating an ovaloid area around the fixation point as an area of fast search. ( 2 ) Mackworth (1965) spoke of a useful field of view, as the area around the retina where information was processed exhaustively. ( 3 ) Flom, Weymouth, and Kahneman ( 1963), Eriksen and Rohrbaugh (1970) and Strangert and Brannsuom (1973) obtained results indicating that contour interference occurred in foveal vision, if letters were separated by less than approximately 0.33' of visual angle. ( 4 ) Eriksen and Hoffman (1972) suggested that the ability of the attention mechanism to focus has a lower limit of approximately 1' of visual angle. ( 5 ) The effect of distance from the fovea has been fairly well explored (cf. Strangert & Brannstrom, 1973; Collins, 1973). However, whether the distance separating two relevant pieces of information (or items) has an effect on the selection process has not been systematically investigated. If a task requires that, out of a number of items only two are to be processed and if the brevity of the stimulus display negates the possibility of more than one fixation, then how does the attentive process accomplish this selection? Given that there are clear physical differences between the relevant and irrelevant items, the preattentive mechanism (Neisser, 1967) should direct focal processing resources onto the relevant items. However, since focal processing is carried out in serial (Neisser, 1967; Eriksen & Colegate, 1971), one of the items will be processed before the other and it is of some interest to decide if 'Requests for reprints should be addressed to R. T. Solman C/-Department of Psychology, School of General Studies, The Australian National University, Box 4, P.O. Canberra, A.C.T. 2600.
the distance separating the targets has any influence on the manner in which they are processed. The study by Eriksen and Colegate ( 1971), where subjects reported two indicated items separated by varying distances, could have been relevant to this issue. However, no attempt was made to extract such information. Also, with the possible exception of the simultaneous indicator condition, a probe technique might not be expected to encourage preattentive selection. In an effort to facilitate initial selection, the present study employed a visual search task where the identity of the target instances did not change. As well as examining the effect of target separation on selective attention, the design of the present study makes it possible to examine the effect of target redundancy on accuracy. The results of previous studies concerned with the effects of redundant critical items have been equivocal, e.g., earlier studies by Wolford, Wessel, and Estes (1968) and Bjork and Estes ( 1971) showed that an increase in the number of critical elements failed to reduce reaction times significantly, and later studies by Kinsbourne and Innis (1972), Estes ( 1972), and Van Der Heijden and Menckenberg (1974) showed that increases in the number of target instances reduced reaction times and increased the percentage of correct detections. In the study reported here, the presence of two target instances should increase the probability that a target will be perceived. It would therefore be predicted that, when corrected for chance, the percentage of correct responses to the first target processed would be greater than the percentage of correct responses in a single-target case.
METHOD The task was visual search, and throughout the experiment a trial consisted of the following steps. After placing a stimulus card in a tachistoscope, E gave a verbal ready signal upon which S fixated a cross and pressed a hand switch. The switch simultaneously removed the cross and initiated a display, which remained present for a pre-set period (20, 40, or 80 msec. for experimental sessions), before being replaced by a 50-msec. mask. The fixation cross returned upon termination of the mask. At the completion of each trial S recorded his result. This was done by marking the positions of the targets, placing a confidence rating beside each, and indicating which target was located first or i f they were both located at the same time. Design The design was three-way ( 2 X 3 X 6 ) with repeated measures on the last rwo factors. The independent variables were, firstly, the order of presenting the display exposures (either 20, 40, 80 or 80, 40, 2 0 ) , secondly display exposure, and thirdly the distance separating the two targets (0.67, 1.27, 1.80, 2.20, 2.43, and 2.53' of visual angle). c- were 4 undergraduates ( 2 males, 2 females) with normal 6/6 vision,
TARGET SEPARATION AND SELECTIVE ATTENTION
757
who received $1 per hour as payment for taking part in the experiment. TWO Ss searched displays where the exposures were ordered from shortest to longest, while the remaining two received the exposures in the reverse order. All SS searched for 2 practice and 9 experimental sessions, each of d ~ r a t i o napproximately 1 hr. At the completion of each trial they recorded the positions of the targets, indicated which was first to be located or if they were both located together, and rated each on a four-point confidence scale ( 4 = sure, 3 = almost sure, 2 = might have been, and 1 = guessing). A target was correctly located if the correct position or one either side was marked; the reasons for adopting this scoring method were outlined by Solman (1974).
Materials and Apparatus The letter F was the target item and G , 0, Q, U, and W were selected as irrelevant items. The stimuli were displayed in a Model G.B. Scientific Prototype Three-channel Tachistoscope. The luminance of the fixation, stimulus and mask channels was 10.8, 17.2, and 38.7 mL respectively. Stimulus displays were constructed by mounting black Letraset Upper Case letters ( 18 pt Grotesque 216) on square white cards of side length 203 mm. Each display contained 12 items positioned at the 12 clock positions on an imaginary circle, the diameter of which subtended a visual angle of 2.53". There were 36 displays constructed containing two Fs, i.e., given the restriction that over the 36 displays each of the 12 positions was equally likely to contain an F, a random selection of 6 of the 12 possible target arrangements was made for each separation ( 6 X 6 = 36). Twenty-four single-target displays were constructed with the target appearing twice at each position. When positioned each display item was 0.27' high with a stroke width of 0.06". The fixation card had a black cross mounted at the centre (arms and stroke width 0.43' and 0.06' respectively) and the mask consisted of a collection of broken, distorted, and whole Letraset letters of the same kind used in the displays. T o allow Ss to record their results sheets were constructed containing six circles with the 12 clock positions numbered 1 to 12. Procedure Upon arrival at the first practice session Ss were told that the task was visual search, the target appeared once on some cards and twice on others, the fixation cross should be in clear focus before the initiation of the display, and the confidence ratings referred to target presence. They were also informed that prior knowledge would be given concerning the number of targets in a display, and that a response, including a confidence rating, must be made to all targets. They were then familiarized with the displays and given 54 practice trials with those containing two targets ( 1 8 at each exposure) and 9 practice trials with the single-target displays ( 3 at each exposure), using exposures of 40, 80, and
160 msec. Two Ss received the exposues in ascending and two in descending order, and this order was maintained throughout the study. The order of presenting the single- and double-target displays was counterbalanced across subject groups. The same procedure was followed for the second practice session, using the experimental exposures (20, 40, and 80 msec.). After completion of practice Ss attended 9 experimental sessions. During the first 8 they searched 6 practice trials (2 at each exposure) before both the double- and single-target conditions and 126 experimental trials. For the doubletarget condition 36 displays were given at each exposure, and for the single-target condition 6. During Session 9 only 42 experimental trials were given (12 at each exposure for the double-target condition and 2 for the single target), which made up a total of 50 responses for each experimental condition.
RESULTSAND DISCUSSION A three-way analysis of variance for stimulus exposure presentation order, stimulus exposure, and the distance separating the two targets, showed that only the main effect of exposure and the interaction between exposure and separation reached statistical significance (F2,4 = 40.72, P < 0.01; PI0,20 = 6.49, P < 0.001). Since order of presentation of the stimulus exposures was not a significant variable, it has been excluded from the results. Table 1 shows the mean number of correctly located targets for the 18 experimental conditions given by the 3 exposures and the 6 distances separating the targets. TABLE 1 MEANNUMBER OF CORRECTLY LOCATED TARGET INSTANCES FOR 18 CONDITIONS GIVEN BY THREEEXPOSURESAND SIX DISTANCES SEPARATTNGTWO OF TARGET INSTANCES Separation (degrees of visual angle) 0.67 1.27 1.80 2.20 2.43 2.53
20
Stimulus Exposure (in msec.) 40
80
62.00 54.75 56.50 54.25 51.00 56.75
70.25 75.00 73.00 64.25 74.50 72.25
76.50 80.50 74.25 75.50 74.50 76.00
The failure of the distance separating the targets to have a significant effect on the total numher located ( F 5 . , ~= 2.35, P > 0.05) suggested that their selection for processing was independent of this variable. In terms of the analogy about a torch beam (Eriksen & Hoffman, 1972), the distance which must be travelled by the beam of processing capacicy as it moves from one target to the other, does not influence the rate and detail of the processing. At first this seems surprising, since it might be expected that moving processing capacity - -
TARGET SEPARATION AND SELECTIVE ATTENTION
759
from one target to the next would require time and that accuracy levels would fall as the distance moved and the time required increased. However, it is not necessary to conclude that the spatial separation of targets in the stimulus environment is represented the same way during processing. An explanation of the significant interaction between target separation and exposure (F10,20= 6.49, P < 0.001) was not obvious from the data. Observation of Table 1 showed that performance at separation 0.67' for 20 msec. and 2.20" for 40 msec. did not conform to the general pattern of the results. Accepting the 1' limit on focussing (Eriksen & Hoffman, 1972), it might be expected that performance would be superior at a separation of 0.67' for all exposures. Why this superiority was only found for an exposure of 20 msec. is not clear. The low level of performance at separation 2.20° for the 40-msec. exposure eludes theoretical interpretation. The significant increase in accuracy with increasing exposure was consistent with the results obtained by Solman (1974), i.e., in the earlier study the results showed a rapid increase in accuracy as exposure increased from 15 to 50 msec., with a tendency toward stable performance for further increases up to 180 msec. In the present study the accuracy changes were greater from 20 to 40 msec. than from 40 to 80 msec. In order to assess the effect of a redundancy of target instances, the data for the first target located were averaged over separation distances and compared with data for a single-target condition. To make the comparison meaningful both sets of results were corrected for chance performance.* The mean number of correct responses for two-target and one-target conditions (original and data corrected for chance), for each of the three levels of stimulus exposure, are shown in Table 2. Although the differences observed in the corrected data ( a higher level of performance for the two-target condition) suggested that target
TABLE 2 MEAN NUMBEROF
CORRECTLYLOCATED FIRSTTARGETS FORm0-TARGET CONDITION A N D SINGLE-TARGET CONDITION -
Stimulus Exposure (msec.)
Note.-Both
Corrected Data T w o targets One target
20 13.20 11.67 40 31.96 28.00 80 36.63 35.00 original and data corrected for chance are shown.
-
original Data T w o targets One target 30.83 42.08 43.50
21.25 33.50 38.75
T h e data for the single-target condition were corrected by subtracting one third of the incorrect responses from the correct responses for each subject, and then calculating the mean for all subjects (see Solman, 1974, Footnote 5 ) . The correction for the data o n the first target located in the two-target condition was carried out by subtracting the single-target factor multiplied by the value of the ratio of chance performance in the two conditions.
760
R. T.SOLMAN
redundancy increased the probability of a target being perceived, an analysis of variance showed that they did not reach significance (F1,2 = 5.57, P 0.15). Why should redundancy of targets increase the probability that a target will be perceived? The salience of the target features in the icon might be expected to determine the probability with which focal processing is directed to these features, and if there are two instances of the one target, this probability should be greater than in the single-target case. In the present study the number of instances of the target item only varied from one to two in a 12-item display, and this increase may not have been sufficient to raise the probability of perceiving a target significantly. Also, since subjects were required to report both targets, there may have been some trade-off in processing efficiency which increased performance for the second target, while slightly decreasing performance for the first.
