Taxonomies of human body representational systems are nearly as numerous as the putative systems themselves. Neuroimaging, neuroclinical, cognitive, and developmental studies over the last 15 years have established that human body knowledge is widely distributed in the adult brain. The most consistently referenced taxonomy of body representations includes at least three partially independent levels (Gallagher Reference Gallagher2005; Reed Reference Reed, Meltzoff and Prinz2002; Sirigu et al. Reference Sirigu, Grafman, Bressler and Sunderland1991): Sensori-motor body representations are primarily responsible for on-line movement and perception of one's own body; visuo-spatial body representations specify the structure and physical appearance of the human body; and lexical-semantic body representations support conceptual, language-based knowledge about the human body.

Dijkerman & de Haan's (D&dH's) model focuses on sensori-motor body representations. Like Sirigu et al. (Reference Sirigu, Grafman, Bressler and Sunderland1991), the target authors distinguish unconscious, own body representations that control motor activity, from own body representations that underpin conscious perception. These perceptual representations are then further distinguished by D&dH as to whether originating from external stimuli (object exploration and recognition) or from internal stimuli (proprioception). This last level is referred to as “body image” in the target article, which is rather unfortunate because this term is already fraught with multiple confusing meanings and references (Gallagher Reference Gallagher2005; Reed Reference Reed, Meltzoff and Prinz2002; Slaughter & Heron Reference Slaughter and Heron2004).

D&dH offer convincing evidence for their taxonomy. Given that multiple body-related representations exist in adults, at least two developmental scenarios are possible. One is that, at birth, humans are equipped with a supramodal representational system that subsumes several types of body representations. Meltzoff and colleagues (Gallagher & Meltzoff Reference Gallagher and Meltzoff1996; Meltzoff & Moore Reference Meltzoff, Moore, Bermúdez, Marcel and Eilan1995) have proposed this sort of a system, mainly to account for neonatal imitation, which, they suggest, could be accomplished only through coordination of visual perceptual information about others' body movements with proprioceptive information about the infant's own body positions and motor representations that control the imitative act. On this view, development of body representations would involve progressive dissociation, from the supramodal format, into the distinct functional representations evident in adulthood.

Another possibility is that at least some of the distinct body representations described by D&dH are differentiated from the start. Neonatal imitation aside (it should be noted that there is a good deal of debate about whether neonatal imitation exists as a flexible behavioural strategy as opposed to a fixed reflex; see Anisfeld Reference Anisfeld1996), the developmental evidence suggests that this is the more likely scenario.

Organized motor activity is evident prenatally (Myowa-Yamakoshi & Takeshita Reference Myowa-Yamakoshi and Takeshita2006) and in the first several months of life. A recent study employed four-dimensional ultrasonography to examine the arm and hand movements that third-trimester foetuses made towards their faces. More than half of the foetal arm movements resulted in the hand touching the mouth, and the majority of reaches to the mouth were preceded by the mouth opening (Myowa-Yamakoshi & Takeshita Reference Myowa-Yamakoshi and Takeshita2006). Neonates spend approximately 20% of their waking hours touching their own face, and almost one-third of these movements result in the infants touching their mouths (Butterworth & Hopkins Reference Butterworth and Hopkins1988; Lew & Butterworth Reference Lew and Butterworth1995). When infants touch their own face, there is no rooting reflex, which suggests that even at this early age the infants are capable of distinguishing external stimulation from self-stimulation (Butterworth & Hopkins Reference Butterworth and Hopkins1988), though this is not yet as complex as the sort of perceptual self-recognition defined by D&dH.

In the target article, perceptual self-recognition is defined as an integration of proprioceptive, kinaesthetic, and visual information. This is not evident until 5 months of age, at the earliest. For instance, when infants were presented with a video playing a contingent view of their own moving legs and a second video playing a non-contingent view of legs (delayed feedback), infants aged 5 months and older preferred to watch the non-contingent view of themselves, as evidenced by longer times spent looking at the unusual display (Bahrick & Watson Reference Bahrick and Watson1985). Furthermore, infants preferred to watch the non-contingent view of themselves even when they could not also see their own legs, which suggests that they could integrate the visual and proprioceptive information. Thus, body representations supporting self-perception appear to develop later than representations that control motor activity.

D&dH distinguish between perceptual representations for self and for objects. The literature on object exploration in infancy suggests that exploration of the surface features of objects becomes increasingly precise over the first year of life. At birth, motor actions applied to different object sizes, textures, and shapes are largely undifferentiated, which has traditionally been interpreted as a consequence of a general inability to recognize the self as a distinct causal entity (Piaget Reference Piaget1953). However, another interpretation is that the somatosensory representations that guide manual exploration do not develop until after 6 months of age. It is only in the second half of the first year that infants show clear evidence of altering their manual exploration in response to specific object properties (Bourgeois et al. Reference Bourgeois, Khawar, Neal and Lockman2005).

Thus, the behavioural evidence from developmental studies is in line with the neuroimaging and neuroclinical evidence presented by D&dH. Whether the distinct processing streams postulated in the target article are evident in infancy and early childhood, is, as yet, unknown, though there is some suggestion that dorsal and ventral streams for object processing are functionally distinct by 8½ months of age (Newman et al. Reference Newman, Atkinson and Braddick2001).

Finally, the development of higher-level body representations – visuo-spatial and lexical-semantic – takes place even later. Infants do not recognize the typical human body shape until at least 12 months of age (Slaughter & Heron Reference Slaughter and Heron2004). Furthermore, a recent study failed to find evidence that the development of representations for recognizing human bodies is related to the maturity or complexity of infants' own sensori-motor representations (Christie & Slaughter, in preparation). Lexical-semantic body knowledge is acquired later still – not until age 15 months at the earliest, when infants first learn names for body parts (Witt et al. Reference Witt, Cermak and Coster1990).

We know bodies from inside and from without; we know them as objects and as vehicles. It is not surprising that there are numerous varied levels on which the brain represents bodies. Taxonomies of body representations can be useful, but to date there is no comprehensive system for organizing our knowledge about body representations, or for communicating that knowledge. This is an important next step.