The neural mechanisms underlying visuomotor transformation are examined based on deficits in tracing and copying, as well as functional neuroimaging studies. The developmental process of copying and tracing, as well as lesion studies with adults showing disability in drawing, are reviewed, then two experiments are introduced. In Experiment 1, a behavioral analysis of copying and tracing by individuals with Williams Syndrome (WS) was presented. In Experiment 2, the brain activity involved in copying and tracing was measured in normal adults using functional magnetic resonance imaging. Based on these findings, we propose a model of visuomotor transformation to explain the neural basis of tracing and copying, as well as to provide a possible neural mechanism underlying the copying deficits and closing-in phenomenon observed in WS.
Figure-drawing behavior, like the language faculty, is one of the unique abilities of human beings. Since ancient times, people have conveyed messages and declared territorial demarcations by drawing figures and have developed relationships with each other through drawing. Figure drawing has enabled people to convert even abstract concepts into tangible forms and to share these concepts with others.
Drawing involves important cognitive functions. There are two types of drawing of visually presented models: tracing over the model in the same space, or copying the model onto a separate space. Although tracing and copying are essentially the same in terms of the accompanying motor output patterns, they differ with respect to the process of coordinate transformation by which the model should be represented. There are primarily two frames of reference for representing objects as well as our body: egocentric (self-centered) and allocentric (object-centered) coordinates. In tracing, the model and the drawing overlap in the same space, thus prompting the use of strategies that continuously minimize the distance between the model and the end effector (pencil or fingertip) in the local space, without considering the whole trajectory. In contrast, copying requires the reproduction of an egocentric representation of the model trajectory relative to the end effector, in a drawing space that is spatially separated from the model. In this case, a transformation from the visually presented model to the drawing space is needed. Therefore, compared with tracing, copying requires both egocentric as well as allocentric representations of the model.
During the human developmental process, construction as well as mutual transformation of different reference frames is considered to be of special importance. After birth, infants acquire their body schema, which is thought to develop by egocentric representation of the body through interactions between the body and the environment, elicited by actions. Then, for imitation of actions, the allocentric representation of others' action must be transformed into and compared with the egocentric representation of one's own movements. For children with autism, known to be impaired in social interaction, reversal errors are common in imitation (Williams, Whiten, Suddendorf, & Perrett, 2001), as are impairments in mirror-image imitation (Avikainen, Wohlschlager, Liuhanen, Hanninen, & Hari, 2003). These are considered to indicate deficits in the perspective transformation between self and others. Furthermore, Werner (1957) considered that a transition from ego-bound things-of-action to ego-distant objects-of-contemplation is an important developmental step for infants, and again this is also related to the ability for self-other discrimination. Taken together, the reciprocal transformation of egocentric and allocentric representation is essential in the development of social interactions. Because this function is also needed in drawing, as mentioned previously, our interest was to investigate copying and tracing abilities in order to develop a model of visuomotor transformation in terms of reference frames.
Development and disorders in drawing
Studies on the development of drawing abilities in normal children have postulated that there is a characteristic change at each age (Piaget & Inhelder, 1956); infants begin to scribble at approximately 15 months of age, and children are copying figures after approximately 3 years of age. However, according to Yamagata (2000), even infants show copying-like behavior (Figure 1), as infants at approximately the first year of life often mimic the drawing behavior of their parents or other people. Initially, infants just move their drawing arm randomly at this time, and the pen point touches on paper only accidently. Infants then begin to make requests of other people for drawings, that is, they encourage others to make models for copying. The important aspect is that the infant's behavior is accompanied by some form of utterance, for example, giving a pen to another person while saying “mamma … (Draw a picture of mamma)” or giving the name “mamma” to a picture drawn by the other person. After a model has been drawn for the infants, they try to copy the model. In other words, they show a kind of learning strategy even at this early age.
At approximately 14 months of age, infants will point to the space on which they desire other people to draw a picture, indicating that they use a self-reference system to draw other people's attention to their desired space, using joint attention. At approximately 15 months of age, infants not only make other people draw a picture, but will also trace the model while giving a name to the model. This behavior appears to be the beginning of tracing. At approximately 18 months of age, infants will give a name to their own drawing and find pleasure in showing it to other people. At approximately 20 months of age, infants try to draw on the specific region of a model drawn by another person, developing this into a more precise description and finally into complete copying.
The neural basis of copying and tracing can be studied by considering how brain damage affects the execution of these tasks. Tracing abilities are included in visuomotor coordination, sharply developing at age 3–4 years based on fully developed visual scanning, as mentioned above. Healthy adults can trace a figure precisely, whereas adult patients with brain damage at any region involving visuomotor coordination show poor performance on a tracing task. When brain damage includes regions involved in motor output and control, such as the motor/premotor cortex, cerebellum, basal ganglia, and projection fibers to the peripheral organs, patients cannot draw with any degree of skill. Brain lesions involving ocular movement and visual perception, such as neuronal nuclei in the brain stem, superior colliculus, pulvinar, frontal eye field, or visual pathway may also cause unskillful tracing. As the proprioception of the hand is also required for tracing, insufficient function of some thalamic nuclei may also affect tracing ability. Because the superior parietal lobules are involved in both visuomotor control and ocular movement, damage to these regions may cause reaching disturbance, which affects tracing ability. Consequently, tracing disabilities are observed very frequently in brain-damaged adult patients, with several etiologies.
Copying disabilities are also part of constructional disabilities, defined as an impaired ability to reproduce the same figure or design of a model in a different space from the model, due to visuospatial disabilities. Clinically, constructional disabilities are observed in block design and copying tasks. Constructional apraxia (CA), proposed by Kleist (1934), is precisely this type of symptom. Clinical studies have indicated that CA is often observed in patients with lesions primarily in the posterior parietal cortex (PPC) (Gainotti, 1985; Grossi & Trojano, 1999). Although Kleist originally proposed that a left parietal lesion may in particular cause CA, recent studies demonstrate that the same impairment may also be caused by right parietal lesions, but in a different way (Laeng, 2006).
This paper aims to investigate the brain mechanisms of visuomotor transformation, with respect to copying and tracing abilities. We used two different methods to approach this goal. In Experiment 1, we investigated several aspects of the drawing behavior of individuals with Williams syndrome (WS), a neurodevelopmental disease characterized by severely impaired visuospatial function. These patients have a particularly noticeable copying disability, but tracing ability is clearly preserved (Nagai, Iwata, Matsuoka, & Kato, 2001). The disorder is characterized by impairments in visuospatial processing as well as disinhibition (e.g. hypersociability), despite a relatively spared ability of other functions such as language. In addition, WS appears to be particularly well suited for exploring the neural basis of tracing and copying, for the following three reasons. First, as recent studies have demonstrated structural and functional abnormalities in the WS brain (Meyer-Lindenberg, Kohn, Mervis, Kippenhan, Olsen, Morris, & Berman, 2004), we could presume a neural mechanism for copying disability, which would differ from behavioral studies with normally developing children. Second, as the atypical cognitive system in this syndrome may be gradually built up during development, which differs from an acutely disrupted system in acquired brain damage, we can estimate the developmental order of the brain function related to tracing and copying. Finally, because WS is a genetically based disease, we may be able to link the neural basis of tracing and copying to some genes in the future. In Experiment 2, we use functional magnetic resonance imaging (fMRI) to investigate the brain regions involved in copying and tracing in normal adults. Based on our findings, a model of the brain mechanisms underlying visuomotor transformation is proposed to explain the neural basis of tracing and copying, as well as to provide a possible neural mechanism underlying the copying deficits and CIP observed in WS.
Experiment 1 Behavioral studies of figure drawing in Williams syndrome
Initially, Nagai et al. (2001) investigated copying, tracing, and drawing from memory in WS individuals, using simple geometric figures, including pentagons and hexagons. Individuals with WS could trace a model figure correctly, but were unable to copy or to draw from memory. There are some functional steps that are required for copying, but not for tracing: (a) feature detection of the target figure; (b) visuospatial working memory during drawing behavior; (c) decision about the drawing space; and (d) coordinate transformation of the target figure onto the new drawing space. As some of these steps appear to be disturbed in WS, we then implemented a copying-tracing task for seven individuals with WS, using a square, rhombus, regular pentagon, regular hexagon, or concave polygon as the target figure (Nagai, 2008).
We conducted the Copying in Real-space Task using a target figure-board vertically linked on the desk, with the drawing board (transparent or opaque) positioned in parallel and at a variable distance (Figure 2). When the distance was 0 cm with the transparent board, the task was just a tracing task; in contrast, when the distance was 10 cm with the opaque board, the task was similar to copying in real space. When the distance was 5 cm with the transparent board, the task was similar to both tracing and copying, depending on the strategy the participant preferred. In other words, if a participant focused on the target figure, a tracing task was performed through the transparent board; conversely, if the participant focused on their own hand movement rather than the target figure, a copying task was performed.
Under the condition that the board was opaque and the distance was 10 cm, no participant was able to draw any figure correctly except the square. Comparing the performance of two conditions (0 cm and 5 cm) with the transparent board, the number of participants who drew correctly at a distance of 5 cm was almost half that at the distance of 0 cm, for all figures except the square (Figure 3). This suggested that nearly half of the participants did not use a tracing strategy for the 5 cm condition, even though they could have used it. In other words, the tracing strategy is externally driven by the target (object), while the copying strategy is internally driven by a participant's motor plan (self). The latter was preferred by WS individuals, suggesting a biased preference for using self-centered representation or an inability to select the appropriate drawing strategy in response to a required task. One possible reason for drawing deficits in WS patients, as suggested by our results, is their impairment in selection or transformation processes between allocentric and egocentric reference frames.
As mentioned in the Introduction, the CIP is observed in both normal developing children and adult brain-damaged patients. Two hypotheses explain this phenomenon: one is a compensatory hypothesis whereby the CIP is viewed as compensatory behavior for insufficient visuospatial working memory; the other is an attraction hypothesis whereby the CIP is viewed as a primitive reaction towards objects that attract the viewer's attention. Recent studies supporting the compensatory hypothesis have investigated adult brain-damaged patients who have posterior lesions (Kwak, 2004; Kwon et al., 2002; Lee et al., 2004; Yamazaki et al., 2006). In contrast, studies supporting the attraction hypothesis that investigated both adult patients with Alzheimer's disease and normal developing children, have concluded that the nature of CIP induced by a dual task (e.g. line drawing and object naming) is the same in both adults and children (Ambron, Della Sala, & McIntosh, 2009; McIntosh et al., 2008). Moreover, fronto-temporal dementia, mainly involved in the fronto-temporal but not in the posterior region, has recently been reported to show CIP, with the same incidence in Alzheimer's disease (Ambron, Allaria, McIntosh, & Della Sala, 2009). Thus, the compensatory and attraction hypotheses focus on an insufficient parietal or frontal function, respectively. Given that CIP is usually observed in patients at the more severe stages of Alzheimer's disease, with lesions extending beyond the temporo-parietal to the frontal lobe, and in preschool children with immature executive function related to the dorsolateral prefrontal cortex, the attraction hypothesis appears to be based on insufficient frontal function.
To investigate whether individuals with WS, which is suspected to involve insufficient parietal function, are liable to show the CIP, we had 10 individuals with WS perform three types of copying tasks (Nagai, Inui, & Iwata, 2008) (Figure 4). In the Filling-in Copying Task, participants were presented with a target figure in the left half of the display and an incompletely copied figure in the right space. The participants were then required to fill in and complete the copy. In this task, individuals with WS showed significantly more overlap types of CIP on both target figures and incompletely copied figures. Fewer approach types of the CIP, such as gradually closing in the target figure during filling in the incompletely copied figures, were seen. The Line-by-line Copying Task presents a target figure line-by-line in the left half of the tablet PC display. The participant is then required to copy the figure by following the presented line in the right half of the display. Participants showed, conversely, significantly more of the approach types than the overlap types of the CIP in this task. In the Copying Task, where there is no visible drawing line, participants showed the least number of CIPs. Because the Filling-in Copying Task requires participants to fill in the blanks of an incomplete copy, without tracing already drawn lines, the inhibition of drawing known to be related to frontal function is necessary, which may cause the overlap type of the CIP. Conversely, the Line-by-line Copying Task required participants to copy lines presented successively into different spaces from those of the target figure. This required motor planning that involved making matches between the presented lines and their own drawing lines. Recent studies have shown that the parietal lobe is involved in these types of matching processes in mental manipulation tasks (Formisano, Linden, Di Salle, Trojano, Esposito, Sack, Grossi, Zanella, & Goebel, 2002; Sack, Camprodon, Pascual-Leone, & Goebel, 2005); thus, the approach type of the CIP appears to relate to parietal insufficiency.
In summary, our two behavioral studies of WS patients revealed that impairments in appropriate selection and transformation between different reference frames, as well as in matching between the presented lines and their own drawing lines in motor planning, are characteristic features underlying drawing deficits in WS. As we found that these functions are also important in healthy individuals, we next performed fMRI experiments in normal adults, to investigate the neural basis of their tracing and copying abilities.
Experiment 2 fMRI study of copying and tracing with normal adults
To investigate the brain areas both commonly and differentially involved in copying and tracing, we used fMRI to measure the brain activity during drawing by copying and tracing with normal adults (Ogawa & Inui, 2009). As stated previously, copying deficits are caused by lesions mainly in the PPC, but the functional roles of the PPC remain unclear. A spared ability to trace with an impaired ability to copy indicates deficits do not lie in low-level visuomotor processing, but in a coordinate transformation involving reproduction of an egocentric representation of model trajectory in the drawing space, which is spatially separated from the model. To avoid potential confounders due to differences in behavioral performances or due to eye movements, a memory-guided condition was introduced in addition to visual guidance. We hypothesized that if the PPC is involved in coordinate transformation, it should show increased activity in copying compared with tracing, irrespective of whether visual- or memory-based guidance was employed.
Participants comprised 28 volunteers (15 men, 13 women) with a mean age of 25 years (range, 20–33 years). A 1.5-T scanner (Shimadzu Magnex, Kyoto, Japan) was used to measure their brain activity. At trial onset, a curved line was displayed as a model on either the upper or lower half of the screen, in a pseudo-randomized manner. During this period, the participant passively observed the model for 3, 4 or 5 s. A mouse cursor was then displayed onscreen with an auditory cue, and the participant started to reproduce model patterns by drawing. A two-way factorial design was employed, with task (tracing (T), copying (C), and rest (R)) as one factor, and presence or absence of the onscreen model during the drawing period (visual (V) and memory (M) guidance) as the other factor, resulting in four drawing and two visual control conditions. First, in the visually guided tracing (VT) trial, the cursor was displayed to the left or right side of the model curve, and the participant traced over the model with the cursor. Second, the visually guided copying (VC) trial was similar to the VT trial, except that the cursor was displayed on the other half of the screen (either above or below) to where the model was displayed, and the participant then copied the model with the cursor at this different location. Third, in the memory-guided tracing (MT) trial, the model disappeared as soon as the mouse cursor appeared at the same location at which the model had previously been displayed, and the participant reproduced the now-absent model from memory using the cursor. Finally, the memory-guided copying (MC) trial was similar to the MT trial, except that the cursor was displayed at a different location to where the model had been located, and the participant then reproduced the model with the cursor in this separate space. The drawing period lasted 5 s, and the participant was instructed to draw the model as accurately as possible, but at sufficient speed to complete the drawing within the given period. The drawing trajectory of the cursor was displayed on the screen as a solid line. In addition, two rest conditions were introduced: the cursor image did not appear onscreen, and the participant did not produce any movement with (visual rest (VR)) or without (memory rest (MR)) a visible model on the screen. These six conditions were pseudo-randomly interleaved within a session. Figure 5 shows representative examples of the screen images used in our experiments.
The result of the whole-brain analysis revealed significantly increased activation in the parieto-premotor and mesial motor areas, as well as in the occipital cortex, for copying compared with tracing. We subsequently conducted a region of interest (ROI) analysis, defining these activated areas as ROIs, and calculated the percentage signal changes, in order to compare their activation patterns in detail. Only the bilateral intraparietal sulcus (IPS) showed significantly increased activations for copying versus tracing under both visual and memory guidance (Figure 6). The IPS therefore appeared to be involved in coordinate transformation for copying, independent of visual or memory guidance. The observed IPS activation is located close to the region identified as the human medial IPS (mIPS or MIP) that has been related to visually guided reaching (Grefkes, Ritzl, Zilles, & Fink, 2004). Although all drawing conditions in our experiment involved visuomotor transformation from model vision to cursor movement, the difference between copying and tracing is that the drawing space is separated from the model space in copying, thus requiring additional transformation from the model to the drawing space. Our results indicate that the IPS is involved in the production of anegocentric representation of a model in a drawing space that is spatially decoupled from the model.
We also observed significantly increased activation of the bilateral ventral premotor area (PMv) in the VC, MT, and MC trials compared with the VT trial, without significant differences between the VC, MT, and MC trials. Participants traced over the visible model as a drawing trajectory in the VT trial, which enabled feedback control by comparing the model and the drawing trajectory in the same space. Participants therefore could consider only local error minimization in the VT trial, without planning the entire drawing trajectory, as was required for the other conditions. In contrast, participants had to produce a drawing trajectory on a blank space for the other conditions. Our results thus indicate the involvement of the PMv in motor planning of a drawing trajectory. Behavioral data also showed significant decreases in both drawing speed and error under the VT condition compared with other conditions. This indicated that the VT trial involved feedback control based on continuous comparisons between the model and the trajectory, resulting in slower movement and smaller errors, whereas other conditions relied mostly on feed-forward motor planning.
Brain mechanisms of visuomotor transformation in tracing and copying
Brain mechanisms of copying and tracing with their neural substrates
Based on prior findings, we propose brain mechanisms of visuomotor transformation to explain the neural basis of tracing and copying regarding frames of reference, as well as to provide a possible neural mechanism underlying clinically observed copying deficits and the CIP.
To reiterate, there are two frames of reference for representing objects as well as the body: egocentric (self-centered) and allocentric (object-centered) coordinates. Previous reports have revealed a relationship between reference frames and parietal laterality. Functional neuroimaging studies using PET have indicated a laterality in parietal activity of egocentric and allocentric imagery of action: the left for first-person, and the right for third-person perspectives (Ruby & Decety, 2001), together with the reciprocal imitation tasks (Decety, Chaminade, Grezes, & Meltzoff, 2002). The fMRI studies also revealed that the right and left PPCs are predominantly involved in mental rotation of object and self motions (Creem, Downs, Wraga, Harrington, Proffitt, & Downs, 2001; Zacks, Vettel, & Michelon, 2003). Our previous fMRI experiment also showed right and left PPC activity for predicting externally versus self-generated movements (Ogawa & Inui, 2007). Based on these prior findings, Inui and Ogawa (2009) proposed that egocentric and allocentric coordinates are dominantly represented in the left and right posterior parietal cortex, respectively. In the present paper, we considered brain mechanisms of tracing and copying in the view of frames of reference.
Figure 7 shows a schematic diagram of our model. In tracing, the model to be drawn and the drawing trajectory both overlap, and participants can thus use strategies to minimize the distance between the model and the end effector in the local space. In other words, tracing could be used mostly to represent the model to be drawn in object-centered coordinates and then drawn with a feedback control strategy: that is to say, tracing could be implemented by estimating the error called the difference vector (DV), by subtracting a target position vector (TPV) with a perceived position vector (PPV) (Bullock, Bongers, Lankhorst, & Beek, 1999). These variables, PPV and DV, are both apparently encoded in the tonic and phasic type neurons in the PPC (area 5), respectively, as their firing patterns are quite similar to the predictions of a computational model of drawing (Grossberg & Paine, 2000). In addition, our previous fMRI experiments of tracing indicated that the differences between the actual and predicted cursor positions are evaluated in the anterior part of the right IPS (aIPS) (Ogawa et al., 2006; Ogawa, Inui, & Sugio, 2007). These neural circuits (depicted on the right side of Figure 7) are viewed as important for successful tracing, which mainly depends on the visual feedback control strategy.
In contrast, a number of components are needed for copying that are not required for tracing. First, copying requires the participant to produce an egocentric representation of the model trajectory relative to the end effector, in a drawing space spatially separated from the model, so that transformation from model to egocentric representation is needed, as depicted on the left side of Figure 7. From our previous experiment (Ogawa & Inui, 2009), these functions of coordinate transformation from object- to self-centered coordinates are mainly located in the medial part of the IPS (mIPS). Neurophysiological studies have showed that mIPS (MIP) neurons are involved in visuomotor transformation during reaching, at the intermediate stage of transforming spatial coordinates of the target into limb-centered coordinates (Cohen & Andersen, 2002). Second, there is a greater requirement for storage of the model into short-term memory for copying compared with tracing, and this is possibly subserved by pre-SMA activity. Third, the PMv observed while drawing on a blank space is related to the planning and generation of a drawing trajectory. A similar role of the premotor area in movement trajectory planning has also been indicated in studies on nonhuman primates (Hocherman & Wise, 1991; Moll & Kuypers, 1977). Neurophysiological studies have also shown that PMv neurons have projections from the IPS (Matelli & Luppino, 2001) and these encode visual targets in effector-centered coordinates (Graziano, Yap, & Gross, 1994). Together with these findings, our model assumes that the PMv plays a role in trajectory planning that is based on coordinate transformation processing in the parietal cortex. Finally, the basal ganglia (BG) are involved in the commanding of GO-signals for starting movements and/or in the determination of drawing size (Grossberg & Paine, 2000). Clinical studies have also reported that patients with forms of Parkinson's disease with BG damage often show reduced pen-stroke size when drawing, known as “micrographia”. This occurs particularly when drawing without external visual cues (Oliveira, Gurd, Nixon, Marshall, & Passingham, 1997), indicating the role of the BG in determining drawing size for internally guided drawing. These areas, including the pre-SMA, PMv, and BG, are depicted in the center part of Figure 7.
Furthermore, tracing that relies only on feedback control cannot generate rapid drawing. Therefore, predictive motor control is also needed. In addition, the drawn trajectory must be compared with the original target, in order to monitor the success of copying. The PPC is also involved in the internal estimation or prediction of self-generated movements, as well as in the comparison of the prediction with the target, as revealed by our previous fMRI studies (Ogawa & Inui, 2007; Ogawa et al., 2006).
Although there have been some models proposed for drawing or handwriting movements (Grossberg & Paine, 2000), these model have focused primarily on tracing on models. Therefore, they cannot explain the dissociation between tracing and copying observed clinically in CA or WS patients. In contrast, our model proposes the operation of partially overlapped, but distinct, neural circuits underlying tracing and copying regarding differences in reference frames. Based on this model, an explanation of selective deficits in copying with spared tracing, as well as the CIP, is presented in the following section.
Neural mechanisms of copying deficits and the CIP
As a final consideration, we explore the possible neural mechanisms for copying deficits as well as for the CIP observed in people with CA or WS. The two types of abnormalities in the brain, structural and functional, can be measured using structural MRI, such as voxel-based morphometry (VBM) and fMRI, respectively. WS patients show a clear structural abnormality in the PPC, mainly in the V6/V6A region, located near the parieto-occipital sulcus (Meyer-Lindenberg et al., 2004). This region, involved in the construction of egocentric coordinates of space, is located posteriorly within the distributed network responsible for visually guided reaching and for eye-hand coordination. This network has a mutual connection with the premotor areas (Caminiti, Genovesio, Marconi, Mayer, Onorati, Ferraina, Mitsuda, Giannetti, Squatrito, Maioli, & Molinari, 1999). Furthermore, functional abnormality (hypo-activation) was also observed in the IPS in WS patients (Meyer-Lindenberg et al., 2004). Because a direct connection from V6A to mIPS (MIP) has been observed in monkeys (Rizzolatti & Matelli, 2003), the functional abnormality in mIPS may be caused by a structural dysfunction in V6A. These findings are in agreement with our model, as well as with the observed deficits in copying in individuals with WS.
Furthermore, we included the superior colliculus (SC) in our model, depicted in the lower left part of Figure 7, to explain the CIP. We consider that the CIP is caused by the dysfunction of inhibition over the tendency to make hands close to the visible target. This is subserved by the subcortical region, possibly the SC, involved in eye-hand coordination (Stuphorn, Bauswein, & Hoffmann, 2000). We also consider the possibility that cortical inhibition over the SC comes from the premotor cortex, because a monkey with a premotor lesion will try to reach straight out to a visible target, having lost the ability to plan complex movement trajectories (Moll & Kuypers, 1977).
In summary, the brain mechanisms of the visuomotor transformation involved in tracing and copying were proposed in this paper. With this model, we have explained the neural basis of tracing and copying regarding frames of reference, and have provided a possible neural mechanism underlying clinically observed copying deficits and the CIP.