No Dissociation Between Perception and Action in Patient DF PDF

The Journal of Neuroscience, February 8, 2012 32(6):2013–2017 2013 Brief Communications No Dissociation between Perception and Action in Patient DF When Haptic Feedback is Withdrawn Thomas Schenk Department of Neurology, University of Erlangen-Nuremberg, Germany Goodale et al. (1991) reported...

The Journal of Neuroscience, February 8, 2012 32(6):2013–2017 2013 Brief Communications No Dissociation between Perception and Action in Patient DF When Haptic Feedback is Withdrawn Thomas Schenk Department of Neurology, University of Erlangen-Nuremberg, Germany Goodale et al. (1991) reported a striking dissociation between vision for perception and action. They examined DF, a human patient who had damage to her ventral visual stream and suffered from visual form agnosia. She was unable to perceive an object’s size but could match the opening of her hand to the object’s size during grasping. It was concluded that grasping relied on a separate representation of visual size in the dorsal stream and required no visual input from the ventral stream. This observation inspired the influential percep- tion–action model, which claimed separate visual streams for perception and action. However, in grasping (but not in corresponding perceptual tasks), participants receive haptic feedback after each trial. Using this feedback, DF might compensate for her impaired size-vision. I reexamined DF’s grasping behavior using a mirror apparatus to dissociate the image of an object from its physical presence. DF’s grasping was only normal when she received haptic feedback. Thus, in grasping, DF can rely on haptic feedback to compensate for her deficit in size-perception. This can explain why her grasping is significantly better than her perceptual performance. The findings emphasize the extent of early interstream interactions and highlight the multimodal nature of sensory processing in the dorsal stream. Introduction good visuomotor behavior is provided (Milner and Goodale, Goodale and Milner (1992) proposed that vision for percep- 2008, Goodale and Milner, 2010, Westwood and Goodale, 2011). tion and action is processed in separate pathways. They ad- Recently, I described an alternative account for DF’s visuomo- opted the distinction between a ventral system, passing tor robustness (Schenk, 2010). I argued that DF’s preserved abil- information from V1 to areas in the temporal lobe, and the ity for visually guided action might reflect the fact that many dorsal system, a visual stream terminating in the parietal lobe visuomotor tasks can be solved on the basis of multiple sensory (Ungerleider and Mishkin, 1982). In this model, the ventral cues. Some of those cues are not even visual and therefore will not system serves object and scene recognition and the dorsal sys- be affected by DF’s visual disorder. Grasping provides an exam- tem uses vision to guide actions (Milner and Goodale, 2006). ple. Our ability to adjust the opening of our hand to the object’s But in recent years, many of the model’s claims were chal- size is informed by visual and haptic information about the ob- lenged (Schenk and McIntosh, 2010; Schenk et al., 2011), most ject’s size (Bingham et al., 2007). Haptic feedback is received at notably, its claim that action is immune to perceptual illusions the end of a grasp. It allows the subject to adjust their initial visual (Smeets and Brenner, 2006; Franz and Gegenfurtner, 2008), its estimate of the object size and to generate more accurate grip-size claim that the dorsal stream does not provide observer-invariant adjustments in future grasping movements. The presence of hap- visual information (Konen and Kastner, 2008) and does not tic feedback during grasping could thus explain one of the most make a contribution to perception (Schenk, 2006), the assertion widely cited studies in support of the perception–action model. that visual form agnosia and optic ataxia constitute a proper In 1991, Goodale and colleagues reported that DF successfully double-dissociation (Pisella et al., 2006), and the assumption that grasps objects whose size she cannot perceive. If the above ac- memory-based action is not processed by the dorsal stream count is correct, I would expect that DF fails to show normal (Himmelbach and Karnath, 2005; Himmelbach et al., 2009). In grasping behavior when haptic feedback is withdrawn. response to this critique, Goodale and colleagues noticed that critics fail to take the evidence from their patient, DF, into ac- Materials and Methods count and argued that one cannot dispute the two-visual stream Participants DF (aged 53 years at the time of testing) and two samples of 10 age- hypothesis unless an alternative account for DF’s surprisingly matched healthy controls took part in the study. The first sample of controls (10 females; mean age, 55 years; SD, 5.6 years) took part in tasks 1 and 2; the second sample (9 females; mean age, 50 years; SD, 14.0 years) took part in tasks 3–5. All subjects were right-handed, had normal or Received July 5, 2011; revised Nov. 22, 2011; accepted Dec. 19, 2011. corrected-to-normal vision, and volunteered for participation after be- Author contributions: T.S. designed research; T.S. performed research; T.S. analyzed data; T.S. wrote the paper. ing informed about the purpose and conduct of the experiment in agree- I thank D.F. and the healthy participants for their participation. The author declares no financial conflict of interest. ment with the Declaration of Helsinki. The study was performed at the Correspondence should be addressed to Thomas Schenk, Neurology, University of Erlangen-Nuremberg, University of Durham, UK and approved by the local ethics committee. Schwabachanlage 6a, 91054 Erlangen, Germany. E-mail: [email protected]. DF suffered a carbon monoxide intoxication in 1988 that led to exten- DOI:10.1523/JNEUROSCI.3413-11.2012 sive bilateral brain damage mainly affecting the lateral aspects of the Copyright © 2012 the authors 0270-6474/12/322013-05$15.00/0 occipital gyri extending into the parasagittal occipitoparietal region (for 2014 J. Neurosci., February 8, 2012 32(6):2013–2017 Schenk No Dissociation between Perception and Action Figure 1. Setup. a, View of the mirror apparatus from behind the participant. The table was separated into two halves by a vertically positioned mirror (gray rectangle). Objects were placed in front of and/or behind the mirror. An occluder (vertical brown rectangle; in reality, the occluder was opaque) prevented direct view of the object. Participants could only see the object’s (and LED’s) reflection in the mirror and perceived the object (and LED) at its mirror-symmetrical position behind the mirror. The participant’s hand was placed behind the mirror, outside of the participant’s view. b, View from above. Objects could be placed at three different positions in front of and at corresponding positions behind the mirror. The positions were 10 cm apart and marked by holes into which a pin at the bottom of the objects (see c) could be fitted. The start-position of the hand was 22 cm to the right and 10 cm in front of the nearest target position. c, Target objects. Three pairs of cylindrically shaped objects were used. All objects had a height of 70 mm and varied in diameter between 35 and 60 mm. The pin at the bottom of each object fitted into the holes defining the three target positions, ensuring precise positioning of the objects. more details, see James et al., 2003). DF suffers from profound visual Standard grasping (task 3). In this condition, participants saw the ob- form agnosia. A detailed description of her perceptual deficits can be ject at the same position where the objects was felt and picked up. I used found in Milner et al. (1991). the middle position (Fig. 1b). Each of the three different objects was presented 15 times, resulting in a total of 45 trials (Fig. 2a). Testing procedure and apparatus Grasping without haptic feedback (task 4). In this condition, there was Except for tasks 1 and 2, the setup (mirror apparatus) presented in Figure no object behind the mirror. In all other respects, the task was identical to 1, a and b, was used. Figure 1c depicts the three different grasping objects. task 3 (Fig. 2b). The grasping movements were recorded with a CMS70-motion registra- Grasping with intermittent haptic feedback (task 5). An object was pres- tion system (spatial resolution, 0.1 mm; sampling frequency, 50 Hz; Ze- ent behind the mirror in half of all trials. Trials with and without physical bris). Participants used their right hand. Three markers were placed on object were randomly interleaved. The presence of a physical object be- the hand, one each at the index finger, thumb, and wrist. At the start of hind the mirror was signaled by a red LED (Fig. 2c,d). This meant that each grasp, participants placed their hand on a start button with their participants knew beforehand whether they would come into contact index finger and thumb in contact. The grasping hand was placed behind with a physical object or not. This condition was introduced to examine the mirror. Participants could, therefore, not see their hand during the whether knowledge about the presence or absence of a physical object experiment. LCD goggles (PLATO; Translucent), worn by all partici- behind the mirror affected grasping performance. In healthy subjects, pants, turned translucent at the start of the trial and became opaque after there is evidence that such intermittent haptic feedback will lead to grasp- trial completion to ensure that participants did not see the experiment- ing performance that is almost indistinguishable from grasping with full er’s manipulations between trials. Participants were instructed to grasp haptic feedback (Bingham et al., 2007). It is worth noting that the influ- the object and to respond as fast and accurately as possible. For each ence of haptic feedback on grasping comes from preceding trials. Since grasping condition, participants were given 20 practice trials. the frequency of haptic feedback in preceding trials will have been the Tasks same for the two types of trials, performance in trials with and without Size-discrimination task (task 1). In each trial, two objects of different object will have benefited from haptic feedback to the same extent. There diameters were presented. Three levels of difficulty were used: easy (small were six different trials (with vs without physical object and three differ- vs big objects), medium (medium vs big objects), and difficult (small vs ent objects); each type of trial was repeated 10 times, producing a total of medium objects). Participants were asked to point to the bigger object; 60 trials presented in a randomized order. their responses were recorded by the experimenter. Each combination Grasping with dissociated positions (task 6). There are two relevant was presented 20 times, thus leading to a total of 60 trials. differences between manual estimation and standard grasping. One is the Manual estimation task (task 2). Participants were asked to indicate availability of haptic feedback in grasping but not manual estimation. with the index finger and thumb of their right hands the diameter of the The second is the fact that in manual estimation, participants direct their presented object. Distance between those fingers, measured with the Ze- pretend-grasp to a position that differs from the position of the target bris System, provided the value for the grip aperture, which was used in object. In task 4, I examined the role of haptic feedback; in task 6, I the computation of grip-size-scaling slopes (see Data analysis, below). I examined the role of dissociating the position of the target object from used the same objects as in task 1 (each object presented 15 times, 45 trials the grasping position. To dissociate target and grasping position, an ob- in total, presented in a randomized sequence). ject was presented in the middle position but asked participants to pick Schenk No Dissociation between Perception and Action J. Neurosci., February 8, 2012 32(6):2013–2017 2015 differed significantly from that of healthy observers (t(9) ⫽ ⫺333,712, p ⬍ 0.00001). Manual estimation (task 2) Healthy observers matched the size of the objects well, producing a mean slope of 0.98 (SD, 0.25). In comparison, DF’s slope was 0.05. The modified t test confirmed the abnormality of DF’s per- formance (t(9) ⫽ ⫺3.585, p ⬍ 0.0029). Standard grasping (task 3) The results are illustrated in Figure 3. This figure presents the maximal grip aperture of one healthy observer (Fig. 3, left) and DF (Fig. 3, right) as a function of object size. The grip-size slope values for all grasping tasks are presented in Figure 4. DF’s grip- size slope was in the normal range (t(9) ⫽ ⫺0.445, p ⬍ 0.33; Fig. 3). The differences between the size discrimination and grasping and between manual estimation and grasping were more pro- nounced for DF than for healthy subjects (size vs grasping: t(9) ⫽ 63.72, p ⬍ 0.000001; manual estimation vs grasping: t(9) ⫽ 2.55, p ⬍ 0.0307). Therefore, both comparisons (standard grasping vs size discrimination and standard grasping vs manual estimation) fulfilled the criteria of a classical dissociation. Role of haptic information (task 4) Without haptic feedback, DF’s performance was abnormal (t(9) ⫽ ⫺3.708, p ⬍ 0.00243; slope measures for healthy partici- pants ranged between 0.76 and 1.56; DF’s grip-size slope was 0.07; Fig. 4). The difference between task 3 (standard grasping) and task 4 (no haptic feedback) was significantly bigger for DF Figure 2. Setup and illustration of the different grasping tasks. a, Standard grasping task than for healthy controls (t(9) ⫽ 3524, p ⬍ 0.00648). (task 3). The visible object (white circle) and the felt object (gray circle) were in matching positions. b, Grasping without haptic feedback (task 4). Same as task 3, but without an object Role of knowledge (task 5) behind the mirror. c, Haptic feedback in 50% of all trials—trial without haptic feedback (task DF’s grip-size scaling in this task with intermittent haptic 5). There was no object behind the mirror and the LED was switched off. d, Haptic feedback in feedback was in the normal range. This was true both for trials 50% of all trials—trial with haptic feedback (task 5). There was an object behind the mirror and without object (t(9) ⫽ ⫺1.695, p ⬍ 0.0621) and trials with the LED (red circle) was switched on. e, Grasping with dissociated positions (task 5). The object object (t(9) ⫽ ⫺1307, p ⬍ 0.118). Admittedly, her perfor- was perceived in the middle position but had to be grasped in the far condition. The target mance was in the lower part of the normal range, but there were position for the grasp was indicated by the position of the LED. An object behind the mirror at also healthy participants with comparably small scaling factors. I the far position provided haptic feedback. conducted a revised standardized difference test (Crawford et al., 2003) to examine whether the performance difference between up a corresponding but invisible object in the far position. This position the condition with and without object was more pronounced for was indicated by a red LED (Fig. 2e). Importantly, a physical object was DF than for healthy controls. No significant difference was found present at the far position (behind the mirror), meaning that participants (t(9) ⫽ 0.435, p ⬍ 0.673), indicating that DF’s grip-size adjust- received veridical haptic feedback. Each of the three objects was pre- ment was not significantly more affected by the knowledge about sented 15 times in a randomized sequence, resulting in a total of 45 trials. the presence or absence of a physical object than that of healthy Data analysis controls. This is consistent with the assumption that haptic feed- Maximum grip aperture (i.e., maximum distance between index and back on the previous trial, which is comparable for trials with and thumb) was calculated for all grasping tasks. To obtain a single measure without object (see Materials and Methods, above), but not for the subjects’ ability to match their hand-opening to the size of the knowledge, which is different for trials with and without object, object, the slope of the grip-size function was computed using the fol- determines DF’s performance. The revised standardized differ- lowing formula: [(GMm ⫺ GMs)/(Wm ⫺ Ws) ⫹ (GMb ⫺ GMm)/(Wb ⫺ ence test was also applied to examine whether the performance Wm)]/2, where GMb, GMm, and GMs are the maximum grip aperture difference between the condition with partial haptic feedback and Wb, Wm, and Ws and the width for big, medium, and small objects, (task 5) and the condition without haptic feedback (task 3) was respectively. To test for the presence of classical dissociations, I used Crawford et al.’s criteria and tests (Crawford and Howell, 1998; Crawford more pronounced for DF than for healthy controls. This differ- and Garthwaite, 2002; Crawford et al., 2003). For task, 1 a binomial test ence was significant (t(9) ⫽ 2.348, p ⬍ 0.0434), indicating that the was used. (partial) introduction of haptic feedback benefited DF signifi- cantly more than healthy controls. Results Size discrimination (task 1) Grasping with dissociated positions (task 6) All healthy observers achieved error-free performance. DF erred DF’s grip-size adjustment (i.e., the slope of the grip-size func- in 35% of all trials. Her performance was only above chance in the tion) was clearly abnormal in this task (t(9) ⫽ ⫺3.39, p ⬍ 0.004). easy condition (easy: 85% correct, p ⬍ 0.003; medium: 65%, p ⬍ Moreover, the difference in DF’s grip-size adjustment for stan- 0.263; difficult: 45%, p ⬍ 0.824). Accordingly, her performance dard grasping and grasping with dissociated positions was signif- 2016 J. Neurosci., February 8, 2012 32(6):2013–2017 Schenk No Dissociation between Perception and Action icantly greater than the difference found for healthy subjects (t(9) ⫽ 2339, p ⬍ 0.044). Thus, there was a classical dissoci- ation between standard grasping and grasping with dissociated positions. Discussion Our findings suggest that the critical dif- ference between manual estimation, size discrimination, and grasping is that one task, namely grasping, provides access to sensory information that is unavailable in Figure 3. Maximum grip aperture (gripmax) as a function of object size in task 3 (standard grasping). The dots correspond to the other tasks. One such source of infor- gripmax values in individual trials; the short horizontal lines correspond to the average value of gripmax for a given object size. The mation is haptic feedback. Without haptic slope of the red line connecting the mean values for the three object sizes was used as our dependent measure for all grasping tasks feedback, DF’s grasping performance was and indicates how well subjects matched their hand-opening (gripmax) to the objects’ diameter. Left, Performance of a healthy not better than her performance in the control subject. Right, Performance of DF. manual estimation task. When intermit- tent haptic feedback was provided, her performance improved, but more importantly, DF’s performance in this task suggests that it is really haptic feedback and not the expectation of encounter- ing a physical object that is important. This is confirmed by the fact that her grasping performance is the same for trials where she expected an object and for those where no object was expected. But haptic feedback is not everything. When haptic feedback was provided but target and grasp positions were dissociated, DF failed to produce normal grip-size scaling. It is not entirely clear why DF’s performance was so poor in this condition. One possi- Figure 4. Size-grip slopes for DF and healthy control subjects. DF’s performance was within the normal range for the standard grasping task (task 3) and the task where haptic feedback was bility is that the dissociation of target and grasp position deprived available in 50% of all trials (task 5). In task 4 (grasping without haptic feedback in all trials) and DF of egocentric cues. In the standard condition (i.e., target and in task 6 [grasping with dissociated positions (dissoc. pos.)], DF’s grasping performance was grasp position are identical), index finger and thumb can be di- outside the normal range. rected to the visible edges of the object (Smeets and Brenner, 1999). In this case, it is not necessary to compute the object width, terstream interactions and stress the important role of multi- it is sufficient to code the position of the object’s grip surfaces modal integration for the successful guidance of actions. relative to the position of the grasping fingers. When target and The fact that the range of conditions under which DF can grasp position are dissociated, the relative distance between the perform normal grip-size scaling is significantly narrower than position of fingers and the grip surfaces of the object no longer that of healthy people suggests that ventral stream areas make a correspond to the amplitude and direction of the required finger critical contribution to the ability of healthy subjects to perform movements and thus a different strategy and different visual cues successful grasping movements under a wide range of sensory are required. The fact that DF was successful in the standard but conditions. In the past, Milner and Goodale (2008) argued that not in the dissociated condition suggests that she was not relying the contribution of the ventral stream to the control of action is on the one visual cue that was available in both conditions, restricted to memory-guided action and aspects of motor plan- namely the object’s width. It seems that DF relies on a combina- ning. In contrast, motor programming was seen as the exclusive tion of different sensory information, such as haptic feedback, domain of the dorsal stream (Goodale and Milner, 2010). While egocentric cues, and probably also visual feedback. While in our this distinction between planning and programming is not with- study, participants could not see their hand and therefore did not out its problems (Schenk, 2010), it is quite uncontroversial that receive visual feedback, visual feedback was available in a study by grip-size scaling should be seen as an example of motor program- Westwood and colleagues (2002). This might explain why in their ming (Dijkerman et al., 2009). The demonstration of deficits in study, DF could adjust her grip-size to 2D objects. Such 2D ob- grip-size scaling after extensive (but not exclusive) ventral stream jects do not provide veridical haptic feedback. DF’s success in damage therefore implies that ventral stream areas also contrib- grasping them might therefore suggest that she does not have to ute to aspects of motor programming. This conclusion is well in rely on haptic feedback but can also use online visual feedback to line with a recent fMRI study. Verhagen and colleagues (2008) guide her fingers to their final grasping positions. examined the brain activity associated with grasping objects pre- In conclusion, it appears that DF’s superior performance in sented at different slants and under different viewing conditions grasping reflects the abundance of sensory information in this (i.e., monocular vs binocular viewing). An object’s orientation in condition. Our findings, therefore, challenge the conventional depth can either be decoded on the basis of binocular cues or on interpretation of one of the most widely cited studies in support the basis of pictorial cues. It is assumed that such pictorial cues of the perception–action model (Goodale et al., 1991). Providing are processed in ventral stream areas. Verhagen and colleagues an alternative account for one piece of evidence does not chal- (2008) found that in the absence of binocular cues, activity of lenge the perception–action model in its entirety, but it suggests ventral stream area LO (lateral occipital area) and its functional that the task that the model’s proponents (for example, see West- connectivity to areas in the dorsal stream significantly increased wood and Goodale, 2011) have assigned to their critics, namely with increasing object slant. The authors interpreted their find- finding alternative accounts for DF’s visuomotor behavior, can ings as evidence for an early interaction of ventral and dorsal be solved. These findings also suggest a greater emphasis on in- stream areas in the sensory control of movements. Schenk No Dissociation between Perception and Action J. Neurosci., February 8, 2012 32(6):2013–2017 2017 The critical role of haptic feedback for DF’s success in grasping Jackson SR (2010) Is the visual dorsal stream really very visual after all? also highlights the multimodal nature of the sensory control of Cogn Neurosci 1:68 – 69. James TW, Culham J, Humphrey GK, Milner AD, Goodale MA (2003) Ven- grasping. It seems that the visual information in the dorsal stream tral occipital lesions impair object recognition but not object-directed about the target object is not enough to generate an accurate grasping: an fMRI study. Brain 126:2463–2475. grasping movement. Either haptic information, visual informa- Konen CS, Kastner S (2008) Two hierarchically organized neural systems tion from the ventral stream, or online visual feedback is also for object information in human visual cortex. Nat Neurosci 11:224 –231. required to produce accurate grasping. This reminds us that the Milner AD, Goodale MA (2006) The visual brain in action, 2 edition. Ox- posterior parietal cortex as part of the dorsal stream is not just a ford: Oxford UP. visuomotor system but a multimodal integration area (Jackson, Milner AD, Goodale MA (2008) Two visual systems reviewed. Neuropsy- chologia 46:774 –785. 2010). Milner AD, Perrett DI, Johnston RS, Benson PJ, Jordan TR, Heeley DW, Betucci D, Mortara F, Mutani R, Terazze E, Davidson DL (1991) Per- References ception and action in ‘visual form agnosia.’ Brain 114:405– 428. Bingham G, Coats R, Mon-Williams M (2007) Natural prehension in trials without haptic feedback but only when calibration is allowed. Neuropsy- Pisella L, Binkofski F, Lasek K, Toni I, Rossetti Y (2006) No double-dissociation chologia 45:288 –294. between optic ataxia and visual agnosia: multiple sub-streams for multiple Crawford JR, Garthwaite PH (2002) Investigation of the single case in neu- visuo-manual integrations. Neuropsychologia 44:2734 –2748. ropsychology: confidence limits on the abnormality of test scores and test Schenk T (2006) An allocentric rather than perceptual deficit in patient D.F. score differences. Neuropsychologia 40:1196 –1208. Nat Neurosci 9:1369 –1370. Crawford JR, Howell DC (1998) Comparing an individual’s test score Schenk T (2010) Visuomotor robustness is based on integration not segre- against norms derived from small samples. Clin Neuropsychol 12:482– gation. Vis Res 50:2627–2632. 486. Schenk T, McIntosh RD (2010) Do we have independent visual streams for Crawford JR, Garthwaite PH, Gray CD (2003) Wanted: fully operational perception and action? Cogn Neurosci 1:52– 63. definitions of dissociations in single-case studies. Cortex 39:357–370. Schenk T, Franz V, Bruno N (2011) Vision-for-perception and vision-for- Dijkerman HC, McIntosh RD, Schindler I, Nijboer TC, Milner AD (2009) action: which model is compatible with the available psychophysical and Choosing between alternative wrist postures: action planning needs per- neuropsychological data? Vis Res 51:812– 818. ception. Neuropsychologia 47:1476 –1482. Smeets JB, Brenner E (1999) A new view on grasping. Motor Control 3: Franz VH, Gegenfurtner KR (2008) Grasping visual illusions: consistent 237–271. data and no dissociation. Cogn Neuropsychol 25:920 –950. Smeets JB, Brenner E (2006) 10 years of illusions. J Exp Psychol Hum Per- Goodale MA, Milner AD (1992) Separate visual pathways for perception cept Perform 32:1501–1504. and action. Trends Neurosci 15:20 –25. Ungerleider LG, Mishkin M (1982) Two cortical visual systems. In: Analysis Goodale MA, Milner AD (2010) Two visual streams: interconnections do of visual behavior (Ingle DJ, Goodale MA, Mansfield RJ, eds). Cambridge, not imply duplication of function. Cogn Neurosci 1:65– 68. MA: MIT. Goodale MA, Milner AD, Jakobson LS, Carey DP (1991) A neurological Verhagen L, Dijkerman HC, Grol MJ, Toni I (2008) Perceptuo-motor in- dissociation between perceiving objects and grasping them. Nature teractions during prehension movements. J Neurosci 28:4726 – 4735. 349:154 –156. Westwood DA, Goodale MA (2011) Converging evidence for diverging Himmelbach M, Karnath HO (2005) Dorsal and ventral stream interaction: pathways: neuropsychology and psychophysics tell the same story. Vis Res contributions from optic ataxia. J Cogn Neurosci 17:632– 640. 51:804 – 811. Himmelbach M, Nau M, Zündorf I, Erb M, Perenin MT, Karnath HO (2009) Westwood DA, Danckert J, Servos P, Goodale MA (2002) Grasping two- Brain activation during immediate and delayed reaching in optic ataxia. dimensional images and three-dimensional objects in visual-form agno- Neuropsychologia 47:1508 –1517. sia. Exp Brain Res 144:262–267.