Imitation: A direct route from vision to action?
Imitation: A direct route from vision to action?
Abstract and Keywords
Liepmann’s idea that imitation probes the transposition of a correct image of the intended action into appropriate motor command has been falsified by the observation of patients who fail imitation of meaningless gestures but can produce meaningful gestures both on command and in imitation. The alternative position, that there is a specialized direct route for imitation whose interruption causes a selective impairment of imitation, has gained plausibility by the detection of “mirror neurons” unifying vision and motor execution of gestures. The directness of the route from vision to motor execution is contradicted by experiments showing the connection to be malleable by contextual cues and by the observation that apraxic patients who fail imitation on their own body have similar problems when trying to replicate the body configuration on a manikin, although the motor actions of manipulating a manikin are fundamentally different from those of execution the gestures.
We start our survey of the evidence with imitation of gestures. Imitation of gestures was introduced into the clinical examination of apraxia by Liepmann (see Chapter 2). To Liepmann, disturbed imitation was important because imitation is essentially a non-verbal test. Once the general instruction is understood, the command to imitate a particular gesture is given without words by demonstration of that gesture. Since disturbed imitation cannot be attributed to deficient verbal comprehension it underlines the independence of apraxia from aphasia. To be sure that performance is not influenced by paresis or other elemental motor deficits patients are asked to imitate with the hand ipsilateral to their lesion. In accordance with the natural tendency to use the hand on the same side of space as the demonstration rather than the nominally same hand (see Chapter 3), the examiner demonstrates “like a mirror,” using the right hand when the patient uses the left and vice versa.
Healthy persons imitate meaningless gestures like those shown in Figure 6.1 swiftly and virtually errorless. By contrast, some patients encounter massive difficulties. Their hand approaches the face but ends up in a position that deviates grossly from the demonstrated target posture. The path of the hand may consist of swift and secure movements leading to the wrong position or, more frequently, of hesitating and searching movement which sometimes ends up in the correct target position but more frequently misses it (Hermsdörfer et al., 1996).
Imitation and the posterior to anterior stream of action control
The importance given by Liepmann to disturbed imitation was not exhausted by the demonstration of the independency of apraxia from aphasia. Remember that in his model of the posterior to anterior stream of action control mental images of intended actions are transformed into motor commands, and that disturbances can affect either the generation of the mental image or its translation into motor command. Liepmann took for granted that the demonstration of the gesture by the examiner secured the patient’s correct mental image of the gesture. Consequently, errors in imitation testified that “the guidance of the left hand by mental images of the shape of the movement is disturbed” (Liepmann, 1908).
(p.89) The assumption that defective imitation indicates defective execution of correctly conceived gestures still prevails in modern literature. For example, the Italian neurologist Ennio de Renzi wrote:
It must be recognized, that it is not always easy to decide whether the patient has the representation of the gesture clear in his/her mind, . . . and this is one of the reasons why imitation tests are often preferable to verbal commands in testing ideomotor apraxia. Since the examiner provides the model of the action and the patient must simply copy it, errors can only be due to an executive deficit. (De Renzi, 1990, p. 246)
Liepmann had proposed that problems with the translation of mental images into motor acts would come to the fore most conspicuously when movement are made wholly from memory rather than resulting from interactions with external objects. This definition applies not only to imitation of meaningless hand postures but also to the production of communicative gestures. The conversion of mental images into motor commands constitutes a common final path for all gestures regardless of whether the mental image preceding them is generated from the examiner’s demonstration for imitation, from knowledge (p.90) about the use of tools for pantomime of tool use, or from knowledge about the conventional shape of emblematic gestures. In the theoretical framework of the posterior to anterior stream of motor control, imitation is supposed to probe the integrity of the final portion of the path regardless of the nature of its origin. This leads to the prediction that patients who fail imitation should also fail the production of communicative gestures on command.
Clinical evidence contradicts this prediction. There are patients who have severe problems with the imitation of gestures but can produce pantomimes of tool use and (p.91) emblematic gestures flawlessly. This condition has been termed “visuo-imitative apraxia” (Mehler, 1987; Goldenberg & Hagmann, 1997; Cubelli et al., 2000; Peigneux et al., 2000; Bartolo et al., 2001). The dissociation between defective imitation and preserved production on command can be quite impressive. For example, the request to imitate the hand posture of touching the temple with the tip of the middle finger of the horizontally extended flat hand may lead to searching movements that finally result in a touch of the forehead with the thumb of the vertically aligned hand, while the request to show a military salute is promptly answered by the same gesture that the patient was unable to produce in imitation.
Two routes for imitation
The gestures shown in Figure 6.1 are meaningless. Thus, the dissociation between their defective imitation and the preserved production of gestures on verbal command is at the same time dissociation between defective production of meaningless, and preserved production of meaningful, communicative gestures. This ambiguity can be resolved by asking patients to imitate meaningful gestures. Indeed, the imitation of meaningful gestures can be perfectly preserved in patients with visuo-imitative apraxia (Goldenberg & Hagmann, 1997; Peigneux et al., 2000; Bartolo et al., 2001). Apparently the problem in visuo-imitative apraxia is not a general defect of imitation but a specific defect of imitation of meaningless gestures.1
Figure 6.3 shows a model which can account for visuo-imitative apraxia (Rothi et al., 1997). It has its roots in Geschwind’s (Geschwind 1975; Geschwind & Damasio, 1985) reinterpretation of the posterior to anterior stream of action control. Geschwind had postulated that defective imitation is due to destruction of a storehouse of learned motor skills. Because meaningless gestures are generally novel, Geschwind explicitly excluded imitation of meaningless gestures from the realm of apraxia (see Chapter 4). Rothi’s model proposes two significant amendments to Geschwind’s schema. First, the “storehouse of learned motor skills” is conceived as an “action lexicon” storing the shape of gestures, and this lexicon is duplicated resulting in an action input and an action output lexicon. Both are connected to a central semantic memory storing knowledge related to the meaning of the gestures. The action input lexicon mediates understanding of the meaning of perceived gestures and the action output lexicon their production for expression of meaning. Imitation of familiar meaningful gestures is accomplished by transfer of the gesture from the (p.92) input to the output lexicon. The transfer can pass via the direct connection between the lexica or take the detour via their meaning stored in the central semantic memory.
The second amendment is the addition of a direct route leading from visual analysis of gestures to the innervatory patterns of their motor replication and bypassing storage of the shape and meaning of familiar gesture in lexicons and semantic memory. This route can accommodate imitation of meaningless gestures. Its interruption would interfere with imitation of meaningless gestures but spare the imitation of meaningful gestures via the lexicons, and thus bring forward visuo-imitative apraxia.
Imitation and mirror neurons
Thus far, we can conclude that the route from visual perception to motor replication of gestures can be direct insofar as it bypasses recognition of the meaning of the gesture. It is another question whether the route is constituted by a direct connection from vision to motor control or includes interpolated stages where the visual information is translated into a format compatible with the direction of motor actions. The idea that the (p.93) connection from vision to motor execution of gestures is direct also in the sense that it includes no interpolated stages has become very popular in recent years. The popularity was nourished by the detection of “mirror neurons” in monkey premotor cortex (Di Pellegrino et al., 1992; Rizzolatti et al., 2002; Rizzolatti & Sinigaglia, 2008). These neurons are similarly active when the monkey sees another monkey or a human perform an action and when it performs the same actions. The conclusion that they provide a substrate mirroring observed actions by their motor replication is tempting but not straightforward. Indeed, the monkeys in whom mirror neurons had been detected do not imitate motor actions (Subiaul et al., 2004). Furthermore, whereas the putative human direct route accommodates novel and meaningless gestures, mirror neurons in monkey react only to biologically meaningful and familiar actions (Ferrari et al., 2005). Nonetheless, the attractiveness of mirror neurons has encouraged the development of theories as to how imitation could be built upon a direct link from vision to motor execution of gestures.
One such account posits that the basis for imitation is laid by children’s observations of their own movements. The co-occurrence of the motor action and its visual perception increases the strength of associations between neurons mediating the visual perception of an action and neurons directing its motor execution (Keysers & Perrett, 2004; Brass & Heyes, 2005). Because the view of oneself performing an action is similar to that of another person doing the same, the association between visual perception and motor execution of actions generalizes to the perception of other person’s actions. Seeing the action re-activates the association between vision and motor control and elicits motor execution of the perceived action, that is, imitation. The motor neurons that become activated by vision of the action and direct its motor replication become “mirror neurons.”
The association between perception and execution of actions lends itself to being activated automatically. Vision of the action elicits activation of mirror neurons by simple association regardless of whether subjects are intending to imitate the action or not. Indeed, automatic activation of motor cortex by mere observation of actions has been brought forward as evidence for the existence of mirror neurons in human (e.g., Fadiga et al., 1995; Buccino et al., 2001; Aziz-Zadeh et al., 2002; Watkins et al., 2003; Clark et al., 2004; Ehrsson et al., 2006). This automatic motor activation was somatotopic. Observation of a movement activated precisely those parts of the motor cortex which control the moving body part. The parallel to the historical theories postulating an influence of the somatotopic layout of motor cortex on body part specificity of apraxia is striking. We will come back to body part specificity of imitation disturbances in Chapter 7.
Rothi’s model (see Figure 6.3) predicts selective impairment of imitation of meaningless gestures when the direct route is interrupted but not necessarily the converse dissociation of selective impairment of imitation of meaningful gestures when the lexical route is disturbed, because failure of the lexical route could be compensated by the direct route treating the meaningful gestures as if they were meaningless and mediating their imitation by (p.94) connecting their visual analysis to their motor execution. There are, however, a few reports of patients who could imitate meaningless gestures but failed the imitation of meaningful gestures (Bartolo et al., 2001; Tessari et al., 2007). A possible explanation of this unexpected dissociation is the exertion of strategic control on the use of either the direct or the indirect route. Particularly when meaningful and meaningless gestures are presented for imitation in a mixed sequence, subjects prefer to treat both kinds of gestures as meaningless and copy their shape via the direct route without paying attention to the meaning and familiarity of the meaningful gestures. This strategic choice is advantageous in that it takes away the need to select the appropriate strategy for each gesture individually, but it veils the possibility of selective disturbance or selective preservation of only the lexical route. When, however, meaningful and meaningless gestures are presented for imitation in separate blocks, subjects are inclined to select the appropriate route for each block and to adhere to their choice through the whole block even when single items give rise to difficulties. This adherence to the strategic choice prevents compensation of insufficient imitation via the lexical route by application of the direct route and may thus reveal selective deficits of imitating meaningful gestures (Tessari & Rumiati, 2004; Tessari et al., 2007).
The conclusion that subjects can choose whether or not they employ the direct route for imitation of meaningful gestures cast doubts on its automatic activation by mere vision of gestures.
The goals of imitation
Further evidence that the route from perception to replication of gestures can be modulated by strategic choices was provided by developmental studies (Bekkering et al., 2000; Gattis et al., 2002; Bekkering et al., 2005). They revived the historical “Hand, Eyes, and Ear” Test (Head, 1920; Gordon, 1922; see Chapter 3). Children were seated opposite the experimenter who demonstrated touching of either the left or the right ear with either the left or the right hand. The children were instructed to imitate the experimenter’s actions like in a mirror. Children always reached to the correct ear but frequently replaced contralateral by ipsilateral reaching, that is, they reached for the right ear with the right hand although the experimenter had touched the spatially opposite left ear with her right hand, and vice versa. When, however, the examiner touched both of her ears simultaneously the children imitated correctly not only when each hand reached for its ipsilateral ear, but also when the hands were crossed for reaching the contralateral ears (see Figure 6.4). The authors speculated that the children had established a hierarchy of goals for the imitation, and that the laterality of the targets of reaching, that is, the ears, were on top of this hierarchy. The laterality of the instruments of reaching, that is, the hands, were lower valued and not protected from being neglected. Thus the unfamiliar and motorically somewhat cumbersome reach of the contralateral hand could be replaced by the more comfortable and faster reaching of the ipsilateral hand (see Schofield (1976) for evidence that reaching across the midline is difficult for children). In the bimanual condition the target of reaching remained constant, as both ears were touched in all trials. Consequently, attention could be withdrawn from the laterality of the touched ears and directed to the position of the arms. The choice of the (p.95) hand now ascended to the top of the goal hierarchy and the children imitated faithfully not only ipsilateral, but also contralateral reaching.
In a further experiment the children were asked to imitate reaching to a spot on the table located either on their right or the left side with either the right or the left hand. There were two conditions: in one the target positions were marked by black dots whereas in the other there was no visible mark and the target of pointing was indicated by the examiner’s demonstration. Children virtually never failed to touch the correct side of the table, but in the dot condition they frequently used the hand ipsilateral to the dot although the experimenter had demonstrated pointing with the contralateral hand. This error occurred only very rarely when there were no visible dots. The authors speculated that in the dot condition the children had assigned the highest position in the goal hierarchy to reaching the correct dots and neglected the alternations between crossed and uncrossed pointing. When, however, there were no visible dots, the type of movement became the dominant goal and crossing or uncrossing of the hand was replicated correctly. The authors concluded that their results speak:
against the notion that imitation involves a direct mapping of a non-composed action pattern. Imitation involves more than a direct mapping between perceptual input and motoric output. Observed behaviours are coded as constituent goals and those goals are the units from which subsequent action patterns are composed. (Gattis et al., 2002, p. 201)
(p.96) In a similar line of research, Franz and colleagues (2007) manipulated the habitual preference of normal subjects for imitation like in a mirror over imitation with the nominally same hand. The preference reversed in favor of imitation with the nominally same hand when salient markers were affixed to the hand used by the experimenter for demonstration and to the nominally same hand of the subject. Apparently the similarity between demonstrated and imitated gesture can be given different specifications. Contextual cues can induce a change of the dominant specification. The malleability of the mapping from perception to execution of actions is difficult to reconcile with the automatic activation of a direct route from vision to motor control.
How direct is the direct route?
The direct routes in Rothi’s model and in the “mirror neuron” account share the seemingly self-evident assumption that the motor act of imitation is the same as the motor act of the demonstration. If the model touches the temple with extended fingers, the imitating person also touches their temple with extended fingers. However, imitation of gestures can be seen differently. It can be classified as one particular instance of the transposition of a body configuration from one body to another.2 In imitation the configuration is transposed to the own body of the person performing the transposition, but this correspondence can be torn apart. For example, the person can be asked to replicate the demonstrated body configuration on a manikin instead of their own body. The manipulation of the manikin is a motor act too, but its features are substantially different from those of the movements and body configurations that have been demonstrated for replication. In order to approach the extended fingers of the manikin to its temple, the person manipulating the manikin must neither touch their own nor the manikin’s temple, and must clench their hand around the hand or the wrist of the manikin rather than extending it.
I have conducted a clinical study based on this paradigm (Goldenberg, 1995). Imitation of hand postures was tested in two conditions. First, the examiner sat opposite the subject and demonstrated the postures and the subject was asked to repeat them. Then, the examiner sat beside a life-sized wooden manikin (without lower body) whose arm and hand could be moved like those of a human. He demonstrated the hand postures again, and the subject was asked to replicate the posture with the manikin’s hand. There were three groups of subjects: patients with left brain damage and aphasia, patients with right brain damage, and healthy controls. Patients imitated with the hand ipsilateral to the lesion and manipulated the same hand of the manikin. Half of the controls used the left and half the right hand, but results did not differ between them.
The results were quite straightforward. Healthy controls and patients with right brain damage imitated the gestures with nearly no errors. By contrast, out of 35 patients with (p.97) left brain damage, 15 made more errors than controls and were classified as apraxic. The results on the manikin replicated those of imitation: while normal controls, patients with right brain damage, and non-apraxic patients with left brain damage made only a few errors, the apraxic patients scored dramatically worse.
Manipulating the distance between model and replication
Transposition of body configuration from one body to another can be tested without any substantial demands on motor action by asking subjects to match photographs of gestures performed by different persons and seen under different angles of view. In a single case study, Alan Sunderland (2007) examined matching of hand postures in a patient with a severe disturbance of imitation caused by a left parietal lesion. The patient had no difficulties in selecting the same postures from photographs showing the postures with identical or mirror-reversed orientation, but committed many errors when asked to match photographs to cartoons showing the same gestures. Sunderland concluded that the patient could exploit visual similarity for matching gestures but failed when the match had to be mediated by a more abstract conceptual representation of the gestures.
The importance of the perceptual distance between demonstration and replication of gestures was demonstrated in a series of experiments probing the imitation of three- to five-step sequences of combined hand and finger postures (e.g., touch table with index—touch table with back of flat hand—touch table with thumb—touch table with fist) (Jason, 1983a, 1983b). Patients with left brain damage, about one-half of which committed errors on clinical testing of imitation, needed more trials than patients with right brain damage to learn such a sequence and performed more slowly and with more errors even after successful learning. When, however, the examiner was seated beside the patient, demonstrated the sequence with the same hand as the patient, and maintained each position until the patient had copied it, imitation of the sequence became errorless. Moreover, when the examiner gradually accelerated the pace of the demonstration, left brain damaged patient could follow as fast and as errorless as the right brain damaged patients. The procedure of simultaneous imitation had minimized the perceptual distance between demonstration and replication, and had thus rendered intermediate cognitive processing superfluous. The observation that patients who have apraxia for imitation on clinical testing perform as well as right brain damaged patients without apraxia on this variant of imitation strongly suggests that their problems concern interpolated processing stages that mediate equivalence between perceptually distant motor actions.
High and low routes to imitation
Before we leave this discussion of imitation we should reconsider its implications for the high- versus low-level dichotomy. Liepmann’s proposal that defective imitation testifies the inability to direct the limbs according to mental images addresses this dichotomy and identifies defective imitation with insufficient control of the low level of motor control by (p.98) the high level of mental images or, in his words, “the governance of the limbs by the mind” (Liepmann, 1913). By contrast, a direct route connecting perception to motor execution of gestures bears unmistakable hallmarks of belonging entirely to the low level of motor physiology: it functions automatically, bypasses semantic knowledge about the meaning of gestures, and is body part specific. Analysis of the empirical evidence, however, casts doubts on the reality and ubiquity of these features. The malleability of the mapping from vision to execution of gestures rimes poorly with an automatic replication of the perceived motor act. Its susceptibility to interpretations of the goals of an experiment and to cues highlighting the correspondence between the nominal laterality of the model and the subject’s hands, strongly suggest the intervention of higher-order cognitive processes in the functioning of the direct route. Finally, the observation that apraxic patients who fail imitation on their own body have similar problems when trying to replicate the body configuration on a manikin implies that imitation involves spatial representations beyond the “innervatory pattern” (Rothi et al., 1991) of motor actions. We will discuss the possible nature of these representations in Chapter 7 where we also return to the topic of body part specificity of disturbed imitation. At this stage we can summarize that an explication based exclusively on low-level neuronal mechanisms and neglecting the intervention of high-level cognitive processes cannot give a satisfactory account for imitation of gestures.
(1) The apparently selective deficit of imitation of meaningless gestures does not exclude the possibility that patients with visuo-imitative apraxia are generally unable to perform meaningless gestures, regardless of whether they are presented for imitation or for performance on verbal command. However, production of meaningless gestures on verbal command is difficult to assess. Since these gestures have no names, the verbal command must consist of a description of the spatial features of the gesture. Comprehension of such a command is not trivial for patients with left brain damage who have some degree of aphasia and restrictions of verbal working memory. In any case it would be difficult to decide whether errors are due to insufficient production of gestures or to insufficient comprehension of the instruction.
(2) Note the similarity of this definition to the historical accounts of constructional apraxia as affecting the translation of an object from one spatial dimension into another (Critchley, 1953) and to Schlesinger’s (1928) proposal that the disturbance which becomes manifest in the imitation of movements and the postures reached by them, is nothing else than “constructional apraxia on the own body” (see Chapter 3).