The associative processes involved in faces-proper names versus animals-common names binding: A comparative ERP study

Abstract

Recognizing people involves creating and retrieving links between distinct representations such as faces and names. In previous research we have shown that the retrieval of face/name associations produced cerebral activities lateralized in the left hemisphere and spreading from posterior to anterior sites after about 300 ms. The present ERP study was performed to compare the specific electrophysiological activities elicited by the retrieval of face/proper name (FP) and animal/common name (AC) associations. Using a subtraction method to isolate the specific binding activities, we showed that both kinds of association produced two main posterior negative/anterior positive complexes, with a more frontal distribution for AC, and bilateral temporal activities. These findings confirm that general associative processes – independent of the kind of association – are not simply the sum of the activities elicited by each stimulus, and that they could involve both unimodal sensory and multimodal convergence regions of the brain.

Introduction

The ability to recognize people's faces and to recall their names is a key feature of our social life. Indeed, faces and names are two major attributes by which we identify people. For healthy individuals, identifying a person is an operation that often seems easy, fast and automatic (Sergent et al., 1992). However, recognizing a face requires fine discriminative perceptive processes to extract an abstract representation of the face, independent for instance of its orientation or emotional expression (at least according to the classical cognitive model of Bruce and Young, 1986). Moreover, it is well known that processing faces and processing names involve distinct brain areas: faces are processed in posterior occipito-temporal cortical regions, with preferential activity in the right hemisphere (Sergent et al., 1992, Kanwisher et al., 1997, Campanella et al., 2001) whereas anterior left temporal regions are more associated with processing names (Damasio et al., 1996, Gorno-Tempini et al., 1998). The question arises as to how the brain manages to create a single coherent representation of a person on the basis of these different attributes, processed by distinct cortical regions.

Research on this issue has mainly focused on the cross-modal interactions between visual and auditory information relative to people. The McGurk effect (whereby a perceived lip movement can change the perception of a concurrent spoken syllable, McGurk and McDonald, 1976) is a classical example of how stimuli in different sensory modalities can interact under some circumstances. But many cross-modal interactions can happen without anomalous or incongruent multisensory combinations. It has also been shown that the ability to understand auditory speech is increased when lip movements are visible (Reisberg, 1987), and that it is easier to pin-point who is speaking when lips movements and speech are presented in combination (Driver, 1996).

At a cerebral level, one solution to this “binding problem” (Von der Malsburg, 1981), is to postulate the existence of heteromodal regions, mainly located in the parietal, temporal and frontal cortex and in some subcortical regions such as the superior colliculus, which receive inputs from the cerebral areas processing each kind of information and bind them together to create coherent and unified representations of the objects in our surroundings. This hypothesis has received strong empirical support, notably from electrophysiological studies that have shown that auditory–visual interactions can appear very soon after the onset of the stimuli and can involve unimodal sensory regions as well as multimodal convergence areas (Giard and Peronnet, 1999, Molholm et al., 2002).

Macaluso and Driver (2005) have recently proposed an extended model of cross-modal interactions in which the classical view of feed-forward convergence from lower-level, sensory-specific regions to higher-order, heteromodal regions is not the only mechanism for functional integration. Based on the observation that congruency of visuo-tactile stimulations led to enhanced activation of the occipital cortex (Macaluso et al., 2000, Macaluso et al., 2002), the authors suggested that this effect could reflect back-projection influences from heteromodal brain regions to multiple unimodal brain regions. Some event-related potential data supports this model. McDonald et al. (2003) showed that the spatial congruency of visual and auditory stimulations produced an enhanced negativity. A reconstruction of the sources at two different time intervals suggested that cross-modal spatial interactions might have first affected multisensory regions in the superior temporal sulcus, and then unimodal ventral occipital cortex regions, as if multisensory regions influenced unimodal regions via feed-back projections.

But the “binding problem” is not necessarily limited to the study of cross-modal interactions between distinct sensory modalities. Within a single modality, and in particular in the visual domain, people are characterized by multiple representations, such as faces and names, which can also interact. We recently investigated the question of the associative processes between faces and names in two studies using PET-scan and ERP (event-related potential) techniques. Both were aimed at examining the cerebral activities elicited by the retrieval of face–name associations (Campanella et al., 2001, Joassin et al., 2004). After a learning phase of 24 face–name associations, forming 12 male/female pairs, participants were confronted with four experimental conditions, requiring the recognition of the learned pairs on the basis of the presentation of name–name (NN), face–face (FF), name–face (NF) or face–name (FN) pairs of stimuli. As is usual in this kind of study (e.g. Giard and Peronnet, 1999), the main analysis consisted of subtracting the non-mixed conditions (NN and FF) from the mixed ones (NF and FN), to isolate the specific associative processes between faces and proper names. The PET-scan study showed that the recognition of face–name associations was associated with the activation of three cerebral areas located in the left hemisphere: the left inferior frontal gyrus (Brodman area BA 45), the left median frontal gyrus (BA 6) and the left inferior parietal lobe (BA 40). This last region was interpreted as the brain region where the association of the distinct representations of faces and names operates. The ERP study revealed two main ERP components: a negative wave peaking at left parieto-occipital sites around 285 milliseconds (ms) and its positive counterpart with a maximum in the left centro-frontal electrodes around 300 ms.

These two experiments show that the retrieval of the representations of face–name associations is sustained by a network of cortical areas lateralized in the left hemisphere and producing cerebral activities spreading from posterior to anterior sites around 300 ms (Campanella et al., 2001, Joassin et al., 2004). A particular emphasis is put on the left inferior parietal cortex, often considered as a multimodal convergence region (Booth et al., 2002, Booth et al., 2003, Niznikiewicz et al., 2000) and interpreted as reflecting the place where the distinct representations of faces and names are brought together.

Nevertheless, these studies raise several questions, at least one of which concerns the specificity of the network of cortical regions involved in face–name associations relative to more general binding processes. It can be hypothesized that the retrieval of face–name associations will elicit cerebral activities distinct from those elicited by the retrieval of associations between objects and common names. Indeed, the perception of objects and faces has been shown to produce distinct cerebral activities. Allison and her collaborators (Allison et al., 1999, McCarthy et al., 1999) compared the intracranial electrophysiological activities recorded after the presentation of different categories of stimuli (faces, cars, flowers, animals, hands and written words). They detected a negative wave around 200 ms in the ventral occipito-temporal cortex, with a significantly larger amplitude for human faces than for any category of objects. These results suggest that, while this wave is elicited by many visual stimulations, its amplitude is specific to the perception of human faces.

Scalp recordings focused on the N170 wave. This component is a negative potential peaking around 150–170 ms after the stimulus onset on lateral occipito-temporal electrodes (Bentin et al., 1996, George et al., 1996, Eimer, 2000). The N170 is associated with the vortex positive potential (VPP) appearing with the same latency on central electrodes (Jeffreys, 1989, Campanella et al., 2000, Joyce and Rossion, 2005). This negative–positive complex is classically observed after the presentation of human faces and is assumed to reflect structural analysis of facial information in order to obtain a configurational representation of a face (Bentin et al., 1996, Jeffreys, 1996). The observation of a larger N170 amplitude in response to faces than to meaningful non-face objects such as birds, cars, or human hands led several scientists to conclude that this component reflects specific visual mechanisms devoted to the processing of human faces (Bentin et al., 1996, Carmel and Bentin, 2002). Other authors challenged this idea by showing that a large N170 can be elicited by other objects than faces (Tanaka and Curran, 2001, Rousselet et al., 2004). For instance, Tanaka and Curran (2001) showed that birds elicited a larger N170 than dogs in bird experts, and vice versa for dog experts. This suggests that the amplitude of the N170 could depend on the level of expertise of the visual mechanisms recruited to identify a given exemplar of a category of objects (Gauthier et al., 1999, Gauthier et al., 2000). However Carmel and Bentin (2002) recently explored the influence of task manipulation and the observer's expertise on the N170. They observed that (1) similar N170s were elicited by human faces, whatever they were task-relevant or not, (2) ape faces (on which the participants were not experts) and human faces elicited similar N170s, which were equally distinct from the N170 elicited by non-face objects (cars). The reaction to cars showed a large effect of the task requirement. Based on the observation that faces are processed more holistically than other objects (Tanaka and Farah, 1993, Sagiv and Bentin, 2001), these results led the authors to suggest that the N170 is face-specific and could sustain strategic processes (such as holistic processing of faces compared to part-based processing of other objects), and that the processing of other objects could depend on more general visual processes, sensitive to the manipulation of tasks and attention.

Studies of brain-damaged patients and cerebral imaging studies confirmed the special status of faces. For instance, several cases of prosopagnosic patients, who were unable to recognize familiar people whereas they were still able to recognize objects, have been described (Buxbaum et al., 1999, De Renzi et al., 1991, McNeil and Warrington, 1993). The inverse case, that is agnosic patients deficient in object recognition but with unimpaired face recognition, has also been described (Feindberg et al., 1994, Moscovitch et al., 1997).

Several studies of healthy participants have shown that, within the ventral and dorsal visual pathways, perception and identification of different categories of objects such as faces, animals and houses activate different cortical regions (Damasio et al., 1996, Grabowski et al., 1998, Kanwisher et al., 1997, Mitchell et al., 2002).

Proper names also seem to have a special status relative to other words, and in particular relative to common names. Dehaene (1995) showed, in an ERP study, that celebrity's names, when compared to animal names and verbs, elicited a left negative inferior temporal activity between 250 and 280 ms. Proverbio et al. (2001) observed that proper names produced a negative left anterior temporal activity around 150 ms, whereas a negative posterior activity around 275 ms was recorded for common names. They also showed that P300 was modulated by the lexical category: it was larger for proper names, which might reflect the greater complexity involved in recalling them as they represent unique entities. Proverbio et al. compared their results with those of several cerebral imaging studies showing that the lexical retrieval of different categories of words (proper names, tools, animals, verbs) was associated with distinct patterns of cerebral activation. In particular, animal names produced more posterior activation than tool names, reflecting the association of animal names with more visual representations. whereas tool names were more strongly associated with sensory-motor representations in the motor and pre-motor areas (Damasio et al., 1996, Martin et al., 1996, Pulvermuller et al., 1999). They also observed that proper names produced cerebral activation in the left temporal pole, a region also involved in episodic memory (Sergent et al., 1994, Gorno-Tempini et al., 1998, Tsukiura et al., 2002). Proverbio et al. (2001) thus proposed that the different patterns of brain activity obtained for common and proper names may be explained by the specific properties of their semantic organization, characterized by distinct associative links with sensory-motor and visual representations and memory.

Neuropsychology reinforces the notion that lexical categories differ according to their semantic organization, as a double dissociation between common and proper names, at least for access to semantics, has been established: (1) the patient APA, examined by Caramazza and Shelton (1998), showed a deficit in the semantic access to people who he could not name without any semantic deficit towards objects; (2) Lyons et al. (2002) studied a patient, FH, who was severely impaired in the production of common and geographical names, with a semantic impairment for the corresponding entities, but was still able to name people and recall semantic information about them.

Based on the literature reviewed above, this study aimed to examine whether associative processing, as explored in our previous research (Campanella et al., 2001, Joassin et al., 2004), is specific to face–name associations, or represents a more general mechanism, independent of the category of objects. For this purpose, we performed an ERP experiment in which participants were confronted either with common name–object associations or with family name–face associations. Six conditions were proposed (three for each kind of association): animal common name alone (C), animal picture alone (A), simultaneous presentation of animal and its name (AC), family name alone (P), face alone (F), simultaneous presentation of face and family name (FP). As described in previous ERP studies (Giard and Peronnet, 1999, Teder-Sälejärvi et al., 2002), this methodology allowed us to perform the subtractions [AC − (A + C)] and [FP − (F + P)], in which (A + C) and (F + P) represent the algebraic sums of the ‘single stimuli’ conditions. These subtractions enabled us to isolate the specific cerebral activities associated with the retrieval of a single and coherent representation of each association, and to compare the results so as to isolate any possible specificity of face–name associative processes.