The Quarterly Journal of Experimental Psychology Section A
ISSN: 0272-4987 (Print) 1464-0740 (Online) Journal homepage: http://www.tandfonline.com/loi/pqja20
Does attention have different effects on line orientation and line arrangement discrimination? MaryLou Cheal , Don R. Lyon & David C. Hubbard To cite this article: MaryLou Cheal , Don R. Lyon & David C. Hubbard (1991) Does attention have different effects on line orientation and line arrangement discrimination?, The Quarterly Journal of Experimental Psychology Section A, 43:4, 825-857, DOI: 10.1080/14640749108400959 To link to this article: http://dx.doi.org/10.1080/14640749108400959
Published online: 29 May 2007.
Submit your article to this journal
Article views: 44
View related articles
Citing articles: 27 View citing articles
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=pqja20 Download by: [University of Nebraska, Lincoln]
Date: 30 May 2016, At: 22:15
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 1991.43A (4) 825-857
Does Attention have Different Effects on Line Orientation and Line Arrangement Discrimination? Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
MaryLou Cheal, Don R. Lyon, and David C. Hubbard University of Dayton Research Institute, Higley, A Z , U.S.A. Visual search and texture segregation studies have led to the inference that stimuli differing in the orientation of their component line segments can be distinguished without focal attention, whereas stimuli that differ only in the arrangement of line segments cannot. In most of this research, the locus of attention has not been explicitly manipulated. In the first experiment presented here, attention was directed to a relevant peripheral target by a cue presented near the target location or at the fovea. Effects of attention on orientation discrimination were assessed in a two-alternative forced-choice task with targets that were either: (1) lines that slanted obliquely to the right or left, or were horizontal or vertical, or (2) Y-like targets that had a short arm leading obliquely right or left of a vertical line. In some groups, a four-alternative forced-choice test with lines at o", 45", 9o", and 135" orientations was used. Discrimination of these targets (i.e. targets that differ in the orientation of component line segments) was only minimally facilitated as the time between the onset of the valid cue and the onset of the target (cue-target stimulus onset asynchrony, SOA) was iricreased from 0 or 17 msec to 267 msec. In contrast, discrimination of targets that did not differ in the orientation of component line segments but differed in line arrangement (T-like characters), was greatly facilitated by longer cue-target SOAs. In Experiment 2, a cue misdirected attention on 20% of the trials. A decrement occurred on incorrectly cued trials in comparison to correctly cued trials for both types of stimuli used (lines and Ts). The results from these experiments suggest that discrimination of line orientation benefits less from focal attention than does discrimination of line arrangement, but that both discriminations suffer when attention must be disengaged from an irrelevant spatial location. Requests for reprints should be sent to M. L. Cheal, University of Dayton Research Institute, P.O. Box 2020, Higley, AZ, U.S.A., 85236-2020. The authors gratefully acknowledge the editorial comments of Drs R. Evans, M. Houck, C. McCullough Howard, J. Kleiss, J. Lindholm, and E. L. Martin on an earlier version of the manuscript. They also thank S. Shoemaker and C. Voltz for computer programming assistance. Groups A and B of Experiment I and Group A of Experiment 2 were conducted by Dr. Cheal while on an Air Force Systems Command University Resident Research Program Fellowship. The research was funded by the Air Force Office of Scientific Research (Life Sciences Task 2313T3) and the Air Force Human Resources Laboratory (Contracts F-33615-84C-0066, F-33615-87-C-0012,and F-33615-90-C-0005).
0 1991 The Experimental Psychology Society
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
826
CHEAL, LYON, HUBBARD
Recent models of visual information processing have postulated muItipIe processes in which attention is a key component (Neisser, 1967; Reeves & Sperling, 1986; Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977; Treisman & Gelade, 1980). Much of this research involves precuing a target location to elicit a shift of visual attention in the absence of eye movements. Although some of this research is controversial, there is no doubt that precuing a stimulus location results in more accurate and more rapid detection of a simple target, such as a spot of light (Posner, 1980; Shulman, Remington, & McLean, 1979), and more accurate and more rapid discrimination of more complex stimuli, such as letters (Eriksen & Collins, 1969; Eriksen & Hoffman, 1972a, 1972b, 1973; Holmgren, 1974; Jonides, 1980; Sperling & Melchner, 1978). In spite of the usually strong effect of location precuing on visual discrimination, in some experiments performance was not improved. For instance, location cuing did not improve determination of whether five letters were a word or a non-word (Hardyck, Chiarello, Dronkers, & Simpson, 1985). Further, large performance decrements are not always found in studies in which a cue shifts attention to an incorrect location (Posner, 1980). For example, Eriksen and Yeh (1985) reported a much smaller decrement from invalid cues when a target digit or letter was displayed in a field of dots than when it was displayed in a field of a constant set of letters. Of particular relevance to the present experiments is the finding by Ambler and Finklea (1976) that location cuing does not improve detection of a tilted T among upright Ts. The discrimination of a tilted T from an upright T, or an L from an X, is most probably made on the basis of line orientation. The discrimination of line orientation could be thought of as an automatic response that is made without the need for focal attention. If it is automatic, then line orientation discrimination would be involuntary, would operate in parallel over the visual field, would be independent of other tasks, and would not benefit greatly from focal attention (Kahneman & Treisman, 1984). This view of orientation discrimination is supported by the results of several other studies. Beck and Ambler (1972) showed that a tilted T was detected better than an upright L when either was embedded in a field of upright Ts. Further, if the field was limited to eight characters and the target location was correctly cued prior to stimulus presentation, there was no significant difference in accuracy of detection of an L or a tilted T (Beck & Ambler, 1973). In another search study, features that allowed rapid texture segregation (such as an L among Xs) were the same features that were detected rapidly when presented with noise elements (Bergen & Julesz, 1983). In contrast, an L among randomly oriented Ts was not detected rapidly. Not all studies support the view that orientation discrimination is automatic, however. For example, Sagi and Julesz (1987) showed that detection of line orientation varies as a function of the number of distractors
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
AlTENTION AND ORIENTATION DISCRIMINATION
827
at some intercharacter densities. If the task were purely automatic, then the number of distractors should not affect performance. The present research tests the effects of spatially focused attention on orientation discrimination by examining the time course of effects of a location precue. Two different kinds of stimuli are used in these experiments: (1) stimuli that, from previous research, may be assumed to be discriminated by preattentive processes (characters that differed in line orientation), and (2) stimuli that were assumed to require focused attention (characters that differed in line arrangement but not line orientation; Beck, 1982). The method used was a variation of one used by Lyon (1987, 1990) to determine the time course of spatial attention effects on discrimination of the conjunction of line segments. This method contains elements of techniques used by Eriksen and Hoffman (1972a), Posner (1980), and Bashinski and Bacharach (1980). The key feature of the present research is that here attention is manipulated directly by presenting a spatial cue in the area of the target a few milliseconds before the target is presented. The cue is essential in order to define one of four possible target locations, all of which contain target-type stimuli. If the usual inference from the earlier studies is correct, and stimuli that differ in the orientation of their component lines are discriminated in the absence of focal attention, then they may not benefit from a spatial cue. This possibility was tested in the first experiment. It was shown that discrimination of line orientation did not benefit from focal attention nearly as much as did discrimination of line arrangement. This was true even when discrimination of the two stimuli was approximately equated for overall difficulty. If, as the results of these experiments suggest, the process of orientation discrimination is relatively unaffected by focal attention to the target area, then it may also be unaffected by the focusing of attention elsewhere in the visual field (Kahneman & Treisman, 1984). On the other hand, decrements in reaction time (RT) are found even in detection of simple stimuli when the target location is incorrectly cued (Briand & Klein, 1987; Posner, 1980; Prinzmetal, Presti, & Posner, 1986). Such data predict decrements in performance in orientation discrimination if attention is first directed to a non-target location. These alternative hypotheses were tested in the second experiment by including some trials in which the cue directed attention away from the target area. This misdirection of attention resulted in a decrement in accuracy for both orientation discrimination and discrimination of line arrangement.
828
CHEAL, LYON, HUBBARD
GENERAL METHOD
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
0bservers Men and women, 22 to 37 years of age, with normal or corrected to normal vision, were paid to participate in multiple sessions of approximately one hour each. Some observers participated in more than one condition, and there were two to five observers tested in each condition. In these experiments, in addition to the hourly salary, observers could earn a bonus based on overall accuracy for each stimulus set. The bonus was used to increase motivation for accuracy.
Apparatus Stimuli were displayed by an IBM-XT on an enhanced colour monitor with phosphors P-22-B, P-22-G, and P-22-R, all with decay to 10% in less than 1 msec. The white pixels were presented against a dark grey background. In Experiment 1, Groups A and B, and in Experiment 2, Group A, luminance was 13.7 cd/m2.Later, changes in the experimental configuration resulted in greater luminance (99.1 cd/m2)of the screen. Other aspects of the apparatus did not change, but the greater luminance resulted in more accurate responses and, thus, for some observers, target durations were reduced. In these experiments, the short duration of stimulus presentation prevented any increase in accuracy due to eye movements. In fact, an eye movement would have reduced response accuracy under some conditions because of saccadic suppression. However, in order to provide confirmation that fixation was maintained by the observer prior to responding, eye movement was monitored with a video camera. Eye monitoring also allowed the experimenter to give feedback to the observers that helped acquisition of the task. An adjustable chin rest helped to maintain head position at a distance of 29 cm (Experiments lA, lB, 2A) or 38 cm (Experiments lC, lD, lE, lF, 2B).
Procedure Observers were seated in front of the computer monitor with room lights on. They were instructed to maintain fixation on a bar of light (0.2" x 0.4") in the centre of the screen throughout each trial (Figure 1). The computer displayed frames of information at the rate of sixty per second. Thus, the duration of each frame was 16.7 msec. After 668 msec, a rectangular peripheral cue (0.7" x 1.2") appeared in one of four locations: 9" above or below, or 8" right or left of fixation for Groups A and B of Experiment 1, and Group A of Experiment 2, and 7"from fixation for Groups D and F of Experiment 1, and Group B of Experiment 2. In Groups C and E of Experiment 1, central cues (arrows from the standard character set, made of 6 x 7 [vertical] or 7 X 5
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
ATTENTION AND ORIENTATION DISCRIMINATION
829
(CI 6. SIDEWAYS T's
50. 67. or 80 msec
FIG. 1 . Sequence of stimulus events in Experiment I . The two types of stimuli (A: lines, B: Ts) were presented in separate blocks of trials. The actual stimuli were composed of white pixels on a dark grey screen.
[horizontal] pixels) were presented at fixation. The duration of the cue in all conditions was one frame (16.7 msec). The peripheral cue was followed by the fixation point, and the central cue was followed by a blank screen for an interval of variable duration, which could include trials with simultaneous presentation of the cue and target screens. The sum of the cue duration and this variable interval was the cue-target stimulus onset asynchrony (SOA). Thirteen SOAs were used for Groups A, C, D, E, and F of Experiment 1 and Group B of Experiment 2; 11 SOAs were used in Group B of Experiment 1; and 4 SOAs were used in Group A of Experiment 2. At the end of the SOA, stimuli appeared at each of the four locations: 7.5" above or below, and 6.4" right or left of fixation for Groups A and B of Experiment 1 and Group A of Experiment 2, and at 6" from fixation for Groups C, D, E, and F of Experiments 1 and Group B of Experiment 2. The target was the stimulus at the location that was cued. The stimuli for the other three locations were randomly chosen with replacement from the four possible targets. The order of presentation of stimulus sets
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
830
CHEAL, LYON, HUBBARD
(T-like characters or characters containing line orientation differences) was counterbalanced across observers and alternated daily for each observer. The stimuli were followed by a mask that remained lit until the observer responded. The mask was constructed separately for each experimental group and was an approximate negative image of the combination of all possible targets for that group. To respond, the observer pressed the directional arrow key on the computer keypad to indicate the orientation of the stimulus in the cued position. After the response was entered, feedback to indicate correct or incorrect response appeared at the fixation position, and the next trial was initiated. Trials were blocked by type of stimulus with independent randomization of all other variables within each block of 100 or 104 trials for each stimulus set. Within each block (Groups A and B of Experiment 1 and Group A of Experiment 2) or each four blocks (Groups C,D, E, and F of Experiment 1 and Group B of Experiment 2), cue-target locations (right, left, above, or below fixation), cue-target SOAs, target durations, and stimulus orientations were independently randomized. In Groups C, D, E, and F of Experiment 1 and Group B of Experiment 2 randomization was accomplished by placing the variables in an array and then drawing them randomly for trials (total of 416 trials for each target set each session). In the initial session, naive observers received training trials with one type of stimulus presented for long target durations to become familiar with the procedure and the stimuli. They then received four or five blocks of 100 or 104 trials each, using the same stimulus set. The observer then received training trials with the other stimulus set presented for long target durations, followed by four or five blocks of 100 or 104 trials each with this stimulus set. In some cases, early blocks of trials were used for training and were not used in the final analyses. Trials reported in Table 1 indicate the total number of trials analysed for each experimental group.
Statistical Analyses The data consisted of categorical factors with correct/incorrect dichotomous responses. These are the types of data for which the log-linear model of crossclassified categorical data was developed (Fienberg, 1980; Grizzle, Starmer, & Koch, 1969). The log-linear model analysis is a generalized linear model approach that provides the same information regarding factor effects and interactions without the necessity of meeting the assumptions of the analysis of variance. Tests of significance were obtained from chi-square tests of partial association. The log-linear model approach (Fienberg, 1980) extends the concept of the chi-square to contingency tables of higher dimensions. As the only interest in the analysis was the effect of the experimental factors
Experiment
Group
Observers
Cue Validity 100% 100% 100% 100% 100% 100% 80% 80 Yo
Cue Location Peripheral Peripheral Central Peripheral Central Peripheral Peripheral Peripheral
Trials 72,000 36,000 24,9 1 1 19,968 79,872 33,280 49,700 39,936
TABLE 1 Experimental Conditions
Target Orientations
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
17, 33, 50 33, 50 17 33, 50 33, 50 33, 50 33, 50 33, 50
Lines
50, 67, 84 67, 84 33, 67 33, 50 33, 50 33, 50 50, 67 33,50
Ts
Durations (msec)
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
832
CHEAL, LYON, HUBBARD
upon the dichotomous response, a logit type of log-linear analysis was performed. This procedure produces the same type of results as a leastsquares analysis of variance (i.e. tests of main effects and interactions). The analysis was performed using SPSS-X (Groups A, B, C of Experiment 1 and Group A of Experiment 2 ) or BMDP4 F (Groups D, E, F of Experiment 1 and Group B of Experiment 2 ) hierarchical log-linear analysis programs (Dixon, 1988). The data for each experimental group were analysed separately. Analysis of the total data was performed with all variables included as grouping factors: stimulus type (line orientations and Ts), target duration, cue-target SOAs, target orientation, target location, and observers. Following the initial analysis for each group, additional analyses were computed dependent on the significance of the main effects and interactions in the first analysis. In addition, inasmuch as these tests were performed with a fixed effects model, separate analyses for each observer were also computed, but not reported. These analyses supported the overall analyses. The effects of target orientation and location (Cheal, Lyon, & Hubbard, 1987) are not relevant to the hypotheses being tested here; therefore, they are not discussed. In addition to the complete analyses described above, an improvement score was computed by subtracting the proportion correct at 17 msec SOA from the mean proportion correct for SOAs equal or greater than 150 msec.
EXPERIMENT 1 Experiment 1 was a test of whether focusing attention on a target location facilitates discrimination of line orientation as much as it does discrimination of line arrangement. To test this question, a location precue was used with a variable interval prior to target appearance. In this experiment, the cue was always valid; it always indicated the correct location for the target. In order to eliminate possible artifacts in interpretation, six groups of observers were tested with somewhat different conditions. Line orientation discrimination was assessed in two-alternative or fouralternative forced-choice discriminations in which the target was either a single line of two or four different orientations, or a Y-like character in which the short arm was oblique to the vertical line ( f45"). The single-line stimuli are similar to those used in texture segregation research. For example, Olson and Attneave (1 970) showed that RTs to locate texture boundaries when the stimuli were left- and right-slanted lines were much faster than the RTs to make the same judgement when the stimuli only differed in line arrangement (Ls in various orientations). Line arrangement discrimination was assessed in the same task with target characters that were composed of two line segments perpendicular to one another and conjoined to form a T-like character. This type of target has
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
ATTENTION AND ORIENTATION DISCRIMINATION
833
been shown to require serial search under certain conditions (Humphreys, Quinlan, & Riddoch, 1989). If orientation discrimination benefits less from attention than does line arrangement discrimination, then there should be less facilitation of discrimination of targets that contain lines of different orientation than of targets that do not (T-like characters). In preliminary tests it was found that, in general, slanted lines were much easier to discriminate than T-like characters. To avoid potential effects of this difference on the measurement of attention effects, in some of the groups the stimuli were approximately equated in discriminability by using more pixels (increasing total luminance) and/or longer durations for the T-like stimuli than for the slanted lines. In every condition, all of the variations of the target for one stimulus type were composed of the same number of line endings and approximately the same number of pixels, so that only the orientation of the target varied. In Croup A , the task involved a two-alternative forced-choice discrimination of slanted lines and Ts (Table 1). Large differences in accuracy as a function of cue-target SOA between discrimination of slanted lines and discrimination of Ts were found. However, this could have been due to other aspects of these targets besides the presence or absence of an orientation difference. The two types of targets differed in several ways that might influence the number of features present in the stimuli: the number of line segments, the number of line junctions, the number of line terminators, and total luminance. The first three of these features have been shown to be important in texture segregation experiments (Julesz, 1981). Therefore, in Croup B, the stimuli were composed of the same number of pixels, line endings, etc. Only the orientation of the short line of Ys or Ts varied: for Ys, the short arm was oblique to the vertical line; for Ts, the short arm was perpendicular to the vertical line. To compare Groups A and B, it was necessary to ensure that both types of stimuli were tested under identical procedures. For that reason, the same cue (a small rectangle) and mask were used for both conditions. One possible problem is that the rectangular cue could have masked the T-like figures more than it masked the slanted lines. Such a possibility is supported by existing data that showed more interference with discrimination of line orientation when noise lines were parallel than when they were perpendicular to the target (Andriessen & Bouma, 1976). Lyon (1990, Experiment l), using the present paradigm with T-like characters as targets, showed that there is a range of SOAs from - 17 msec to + 17 msec for which there may be some cue-target masking. For longer SOAs, however, performance was better with cues closer to the target (1" away) than with more distant cues (2' away), a result that is not consistent with a masking explanation. However, as existing data allow for the possibility of some cue-target masking at some SOAs, the question remains as to whether the cue might differentially mask
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
834
CHEAL. LYON, HUBBARD
the Ts versus the slanted lines. Therefore, additional experiments were conducted in which the possibility of masking effects was minimized. The peripheral cue was used in the present experiments in order to reduce delays due to deciding where to attend. However, precue effects also have been found with central cues (such as an arrow) (Eriksen & Collins, 1969; Remington, 1980; Remington & Pierce, 1984; Shulman et al., 1979) for both detection (as of light increment, Posner, Snyder, & Davidson, 1980) and discrimination (i.e. a letter, Jonides, 1981). Cheal and Lyon (1989) used both peripheral and central cues in facilitating four-alternative forced-choice discrimination of T-like figures as a function of eccentricity. The main difference in the data for these two conditions was a delay in onset of effects. The fact that similar effects on discrimination of T-like characters occur following either kind of cue suggests that these effects are not due to masking of the target by the cue, or to some other local cue-target interaction. It is doubtful that a central cue, presented at fixation, could mask the target. Therefore, in Group C , the peripheral rectangular cue was replaced by a central arrow cue. The 6" separation between target and cue should be sufficient to prevent any differential masking. In the Andriessen and Bouma study (1976), differential interference from parallel vs. perpendicular noise elements occurred at small angular distances. It was found that interference depended on eccentricity of the target such that if the target was at 6", there would be little interference if noise was more than 2.4" away. In previous studies using this paradigm (Cheal & Lyon, 1989; Lyon, 1987), it was easy to show improvement in accuracy with SOA when a central cue was used, because in those studies chance performance was 25%, and, thus, performance could improve by 75%. In Groups A, B, and C, chance performance is 50%, and therefore there is less room for improvement. Because of this fact, any delay in onset of attentional effects found with a central cue (Cheal & Lyon, this issue) in Group C could interact with the smaller possible amount of improvement, and thus, any differences between curves for slants and Ts would be smaller with the central cue than with the peripheral cue. Therefore, in order to look for larger differences between discrimination of line orientation and discrimination of line arrangement, Group D repeated the task of Group A, but with a four-alternative discrimination rather than the two-alternative discrimination that was used in the first three groups. Again, differencesbetween discrimination of lines and Ts at short SOAs in Group D could have been due to differential masking as suggested for Group A. Therefore, in Group E, the four-alternative discrimination was conducted with the foveal arrow cue as in Group C. It has been suggested that the visual system operates differently in perception of horizontal and vertical lines in comparison to perception of
ATTENTION AND ORIENTATION DISCRIMINATION
835
oblique lines (Pettigrew, Nikara, & Bishop, 1968; Vogels & Orban, 1986). Although there is no reason to assume that effects of attention would be affected by the obliqueness of the targets, some observers were tested with oblique targets. Group F also provided a final test of possible effects of cue/ target interactions that might have resulted in poorer performance at short SOAs. In this condition the lines and Ts were both rotated obliquely, but the cue was the same rectangle used for the other peripheral cue groups.
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
Procedure Group A: Two-Alternative Forced-Choice Discrimination of (1) Slanted Lines and ( 2 ) Sidewuys Ts; Peripheral Cue. An extended character set was generated in order to present the characters used as stimuli. The slanted lines were composed of a I" straight line, made of 5 pixels, slanted at an oblique angle, with the top of the line pointing either clockwise (right) or counterclockwise (left) (Figure 1A). The Ts were formed by a 0.8"horizontal line, composed of 2 rows of 4 pixels that extended either right or left from the centre of a 0.8" vertical line, composed of 8 rows of 2 pixels (total number of pixels=24; Figure 1B). The mask is illustrated in Figure 1. The three observers each completed 24 sessions of 1000 trials. For other particulars of each group, see Table 1. Group B: Two-Alternative Forced-Choice Discrimination of ( I ) Ys and ( 2 ) Sideways Ts; Peripheral Cue. Each stimulus character was composed of a vertical line of 9 pixels and a line of 4 pixels that conjoined the vertical line at the 5th pixel. For the orientation discrimination, the conjoining line led upward at an oblique angle right or left of the vertical line, forming a Ylike figure. For the line arrangement discrimination, the conjoining line was horizontal to either the right or the left of the vertical line, forming a sideways T. Each stimulus had three terminating ends. For Ys, discrimination could be made by line orientation as in Experiment I , or it could be made by the arrangement of the two conjoining line segments. For Ts, on the other hand, the only basis for discrimination was the arrangement of the two line segments. A new mask was constructed by making an outline of one pixel thickness around the compound character created by superimposing the four possible targets (right and left Ts and right and left Ys). Each observer completed 12 sessions of 1000 trials. Group C: Two-Alternative Forced-Choice Discrimination of ( 1 ) Slanted Lines and ( 2 ) Sideways Ts; Central Arrow Cue. The stimuli were the same as those used for Group A. The fixation point was removed when the cue appeared and only reappeared for the next trial. The mask was an outline of the compound stimuli. Each observer completed 20 sessions of 832 trials.
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
836
CHEAL, LYON, HUBBARD
Group D: Four-Alternative Forced-Choice Discrimination of ( I ) Lines and ( 2 ) T-like Characters; Peripheral Cue. For line orientation discrimination, two slanted lines (right & left), a vertical line (respond up), and a horizontal line (respond down) were composed of 5-pixel straight lines. For the line arrangement discrimination, the sideways Ts were constructed of a 12-pixel vertical line with a 2 x 3-pixel perpendicular line conjoining at the centre pixels of the vertical line (18 pixels). The other two Ts were composed of a 2 x 7-pixel horizontal line with a 5-pixel vertical line conjoining at the centre of the horizontal line (19 pixels). The mask was an outline of the composite of the eight possible targets. Each observer completed 8 sessions of 832 trials. Group E: Four-Alternative Forced-Choice Discrimination of ( 1 ) Slanted Lines and ( 2 ) Sideways Ts of Same Dimensions as Group D; Central Arrow Cue. The fixation point was removed when the cue appeared and only reappeared for the next trial. Each observer completed 24 sessions of 832 trials. Group F: Two-Alternative Forced-Choice Discrimination of ( I ) Horizontal and Vertical Lines and ( 2 ) Oblique Ts; Peripheral Cue. Each horizontal or vertical straight line was composed of 4 pixels. Ts were composed of a 5-pixel oblique line with 2 pixels leading up and right or down and left at approximately a right angle to the longer line. The monitor did not allow an actual right angle for these Ts, but discrimination was based on line arrangement because the lines of the two variations did not differ in orientation. The mask was an outline of the compound stimuli enclosed in a rectangle. Each observer completed 8 sessions of 832 trials.
R esu1ts Overall Analyses. The log-linear model test revealed significant effects of all variables and many significant interactions for each group. As expected, proportion of correct responses was well above chance: for example, for Group A, where chance was 0.50, total probability correct was 0.76; x2(1)=21,631, p
ISSN: 0272-4987 (Print) 1464-0740 (Online) Journal homepage: http://www.tandfonline.com/loi/pqja20
Does attention have different effects on line orientation and line arrangement discrimination? MaryLou Cheal , Don R. Lyon & David C. Hubbard To cite this article: MaryLou Cheal , Don R. Lyon & David C. Hubbard (1991) Does attention have different effects on line orientation and line arrangement discrimination?, The Quarterly Journal of Experimental Psychology Section A, 43:4, 825-857, DOI: 10.1080/14640749108400959 To link to this article: http://dx.doi.org/10.1080/14640749108400959
Published online: 29 May 2007.
Submit your article to this journal
Article views: 44
View related articles
Citing articles: 27 View citing articles
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=pqja20 Download by: [University of Nebraska, Lincoln]
Date: 30 May 2016, At: 22:15
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 1991.43A (4) 825-857
Does Attention have Different Effects on Line Orientation and Line Arrangement Discrimination? Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
MaryLou Cheal, Don R. Lyon, and David C. Hubbard University of Dayton Research Institute, Higley, A Z , U.S.A. Visual search and texture segregation studies have led to the inference that stimuli differing in the orientation of their component line segments can be distinguished without focal attention, whereas stimuli that differ only in the arrangement of line segments cannot. In most of this research, the locus of attention has not been explicitly manipulated. In the first experiment presented here, attention was directed to a relevant peripheral target by a cue presented near the target location or at the fovea. Effects of attention on orientation discrimination were assessed in a two-alternative forced-choice task with targets that were either: (1) lines that slanted obliquely to the right or left, or were horizontal or vertical, or (2) Y-like targets that had a short arm leading obliquely right or left of a vertical line. In some groups, a four-alternative forced-choice test with lines at o", 45", 9o", and 135" orientations was used. Discrimination of these targets (i.e. targets that differ in the orientation of component line segments) was only minimally facilitated as the time between the onset of the valid cue and the onset of the target (cue-target stimulus onset asynchrony, SOA) was iricreased from 0 or 17 msec to 267 msec. In contrast, discrimination of targets that did not differ in the orientation of component line segments but differed in line arrangement (T-like characters), was greatly facilitated by longer cue-target SOAs. In Experiment 2, a cue misdirected attention on 20% of the trials. A decrement occurred on incorrectly cued trials in comparison to correctly cued trials for both types of stimuli used (lines and Ts). The results from these experiments suggest that discrimination of line orientation benefits less from focal attention than does discrimination of line arrangement, but that both discriminations suffer when attention must be disengaged from an irrelevant spatial location. Requests for reprints should be sent to M. L. Cheal, University of Dayton Research Institute, P.O. Box 2020, Higley, AZ, U.S.A., 85236-2020. The authors gratefully acknowledge the editorial comments of Drs R. Evans, M. Houck, C. McCullough Howard, J. Kleiss, J. Lindholm, and E. L. Martin on an earlier version of the manuscript. They also thank S. Shoemaker and C. Voltz for computer programming assistance. Groups A and B of Experiment I and Group A of Experiment 2 were conducted by Dr. Cheal while on an Air Force Systems Command University Resident Research Program Fellowship. The research was funded by the Air Force Office of Scientific Research (Life Sciences Task 2313T3) and the Air Force Human Resources Laboratory (Contracts F-33615-84C-0066, F-33615-87-C-0012,and F-33615-90-C-0005).
0 1991 The Experimental Psychology Society
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
826
CHEAL, LYON, HUBBARD
Recent models of visual information processing have postulated muItipIe processes in which attention is a key component (Neisser, 1967; Reeves & Sperling, 1986; Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977; Treisman & Gelade, 1980). Much of this research involves precuing a target location to elicit a shift of visual attention in the absence of eye movements. Although some of this research is controversial, there is no doubt that precuing a stimulus location results in more accurate and more rapid detection of a simple target, such as a spot of light (Posner, 1980; Shulman, Remington, & McLean, 1979), and more accurate and more rapid discrimination of more complex stimuli, such as letters (Eriksen & Collins, 1969; Eriksen & Hoffman, 1972a, 1972b, 1973; Holmgren, 1974; Jonides, 1980; Sperling & Melchner, 1978). In spite of the usually strong effect of location precuing on visual discrimination, in some experiments performance was not improved. For instance, location cuing did not improve determination of whether five letters were a word or a non-word (Hardyck, Chiarello, Dronkers, & Simpson, 1985). Further, large performance decrements are not always found in studies in which a cue shifts attention to an incorrect location (Posner, 1980). For example, Eriksen and Yeh (1985) reported a much smaller decrement from invalid cues when a target digit or letter was displayed in a field of dots than when it was displayed in a field of a constant set of letters. Of particular relevance to the present experiments is the finding by Ambler and Finklea (1976) that location cuing does not improve detection of a tilted T among upright Ts. The discrimination of a tilted T from an upright T, or an L from an X, is most probably made on the basis of line orientation. The discrimination of line orientation could be thought of as an automatic response that is made without the need for focal attention. If it is automatic, then line orientation discrimination would be involuntary, would operate in parallel over the visual field, would be independent of other tasks, and would not benefit greatly from focal attention (Kahneman & Treisman, 1984). This view of orientation discrimination is supported by the results of several other studies. Beck and Ambler (1972) showed that a tilted T was detected better than an upright L when either was embedded in a field of upright Ts. Further, if the field was limited to eight characters and the target location was correctly cued prior to stimulus presentation, there was no significant difference in accuracy of detection of an L or a tilted T (Beck & Ambler, 1973). In another search study, features that allowed rapid texture segregation (such as an L among Xs) were the same features that were detected rapidly when presented with noise elements (Bergen & Julesz, 1983). In contrast, an L among randomly oriented Ts was not detected rapidly. Not all studies support the view that orientation discrimination is automatic, however. For example, Sagi and Julesz (1987) showed that detection of line orientation varies as a function of the number of distractors
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
AlTENTION AND ORIENTATION DISCRIMINATION
827
at some intercharacter densities. If the task were purely automatic, then the number of distractors should not affect performance. The present research tests the effects of spatially focused attention on orientation discrimination by examining the time course of effects of a location precue. Two different kinds of stimuli are used in these experiments: (1) stimuli that, from previous research, may be assumed to be discriminated by preattentive processes (characters that differed in line orientation), and (2) stimuli that were assumed to require focused attention (characters that differed in line arrangement but not line orientation; Beck, 1982). The method used was a variation of one used by Lyon (1987, 1990) to determine the time course of spatial attention effects on discrimination of the conjunction of line segments. This method contains elements of techniques used by Eriksen and Hoffman (1972a), Posner (1980), and Bashinski and Bacharach (1980). The key feature of the present research is that here attention is manipulated directly by presenting a spatial cue in the area of the target a few milliseconds before the target is presented. The cue is essential in order to define one of four possible target locations, all of which contain target-type stimuli. If the usual inference from the earlier studies is correct, and stimuli that differ in the orientation of their component lines are discriminated in the absence of focal attention, then they may not benefit from a spatial cue. This possibility was tested in the first experiment. It was shown that discrimination of line orientation did not benefit from focal attention nearly as much as did discrimination of line arrangement. This was true even when discrimination of the two stimuli was approximately equated for overall difficulty. If, as the results of these experiments suggest, the process of orientation discrimination is relatively unaffected by focal attention to the target area, then it may also be unaffected by the focusing of attention elsewhere in the visual field (Kahneman & Treisman, 1984). On the other hand, decrements in reaction time (RT) are found even in detection of simple stimuli when the target location is incorrectly cued (Briand & Klein, 1987; Posner, 1980; Prinzmetal, Presti, & Posner, 1986). Such data predict decrements in performance in orientation discrimination if attention is first directed to a non-target location. These alternative hypotheses were tested in the second experiment by including some trials in which the cue directed attention away from the target area. This misdirection of attention resulted in a decrement in accuracy for both orientation discrimination and discrimination of line arrangement.
828
CHEAL, LYON, HUBBARD
GENERAL METHOD
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
0bservers Men and women, 22 to 37 years of age, with normal or corrected to normal vision, were paid to participate in multiple sessions of approximately one hour each. Some observers participated in more than one condition, and there were two to five observers tested in each condition. In these experiments, in addition to the hourly salary, observers could earn a bonus based on overall accuracy for each stimulus set. The bonus was used to increase motivation for accuracy.
Apparatus Stimuli were displayed by an IBM-XT on an enhanced colour monitor with phosphors P-22-B, P-22-G, and P-22-R, all with decay to 10% in less than 1 msec. The white pixels were presented against a dark grey background. In Experiment 1, Groups A and B, and in Experiment 2, Group A, luminance was 13.7 cd/m2.Later, changes in the experimental configuration resulted in greater luminance (99.1 cd/m2)of the screen. Other aspects of the apparatus did not change, but the greater luminance resulted in more accurate responses and, thus, for some observers, target durations were reduced. In these experiments, the short duration of stimulus presentation prevented any increase in accuracy due to eye movements. In fact, an eye movement would have reduced response accuracy under some conditions because of saccadic suppression. However, in order to provide confirmation that fixation was maintained by the observer prior to responding, eye movement was monitored with a video camera. Eye monitoring also allowed the experimenter to give feedback to the observers that helped acquisition of the task. An adjustable chin rest helped to maintain head position at a distance of 29 cm (Experiments lA, lB, 2A) or 38 cm (Experiments lC, lD, lE, lF, 2B).
Procedure Observers were seated in front of the computer monitor with room lights on. They were instructed to maintain fixation on a bar of light (0.2" x 0.4") in the centre of the screen throughout each trial (Figure 1). The computer displayed frames of information at the rate of sixty per second. Thus, the duration of each frame was 16.7 msec. After 668 msec, a rectangular peripheral cue (0.7" x 1.2") appeared in one of four locations: 9" above or below, or 8" right or left of fixation for Groups A and B of Experiment 1, and Group A of Experiment 2, and 7"from fixation for Groups D and F of Experiment 1, and Group B of Experiment 2. In Groups C and E of Experiment 1, central cues (arrows from the standard character set, made of 6 x 7 [vertical] or 7 X 5
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
ATTENTION AND ORIENTATION DISCRIMINATION
829
(CI 6. SIDEWAYS T's
50. 67. or 80 msec
FIG. 1 . Sequence of stimulus events in Experiment I . The two types of stimuli (A: lines, B: Ts) were presented in separate blocks of trials. The actual stimuli were composed of white pixels on a dark grey screen.
[horizontal] pixels) were presented at fixation. The duration of the cue in all conditions was one frame (16.7 msec). The peripheral cue was followed by the fixation point, and the central cue was followed by a blank screen for an interval of variable duration, which could include trials with simultaneous presentation of the cue and target screens. The sum of the cue duration and this variable interval was the cue-target stimulus onset asynchrony (SOA). Thirteen SOAs were used for Groups A, C, D, E, and F of Experiment 1 and Group B of Experiment 2; 11 SOAs were used in Group B of Experiment 1; and 4 SOAs were used in Group A of Experiment 2. At the end of the SOA, stimuli appeared at each of the four locations: 7.5" above or below, and 6.4" right or left of fixation for Groups A and B of Experiment 1 and Group A of Experiment 2, and at 6" from fixation for Groups C, D, E, and F of Experiments 1 and Group B of Experiment 2. The target was the stimulus at the location that was cued. The stimuli for the other three locations were randomly chosen with replacement from the four possible targets. The order of presentation of stimulus sets
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
830
CHEAL, LYON, HUBBARD
(T-like characters or characters containing line orientation differences) was counterbalanced across observers and alternated daily for each observer. The stimuli were followed by a mask that remained lit until the observer responded. The mask was constructed separately for each experimental group and was an approximate negative image of the combination of all possible targets for that group. To respond, the observer pressed the directional arrow key on the computer keypad to indicate the orientation of the stimulus in the cued position. After the response was entered, feedback to indicate correct or incorrect response appeared at the fixation position, and the next trial was initiated. Trials were blocked by type of stimulus with independent randomization of all other variables within each block of 100 or 104 trials for each stimulus set. Within each block (Groups A and B of Experiment 1 and Group A of Experiment 2) or each four blocks (Groups C,D, E, and F of Experiment 1 and Group B of Experiment 2), cue-target locations (right, left, above, or below fixation), cue-target SOAs, target durations, and stimulus orientations were independently randomized. In Groups C, D, E, and F of Experiment 1 and Group B of Experiment 2 randomization was accomplished by placing the variables in an array and then drawing them randomly for trials (total of 416 trials for each target set each session). In the initial session, naive observers received training trials with one type of stimulus presented for long target durations to become familiar with the procedure and the stimuli. They then received four or five blocks of 100 or 104 trials each, using the same stimulus set. The observer then received training trials with the other stimulus set presented for long target durations, followed by four or five blocks of 100 or 104 trials each with this stimulus set. In some cases, early blocks of trials were used for training and were not used in the final analyses. Trials reported in Table 1 indicate the total number of trials analysed for each experimental group.
Statistical Analyses The data consisted of categorical factors with correct/incorrect dichotomous responses. These are the types of data for which the log-linear model of crossclassified categorical data was developed (Fienberg, 1980; Grizzle, Starmer, & Koch, 1969). The log-linear model analysis is a generalized linear model approach that provides the same information regarding factor effects and interactions without the necessity of meeting the assumptions of the analysis of variance. Tests of significance were obtained from chi-square tests of partial association. The log-linear model approach (Fienberg, 1980) extends the concept of the chi-square to contingency tables of higher dimensions. As the only interest in the analysis was the effect of the experimental factors
Experiment
Group
Observers
Cue Validity 100% 100% 100% 100% 100% 100% 80% 80 Yo
Cue Location Peripheral Peripheral Central Peripheral Central Peripheral Peripheral Peripheral
Trials 72,000 36,000 24,9 1 1 19,968 79,872 33,280 49,700 39,936
TABLE 1 Experimental Conditions
Target Orientations
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
17, 33, 50 33, 50 17 33, 50 33, 50 33, 50 33, 50 33, 50
Lines
50, 67, 84 67, 84 33, 67 33, 50 33, 50 33, 50 50, 67 33,50
Ts
Durations (msec)
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
832
CHEAL, LYON, HUBBARD
upon the dichotomous response, a logit type of log-linear analysis was performed. This procedure produces the same type of results as a leastsquares analysis of variance (i.e. tests of main effects and interactions). The analysis was performed using SPSS-X (Groups A, B, C of Experiment 1 and Group A of Experiment 2 ) or BMDP4 F (Groups D, E, F of Experiment 1 and Group B of Experiment 2 ) hierarchical log-linear analysis programs (Dixon, 1988). The data for each experimental group were analysed separately. Analysis of the total data was performed with all variables included as grouping factors: stimulus type (line orientations and Ts), target duration, cue-target SOAs, target orientation, target location, and observers. Following the initial analysis for each group, additional analyses were computed dependent on the significance of the main effects and interactions in the first analysis. In addition, inasmuch as these tests were performed with a fixed effects model, separate analyses for each observer were also computed, but not reported. These analyses supported the overall analyses. The effects of target orientation and location (Cheal, Lyon, & Hubbard, 1987) are not relevant to the hypotheses being tested here; therefore, they are not discussed. In addition to the complete analyses described above, an improvement score was computed by subtracting the proportion correct at 17 msec SOA from the mean proportion correct for SOAs equal or greater than 150 msec.
EXPERIMENT 1 Experiment 1 was a test of whether focusing attention on a target location facilitates discrimination of line orientation as much as it does discrimination of line arrangement. To test this question, a location precue was used with a variable interval prior to target appearance. In this experiment, the cue was always valid; it always indicated the correct location for the target. In order to eliminate possible artifacts in interpretation, six groups of observers were tested with somewhat different conditions. Line orientation discrimination was assessed in two-alternative or fouralternative forced-choice discriminations in which the target was either a single line of two or four different orientations, or a Y-like character in which the short arm was oblique to the vertical line ( f45"). The single-line stimuli are similar to those used in texture segregation research. For example, Olson and Attneave (1 970) showed that RTs to locate texture boundaries when the stimuli were left- and right-slanted lines were much faster than the RTs to make the same judgement when the stimuli only differed in line arrangement (Ls in various orientations). Line arrangement discrimination was assessed in the same task with target characters that were composed of two line segments perpendicular to one another and conjoined to form a T-like character. This type of target has
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
ATTENTION AND ORIENTATION DISCRIMINATION
833
been shown to require serial search under certain conditions (Humphreys, Quinlan, & Riddoch, 1989). If orientation discrimination benefits less from attention than does line arrangement discrimination, then there should be less facilitation of discrimination of targets that contain lines of different orientation than of targets that do not (T-like characters). In preliminary tests it was found that, in general, slanted lines were much easier to discriminate than T-like characters. To avoid potential effects of this difference on the measurement of attention effects, in some of the groups the stimuli were approximately equated in discriminability by using more pixels (increasing total luminance) and/or longer durations for the T-like stimuli than for the slanted lines. In every condition, all of the variations of the target for one stimulus type were composed of the same number of line endings and approximately the same number of pixels, so that only the orientation of the target varied. In Croup A , the task involved a two-alternative forced-choice discrimination of slanted lines and Ts (Table 1). Large differences in accuracy as a function of cue-target SOA between discrimination of slanted lines and discrimination of Ts were found. However, this could have been due to other aspects of these targets besides the presence or absence of an orientation difference. The two types of targets differed in several ways that might influence the number of features present in the stimuli: the number of line segments, the number of line junctions, the number of line terminators, and total luminance. The first three of these features have been shown to be important in texture segregation experiments (Julesz, 1981). Therefore, in Croup B, the stimuli were composed of the same number of pixels, line endings, etc. Only the orientation of the short line of Ys or Ts varied: for Ys, the short arm was oblique to the vertical line; for Ts, the short arm was perpendicular to the vertical line. To compare Groups A and B, it was necessary to ensure that both types of stimuli were tested under identical procedures. For that reason, the same cue (a small rectangle) and mask were used for both conditions. One possible problem is that the rectangular cue could have masked the T-like figures more than it masked the slanted lines. Such a possibility is supported by existing data that showed more interference with discrimination of line orientation when noise lines were parallel than when they were perpendicular to the target (Andriessen & Bouma, 1976). Lyon (1990, Experiment l), using the present paradigm with T-like characters as targets, showed that there is a range of SOAs from - 17 msec to + 17 msec for which there may be some cue-target masking. For longer SOAs, however, performance was better with cues closer to the target (1" away) than with more distant cues (2' away), a result that is not consistent with a masking explanation. However, as existing data allow for the possibility of some cue-target masking at some SOAs, the question remains as to whether the cue might differentially mask
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
834
CHEAL. LYON, HUBBARD
the Ts versus the slanted lines. Therefore, additional experiments were conducted in which the possibility of masking effects was minimized. The peripheral cue was used in the present experiments in order to reduce delays due to deciding where to attend. However, precue effects also have been found with central cues (such as an arrow) (Eriksen & Collins, 1969; Remington, 1980; Remington & Pierce, 1984; Shulman et al., 1979) for both detection (as of light increment, Posner, Snyder, & Davidson, 1980) and discrimination (i.e. a letter, Jonides, 1981). Cheal and Lyon (1989) used both peripheral and central cues in facilitating four-alternative forced-choice discrimination of T-like figures as a function of eccentricity. The main difference in the data for these two conditions was a delay in onset of effects. The fact that similar effects on discrimination of T-like characters occur following either kind of cue suggests that these effects are not due to masking of the target by the cue, or to some other local cue-target interaction. It is doubtful that a central cue, presented at fixation, could mask the target. Therefore, in Group C , the peripheral rectangular cue was replaced by a central arrow cue. The 6" separation between target and cue should be sufficient to prevent any differential masking. In the Andriessen and Bouma study (1976), differential interference from parallel vs. perpendicular noise elements occurred at small angular distances. It was found that interference depended on eccentricity of the target such that if the target was at 6", there would be little interference if noise was more than 2.4" away. In previous studies using this paradigm (Cheal & Lyon, 1989; Lyon, 1987), it was easy to show improvement in accuracy with SOA when a central cue was used, because in those studies chance performance was 25%, and, thus, performance could improve by 75%. In Groups A, B, and C, chance performance is 50%, and therefore there is less room for improvement. Because of this fact, any delay in onset of attentional effects found with a central cue (Cheal & Lyon, this issue) in Group C could interact with the smaller possible amount of improvement, and thus, any differences between curves for slants and Ts would be smaller with the central cue than with the peripheral cue. Therefore, in order to look for larger differences between discrimination of line orientation and discrimination of line arrangement, Group D repeated the task of Group A, but with a four-alternative discrimination rather than the two-alternative discrimination that was used in the first three groups. Again, differencesbetween discrimination of lines and Ts at short SOAs in Group D could have been due to differential masking as suggested for Group A. Therefore, in Group E, the four-alternative discrimination was conducted with the foveal arrow cue as in Group C. It has been suggested that the visual system operates differently in perception of horizontal and vertical lines in comparison to perception of
ATTENTION AND ORIENTATION DISCRIMINATION
835
oblique lines (Pettigrew, Nikara, & Bishop, 1968; Vogels & Orban, 1986). Although there is no reason to assume that effects of attention would be affected by the obliqueness of the targets, some observers were tested with oblique targets. Group F also provided a final test of possible effects of cue/ target interactions that might have resulted in poorer performance at short SOAs. In this condition the lines and Ts were both rotated obliquely, but the cue was the same rectangle used for the other peripheral cue groups.
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
Procedure Group A: Two-Alternative Forced-Choice Discrimination of (1) Slanted Lines and ( 2 ) Sidewuys Ts; Peripheral Cue. An extended character set was generated in order to present the characters used as stimuli. The slanted lines were composed of a I" straight line, made of 5 pixels, slanted at an oblique angle, with the top of the line pointing either clockwise (right) or counterclockwise (left) (Figure 1A). The Ts were formed by a 0.8"horizontal line, composed of 2 rows of 4 pixels that extended either right or left from the centre of a 0.8" vertical line, composed of 8 rows of 2 pixels (total number of pixels=24; Figure 1B). The mask is illustrated in Figure 1. The three observers each completed 24 sessions of 1000 trials. For other particulars of each group, see Table 1. Group B: Two-Alternative Forced-Choice Discrimination of ( I ) Ys and ( 2 ) Sideways Ts; Peripheral Cue. Each stimulus character was composed of a vertical line of 9 pixels and a line of 4 pixels that conjoined the vertical line at the 5th pixel. For the orientation discrimination, the conjoining line led upward at an oblique angle right or left of the vertical line, forming a Ylike figure. For the line arrangement discrimination, the conjoining line was horizontal to either the right or the left of the vertical line, forming a sideways T. Each stimulus had three terminating ends. For Ys, discrimination could be made by line orientation as in Experiment I , or it could be made by the arrangement of the two conjoining line segments. For Ts, on the other hand, the only basis for discrimination was the arrangement of the two line segments. A new mask was constructed by making an outline of one pixel thickness around the compound character created by superimposing the four possible targets (right and left Ts and right and left Ys). Each observer completed 12 sessions of 1000 trials. Group C: Two-Alternative Forced-Choice Discrimination of ( 1 ) Slanted Lines and ( 2 ) Sideways Ts; Central Arrow Cue. The stimuli were the same as those used for Group A. The fixation point was removed when the cue appeared and only reappeared for the next trial. The mask was an outline of the compound stimuli. Each observer completed 20 sessions of 832 trials.
Downloaded by [University of Nebraska, Lincoln] at 22:15 30 May 2016
836
CHEAL, LYON, HUBBARD
Group D: Four-Alternative Forced-Choice Discrimination of ( I ) Lines and ( 2 ) T-like Characters; Peripheral Cue. For line orientation discrimination, two slanted lines (right & left), a vertical line (respond up), and a horizontal line (respond down) were composed of 5-pixel straight lines. For the line arrangement discrimination, the sideways Ts were constructed of a 12-pixel vertical line with a 2 x 3-pixel perpendicular line conjoining at the centre pixels of the vertical line (18 pixels). The other two Ts were composed of a 2 x 7-pixel horizontal line with a 5-pixel vertical line conjoining at the centre of the horizontal line (19 pixels). The mask was an outline of the composite of the eight possible targets. Each observer completed 8 sessions of 832 trials. Group E: Four-Alternative Forced-Choice Discrimination of ( 1 ) Slanted Lines and ( 2 ) Sideways Ts of Same Dimensions as Group D; Central Arrow Cue. The fixation point was removed when the cue appeared and only reappeared for the next trial. Each observer completed 24 sessions of 832 trials. Group F: Two-Alternative Forced-Choice Discrimination of ( I ) Horizontal and Vertical Lines and ( 2 ) Oblique Ts; Peripheral Cue. Each horizontal or vertical straight line was composed of 4 pixels. Ts were composed of a 5-pixel oblique line with 2 pixels leading up and right or down and left at approximately a right angle to the longer line. The monitor did not allow an actual right angle for these Ts, but discrimination was based on line arrangement because the lines of the two variations did not differ in orientation. The mask was an outline of the compound stimuli enclosed in a rectangle. Each observer completed 8 sessions of 832 trials.
R esu1ts Overall Analyses. The log-linear model test revealed significant effects of all variables and many significant interactions for each group. As expected, proportion of correct responses was well above chance: for example, for Group A, where chance was 0.50, total probability correct was 0.76; x2(1)=21,631, p
