- Email: [email protected]
Delayed pointing movements to masked Müller–Lyer figures are affected by target size but not the illusion
Neuropsychologia 49 (2011) 1903–1909
Contents lists available at ScienceDirect
Neuropsychologia journal homepage: www.elsevier.com/locate/neuropsychologia
Delayed pointing movements to masked Müller–Lyer figures are affected by target size but not the illusion Matthieu de Wit a,∗ , John van der Kamp a,b , Rich S.W. Masters a a b
Institute of Human Performance, University of Hong Kong, 111-113 Pokfulam Road, Hong Kong SAR, China Research Institute MOVE, Faculty of Human Movement Sciences, VU University Amsterdam, The Netherlands
a r t i c l e
i n f o
Article history: Received 9 April 2010 Received in revised form 4 March 2011 Accepted 14 March 2011 Available online 21 March 2011 Keywords: Vision for action Vision for perception Egocentric information Allocentric information Müller–Lyer illusion Visual masking Delayed action
a b s t r a c t There is ongoing debate with respect to interpretation of the finding that, in contrast to perceptual size judgments, actions are relatively unaffected by the Müller–Lyer illusion. In normal unrestricted viewing situations observers cannot perform an action directed at an object without simultaneously perceiving the object – this makes it difficult to unequivocally establish whether observed effects are a function of vision for perception, vision for action, a combination of both, or of a single all-purpose visual system. However, there is evidence that observers are capable of performing actions towards objects of which they are not consciously aware, implying that two distinct visual thresholds may exist; one accompanying vision for action and one accompanying vision for perception. To investigate this possibility we created a situation in which visual information was presented below the perception threshold, but above the purported action threshold, allowing examination of action responses independent of contributions from vision for perception. Following a perceptual categorization task, participants performed delayed pointing movements towards briefly exposed masked Müller–Lyer targets of different sizes. When the targets were presented below the perception threshold, participants were unable to discriminate between them, yet their delayed pointing movements were affected by target size (but not the illusion). The results imply that vision for action is functional even after a delay and/or that the pickup of egocentric information is associated with a lower visual threshold than the pickup of allocentric information. © 2011 Elsevier Ltd. All rights reserved.
1. Introduction Goodale and Milner (1992) and Milner and Goodale (1995, 2008) proposed that the perception of objects and the visual control of actions directed at those objects are mediated by two functionally and anatomically distinct visual systems (i.e., vision for perception and vision for action). In a paradigm that has been used extensively to examine this proposal, participants perform perception and action responses upon targets embedded in geometrical illusions such as the Müller–Lyer illusion. Meta-analyses indicate that unlike perceptual judgments, actions performed in unrestricted viewing situations are relatively unaffected by (but not immune to) the Müller–Lyer illusion (e.g., Bruno, Bernardis, & Gentilucci, 2008; Bruno & Franz, 2009). Milner and Goodale (1995, 2008; Ganel, Tanzer, & Goodale, 2008; Goodale & Haffenden, 1998) explain these findings by arguing that vision for action and vision for perception exploit distinct types of information to perform their tasks; while vision for action uses egocentric (i.e., body-centered) information, vision for perception relies mainly on allocentric (i.e.,
∗ Corresponding author. Tel.: +852 2589 0585; fax: +852 2855 1712. E-mail address: [email protected] (M. de Wit). 0028-3932/$ – see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.neuropsychologia.2011.03.017
world-centered) information. Gentilucci, Chieffi, Daprati, Saetti, and Toni (1996; see also Hu & Goodale, 2000; Westwood & Goodale, 2003; Westwood, Heath, & Roy, 2000) examined the kinematics of delayed pointing movements directed at occluded Müller–Lyer figures. Restriction of vision by occluding the goal target led to an increase in the effect of the illusion to a level comparable to effects normally associated with perceptual judgments. This use of (retained) allocentric information has been taken to imply that delayed actions are mediated by vision for perception and that vision for action only engages in guiding actions when they are performed online (i.e., in real time) and in full vision (Goodale, Westwood, & Milner, 2004). Ongoing controversy exists with respect to the interpretation of the differential effect of geometrical illusions on perception and action. Several authors have proposed that the relative immunity of actions to geometrical illusions when compared to perceptual measures can be explained by differences in task characteristics between commonly used action and perception tasks, and does not implicate different visual processes (i.e., vision for action and vision for perception) underlying the execution of those tasks. Smeets, Brenner, de Grave, and Cuijpers (2002; see Schenk, 2006 for a related argument) argued that the spatial attributes that are used to perform a task determine susceptibility to an illu-
1904
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909
sion. de Grave, Brenner, and Smeets (2004), for example, found that pointing movements along the shaft of a Müller–Lyer figure (emphasizing the use of size information, which is affected by the illusion) were influenced by the illusion whereas pointing movements perpendicular to the shaft (emphasizing endpoint position information, which is unaffected by the illusion) were not. Franz, Hesse, and Kollath (2009), varied the amount of visual feedback available to participants during the execution of grasps directed at Müller–Lyer targets. Removing feedback at movement onset, at 1/3 of the transport phase and at 2/3 of the transport phase led to a gradual reduction of the illusion effect that was directly related to the amount of available visual feedback. Based on these results, Franz et al. (2009) argued that the increased effect of illusion after a delay is caused by the availability of visual feedback leading to online corrections of the movement (and not by a shift in control from vision for action to vision for perception). In unrestricted viewing situations, observers cannot perform an action directed at an object without simultaneously perceiving the object (Milner & Goodale, 2008; see also Enns & Liu, 2009; van Doorn, van der Kamp, de Wit, & Savelsbergh, 2009). This makes it difficult to unequivocally determine whether observed illusion effects are a function of vision for perception, vision for action, a combination of both, or of a single all-purpose visual system. However, a possible resolution to this problem might exist. There is evidence that vision for action has quicker access to visual information than vision for perception. Pisella, Arzi, and Rossetti (1998; see also Veerman, Brenner, & Smeets, 2008) asked participants to perform reach movements towards stimuli that could be perturbed in either location or color during the ongoing movement. In case of a perturbation, participants were required to stop their movement. Results showed that stop-responses to perturbations of color, an object property that would arguably be picked up by vision for perception, were initiated about 80 ms later than stop-responses to location changes, which are arguably guided by vision for action (Rossetti, Pisella, & Pelisson, 2000). Heath, Maraj, Godbolt, and Binsted (2008; Binsted, Brownell, Vorontsova, Heath, & Saucier, 2007; Heath, Neely, Yakimishyn, & Binsted, 2008; and see Cressman, Franks, Enns, & Chua, 2007 for a related experiment) asked participants to perform pointing movements towards masked briefly exposed (i.e., 13 ms) targets of different sizes. Although participants were unable to perceptually discriminate between targets at above chance levels, Fitts’ law (1954) was preserved in that the movement time for pointing movements towards smaller targets was longer. Notably, these movements were not performed online but after a delay of up to 2 s. These findings suggest that (1) vision for action may remain functional at lower minimum stimulus exposure times than vision for perception and (2) vision for action may be capable of guiding actions performed after a delay. The current experiment was designed to exploit this potential difference in visual threshold between vision for action and vision for perception by creating a situation in which participants performed an action task and a perception task in response to Müller–Lyer targets that were presented below the vision for perception threshold but above the vision for action threshold. This would allow for an examination of action responses independent of contributions from vision for perception. Participants were exposed to masked targets for brief (12 ms) or long (1500 ms) durations. Targets consisted of ‘wings out’ and ‘wings in’ Müller–Lyer figures and neutral (i.e., without wings) figures of three different lengths. Based on the finding by Heath and colleagues (i.e., Binsted et al., 2007; Heath, Maraj, et al., 2008; Heath, Neely, et al., 2008) that Fitts’ law was preserved for action responses to perceptually indiscriminable masked briefly exposed (i.e., 13 ms) targets, it was hypothesized that the briefly exposed targets would exceed the vision for action threshold but not the vision for perception threshold. To verify whether targets were
indeed presented below the vision for perception threshold, participants were first required to indicate the size of the briefly exposed targets in a perception task. In four subsequent action conditions, participants made pointing movements along the shaft of the targets that were presented for brief (i.e., 12 ms) or long durations (i.e., 1500 ms) either as soon as possible after target stimulus offset (i.e., reaction time (RT) delay) or after a delay of 2 s. We expected pointing movements to be scaled to target length regardless of whether they were briefly presented (i.e., below the vision for perception threshold) or not – as long as target stimulus duration exceeded the vision for action threshold. We also had specific expectations with respect to the effect of the illusion on pointing movements directed at the briefly exposed Müller–Lyer targets. Based on the evidence that vision for perception relies mainly on the use of allocentric information (e.g., as implicated by the relatively large effects of the Müller–Lyer illusion on perceptual measures, see Bruno et al., 2008; Bruno & Franz, 2009; Ganel et al., 2008) together with the assumption that the briefly exposed targets were not expected to exceed the vision for perception threshold, we only expected an effect of illusion on the delayed pointing movements for the targets exposed for long durations, not for targets exposed for brief durations. 2. Methods 2.1. Participants Seventeen right-handed participants (8 females) aged 23–60 years (33 ± 10) with normal or corrected-to-normal vision participated in the experiment. They were naïve with regard to the purpose of the experiment and were treated in accordance with the ethical guidelines of the local Institution. 2.2. Materials Stimuli were presented on a 19 in. CRT-monitor (Philips Brilliance 109P4) with a refresh rate of 85 Hz and a resolution of 1024 × 768 pixels using E-Prime 2.0 presentation software (Psychology Software Tools, Pittsburg, PA). An Optotrak 3020 motion analysis system (Northern Digital, Waterloo, Ontario) was used to measure the extent of the pointing movements by recording the position of an infrared light emitting diode placed on the tip of the right index finger with a frequency of 200 Hz. Stimuli consisted of three different lengths (short: 11.5 cm, medium: 14.5 cm, long: 17.5 cm) of wings out and wings in Müller–Lyer figures and neutral figures (i.e., without wings). The lines that made up the figures (shaft and wings) were 5 mm wide. The wings had a length of 3 cm and an angle of 45◦ relative to the shaft. Stimuli were presented randomly in one of six locations (top-left, center-left, bottom-left, top-right, center-right, bottom-right) but appeared equally often on the left and right side and top, center and bottom of the computer screen (see Fig. 1). 2.3. Procedure and design Participants performed an action task and a perception task. In the action task, participants were instructed to place the tip of their right index finger on a fixation dot that was positioned at one end of the to-be-presented target shaft and to fixate their gaze on the tip of their finger (see Fig. 1). If the dot appeared at the left side of the screen, a movement from the left end of the target shaft to the right end of the target shaft was required, and vice versa. The dot was present for 3 s after which the target stimulus was presented. Stimulus exposure duration was either brief (12 ms) or long (1500 ms) and pointing movements had to be made either as quickly as possible after stimulus offset (RT delay) or after a delay of 2000 ms following stimulus offset, in both cases indicated by an auditory start signal (i.e., for RT delays the start signal sounded at stimulus offset and for 2000 ms delays the start signal sounded 2000 ms after stimulus offset). Stimulus presentation was always followed by a mask that was presented for 200 ms and consisted of an array of scrambled target stimuli (see Fig. 1). Participants were required to maintain gaze fixation upon the tip of their finger until they heard the start signal. For the pointing movement, the instruction was to “point as fast and accurately as possible to the other end of the horizontal shaft line when you hear the beep”. The combination of the factors stimulus exposure duration and movement delay led to a total of four action conditions (i.e., 12 ms stimulus exposure × RT/2000 ms movement delay and 1500 ms stimulus exposure × RT/2000 ms movement delay) which were performed in separate blocks by each participant in counterbalanced order. To assess whether participants were able to categorize the target stimuli at 12 ms stimulus exposure duration, participants performed a perception task both before and after the four action blocks. The conditions in the perception task were identical to those in the action task but instead of making a pointing movement participants
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909
1905
(i.e., whether pointing distance is scaled to actual target length) and of stimulus type (i.e., whether pointing distance is affected by the illusion), by submitting the raw data to separate hierarchical regression analyses for each condition. The contribution of the following three predictor variables was tested: (1) actual target length (treated as a numerical variable with 3 levels; short, medium, long), (2) stimulus type (a dummy coded categorical variable with 3 levels; wings out, wings in, no wings) and (3) an interaction term for actual target length and stimulus type (the product of the numerical and the dummy coded variable). We first fitted a one-predictor model with actual target length as the predictor to the data of each condition. A significant fit of this model showed that pointing distance was scaled to actual target length, so we proceeded to test whether, next to actual target length, stimulus type was also a significant predictor of pointing distance (indicating an effect of illusion). To this end, we fitted a two-predictor model with actual target length and stimulus type as predictors to the raw data for each condition. A significant increase in the proportion of explained variance from model one (one predictor) to model two (two predictors) would indicate that next to actual target length, stimulus type also affected pointing distance. To allow for comparison of effects of illusion between conditions, we also computed the percent corrected illusion effect (see Bruno et al., 2008) for each condition using Eq. (2) where ‘expand’ and ‘compress’ stand for pointing distance directed to wings out and wings in stimuli and ‘baseline’ stands for pointing distance directed to the no wings stimuli. Mean effects were subsequently submitted to paired t-tests to examine possible differences between conditions.
%= Fig. 1. Schematic representation of stimulus presentation. In this example the target stimulus is presented at the top-left position, consists of a medium length wings out figure and requires a movement from left to right.
were now required to indicate whether they had seen a short, medium or long target stimulus by pressing a corresponding key on the keyboard. Participants were encouraged to respond fairly quickly (i.e., within 1–3 s after stimulus offset). In each action and perception block, each stimulus type (wings out/wings in/no wings) in combination with each length (short/medium/long) was randomly presented two times leading to a total of 18 experimental trials (i.e., 3 stimulus types × 3 lengths × 2 repetitions) per block. To maintain active engagement in the experiment despite the very high task difficulty in the 12 ms stimulus duration action blocks and the perception blocks, experimental trials were randomly interspersed with trials in which the stimulus duration was 36 ms instead of 12 ms. In these trials, each stimulus type in combination with the short and long target length was presented once (total 6 trials). Similarly, in the long stimulus duration action blocks experimental trials were interspersed with trials in which the stimulus duration was 1524 ms instead of 1500 ms. Each block thus consisted of 18 + 6 = 24 trials of which only the 18 experimental trials were analyzed. To familiarize participants with the different stimulus lengths and types and with the task requirements, the experimental blocks were preceded by a practice block of 18 trials in which all stimuli were presented twice and the stimulus duration was 3 s. In this block, the experimenter verbally indicated the length of the presented target in each trial. Participants performed a total of 162 trials. There was a 5 min break between the third and the fourth block and the experiment was completed in approximately 1 h. 2.4. Data analysis For the perception task, we first computed the percentages of correct responses. To assess whether group performance in the perception task was at chance levels, the percentages of correct responses were submitted to a one-sample t-test and tested against a test value of 33.3% (i.e., chance). Furthermore, to examine whether individual participants were able to discriminate between stimulus lengths at above chance levels, for each participant the number of correct responses were entered in a Z-score proportions test (see Eq. (1), where c is the number of correct responses). The critical Z-value for ˛ = .05 lies between 1.64 and 1.65. A Z-score greater than 1.65 was therefore interpreted as an indication that the participant was performing above chance levels. To examine whether errors made by participants were influenced by the illusion-inducing wings that were present in a subset of the stimuli, we recorded the direction of the errors made in response to the Müller–Lyer stimuli. For both wings in and wings out stimuli the number of errors that corresponded with the direction of the illusion (i.e., the stimulus was categorized respectively as smaller than the presented stimulus and as larger than the presented stimulus) and the number of errors that did not correspond with the direction of the illusion was determined. To test for an effect of illusion on the errors, the number of corresponding and noncorresponding errors was submitted to a paired-samples t-test. Z=
c/(36 − .33)
(1)
expand − compress baseline
× 100
(2)
3. Results 3.1. Perception task First, the percentages of correct responses in the perception block that was completed before the action blocks and in the perception block that was completed after the action blocks were compared. A lack of difference between the two blocks (t(16) = 1.23, p = .236) indicates that no perceptual learning occurred over the course of the experiment (39.2 and 43.5% correct for the first and second perception block, respectively). Group categorization performance in the two perception blocks differed significantly from chance level performance (t(16) = 2.81, p < 0.05). We therefore entered the score of each individual into a proportions test (Z-test) to identify participants who were able to categorize the stimuli above chance levels at 12 ms stimulus exposure time. Because an increase in the number of trials leads to more reliable estimates of proportions and there was no difference in performance between the first and the second perception block, the total amount of correct responses of the 36 trials from the two perception blocks were entered in the proportions test. Out of the seventeen participants, six (3 females) performed above chance levels (see Fig. 2). For these participants the brief stimulus duration (12 ms) exceeded the vision for perception threshold. As we were interested in pointing movements towards stimuli presented below the vision for perception threshold, we excluded these participants from the pointing movement analysis. To assess whether the illusion-inducing wings that surrounded a subset of the stimuli influenced the categorization errors of the eleven participants who performed at chance levels in the perception task, we compared the number of errors that corresponded with the direction of the illusion with the number of errors that did not correspond with the direction of the illusion. No significant differences existed (t(10) = .1.44, p = .181; 8.2 corresponding errors and 6.9 noncorresponding errors, on average), indicating that the presence of wings did not have an effect on categorization performance in the perception task. 3.2. Action task
(.33 − .66)/36
Pointing distance in the action task was determined by subtracting the difference (in mm) between the start position and the end position of the pointing movement for each trial. Following Bruno and Bernardis’ (2003) approach, we examined the extent to which pointing movements varied as a function of target length
The pointing movement analyses were performed on the eleven participants who performed at chance levels in the perception task. We first conducted a boxplot analysis and identified 12 outliers (corresponding to 1.5% of the total number of data points), which
1906
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909
Fig. 2. Percentage of correct responses in the perception task for each participant. The black dashed line indicates the criterion for chance level performance based on the proportions test. The grey dashed line indicates the 33% level. Participants 7, 9, 10, 12, 14 and 17 were excluded from the pointing movement analysis.
Fig. 3. The two-predictor (actual target length, stimulus type) model fitted to the pointing distances of the 1500 ms stimulus exposure (SE) − RT delay (left panel) and 2000 ms delay (right panel) conditions. Circles correspond to individual data points. Wings out and wings in data points are slightly displaced to the left and right to better reveal the pattern in the data.
were subsequently removed from the data set. Fig. 3 shows the pointing distances to the different stimuli in the 1500 ms stimulus exposure conditions. Individuals are not identified by different symbols as there was a high consistency in their pattern of behavior (i.e., if an individual tended to undershoot or overshoot at one target length he or she would do so at other target lengths as well). As one would expect, pointing distance in these conditions was significantly scaled to actual target length. The fit of the one-predictor (actual target length) model was highly significant for both the RT delay condition, F(1, 190) = 416.15, p < 0.0001, R2 = 0.687 (slope = 0.917), and the 2000 ms delay condition, F(1, 190) = 403.19, p < 0.0001, R2 = 0.680 (slope = 0.851). To establish whether, next to actual target length, stimulus type was also a significant predictor of pointing distance (indicating an effect of illusion), we subsequently fitted a two-predictor (actual target length, stimulus type) model to the data of the 1500 ms stimu-
lus exposure conditions. Adding the second predictor to the model led to a significant change in the proportion of explained variance for both conditions, indicating a substantial improvement of the model fit compared to the one-predictor model.1 For the RT delay condition, R2 increased from 0.687 to 0.730 (F(2, 188) = 15.29, p < 0.0001), while for the 2000 ms delay condition R2 increased from 0.680 to 0.725 (F(2, 188) = 15.66, p < 0.0001). This indicates the presence of an effect of illusion in these conditions, which
1 We also fitted a three-predictor model that additionally included an interaction term between actual stimulus length and stimulus type to the data of each condition. This model was used to examine whether the effect of illusion, if any, was modulated by actual target size (i.e., different steepness of slopes for wings in and wings out stimuli). However, because the interaction term did not contribute significantly to the model in any of the conditions, we do not report the statistics of this model.
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909
is further supported by the fact that the t-test for the coefficient reflecting the extent to which pointing distance differed between wings out and wings in stimuli was significant in both conditions; RT delay, t(10) = 5.29, p < 0.0001 (slope wings out = 0.968, wings in = 0.869), 2000 ms delay, t(10) = 5.58, p < 0.0001 (slope wings out = 0.930, wings in = 0.846). A t-test comparing the mean percent corrected illusion effects of the two 1500 ms stimulus exposure conditions revealed no differences between these conditions (RT delay: 10.47%, SD = 7.41, 2000 ms delay: 10.79%, SD = 7.74). Analysis of the pointing distances in the 12 ms stimulus exposure conditions revealed a rather different pattern (see Fig. 4). Although the one-predictor model explained only a small amount of the variance in these conditions, this model nevertheless showed a significant fit to the data of both the RT delay condition, F(1, 186) = 5.91, p < 0.05, R2 = 0.031 (slope = 0.209) and the 2000 ms delay condition, F(1, 192) = 10.72, p < 0.01, R2 = 0.053 (slope = 0.264), indicating that pointing movements were significantly scaled to actual target length. However, for these conditions, adding the second predictor to the model did not lead to a significant change in the proportion of explained variance, indicating that at 12 ms stimulus exposure duration, the presence of wings did not have an effect on pointing distance. In fact, for the RT delay condition, adding the predictor stimulus type rendered the model’s fit nonsignificant (Model 1, p = 0.012, Model 2, p = 0.121) while for the 2000 ms delay condition, adding the second predictor merely reduced the model’s significance (Model 1, p = 0.001, Model 2, p = 0.008). The absence of an effect of illusion at 12 ms stimulus exposure is further supported by the fact that the t-test for the coefficient reflecting the extent to which pointing distance differed between wings out and wings in stimuli was not significant in both conditions. Slopes in the RT delay condition measured 0.088 and 0.305 for wings out and wings in stimuli respectively. In the 2000 ms delay condition they measured 0.334 and 0.428 for wings out and wings in stimuli. The mean percent corrected illusion effect measured 1.50% (SD = 15.82) in the RT delay condition and 2.94% (SD = 13.43) in the 2000 ms delay condition2 . These effects were not significantly different from each other or from 0%. 4. Discussion Under the assumption that vision for perception and vision for action remain functional at different minimum stimulus exposure times, we attempted to present target stimuli below the vision for perception threshold but above the vision for action threshold. In a perception and an action task, participants were exposed to masked Müller–Lyer figures (wings out, wings in, no wings) of different lengths for brief (12 ms) or long (1500 ms) durations. Analysis of performance levels in the perception task showed that 6 out of the 17 participants were able to discriminate between stimulus lengths at above chance levels, implying that for these participants the briefly exposed stimuli exceeded the vision for perception threshold. Strikingly, analysis of pointing distances of the remaining 11 participants, for whom the briefly presented stimuli did not exceed
2 Bruno and Franz (2009) have argued that a correction formula incorporating the slope of the function that relates movement distance to target length in the denominator of the fraction might be appropriate when there are differences in scaling between conditions, as is the case for the present 12 and 1500 ms stimulus exposure conditions. Doing so returns corrected percent measures for the 12 ms stimulus exposure time conditions of 9.2 % (SD = 74.1) and 12.8 % (SD = 70.2) for the RT - and 2000 ms Delay conditions, respectively. However, note the high standard deviations. Clearly, with the relatively high variability in pointing distance at 12 ms stimulus exposure and the lack of a consistent pattern in the relationship between pointing distance and wing orientation, this correction approach produces somewhat misleading percent measures. Accordingly, t-tests show that, despite their apparently large size, the corrected percent measures do not differ from 0 % (p = 0.310 and 0.194, respectively).
1907
the vision for perception threshold, showed that their actions were nevertheless scaled to actual stimulus length. Thus, while participants were exposed to the same set of stimuli, the perception task was accompanied by a higher visual threshold than the action task. Several studies (i.e., Franz et al., 2009; Gentilucci et al., 1996; Hu & Goodale, 2000; Westwood & Goodale, 2003; Westwood et al., 2000) have reported that the introduction of a delay causes a profound increase in the effect of illusion on actions; this is typically explained as an effect of vision for perception supplanting vision for action (but see Franz et al., 2009). The current study replicates these findings. Pointing movements after a delay were significantly affected by the illusion for targets presented above the vision for perception threshold. This effect was the same for movements performed after a short (i.e., start signal at stimulus offset, RT delay) or a long delay (i.e., start signal 2 s after stimulus offset). However, when targets were presented below the vision for perception threshold, the illusion did not affect pointing distance, both when movements were performed at RT delay and after a delay of 2 s. Together with our finding that pointing movements were nevertheless scaled to physical target length, this may suggest that vision for action is not fully disrupted when movements are performed after a delay. One might argue that because a subset of the stimuli was surrounded by illusion-inducing wings, the possibility exists that an effect of illusion rather than an inability to perceive the stimuli may have caused the participants to perform at chance levels in the perception task. However, because on average the size of perceptual illusion effects is between 10 and 20% of physical target length (see Bruno et al., 2008; Bruno & Franz, 2009) participants should have been able to discriminate between the targets in our experiment regardless of whether wings were present or not. Moreover, under this assumption the perceptual errors made by participants should have been made in accordance with the direction of the illusion. That is, the size of wings out stimuli should have been overestimated more often than the size of wings in stimuli, and the size of wings in stimuli should have been underestimated more often than the size of wings out stimuli, which was not the case. In addition, participants frequently reported no experience of seeing the stimuli. Notably, the possibility that some participants were in fact able to perceptually discriminate between briefly presented stimuli does potentially apply to the studies by Heath and colleagues (e.g. Binsted et al., 2007; Heath, Maraj, et al., 2008; Heath, Neely, et al., 2008; see also Cressman et al., 2007)3 who found that Fitts’ law (1954) was preserved for delayed pointing movements towards masked stimuli. In sum, our results provide solid evidence for the existence of distinct task-dependent visual thresholds for the pickup of visual information. Our results uncover a problem with the common interpretation of the increased effects of illusion on delayed action in which it is argued that delayed actions are necessarily guided by vision for perception (as implied by their dependence on allocentric information; see Franz et al., 2009; Goodale et al., 2004; Hu & Goodale, 2000; Westwood & Goodale, 2003 for a related argument). To our knowledge, this interpretation has never been tested by preventing vision for perception from contributing to actions performed after a delay. The claim is therefore an indirect one and, as evidenced by our finding that delayed pointing movements are scaled to the
3 In stead of examining the discrimination performance of individual participants with proportions tests, these authors examined discrimination performance at the group level with (one-sample) t statistics or analysis of variance. It is possible that while these tests did not show an overall effect, some participants may in fact have been able to discriminate between the masked stimuli above chance levels. In the current experiment, we removed such participants from the pointing movement analysis.
1908
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909
Fig. 4. The one-predictor (actual target length) model fitted to the pointing distances of the 12 ms stimulus exposure (SE) − RT delay (left panel) and 2000 ms delay (right panel) conditions. Plotting conventions are the same as in Fig. 3.
veridical size of Müller–Lyer targets presented below the vision for perception threshold, may have to be revisited. Evidence against the notion that vision for action is capable of guiding delayed actions that is not subject to the above criticism comes from a study of patient D.F., who has damage to the ventral stream (i.e., the anatomical substrate of vision for perception) but has an intact dorsal stream (i.e., the anatomical substrate of vision for action). Although she is well able to grasp visible objects, her ability to accurately grasp objects after a delay is completely disrupted (Goodale, Jakobson, & Keillor, 1994). However, in a recent study D.F. displayed preserved performance in a perception task that depended on the use of egocentric information and impaired performance in an action task that depended on the use of allocentric information, implying that her deficit may be task (i.e., perception or action) independent and instead related to an inability to exploit allocentric information (Schenk, 2006). A similar, information based (as opposed to task based), distinction may account for the results obtained in the current experiment. Closer inspection of our perception task shows that it could only be performed by relying on allocentric information. Categorization necessarily required a comparison of the length of the presented stimulus with the recalled length of previously presented stimuli, thus leaving participants no choice but to use allocentric information. Both the chance level categorization performance and the absence of an effect of wing orientation on the categorization errors imply that participants were unable to exploit allocentric information in the perception task. The action task, on the other hand, did not require memory based length comparisons. In this task participants made an absolute movement along the stimulus shaft, allowing them to potentially base their movement both on egocentric information and on the allocentric information that was present within the stimuli (i.e., the wings). The fact that at 12 ms stimulus exposure pointing distance was scaled to actual target length but not affected by the presence of wings implies that in this task the allocentric information was presented below threshold but the egocentric information was not. It is thus unclear whether the presence of distinct visual thresholds that is evidenced by our findings indicates that vision for action is associated with a lower visual threshold than vision for perception or instead whether the pickup of egocentric informa-
tion is associated with a lower visual threshold than the pickup of allocentric information, possibly irrespective of whether a task is perceptual or motoric in nature. Under the latter assumption, vision for perception may have informed the delayed pointing movements in our experiment, but was depending on the use of egocentric information. Vision for perception is commonly assumed to rely mainly on the use of allocentric information (Ganel et al., 2008; Milner & Goodale, 1995, 2008, but see Briscoe, 2009). However, there is evidence indicating that vision for perception has the potential to exploit egocentric information as well. Predebon (2006), for example, showed that the effect of the Müller–Lyer illusion on perceptual length estimations disappears when participants are instructed to ignore the wings (i.e., the allocentric information). van Doorn et al. (2009) examined the gaze behavior of participants engaged in a perceptual length estimation task and a grasping task directed at Müller–Lyer stimuli. Results showed a significant negative correlation between the amount of time that was spent looking at the target shaft (containing the pertinent sources of egocentric information for grasping) and the size of the illusion effect on hand aperture in both the grasping task and the perceptual length estimation task, showing that in either task participants had the potential to rely on egocentric information. As a third and final alternative, our findings are in accordance with the possibility that there exists both a distinction in threshold between vision for perception and vision for action and a distinction in threshold for the pickup of allocentric information and egocentric information. Future experiments are needed to distinguish between the described possibilities and to further examine the interrelationships between vision for perception and vision for action and the pickup of allocentric and egocentric information. References Binsted, G., Brownell, K., Vorontsova, Z., Heath, M., & Saucier, D. (2007). Visuomotor system uses target features unavailable to conscious awareness. Proceedings of the National Academy of Sciences of the United States of America, 104(31), 12669–12672. Briscoe, R. (2009). Egocentric spatial representation in action and perception. Philosophy and Phenomenological Research, 79(2), 423–460. Bruno, N., & Bernardis, P. (2003). When does action resist visual illusions? Effector position modulates illusory influences on motor responses. Experimental Brain Research, 151, 225–237.
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909 Bruno, N., Bernardis, P., & Gentilucci, M. (2008). Visually guided pointing, the Müller–Lyer illusion, and the functional interpretation of the dorsal–ventral split: Conclusions from 33 independent studies. Neuroscience and Biobehavioral Reviews, 32(3), 423–437. Bruno, N., & Franz, V. H. (2009). When is grasping affected by the Müller–Lyer illusion? A quantitative review. Neuropsychologia, 47(6), 1421–1433. Cressman, E. K., Franks, I. M., Enns, J. T., & Chua, R. (2007). On-line control of pointing is modified by unseen visual shapes. Consciousness and Cognition, 16, 265–275. de Grave, D. D. J., Brenner, E., & Smeets, J. B. J. (2004). Illusions as a tool to study the coding of pointing movements. Experimental Brain Research, 155, 56–62. Enns, J. T., & Liu, G. (2009). Attentional limits and freedom in visually guided action. Progress in Brain Research, 176, 215–226. Fitts, P. M. (1954). The information capacity of the human motor system in controlling the amplitude of movement. Journal of Experimental Psychology, 47(6), 381–391. Franz, V. H., Hesse, C., & Kollath, S. (2009). Visual illusions, delayed grasping, and memory: No shift from dorsal to ventral control. Neuropsychologia, 47(6), 1518–1531. Ganel, T., Tanzer, M., & Goodale, M. A. (2008). A double dissociation between action and perception in the context of visual illusions: Opposite effects of real and illusory size. Psychological Science, 19, 221–225. Gentilucci, M., Chieffi, S., Daprati, E., Saetti, M. C., & Toni, I. (1996). Visual illusion and action. Neuropsychologia, 34(5), 369–376. Goodale, M. A., & Haffenden, A. (1998). Frames of reference for perception and action in the human visual system. Neuroscience and Biobehavioral Reviews, 22(2), 161–172. Goodale, M. A., Jakobson, L. S., & Keillor, J. M. (1994). Differences in the visual control of pantomimed and natural grasping movements. Neuropsychologica, 32, 1159–1178. Goodale, M. A., & Milner, A. D. (1992). Separate visual pathways for perception and action. Trends in Neurosciences, 15(1), 20–25. doi:10.1016/ 0166-2236(92)90344-8 Goodale, M. A., Westwood, D. A., & Milner, A. D. (2004). Two distinct modes of control for object-directed action. Progress in Brain Research, 144, 131–144.
1909
Heath, M., Maraj, A., Godbolt, B., & Binsted, G. (2008). Action without awareness: Reaching to an object you do not remember seeing. PLoS ONE, 3(10). Heath, M., Neely, K. A., Yakimishyn, J., & Binsted, G. (2008). Visuomotor memory is independent of conscious awareness of target features. Experimental Brain Research, 188(4), 517–527. Hu, Y., & Goodale, M. A. (2000). Grasping after a delay shifts size-scaling from absolute to relative metrics. Journal of Cognitive Neuroscience, 12(5), 856–868. Milner, A. D., & Goodale, M. A. (1995). The visual brain in action. Oxford: Oxford University Press. Milner, A. D., & Goodale, M. A. (2008). Two-visual systems re-viewed. Neuropsychologia, 46(3), 774–785. Pisella, L., Arzi, M., & Rossetti, Y. (1998). The timing of color and location processing in the motor context. Experimental Brain Research, 121(3), 270–276. Predebon, J. (2006). Decrement of the Müller–Lyer and Poggendorff illusions: The effects of inspection and practice. Psychological Research, 70, 384–394. Rossetti, Y., Pisella, L., & Pelisson, D. (2000). New insights on eye blindness and hand sight: Temporal constraints of visuo-motor networks. Visual Cognition, 7(6), 785–808. Schenk, T. (2006). An allocentric rather than perceptual deficit in patient D.F. Nature Neuroscience, 9, 1369–1370. Smeets, J. B. J., Brenner, E., De Grave, D. D. J., & Cuijpers, RH. (2002). Illusions in action: Consequences of inconsistent processing of spatial attributes. Experimental Brain Research, 147, 135–144. van Doorn, H., van der Kamp, J., de Wit, M., & Savelsbergh, G. J. P. (2009). Another look at the Müller–Lyer illusion: Different gaze patterns in vision for action and perception. Neuropsychologia, 47(3), 804–812. Veerman, M. M., Brenner, E., & Smeets, J. B. J. (2008). The latency for correcting a movement depends on the visual attribute that defines the target. Experimental Brain Research, 187, 219–228. Westwood, G. A., & Goodale, M. A. (2003). Perceptual illusion and the real-time control of action. Spatial Vision, 16(3–4), 243–254. Westwood, D. A., Heath, M., & Roy, E. A. (2000). The effect of a pictorial illusion on closed-loop and open-loop prehension. Experimental Brain Research, 134(4), 456–463.
Contents lists available at ScienceDirect
Neuropsychologia journal homepage: www.elsevier.com/locate/neuropsychologia
Delayed pointing movements to masked Müller–Lyer figures are affected by target size but not the illusion Matthieu de Wit a,∗ , John van der Kamp a,b , Rich S.W. Masters a a b
Institute of Human Performance, University of Hong Kong, 111-113 Pokfulam Road, Hong Kong SAR, China Research Institute MOVE, Faculty of Human Movement Sciences, VU University Amsterdam, The Netherlands
a r t i c l e
i n f o
Article history: Received 9 April 2010 Received in revised form 4 March 2011 Accepted 14 March 2011 Available online 21 March 2011 Keywords: Vision for action Vision for perception Egocentric information Allocentric information Müller–Lyer illusion Visual masking Delayed action
a b s t r a c t There is ongoing debate with respect to interpretation of the finding that, in contrast to perceptual size judgments, actions are relatively unaffected by the Müller–Lyer illusion. In normal unrestricted viewing situations observers cannot perform an action directed at an object without simultaneously perceiving the object – this makes it difficult to unequivocally establish whether observed effects are a function of vision for perception, vision for action, a combination of both, or of a single all-purpose visual system. However, there is evidence that observers are capable of performing actions towards objects of which they are not consciously aware, implying that two distinct visual thresholds may exist; one accompanying vision for action and one accompanying vision for perception. To investigate this possibility we created a situation in which visual information was presented below the perception threshold, but above the purported action threshold, allowing examination of action responses independent of contributions from vision for perception. Following a perceptual categorization task, participants performed delayed pointing movements towards briefly exposed masked Müller–Lyer targets of different sizes. When the targets were presented below the perception threshold, participants were unable to discriminate between them, yet their delayed pointing movements were affected by target size (but not the illusion). The results imply that vision for action is functional even after a delay and/or that the pickup of egocentric information is associated with a lower visual threshold than the pickup of allocentric information. © 2011 Elsevier Ltd. All rights reserved.
1. Introduction Goodale and Milner (1992) and Milner and Goodale (1995, 2008) proposed that the perception of objects and the visual control of actions directed at those objects are mediated by two functionally and anatomically distinct visual systems (i.e., vision for perception and vision for action). In a paradigm that has been used extensively to examine this proposal, participants perform perception and action responses upon targets embedded in geometrical illusions such as the Müller–Lyer illusion. Meta-analyses indicate that unlike perceptual judgments, actions performed in unrestricted viewing situations are relatively unaffected by (but not immune to) the Müller–Lyer illusion (e.g., Bruno, Bernardis, & Gentilucci, 2008; Bruno & Franz, 2009). Milner and Goodale (1995, 2008; Ganel, Tanzer, & Goodale, 2008; Goodale & Haffenden, 1998) explain these findings by arguing that vision for action and vision for perception exploit distinct types of information to perform their tasks; while vision for action uses egocentric (i.e., body-centered) information, vision for perception relies mainly on allocentric (i.e.,
∗ Corresponding author. Tel.: +852 2589 0585; fax: +852 2855 1712. E-mail address: [email protected] (M. de Wit). 0028-3932/$ – see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.neuropsychologia.2011.03.017
world-centered) information. Gentilucci, Chieffi, Daprati, Saetti, and Toni (1996; see also Hu & Goodale, 2000; Westwood & Goodale, 2003; Westwood, Heath, & Roy, 2000) examined the kinematics of delayed pointing movements directed at occluded Müller–Lyer figures. Restriction of vision by occluding the goal target led to an increase in the effect of the illusion to a level comparable to effects normally associated with perceptual judgments. This use of (retained) allocentric information has been taken to imply that delayed actions are mediated by vision for perception and that vision for action only engages in guiding actions when they are performed online (i.e., in real time) and in full vision (Goodale, Westwood, & Milner, 2004). Ongoing controversy exists with respect to the interpretation of the differential effect of geometrical illusions on perception and action. Several authors have proposed that the relative immunity of actions to geometrical illusions when compared to perceptual measures can be explained by differences in task characteristics between commonly used action and perception tasks, and does not implicate different visual processes (i.e., vision for action and vision for perception) underlying the execution of those tasks. Smeets, Brenner, de Grave, and Cuijpers (2002; see Schenk, 2006 for a related argument) argued that the spatial attributes that are used to perform a task determine susceptibility to an illu-
1904
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909
sion. de Grave, Brenner, and Smeets (2004), for example, found that pointing movements along the shaft of a Müller–Lyer figure (emphasizing the use of size information, which is affected by the illusion) were influenced by the illusion whereas pointing movements perpendicular to the shaft (emphasizing endpoint position information, which is unaffected by the illusion) were not. Franz, Hesse, and Kollath (2009), varied the amount of visual feedback available to participants during the execution of grasps directed at Müller–Lyer targets. Removing feedback at movement onset, at 1/3 of the transport phase and at 2/3 of the transport phase led to a gradual reduction of the illusion effect that was directly related to the amount of available visual feedback. Based on these results, Franz et al. (2009) argued that the increased effect of illusion after a delay is caused by the availability of visual feedback leading to online corrections of the movement (and not by a shift in control from vision for action to vision for perception). In unrestricted viewing situations, observers cannot perform an action directed at an object without simultaneously perceiving the object (Milner & Goodale, 2008; see also Enns & Liu, 2009; van Doorn, van der Kamp, de Wit, & Savelsbergh, 2009). This makes it difficult to unequivocally determine whether observed illusion effects are a function of vision for perception, vision for action, a combination of both, or of a single all-purpose visual system. However, a possible resolution to this problem might exist. There is evidence that vision for action has quicker access to visual information than vision for perception. Pisella, Arzi, and Rossetti (1998; see also Veerman, Brenner, & Smeets, 2008) asked participants to perform reach movements towards stimuli that could be perturbed in either location or color during the ongoing movement. In case of a perturbation, participants were required to stop their movement. Results showed that stop-responses to perturbations of color, an object property that would arguably be picked up by vision for perception, were initiated about 80 ms later than stop-responses to location changes, which are arguably guided by vision for action (Rossetti, Pisella, & Pelisson, 2000). Heath, Maraj, Godbolt, and Binsted (2008; Binsted, Brownell, Vorontsova, Heath, & Saucier, 2007; Heath, Neely, Yakimishyn, & Binsted, 2008; and see Cressman, Franks, Enns, & Chua, 2007 for a related experiment) asked participants to perform pointing movements towards masked briefly exposed (i.e., 13 ms) targets of different sizes. Although participants were unable to perceptually discriminate between targets at above chance levels, Fitts’ law (1954) was preserved in that the movement time for pointing movements towards smaller targets was longer. Notably, these movements were not performed online but after a delay of up to 2 s. These findings suggest that (1) vision for action may remain functional at lower minimum stimulus exposure times than vision for perception and (2) vision for action may be capable of guiding actions performed after a delay. The current experiment was designed to exploit this potential difference in visual threshold between vision for action and vision for perception by creating a situation in which participants performed an action task and a perception task in response to Müller–Lyer targets that were presented below the vision for perception threshold but above the vision for action threshold. This would allow for an examination of action responses independent of contributions from vision for perception. Participants were exposed to masked targets for brief (12 ms) or long (1500 ms) durations. Targets consisted of ‘wings out’ and ‘wings in’ Müller–Lyer figures and neutral (i.e., without wings) figures of three different lengths. Based on the finding by Heath and colleagues (i.e., Binsted et al., 2007; Heath, Maraj, et al., 2008; Heath, Neely, et al., 2008) that Fitts’ law was preserved for action responses to perceptually indiscriminable masked briefly exposed (i.e., 13 ms) targets, it was hypothesized that the briefly exposed targets would exceed the vision for action threshold but not the vision for perception threshold. To verify whether targets were
indeed presented below the vision for perception threshold, participants were first required to indicate the size of the briefly exposed targets in a perception task. In four subsequent action conditions, participants made pointing movements along the shaft of the targets that were presented for brief (i.e., 12 ms) or long durations (i.e., 1500 ms) either as soon as possible after target stimulus offset (i.e., reaction time (RT) delay) or after a delay of 2 s. We expected pointing movements to be scaled to target length regardless of whether they were briefly presented (i.e., below the vision for perception threshold) or not – as long as target stimulus duration exceeded the vision for action threshold. We also had specific expectations with respect to the effect of the illusion on pointing movements directed at the briefly exposed Müller–Lyer targets. Based on the evidence that vision for perception relies mainly on the use of allocentric information (e.g., as implicated by the relatively large effects of the Müller–Lyer illusion on perceptual measures, see Bruno et al., 2008; Bruno & Franz, 2009; Ganel et al., 2008) together with the assumption that the briefly exposed targets were not expected to exceed the vision for perception threshold, we only expected an effect of illusion on the delayed pointing movements for the targets exposed for long durations, not for targets exposed for brief durations. 2. Methods 2.1. Participants Seventeen right-handed participants (8 females) aged 23–60 years (33 ± 10) with normal or corrected-to-normal vision participated in the experiment. They were naïve with regard to the purpose of the experiment and were treated in accordance with the ethical guidelines of the local Institution. 2.2. Materials Stimuli were presented on a 19 in. CRT-monitor (Philips Brilliance 109P4) with a refresh rate of 85 Hz and a resolution of 1024 × 768 pixels using E-Prime 2.0 presentation software (Psychology Software Tools, Pittsburg, PA). An Optotrak 3020 motion analysis system (Northern Digital, Waterloo, Ontario) was used to measure the extent of the pointing movements by recording the position of an infrared light emitting diode placed on the tip of the right index finger with a frequency of 200 Hz. Stimuli consisted of three different lengths (short: 11.5 cm, medium: 14.5 cm, long: 17.5 cm) of wings out and wings in Müller–Lyer figures and neutral figures (i.e., without wings). The lines that made up the figures (shaft and wings) were 5 mm wide. The wings had a length of 3 cm and an angle of 45◦ relative to the shaft. Stimuli were presented randomly in one of six locations (top-left, center-left, bottom-left, top-right, center-right, bottom-right) but appeared equally often on the left and right side and top, center and bottom of the computer screen (see Fig. 1). 2.3. Procedure and design Participants performed an action task and a perception task. In the action task, participants were instructed to place the tip of their right index finger on a fixation dot that was positioned at one end of the to-be-presented target shaft and to fixate their gaze on the tip of their finger (see Fig. 1). If the dot appeared at the left side of the screen, a movement from the left end of the target shaft to the right end of the target shaft was required, and vice versa. The dot was present for 3 s after which the target stimulus was presented. Stimulus exposure duration was either brief (12 ms) or long (1500 ms) and pointing movements had to be made either as quickly as possible after stimulus offset (RT delay) or after a delay of 2000 ms following stimulus offset, in both cases indicated by an auditory start signal (i.e., for RT delays the start signal sounded at stimulus offset and for 2000 ms delays the start signal sounded 2000 ms after stimulus offset). Stimulus presentation was always followed by a mask that was presented for 200 ms and consisted of an array of scrambled target stimuli (see Fig. 1). Participants were required to maintain gaze fixation upon the tip of their finger until they heard the start signal. For the pointing movement, the instruction was to “point as fast and accurately as possible to the other end of the horizontal shaft line when you hear the beep”. The combination of the factors stimulus exposure duration and movement delay led to a total of four action conditions (i.e., 12 ms stimulus exposure × RT/2000 ms movement delay and 1500 ms stimulus exposure × RT/2000 ms movement delay) which were performed in separate blocks by each participant in counterbalanced order. To assess whether participants were able to categorize the target stimuli at 12 ms stimulus exposure duration, participants performed a perception task both before and after the four action blocks. The conditions in the perception task were identical to those in the action task but instead of making a pointing movement participants
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909
1905
(i.e., whether pointing distance is scaled to actual target length) and of stimulus type (i.e., whether pointing distance is affected by the illusion), by submitting the raw data to separate hierarchical regression analyses for each condition. The contribution of the following three predictor variables was tested: (1) actual target length (treated as a numerical variable with 3 levels; short, medium, long), (2) stimulus type (a dummy coded categorical variable with 3 levels; wings out, wings in, no wings) and (3) an interaction term for actual target length and stimulus type (the product of the numerical and the dummy coded variable). We first fitted a one-predictor model with actual target length as the predictor to the data of each condition. A significant fit of this model showed that pointing distance was scaled to actual target length, so we proceeded to test whether, next to actual target length, stimulus type was also a significant predictor of pointing distance (indicating an effect of illusion). To this end, we fitted a two-predictor model with actual target length and stimulus type as predictors to the raw data for each condition. A significant increase in the proportion of explained variance from model one (one predictor) to model two (two predictors) would indicate that next to actual target length, stimulus type also affected pointing distance. To allow for comparison of effects of illusion between conditions, we also computed the percent corrected illusion effect (see Bruno et al., 2008) for each condition using Eq. (2) where ‘expand’ and ‘compress’ stand for pointing distance directed to wings out and wings in stimuli and ‘baseline’ stands for pointing distance directed to the no wings stimuli. Mean effects were subsequently submitted to paired t-tests to examine possible differences between conditions.
%= Fig. 1. Schematic representation of stimulus presentation. In this example the target stimulus is presented at the top-left position, consists of a medium length wings out figure and requires a movement from left to right.
were now required to indicate whether they had seen a short, medium or long target stimulus by pressing a corresponding key on the keyboard. Participants were encouraged to respond fairly quickly (i.e., within 1–3 s after stimulus offset). In each action and perception block, each stimulus type (wings out/wings in/no wings) in combination with each length (short/medium/long) was randomly presented two times leading to a total of 18 experimental trials (i.e., 3 stimulus types × 3 lengths × 2 repetitions) per block. To maintain active engagement in the experiment despite the very high task difficulty in the 12 ms stimulus duration action blocks and the perception blocks, experimental trials were randomly interspersed with trials in which the stimulus duration was 36 ms instead of 12 ms. In these trials, each stimulus type in combination with the short and long target length was presented once (total 6 trials). Similarly, in the long stimulus duration action blocks experimental trials were interspersed with trials in which the stimulus duration was 1524 ms instead of 1500 ms. Each block thus consisted of 18 + 6 = 24 trials of which only the 18 experimental trials were analyzed. To familiarize participants with the different stimulus lengths and types and with the task requirements, the experimental blocks were preceded by a practice block of 18 trials in which all stimuli were presented twice and the stimulus duration was 3 s. In this block, the experimenter verbally indicated the length of the presented target in each trial. Participants performed a total of 162 trials. There was a 5 min break between the third and the fourth block and the experiment was completed in approximately 1 h. 2.4. Data analysis For the perception task, we first computed the percentages of correct responses. To assess whether group performance in the perception task was at chance levels, the percentages of correct responses were submitted to a one-sample t-test and tested against a test value of 33.3% (i.e., chance). Furthermore, to examine whether individual participants were able to discriminate between stimulus lengths at above chance levels, for each participant the number of correct responses were entered in a Z-score proportions test (see Eq. (1), where c is the number of correct responses). The critical Z-value for ˛ = .05 lies between 1.64 and 1.65. A Z-score greater than 1.65 was therefore interpreted as an indication that the participant was performing above chance levels. To examine whether errors made by participants were influenced by the illusion-inducing wings that were present in a subset of the stimuli, we recorded the direction of the errors made in response to the Müller–Lyer stimuli. For both wings in and wings out stimuli the number of errors that corresponded with the direction of the illusion (i.e., the stimulus was categorized respectively as smaller than the presented stimulus and as larger than the presented stimulus) and the number of errors that did not correspond with the direction of the illusion was determined. To test for an effect of illusion on the errors, the number of corresponding and noncorresponding errors was submitted to a paired-samples t-test. Z=
c/(36 − .33)
(1)
expand − compress baseline
× 100
(2)
3. Results 3.1. Perception task First, the percentages of correct responses in the perception block that was completed before the action blocks and in the perception block that was completed after the action blocks were compared. A lack of difference between the two blocks (t(16) = 1.23, p = .236) indicates that no perceptual learning occurred over the course of the experiment (39.2 and 43.5% correct for the first and second perception block, respectively). Group categorization performance in the two perception blocks differed significantly from chance level performance (t(16) = 2.81, p < 0.05). We therefore entered the score of each individual into a proportions test (Z-test) to identify participants who were able to categorize the stimuli above chance levels at 12 ms stimulus exposure time. Because an increase in the number of trials leads to more reliable estimates of proportions and there was no difference in performance between the first and the second perception block, the total amount of correct responses of the 36 trials from the two perception blocks were entered in the proportions test. Out of the seventeen participants, six (3 females) performed above chance levels (see Fig. 2). For these participants the brief stimulus duration (12 ms) exceeded the vision for perception threshold. As we were interested in pointing movements towards stimuli presented below the vision for perception threshold, we excluded these participants from the pointing movement analysis. To assess whether the illusion-inducing wings that surrounded a subset of the stimuli influenced the categorization errors of the eleven participants who performed at chance levels in the perception task, we compared the number of errors that corresponded with the direction of the illusion with the number of errors that did not correspond with the direction of the illusion. No significant differences existed (t(10) = .1.44, p = .181; 8.2 corresponding errors and 6.9 noncorresponding errors, on average), indicating that the presence of wings did not have an effect on categorization performance in the perception task. 3.2. Action task
(.33 − .66)/36
Pointing distance in the action task was determined by subtracting the difference (in mm) between the start position and the end position of the pointing movement for each trial. Following Bruno and Bernardis’ (2003) approach, we examined the extent to which pointing movements varied as a function of target length
The pointing movement analyses were performed on the eleven participants who performed at chance levels in the perception task. We first conducted a boxplot analysis and identified 12 outliers (corresponding to 1.5% of the total number of data points), which
1906
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909
Fig. 2. Percentage of correct responses in the perception task for each participant. The black dashed line indicates the criterion for chance level performance based on the proportions test. The grey dashed line indicates the 33% level. Participants 7, 9, 10, 12, 14 and 17 were excluded from the pointing movement analysis.
Fig. 3. The two-predictor (actual target length, stimulus type) model fitted to the pointing distances of the 1500 ms stimulus exposure (SE) − RT delay (left panel) and 2000 ms delay (right panel) conditions. Circles correspond to individual data points. Wings out and wings in data points are slightly displaced to the left and right to better reveal the pattern in the data.
were subsequently removed from the data set. Fig. 3 shows the pointing distances to the different stimuli in the 1500 ms stimulus exposure conditions. Individuals are not identified by different symbols as there was a high consistency in their pattern of behavior (i.e., if an individual tended to undershoot or overshoot at one target length he or she would do so at other target lengths as well). As one would expect, pointing distance in these conditions was significantly scaled to actual target length. The fit of the one-predictor (actual target length) model was highly significant for both the RT delay condition, F(1, 190) = 416.15, p < 0.0001, R2 = 0.687 (slope = 0.917), and the 2000 ms delay condition, F(1, 190) = 403.19, p < 0.0001, R2 = 0.680 (slope = 0.851). To establish whether, next to actual target length, stimulus type was also a significant predictor of pointing distance (indicating an effect of illusion), we subsequently fitted a two-predictor (actual target length, stimulus type) model to the data of the 1500 ms stimu-
lus exposure conditions. Adding the second predictor to the model led to a significant change in the proportion of explained variance for both conditions, indicating a substantial improvement of the model fit compared to the one-predictor model.1 For the RT delay condition, R2 increased from 0.687 to 0.730 (F(2, 188) = 15.29, p < 0.0001), while for the 2000 ms delay condition R2 increased from 0.680 to 0.725 (F(2, 188) = 15.66, p < 0.0001). This indicates the presence of an effect of illusion in these conditions, which
1 We also fitted a three-predictor model that additionally included an interaction term between actual stimulus length and stimulus type to the data of each condition. This model was used to examine whether the effect of illusion, if any, was modulated by actual target size (i.e., different steepness of slopes for wings in and wings out stimuli). However, because the interaction term did not contribute significantly to the model in any of the conditions, we do not report the statistics of this model.
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909
is further supported by the fact that the t-test for the coefficient reflecting the extent to which pointing distance differed between wings out and wings in stimuli was significant in both conditions; RT delay, t(10) = 5.29, p < 0.0001 (slope wings out = 0.968, wings in = 0.869), 2000 ms delay, t(10) = 5.58, p < 0.0001 (slope wings out = 0.930, wings in = 0.846). A t-test comparing the mean percent corrected illusion effects of the two 1500 ms stimulus exposure conditions revealed no differences between these conditions (RT delay: 10.47%, SD = 7.41, 2000 ms delay: 10.79%, SD = 7.74). Analysis of the pointing distances in the 12 ms stimulus exposure conditions revealed a rather different pattern (see Fig. 4). Although the one-predictor model explained only a small amount of the variance in these conditions, this model nevertheless showed a significant fit to the data of both the RT delay condition, F(1, 186) = 5.91, p < 0.05, R2 = 0.031 (slope = 0.209) and the 2000 ms delay condition, F(1, 192) = 10.72, p < 0.01, R2 = 0.053 (slope = 0.264), indicating that pointing movements were significantly scaled to actual target length. However, for these conditions, adding the second predictor to the model did not lead to a significant change in the proportion of explained variance, indicating that at 12 ms stimulus exposure duration, the presence of wings did not have an effect on pointing distance. In fact, for the RT delay condition, adding the predictor stimulus type rendered the model’s fit nonsignificant (Model 1, p = 0.012, Model 2, p = 0.121) while for the 2000 ms delay condition, adding the second predictor merely reduced the model’s significance (Model 1, p = 0.001, Model 2, p = 0.008). The absence of an effect of illusion at 12 ms stimulus exposure is further supported by the fact that the t-test for the coefficient reflecting the extent to which pointing distance differed between wings out and wings in stimuli was not significant in both conditions. Slopes in the RT delay condition measured 0.088 and 0.305 for wings out and wings in stimuli respectively. In the 2000 ms delay condition they measured 0.334 and 0.428 for wings out and wings in stimuli. The mean percent corrected illusion effect measured 1.50% (SD = 15.82) in the RT delay condition and 2.94% (SD = 13.43) in the 2000 ms delay condition2 . These effects were not significantly different from each other or from 0%. 4. Discussion Under the assumption that vision for perception and vision for action remain functional at different minimum stimulus exposure times, we attempted to present target stimuli below the vision for perception threshold but above the vision for action threshold. In a perception and an action task, participants were exposed to masked Müller–Lyer figures (wings out, wings in, no wings) of different lengths for brief (12 ms) or long (1500 ms) durations. Analysis of performance levels in the perception task showed that 6 out of the 17 participants were able to discriminate between stimulus lengths at above chance levels, implying that for these participants the briefly exposed stimuli exceeded the vision for perception threshold. Strikingly, analysis of pointing distances of the remaining 11 participants, for whom the briefly presented stimuli did not exceed
2 Bruno and Franz (2009) have argued that a correction formula incorporating the slope of the function that relates movement distance to target length in the denominator of the fraction might be appropriate when there are differences in scaling between conditions, as is the case for the present 12 and 1500 ms stimulus exposure conditions. Doing so returns corrected percent measures for the 12 ms stimulus exposure time conditions of 9.2 % (SD = 74.1) and 12.8 % (SD = 70.2) for the RT - and 2000 ms Delay conditions, respectively. However, note the high standard deviations. Clearly, with the relatively high variability in pointing distance at 12 ms stimulus exposure and the lack of a consistent pattern in the relationship between pointing distance and wing orientation, this correction approach produces somewhat misleading percent measures. Accordingly, t-tests show that, despite their apparently large size, the corrected percent measures do not differ from 0 % (p = 0.310 and 0.194, respectively).
1907
the vision for perception threshold, showed that their actions were nevertheless scaled to actual stimulus length. Thus, while participants were exposed to the same set of stimuli, the perception task was accompanied by a higher visual threshold than the action task. Several studies (i.e., Franz et al., 2009; Gentilucci et al., 1996; Hu & Goodale, 2000; Westwood & Goodale, 2003; Westwood et al., 2000) have reported that the introduction of a delay causes a profound increase in the effect of illusion on actions; this is typically explained as an effect of vision for perception supplanting vision for action (but see Franz et al., 2009). The current study replicates these findings. Pointing movements after a delay were significantly affected by the illusion for targets presented above the vision for perception threshold. This effect was the same for movements performed after a short (i.e., start signal at stimulus offset, RT delay) or a long delay (i.e., start signal 2 s after stimulus offset). However, when targets were presented below the vision for perception threshold, the illusion did not affect pointing distance, both when movements were performed at RT delay and after a delay of 2 s. Together with our finding that pointing movements were nevertheless scaled to physical target length, this may suggest that vision for action is not fully disrupted when movements are performed after a delay. One might argue that because a subset of the stimuli was surrounded by illusion-inducing wings, the possibility exists that an effect of illusion rather than an inability to perceive the stimuli may have caused the participants to perform at chance levels in the perception task. However, because on average the size of perceptual illusion effects is between 10 and 20% of physical target length (see Bruno et al., 2008; Bruno & Franz, 2009) participants should have been able to discriminate between the targets in our experiment regardless of whether wings were present or not. Moreover, under this assumption the perceptual errors made by participants should have been made in accordance with the direction of the illusion. That is, the size of wings out stimuli should have been overestimated more often than the size of wings in stimuli, and the size of wings in stimuli should have been underestimated more often than the size of wings out stimuli, which was not the case. In addition, participants frequently reported no experience of seeing the stimuli. Notably, the possibility that some participants were in fact able to perceptually discriminate between briefly presented stimuli does potentially apply to the studies by Heath and colleagues (e.g. Binsted et al., 2007; Heath, Maraj, et al., 2008; Heath, Neely, et al., 2008; see also Cressman et al., 2007)3 who found that Fitts’ law (1954) was preserved for delayed pointing movements towards masked stimuli. In sum, our results provide solid evidence for the existence of distinct task-dependent visual thresholds for the pickup of visual information. Our results uncover a problem with the common interpretation of the increased effects of illusion on delayed action in which it is argued that delayed actions are necessarily guided by vision for perception (as implied by their dependence on allocentric information; see Franz et al., 2009; Goodale et al., 2004; Hu & Goodale, 2000; Westwood & Goodale, 2003 for a related argument). To our knowledge, this interpretation has never been tested by preventing vision for perception from contributing to actions performed after a delay. The claim is therefore an indirect one and, as evidenced by our finding that delayed pointing movements are scaled to the
3 In stead of examining the discrimination performance of individual participants with proportions tests, these authors examined discrimination performance at the group level with (one-sample) t statistics or analysis of variance. It is possible that while these tests did not show an overall effect, some participants may in fact have been able to discriminate between the masked stimuli above chance levels. In the current experiment, we removed such participants from the pointing movement analysis.
1908
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909
Fig. 4. The one-predictor (actual target length) model fitted to the pointing distances of the 12 ms stimulus exposure (SE) − RT delay (left panel) and 2000 ms delay (right panel) conditions. Plotting conventions are the same as in Fig. 3.
veridical size of Müller–Lyer targets presented below the vision for perception threshold, may have to be revisited. Evidence against the notion that vision for action is capable of guiding delayed actions that is not subject to the above criticism comes from a study of patient D.F., who has damage to the ventral stream (i.e., the anatomical substrate of vision for perception) but has an intact dorsal stream (i.e., the anatomical substrate of vision for action). Although she is well able to grasp visible objects, her ability to accurately grasp objects after a delay is completely disrupted (Goodale, Jakobson, & Keillor, 1994). However, in a recent study D.F. displayed preserved performance in a perception task that depended on the use of egocentric information and impaired performance in an action task that depended on the use of allocentric information, implying that her deficit may be task (i.e., perception or action) independent and instead related to an inability to exploit allocentric information (Schenk, 2006). A similar, information based (as opposed to task based), distinction may account for the results obtained in the current experiment. Closer inspection of our perception task shows that it could only be performed by relying on allocentric information. Categorization necessarily required a comparison of the length of the presented stimulus with the recalled length of previously presented stimuli, thus leaving participants no choice but to use allocentric information. Both the chance level categorization performance and the absence of an effect of wing orientation on the categorization errors imply that participants were unable to exploit allocentric information in the perception task. The action task, on the other hand, did not require memory based length comparisons. In this task participants made an absolute movement along the stimulus shaft, allowing them to potentially base their movement both on egocentric information and on the allocentric information that was present within the stimuli (i.e., the wings). The fact that at 12 ms stimulus exposure pointing distance was scaled to actual target length but not affected by the presence of wings implies that in this task the allocentric information was presented below threshold but the egocentric information was not. It is thus unclear whether the presence of distinct visual thresholds that is evidenced by our findings indicates that vision for action is associated with a lower visual threshold than vision for perception or instead whether the pickup of egocentric informa-
tion is associated with a lower visual threshold than the pickup of allocentric information, possibly irrespective of whether a task is perceptual or motoric in nature. Under the latter assumption, vision for perception may have informed the delayed pointing movements in our experiment, but was depending on the use of egocentric information. Vision for perception is commonly assumed to rely mainly on the use of allocentric information (Ganel et al., 2008; Milner & Goodale, 1995, 2008, but see Briscoe, 2009). However, there is evidence indicating that vision for perception has the potential to exploit egocentric information as well. Predebon (2006), for example, showed that the effect of the Müller–Lyer illusion on perceptual length estimations disappears when participants are instructed to ignore the wings (i.e., the allocentric information). van Doorn et al. (2009) examined the gaze behavior of participants engaged in a perceptual length estimation task and a grasping task directed at Müller–Lyer stimuli. Results showed a significant negative correlation between the amount of time that was spent looking at the target shaft (containing the pertinent sources of egocentric information for grasping) and the size of the illusion effect on hand aperture in both the grasping task and the perceptual length estimation task, showing that in either task participants had the potential to rely on egocentric information. As a third and final alternative, our findings are in accordance with the possibility that there exists both a distinction in threshold between vision for perception and vision for action and a distinction in threshold for the pickup of allocentric information and egocentric information. Future experiments are needed to distinguish between the described possibilities and to further examine the interrelationships between vision for perception and vision for action and the pickup of allocentric and egocentric information. References Binsted, G., Brownell, K., Vorontsova, Z., Heath, M., & Saucier, D. (2007). Visuomotor system uses target features unavailable to conscious awareness. Proceedings of the National Academy of Sciences of the United States of America, 104(31), 12669–12672. Briscoe, R. (2009). Egocentric spatial representation in action and perception. Philosophy and Phenomenological Research, 79(2), 423–460. Bruno, N., & Bernardis, P. (2003). When does action resist visual illusions? Effector position modulates illusory influences on motor responses. Experimental Brain Research, 151, 225–237.
M. de Wit et al. / Neuropsychologia 49 (2011) 1903–1909 Bruno, N., Bernardis, P., & Gentilucci, M. (2008). Visually guided pointing, the Müller–Lyer illusion, and the functional interpretation of the dorsal–ventral split: Conclusions from 33 independent studies. Neuroscience and Biobehavioral Reviews, 32(3), 423–437. Bruno, N., & Franz, V. H. (2009). When is grasping affected by the Müller–Lyer illusion? A quantitative review. Neuropsychologia, 47(6), 1421–1433. Cressman, E. K., Franks, I. M., Enns, J. T., & Chua, R. (2007). On-line control of pointing is modified by unseen visual shapes. Consciousness and Cognition, 16, 265–275. de Grave, D. D. J., Brenner, E., & Smeets, J. B. J. (2004). Illusions as a tool to study the coding of pointing movements. Experimental Brain Research, 155, 56–62. Enns, J. T., & Liu, G. (2009). Attentional limits and freedom in visually guided action. Progress in Brain Research, 176, 215–226. Fitts, P. M. (1954). The information capacity of the human motor system in controlling the amplitude of movement. Journal of Experimental Psychology, 47(6), 381–391. Franz, V. H., Hesse, C., & Kollath, S. (2009). Visual illusions, delayed grasping, and memory: No shift from dorsal to ventral control. Neuropsychologia, 47(6), 1518–1531. Ganel, T., Tanzer, M., & Goodale, M. A. (2008). A double dissociation between action and perception in the context of visual illusions: Opposite effects of real and illusory size. Psychological Science, 19, 221–225. Gentilucci, M., Chieffi, S., Daprati, E., Saetti, M. C., & Toni, I. (1996). Visual illusion and action. Neuropsychologia, 34(5), 369–376. Goodale, M. A., & Haffenden, A. (1998). Frames of reference for perception and action in the human visual system. Neuroscience and Biobehavioral Reviews, 22(2), 161–172. Goodale, M. A., Jakobson, L. S., & Keillor, J. M. (1994). Differences in the visual control of pantomimed and natural grasping movements. Neuropsychologica, 32, 1159–1178. Goodale, M. A., & Milner, A. D. (1992). Separate visual pathways for perception and action. Trends in Neurosciences, 15(1), 20–25. doi:10.1016/ 0166-2236(92)90344-8 Goodale, M. A., Westwood, D. A., & Milner, A. D. (2004). Two distinct modes of control for object-directed action. Progress in Brain Research, 144, 131–144.
1909
Heath, M., Maraj, A., Godbolt, B., & Binsted, G. (2008). Action without awareness: Reaching to an object you do not remember seeing. PLoS ONE, 3(10). Heath, M., Neely, K. A., Yakimishyn, J., & Binsted, G. (2008). Visuomotor memory is independent of conscious awareness of target features. Experimental Brain Research, 188(4), 517–527. Hu, Y., & Goodale, M. A. (2000). Grasping after a delay shifts size-scaling from absolute to relative metrics. Journal of Cognitive Neuroscience, 12(5), 856–868. Milner, A. D., & Goodale, M. A. (1995). The visual brain in action. Oxford: Oxford University Press. Milner, A. D., & Goodale, M. A. (2008). Two-visual systems re-viewed. Neuropsychologia, 46(3), 774–785. Pisella, L., Arzi, M., & Rossetti, Y. (1998). The timing of color and location processing in the motor context. Experimental Brain Research, 121(3), 270–276. Predebon, J. (2006). Decrement of the Müller–Lyer and Poggendorff illusions: The effects of inspection and practice. Psychological Research, 70, 384–394. Rossetti, Y., Pisella, L., & Pelisson, D. (2000). New insights on eye blindness and hand sight: Temporal constraints of visuo-motor networks. Visual Cognition, 7(6), 785–808. Schenk, T. (2006). An allocentric rather than perceptual deficit in patient D.F. Nature Neuroscience, 9, 1369–1370. Smeets, J. B. J., Brenner, E., De Grave, D. D. J., & Cuijpers, RH. (2002). Illusions in action: Consequences of inconsistent processing of spatial attributes. Experimental Brain Research, 147, 135–144. van Doorn, H., van der Kamp, J., de Wit, M., & Savelsbergh, G. J. P. (2009). Another look at the Müller–Lyer illusion: Different gaze patterns in vision for action and perception. Neuropsychologia, 47(3), 804–812. Veerman, M. M., Brenner, E., & Smeets, J. B. J. (2008). The latency for correcting a movement depends on the visual attribute that defines the target. Experimental Brain Research, 187, 219–228. Westwood, G. A., & Goodale, M. A. (2003). Perceptual illusion and the real-time control of action. Spatial Vision, 16(3–4), 243–254. Westwood, D. A., Heath, M., & Roy, E. A. (2000). The effect of a pictorial illusion on closed-loop and open-loop prehension. Experimental Brain Research, 134(4), 456–463.