The way we use information determines the way in which we encode it. Dijkerman & de Haan (D&dH) illustrate this general principle with regard to somatosensory processing. Like vision and audition, proprioception can be either action-oriented or recognition-oriented. However, what characterizes proprioception from other sensory modalities is the role played by body representations. D&dH conclude that the dual somatosensory pathways give rise to two kinds of body representations: the body schema and the body image. But what is the basis of this dualistic distinction? How many body representations do we really have?

One representation that integrates all of the different types of information into a unified neuromatrix (Melzack Reference Melzack1992)? Two representations that are either based on functional criteria distinguishing the body image for recognition and the body schema for action (Gallagher Reference Gallagher2005; Paillard Reference Paillard, Gantchev, Mori and Massion1999), or based on temporal criteria distinguishing the actual body from the habitual body (Merleau-Ponty Reference Merleau-Ponty1945) or short-term and long-term body images (O'Shaughnessy Reference O'Shaughnessy, Bermudez, Marcel and Eilan1995)? Three representations that, for a more fine-grained distinction within the body image, take apart visuo-spatial body map and body semantics (Schwoebel & Coslett Reference Schwoebel and Coslett2005; Sirigu et al. Reference Sirigu, Grafman, Bressler and Sunderland1991)?

The evidence provided to support the distinction between different kinds of body representations relies mainly on neuropsychological dissociations: between deafferentation (disruption of body schema) and neglect or numbsense (disruption of body image); or between apraxia (disruption of body schema), autotopagnosia (disruption of structural body description), and body-specific aphasia (disruption of body semantics). However, neuropsychological evidence is open to interpretation. First, there are so many body deficits, and therefore so many possible dissociations, that one may end up with almost an infinite list of body representations. For instance, some patients are unable to identify their own body parts only, whereas others are unable to identify exclusively someone else's body parts (Felician et al. Reference Felician, Ceccaldi, Didic, Thinus-Blanc and Poncet2003). One would then have to make a further distinction between two types of structural description: for one's own body and for other people's bodies. Second, without a clear definition of what each type of body representation involves, it is hard to find conclusive clinical tests to assess the different levels of body representations in patients. For instance, autotopagnosia has been diagnosed by asking patients to point toward body parts, although one could argue that pointing requires not only the body structural description, but also the body schema (Schwoebel & Coslett Reference Schwoebel and Coslett2005).

A further problem in validating the distinction between multiple body representations is that there is almost a complete lack of experimentation outside of neuropsychology (except, e.g., Kammers et al. Reference Kammers, van der Ham and Dijkerman2006). D&dH emphasize the distinct neuroanatomical bases of somatosensory processing. However, the differences in activations that they describe may depend on the tasks, rather than on the underlying distinct body representations (Holmes & Spence Reference Holmes, Spence, Knoblich, Thornton, Grosjean and Shiffrar2006). We have to disentangle two alternatives: (a) one and the same representation used for different functions, or (b) different representations specific to each function. In order to prove the latter hypothesis, one would need to show that what differs is not only the function, but also the content of the representations. The function of body representation, whether it is action-oriented or recognition-oriented, might not provide such a clear criterion, contrary to what D&dH assume.

Action involves many different types of information about the body, which may not be encoded in the same format. On the one hand, one could distinguish the body as a target or as the mean (grouped together in the body schema in Figure 1 of the target article). Whereas the former is encoded in an egocentric frame of reference, like any other goal, this might not be true for the latter (Bermudez Reference Bermudez1998). They are both action-oriented, but their spatial perspectives are different. On the other hand, one could distinguish short-term information about body posture and long-term information about bodily constraints, such as the size and the strength of the limbs (de Vignemont Reference de Vignemont2006). Again, both types of information are necessary to plan a movement, but the dynamics are different.

Recognition covers an even wider scope of bodily information, including body posture and body size. The body image is said to include body percept, body concept, and body affect (Gallagher Reference Gallagher2005). The dynamics may vary, from short-term bodily sensations to long-term bodily properties (O'Shaughnessy Reference O'Shaughnessy, Bermudez, Marcel and Eilan1995). A further level of complexity occurs for the body image, as it can be applied both to one's own body and to someone else's body. Why are all of these aspects part of one single category? The unitary function of the body image is far from obvious. It is often described as what is left over after the body schema. Breaking down the body image into pieces is not the optimal solution either. Indeed, the triadic distinction leaves out the emotional component of the body image. Does that mean that there would be four kinds of body representations, or even more?

Shall we then give up on drawing distinctions between bodyrepresentations altogether? No, but one must be careful to avoid three main obstacles: (1) a lack of unity within each kind of body representation, (2) a lack of positive definition, and (3) a risk of infinite multiplication. D&dH have taken the first step by describing in detail the distinction for somatosensory processing. Further work needs to be done for body representations. To establish a taxonomy of body representations, one needs to describe not only the specific functions of body representations, but also their input (e.g., Does the weighting of each sensory modality in multimodal integration depend on the type of body representation?), their content (e.g., Are some body representations self-specific, and others shared between self and other?), their dynamics (e.g., How quickly can body representations adjust to body changes?), and their spatial frame of reference (e.g., Does the body schema/body image distinction overlap with the egocentric/allocentric distinction?). One may then be able to provide a full account of body representations.