In this paper, we report four experiments that explore the maintenance of a body representation in two individuals living with chronic or congenital absence of somatosensation. We used perceptual tasks to probe judgements about hand shape, (Experiments 1A and 1B), arm reach (Experiment 2), and to test for an attentional bias in peri-personal space (Experiment 3).
Experiments 1 and 2 required conscious judgements about body size and shape, or the location of landmarks on the hand. We found strikingly different degrees of distortion between controls, and the participants, suggesting that the visual representation of hand shape may be quite distinct from metrical knowledge (Longo and Haggard 2010, 2012). Both IW, who has lived nearly 50 years without large fibre somatosensation from age 19, and KS, who was born and has lived 40 years without both large and small fibre somatic input from her body, performed these tasks well. While there were differences in their estimates across conditions, which we will discuss below, they were able to make consistent decisions in these tasks, and hence do have a conscious representation of their unfelt bodies. Vision seems to be sufficient for them to make reliable judgements in these circumstances. The absence of small diameter fibres in KS is not expected to be relevant here as these are not thought to contribute greatly to the body image or schema (Proske and Gandevia 2012). Thus, our study of two rare individuals demonstrates that the conscious body image can be developed and maintained even when without somatic sensation.
Hand shape is accurately recognized with vision
In Experiment 1A, participants selected among a set of images of their own hand, distorted in length or width, the one closest to their perception of its real shape. The majority of control participants showed a systematic bias, choosing an image with reduced length/width ratio, on average 5% shorter and 5% wider, than the actual. While the results from the two test participants exposed differences between their two hands, in all cases, IW and KS showed less of a shortening than did controls, with KS even showing an overestimate of the length of her non-dominant hand. Thus, they both make “depictive” assessments (Longo and Haggard 2012) of the shape of their unseen hands that are—for the non-dominant hand—significantly more accurate than the controls. Whether this task assesses own-body image is, however, debatable and, in retrospect, it would have been useful to test participant’s sensitivity to distortion of others’ hands versus their own. If the identity of the hand is not critical, it may be that the task instead probes a form of mental imagery (Sirigu and Duhamel 2001; ter Horst et al. 2012).
Reporting landmark positions on the hand reflects experience
In Experiment 1B, participants reported the location of landmarks on their unseen hands. The controls showed the previously described severe underestimation of finger lengths (Longo and Haggard 2010, 2012). We did not, however, reproduce the graduated and increasing underestimation from thumb to little finger that Longo and colleagues report, possibly due to a difference in instructions. Longo and colleagues instructed their participants to report the felt location of the hand landmarks; since KS and IW would not be able to feel their hands, our instruction was to place the cursor onto the hand landmark, and we did not mention the felt position.
Interestingly, Ganea and Longo (2017) showed that the maps were robust to whether the hand was directly under the report surface or held on the lap. Our tests with IW and KS confirm that the reported positions are similar regardless of actual hand location. This suggests that the task is heavily reliant on a visual (KS and IW) or visuo-somatic representation (controls). Longo and Haggard (2010, 2012) have suggested that the “metrical” graduated distortion might be related to the different density of cutaneous receptors across the hand. Since the conscious representation of body shape is normally multi-modal, integrating visual and somatosensory inputs, the contribution of these two sources might alter across the cortical hand representation, affecting overestimation in their judgements. Instead, we found that in both control groups, all-finger digits were estimated at between 35 and 40% of their actual length (although the thumb was less distorted, at about 45%), while hand width was also slightly underestimated at about 90% of true width. The overall length/width ratio is consistent with—but much more marked than—the average selected ratios of less than unity in Experiment 1A. Thus, granular support for the influence of receptor density on hand representation was not found.
Cocchini et al. (2018) have reported that hand maps reflect dextrous hand use and are more accurate in trained magicians. The reduced error for thumb length that we observed is consistent with this idea, given the thumb’s importance in everyday actions and its priority in placement for grasp and manipulation of objects (Smeets and Brenner 1999; Smeets et al. 2002). The dominance of the thumb is further evident in both motor and sensory homunculi (Penfield and Boldrey 1937). There is also typically less individuated use of the lateral fingers (Miall et al. 2019), and in a grasping action, these can be guided by somatic rather than visual control—potentially leading to greater distortion in our visually based mapping experiment.
IW and KS differed somewhat from controls in their distortion patterns. They underestimated digit lengths, though by less than the controls, meaning that they had greater accuracy, while slightly overestimating hand width. These metric distortions were evident and consistent for both hands, with a suggestion of greater accuracy for the non-dominant hand. We had expected that due to their greater reliance on vision for control of action and their lack of access to a topographically skewed somatosensory input, IW and KS might be more veridical in their reports of the individual digit lengths than controls. This was indeed the case for IW. On the other hand, the shortening of digit lengths that KS displayed was similar to that seen in the controls, calling into question a simple somatosensory representational argument for the under-representation of digit length.
The reporting procedure used by KS and IW, verbally instructing the experimenter to steer the on-screen cursor, was clearly distinct from the self-driven joystick movement that the controls used. Note, however, that the joystick provided velocity control over the cursor position, so the controls gained no direct proprioceptive feedback of selected position. In addition, Longo (2018) compared the maps generated by participants using either a long pointer to indicate the landmarks, or verbal instruction of the experimenter who held the pointer; the maps were similar, and the pattern of distortion equivalent. Hence, we do not think the mode of reporting explains differences between IW, KS, and the controls.
The slight underestimates of hand width (between the primary knuckles for the index and little fingers) seen in our control groups are in contrast to the overestimation reported by Longo and colleagues (Longo and Haggard 2010, 2012; Ganea and Longo 2017). They asked participants to use (or to guide) a long-thin pointer to mark a position on a surface a few centimetres above the hand, and to move it to the lateral edge of the board between trials. We used a cursor that always originated at the centre of the lower edge of the screen, and the screen image was coplanar with the table top on which the hand rested. However, Longo and Haggard (2010) have shown that hand maps are unaffected if the hand is rotated 90 degrees, ruling out perspective biases. It may be therefore that a key difference is in guiding a visual cursor to each landmark, rather than using a pointer.
A surprise was that KS was initially uncertain about the names of her fingers. This suggests that she has paid little attention to her hands. KS does not use her hands much and rarely uses the lateral digits (middle, ring, and little). There are clear abnormalities in the musculoskeletal arrangement of her hands, including an inability to fully extend at the wrist, and in their central control, since she cannot independently move the middle, ring, or little fingers on either hand. There are also clear differences between IW and KS in their activities of daily living (see ABILHAND scores, Methods). She does not perform most dextrous tasks such as cutting up food, buttoning, brushing teeth, and combing her hair. In tasks such as using a spoon (which KS does daily) or writing (which she does rarely) KS uses an all-finger power grasp. In tasks such as picking up cards or a jigsaw puzzle piece, she uses a precision grip between thumb and index finger.
In contrast to KS, IW has used his hands extensively since an intense period of re-learning and rehabilitation just months after the onset of his neuropathy (Cole 1995). He daily performs tasks such as cooking, dressing, and so on. He is competent in many grasp actions, albeit with modifications in hand posture to improve object stability: he mainly uses his thumb, index, and middle fingers and often actively excludes his lateral fingers from the grasp, either extending them or flexing them into the palm of his hand (Miall et al. 2019). It is striking that the digit length estimates derived from his hand maps were more accurate than KSs and the controls (Fig. 4). He also had a more accurate spatial representation of his hands (Figs. 6 vs 5), for example with the angles between the knuckles closer to reality than for most controls. Hence, it is possible that chronic absence of somatosensory inputs together with visually controlled hand movement led to an accurate, visually based representation in IW, compared to controls whose body image is distorted by somatosensory inputs. Supporting this, Longo et al. (2012) reported that a subject born without one arm estimated the digit lengths more accurately for her phantom than her intact hand, consistent with a lack of distorting somatosensation from the missing limb facilitating veridical body image. Knowledge of hand shape gained from daily use of the contralateral, intact hand may have further augmented a veridical image. Interestingly, IW is strongly left-hand dominant, and yet this did not lead to perceptual differences between hands.
One explanation for the limited difference in KS’s hand maps from controls is the opposing effects of her lack of somatic input, taking her closer to veridical (as in IW) and her lack of motoric use, potentially augmenting the distance from veridical. Hence, we suggest that it may be KS’s impoverished hand use that has led to her degraded metrical knowledge of shape and size of her hands compared to IW. Her foreshortened hand map reflects her lack of dextrous experience tempered by a lack of somatosensory inputs, whereas the even greater foreshortening in controls is due to the presence of somatosensory distortions.
Overestimation of arm length
Healthy controls typically overestimate the target distance they can reach (Carello et al. 1989; Bootsma et al. 1992; Heft 1993; Rochat and Wraga 1997; Mark et al. 1997; Leclere et al. 2019). In the judgement of arm length, estimated in Experiment 2, we found that this overestimation was even greater for IW than for his controls, whereas the overestimation for KS was comparable to that of younger controls. The high variability of the original control group for KS, and the low mean bias of the older control group, was highlighted when testing a large group of young undergraduates. In that comparison, both IW’s and KS’s reach estimation biases were similar to the mean of both the undergraduate and older control groups (Fig. 9a).
Both KS and IW reported thinking through and attempting to use surrounding landmarks in making these judgements. IW’s precision in his reaching judgements, based on the slope of the psychometric curve, was slightly lower than that of the controls, i.e., the JND was larger, even more so for his non-dominant arm. This might again reflect a motoric component; his visual calibration of reach distance is better for his more used arm. KS’s JND values were not significantly different from the controls, however. When we repeated the measurements after adding a wider surround to the reaching arena, so that any obvious landmarks were distant from the visual target, their accuracy was unchanged, albeit this was only tested after they had familiarisation with the paradigm. It is also interesting to note that despite her self-reported poor depth perception, KS performed as well as controls. It remains to be seen how extra-personal depth perception in KS and IW compares to controls.
Attentional bias to peri-personal space
In Experiment 3, testing reaction times to detect visual targets, we found the expected RT advantages in our control participants for targets ipsilateral to the visible hand in peri-personal space relative to contralateral targets. This is consistent with the previously reported attentional advantage for targets close to the hand, within peri-personal space (Reed et al. 2006; Brown et al. 2015), that is thought to depend on multi-modal integration of visual, haptic, and proprioceptive representations in parietal cortex. Unexpectedly, and unlike previous reports (Reed et al. 2006; Brown et al. 2015), we found this effect reversed in extra-personal space, beyond reaching distance, for both control groups. It also reversed when the hand was hidden, although the depth of attentional modulation was less (Fig. 11, right side). We cannot yet explain these reversals, which are not expected in an account based simply on hand–target proximity. However, others have suggested that competitive attentional processes are at play. Hand–target proximity reduces reaction times in a visual search paradigm, but it simultaneously increases the difficulty of disengagement and relocation of attention from one place to another (Thomas and Sunny 2017).
Regardless of the attentional mechanism, we can safely assume that differences in RT depend on the relative distance between the target and the hand (which was not moved in our paradigm), as they were found for the controls when the hand was both visible and hidden. Thus, we can use the task to probe whether IW and KS also display these differences. KS showed significant RT difference in the visual but not non-visual conditions; IW showed no significant differences in either condition. These data suggest that KS has a visually based body representation or schema, whereas IW has no discernible body-schema-based representation of peri-personal space.
Two features of KS’s data are striking. First, her reaction times were uniformly short, when compared to the young controls, and second, as mentioned above, in the hand-visible conditions she showed the same pattern of modulated RTs as the controls. Fast responses in these attentional detection tasks are thought to reflect implicit and bottom–up processes (Risko and Stolz 2010). She may have used a predominantly bottom-up process because, unlike the other tasks, the visual detection task did not require mental imagery of her body. Additionally, a simple and consistent verbal response was sufficient, rendering her performance unaffected by her dextrous inexperience. That her RTs were modulated by target–hand distance suggests this bottom–up approach is influenced within peri-personal space, presumably by enhanced visual representation (Brown et al. 2015). This interpretation is further supported by the finding that RT modulation was only present when KS’s hand was visible: when her hand was occluded, RTs were unmodulated. These data argue that KS has a vision-based subconscious representation of space around her body.
In contrast to KS, IW was slow in his reaction times, even compared to the older control group. Additionally, he did not show any significant modulation of RTs across the tested conditions (other than the typical slow responses for invalid cue conditions, which we do not report here). We suggest that he was more strategic, top–down, in his approach—and possibly more cautious about the invalid and catch trials—as he was in a previous attentional task (Nougier et al. 1994). IW prefers a cautious approach in tasks that he wishes to perform as well as possible (Renault et al. 2018). There were certainly no signs of any subconscious proximity advantage for IW.
Visual proprioception versus visual control
Another way of framing the distinction between IW and KS is that KS has a strong sense of “visual proprioception” (Lee and Lishman 1975), the unconscious visual representation of the body, whereas IW does not. Again, while speculative, it is possible that visual inputs have replaced somatic inputs in KS’s central representations at some point in her development, and she may be able to use such alternative pathways (cerebellar and/or cerebral) without need for cognitive attention. In contrast, such a replacement would not have occurred in IW who matured into adulthood with intact somatosensation. Instead, IW appears to have replaced his loss of somatic input with conscious strategic control. Revised versions of Experiments 1 and 2 in which judgements of body shape and size are made implicit might be able to determine whether KS, unlike IW, may also have developed the capacity for visually controlled non-conscious motoric judgements. Other recent experiments support the idea that KS has more automaticity and less conscious visual control of hand movements than IW (Miall et al., in revision).
While visual proprioception may fuel KS’s automaticity and rapid responses, it is not sufficient to produce representational accuracy, which may in turn be reduced by her limited motor experience. In contrast, IW’s awareness of his body is entirely top–down, constructed through comparatively slow information processing traversing (we propose) conscious visual streams. IW is clear about his need to be consciously aware of his body position to control movement, and its dependence on conscious vision: “everything is through vision” (Cole 2016).
Finally, we note that our assessments of hand configuration and arm reach (Experiments 1 and 2) provide objective evidence that both IW and KS have developed and maintained a perceptually accessible representation of the body, or a body image. As IW relates, “rather than being disembodied, I am completely, totally, embodied. If I was not I would not know where I am. I re-associate and reconnect constantly.” (Cole 2016). KS has been asked repeatedly about her sense of the body. She never hesitates and has always maintained that she has one; when she closes her eyes, the world goes away, but she does not. To illustrate this difference, upon awakening and opening their eyes in the morning, IW goes through a process of re-establishing where his body is, whereas KS simply welcomes back the world to her embodied self.
In sum, and returning to the conceptual framing provided by body image and body schema, our data lend support to the idea that KS has developed a low-fidelity, automated (fast) motor representation (or schema) whereas IW uses a slow, high-fidelity, cognition-dependent representation for movement control.