Oculomotor control and the maintenance of spatially and temporally distributed events in visuo-spatial working memory

Transcription

1 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2003, 56A (7), Oculomotor control and the maintenance of spatially and temporally distributed events in visuo-spatial working memory David G. Pearson and Arash Sahraie University of Aberdeen, Scotland, UK Previous studies have demonstrated that working memory for spatial location can be significantly disrupted by concurrent eye or limb movement (Baddeley, 1986; Smyth, Pearson, & Pendleton, 1988). Shifts in attention alone can also interfere with spatial span (Smyth & Scholey, 1994), even with no corresponding movement of the eyes or limbs (Smyth, 1996). What is not clear from these studies is how comparable is the magnitude of effect caused by different forms of spatial disrupter. Recently, it has been demonstrated that limb movements produce as much interference with spatial span as do reflexive saccades (Lawrence, Myerson, Oonk, & Abrams, 2001). In turn this has led to the hypothesis that all spatially directed movement can produce similar effects in visuospatial working memory. This paper reports the results of five experiments that have contrasted the effect of concurrent eye movement, limb movement, and covert attention shifts on participants working memory for sequences of locations. All conditions involving concurrent eye movement produced significantly greater reduction in span than equivalent limb movement or covert attention shifts with eyes fixated. It is argued that these results demonstrate a crucial role for oculomotor control processes during the rehearsal of location-specific representations in working memory. In the case of the verbal working memory there is a well-documented account of rehearsal in the form of an active articulatory mechanism that revives auditory memory traces stored within a temporary phonological store (Baddeley, 1986, 2000). In contrast, accounts of rehearsal in visuo-spatial working memory are considerably less well developed (Pearson, 2001). Potential candidates for rehearsal in the visuo-spatial domain are processes associated with eye movement, which were initially investigated in a series of studies carried out by Idzikowski, Baddeley, Dimberly, and Park (reported in Baddeley, 1986). They examined the effect on working memory for spatial and verbal material of asking participants to track a sinusoidally moving target with their eyes. They found that eye movements produced a highly significant decline in spatial working memory in contrast to a verbal control and a condition in Requests for reprints should be sent to David G. Pearson, Vision Research Laboratories, Department of Psychology, William Guild Building, University of Aberdeen, Aberdeen, AB24 2UB, Scotland, UK. d.g.pearson@abdn.ac.uk 2003 The Experimental Psychology Society DOI: /

2 1090 PEARSON AND SAHRAIE which the background to the target moved but the eyes remained stationary. In addition, the interference effect occurred at both encoding and/or retrieval of the spatial material. On the basis of these results Baddeley suggested that rehearsal in visuo-spatial working memory could occur via an active control process linked to oculomotor control that refreshed visuospatial material represented in a passive perceptual input store, in a structure that mirrored the relationship between the phonological store and articulatory loop that had been proposed for verbal working memory. However, other forms of concurrent body movement have also been shown to produce interference in spatial working memory, including sequential tapping of keys (Logie & Marchetti, 1991; Pearson, Logie, & Gilhooly, 1999; Smyth, Pearson, & Pendleton, 1988; Smyth & Pendleton, 1989) and arm movements across an unseen matrix (Quinn, 1991; Quinn & Ralston, 1986). In addition, Johnson (1982) found that simply asking participants to imagine making an arm movement disrupted temporary memory for location, suggesting that the planning of movement was sufficient to produce interference irrespective of whether the movement was actually executed. A possible overlap between spatial rehearsal and the cognitive processes involved during the planning and production of movement has been proposed by Logie (1995), who implicates a spatial inner scribe mechanism in the extraction of information from a visual cache to allow for targeted movement. In Logie s model of visuospatial working memory the production of physical movement, or even just planned production, should interfere with any task that also requires the operation of the inner scribe, including the retention of sequential locations or movements (Logie, 1995; pp ). However, the extent to which actual explicit or implicit motor processes underlie the operation of spatial working memory, particularly regarding the temporary rehearsal of sequences of spatial locations, remains controversial. An alternative explanation is that spatial working memory is disrupted by shifts in spatial attention caused by eye or hand movements rather than the movements themselves. A key study carried out by Smyth and Scholey (1994) demonstrated that spatial recall span was significantly impaired if participants were exposed to visual or auditory signals that appeared in different spatial locations, and that this impairment was increased if they had to point to the location of the signal or make left/right categorical judgements. A follow-up study that controlled for eye movements verified that shifts in spatial attention alone could produce significant disruption of spatial span in the absence of any overt movement of the eyes or limbs (Smyth, 1996). Smyth and colleagues interpreted these results as providing evidence that maintenance of location in spatial working memory involves shifts in spatial attention rather than implicit motor processes. A more detailed theory of spatial rehearsal has since been proposed by Awh and colleagues (Awh, Anllo-Vento, & Hillyard, 2000; Awh & Jonides, 2001). They have argued that spatial rehearsal involves a frontal-parietal system that actively maintains spatial information by means of focal shifts of spatial attention to memorized locations held in working memory. Awh, Jonides, and Reuter-Lorenz (1998) found that when participants were asked to hold a location in working memory this facilitated the visual processing of stimuli appearing in that location relative to stimuli appearing in nonmemorized locations. They also found that disrupting the focus of spatial location during a retention interval significantly impaired participants accuracy on a spatial memory task in comparison to a condition in which the focus of attention was static. On the basis of these results Awh and colleagues have argued that the process of visual selection is a key component of rehearsal in spatial

3 OCULOMOTOR CONTROL AND SPATIAL WORKING MEMORY 1091 working memory, with the cognitive mechanisms involved in spatial selective attention also being used to provide functional markers for location-specific representations in working memory (Awh & Jonides, 2001). Importantly, their account of spatial rehearsal assumes a supramodal basis for attention, in which covert shifts of attention involve processes that are functionally independent from those that mediate oculomotor programming (Awh et al., 1998, 2000). However, other evidence suggests that prefrontal executive processes responsible for the control of attention may be involved in the initiation of planned saccades to a greater extent than a supramodal theory might predict. Roberts, Hager, and Heron (1994) have shown that concurrent mental arithmetic interferes with the execution of antisaccades, but does not appear to interfere with reflexive prosaccades to similar targets. However, a study by Stuyven et al. found that prosaccades could also be vulnerable to interference from a concurrent executive task if the saccades were controlled rather than automatic, suggesting that prefrontal executive processes could be involved in controlled saccade execution across a wider range of tasks than just antisaccades (Stuyven, Van der Goten, Vandierendonck, Claeys, & Crevits, 2000). It is clear from the published literature on spatial working memory that the temporary maintenance of location can be disrupted by a wide range of different secondary tasks, including eye movements, limb movements, and shifts of spatial attention that occur in the absence of eye or limb movements. What is less clear, however, is how comparable these different forms of spatial disrupter are. The magnitude of the different effects could be the same, or alternatively one form of secondary task could produce much higher levels of disruption. In addition, the extent to which the interference caused by different procedures is due to a common cognitive mechanism is unclear. For example, the disruption caused by concurrent eye and limb movement could be entirely due to related shifts in spatial attention, with the processes involved during the control of the movement being largely irrelevant. Some of these issues have been addressed in a recent study carried out by Lawrence, Myerson, Oonk, and Abrams (2001) who examined the effect of saccadic eye movement, limb movement, and saccade inhibition on memory span for spatial locations. They found that limb movements that were performed while maintaining fixation produced as much interference with spatial span as did reflexive saccades, and they concluded from this that the interference caused by eye movement was not due to its visual consequences. Instead Lawrence et al. argue that all spatially directed movement produces similar effects on visuospatial working memory, irrespective of the type of movement being executed. They speculate that the interference caused by spatially directed movements can be due either to interruption of a rehearsal sequence such as a positive feedback loop proposed to link the prefrontal and posterior cortices (Chafee & Goldman-Rakic, 2000; Fuster, 1995), or alternatively to disruption of an attention-based rehearsal process such as that put forward by Awh and colleagues (Awh & Jonides, 2001; Awh et al., 1998). However, there are some aspects of the procedure adopted by Lawrence et al. (2001) that may have mitigated against finding differential effects of eye and limb movement on spatial span. They used a procedure in which the presentation of locations in the primary task is interleaved with performance of the secondary task that is, the first location is presented, participants execute a single reflexive saccade, the second location is presented, another single saccade is executed, and so on. A similar procedure was used for limb movement, in which participants perform single key presses interleaved with presentation of the locations in the

4 1092 PEARSON AND SAHRAIE primary task. Such a procedure has been rarely used in the majority of other studies that have examined visuo-spatial working memory, in which performance of the secondary task is typically confined to a retention interval that occurs after the encoding stage (Baddeley, 1986; Logie, 1995). This has the advantage of allowing more precise interpretation of any effects of intervening task on encoding and retention processes. A single eye or limb movement may not produce comparable effects to those when a succession of movements has to be generated. The interleaving technique may allow participants to carry out covert rehearsal, as can occur with the effects of articulatory suppression on verbal span if the rate of articulation is too slow (Baddeley, 1986). In addition, the interleaving of primary and secondary tasks may cause general interference at the executive level that may contribute to masking differential effects of eye and limb movements. The current paper presents a series of experiments that have attempted to comprehensively determine the comparative effects of eye movements, covert attention shifts with maintained fixation, and limb movements, on memory span for sequentially presented spatial locations. The main objective of the study was to compare the level of interference caused by saccadic and smooth pursuit eye movement with that caused by discrete and continuous covert attention shifts. Although previous studies (i.e., Awh et al., 1998; Smyth & Scholey, 1994) have demonstrated that covert shifts in attention alone can significantly disrupt spatial memory, it is unclear whether the magnitude of this disruption is equivalent in behavioural terms to that caused by concurrent eye or limb movements. It is also unclear whether smooth pursuit and saccadic eye movements can be treated as equivalent in terms of the levels of interference produced in spatial memory (Idzihowski et al., in Baddeley, 1986; Hale, Myerson, Rhee, Weiss, & Abrams, 1996; Lawrence et al., 2001; Quinn & McConnell, 1996). Although it might be expected that pursuit and saccadic movements produce similar attentional demands, functional imaging studies have demonstrated differences between the levels of activity generated by the two types of eye movement within the frontal eye fields (Petit, Clark, Ingeholm, & Haxby, 1997). It has also been shown that accuracy of pointing to justextinguished visual targets is strongly influenced by the type of eye movement (saccadic or pursuit) that is made to the target (Honda, 1984). This could result in differences in the behavioural effects of pursuit and saccadic movements on spatial recall span. Comparison of the effect of different types of secondary task on spatial span can be seen as an essential step in understanding the cognitive mechanisms that underlie temporary memory for spatial location. Although eye movement has been shown to be related to spatial memory, the exact nature of this relationship is unclear, and there is a need for the relative effects of eye movements, limb movements, and covert attention shifts on spatial span to be clearly demonstrated. In an attempt to address these issues five experiments were carried out using a repeated measures design. Participants GENERAL METHOD FOR EXPERIMENTS 1 5 The use of repeated measures in which participants act as their own control provides the most effective means to examine the comparative effects of different secondary tasks, as differential strategy use or other confounding factors are considerably reduced relative to a mixed or between-subjects design. In

5 OCULOMOTOR CONTROL AND SPATIAL WORKING MEMORY 1093 light of this it was decided that the same volunteers should participate in all five planned experiments. The experiments were carried out separately over a 15-month period in the order reported, with each testing session lasting 2 3 hours per participant. The participants were 12 right-handed females with normal or corrected-to-normal vision. The age range was 22 to 43 years (mean age years). All participants were postgraduate students at the University of Aberdeen who took part in the experiments on a voluntary basis. Informed consent was obtained from all the participants, and ethical approval was obtained from the Departmental Ethics Committee at the University of Aberdeen. Apparatus and method Participants spatial span was measured using an irregular array of nine blocks based on the Corsi Blocks test (De Renzi & Nichelli, 1975), and modified for display on a PC monitor. The method allowed the presentation of spatially distributed and temporally separated events, with each event consisting of a transient change in luminance of a block from grey to black. Participants viewed binocularly, a uniform grey background (26 cd/m 2 ) from a viewing distance of 57 cm. A headrest was used to reduce irrelevant movement. The monitor screen measured 37 cm (w) 26 cm (h), and the Corsi blocks display (28 cm 21 cm) was positioned centrally on the uniform field. Each block measured 3.6 cm square and was defined by a 2-mm border. For each trial the sequence of events was as follows: The array of blocks appeared, and participants fixated on a cross positioned at the centre of the screen. The series of transient visual events was denoted by a momentary change of square luminance to 0.65 cd/m 2 for 250 ms followed by return to the background luminance for the same period. The selection of blocks to be highlighted was carried out randomly by the computer with the restriction that no block was highlighted twice in each trial. At the end of the sequence the blocks disappeared signalling the start of a 5-s retention interval. The reappearance of the blocks signalled the end of the retention interval, which was also signalled by a short audio tone. Participants recalled the sequence of presented events by pointing and pressing the mouse key on previously highlighted blocks in an identical order to that of presentation. A summary of these events is given in Figure 1. The recall of the sequence always took place with no restrictions on eye movements. To ensure compliance with the experimental instructions both horizontal and vertical eye movements were monitored and recorded using electro-oculograms (EOG). Some examples of EOG traces from different experimental conditions are illustrated in Figure 2. The EOG signals were displayed on a separate computer screen throughout the testing session. Calibration of exact position of gaze was accomplished by the measurement of the EOG signals when participants sequentially fixated on four calibration points on the screen at the start of the experimental run. Accurate and consistent measurement of spatial span was integral to the investigations reported here. Span was calculated using a staircase procedure adapted from previously published studies (Hulme, Maughan, & Brown, 1991; Logie, Della Sala, Laiacona, Chalmers, & Wynn, 1996; Logie & Pearson, 1997). We also took multiple estimates of span for each condition to reduce any confounding effects produced by the random highlighting of blocks for each trial (Berch, Krikorian, & Huha, 1998; Kemps, 1999; Smirni, Villardita, & Zappala, 1983). At the start of each condition, participants carried out a series of practice trials to ensure understanding and compliance with the experimental instructions. Once performance was judged to be stable they continued with five separate runs for every condition. For each experimental run trials were blocked in groups of three. In the first block only two sequential transient events were presented per trial. Following the retention interval the participant reported the position and order of the visual events. If at least two out of the three trials were correctly identified then an additional block was presented on each trial. The number of transient visual events was increased sequentially in the consecutive blocks, and the experimental run was terminated if more than two trials were reported incorrectly in any one block.

6 1094 PEARSON AND SAHRAIE Figure 1. Diagram of screenshots depicting the order and timing of events in a two-event trial in the baseline condition used for Experiments 1 to 5. Events run in direction of arrow and represent (1) initial presentation of array of blocks, (2) presentation of first event indicated by transient change in luminance of block (250 ms), (3) presentation of second event (interstimulus time of 250 ms), (4) 5-s retention interval with presentation of uniform field with fixation point (screenshots from retention intervals in continuous pursuit, Experiment 1, and visual noise, Experiment 3, conditions are also illustrated), (5) block array reappears and participants indicate position and order of presented locations using the PC mouse. The span for an individual experimental run was calculated as the average of the number of events per trial in the last three correctly identified trials. The participant s span for a condition was then calculated as the average of the results from the last four out of five experimental runs recorded (the first was treated as a practice run and not included in the analysis). The control condition for each of the five experimental sessions was when the participant fixated on a fixation crosshair placed on a grey background during the retention interval. Each experiment started by baseline measurements followed by two or three experimental conditions. The order of presentation of conditions was counterbalanced across participants. EXPERIMENT 1 Effects on spatial span of smooth pursuit eye movements and continuous attention shifts with maintained fixation The first experiment examined the effect on spatial span of smooth pursuit eye movements during the retention interval, compared with covert attention shifts with maintained fixation. Smooth pursuit eye movements have been shown previously to disrupt spatial memory (Baddeley, 1986). It has also been demonstrated that attention shifts with maintained fixation

7 OCULOMOTOR CONTROL AND SPATIAL WORKING MEMORY 1095 Figure 2. Examples of horizontal EOG traces for all eye movement conditions, as well as baseline and no eye movement with eyes shut conditions. also disrupt spatial memory (Awh et al., 1998; Smyth, 1996). The aim of the experiment was to compare the magnitude of these two effects against each other. Method Secondary tasks In the smooth pursuit condition participants were asked to track an oscillating ball with sinusoidally modulated speed moving in the centre of the screen. The ball appeared in the same position as the fixation point at the beginning of the retention interval and moved right or left on a randomly determined basis. Participants tracked the ball continuously over the duration of the 5-s retention interval. The ball took one second to move from one side of the screen to the other. In the attention shift with maintained fixation condition the ball moved in an identical fashion at the top of the screen. The crosshair fixation point was present throughout the retention interval. Participants were instructed to maintain fixation while focusing their attention on the ball at the top of the screen. A screenshot from the pursuit movement condition is presented in Figure 1. Participants EOG traces were monitored continuously during both conditions to ensure compliance with the experimental instructions. Any deviation from the instructions would result in the given trial being re-run. The advantage of this procedure was that it explicitly prevented participants from protecting primary task performance by trading off performance in the secondary task.

8 1096 PEARSON AND SAHRAIE Results and discussion Mean spans in each condition were 5.09 (SD = 0.85) for baseline, 3.43 (SD = 1.19) for pursuit, and 4.29 (SD = 0.88) for continuous attention shifts. Participants compliance with experimental instructions was excellent, and it was not necessary to omit or re-run any trials. The results were analysed by a one-way analysis of variance (ANOVA) that revealed a main effect of condition, F(2, 22) = 33.22, p <.001. Post hoc comparisons by Tukey HSD tests revealed that all differences were significant at the.001 level. The results of the first experiment show that smooth pursuit eye movement during a retention interval produces significantly greater disruption in spatial span than do continuous attention shifts carried out while the eyes remain fixated. However, unlike the smooth pursuit condition there was no objective measure of participants compliance with experimental instructions in the attention shift condition. Despite the fact that all of the participants indicated that they had complied with the experimental instructions, it remained a possibility that on some or all of the trials they may not have attended to the moving target. Experiment 2 addressed this issue by repeating the procedure of Experiment 1, but with the addition of requiring participants to monitor and respond to colour changes of the target in both pursuit and attention shift conditions. EXPERIMENT 2 Effects on spatial span of smooth pursuit eye movements and continuous attention shifts with additional monitoring and report of target colour changes Method Secondary tasks The secondary tasks used in Experiment 1 were modified to include random colour changes of the target from white to red. Each colour change lasted 200 ms, with up to three changes occurring over the 5-s retention interval. There was at least a 750-ms delay between target changes. Participants were instructed to make a mouse click as quickly as possible each time a colour change occurred. These responses were recorded to allow a measure of participants compliance with the experimental instructions in the continuous attention shift only condition. All other aspects of the procedure were identical to those of Experiment 1. Results and discussion Mean spans in each condition were 4.99 (SD = 0.49) for baseline, 3.20 (SD = 0.96) for pursuit, and 3.87 (SD = 1.05) for continuous attention shifts. It was not necessary to re-run or omit any trials. The results were analysed by a one-way ANOVA that revealed a main effect of condition, F(2, 22) = 28.01, p <.001. Post hoc Tukey comparisons revealed that all differences were significant at the.02 level. Mean accuracy of detecting target colour changes in the attention shift condition was 98.66% and 100% in the smooth pursuit condition, confirming that participants were attending to the moving target as instructed. The classification of errors included false alarms as well as failure to respond to a target change.

9 OCULOMOTOR CONTROL AND SPATIAL WORKING MEMORY 1097 These results replicate the finding reported in Experiment 1, in which smooth pursuit eye movement during a retention interval produced significantly greater disruption of spatial span than continuous attention shifts carried out with the eyes fixated. Participants high level of accuracy for responding to target changes in the continuous attention shift condition demonstrates that this difference cannot be explained by a failure to comply with the experimental instructions. EXPERIMENT 3 Effects on spatial span of articulatory suppression and dynamic visual noise The results of the previous two experiments have shown that continuous eye movement produces significantly greater disruption of spatial span than do continuous covert shifts of attention with eyes fixated. What is not clear, however, is whether the eye movements themselves are a crucial factor in producing these results. An alternative explanation is that any requirement of participants to produce a rapid concurrent response will lead to a significant reduction in span. In order to address this, a third experiment was carried out in which participants carried out articulatory suppression during the retention interval. Articulatory suppression is a widely used technique for disrupting the operation of verbal working memory (Andrade, 2001; Baddeley, 1986), but it is believed to have a minimal effect on visuo-spatial working memory (Logie, 1995; Smyth et al., 1988). Therefore, if continuous eye movement only disrupts the operation of spatial working memory, it can be predicted that there should be no comparable effect produced by concurrent articulatory suppression. A second alternative to eye movement as the cause of disruption is that movement of the eye produces a change in the retinal image. This change alone could reduce span, either by a masking effect (Sperling, 1967; Turvey, 1973), or perhaps by potentially disrupting the operation of a cognitive retention strategy based on visual mental imagery (Andrade, Kemps, Werniers, May, & Szmalec, 2002; Smyth & Pendleton, 1989). In order to investigate this a secondary task condition was added in which a rapidly changing visual noise display filled the screen during the retention interval while participants maintained fixation. In addition to acting as a perceptual mask, dynamic visual displays of this type have also been shown to significantly disrupt tasks that rely on the formation of mental images (McConnell & Quinn, 2000; Quinn & McConnell, 1996; Smyth & Waller, 1998). Method Secondary tasks Articulatory suppression consisted of participants repeating the word go once every half a second, following procedures previously reported in the literature (i.e., Baddeley, 1986; Murray, 1968). The audio signals from a metronome were used to demonstrate to participants the correct rate at which to perform the suppression during the practice trials. No audio cue was provided during the experiments, and the experimenters used the visual signals from the metronome to ensure participant compliance. Any trials in which performance of the secondary task fell below 100% were re-run. Participants were instructed to begin suppression immediately at the start of the retention interval and to perform it continuously throughout the 5-s duration.

10 1098 PEARSON AND SAHRAIE The dynamic visual noise condition was based on previous displays reported in the literature (Pearson et al., 1999; Quinn & McConnell, 1996). During the retention interval the screen was filled with a display of approximately 2924 black and white circles per second (see Figure 1). The display ran at a frame rate of 85 Hz, with approximately 34 new items appearing per frame at random locations excluding the fixation crosshair. The fixation point was visible throughout, and participants were instructed to maintain fixation and ignore the changing display as much as possible. In all other respects the procedure for both secondary task conditions was identical to that for the baseline condition described previously. Performance of the secondary tasks was monitored by the experimenters throughout testing to ensure compliance with the experimental instructions and prevent any primary task/secondary task trade-off. Results and discussion Mean spans in each condition were 4.95 (SD = 0.57) for baseline, 4.85 (SD = 0.98) for articulatory suppression, and 5.01 (SD = 0.67) for dynamic visual noise. It was not necessary to re-run or omit any trials. The results were analysed by a one-way ANOVA that revealed no significant effect of condition (F <1,ns). Neither concurrent articulatory suppression during the retention interval nor the presence of dynamic visual noise led to any significant reduction in span in comparison to the baseline condition. The results of Experiment 3 suggest that the reduction in spatial span caused by continuous eye movement cannot be accounted for simply in terms of the demands caused by producing a regular response, or by changes in the retinal image during the retention interval. The absence of any significant effect by suppression and visual noise also suggests that participants are not relying on a visual imagery strategy or verbal recoding in order to remember the spatial sequences (Avons & Mason, 1999; Smyth & Pendleton, 1989). EXPERIMENT 4 Effects on spatial span of saccadic eye movements, discrete attention shifts, and limb movements The previous experiments have shown that continuous eye movements produce significantly greater reduction in spatial span than do equivalent covert attention shifts with eyes fixated. However, it is not clear whether other forms of eye movement can also produce this effect. As discussed in the Introduction, functional imaging studies have demonstrated differences in underlying neuronal mechanisms mediating saccadic and smooth pursuit eye movements (Petit et al., 1997). Differences between the different types of eye movement have also been shown for pointing accuracy to just-extinguished visual targets (Honda, 1984). Therefore, it is of interest to investigate whether the saccadic eye movements can produce the same effect on spatial span as can continuous eye movements. In order to investigate this, a fourth experiment was carried out in which continuous eye movements and attention shifts were replaced by saccadic eye movements and equivalent discrete attention shifts with the eyes fixated. A third secondary task was added in the form of concurrent limb movement in which participants tapped between two buttons using their preferred hand. According to Lawrence et al. (2001) all spatially directed movement should produce similar effects in spatial working

11 memory, irrespective of whether the eyes or limbs are being moved. This experiment allows a direct test of this hypothesis. If Lawrence et al. (2001) are correct, then it should be expected that there would be no significant difference in the level of disruption caused by saccadic eye movement and limb movement. Alternatively, if it is the case that eye movements produce a unique form of interference in spatial working memory, then it can be predicted that the saccadic eye movement condition will produce a significant or even a greater reduction in span than either concurrent limb movements or covert discrete shifts of attention. Method Secondary tasks In the saccadic eye movement condition participants were asked to track a ball that appeared alternating in position left and right of the screen. The ball appeared in the same position as the fixation point at the beginning of the retention interval and then appeared on the right or left on a randomly determined basis. Participants were instructed to keep their eyes focused on the ball throughout the duration of the retention interval. As soon as the ball disappeared from one side of the screen and reappeared on the other, participants had to shift their eyes as quickly as possible to the new position. The ball s movement mimicked the amplitude and frequency of the continuous tracking condition used in Experiments 1 and 2. The discrete attention shift condition was identical to the saccadic movement condition, except that the crosshair fixation point was present throughout the retention interval. Participants were instructed to maintain fixation while focusing their attention on the ball as it alternated position between the left and right of the screen. In both the saccadic and discrete attention shift conditions participants were required to monitor and report changes in the target colour from white to red as described in the procedure for Experiment 2. In the limb movement condition two blocks were positioned in front of the computer monitor at the same spatial separation as the amplitude of the moving ball in the saccadic and attention shift conditions. Participants were instructed to tap between the two blocks with their preferred hand at the same frequency as the moving ball in the other conditions. As in Experiment 3, a metronome was used to aid participants in establishing the correct rate of movement, but this was not audible during the actual experimental trials. During the retention interval the monitor screen remained blank except for the fixation point. Participants were required to maintain fixation throughout performance of the limb movements to ensure that any interference caused by the tapping was not due to extraneous visual processing. As in the previous experiments, participants EOG traces were monitored continuously during both conditions to ensure compliance with the experimental instructions. The experimenters also monitored participants secondary task performance in the limb movement condition using a light metronome (which was not visible to participants). Any deviation from the instructions or fall in secondary task performance from 100% would result in the given trial being re-run. Results and discussion OCULOMOTOR CONTROL AND SPATIAL WORKING MEMORY 1099 Mean spans in each condition were 5.21 (SD = 0.57) for baseline, 2.52 (SD = 1.20) for saccadic, 4.20 (SD = 0.81) for limb movement, and 3.82 (SD = 0.90) for discrete attention shifts. It was not necessary to re-run or omit any trials. The results were analysed by a one-way ANOVA that revealed a main effect of condition, F(3,33) = 41.05; p <.001. Post hoc Tukey comparisons revealed that there was no significant difference between the discrete attention shift and limb movement conditions (p >.41). All other comparisons were significant at the

12 1100 PEARSON AND SAHRAIE.001 level. Mean accuracy of detecting target colour changes in both the discrete attention shift and saccadic conditions was 100%, confirming that participants were attending to the moving target as instructed. These results show that saccadic eye movements performed during a retention interval produce significantly greater reduction in span than do either discrete covert attention shifts or limb movements that are performed at the same amplitude and frequency as the eye movements. These findings do not support the claim made by Lawrence et al. (2001) that all spatially directed movement produces similar effects in visuo-spatial working memory. However, the absence of any significant difference between the discrete attention shift and limb movement conditions provides support for the hypothesis that the disruption caused by limb movement is due to related shifts in attention rather performance of the movement itself (Awh & Jonides, 2001; Awh et al., 1998; Lawrence et al., 2001; Smyth, 1996). EXPERIMENT 5 Effects on spatial span of saccadic eye movement with eyes closed, eyes closed with no movement, and free eye movement conditions The previous experiments have established that concurrent eye movement during a retention interval produces significantly greater reduction in spatial span than do comparable conditions in which limb movement or attention shifts are performed with the eyes fixated. However, it remains uncertain as to the extent to which the eye movements themselves are responsible for this effect rather than a related factor. The use of dynamic visual noise in Experiment 3 established that changes to the retinal image during the retention interval did not lead to a reduction in span. What is not clear, however, is whether eye movements performed without any corresponding shift in retinal image will still lead to a reduction in spatial span. To investigate this, the final experiment required participants to execute saccadic movements with their eyes closed during the retention interval. In addition to removing any shift in retinal image as the eyes move, this condition also had the advantage that the eye movements were not executed in response to visual targets. A large number of studies have demonstrated that shifts of attention precede saccadic eye movements to visual targets (Shepherd, Findlay, & Hockey, 1986; Chelazzi, Biscaldi, Corbetta, Peru, & Berlucchi, 1995). In the present experiment the saccades were self-generated by the participant and were not directed to specific targets in their visual field. Of course, requiring participants to shut their eyes during the retention interval might in itself disrupt performance. To control for this a condition was added in which participants were instructed to close their eyes but to attempt to keep them as still as possible. The final condition in the experiment was one in which the fixation point was removed, and participants were allowed to make any voluntary eye movements they wished during the retention interval. As Baddeley (1986) has suggested that participants may use overt eye movement as a rehearsal mechanism to maintain sequences of locations, requiring participants to fixate during the retention interval might in itself lead to a reduction in spatial span. The final condition was designed to test this.

13 Method Secondary tasks OCULOMOTOR CONTROL AND SPATIAL WORKING MEMORY 1101 In the saccadic movement with eyes closed condition participants were initially trained using a metronome to make saccades with the same frequency as that used in the saccadic movement condition in the previous experiment. EOG signals were monitored to ensure that participants executed the eye movements correctly. During the actual experimental trials the metronome was not present, and participants initiated each saccade themselves, with EOG signal monitoring ensuring that the correct number of saccades were carried out for each trial. Participants were instructed to close their eyes and begin saccadic movement as soon as the retention interval began, and to continue this until the short audio tone indicated the end of the retention interval. The second condition was identical to the first, except that participants were instructed to keep their eyes as still as possible during the retention interval. Again, the EOG signals were monitored to ensure compliance with these instructions. Finally, in the free eye movement condition the crosshair fixation point was removed during the retention interval, leaving a blank screen. Participants were told that they were free to make any eye movements they wished during the interval. However, EOG signals were still recorded so as to provide an indication of how participants responded to these instructions. All other aspects of the procedure were the same as those in the previous experiments. Results and discussion Mean spans in each condition were 5.30 (SD = 0.67) for baseline, 3.13 (SD = 1.13) for saccades with eyes closed, 4.93 (SD = 0.51) for eyes closed with no movement, and 5.06 (SD = 0.61) for free eye movement. It was not necessary to re-run or omit any trials. The results were analysed by a one-way ANOVA that revealed a main effect of condition, F(3, 33) = 41.79; p <.001. Post hoc Tukey comparisons revealed that the eyes shut with no movement and free eye movement conditions did not significantly differ either from each other (p >.93) or from the baseline condition (p >.33). All other comparisons were significant at the.001 level. These results demonstrate that saccadic eye movements significantly reduce span even when they are not directed towards targets that are present in the visual field. The lack of any effect from the eyes shut with no movement condition demonstrates that this disruption is not simply due to participants closing their eyes. The results from the free eye movement condition show that the presence of a fixation point during the retention interval does not in itself reduce participants spatial span. However, analysis of the EOG recordings showed that participants did not appear to be making any form of overt eye movement during the retention interval, even though they were free to do so. Although there was some evidence in the practice trials that participants started by adopting a strategy in which they used eye movements to attempt to retain the locations, this was rapidly abandoned in favour of a strategy in which the eyes were kept as still as possible. Indeed, the EOG signals in this condition were indistinguishable from the signals recorded in the baseline condition that required fixation. In addition, when questioned participants indicated it was their subjective experience that overt eye movements were actually making it harder to retain the presented locations accurately. These results are discussed further in the General Discussion section below.

14 1102 PEARSON AND SAHRAIE COMPARISON OF CONDITIONS ACROSS EXPERIMENTS 1 TO 5 In all five of the experiments described previously the eye movement conditions produced significantly greater reduction in span than did comparable attention shift or limb movement conditions. However, as discussed in the introduction, it is important to compare the relative size of these decrements across different experiments. Figure 3 shows the mean baseline performance for the 12 participants across the five experimental testing sessions. ANOVA revealed no significant change in baseline span across different sessions either for participants as a group (F <1,ns) or when considering the performance of each participant separately: for each participant, 0.43 < F(4, 15) < 3.01; p = ns. There is no indication of a practice effect developing across experiments. These results indicate that the procedure adopted for sampling participants spatial span was reliable. Following this a percentage change in performance for each condition was calculated by subtracting the mean span for each condition from the baseline span established in the same testing session. These scores are presented in Figure 4. Dependent samples t tests were carried out to compare performance across conditions. Interference caused by the saccadic eye movement condition was significantly greater than both the smooth pursuit condition, t(11) = 3.133, p <.01, and the eyes shut and saccadic movement condition, t = 2.26, p <.05. Performance in the saccadic movement with eyes closed condition did not significantly differ from the smooth pursuit eye movement condition, t(11) = 0.410, p >.69, but did cause significantly greater reduction in span than did continuous shifts in attention, t(11) = 2.628, p <.02, discrete attention shifts, t(11) = 2.564, p <.02, and concurrent limb movement, t(11) = 5.535, p <.001. Figure 3. Illustration of the mean span, standard deviation, and standard error of participants baseline performance across Experiments 1 to 5.

15 OCULOMOTOR CONTROL AND SPATIAL WORKING MEMORY 1103 Figure 4. Illustration of the percentage change from baseline for all conditions across Experiments 1 to 5. Conditions are identified as follows: (A) free eye movement; (B) eyes shut with no movement; (C) visual noise; (D) articulatory suppression; (E) continuous covert attention shifts; (F) continuous covert attention shift plus colour change monitoring and report; (G) discrete covert attention shifts; (H) spatial tapping; (I) smooth pursuit eye movement; (J) smooth pursuit eye movement plus colour change monitoring and report; (K) saccadic eye movement with eyes shut; (L) saccadic eye movement with eyes open. The interference caused by concurrent limb movement was not significantly different from performance in the discrete attention shift condition, t(11) = 0.257, p >.802, and continuous and discrete attention shift conditions did not significantly differ from each other, t(11) =.257, p >.802. Finally, there was no significant difference in performance between continuous eye movements with and without response to target colour change, t(11) = 0.194, p >.849, and continuous attention shifts with and without responses, t(11) = 1.386, p >.193. This demonstrates that any additional executive load imposed by the requirement to monitor target colour changes was not in itself disruptive to participants performance. Comparison across the conditions reveals a pyramid of effects that is summarized in Figure 5. All conditions that involved eye movement (I, J, K, and L) produced significantly greater reduction in spatial span than did conditions that involved shifts in attention with the eyes fixated (E, F, G, and H). The greatest reduction from baseline occurred in the saccadic eye movement condition, which produced significantly greater disruption in comparison to all other conditions. In an attempt to explore these findings further, an examination was carried out on the effect of different secondary tasks on the types of error made by participants during recall of the spatial sequences. This was done by categorizing the first recall error in a trial as either a spatial (S) or a temporal (T, order) error. In any given trial participants were recalling a sequence of spatially and temporally distributed events. If the first error in the trial was due to the participant reporting the incorrect temporal order of that event, then it was termed a temporal error (as they had still reported a location that was part of the correct sequence, but not the correct

16 1104 PEARSON AND SAHRAIE Figure 5. Illustration of the hierarchy of conditions and corresponding distribution of temporal and spatial errors occurring for the last two incorrectly recalled trials in each experimental run. order in which it had occurred). On the other hand, if the participant reported a block that had not been presented in the original sequence (i.e., a location error), it was termed a spatial error. Any subsequent errors after the first one could not be reliably categorized as being spatial or temporal and were ignored. Errors for all conditions were categorized in this way for the last two incorrect trials that had caused a particular experimental run to be terminated. The results of this analysis are also summarized in Figure 5, which shows that the distribution of type of error maps on to the pyramid of effects previously described. Repeated measures one-way ANOVAs were used to analyse these results. The proportions of spatial and temporal errors in conditions A, B, C, and D were 25% (S) and 75% (T), which were not significantly different from those found in the baseline conditions (F <1;ns). The ratio of errors for conditions involving covert shifts in attention was 42.4% (S) and 57.6% (T). Although there were no significant differences in these ratios between conditions E, F, and G, involving shifts in attention and limb movements, there were significantly more spatial errors than in the baseline condition and conditions A, B, C, and D, F(1, 82) = 15.3; p <.001. In turn, eye movement conditions produced higher spatial errors than those including shifts in attention alone. These ratios were 51.6% (S) and 48.4% (T) for the smooth pursuit conditions and saccadic movement with eyes shut (I, J, and K), and 61.5% (S) and 38.5% (T) for saccadic eye movements (L). Again, the incidence of spatial error for eye movement conditions was significantly higher than those including shifts in spatial attention alone, F(1, 170) = 6.27; p <.01. In all conditions, the first error of recall occurred nearly a third of the way through the sequence (grand average mean 3.11, SD = 0.11). There were no significant differences across any of the conditions for the position in the sequence of the first error. The theoretical probability of making a spatial error in the saccadic eye movement condition was also determined. On average, three visual events had been presented for these