Documenta Ophthalmologica 99: 103–121, 1999. © 2000 Kluwer Academic Publishers. Printed in the Netherlands.
Saccadic latency during perceptual processing and sequence learning JAMES G. MAY, MARIA L. BERG and LAURIE A. ZEBLEY Department of Psychology, University of New Orleans, Lakefront, New Orleans, LA 70148, USA Accepted 19 January 2000
Abstract. The difference between simple and choice manual reaction time (RT) has been taken to be a measure of the time necessary for various cognitive operations. In contrast, simple and choice saccadic latencies (SL – time elapsing from stimulus onset to saccade initiation) are quite similar, suggesting that such responses may be more automated. In the present investigation, SL and saccadic reaction times (SR – time elapsing from stimulus onset to saccade completion) were measured for targets appearing in the same and different locations, and to different ends of compound stimuli (big arrows) composed of small elements (little arrows) using either the global figure or the local elements as indicators of required saccade direction. In addition, measures of sequence learning were obtained behaviorally over iterative trials (decreases in response time) and with post-test interrogation. The results indicated that local response times were significantly slower than choice or global response times. Both global precedence and consistency effects were observed. Robust sequence learning was observed under the local condition, but only in the choice condition were all subjects able to recall the sequence correctly. These results are discussed in terms of proposed models of visual perception and saccade generation based on parallel processing. Key words: consistency effects, global precedence, saccadic latency, sequential learning, reaction time
Introduction Many patients with multiple sclerosis develop oculomotor dysfunction at some stage throughout the course of the disease. Clinical oculography has been employed to characterize these anomalies and to aid in the diagnosis of this disorder (Mastaglia, Black and Collins, 1979; Mastaglia, Black, Theikbroom and Collins, 1982; Reulen, Sanders and Hogenhuis, 1983; Meienberg, Muri and Robineau, 1986; Flipse, Staathof, Van der Steen, Van Leeuwen, Van Doorn, Van der Meche and Collewijn, 1997). The tasks employed to elicit eye movements have been quite simplistic, however, and may not take advantage of the growing knowledge of neuro-physiological substrates of human performance. The present investigation with normal subjects reveals the
104 considerable variability in saccadic latency that occurs across various tasks and reviews our understanding of the neurological pathways involved. Donders (1868) used manual RT in an attempt to measure the time necessary for mental processes such as detection, discrimination and choice. He employed an experimental paradigm in which different tasks were required, beginning with the presentation of a stimulus and ending with the performance of a response. In this way the latency between the stimulus and the response would be a measure of the time required for various mental processes. He compared three tasks: simple (detection) RT (react to a single, repeated stimulus), choice RT (e.g. react to red with the right hand, to blue with the left hand), and go/no-go (discrimination) RT (e.g. react to red, but not blue). In contrast to the simple RT experiment, in which subjects responded to a single stimulus that always appeared in the same place, the choice RT experiment was designed such that the subject did not know where the target stimulus would occur and thus had to choose the correct response based on a number of stimuli or stimulus locations. Donders found that choice RT was about 100 msec longer than simple RT (and slightly longer than the go/no-go RT) and regarded this difference as the time required for the mental processes of choice and discrimination. It has also been shown (Merkel, 1885) that RT increases as a function of the number of alternatives. While these basic findings have been found to be reliable and ubiquitous for manual responses, eye movement research suggests that some choice tasks might be mediated by processes that operate below the level of conscious awareness (Maljkovic and Nakayama, 1994; Lambert and Sumich, 1996). Since this early work, great strides have been made in our understanding of what parts of the brain are involved in various forms of cognitive processing and how such processing is manifest in reflexive and volitional eye movements. Fast choice RTs The simple/choice difference in RT has been found for most motor responses, except for: shadowing of speech (Stetson, 1951; Netsell and Daniel, 1974; Fry, 1975) and eye movements (Saslow, 1967b). Saslow (1967b) found that choice SLs for four alternatives were about the same as simple SLs for only one alternative and SL was unaffected by differences in the probability that the stimulus would occur at a given location. Thus, some eye movements have been thought to be reflexive and not to require such mental processes as discrimination and choice. Yet, it is clear that volitional control of eye movements is also possible and employed readily in tasks like visual search (Gould, 1973) and reading (Rayner, 1995). Under these conditions, SL is heavily influenced by the information processing load in the fovea.
105 To say that some saccades are reflexive in the classical sense of the word implies that they occur without volition. Yet it is difficult to define such a saccade in conscious subjects after the subject has been instructed to fixate targets appearing in various locations. The subject has been given a purpose and is voluntarily (not reflexively) cooperating. Saslow (1967a) found that saccade latency was significantly less when a fixation point was extinguished 100–200 msec prior to target presentation, suggesting that having something to look at in one location, impedes the ability to look at something which appears in another location. While Ross and Ross (1980, 1981) point out that this could simply be a warning stimulus effect, others (Reuter-Lorenz et al., 1991, 1995; Kingstone and Klein, 1993; Forbes and Klein, 1996) suggest that this ‘gap’ effect is composed of two components: one serving a warning function (as in manual RT), but another having to do with release from ocular fixation. Two recent reports examining the ‘gap’ effect attempt to differentiate saccades as exogenous (reflexive) or endogenous (voluntary) based on the paradigm used (Forbes and Klein, 1996; Abrams, Oonk and Pratt, 1998). Exogenous saccades are seen as stimulus driven saccades to the onset of a visual stimulus in the visual field (however, subjects are typically instructed as to the task rules). Endogenous saccades are seen as saccades made to different continuously depicted locations according to a cognitively defined rule (e.g. Look at the left box if you hear a high tone and the right box if you hear a low tone). Despite the fact that subjects volitionally comply with the instruction, exogenous saccades are found to be significantly faster (∼70 msec) than endogenous saccades. Thus, it appears that both fixation offset and target onset can result in faster saccades, even if the target location is unpredictable. The global and consistency effects Manual choice RTs have been found to vary depending on the task and the visual stimuli involved. Navon (1977, 1981) had subjects respond to complex stimuli (e.g. large letters made up of smaller letters) which were either consistent (e.g. large H made of small Hs) or inconsistent (e.g. large H made of small Ss). He found that when subjects were asked to attend to the global level (large compound stimuli) RTs are faster than those obtained when they attend to the local level (small component stimuli). He named this phenomenon global precedence. He also found what has become known as the consistency effect. This is the finding that when attending to local stimuli, RTs are slower when the global and local stimuli are inconsistent, but this does not occur when attending to the global stimuli. Much of the research pertaining to global precedence has explored how manipulation of the stimuli affects responding. The advantage of the global elements has been found to diminish
106 with sparsity of local elements (Martin, 1979), and high pass spatial filtering (Hughes, Layton, Baird and Lester, 1984; Badcock, Whitworth, Badcock and Lovegrove, 1990; LaGasse, 1993). These manipulations result in reductions of low spatial frequency amplitude. In addition, whole field flicker was found to diminish the global precedence effect by slowing responses to global elements (Lovegrove, Lehmkuble, Baro and Garzia, 1990). Badcock et al. argue that both the global precedence and consistency effects derive from the same mechanism, but May, Gutierrez and Harsin (1995) suggest that the consistency effect is the result of higher level processing concerned with response selection. Sequence learning When tasks involve the repeated performance of RT responses in a particular sequence, RTs may become faster because the subject is learning the serial order of the sequence. Howard, Howard and Mutter (1992) found that subjects learned more about the repeated sequence of stimuli through observation (which involved “looking at a series of stimuli”) than they did when making responses to them. This suggests that knowledge of the serial order of stimuli can develop through simple perceptual experience, and this is actually more available to deliberate recall than is knowledge acquired while responding manually. This finding is in conflict with the previous idea that eye movements are more automatic and that they do not reflect higher mental processes. However, Epelboim, Steinman, Kowler, Edwards, Pizlo, Erkelens and Collewijn (1995) found that in a sequence learning paradigm where subjects were asked to look at a series of targets or manually point to a series of targets, the looking-only sequences took longer to complete and did not benefit from practice. Traditionally, manual choice RT has been used to study implicit memory (learning without explicit knowledge of what is learned) in sequence learning. Nissen and Bullemer (1987) created a sequential pattern acquisition paradigm in which subjects responded, as quickly and accurately as possible, to an asterisk that appeared in one of four quadrants of a computer screen. They responded by pressing one of four keys on a computer keyboard, which represented the stimulus locations. There were two experimental conditions: one with randomly sequenced stimuli and one with particular sequence of trials repetitively presented. At the end of the series of trials subjects were asked whether or not they detected a pattern in the order of stimulus presentations. Results from sequential pattern acquisition studies show a decrease in RTs over trials in conditions where the sequence is repeated. Though the change in RTs indicates sequential learning, when queried the subjects replied that they had not recognized a pattern (Nissen and Bullemer, 1987; Willingham,
107 Nissen and Bullemer, 1989; Cohen, Ivry and Keele, 1990), suggesting that this sequential learning was implicit. One of the purposes of the present investigation was to determine if the influence of various perceptual processing on manual RT is also reflected when eye movements (SL and SRT) are examined. Three choice paradigms were employed to determine the degree to which such differences in task demands effected these eye movement responses. A second purpose of the current research was to determine whether global precedence and consistency effects could be obtained with both SL and SRT employed as the behavioral responses. A third purpose of this study was to determine whether sequence learning found with manual responses also occur with SL and SRT, and if it does, whether it is implicit or explicit.
Method Subjects. Fifteen university students and members of the community served as subjects. Eight were female and seven were male. All were naive as to the hypothesis under investigation. Subjects ranged in age from 18–30. All had normal or corrected to normal visual acuity (20/20). Informed consent and subject protection was in adherence with NIH guidelines. Stimuli and apparatus. The experiments were carried out with two microprocessors, one that controlled stimulus presentation (Everex, Model 286/16) and another that collected eye movement recordings (Everex, Model EX1800). The stimuli were displayed on a high-resolution monitor (Zenith, Model CV 2562). Horizontal eye movements were recorded for each eye with two channels of an eye tracker (Applied Science Laboratories, Model 106) and digitized at a rate of 1000 Hz with an A to D converter (Modular Instruments Inc., Model M-100). The microprocessor controlling the display provided sync pulses to the response recording microprocessor. Subjects were randomly assigned to one of three conditions: choice, global or local. Prior to performing these complex tasks, simple response times were obtained in a paradigm that required saccades from the central fixation point to a target appearing 10◦ to the right of fixation (See Figure 1 – left panel SIMPLE). A variable inter-trial interval of 700–1100 msec was employed to mitigate against anticipatory responses. These responses served as a baseline against which to assess the results with more complex tasks. In the choice condition, the asterisk was presented 10◦ to the right or left of fixation (see Figure 1 – left panel CHOICE). For the other global and local conditions the same central fixation point was viewed, but it was followed by stimuli consisting of global characters (either arrows pointing to the left or to the right), made
108
Figure 1. The stimulus conditions employed in the experiment are summarized in the left panel of this figure. F1 is the first frame viewed under all conditions – a fixation point in the middle of the screen. For the simple RT condition (which all subjects performed) F2 (the second frame) contained an ‘X’ positioned 10◦ to the right of fixation and the subject made a saccade to that position. For the Choice condition, F2 contained an ‘X’ positioned 10◦ to the right or left of fixation and the subject made a saccade to the appropriate position. The fixation point (an asterisk) and the target (a letter X) both had a luminance of 94.65 cd m2 , against a background luminance of 0.012 cd m2 . The asterisk subtended 0.43◦ (vertical) by 0.60◦ (horizontal) and the ‘X’ subtended 0.60◦ (vertical) by 0.34◦ (horizontal). For the local condition, F2 contained a compound arrow and the subject made a saccade to the end of the arrow to which the small elements pointed. For the Global condition, F2 contained compound arrow and the subject made a saccade to the end of the arrow to which the large arrow pointed. The trial sequence for all conditions was the same and is depicted in the right panel of this figure. There were 31 trial blocks of 16 trials each. The first 30 blocks contained the same sequence of 8 repeated twice. The L and R within each block refers to the direction of each saccade required on that trial. Eye movements were recorded in trial blocks 2, 9, 16, 23, 30 and 31. During trial block 31, a new randomization of saccade directions was employed. Immediately after trial 31, subjects were queried as to what the sequence was in the first 30 trial blocks.
up of local characters, smaller arrows which were either consistent (pointing in the same direction) or inconsistent (pointing in the opposite direction) as the global character (see Figure 2). Subjects sat 130 cm from the monitor. The inter-trial interval for the experimental conditions was fixed at 1.0 sec. The experimental conditions. Subjects were randomly assigned to one of three groups (n=5). Each group participated in only one experimental condition. In the choice condition, subjects were asked to focus on a fixation
109
Figure 2. An example of the stimuli used in the local and global conditions. The Michaelson contrast was 100% for all stimuli. The local stimuli subtended visual angles of 0.55◦ (vertical) by 0.41◦ (horizontal) and the global stimuli subtended visual angles of 20◦ (horizontal) by 6◦ (vertical).
point until a stimulus was presented on the screen (an ‘X’ 10◦ to the right or left of the fixation point) and then to move their eyes to the location of the stimulus as quickly as possible and not to move their eyes back until the fixation point reappeared. In the second and third groups, subjects respond to the direction of either the global or local elements of the stimuli (see Figures 1 and 2). These experimental tasks also involved a choice RT task, but in these cases subjects indicated the direction that the arrows pointed (left or right) for a given level (global or local) by moving their eyes to the end of the global arrow. For the global conditions (see Figure 1 – left panel GLOBAL), subjects looked at the tip of the global arrow head, and for the local conditions
110 Table 1. Summary table for the analysis of variance concerning task differences
1
2
3
4
Source
Sum of squares
D.F.
Mean square
F
Probability
Mean Group Error Response RXG Error Blocks BXG Error RXB RXBXG Error
5485920.44 214322.02 56091.73 98796.98 337.52 7562.64 220553.85 105736.19 22325.06 710.82 1476.16 10190.56
1 2 12 1 2 12 2 4 24 2 4 24
5485920.44 107161.01 4674.31 98796.98 168.76 630.22 110276.93 26434.05 930.21 355.41 369.04 424.61
1173.63 22.93
>0.001 >0.001
156.77 0.27
>0.001 >0.770
118.55 28.42
>0.001 >0.001
0.84 0.87
>0.445 >0.497
(see Figure 1 – left panel LOCAL) subjects looked at the end of the global arrow to which the local elements pointed. Each condition consisted of 31 blocks of 16 stimulus presentations (see Figure 1 – right panel). The last set of 16 stimuli were in a random sequence while the other 30 sets of 16 stimuli contained two repeated sequences of eight representing each of the stimuli four times. The same repeating sequence was used for all of the experimental conditions. At the end of the session the subjects were asked if they had recognized a repeated sequence in the presentation of the stimuli. All answers were recorded by the experimenter. If subjects responded that they had not detected a sequence they were told that there had been a repeating sequence of eight stimuli during the task and asked to guess as to the sequence. This acted as a ‘probe’ to determine whether subjects had explicitly learned the trial sequence. SL and SRT measures. The time elapsing from the onset of the test stimulus to the initiation of an eye movement (SL) and to the cessation of an eye movement (SRT) was recorded to the nearest 1.0 ms with a signal analysis program. SL and SRT were recorded on each trial during 6 trial blocks throughout the session, during iterative blocks 2, 9, 16, 23, and 30, and the final random block. Trials including blinks were discarded. The mean SL and SRT for each block were computed for each subject. In addition, these measures were computed separately for trials on which stimuli were consistent and inconsistent.
111
Figure 3. Mean response times for SL (left panel) and SRT (right panel) under the simple and complex conditions for each group. Vertical bars=±1 S.E.
Results Task differences Task differences were assessed by examining the response times on the first iterative and last random blocks (during which the subjects had no opportunity to be effected by sequence learning) for each group in relation to the simple condition. The means for simple SL (white bars) for each group are compared to the mean for the first block (gray bars) and the last block (black bars) in each condition in the left panel of Figure 3. A similar comparison is provided for SRT in the right panel of Figure 3. It is clear from these results that the simple response times were quite similar for each group and that the more complex tasks performed by each group gave rise to increased response times, but were similar across trial blocks. The mean SL and SRT for each subject in each group for the simple and complex conditions was tabulated and these data were submitted to a mixed three factor analysis of variance
112 Table 2. Summary table for the analysis of variance concerning global precedence and consistency effects
1
2
3
4
Source
Sum of squares
D.F.
Mean square
F
Probability
Mean Group Error Response RXG Error Consistency CXG Error RXC RXCXG Error
4098752.46 239955.58 10306.39 16568.97 2447.66 5182.62 7994.76 7994.76 2313.66 130.68 32.58 1923.77
1 1 8 1 1 8 1 1 8 1 1 8
4098752.46 239955.58 1288.30 16568.97 2447.66 647.83 7994.76 7731.18 289.21 130.68 32.58 240.47
3181.52 186.26
>0.001 >0.001
25.58 3.78
>0.001 >0.088
27.64 26.73
>0.001 >0.001
0.54 0.14
>0.482 >0.722
[group (choice, global or local) X response type (SL or SRT) X task (simple, complex 1 or complex 2)] design, with response and task as repeated measures (see Table 1). Subsequent tests (Newman-Kuels) revealed no significant differences between the simple responses across groups, and only marginally significant differences between simple and complex conditions for the choice group (p <0.06). Significant differences were obtained between the simple and complex conditions for the global (p <0.03), and local (p <0.00001) groups, and significant differences between the complex responses across groups (p <0.05 or less). There were no significant differences between the first and final blocks for any of the conditions. Global precedence and consistency effects Response times for the (first) iterative and (final) random blocks (which precluded the effects of sequential learning) were sorted as to whether the stimuli were consistent or inconsistent for each subject in each group, and group means were derived from the means of the subjects in each group. The mean response times (SL – left panel and SRT – right panel) for consistent and inconsistent stimuli are presented for local and global conditions in Figure 4. It is clear that response times are faster for global than local conditions (the global precedence effect) and that the mean response times are faster for consistent conditions than for inconsistent conditions for subjects responding to the local level (the consistency effect). These data were submitted to
113
Figure 4. Mean SL and SRT as a function of stimulus condition (consistent or inconsistent) for the local and global groups. Vertical bars=±1 S.E.
mixed analysis of variance design [Groups (global or local) X stimulus type (consistent or inconsistent) X response type (SL or SRT)], with stimulus type and response type as a repeated measures (see Table 2). Tests subsequent to analysis of variance (Newman Kuels) indicated that both means for the global condition were significantly less than those for the local condition (p <0.00001), but not significantly different from each other. For the local conditions, the mean for consistent trials was significantly less than the mean for the inconsistent condition (p <0.0006). Sequence learning The mean response time as a function of trial blocks for each group is presented in Figure 5 (SL – left panel, SRT – right panel). The results for both measures are quite similar. The horizontal dotted lines indicate the average simple response time for all 15 subjects. For the choice condition, response times were almost as fast as the simple case. There was little decrease in response times over trial blocks and the increase in response times from the last iterative to the final random block is modest. For the global condition, response times were somewhat slower than the simple case, but there was little decline as a function of trial blocks and the difference in response time
114 Table 3. Summary table for the analysis of variance concerned with sequence learning (Blocks 1–5)
1
2
3
4
Source
Sum of squares
D.F.
Mean square
F
Probability
Mean Group Error Response RXG Error Blocks BXG Error RXB RXBXG Error
10899620.82 903119.67 117777.17 196665.20 857.74 5950.67 14786.96 13736.53 50883.04 808.68 1655.42 15459.50
1 2 12 1 2 12 4 8 48 4 8 48
10899620.82 451559.83 9814.77 196665.20 428.87 495.89 3696.74 1717.07 1060.06 202.17 206.93 322.07
1110.53 46.01
>0.001 >0.001
396.59 0.86
>0.001 >0.446
3.49 1.62
>0.014 >0.144
0.63 0.64
>0.645 >0.738
from the last iterative to the final random block is slight. In contrast, the response times for the local condition declined over trial blocks and there was a considerable increase from the last iterative to the final random block. The response time data for the first 5 iterative blocks were submitted to a mixed analysis of variance [group (choice, global or local) X response type (SL or SRT) X trial blocks (1–5)] design with response type and trial blocks as repeated measures (see Table 3). Tests subsequent to analysis of variance (Newman Kuels) indicated significant differences between the choice and local groups (p <0.00003), and the global and local groups (p <0.0003), but only marginally significant differences between the choice and global groups (p <0.055). The SL responses were, of course, significantly faster than the SRT responses (p <0.00001], and responses for the first iterative trial block were significantly slower than the last iterative trial block (p <0.00001). No other mean comparisons were significant. The response times for the last iterative and the final random block were entered into a similar analysis (see Table 4). Subsequent tests (NewmanKuels) revealed that significant differences between the last two trial blocks occurred for the choice (p <0.01) and local (p <0.05) groups, but not for the global group. The results of the post-test inquiries as to sequence order are presented in Table 5. All subjects in the choice group displayed perfect knowledge of
115
Figure 5. Mean SL and RT as a function of trial blocks (1–6) for the choice, local and global groups. The dashed horizontal lines represent the mean SL or SRT in the simple baseline condition. Closed symbols depict iterative trial blocks and open symbols depict a new randomization. Vertical bars=±1 S.E.
the stimulus sequence while only one subject in the global group and two subjects in the local group acquired that knowledge. Discussion It is clear that tasks employed lead to quite different response times. The simple and choice conditions were similar as first reported by Saslow (1967), but the choice response times were only marginally slower than the simple responses. With either measure of eye movement response time (SL or RT), the global precedence and consistency effects are evident, as is the case with manual responses (Navon 1977, 1981; Berg and May, 1998). These findings imply that different eye movement tasks are mediated by different physiological mechanisms that account for the differences in rapidity of response. In addition, it is clear that sequence learning occurred explicitly in the choice group, but only implicitly in the local group. We consider below two current models of saccade generation and perceptual processing which account for
116 Table 4. Summary table for analysis of variance concerning sequence learning (Blocks 5 and 6)
1
2
3
4
Source
Sum of squares
D.F.
Mean square
F
Probability
Mean Group Error Response RXG Error Blocks BXG Error RXB RXBXG Error
4293519.43 266434.92 81611.74 78976.48 785.95 10611.23 11736.21 6486.25 8181.75 55.01 690.41 2493.89
1 2 12 1 2 12 1 2 12 1 2 12
4293519.43 133217.46 6800.98 78976.48 392.98 884.27 11736.21 3243.13 681.81 55.01 345.20 207.82
631.31 19.59
>0.001 >0.001
89.31 0.44
>0.001 >0.651
17.21 4.76
>0.001 >0.030
0.26 1.66
>0.616 >0.231
Table 5. Subjects’ attempts to identify the sequence used for each experimental group. The correct sequence was RLRLRRLL. Asterisks indicate correct responses Subject number
Local response
Global response
Choice response
1 2 3 4 5
LLRLRR RLRLRLRL LRLRLRRL∗ – RRLLRLRL∗
– – – LLRLR RRLLRLRL∗
RLRLRRLL∗ RRLLRLRL∗ RRLLRLRL∗ RLRLRRLL∗ RLRLRRLL∗
the differences in response time and suggest how we might predict response speed from task demands. The perceptual processing model also suggests how different tasks might result in differences in explicit sequence learning. The what and when model of saccade generation In a recent review article, Findley and Walker (1999) propose a model of saccade generation based on parallel processing and competitive inhibition. The model involves two physiologically based pathways concerned with the governance of ‘when’ an eye movement occurs and to ‘where’ the saccade is directed. It accounts for automatic responses to visual onsets via lower
117 level mechanisms and more cognitively based, directed, eye movements via higher level mechanisms. The basic competition occurs at the lowest, motor command level where the decision is to maintain fixation or to move the eyes. This level is influenced in turn by two antagonistic centers (‘fixate’ and ‘move’) at the second, movement decision, level. These two centers receive information: about visual onsets in the fovea or periphery from a third, automatic, level which can give rise to reflexive (or exogenous) responses; from various temporally prepared cognitive processes from the fourth, automated level which provide for responses based on spatial selection or visual features; from higher decision centers at the fifth, voluntary level which may override the influences generated at the automatic and automated levels. Within the framework of this model, the fastest simple SL or SRT is governed by the visual onset of a single target in the periphery (via level 3) and it is facilitated by automated spatial selection (via level 4). The somewhat slower choice responses (in the present investigation) involve visual onset of a single target level 3), but the automated spatial selection rule imposed by level 4 is more complex and a spatial decision process is required (level 5). Global responses suffered because numerous small targets were presented in the fovea and the periphery (creating competition in the salience map at the automatic level), the automated spatial selection rule was more complex, and the voluntary decision was based on both spatial position and global visual features. The slowest local responses suffered from the same competing influences at the automatic and automated levels, but with the added intrinsic salience of the foveal stimulus (favoring a fixation response), and the voluntary decision was based on both spatial position and local visual features. Thus, the explanation of the When and Where Model provides at least qualitative agreement with the data and suggests that the response time of saccades depends on the interaction of numerous processes at various levels. The what and where streams in perceptual processing Another current conceptualization of parallel processing in visual perception provides some rationale for why local responses were so much slower than those in the other conditions. In the early eighties, visual search experiments (Triesman and Gelade, 1980; Julesz, 1980; Bergan and Julesz, 1983) led to the contention that visual features of an object are discerned via a time consuming, focal attention process, while detecting the position of objects is achieved rapidly via preattentive, parallel processing (Sagi and Julesz, 1985). Later lesion work in primates (Desimone and Ungerleider, 1989; Morel and Bullier, 1990; Baizer, Ungerleider and Desimone, 1991) and neuropsychological investigations in human patients (Holmes, 1919; Cole, Schutta and Warrington, 1962; Ratcliff and Davies-Jones, 1972; Benton, Han-
118 nay and Varney, 1975; Meerwaldt and Van Harskamp, 1982; Hess, Baker and Zihl, 1989; Vaina, LeMay, Bienfang, Choi and Nakayama, 1990) suggested that the dorsal stream (which courses from the primary visual cortex toward the parietal lobe) is specialized for the processing of spatial information (‘where’) and the ventral stream (which courses from the primary visual cortex toward to inferior temporal lobe) is involved with the perception of object properties (‘what’) (Mishkin et al., 1983). As summarized above, the global precedence effect has been explained in terms of the response properties of retinal ganglion cells, and this notion is compatible with the what and where distinction, since m-cells have input to motion detection (DeYoe and Van Essen, 1988) and may provide information to eye movement control centers in the dorsal stream. The large differences in response time between the local and global conditions reflect the increased time needed to process ‘what’ information and carry out the required response. To perform the global response, subjects did not have to concern themselves with the properties of the large arrow (e.g. a large arrow pointing to the left composed of small arrows pointing to the right). Instead, they merely had to locate the extra elements comprising the arrowhead and saccade to that side. In the local condition, in order to decide in which direction to saccade, subjects had to first discern what the small element was and what it meant in terms of the response alternatives. The voluntary responses in the former case would be facilitated by reflexive attentional mechanisms, while those in the latter case would require directed attention, which has been shown to require more time (Posner, 1990). The notion that the consistency effect might derive from differences in motor performance was not supported by the present findings in that the pattern of results for SL and SRT did not differ. Thus, the mechanisms that give rise to the differences in response times are clearly pre-motor, but may derive from different stages in the perceptual and response selection process. If ‘where’ tasks are carried out preattentively, then it implies that the focal or directed attentional mechanisms are free to process other, perhaps incedental, aspects of the task. This might account for why all the members of the choice group had explicit knowledge of the iterative sequence, while fewer members of the local group attained this knowledge. It does not explain the differences between the last iterative and final random block in the local group. This suggests that the local group acquired some implicit knowledge of the sequence while the global group did not. These findings are at odds with a previous finding that implicit learning occurred with the global, but not local conditions (Berg and May, 1998). It is conceivable that future work with carefully designed tasks will provide more sophisticated diagnostic criteria for use with clinical populations. Eye
119 movement anomalies with paradigms which elicit short latency saccades could indicate lesions in lower level areas concerned with reflexive or automated eye movements, while anomalies uncovered with tasks associated with longer latency responses would be more indicative of cognitive dysfunction caused by cortical disruption.
References Abrams RA, Onnk HM, Pratt J. Fixation point offsets facilitate endogenous saccades. Perception & Psychophysics 1998; 60: 201–8. Badcock JC, Whitworth FA, Badcock DR, Lovegrove WJ. Low-frequency filtering and the processing of local-global stimuli. Perception 1990; 19: 617–29. Baizer JS, Ungerleider LG, Desimone R. Organization of visual inputs to the inferior temporal and posterior parietal cortex in macaques. J Neurosci 1991; 11: 168–90. Benton A, Hannay H, Varney N. Visual perception of line direction in patients with unilateral brain disease. Neurology 1975; 25: 907–10. Berg ML, May JG. Parallel processing in visual perception and memory: What goes where and when? Current Psychology 1997; 16: 247–83. Bergen JR, Julesz B. Parallel versus serial processing in rapid pattern discrimination. Nature 1983; 303(5919): 696–8. Cohen A, Ivry RI, Keele SW. Attention and structure in sequence learning. J Exper Psychol: Learning. Memory and Cognition 1990; 16(1): 17–30. Cole M, Schutta H, Warrington E. Visual disorientation in homonymous half fields. Neurology 1962; 12: 257–63. Desimone R, Ungerleider LG. Neural mechanisms of visual processing in monkeys. In: Boller F, Grafman eds. Handbook of neuropsychology. vol. 2. New York: Elsevier, 1989: 267–99. DeYoe EA, Van Essen DC. Concurrent processing streams in monkey visual cortex. Trends Neurosci 1988; 11: 219–26. Donders FC. Die Schnelligkeit psychischer Processe. Archiv für Anatomie und Physiologie 1868; 657–81. Epelboim J, Steinman RM, Kowler E, Edwards M, Pizlo Z, Erkelens CJ, Collewijn H. The function of visual search and memory in sequential looking tasks. Vision Res 1995; 135: 3401–422. Flipse JP, Staathof CSM, Van der Steen J, Van Leeuwen AF, Van Doorn PA, Van der Meche FGA, Collewijn H. Binocular saccadic eye movements in multiple sclerosis. J Neurological Sci 1997; 148: 53–65. Findley JM, Walker R. A model of saccade generation based on parallel processing and competitive inhibition. Behav Brain Sci (in press). Forbes B, Klein RM. The magnitude of the fixation offset effect with endogenously and exogenously controlled saccades. J Cognitive Neurosci 1996; 8: 344–52. Fry DB. Simple reaction times to speech and non-speech stimuli. Cortex 1975; 11: 355–60. Gould JD. Eye movements during visual search and memory search. J Exper Psychol 1973; 98: 184–95. Goldberg ME, Eggers HM, Gouras P. The ocular motor system. In: Kandel ER, Schwartz JH, Jessel TM, eds. Principles of Neural Science. East Norwalk, CO: Appleton & Lange, 1991.
120 Helmholtz HLF. On the rate of transmission of the nerve impulse. Berichtn. König. Preussische Akadamie der Wissenschaften. Berlin 1850; 14–5. Hess R, Baker C, Zihl J. The motion-blind patient: Low-level spatial and temporal filters. J Neurosci 1989; 9: 1628–40. Holmes G. Disturbances of visual orientation. British J Ophthalmol 1919; 2: 449–68, 506–18. Howard SH Jr, Howard SA, Mutter DV. Serial pattern learning by event observation. J Exper Psychol: Learning, Memory and Cognition 1992; 92: 1029–39. Hughes HC, Layton WM, Baird GC, Lester LS. Global precedence in visual pattern recognition. Perception and Psychophysics 1984; 35(4): 361–71. Julesz B. Textons, the elements of texture perception, and their interactions. Nature 1981; 290(5802): 91–7. Kingstone A, Klein RM. Visual offset facilitates saccade latency: does pre-disengagement of attention mediate this gap effect? J Exper Psychol, Human Perception and Performance 1993; 19: 251–65. LaGasse LL. Effects of good form and spatial frequency on global precedence. Perception and Psychophysics 1993; 53(1): 89–105. Lambert AJ, Sumich AJ. Spatial orienting controlled without awareness: a semantically based implicit learning effect. Quarterly J Exper Psychol 1996; 49A: 490–518. Lovegrove WJ, Lehmkuhle S, Baro JA, Garzia R. The effects of uniform field flicker and blurring on the global precedence effect. Bulletin Psychonomic Soc 1990; 29(4): 289–381. Maljkovic V, Nakayama K. Priming of pop-out 1. Role of features. Memory Cognition 1994; 22: 657–72. Martin M. Local and global processing: The role of sparsity. Memory Cognition 1979; 7(6): 476–84. Mastaglia FL, Black JL, Collins DWK. Quantitative studies of saccadic and pursuit eye movements in multiple sclerosis. Brain 1979; 102: 817–34. Mastiglia FL, Black JL, Thiekbroom G, Collins DWK. Saccadic eye movements in multiple sclerosis. Neuro-Ophthalmology 1982; 2: 225–36. May JG, Gutierrez C, Harsin CA. The time-course of global precedence and consistency effects. Internat J Neurosci 1995; 80: 237–45. Meerwaldt J, Van Harskamp F. Spatial disorientation in right-hemisphere infarction. J Neurology, Neurosurgery and Psychiatry 1982; 45: 586–90. Meienberg O, Muri B, Rabineau PA. Clinical and oculographic examinations of saccadic eye movements in the diagnosis of multiple sclerosis. Arch Neurology 1986; 43: 438–43. Morel A, Bullier J. Anatomical segregation of two cortical visual pathways in the macaque monkey. Vis Neurosci 1990; 4: 555–8. Merkel J. Die zeitlichen Verhältnisse der Willensthätigkeit. Philos St 1885; 2: 73–127. Mishkin M, Ungerleider LG, Macko KA. Object vision and spatial vision: Two cortical pathways. Trends Neurosci 1983; 6: 414–7. Navon D. Forest before trees: the precedence of global features in visual perception. Cogn Psychol 1977; 9: 353–83. Navon D. The forest revisited: More on global precedence. Psychol Res 1981; 43: 1–32. Netsell R, Daniel B. Neural and mechanical response time for speech production. J Speech and Hearing Res 1974; 17: 608–18. Nissen MJ, Bullemer P. Attentional requirements of learning: Evidence from performance measures. Cogn Psychol 1987; 19: 1–32. Posner MI, Petersen SE. The attentional system of the human brain. The Annual Review of Neuroscience 1990; 13: 25–42.
121 Ratcliff G, Davies-Jones G. Defective visual localization in focal brain wounds. Brain 1972; 95: 49–60. Rayner K. Eye movements and cognitive processes in reading, visual search, and scene perception. In: Findlay JM, Walker R, Kentridge RW, eds. Eye movement research: mechanisms, processes and applications. North Holland, 1995. Reuter-Lorenz PA, Hughes HC, Fendrich R. The reduction of saccadic latency by prior offset of the fixation point: an analysis of the gap effect. Perception and Psychophysics 1991; 49: 167–75. Reuter-Lorenz PA, Oonk HM, Barnes LL, Hughes HC. Effects of warning signals and fixation point offsets on the latencies of pro- versus antisaccade: implications for an interpretation of the gap effect. Exper Brain Res 1995; 103: 287–93. Ross LE, Ross SM. Saccade latency and warning signal: stimulus onset, offset and change as warning events. Perception and Psychophysics 1980; 27: 251–7. Saslow MG. Effects of components of displacement-step stimuli upon latency for saccadic eye movement. J Optical Soc Amer 57: 1024–9. Saslow MG. Latency for saccadic eye movements. J Optical Soc Amer 1967b; 57: 1030–3. Stetson RH. Motor Phonetics: A Study of Speech, Movement, and Action. Amsterdam: North Holland Publishing Company, 1951. Treisman AM, Gelade G. A feature-integration theory of attention. Cogn Psychol 1980; 12(1): 97–136. Vaina L, LeMay M., Bienfang D, Choi A, Nakayama K. Intact ‘biological motion’ and ‘structure from motion’ perception in a patient with impaired motion mechanisms. A case study. Vis Neurosci 1990; 5: 353–69. Willingham D, Nissen M, Bullemer P. On the development of procedural knowledge. J Exper Psychol: Learning, Memory and Cognition 1989; 15: 1047–60. Address for correspondence: J. G. May, Department of Psychology, University of New Orleans, Lakefront, New Orleans, LA 70148, USA Phone: (504) 280-6770; E-mail:
[email protected]