Dijkerman & de Haan's (D&dH's) target article can be seen as an effort to apply the kind of dissociative organization proposed by Milner and Goodale for the visual system (see Goodale & Milner Reference Goodale and Milner1992; Milner & Goodale Reference Milner and Goodale1995), to somatosensory processes. Such a strategy is an example of a decompositional approach to neuroscientific explanation: It starts with a functional analysis of a cognitive process and decomposes it into different subsystems (each one supposedly responsible for processing different kinds of information); and then tries to determine the neural bases of each subsystem. The proposal is considered successful if these subsystems can be differentially correlated with different neural paths and regions, either by neuroimaging studies or by neuropsychological (double) dissociations.

However, a common problem of such explanations is that they are not complete until they can show how these different subsystems integrate and work together in normal, healthy subjects. The idea that the brain processes different information in different places at different speeds is fine, but the model must also offer an account of the time course of the different processing paths, how these different paths may conflict and integrate, how the internal activity evolves in interaction with the specific kinds of information available in the environment, and how the brain is able to organize its resources, so that on a given occasion, a given informational pathway may take over (Gomila & Calvo, forthcoming).

In this case, however, we think there are grounds for doubting D&dH's functional analysis in the first place. Their article proposes that tactile object recognition and body image constitute a different system from the one involved in proprioception. This implies that somatosensory processes involved in perception are different from somatosensory processes involved in action. Unfortunately, this distinction does not take into account results, like those obtained in the field of dynamic touch (Turvey Reference Turvey1996), that support a more comprehensive view on touch. In fact, it only focuses on what, following Jones and Lederman (Reference Jones and Lederman2006), may be called tactile sensing, as opposed to haptic sensing (which involves exploratory procedures, that is, hand movements for haptic object perception (Lederman & Klatzky Reference Lederman and Klatzky1990). There is plenty of evidence obtained from these more comprehensive approaches that shows, on the one hand, that proprioceptive information is relevant to object recognition (Blanco & Travieso Reference Blanco and Travieso2003; Pagano & Turvey Reference Pagano and Turvey1998; Pagano et al. Reference Pagano, Carello and Turvey1996), and, on the other hand, that cutaneous stimulation may contribute to motor control through the body schema (Travieso & Lederman, under revision; Westling & Johansson Reference Westling and Johansson1984).

To illustrate the relevance of proprioceptive information in object recognition, we will focus on a concise length and weight perception task. In such a task (Amazeen & Turvey Reference Turvey1996), a subject is asked to estimate the length and weight of a non-visible object which cannot be grasped completely (it can only be handled). What Turvey's group has been able to prove is that, in order to perform this task, the subject needs to extract the rotational inertial properties of the object. These are obtained from the antigravitational force required by the muscles to hold the object (their rotational dynamics). The invariant tensor of inertia, which according to Turvey (Reference Turvey1996) is the physical quantity we extract, can be obtained only from the muscular effort required to move the object. Hence, this example clearly shows that intentional movement is necessary for haptic length or weight object perception. Furthermore, the same information is used to perceive both our own body and the external object (Pagano & Turvey Reference Pagano and Turvey1998; Pagano et al. Reference Pagano, Carello and Turvey1996), which contradicts D&dH's functional separation.

With respect to the contribution of cutaneous information to motor control, several studies have shown that cutaneous information contributes to proprioception. It is well known that the perception of hand position and orientation is mediated by cutaneous information, but recent studies show, as well, that the contribution of cutaneous information also extends to the perception of other body parts (Collins et al. Reference Collins, Refshauge, Todd and Gandevia2005). On the other hand, cutaneous deficits also affect motor dexterity (Moberg Reference Moberg1983; Reference Moberg1991). For example, Travieso and Lederman (under revision) have found inverse correlations between the spatial resolution thresholds at the fingertips and the time needed to perform motor-dexterity tests and to recognize three-dimensional objects, together with deficits in texture perception (Travieso & Lederman, under revision; Travieso et al. Reference Travieso, Blanco, Rosa, Burton and Schmidt2001). These results also contradict the separation of different somatosensory processes for perception and action proposed by D&dH.

As a last point, we want to highlight that the authors themselves implicitly acknowledge in the target article that their proposed functional decomposition does not really fit with the neuroanatomical data, given that the posterior parietal cortex plays a role both in recognition and in action, which rather suggests common processing. That their functional analysis is problematic can also be seen, for example, in how they interpret the performance of patient G.L. in a pointing task (Fig. 3). In section 3.2, paragraph 3, D&dH review different studies that claim that different representations of the body are used for the guidance of movements and for perceptual judgments. In this context, they interpret the results obtained by Paillard (Reference Paillard, Gantchev, Mori and Massion1999) when studying patient G.L. as showing a dissociation between a normal “perceptual” representation (because she is able to recognize objects and point to a picture of her hand) and an impaired “sensorimotor” representation (because she has problems when pointing towards haptic stimuli). However, what we think the results show is that G.L. was simply poor at pointing without vision: she was equally good at pointing with vision towards her hand or towards a picture. Her problem was not with the action itself, but with the use of haptic information alone (without vision). This example shows a problem shared by all those models which try to establish strict functional separations between perception and action: When looking for experimental support for the dissociation, it is never clear whether a task is really a “perceptual task” or an “action task” (for an example, see Franz Reference Franz2003), or whether the tasks chosen are really equivalent and comparable (for a discussion of this issue in the case of vision, see, e.g., Franz Reference Franz2001; Franz et al. Reference Franz, Fahle, Bulthoff and Gegenfurtner2001). This is so because perception and action are closely intertwined in our interaction with the world, as our example of haptic length and weight perception shows.

In summary, we think that a comprehensive view on touch cannot be reduced to a twofold perception-action system with corresponding independent neural pathways. In particular, the dynamic properties of perception-action loops must be taken into account to provide a satisfactory explanation of the haptic system.