The relations between vision and touch in infancy have been studied according to two paradigms: (1) Fine descriptions of the visuo-prehensile coordination when infants attempt to reach and grasp objects near them. (2) Looking or holding times obtained in cross-modal transfer tasks between touch and vision. The analysis of evidence obtained from the separate bodies of literature (see Streri Reference Streri1993) supports Dijkerman & de Haan's (D&dH's) argument regarding dissociation between somatosensory processing for perception and for action. On one hand, cross-modal transfer of shape is not bi-directional from birth. In these tasks, newborns were familiarized with an object in one modality (touch or vision) and then, in the test phase, they received the familiarized object and the novel one in the other modality. A longer holding or looking time for the novel object is taken as evidence of cross-modal recognition. Newborns visually recognize a shape held previously (Streri & Gentaz Reference Streri and Gentaz2003; Reference Streri and Gentaz2004), but they fail to recognize tactilely a shape seen previously (Sann & Streri Reference Sann and Streri2007). On the other hand, visuo-prehensile coordination is also present at birth. The newborns, firmly seated on a chair inclined at 50° (with the chair supporting their head and trunk) are able to move their arms towards a target. The direction of movement is always correct, and this gesture is considered as visually triggered. However, contact with the object is rarely achieved (von Hofsten Reference von Hofsten1982). Neonatal visuo-prehensile coordination is also uni-directional. When newborns hold an object, they never move it towards the eyes (Streri & Gentaz Reference Streri and Gentaz2003; Reference Streri and Gentaz2004).

In short, when newborns hold an object, they do not attempt to look at it, but are able to recognize it visually in a cross-modal transfer task. When newborns look at an object, they do not recognize it tactilely, but attempt to reach and grasp it. This neonatal reaching is visually triggered, whereas perceptual recognition is tactilely supported. During the course of development, this scenario remains unchanged. Concerning cross-modal transfer, the failure from vision to touch has been demonstrated in 2-month-old infants (Streri Reference Streri1987) and in 1-year-old infants (Rose and Orlian Reference Rose and Orlian1991). Also, adults' performances are worse in the visual-tactile (V-T) modality than in the tactile-visual (T-V) modality (Connolly & Jones Reference Connolly and Jones1970; Jones & Connolly Reference Jones and Connolly1970; Juurmaa & Lehtinen-Railo Reference Juurmaa and Lehtinen-Railo1988; Newham & McKenzie Reference Newham and McKenzie1993; cf. Hatwell Reference Hatwell, Richelle, Requin and Robert1994). Thus, from birth and throughout life, the T-V modality appears to be performed better than the V-T modality. Regarding visuo-prehensile coordination, by the age of about 4–5 months, a child's hand approaching an object is guided visually and the focus of gaze switches back and forth from the hand to the object to be grasped. Infants are able to anticipate the trajectory of a moving target, and the hand is prepared to receive visually gained information. This action has several characteristics of the adult system (Jeannerod Reference Jeannerod, Whiting and Wade1986; von Hofsten & Rönnqvist Reference von Hofsten and Rönnqvist1988).

We suggest that visuo-prehensile coordination is devoted mainly to spatial process (Where). Because cross-modal transfers require memory and recognition processes, they are mainly devoted to the identification of the shape of objects (What). This dissociation is supported by the neuroimaging data obtained from studies on adults (Reed et al. Reference Reed, Klatzky and Halgren2005). Reed et al.'s data provide direct evidence that neural regions for somatosensory “what” are different from “where.”

Another dissociation in the somatosensory system can be evoked by the processing of the characteristics of objects. The somatosensory processing of microgeometric properties (such as texture) has to be distinguished from the somatosensory processing of macrogeometric properties (such as shape). Support for this idea is based on two lines of evidence: (1) our studies on newborns' perception of both properties of an object in bidirectional cross-modal transfer tasks between vision and touch (Sann & Streri Reference Sann and Streri2007), and (2) neuroimaging data obtained in adults. The results of the former revealed that cross-modal transfer of shape is not bidirectional. In contrast, cross-modal transfer of texture is bidirectional. Vision and touch appear to be equivalent for gathering, exchanging, and processing information about texture, but not about shape. Regarding neuroimaging data obtained in adults, Roland et al. (Reference Roland, O'Sullivan and Kawashima1998) suggest that a functional separation may exist in the cortical processing of microgeometric and macrogeometric cues: Shape and length discrimination activate the anterior part of intraparietal sulcus more than does roughness discrimination. Merabet et al. (Reference Merabet, Thut, Murray, Andrews, Hsiao and Pascual-Leone2004), using repetitive transcranial magnetic stimulation (rTMS), showed that a transient disruption in the somatosensory cortex impairs microgeometric (texture) judgment, while a transient disruption in the visual cortex impairs macrogeometric (distance) judgment. More recently, a functional magnetic resonance imaging (fMRI) study investigated the activity and the connectivity of parietal and occipital cortical areas during haptic, and (in separate runs) visual, shape or texture perception (Peltier et al. Reference Peltier, Stilla, Mariola, LaConte, Hu and Sathian2007). The results revealed that a network of cortical regions is selectively activated during haptic shape perception compared to haptic texture perception. The intraparietal sulcus and the lateral occipital complex appear to be shape-selective in the tactile modality, as well in the visual modality.

Taken together, these data suggest that, from birth, somatosensory processing is organized into several functional streams, and its evolution appears stable. Neonate behavioral data and adult neuroimaging data provide an answer to the major question concerning the visual influences on somatosensory processing. These influences differ for action and for perceptual recognition. They also differ according to the perceptual processing of object properties.