Perceptual and Motor SkilIs, 1991, 73, 147-161. O Perceptual and Motor Skills 1991
BINARY-CHOICE DECISION TIME DEPENDS UPON CEREBRAL HEMISPHERE AND NATURE O F TASK LAURENCE R. SCHWEITZER Universiq of Texas Medical Branch at Galveston Summary.-Binary-choice paradigms are classificatory problems of basic importance to the understanding of elementary decision processes. Generally when subjects decide if two visual stimuli are identical or differ by as Little as one element, the decision of "Different" takes longer. This finding is unexpected as decisions of "Different" should not require an exhaustive matching of elements. Using stimulus presentation to the right and left cerebral hemispheres, the right hemisphere initiated fast selections of "Same" for figural material and alone was responsible for the "Same"/ "Different" response differential. Exp. 1 (n = 22) gave no differences for same-different, unilateral-bilateral stimulation, and left-right hemispheres. Exp. 2, using word meaning as the binary-choice task, also showed faster decisions for "Same" but a different left-hemisphere-dependent strategy. The nature of information processing in relation to binary-choice tasks is discussed and the utility of bihemispheric paradigms is demonstrated.
Binary-choice decisions are elementary tasks on which subjects decide whether two or more stimuli (generally presented in the visual mode) are identical or have the same meaning. These tasks require that subjects partition stimuli into mutually exclusive and exhaustive classes in some specified time. Because they are simple and easy to manipulate, binary-choice tasks have been widely used to study elementary decision processes. By varying the number of elements or the attributes upon which decisions depend, one can measure the information capacity of a single glance, decision time as a function of item number, the relationship of attribute to decision time, and the trade-off between accuracy and speed. In such experiments, response latency for "Same" choices is generally shorter than "Different" ("Yes" faster than "No") (Audley, 1973; Falmagne, 1965; Falrnagne, Cohen, & Dwividi, 1975; Nickerson, 1969, 1971, 1972, 1973; Ollman, 1966; Proctor, 1981; Proctor & Rao, 1983; Yellott, 1967, 1971). This "Same"-"Different" differential is puzzling since judgments of "Same" would appear to require an examination of all stimulus elements, whereas the selection of "Different" may occur after the first recognized difference. In 1973 Audley dichotomized models of this response differential into (1) mixed processes (i.e., both serial and parallel modes of processing) or (2) statistical sampling processes (i.e., decisions based upon a continuous flow of data moving toward a "Same" or "Different" boundary). As examples of the 'Address correspondence to Laurence Schweitzer, Department of Ps chiatry and Behavioral Sciences, University of Texas Medical Branch at Galveston, Galveston, $X 77550-2774.
148
L. R. SCHWEITZER
mixed model, Ollman (1966) and Yellott (1967, 1971) . proposed a two-state noise-free decision process (i.e., one in which the stimulus is represented centrally without distortion). In one state, subjects are assumed to respond only after recognizing a stimulus, whereas in the second state, subjects make fast guesses which, more often than not, are "Same." Basically, this model predicts a reduction in "Same" latency along with an increase in "Same" errors. In a related model, Falmagne (1965) and Falmagne, et a/. (1975) suggested that subjects tend to expect "Same" stimuli. This decreases latency for the expected same stimulus, while prolonging it for the unexpected different stimulus. In this model no predictions are apparent for errors. Audey's (1973) second class of model, a statistical sampling process (Nickerson, 1969, 1971, 1972; Stone, 1960), is independent of guesses or an expectation bias to generate a response differential. Here, a single processor is assigned the task of cumulating differences in a register until a predetermined number is or is not attained in an allotted time. A fundamental assumption was that internal noise produced random distortions in the mental representation of stimuli introducing additional differences or converting same stimuli into different percepts. To account for this systematic error (i.e., increased frequency of "Different" errors) the stimulus or its mental representation are rechecked by repetitive scanning (Chignell & Krueger, 1984; Krueger, 1973, 1978; Krueger & Chignell, 1985; Stone, 1960). While the predetermined number of differences necessary to initiate rechecking is arbitrary, clearly, any rechecking will increase latency for "Different" responses. Thus, when subjects are time limited (i.e., restricted to one glance), this model predicts an increase in "Different" errors along with an increase in latency for "Different" selections. The validity of either class of model remains undecided, but since each predicts a different and measurable error, experimental verification would appear possible. Unfortunately, this task is complicated since binary-choice tasks are associated with an inverse speed for accuracy trade-off when subjects work close to their maximum accuracy. To minimize these fast response errors, most studies employ well learned and rather simple tasks to decrease the dependence of accuracy upon speed. This practice, however, attenuates the value of an error analysis. Evaluation of errors has yielded inconclusive results. The prediction of fast "Same" guesses by mixed models has not been borne out (Ratcliff & Hacker, 1981; Swensson, 1972). More important, sequential effects have been observed, i.e., latency to a stimulus is faster if the stimulus is of the same type as the preceding one (Krueger, 1973; Nickerson, 1973). These findings are incompatible with a simple two-state description of response behavior and have led to revisions of mixed models. Conversely, no studies have shown increased latency for correct "Different" responses along with a con-
BINARY-CHOICE DECISION TIME
149
current increase in "Different" errors as required by statistical sampling models. Attention has also focused on the processes involved in stimulus matching. Again, there is controversy as to whether the response differential implies separate processors for each decision type or a single processor with variable set points for "Same" and "Different." The mixed models of Bamber (1969) and Taylor (1976) require two processors, with responses to "Same" made on the basis of a fast but dirty holistic match and responses of "Different" by a slower comparison of item-sequences. In contrast, Krueger (1973, 1978), Krueger and Chignell (1985), and Chignell and Krueger (1984) have proposed a single-processor model. Here, data are registered in shortterm memory, and successive independent looks or passes of about 200 msec. duration are made of these data. Individual features of the stimulus pattern are compared by overlap (an holistic process), and differences are cumulated without error over successive trials. Krueger (1978) specifically attributes the response differential to slower decisions of "Different" that arise from rechecking. F i n d y , Ratcliff (1985) has proposed a random walk accumulator (diffusion) model with variable set points for "Same" and "Different." This relatively complicated single-processor model contains three independent parameters, set by the subject, which simulate a changing stimulus expectation (i.e., a bias condition) by raising or lowering either decision threshold. While Ratcliff's model (1981, 1985) has decisional similarities to mixed models and does not rely on rechecking, its single processor approach is most fundamentally similar to Krueger's (1978) model. Single-processor models can be tested directly by examining binarychoice latencies for (1) decisions made within each cerebral hemisphere and (2) decisions that require the transfer of data between the hemispheres. Single-processor models predict: (I) a response differential in the two cerebral hemispheres, (2) relatively faster and more accurate decisions of "Same" and "Different" in the right hemisphere because that hemisphere shows an inductive and holistic propensity (Levy-Agresti & Sperry, 1968; Levy-Agresti, Nebes, & Sperry, 1971; Nebes, 1972), (3) decisions of "Different" w d be longer than those of "Same" in both hemispheres, and (4) more frequent incorrect responses of "Different" than incorrect ones of "Same." If a response differential is lateralized, single-processor models predict increased latency for "Different" in one hemisphere; latencies for "Different" equal to those for "Same" in the other hemisphere, and responses of "Same" equal in both hemispheres. Single-processor models also predict that bilateral hemifield presentations will take longer than unilateral presentations. In this model, a bilateral decision requires: (1) data transmission from the hemisphere contralateral to
150
L. R. SCHWEITZER
the output side, (2) the formation of a composite stimulus in short-term memory, and (3) the scanning of the memory store. On the other hand, mixed models do not depend upon a process of comparative mapping in short-term memory. Thus, categorical information may be extracted prior to callosal transfer, making mixed models potentially faster for bilateral than for unilateral presentations. In this study, unilateral and bilateral visual-hemifield presentations of shapes were used to test the validity of single-processor models. The working hypotheses were (1) that the right hemisphere would be faster than the left one in determining "Same" and "Different," (2) a response differential would be present in one or both cerebral hemispheres, (3) if the hemispheres differed in response time, responses of "Different" would be longer than those to "Same" in one hemisphere while responses of "Same" would be equal in both hemispheres and equal to "Different" in the hemisphere with the briefer over-all latency, and (4) bilateral presentations would be longer than unilateral presentations. METHOD Twenty-two dextral, healthy subjects (eleven of whom were men) had normal or corrected vision; all were medical students, nurses, or allied health professionals, and participated in this study after informed consent was obtained as required by the Baylor College of Medicine and The Methodist Hospital human subject research committees. Handedness was established on the basis of the Edinburgh Inventory (Oldfield, 1971). The experimental task (see Fig. 1) required that subjects determine whether three shapes, projected onto a cathode ray tube for 150 msec., were identical. The shapes were triangles, squares, and rectangles. The stimulus set contained eight arrangements of these shapes, four with identical elements and four with a single differing shape. These stimuli were presented in one of two ways, (1) three elements in one visual hemifield, i.e., a unilateral (within-hemisphere) presentation or (2) two shapes in one visual hemifield and one in the field on the opposite side, i.e., a bilateral (between-hemispheres) presentation. When a different element occurred in a bilateral presentation, it was always alone and required transfer to the output hemisphere prior to a decision. An Apple IIe computer displayed the stimuli (4" to 6" off visual fixation) on a fast decay phosphor CRTenclosed in a baffle box, situated 36 in. from the subject's forehead. Images from the left half of the CRT were projected to a subject's right hemiretina and thence to the right visual cortex and vice versa. Luminances could not be measured accurately so they are not reported. A detailed description of the apparatus is given by Schweitzer (1982). Each stimulus presentation was preceded by a 1-sec. preparatory interval
BINARY-CHOICE DECISION TIME
151
of visual fixation to an illuminated square in the center of the CRT. A dark period of random duration (50 to 250 msec.) then preceded a stimulus presentation of 150 msec. B WRH A WLH I
FIG.1. Schematic depiction of the experimental design. Frames A and B illustrate within hemisphere (WLH, WRH) presentations while Frames C and D indicate the transfer of information via the corpus callosum from right to left (BRL) and left to right (BLR). Frames A and C depict the first and third blocks of presentations and B and D the second and fourth blocks under each condition.
The complete experiment consisted of four conditions. The first, a farmliarization trial called normal pace, allowed subjects 1.0 sec. to register a decision. Under the next condition, accuracy, subjects were told that accuracy was three times as important as speed in a score that the computer was tallying. Subjects were d o w e d 1.5 sec. to respond and were encouraged to take their time to be accurate. In the third condition, speed was three times as important as accuracy, and response time was halved to 0.75 sec. The last condition required speed and accuracy. Subjects were encouraged to be as fast and as accurate as possible and to balance each equally. Again, response time was 0.75 sec. Each condition consisted of four blocks of 16 presentations. Subjects began each condition using the right hand. Decisions of "Same" and "Different" were registered by pushing a microswitch forward or backward, respectively. Subjects alternated hands for each block.
152
L. R. SCHWEITZER
Half of all trials were bilateral presentations, and half of these included a different shape which appeared alone and was projected to the output hemisphere, i.e., controlling the hand used (see Fig. 1). Half of the bilateral presentations required information to be transmitted from right to left and the remaining 25% from left to right. Thus, the first block of 16 presentations under t h e first condition measured response time for unilateral and bilateral left-hemisphere presentations (see Fig. 1A and C). Similarly, the second block of presentations used the left hand and right hemisphere (see Fig. 1B and D). The third and fourth blocks duplicated this sequence and completed the first condition, normal-pace. Between blocks subjects rested for several minutes and then for five minutes between conditions. The entire experiment (256 trials) occurred in one session and required 50 to 60 minutes to complete. When a response exceeded the time limit, the message That was too slow, please try harder was flashed and the trial replaced. Responses less than 130 msec. were also discarded as anticipatory; the subject was instructed Try to relax a bit, you are anticipating, and the trial was replaced. Following completion of the four conditions, subjects performed a test of simple reaction time. The appearance of a square of light initiated an index-finger lift from a telegraph key positioned in the midline at the same level as the switches. Again, stimuli appeared randomly in both visual fields and hands were alternated, providing latency measures for the left and right hemispheres under u d a t e r a l and bilateral presentation. There were 16 trials for each condition. It is important to note that spatial stimuli presented to the left hemisphere may actually be transferred for processing to the spatially more proficient right hemisphere. Thus, a strategy was required to decrease the likelihood of such transfers. This was accomplished by selecting an easy task and placing a premium on speed. Under these conditions any decision strategy involving data transfer to the right hemisphere for processing imposes an unacceptable latency overhead of two callosal transfers.
Shape Similarities The primary statistical tool was a fixed-model analysis of variance with repeated measures on each of the main effects. All computations were performed using SASISTAT Version 6 with post hoc t tests reported only when significant effects were identified by an analysis of variance. Simple reaction time.-The assumption that response Iatency may be decomposed into the temporal sum of (1) stimulus registration, (2) central processing, and (3) motor output was central to this study (Sternberg, 1969). However, there is no a priori reason to assume that these components are
BINARY-CHOICE DECISION TIME
153
balanced equally between hemispheres. Since reaction time (i.e., finger lift latency) measures two of these components (stimulus registration + motor output), an analysis of variance was performed on these data. No significant main or interactive effects were found for left, right, unilateral, or bilateral hemisphere presentations. Thus, reaction time was not biased, and a response differential will necessarily originate from central processing. An analysis of variance was computed for the raw and log transformed latencies. Subjects (n = 22), "same"-"different," udateral-bilateral, and leftright (hemispheres) were independent main factors. As the results of both analyses were statistically indistinguishable, the untransformed data are presented throughout. Error analysis.-To answer the question "Did fast guesses of "Same" generate a "SameH-"Different" response differential?", one must compare latencies of correct with incorrect responses. Unfortunately, the small (8%) over-all error rate left many subjects with no errors to compare with correct responses. As an alternative, correct responses were compared with responses containing errors (i.e., all responses). An analysis of variance failed to disclose a group difference or an interaction with "Samew-"Different," unilateral-bilateral, or left-right. In contrast to earlier work (Bamber, 1969; Falmagne, 1965; Falrnagne, et al., 1975; Yellott, 1967, 19711, mean error latency (480 msec.) was not shorter than mean time for correct responses (480 msec.) . The differential expectation and statistical sampling models also predict specific patterns of errors. Thus, the frequency of each type of error was examined by an analysis of variance using percent correct as the dependent variable; see Table 1. A nonsignificant 3% increase in errors was found for different stimuli, i.e., incorrect responses of "Same" (F,,,,= 3.03, p = .09). However, this €inding conflicts with the statistical sampling model which predicted increased errors for same stimuli (i.e., increased incorrect responses of "Different"). O n the other hand, an increase in errors of "Same" is consistent with the differential expectation or hypothesis about a fast guess of "Same", if guesses of "Same" were actually fast. Since mean error latency was not faster than the mean time for correct responses (both were 481 msec.), errors cannot account for a response differential. Latency.-Responses of "Same" were faster than those of "Different" over-all and under normal pace and accuracy conditions; see Table 1. The fact that this differential was not significant under speed and the combined speed and accuracy conditions probably reflects the learning of a simple and undemanding task. Left-right, the decision and output hemisphere variable, was also significant, with the right hemisphere being faster than the left one over-all and under normal pace. Finally, the condition variable was signifi-
154
L. R. SCHWEITZER TABLE 1 SHAPEDETERMINATION
AU
Normal Pace
Accuracy
Speed
Speed and Accuracy
480 96 4671492 186/201 4881472 1971192
509 90 4911527 881 90 5291489 901 87
647 149 6271666 1501146 6601634 1441153
375 70 3701379 731 66 3721377 661 73
388 73 3831394 721 74 3901386 741 72
4781482 1821202
5131504 931 89
6521642 1531145
3731376 661 74
3881388 691 76
< .03 ns
ns ns
ns ns
ns
ns
ns
ns
BINARY-CHOICE DECISION TIME DEPENDS UPON CEREBRAL HEMISPHERE AND NATURE O F TASK LAURENCE R. SCHWEITZER Universiq of Texas Medical Branch at Galveston Summary.-Binary-choice paradigms are classificatory problems of basic importance to the understanding of elementary decision processes. Generally when subjects decide if two visual stimuli are identical or differ by as Little as one element, the decision of "Different" takes longer. This finding is unexpected as decisions of "Different" should not require an exhaustive matching of elements. Using stimulus presentation to the right and left cerebral hemispheres, the right hemisphere initiated fast selections of "Same" for figural material and alone was responsible for the "Same"/ "Different" response differential. Exp. 1 (n = 22) gave no differences for same-different, unilateral-bilateral stimulation, and left-right hemispheres. Exp. 2, using word meaning as the binary-choice task, also showed faster decisions for "Same" but a different left-hemisphere-dependent strategy. The nature of information processing in relation to binary-choice tasks is discussed and the utility of bihemispheric paradigms is demonstrated.
Binary-choice decisions are elementary tasks on which subjects decide whether two or more stimuli (generally presented in the visual mode) are identical or have the same meaning. These tasks require that subjects partition stimuli into mutually exclusive and exhaustive classes in some specified time. Because they are simple and easy to manipulate, binary-choice tasks have been widely used to study elementary decision processes. By varying the number of elements or the attributes upon which decisions depend, one can measure the information capacity of a single glance, decision time as a function of item number, the relationship of attribute to decision time, and the trade-off between accuracy and speed. In such experiments, response latency for "Same" choices is generally shorter than "Different" ("Yes" faster than "No") (Audley, 1973; Falmagne, 1965; Falrnagne, Cohen, & Dwividi, 1975; Nickerson, 1969, 1971, 1972, 1973; Ollman, 1966; Proctor, 1981; Proctor & Rao, 1983; Yellott, 1967, 1971). This "Same"-"Different" differential is puzzling since judgments of "Same" would appear to require an examination of all stimulus elements, whereas the selection of "Different" may occur after the first recognized difference. In 1973 Audley dichotomized models of this response differential into (1) mixed processes (i.e., both serial and parallel modes of processing) or (2) statistical sampling processes (i.e., decisions based upon a continuous flow of data moving toward a "Same" or "Different" boundary). As examples of the 'Address correspondence to Laurence Schweitzer, Department of Ps chiatry and Behavioral Sciences, University of Texas Medical Branch at Galveston, Galveston, $X 77550-2774.
148
L. R. SCHWEITZER
mixed model, Ollman (1966) and Yellott (1967, 1971) . proposed a two-state noise-free decision process (i.e., one in which the stimulus is represented centrally without distortion). In one state, subjects are assumed to respond only after recognizing a stimulus, whereas in the second state, subjects make fast guesses which, more often than not, are "Same." Basically, this model predicts a reduction in "Same" latency along with an increase in "Same" errors. In a related model, Falmagne (1965) and Falmagne, et a/. (1975) suggested that subjects tend to expect "Same" stimuli. This decreases latency for the expected same stimulus, while prolonging it for the unexpected different stimulus. In this model no predictions are apparent for errors. Audey's (1973) second class of model, a statistical sampling process (Nickerson, 1969, 1971, 1972; Stone, 1960), is independent of guesses or an expectation bias to generate a response differential. Here, a single processor is assigned the task of cumulating differences in a register until a predetermined number is or is not attained in an allotted time. A fundamental assumption was that internal noise produced random distortions in the mental representation of stimuli introducing additional differences or converting same stimuli into different percepts. To account for this systematic error (i.e., increased frequency of "Different" errors) the stimulus or its mental representation are rechecked by repetitive scanning (Chignell & Krueger, 1984; Krueger, 1973, 1978; Krueger & Chignell, 1985; Stone, 1960). While the predetermined number of differences necessary to initiate rechecking is arbitrary, clearly, any rechecking will increase latency for "Different" responses. Thus, when subjects are time limited (i.e., restricted to one glance), this model predicts an increase in "Different" errors along with an increase in latency for "Different" selections. The validity of either class of model remains undecided, but since each predicts a different and measurable error, experimental verification would appear possible. Unfortunately, this task is complicated since binary-choice tasks are associated with an inverse speed for accuracy trade-off when subjects work close to their maximum accuracy. To minimize these fast response errors, most studies employ well learned and rather simple tasks to decrease the dependence of accuracy upon speed. This practice, however, attenuates the value of an error analysis. Evaluation of errors has yielded inconclusive results. The prediction of fast "Same" guesses by mixed models has not been borne out (Ratcliff & Hacker, 1981; Swensson, 1972). More important, sequential effects have been observed, i.e., latency to a stimulus is faster if the stimulus is of the same type as the preceding one (Krueger, 1973; Nickerson, 1973). These findings are incompatible with a simple two-state description of response behavior and have led to revisions of mixed models. Conversely, no studies have shown increased latency for correct "Different" responses along with a con-
BINARY-CHOICE DECISION TIME
149
current increase in "Different" errors as required by statistical sampling models. Attention has also focused on the processes involved in stimulus matching. Again, there is controversy as to whether the response differential implies separate processors for each decision type or a single processor with variable set points for "Same" and "Different." The mixed models of Bamber (1969) and Taylor (1976) require two processors, with responses to "Same" made on the basis of a fast but dirty holistic match and responses of "Different" by a slower comparison of item-sequences. In contrast, Krueger (1973, 1978), Krueger and Chignell (1985), and Chignell and Krueger (1984) have proposed a single-processor model. Here, data are registered in shortterm memory, and successive independent looks or passes of about 200 msec. duration are made of these data. Individual features of the stimulus pattern are compared by overlap (an holistic process), and differences are cumulated without error over successive trials. Krueger (1978) specifically attributes the response differential to slower decisions of "Different" that arise from rechecking. F i n d y , Ratcliff (1985) has proposed a random walk accumulator (diffusion) model with variable set points for "Same" and "Different." This relatively complicated single-processor model contains three independent parameters, set by the subject, which simulate a changing stimulus expectation (i.e., a bias condition) by raising or lowering either decision threshold. While Ratcliff's model (1981, 1985) has decisional similarities to mixed models and does not rely on rechecking, its single processor approach is most fundamentally similar to Krueger's (1978) model. Single-processor models can be tested directly by examining binarychoice latencies for (1) decisions made within each cerebral hemisphere and (2) decisions that require the transfer of data between the hemispheres. Single-processor models predict: (I) a response differential in the two cerebral hemispheres, (2) relatively faster and more accurate decisions of "Same" and "Different" in the right hemisphere because that hemisphere shows an inductive and holistic propensity (Levy-Agresti & Sperry, 1968; Levy-Agresti, Nebes, & Sperry, 1971; Nebes, 1972), (3) decisions of "Different" w d be longer than those of "Same" in both hemispheres, and (4) more frequent incorrect responses of "Different" than incorrect ones of "Same." If a response differential is lateralized, single-processor models predict increased latency for "Different" in one hemisphere; latencies for "Different" equal to those for "Same" in the other hemisphere, and responses of "Same" equal in both hemispheres. Single-processor models also predict that bilateral hemifield presentations will take longer than unilateral presentations. In this model, a bilateral decision requires: (1) data transmission from the hemisphere contralateral to
150
L. R. SCHWEITZER
the output side, (2) the formation of a composite stimulus in short-term memory, and (3) the scanning of the memory store. On the other hand, mixed models do not depend upon a process of comparative mapping in short-term memory. Thus, categorical information may be extracted prior to callosal transfer, making mixed models potentially faster for bilateral than for unilateral presentations. In this study, unilateral and bilateral visual-hemifield presentations of shapes were used to test the validity of single-processor models. The working hypotheses were (1) that the right hemisphere would be faster than the left one in determining "Same" and "Different," (2) a response differential would be present in one or both cerebral hemispheres, (3) if the hemispheres differed in response time, responses of "Different" would be longer than those to "Same" in one hemisphere while responses of "Same" would be equal in both hemispheres and equal to "Different" in the hemisphere with the briefer over-all latency, and (4) bilateral presentations would be longer than unilateral presentations. METHOD Twenty-two dextral, healthy subjects (eleven of whom were men) had normal or corrected vision; all were medical students, nurses, or allied health professionals, and participated in this study after informed consent was obtained as required by the Baylor College of Medicine and The Methodist Hospital human subject research committees. Handedness was established on the basis of the Edinburgh Inventory (Oldfield, 1971). The experimental task (see Fig. 1) required that subjects determine whether three shapes, projected onto a cathode ray tube for 150 msec., were identical. The shapes were triangles, squares, and rectangles. The stimulus set contained eight arrangements of these shapes, four with identical elements and four with a single differing shape. These stimuli were presented in one of two ways, (1) three elements in one visual hemifield, i.e., a unilateral (within-hemisphere) presentation or (2) two shapes in one visual hemifield and one in the field on the opposite side, i.e., a bilateral (between-hemispheres) presentation. When a different element occurred in a bilateral presentation, it was always alone and required transfer to the output hemisphere prior to a decision. An Apple IIe computer displayed the stimuli (4" to 6" off visual fixation) on a fast decay phosphor CRTenclosed in a baffle box, situated 36 in. from the subject's forehead. Images from the left half of the CRT were projected to a subject's right hemiretina and thence to the right visual cortex and vice versa. Luminances could not be measured accurately so they are not reported. A detailed description of the apparatus is given by Schweitzer (1982). Each stimulus presentation was preceded by a 1-sec. preparatory interval
BINARY-CHOICE DECISION TIME
151
of visual fixation to an illuminated square in the center of the CRT. A dark period of random duration (50 to 250 msec.) then preceded a stimulus presentation of 150 msec. B WRH A WLH I
FIG.1. Schematic depiction of the experimental design. Frames A and B illustrate within hemisphere (WLH, WRH) presentations while Frames C and D indicate the transfer of information via the corpus callosum from right to left (BRL) and left to right (BLR). Frames A and C depict the first and third blocks of presentations and B and D the second and fourth blocks under each condition.
The complete experiment consisted of four conditions. The first, a farmliarization trial called normal pace, allowed subjects 1.0 sec. to register a decision. Under the next condition, accuracy, subjects were told that accuracy was three times as important as speed in a score that the computer was tallying. Subjects were d o w e d 1.5 sec. to respond and were encouraged to take their time to be accurate. In the third condition, speed was three times as important as accuracy, and response time was halved to 0.75 sec. The last condition required speed and accuracy. Subjects were encouraged to be as fast and as accurate as possible and to balance each equally. Again, response time was 0.75 sec. Each condition consisted of four blocks of 16 presentations. Subjects began each condition using the right hand. Decisions of "Same" and "Different" were registered by pushing a microswitch forward or backward, respectively. Subjects alternated hands for each block.
152
L. R. SCHWEITZER
Half of all trials were bilateral presentations, and half of these included a different shape which appeared alone and was projected to the output hemisphere, i.e., controlling the hand used (see Fig. 1). Half of the bilateral presentations required information to be transmitted from right to left and the remaining 25% from left to right. Thus, the first block of 16 presentations under t h e first condition measured response time for unilateral and bilateral left-hemisphere presentations (see Fig. 1A and C). Similarly, the second block of presentations used the left hand and right hemisphere (see Fig. 1B and D). The third and fourth blocks duplicated this sequence and completed the first condition, normal-pace. Between blocks subjects rested for several minutes and then for five minutes between conditions. The entire experiment (256 trials) occurred in one session and required 50 to 60 minutes to complete. When a response exceeded the time limit, the message That was too slow, please try harder was flashed and the trial replaced. Responses less than 130 msec. were also discarded as anticipatory; the subject was instructed Try to relax a bit, you are anticipating, and the trial was replaced. Following completion of the four conditions, subjects performed a test of simple reaction time. The appearance of a square of light initiated an index-finger lift from a telegraph key positioned in the midline at the same level as the switches. Again, stimuli appeared randomly in both visual fields and hands were alternated, providing latency measures for the left and right hemispheres under u d a t e r a l and bilateral presentation. There were 16 trials for each condition. It is important to note that spatial stimuli presented to the left hemisphere may actually be transferred for processing to the spatially more proficient right hemisphere. Thus, a strategy was required to decrease the likelihood of such transfers. This was accomplished by selecting an easy task and placing a premium on speed. Under these conditions any decision strategy involving data transfer to the right hemisphere for processing imposes an unacceptable latency overhead of two callosal transfers.
Shape Similarities The primary statistical tool was a fixed-model analysis of variance with repeated measures on each of the main effects. All computations were performed using SASISTAT Version 6 with post hoc t tests reported only when significant effects were identified by an analysis of variance. Simple reaction time.-The assumption that response Iatency may be decomposed into the temporal sum of (1) stimulus registration, (2) central processing, and (3) motor output was central to this study (Sternberg, 1969). However, there is no a priori reason to assume that these components are
BINARY-CHOICE DECISION TIME
153
balanced equally between hemispheres. Since reaction time (i.e., finger lift latency) measures two of these components (stimulus registration + motor output), an analysis of variance was performed on these data. No significant main or interactive effects were found for left, right, unilateral, or bilateral hemisphere presentations. Thus, reaction time was not biased, and a response differential will necessarily originate from central processing. An analysis of variance was computed for the raw and log transformed latencies. Subjects (n = 22), "same"-"different," udateral-bilateral, and leftright (hemispheres) were independent main factors. As the results of both analyses were statistically indistinguishable, the untransformed data are presented throughout. Error analysis.-To answer the question "Did fast guesses of "Same" generate a "SameH-"Different" response differential?", one must compare latencies of correct with incorrect responses. Unfortunately, the small (8%) over-all error rate left many subjects with no errors to compare with correct responses. As an alternative, correct responses were compared with responses containing errors (i.e., all responses). An analysis of variance failed to disclose a group difference or an interaction with "Samew-"Different," unilateral-bilateral, or left-right. In contrast to earlier work (Bamber, 1969; Falmagne, 1965; Falrnagne, et al., 1975; Yellott, 1967, 19711, mean error latency (480 msec.) was not shorter than mean time for correct responses (480 msec.) . The differential expectation and statistical sampling models also predict specific patterns of errors. Thus, the frequency of each type of error was examined by an analysis of variance using percent correct as the dependent variable; see Table 1. A nonsignificant 3% increase in errors was found for different stimuli, i.e., incorrect responses of "Same" (F,,,,= 3.03, p = .09). However, this €inding conflicts with the statistical sampling model which predicted increased errors for same stimuli (i.e., increased incorrect responses of "Different"). O n the other hand, an increase in errors of "Same" is consistent with the differential expectation or hypothesis about a fast guess of "Same", if guesses of "Same" were actually fast. Since mean error latency was not faster than the mean time for correct responses (both were 481 msec.), errors cannot account for a response differential. Latency.-Responses of "Same" were faster than those of "Different" over-all and under normal pace and accuracy conditions; see Table 1. The fact that this differential was not significant under speed and the combined speed and accuracy conditions probably reflects the learning of a simple and undemanding task. Left-right, the decision and output hemisphere variable, was also significant, with the right hemisphere being faster than the left one over-all and under normal pace. Finally, the condition variable was signifi-
154
L. R. SCHWEITZER TABLE 1 SHAPEDETERMINATION
AU
Normal Pace
Accuracy
Speed
Speed and Accuracy
480 96 4671492 186/201 4881472 1971192
509 90 4911527 881 90 5291489 901 87
647 149 6271666 1501146 6601634 1441153
375 70 3701379 731 66 3721377 661 73
388 73 3831394 721 74 3901386 741 72
4781482 1821202
5131504 931 89
6521642 1531145
3731376 661 74
3881388 691 76
< .03 ns
ns ns
ns ns
ns
ns
ns
ns
