Pergamon PII: S0028 3932(97)0(1(10- \cuHJp,vvclz��doqio.\%l ~5. N o ~),pp 12gtJ 1310, 19L)7 I [997 Fl',cxici Science 1 Id. All iighP, rc',elvcd l'rintcd m ( h e a l Britain O02S 3c1"~2 ~)- S]7,00 ~ (I.00 Visual motion perception after brain damage: II. Deficits in form-from-motion perception THOMAS SCHENK* and JOSEF ZIHL) *Neurologische Forschung, Klinikum Groghadern, Ludwig-Maximilians-UniversitLit, Marchioninistrasse 23. I)-81377 Munich, Germany and tMax Planck Institute for Psychiatry. Kraepelinstrassc 10. D-80804 Munich. Gcrnlany (Receit'ed 9 Mar 1996 accepted 26 ~'metnh~ r 1996) Abstract We investigated form-from-motion perception (FFM perception) in a sample of 39 patients with acquired brain damage. Pronounced FFM deficits were found in two patients (FM1 and FM2) ,aith biparietal lesions. Both patients were able to identify the relevant figure, when it was not embedded in obstructive texture. Moreover, the3 could localize the figures in the FFM condition, although they could not reliably identify them. The two patients had normal motion coherence thresholds. Their perfommnce in a static figure ground task did not differ flom that of other patients. These findings imply that the FFM deficits are not caused by impairment of basic visual motion or form perception but are the consequence of damage to a parietal brain structure revolved in the combined analysis of visual motion and form information. The nature and functional role of this brain structure as well as the implications of our results for models of FFM perception are discussed, i 1997 Elsevier Science Lid Key Words: akinetopsia: attention biological motion perception: figure ground perception: V5: MT. Introduction V5 is widely regarded as the main cortical constituent of the visual m o t i o n system in primate visual cortex [30, 34]. However, since V5 contains only a few orientation- and no form-selective cells, it seems unsuited for m o t i o n pet- ception involving forna analysis [32]. A n example of this kind o1" motion perception is the perception of form- f r o m - m o t i o n ( F F M ) . In F F M , the visual system uses motion cues to segregate a figure from the b a c k g r o u n d [2, 5]. V5 may play a role in F F M perception in particular, it may be involved in detecting and localizing the m o t i o n b o u n d a r i e s of a two-dimensional figure. However, for the integration of these b o u n d a r i e s into a coherent c o n t o u r that can be identified later as a particular lk)rm, some kind of interaction between a motion-extracting mech- anism (c.g. VS) and a neural mechanism suited for form analysis seems necessary. Two different mechanisms are * To whom all correspondence should be addressed: liix: 0049-89-70906- l 0 l. :~ We haxe chosen V3 and V4 as two possible candidates for areas inxolved in form analysis. Other areas might be involved as well or might be even more important for this perceptual function. Hox,~e',er, as long as it may be assurned that there are different areas l\~r motion and form analysis, the actual areas ilwolved are of no particular importance for the purpose of our a r g u n l e n t . suggested to facilitate this interaction: a distributed solu- tion and a local solution. The local sohition postulates the existence of a cortical brain region m which neural mechanisms For the analysis of visual form and motion are combined. D a m a g e to this area might result in a selective deficit in F F M perception. In the distributed solution, it is assumed that the integration of visual m o t i o n and form information is achieved by a dynamic interaction between different areas that are specialized for either motion (e.g. V5) or form analysis (e,g. V3 V4).:] The local and distributed solutions predict different effects of cortical lesions. A c c o r d i n g to the distributed solution, cortical lesions may impair F F M perception by d a m a g e either to the m o t i o n area V5 or to I'ornl-relewint areas, e.g. V3 or V4. Thus, the distributed solution pre- dicts that an FE'M deficit will not occur in isolation, but only in c o m b i n a t i o n with other deficit, in visual motion or l\)rm perception. The existence of patients with an isolated deficit in F F M perception alter a cortical lesion would constitute evidence for the presence of an area that is, in some x~.ay, specialized for F F M stimuli. ] h i s would support the local solution. Two recent neuropsychological studies reported on patients with deficits in F F M perception [19, 28]. However, both litiled to show a clear dissociation between perlk~rmance in an F F M task vs static form (a function of I\~rm-relevant areas, possibly V3 arid V4) and global 1299

1300 T. Schenk and J. Zihl/Deficits in form-from-motion perception visual motion perception (a V5 function). In the study by Vaina [28], neither form nor figure ground perception was tested for static stimuli. Regan et al. [19] used a visual motion paradigm (full-coherence motion, measuring vel- ocity thresholds), whose potential to detect V5 damage had not been established in prior studies on V5-1esioned animals or patients with impaired visual motion percep- tion. F F M perception was tested in a group of 39 brain- damaged patients. Form, static figure ground, and visual motion perception were also tested. As a measure of visual motion perception, motion coherence thresholds were determined. Several studies have shown that motion coherence thresholds are a sensitive measure of damage to V5 [3, 4, 12, 21]. Two patients with biparietal lesions showed evidence of a clear dissociation between per- formance in F F M perception and performance in all other tests. Experiment 1 In Experiment l, figures walking in one of four possible directions (to the right, to the left, forward, or backward) were presented. This type of stimulus was introduced into vision research by Johansson [8]. The figures were generated on a computer, recorded on video, and dis- played on a TV screen. Half of the figures simulated a walking person (human figure), and the other half were scrambled versions of human figures (scrambled figures). The figures were presented under three different con- ditions: the background was either dark (Experiment 1A), filled with a static (Experiment 1B), or dynamic random-dot pattern (Experiment IC). Under all three conditions, subjects were asked to identity the presented figure. In Experiments 1B and 1 C, they were also required to point to where they assumed the figure to be. Experi- ment I A served as a screening test. Subjects who failed this test were excluded from all following experiments. Experiments 1B and 1C were F F M tasks. In both experi- ments, motion cues had to be used to separate the figure from the background. In Experiment I B, the relevant cue was simply the presence or absence of motion. In Experiment I C, the relevant cue was the specific kind of motion. We questioned whether different results would be obtained under the two different F F M conditions. Two patients (see below) took part in an additional experiment (Experiment I D). In Experiment 1D, only walking h u m a n figures were presented. However, in con- trast to Experiments 1A 1C, where the figures never changed position, the figures in this experiment were slowly shifted either to the left, to the right, forward or backward. Subjects perceived these positional shifts as slow motion. The forward and backward movements were actually simulated by increasing or decreasing, respectively, the size of the figure. The positional shift occurred either in the same direction as the walking motion (compatible condition, e.g. walking to the right, positional shift to the right) or in the opposite direction (incompatible condition, e.g. walking to the right, pos- itional shift to the left). The subjects' task was to decide whether the walking motion and the positional shift were compatible or incompatible. Method Subjects. Most of the patients (31 patients) had already taken part in our study on global motion perception [21]. However, this time, patients with symmetrical bilateral lesions and those with complete hemianopic deficits were also admitted to the study otherwise, the patients had to fulfil the same criteria as in our first study [21]. We tested 26 male and 13 female patients, whose age ranged between 30 and 82 years (mean: 56 years). The lesions of all 39 patients were documented by CT or MRI scans (for details on the lesion analysis, see [21]). Seventeen patients had left-hemisphere lesions, 18 had right-hemisphere lesions and five patients had bilateral lesions. The lesions had been sustained from 160 to 2 weeks prior to examination (mean: 14 weeks). Six patients suffered from an incomplete hom- onymous visual field deficit, and two patients had a complete or almost complete homonymous hemianopsia. The patients" performance in tasks testing visual exploratory behaviour was unimpaired. None of the patients showed any signs of a restric- ted attentional field or a Balint-Holmes syndrome. Twenty- seven patients suffered from mild attention deficits, as measured with the ZVT (Zahlenverbindungstest: this test measures the time required to connect a set of spatially distributed numbers with a pencil [13]). Other cognitive deficits (mnestic, language or problem-solving deficits) were found in 18 patients. However, since it may be assumed that these deficits did not interfere with the patients' performance in our experiments, we will not describe these deficits in detail here. Further details on the 39 patients can be found in Table 1. Seven normal male and eight normal female subjects, aged between 27 and 67 years (mean: 53 years), served as controls. Material and procedure. Stimuli were stored on video and presented on a TV screen (diameter: 64cm). The stimuli are illustrated in Fig. 1. Subjects sat 2.5 m from the front of the screen. The size of the presented figures varied between 3.2 and 6.8' in height and 1.1 and 2.3' in width. The figures were presented at random positions within a square-shaped frame (length =9.8). The limbs of the figures moved in a walking-like fashion: however, their position remained the same throughout the presentation in subtests A C. The random-dot background in subtests B and C contained 100 dots distributed across the screen at a density of 0.05 dots per cm 2. Each stimulus pres- entation lasted for 5 sec. Each subtest began with 15 practice trials, followed by 25 experimental trials. Fifty per cent of the trials contained the target [e.g. a human figure (subtests A C), compatible condition (subtest D)], and 50% contained the distractor [e.g. scrambled figure (subtests A C), incompatible condition (subtest D)]. The trials of the subtests were arranged in a random sequence. Subjects had to indicate whether the trial contained a target or a distractor. This task was called the identification task. In subtests B and C, subjects were addition- ally required to indicate the position of the figure by using a light pointer (localization task). The subjects' responses were recorded by the examiner. Earlier studies on FFM perception used geometric forms or letters [2, 5, 19]. By using more complex forms (i.e. human figures and their scrambled versions), we increased the difficulty of the task, thereby hoping to raise the sensitivity of this test. Statistical analysis was done by computing separately the 95% confidence interval for the results of the control and the patient sample. The performance of a subject (measured in number of