Journal of Vision (2003) 3, 751-760 http://journalofvision.org/3/11/9/ 751

Differential effects of the Müller-Lyer illusion on reflexive and voluntary

Department of Psychology Jason S. McCarley Mississippi State University, Mississippi State, MS, USA Beckman Institute and Department of Psychology, Arthur F. Kramer University of Illinois at Urbana-Champaign, Urbana, IL, USA

Department of Experimental Psychology, Gregory J. DiGirolamo University of Cambridge, Cambridge, England

Research has produced conflicting evidence as to whether programming is or is not biased by perceptual illusions. However, previous studies have generally not distinguished between effects of illusory percepts on reflexive saccades, programmed automatically in response to an external visual signal, and voluntary saccades, programmed purposively to a location where no signal has occurred. Here we find that voluntary and reflexive saccades are differentially susceptible to the Müller-Lyer illusion; reflexive movements are reliably but modestly affected by the illusion, whereas voluntary movements show an effect similar to that of perceptual judgments. Results suggest that voluntary saccade programming occurs within a non-retinotopic spatial representation similar to that of visual consciousness, whereas reflexive saccade programming occurs within a representation integrating retinotopic and higher level spatial frames. The effects of the illusion on reflexive saccades are not subject to endogenous control, nor are they modulated by the strength of an exogenous target signal.

Keywords: saccades, visually guided behavior, illusions

DeSouza, and Goodale (1995), for example, found large Introduction effects of the Titchener illusion on psychophysical judgments of object size, but found no effect on grip Perception and Action: Separate or scaling. Bridgeman and colleagues, similarly, found that perceptual judgments but not pointing responses were Common Representations? biased by induced motion of a target object (Bridgeman, Despite much study, researchers have yet to reach Kirch, & Sperling, 1981). Such findings are consistent consensus as to whether and how conscious perception with the hypothesis that perception and action are served and the control of action are linked. As described by by distinct spatial representations, with only the former Franz and colleagues (Franz, Fahle, Bülthoff, & representation being susceptible to illusion. Gegenfurtner, 2001), the relationship between perception Neurophysiological and neuropsychological evidence for and action has been hypothesized to take three different distinct visual cortical streams, one primarily responsible forms. The strong separate representation model proposes for object recognition and the other for spatial that distinct neural representations underlie motor representation and visuomotor control (Goodale & behavior and conscious . The weak Milner, 1992), provides a plausible biological basis for separate representation model likewise posits that separate this dissociation. spatial representations exist in the brain, but allows for Other data, however, cast doubt on the strong crosstalk, modulated by task demands, to occur between separate representation model. Consistent with a weak them. The common representation model, finally, holds that separate representation model, some findings suggest that conscious perception and visuomotor control proceed illusion affects some aspects of visually guided movement from the same mental representation.1 (e.g., velocity of reach and grip force) but not others (e.g., Research in healthy observers has attempted to grip scaling) (Brenner & Smeets, 1996; Jackson & Shaw, distinguish among these models by measuring the effects 2000), or that dissociations between perception and of perceptual illusions on visually guided behavior. visually guided action are mediated by task demands Though counterintuitive, the strong separate (Bridgeman, Peery, & Anand, 1997; Gentilucci, Chieffi, representation model has received support from findings Daprati, Saetti, & Toni, 1996; Wraga, Creem, & Proffitt, suggesting that visually guided behavior is uninfluenced 2000). Other data appear to support a common by illusions that are evident in subjective reports. Aglioti, representation model. Experiments by Franz and

doi:10.1167/3.11.9 Received April 4, 2003; published December 4, 2003 ISSN 1534-7362 © 2003 ARVO

Downloaded from jov.arvojournals.org on 10/02/2021 McCarley, Kramer, & DiGirolamo 752

colleagues (Franz et al., 2001; Franz, Gegenfurtner, by data from an alternative line of research. Studies Bülthoff, & Fahle, 2000) found similar effects of the beginning as early as 1897 (Delabarre) have consistently Titchener and Müller-Lyer (M-L) illusions on perception demonstrated that saccadic eye movements are susceptible and grasping, while a study by Lopez-Moliner et al. (2003) to misperceptions induced by the M-L illusion; when demonstrated equal effects of a pictorial depth illusion on observers are asked to saccade toward the endpoint of an perception and manual tracking. A study by Vishton, Read, Cutting and Nunez (1999) found that subjective M-L figure, or to saccade from one endpoint of the figure reports and grip scaling were both susceptible to the to another, landing positions are reliably horizontal-vertical illusion, but only when observers were biased in the direction of illusory changes in the figure’s induced to encode the horizontal and vertical dimensions length (Binsted & Elliott, 1999; Delabarre, 1897; of the stimulus relative to each other. When observers Festinger, White, & Allyn, 1968; Stratton, 1906; Yarbus, were induced to encode the size of one dimension, 1967). This effect obtains even when the stimulus figure ignoring the orthogonal dimension, no illusion was remains visible throughout saccade execution and evident in either form of response. These findings are preparation, and as such is not the product of memory- consistent with a common representation model in which guided targeting. the frame of reference induced by task demands Visually guided saccades thus appear to be susceptible determines what effect an illusion will have on a response, independent of the response mode. to illusion under some circumstances and resistant under A complicating factor in the effort to distinguish others. Given large differences between the methodology between strong separate, weak separate, and common of Wong and Mack (1981) and that of Delabarre (1897), representation models using perception-action Yarbus (1967), Festinger (1968) and others, in addition to dissociations, it should be noted, is that different illusions the inherent differences between the induced motion and may arise at different points within the M-L illusions, it is not immediately obvious what the stream. An illusion that arises early in processing and is relevant circumstances are. One mediating factor, propagated forward might therefore affect visually guided however, may be the difference between behavior even if perception and visuomotor behavior are reflexive/reactive/exogenously triggered and ultimately based on separate representations. Visuomotor voluntary/volitional/endogenously triggered saccades (Deubel, behavior will be resistant to an illusion only if the representations supporting perception and action are 1995; Erkelens & Hulleman, 1993; Klein, Kingstone, & indeed separate, and only then if the illusion has its locus Pontefract, 1992; Klein & Shore, 2000). Reflexive beyond the point at which those representations diverge saccades, although they can be inhibited or modified by (Dyde & Milner, 2002). top-down control (Machado & Rafal, 2000a, 2000b; Rafal, Machado, Ro, & Ingle, 2000), are programmed Effects of Illusions on Oculomotor automatically, presumably by collicular mechanisms Behavior (Rafal, Smith, Krantz, Cohen, & Brennan, 1990), in response to a transient external signal at a saccade target Like the studies of reaching and grasping described location. In essence, they reflect a visual grasp reflex above, studies of oculomotor behavior have produced (Rafal et al., 2000). Voluntary saccades are programmed apparently conflicting evidence for perception/action purposively, apparently by cortical mechanisms (Henik, dissociations. A seminal study by Wong and Mack (1981) Rafal, & Rhodes, 1994), in the absence of a transient suggested that visually guided saccade programming is signal to mark the target location. Thus, while exogenous immune to illusion. Observers were asked to saccade and endogenous influences may interact to shape saccadic toward a stepped target, then to saccade from memory behavior in a particular task (Godjin & Theeuwes, 2002; back to the location of the original fixation (where no Kopecz, 1995; Trappenberg, Dorris, Munoz, & Klein, fixation marker remained). Motion of a surrounding 2001), the presence or absence of an external signal to frame was used to induce illusory changes in the target directly specify a movement’s direction and amplitude step. Data revealed that eye movements toward the target appears to distinguish between qualitatively different were impervious to illusory motion; saccade targeting was subclasses of movement (Deubel, 1995; Erkelens & accurate despite changes in the perceived length or even Hulleman, 1993). direction of the target step. Only when observers made Past research into the effects of the M-L illusion on memory-guided return movements to the location of the saccadic eye movements has generally asked subjects to original saccade launch point was any influence of the saccade from end-to-end of an M-L figure, with no illusion evident in oculomotor behavior. transient signal to mark saccade target locations. As such, The conclusion that saccades toward a visible movements were necessarily voluntary. In contrast, in the stimulus are resistant to illusion, however, is challenged

Downloaded from jov.arvojournals.org on 10/02/2021 McCarley, Kramer, & DiGirolamo 753

experiment described by Wong and Mack (1981), visually and to compare the magnitude of these effects to that of guided eye movements were toward a stepped target—a the illusion on conscious perception. Toward that end, transient signal—and as such were likely to have been observers were asked to execute voluntary and reflexive largely reflexive. This suggests that the effects of a visual saccades from the center to the endpoint of an M-L illusion may in fact differ for voluntary and reflexive configuration or a size-matched control figure. Stimulus parameters were chosen so that saccade targets within M- saccades, and more specifically that reflexive but not L figures appeared to be equidistant from the movement voluntary saccades may be resistant to illusion. The goal launch point, when in reality one target was 1.3° farther of the present research was to explore this possibility (see than the other. By comparing saccades toward targets of also DiGirolamo, McCarley, & Kramer, 2001). The different physical distances, therefore, it was possible to observers’ task in the current experiments was to saccade assess the degree to which movement amplitudes deviated from a central fixation point to one endpoint of a from perceived target distances. Brentano-style M-L configuration or a size-matched control stimulus (Figure 1). Reflexive saccades were Method produced by presentation of a transient go-signal at the saccade target location. Voluntary saccades were produced Observers by use of a spoken go-signal. Experiment 1 revealed that Observers were eight young adults recruited from the both reflexive and voluntary saccades are susceptible to community of the University of Illinois at Urbana- the M-L illusion, but to different degrees; while reflexive Champaign. All observers had normal or corrected-to- normal vision. Observers were paid for participating. One saccades are modestly affected, voluntary saccades show observer, whose anticipatory movement rates were effects similar to those of subjective judgments. unacceptably high (53% in the reflexive saccade Experiment 2 demonstrated that the modest effects of the condition), was replaced. This did not change the pattern M-L illusion on reflexive movements were not the result of results described below. of endogenous or anticipatory saccade planning, and were not modulated by the strength of the saccade go-signal. In Apparatus total, results suggest that reflexive saccade programming Stimuli were presented on a 21” monitor at a occurs within a representation that is a weighted resolution of 800 x 600 pixels and a refresh rate of 85 Hz. Eye movements were recorded with an Eyelink eye tracker combination of retinotopic and higher level spatial (SR Research Ltd.) with a temporal resolution of 250 Hz frames, whereas voluntary saccade programming occurs and a spatial resolution of 0.2°. An eye movement was within a reference frame more similar to that of visual classified as a saccade when its distance exceeded 0.2° and consciousness. its velocity reached 30°/s, or when its length exceeded 2 Illusion 0.2° and its acceleration had reached 9500°/s . Subjects viewed displays from a distance of 91 cm, with viewing distance controlled by a chin rest. Stimuli Control Stimuli were horizontally oriented Brentano-style M-L figures and size-matched control figures with vertical lines as wings. The middle wing of each stimulus divided the horizontal shaft into two segments of different lengths, one of 8.9° and the other of 7.6°. Within M-L configurations, the physically longer segment was always Figure 1. Illustration of the stimuli of Experiment 1. Stimulus wings-in (illusively short), and the physically shorter dimensions were chosen such that wings-in and wings-out segment always wings-out (illusively long). Stimulus segments of the M-L stimulus were approximately equal in dimensions were chosen on the basis of a psychophysical apparent length. Stimuli for Experiment 2 were identical to the pilot experiment (n = 2, method of constant stimuli) to M-L figures of Experiment 1, except that wings-in and wings- produce the perception that wings-in and wings-out out segments were matched in physical length and differed in segments of the illusion were of approximately equal apparent length. extent (DiGirolamo, McCarley, & Kramer, 2001). Stimulus orientation (long segment on the left side vs. long segment on the right side) was counterbalanced Experiment 1 within observers. The line segments that formed the ° The aim of Experiment 1 was to examine the effects wings of the stimuli were 2.1 in length. In the M-L ° of the M-L illusion on voluntary and reflexive saccades, figures, wings were angled at 30 relative to the shaft of

Downloaded from jov.arvojournals.org on 10/02/2021 McCarley, Kramer, & DiGirolamo 754

the figure. In the control figures, they were at right angles 750 ms. This resulted in the loss of 5% of all trials, relative to the shaft of the figure. Figures were drawn in approximately equivalent across reflexive saccade and gray (19.6 cd/m2) against a black (3.6 cd/m2) background. voluntary saccade conditions. Also excluded from analysis were trials on which the saccade was not in the Procedure appropriate direction. As expected, saccade direction Procedure is illustrated in Figure 2. The observer errors were less frequent for reflexive than for voluntary began each trial by pressing a key while fixating a filled saccade trials (3% vs. 22%), and saccade latencies were gray circle (0.12° in diameter) in the center of an shorter (182 ms vs. 365 ms). Neither error rates nor otherwise empty display. The stimulus figure for that trial latencies varied as a function of saccade target distance or appeared immediately thereafter, with the vertex of the stimulus configuration. middle wing centered on the fixation mark. A go-signal followed after a delay of 506 ms. On reflexive saccade Saccade Amplitudes trials, the go-signal was a filled white circle (90.7 cd/m2, Figure 3 presents mean horizontal amplitudes for 0.15° diameter), which appeared for 59 ms at one end of reflexive (top) and voluntary (bottom) saccades. Note that the stimulus figure. The observer’s task was to saccade to within M-L figures, line segments of different physical the end of the horizontal shaft at which the go-signal lengths were of approximately the same apparent length. appeared. On voluntary saccade trials, the go-signal was a Thus, to the extent that saccade amplitudes were spoken word “left” or “right,” presented through the modulated by the M-L illusion, the difference between experimental computer’s speakers. The observer’s task was saccade amplitudes for near and far targets should have to saccade to the end of the horizontal shaft specified by been smaller than in the control conditions. the auditory signal. Reflexive and voluntary saccade trials 10.0 were run in two separate blocks, with order of blocks Reflexive Saccades counterbalanced across observers. Each block was composed of 20 randomly chosen practice trials, then 100 9.0 randomly ordered experimental trials for each combination of saccade target distance (8.9° vs. 7.6°) and stimulus configuration (M-L or control). 8.0

Reflexive Voluntary 7.0

Fixation M-L Figure Control Figure 6.0 7.6 8.9 10.0 Voluntary Saccades Go-signal 9.0 Saccade Amplitude (degrees) "left" 8.0 Saccade

7.0

M-L Figure Control Figure Figure 2. Illustration of the procedure of Experiment 1. In 6.0 reflexive saccade conditions, observers made eye movements 7.6 8.9 toward a flashed go-signal. In voluntary saccade conditions, Saccade Target Distance (degrees) observers made eye movements toward the end of the stimulus figure specified by a spoken go-signal. Figure 3. Reflexive (top) and voluntary (bottom) saccade amplitudes for Experiment 1. Results For omnibus analysis, amplitudes were submitted to a 2 X 2 X 2 ANOVA with saccade type (reflexive vs. Data Loss, Error Rates, and Saccade Latencies voluntary), saccade target distance (7.6° vs. 8.9°), and Trials on which shifted prior to onset of the go- stimulus configuration (M-L vs. control) as within-subjects signal were excluded from analysis, as were trials on factors. As can be seen in Figure 3, amplitudes for both which saccade latency was less than 50 ms or greater than reflexive and voluntary saccades were modulated by the

Downloaded from jov.arvojournals.org on 10/02/2021 McCarley, Kramer, & DiGirolamo 755

M-L illusion; for both classes of movement, the difference short latency movements are less susceptible than long in amplitude for near and far targets was smaller within latency movements. A related possibility is that the the M-L configurations than within control figures. seeming effect of the M-L illusion on reflexive saccade Effects of the illusion, however, were greater for voluntary amplitudes might actually have resulted from voluntary than reflexive movements, with reflexive saccade movements occasionally produced in response to a visual amplitudes being biased by a mean of .29°, SE = .03, in go-signal. The distribution of saccade amplitudes the direction of the perceived changes in target distance produced in response to a visual go-signals, in other (compared to saccade amplitudes within control figures) words, might have comprised a mixture of voluntary and voluntary saccade amplitudes being biased by a mean saccade amplitudes showing a relatively large effect of the of .77°, SE = .08. Voluntary amplitudes were in fact illusion and reflexive saccade amplitudes showing no slightly larger for targets that were 7.6° distant than for effect. This hypothesis, like the differential-latency targets that were 8.9° distant. This finding may indicate hypothesis noted above, suggests that the effects of either that perceived target distances within the M-L stimulus configuration on reflexive saccade amplitude stimuli were not perfectly matched, or that the effects of should be larger for long-latency than for short-latency the illusion were stronger on voluntary movements than saccades. on perception. In either case, it is clear that reflexive An additional concern is that differential effects of saccade programming was based on a weighted average of wings-in and wings-out patterns on movement amplitudes actual and perceived target distances, whereas voluntary might reflect a center-of-gravity tendency in saccade saccade programming was dominated by perceived target targeting (Coren & Hoenig, 1972; Findlay, 1982; He & distance. Statistical analysis confirmed these observations. Kowler, 1989), rather than the influence of an illusory A reliable main effect of saccade target distance [F(1, 7) = percept, per se. Data would then indicate that reflexive 120.157, p < .001, MSE = .057] indicated that on average saccades are more resistant to the center-of-gravity saccade amplitudes were larger for more distant targets. tendency than are voluntary saccades, but would not This effect was qualified, however, by a two-way speak to the effects of illusion on oculomotor interaction of target distance by stimulus configuration programming. Existing data cast doubt on this possibility; [F(1, 7) = 130.846, p < .001, MSE = .033], confirming that although center-of-gravity targeting appears to be the saccade amplitudes were reliably modulated by perceived default strategy in eye movement programming, observers target distance, and by a three-way interaction of saccade can easily target non-central positions within a shape type by target distance by stimulus configuration [F(1, 7) = when instructed to do so (He & Kowler, 1991). It thus 34.830, p = .001, MSE = .027], confirming that the effects seems unlikely that a center-of-gravity tendency would of perceived target distance were larger for voluntary than have strongly influenced performance in the current for reflexive saccades. Nonetheless, two-way ANOVAs experiment, where observers were instructed to target a with stimulus configuration and target distance as factors specific and well-defined target position (the end of the produced reliable interactions for both voluntary and horizontal stimulus shaft). Nonetheless, it is useful to seek reflexive movements [F(1, 7) = 84.425, p < .001, MSE = additional evidence against the center-of-gravity .056 and F(1, 7) = 127.746, p < .001, MSE = .049], hypothesis. Evidence indicates, notably, that center-of- confirming that both classes of saccade were susceptible gravity effects are modulated by saccade latency. More to the illusion. Post hoc t tests revealed that reflexive specifically, center-of-gravity effects are larger for short- saccade amplitudes were shorter for near targets than for latency saccades than for long-latency movements (Coëffé far targets within M-L figures [t(7) = -13.424, p < .001, SE & O'Regan, 1987; Deubel, 1996; Ottes, van Gisbergen, = .048], whereas voluntary saccade amplitudes were & Eggermont, 1985). The center-of-gravity account of the shorter for far targets [t(7) = 2.674, p = .032, SE = .145]. current data therefore predicts that the M-L illusion should affect short-latency movements more strongly than Control Analyses long-latency movements. The results described above suggest that both To examine these various possibilities, saccade reflexive and voluntary saccades are modulated by amplitude data within each cell of the design were changes in perceived target distance produced by the M-L subjected to a median split on the basis of movement illusion, but that the effects of the illusion are larger for latency, and were reanalyzed with a 2 x 2 x 2 x 2 ANOVA voluntary movements. An additional analysis was that included saccade latency (below median vs. above conducted to rule out alternative interpretations of these median), saccade type, stimulus configuration, and data. One possibility is that these results are not evidence saccade target distance as within-subjects variables. of a functional difference between voluntary and reflexive Consistent with the possibility that the influence of the saccades, but were produced by differences in saccade illusion might vary with saccade latency, the analysis latency. As noted, voluntary latencies were considerably revealed marginally reliable interactions of saccade latency longer than reflexive latencies. Results might therefore be by saccade type by target distance [F(1, 7) = 5.196, p = taken to indicate not that reflexive saccades are less .057, MSE = .067] and of saccade latency by saccade type susceptible to illusion than voluntary saccades, but that by stimulus figure by target distance [F(1, 7) = 5.207, p =

Downloaded from jov.arvojournals.org on 10/02/2021 McCarley, Kramer, & DiGirolamo 756

.056, MSE = .105]. Upon examination, however, these & Williams, 1995; Dorris & Munoz, 1998). Such effects did not contradict any of the conclusions drawn endogenous saccade planning might be expected to have above. For closer analysis, voluntary and reflexive either of two contrary effects. The results of Experiment 1 movements were submitted to separate three-way indicate that the influence of the M-L illusion is larger for ANOVAs with saccade latency, stimulus configuration, endogenously generated than for exogenously generated and target distance as factors. A reliable three-way movements. This suggests that anticipatory movement interaction indicated that the effect of the M-L illusion on planning might exacerbate the effects of the illusion on voluntary saccades was stronger for long-latency than for reflexive saccades. Alternatively, it is possible that short-latency movements [F(1,7) = 7.464, p = .029, MSE = endogenous saccade preparation might make saccade .119] with mean magnitude of the bias produced by M-L targeting more accurate, reducing the illusion’s influence wings increasing from .58°, SE = .12, for saccades of much as foreknowledge of target position attenuates the below-median latency, to 1.05°, SE = .10, for saccades of center-of-gravity effect in saccade programming (Coëffé & above-median latency. Conversely, a nonsignificant three- O'Regan, 1987). Indeed, a failure to find that endogenous way interaction within the reflexive saccade data [F(1, 7) = preparation reduced the M-L configurations’ effects .205, p = .665, MSE = .005] suggested that latency did would provide additional evidence against a center-of- little to modulate the influence of the M-L illusion on gravity account of the above data. A third potential externally triggered movements, with mean effect outcome, of course, is that the effect of the illusion on magnitude being .33°, SE = .08, for saccades of below- reflexive saccade programming is immune to top-down median latency, and .24°, SE = .05, for saccades of above- influences and thus unaffected by anticipatory saccade median latency. Thus, contrary to a center-of-gravity preparation. To test these possibilities, Experiment 2 explanation, the influence of the M-L illusion did not employed a reflexive saccade task similar to that of tend to decrease as saccade latencies increased. Contrary Experiment 1, but manipulated target location to the possibility that differences in mean latency might probabilities to induce observers to prepare movements account for the differential susceptibility of reflexive and prior to onset of the go-signal. voluntary movements to the M-L illusion, or that a small The second goal of Experiment 2 was to examine the number of voluntary movements might have role of stimulus-driven or exogenous factors in contaminated the reflexive saccade data, latency did modulating the influence of the illusion on reflexive nothing to modulate the influence of the illusion on saccades. Past research has indicated that the strength or reflexive saccades. contrast of a flashed target has little effect on saccade accuracy (Deubel, 1996). This suggests that in the current Discussion reflexive saccade task, contrast of the go-signal should do little to modulate the effects of the M-L illusion. The results of Experiment 1 indicate that voluntary saccades are as susceptible to the M-L illusion as perceptual judgments. Reflexive saccades show far smaller Method effects. These results do not appear to have been Observers produced by a confound of saccade type with saccade latency, or by voluntary movements contaminating the Observers were 10 young adults recruited from the reflexive saccade distributions, or by a center-of-gravity community of the University of Illinois at Urbana- effect. Champaign. All observers had normal or corrected-to- In total, results are consistent with the hypothesis normal vision. Observers were paid for participating. that voluntary and reflexive saccades are differentially Apparatus susceptible to illusive changes in target distance. Contrary to expectations, however, reflexive movements in Apparatus were identical to those of Experiment 1. Experiment 1 were not fully resistant to illusion. A Stimuli second experiment was conducted to further investigate Stimuli were similar to those of Experiment 1, except the effects of the M-L illusion on reflexive saccade for the following exceptions made to simplify data programming. analysis. First, all stimuli were M-L configurations; vertical-wing control stimuli like those of Experiment 1 Experiment 2 were not used. Second, wings-in and wings-out segments were now of the same physical length (8.25°), and thus Experiment 1 revealed a modest but reliable influence differed in apparent length. of the M-L illusion on reflexive saccade programming. The first aim of Experiment 2 was to assess the role of Procedure top-down or endogenous factors in this effect. It is well Procedure was similar to that of Experiment 1, with known that observers can pre-plan saccades, anticipating the following exceptions. First, all trials used visual go- and preparing for upcoming movements (e.g., Carpenter signals. Thus, only reflexive saccades were studied.

Downloaded from jov.arvojournals.org on 10/02/2021 McCarley, Kramer, & DiGirolamo 757

Second, the luminance transients used as go-signals were strength by perceived target distance [F(1, 9) = 4.482, p = now of two different intensities. On weak signal trials, the .063, MSE = 357.439], suggesting that effects of perceived go-signal was relatively dim (49.2 cd/m2). On strong signal distance on saccade latency were significant only with trials, the go-signal was more intense (90.7 cd/m2). Third, weak exogenous signals. the go-signal was no longer equiprobable on both sides of 300 the display. Rather, the signal appeared on one side of the Weak signal, Low probability display on 80% of all trials, and on the opposite side on Weak signal, High probability only the remaining 20% of trials. The assignment of left 280 Strong signal, Low probability and right sides to high- and low-probability conditions Strong signal, High probability was counterbalanced across observers. Each observer (ms) 260

completed 10 randomly chosen practice trials, then 400 ency randomly ordered experimental trials, 200 with a high- 240 contrast go-signal and 200 with a low-contrast go-signal. 220 ccade Lat

Results a S 200 Data Loss and Error Rates Data from trials on which gaze shifted prior to onset 180 of the go-signal were discarded, as were trials on which Wings-in Wings-out saccade latency was <50 ms or >750 ms. This resulted in 10.0 the loss of 6% of all data. Saccade direction errors 9.5 occurred on <1% of all trials. Data from these trials were also excluded from the analyses reported below. 9.0 Saccade Latencies 8.5

To draw conclusions about the effects of cue contrast 8.0

and location probability on saccade amplitudes, it is first plitude (degrees) necessary to examine saccade latency data for m 7.5 independent evidence that manipulations of these factors were effective. Mean saccade latencies are presented in 7.0 ccade A

Figure 4 (top). For statistical analysis, latency data were a 6.5 submitted to a within-subjects ANOVA with signal S strength (weak vs. strong), target side (high probability vs. 6.0 low probability), and perceived target distance (near vs. Wings-in Wings-out far) as factors. Of foremost importance, a main effect of Figure 4. Saccade latencies (top) and amplitudes signal contrast indicated that latencies were shorter for (bottom) for Experiment 2. high-contrast than for low-contrast signals [F(1, 9) = 71.216, p < .001, MSE = 302.133], and a main effect of Saccade Amplitudes target side revealed that latencies were shorter for targets Figure 4 (bottom) presents mean horizontal saccade on the high-probability side than for targets on the low- amplitudes. Note that because saccade target distances probability side [F(1, 9) = 5.510, p = .043, MSE = were physically equivalent in the current experiment, 849.633]. Data thus confirmed that high-contrast signals effects of the illusion on saccade programming are evident served as more salient cues for reflexive saccade here as a difference in movement amplitude for wings-in generation and that observers anticipated and prepared and wings-out targets. Statistical analysis was identical to for saccades toward the high-probability side of the that for latency data. As in Experiment 1, saccade display. Manipulations of signal contrast and of location amplitudes were reliably biased by illusory changes in probability, in other words, were effective. Additional target distance, with saccade amplitudes being larger for effects in the saccade latency data included an unexpected targets which appeared to be farther from the saccade reliable main effect of perceived target distance [F(1, 9) = launch point [F(1, 9) = 34.486, p < .001, MSE = .112]. 9.309, p = .014, MSE = 124.089], indicating that saccade This effect, however, was independent of both exogenous latencies were shorter for illusively far targets than for and endogenous manipulations; no interactions illusively near targets, a reliable interaction of signal approached significance (all Fs < 1). strength by target side [F(1, 9) = 5.103, p = .050, MSE = 125.117], indicating that the benefits of endogenous Discussion cuing were larger for saccades toward low-contrast targets, and a marginally reliable interaction of exogenous signal Experiment 2 was designed to answer two questions. First, is the effect of the M-L illusion on reflexive saccades

Downloaded from jov.arvojournals.org on 10/02/2021 McCarley, Kramer, & DiGirolamo 758

modulated by anticipatory or endogenous planning of eye and not modulated either by top-down saccade movements? Second, is the influence of the M-L illusion preparation or by bottom-up strength of a retinotopically on reflexive saccades modulated by the strength of the coded visual signal. Additional work, again using different visual go-signal? In both cases, the answer is no. Effects of forms of visual illusion, will be necessary to determine the illusion on reflexive saccade amplitudes were similar whether or not the higher level representation that biases whether movements were made toward a likely or an reflexive saccade programming is identical to the unlikely target location. Similarly, the illusion was of the representation that underlies conscious experience. same magnitude for low-contrast target signals as for high- Finally, it is useful to consider the apparent contrast signals. contradiction between the current results and those of Wong and Mack (1981), who found no effect of an induced motion illusion on reflexive saccade targeting. General Discussion One potential explanation for this discrepancy lies in the Earlier findings have suggested that reflexive and different temporal characteristics of the M-L and induced voluntary saccades might be differentially susceptible to motion illusions. The M-L illusion is extended in time, visual illusions. The present results indicate that, at least persisting for as long as the stimulus figure remains visible in the case of the M-L illusion, this is so. Voluntary (ignoring gradual declines of the illusion that may occur saccades show effects of the illusion similar in magnitude with very long periods of extended viewing, e.g., Festinger to those that are evident in subjective perceptual et al., 1968). In the current experiments, the illusive judgments. Reflexive saccades show effects of the illusion percept, therefore, began each trial well before that, while reliable, are far smaller. The results of presentation of the saccade go-signal, and remained Experiment 2 indicate that the influence of the illusion throughout saccade preparation and execution. In on reflexive saccades is not modulated by endogenous contrast, the induced motion illusion employed by Wong saccade preparation or by strength of the transient signal and Mack (1981) would have affected the perceived marking the target location. movement of the stepped target stimulus, but would not How should these results be explained? As noted in have existed either before that movement began or after it the Introduction, past research has indicated that had been completed. The illusion may therefore have voluntary eye movements are programmed and initiated existed either too briefly or at the wrong moment to bias purposively by cortical mechanisms (Henik et al., 1994), reflexive saccade programming. An alternative possibility, whereas reflexive saccades are programmed automatically as noted in the preceding paragraph, is that the higher within the superior colliculus in response to transient level spatial representation involved in reflexive saccade visual signals (Rafal et al., 1990). The present data suggest programming is not perfectly matched to a subjective that cortical planning of voluntary saccades occurs within frame of reference. The signals that bias reflexive saccade a representation whose frame of reference is — in at least programming away from a perfectly retinotopic spatial some aspects — similar or identical to that of conscious frame, for example, might arise from a cortical visual perception. This conclusion is compatible with a representation that precedes the locus of the induced common representation model, or alternatively, with a motion illusion. Once more, further research will be separate representation model in which the M-L illusion needed to test these possibilities. arises prior to the bifurcation of the perception/action representations. Further research, employing different Acknowledgments forms of visual illusion, will be necessary to distinguish these possibilities. Portions of this work were completed while J.S.M. In contrast, the current results support a weak was a postdoctoral Fellow at the Beckman Institute, separate representation model of reflexive saccade University of Illinois at Urbana-Champaign. Thanks to control. More specifically, findings suggest that reflexive Eric Vidoni and Russell Smith for assistance with data saccade programming occurs within a representation that collection. Commercial relationships: none. is a weighted average of a retinotopic map and a higher level spatial representation more closely matched to conscious perception. One possibility is that the Footnotes integration of spatial frames is effected by the 1This work is concerned primarily with visually guided convergence of feedforward retinal signals and feedback movements. It is widely acknowledged that memory- cortical signals on the superior colliculus; the colliculus is guided movements share a mental representation with known to receive input directly from the , and from conscious perception (e.g., Bridgeman, 1999; Gentilucci the striate and extrastriate cortex (for a review, see et al., 1996; Wong & Mack, 1981). Deubel, 1996). The results of Experiment 2 suggest that the weighting of the retinotopic and higher level information within the integrated representation is fixed,

Downloaded from jov.arvojournals.org on 10/02/2021 McCarley, Kramer, & DiGirolamo 759

Dorris, M. C., & Munoz, D. P. (1998). Saccadic References probability influences motor preparation signals and time to saccadic initiation. Journal of Neuroscience, 18, Aglioti, S., DeSouza, J. F. X., & Goodale, M. A. (1995). 7015-7026. [PubMed] Size-contrast illusions deceive the eye but not the hand. Current Biology, 5, 679-685. [PubMed] Dyde, R. T., & Milner, D. A. (2002). Two illusions of perceived orientation: One fools all of the people Binsted, G., & Elliott, D. (1999). The Müller-Lyer illusion some of the time; the other fools all of the people all as a perturbation to the saccadic system. Human of the time. Experimental Brain Research, 144, 518- Movement Science, 18.(1), 103-117. 527. [PubMed] Brenner, E., & Smeets, J. B. J. (1996). Size illusion Erkelens, C. J., & Hulleman, J. (1993). Selective influences how we lift but not how we grasp an adaptation of internally triggered saccades to visual object. Experimental Brain Research, 111, 473-476. targets. Experimental Brain Research, 93, 157-164. [PubMed] [PubMed] Bridgeman, B. (1999). Separate representations of visual Festinger, L., White, C. W., & Allyn, M. R. (1968). Eye space for perception and visually guided behavior. In movements and decrement in the Müller-Lyer G. Aschersleben, T. Bachmann & J. Müsseler (Eds.), illusion. Perception & Psychophysics, 3, 376-382. Cognitive Contributions to the perception of spatial and temporal events (pp. 3-13). Amsterdam: Elsevier. Findlay, J. M. (1982). Global visual processing for saccadic eye movements. Vision Research, 22, 1033- Bridgeman, B., Kirch, M., & Sperling, A. (1981). 1045. [PubMed] Segregation of cognitive and motor aspects of visual function using induced motion. Perception & Franz, V. H., Fahle, M., Bülthoff, H. H., & Gegenfurtner, Psychophysics, 29, 336-342. K. R. (2001). Effects of visual illusions on grasping. Journal of Experimental Psychology: Human Perception & Bridgeman, B., Peery, S., & Anand, S. (1997). Interaction Performance, 27, 1124-1144. [PubMed] of cognitive and sensorimotor maps of visual space. Perception and Psychophysics, 59, 456-469. [PubMed] Franz, V. H., Gegenfurtner, K. R., Bülthoff, H. H., & Fahle, M. (2000). Grasping visual illusions: No Carpenter, R. H., & Williams, M. L. (1995). Neural evidence for a dissociation between perception and computation of log likelihood in control of saccadic action. Psychological Science, 11. [PubMed] eye movements. Nature, 377, 59-62. [PubMed] Gentilucci, M., Chieffi, S., Daprati, E., Saetti, M. C., & Coëffé, C., & O'Regan, J. K. (1987). Reducing the Toni, I. (1996). Visual illusion and action. influence of non-target stimuli on saccade accuracy: Neuropsychologia, 34, 369-376. [PubMed] Predictability and latency effects. Vision Research, 27, 227-240. [PubMed] Goodale, M. A., & Milner, A. D. (1992). Separate visual pathways for perception and action. Trends in Coren, S., & Hoenig, P. (1972). Effect of non-target Neuroscience, 15, 20-25. [PubMed] stimuli on the length of voluntary saccades. Perceptual and Motor Skills, 34. 499-508. [PubMed] Godjin, R., & Theeuwes, J. (2002). Programming of endogenous and exogenous saccades: Evidence for a Delabarre, E. B. (1897). A method of recording eye competitive integration model. Journal of Experimental movements. American Journal of Psychology, 9, 572- Psychology: Human Perception and Performance, 28, 574. 1039-1054. [Pubmed] Deubel, H. (1995). Separate adaptive mechanisms for the He, P., & Kowler, E. (1989). The role of location control of reactive and volitional saccadic eye probability in the programming of saccades: movements. Vision Research, 35, 3529-3540. Implications for "center-of-gravity" tendencies. Vision [PubMed] Research, 9, 1165-1181. [PubMed] Deubel, H. (1996). Visual processing and cognitive factors He, P., & Kowler, E. (1991). Saccadic localization of in the generation of saccadic eye movements. In W. eccentric forms. Journal of the Optical Society of Prinz & B. Bridgeman (Eds.), Handbook of perception America, 8, 440-449. [PubMed] and action: Volume 1. Perception (Vol. 1, pp. 143-189). London: Academic Press. Henik, A., Rafal, R., & Rhodes, D. (1994). Endogenously generated and visually guided saccades after lesions DiGirolamo, G. J., McCarley, J. S., & Kramer, A. F. of the human . Journal of Cognitive (2001). Endogenously and exogenously driven eye Neuroscience, 6, 400-411. movements to illusory locations. Poster presented at the 42nd Annual Meeting of the Psychonomic Society, Orlando, FL.

Downloaded from jov.arvojournals.org on 10/02/2021 McCarley, Kramer, & DiGirolamo 760

Jackson, S. R., & Shaw, A. (2000). The Ponzo illusion Rafal, R., Machado, L., Ro, T., & Ingle, H. (2000). affects grip-force but not grip-aperture scaling during Looking forward to looking: Saccade preparation prehension movements. Journal of Experimental and control of the visual grasp reflex. In S. Monsell Psychology: Human Perception & Performance, 26, 418- & J. Driver (Eds.), Attention & performance XVIII (pp. 423. [PubMed] 155-174). Cambridge, MA: MIT Press. Klein, R., Kingstone, A., & Pontefract, A. (1992). Rafal, R., Smith, J., Krantz, J., Cohen, A., & Brennan, C. Orienting of visual attention. In K. Rayner (Ed.), Eye W. (1990). Extrageniculate vision in hemianopic movements and visual cognition: Scene perception and humans: Saccade inhibition by signals in the blind reading (pp. 46-65). New York: Springer-Verlag. field. Science, 250, 118-121. [PubMed] Klein, R., & Shore, D. I. (2000). Relations among modes Stratton, G. M. (1906). Symmetry, linear illusions, and of visual orienting. In S. Monsell & J. Driver (Eds.), the movements of the eye. Psychological Review, 13, Attention & performance XVIII (pp. 195-208). 82-96. Cambridge, MA: MIT Press. Trappenberg, T. P., Dorris, M. C., Munoz, D. P., & Kopecz, K. (1995). Saccadic reaction times in gap/overlap Klein, R. M. (2001). A model of saccade inititiation paradigms: A model based on integration of based on the competitive integration of exogenous intentional and visual information on neural, and endogenous signals in the superior colliculus. dynamic fields. Vision Research, 35, 2911-2925. Journal of Cognitive Neuroscience, 13, 256-271. [PubMed] [PubMed] López-Moliner, J., Smeets, J. B. J., & Brenner, E. (2003). Vishton, P. M., Rea, J. G., Cutting, J. E., & Nunez, L. N. Comparing the sensitivity of manual pursuit and (1999). Comparing effects of the horizontal-vertical perceptual judgments to pictorial depth effects. illusion on grip scaling and judgment: Relative vs. Psychological Science, 14, 232-236. [PubMed] absolute, not perception vs. action. Journal of Machado, L., & Rafal, R. (2000a). Control of eye Experimental Psychology: Human Perception & movement reflexes. Experimental Brain Research, 135, Performance, 25, 1659-1672. [PubMed] 73-80. [PubMed] Wong, E., & Mack, A. (1981). Saccadic programming and Machado, L., & Rafal, R. (2000b). Strategic control over perceived location. Acta Psychologica, 48, 121-131. saccadic eye movements: Studies of the fixation [PubMed] offset effect. Perception and Psychophysics, 62, 1236- Wraga, M., Creem, S. H., & Proffitt, D. R. (2000). 1242. [PubMed] Perception-action dissociations of a walkable Müller- Ottes, F. P., van Gisbergen, J. A. M., & Eggermont, J. J. Lyer configuration. Psychological Science, 11, 239-243. (1985). Latency dependence of colour-based target [PubMed] versus non-target discrimination by the saccadic Yarbus, A. L. (1967). Eye movements and vision (B. Haigh, system. Vision Research, 24, 826-849. [PubMed] Trans.). New York: Plenum Press.

Downloaded from jov.arvojournals.org on 10/02/2021