Human brain potential correlates of repetition priming in face and name recognition

Abstract

We investigated repetition priming in the recognition of famous people by recording event-related brain potentials (ERPs) and reaction times (RTs). Participants performed speeded two-choice responses depending on whether or not a stimulus showed a famous person. In Experiment 1, a facilitation was found in RTs to famous (but not to unfamiliar) faces when primed by the same face shown in an earlier priming phase of the experiment. In ERPs, an influence of repetition priming was observed neither for the N170 nor for a temporal N250 component which in previous studies had been shown to be sensitive to immediate face repetitions. ERPs to primed unfamiliar faces were more negative over right occipitotemporal areas than those to unprimed faces, but this effect was specific for repetitions of the same image, consistent with recent findings. In contrast, ERPs to primed familiar faces were more positive than those to unprimed faces at parietal sites from 500–600 ms after face onset, and these priming effects were comparable regardless of whether the same or a different image of the celebrity had served as prime. In Experiment 2, similar results were found for name recognition—a facilitation in RTs to primed familiar but not unfamiliar names, and a parietal positivity to primed names around 500–600 ms. ERP repetition effects showed comparable topographies for faces and names, consistent with the idea of a common underlying source. With reference to current models of face recognition, we suggest that these ERP repetition effects for familiar stimuli reflect a change in post-perceptual representations for people, rather than a neural correlate of recognition at a perceptual level.

Introduction

Human faces are very rich in social information, in providing cues not only to a person’s identity, but also to emotional state, age, gender and so forth. It is, therefore, not surprising that face perception has attracted a lot of interest in current research. Indeed, it is widely believed that face recognition is subserved by specific processes that are qualitatively different from processes underlying the recognition of other types of objects [1], [20], [30], [31], [34], [40], [41].

The present paper investigates repetition priming in the recognition of famous faces and names by recording event-related brain potentials (ERPs) and reaction times (RTs). Research on priming has contributed to our understanding of the visual recognition of words [49], [50], pictures of objects [70], or faces [12], [25], [27], [65]. Repetition priming refers to the observation that the processing of stimuli can be altered, and often enhanced, when the same stimuli have been previously encountered. Priming is often observed in terms of decreased RTs or error rates in responding to repeated, as compared with novel, stimuli and priming can be present in the absence of explicit memory for the study items [32]. Moreover, the effects of repetition often exhibit a degree of perceptual specificity, in that they are reduced when the perceptual format of the stimuli is changed between a study and a test phase [27], [39], [76].

Although priming effects have usually been investigated with behavioural measures, ERP studies have contributed significantly to our understanding of processes related to memory and priming [51]. ERPs have been widely used to study face perception, with many of the older studies focussing on various ERP components related to identity matching [3] or semantic matching [2], memory encoding [75], or hemispheric asymmetries for faces [68], [69], [73]. In contrast, many of the very recent ERP studies on neural correlates of face perception have focussed on one particular component, the N170, an electrically negative wave over occipito-temporal areas, approximately 170 ms after the onset of a face. The N170 is prominent for faces but absent, or strongly attenuated, for visual stimuli other than faces [6]. However, although the N170 may signal face specificity (for controversial issues, see [21]), subsequent research has shown that this component is not influenced by the familiarity of faces [7], [22], [60]. Similarly, the N170 is insensitive to repetition priming [5], [17], [66]. The available evidence, therefore, suggests that the N170 is related to early structural encoding of faces, rather than to the individual recognition of familiar faces [7], [23], [61].¹

Rather than focussing on whether or not faces are special, the present study aims at identifying processes that are involved in priming and in the recognition of individual faces. Given the insensitivity of the N170 to face familiarity, it seems likely that other ERP components would be better qualified as candidates for a neural correlate of face recognition and priming. The evidence on this issue is limited, however. In one recent study by Paller et al. [53] some faces were experimentally familiarized and the ERP differences between learned and new faces were assessed. Some faces, ‘unnamed faces’ were learned purely visually whereas others ‘named faces’ were supplemented with biographical and name information during learning. Compared with new faces, named faces elicited more electrical positivity at anterior and posterior scalp locations between 300 and 600 ms in the recognition test. For unnamed faces, however, this effect was confined to posterior sites. In another recent study by Eimer [22] famous faces were found to elicit an increased negativity between 300 and 500 ms and an increased positivity between 500 and 700 ms when compared with unfamiliar faces. Furthermore, these differences were attenuated for repeated stimuli. Although the reasons for the discrepancy between the results by Paller et al. [53] and Eimer [22] are unclear at present, these studies were designed to investigate differing issues within face recognition. The Paller et al. study used experimentally familiarized rather than famous faces, and Eimer’s study required participants to identify non-face stimuli, or immediate stimulus repetitions, rather than make familiarity decisions. Thus, task or stimulus differences may account for the different findings.

The time range of 300–700 ms in which ERP modulations due to face familiarity were identified by both Eimer [22] and Paller et al. [53] is comparable to the time range in which ERP modulations of associative priming (e.g. Gorbachev’s face as a prime for Yeltsin’s face) are observed in face recognition [63], [66]. Cognitive models of face recognition assume that associative priming affects post-perceptual processing, and in particular interactions between stimulus-independent representations for personal identity (so-called “person identity nodes” or PINs) and semantic representations [15]. ERP correlates of these processes should therefore, be independent of the stimulus type (e.g. faces, personal names, voices) in which a famous person is presented. While it is unclear to what extent the familiarity effects found by Eimer [22] and Paller et al. [53] would be face-specific, there is good evidence that ERP effects of associative priming are independent of the stimulus type [63], and are thus confined to post-perceptual processing.

Two classes of ERP phenomena for faces have been discussed so far. The N170 exhibits a degree of specificity for faces, but at the same time it is insensitive to face familiarity. A class of later ERP modulations within the 300–700 ms time range appear sensitive to face familiarity, but it is unclear whether these modulations are specifically related to face processing, or whether they are governed more by semantic processing of a person’s identity.

Various groups have reported ERP modulations elicited by the repetition of faces. The most consistent finding is that ERPs elicited by repeated faces evoke more positivity (or less negativity) around 300–600 ms at parietal-central locations than ERPs elicited by new faces [3], [33], [35], [52], [56], [57], [72]. This effect is referred to by different groups as either the ‘ERP repetition effect’ [35], [54], [56], [62], [66], ‘old/new effect’ or as a modulation of the N400 [3], [8], [56]. However, it is important to note that different ERP effects of face repetitions may be elicited, depending on factors such as the familiarity of faces [66], the lag between repetitions [54], or the way in which memory for the prime face is accessed (direct or indirect [33]). While the ERP repetition effects between 300 and 600 ms at parietal-central locations may well be functionally similar to the N400, the standard N400 is typically elicited in rather different experimental settings. In the present study, we will therefore use the term ‘ERP repetition effect’ as a theoretically neutral and convenient shorthand label.

Of particular interest for the present study, recent research has identified a distinct ERP modulation as a result of the immediate repetition of faces. This effect is seen in the time range in between the N170 and the above-mentioned later modulations beyond 300 ms. Schweinberger et al. [66] found that immediate repetition of faces caused an increased negativity between 180 and 290 ms at inferior temporal sites. This relatively early ERP repetition effect peaked around 250 ms ‘N250r’ and was strongly lateralized to the right hemisphere (for similar results, see also [5]). This effect could only partially be related to representations for familiar faces, because the effect was observed for unfamiliar faces as well, though with reduced amplitude. However, subsequent research showed that when between two and four other faces intervened between repetitions, repetition priming still caused increased negativity at right inferior temporal sites, but only for repeated familiar faces [54]. Relative to immediate repetitions, this effect was smaller in amplitude and had a slightly longer peak latency (around 280 ms). Repetition of written personal names also caused increased negativity at inferior temporal electrodes for familiar but not unfamiliar names. However, the effect for names was observed predominantly over left hemisphere temporal areas (for further evidence that ERPs can differentiate brain systems subserving memory for faces and names, see [74]). These two features of this ERP effect—specificity for familiar stimuli and different topography for faces and names—suggest that it does not reflect a general facilitation of perceptual encoding, nor does it reflect a stimulus-independent facilitation of semantic processing. In terms of cognitive models of face recognition [13], [14], [15], [81], this ERP effect may therefore, reflect the transient activation of perceptual recognition units for faces and names (FRUs and NRUs, respectively).

The aim of this study is to investigate ERP correlates of repetition priming over longer time intervals (i.e. more than 15 min) and in a situation in which many more (i.e. several hundreds) stimuli intervene between any repetitions. A large body of experimental studies has demonstrated behavioural repetition priming under those conditions (e.g. [11], [12], [25], [27], [28]). For some time, researchers tended to explain repetition priming in terms of increased activation levels, or decreased thresholds, at the level of recognition units [12], [50]. However, more recent models of face recognition argue that changes at the level of recognition units are too transient in nature to explain the remarkable longevity (e.g. [11], [28]) of repetition priming in face recognition. Instead, it is held that priming involves a strengthening of the connections of perceptual nodes to post-perceptual memory nodes (specifically, links from FRUs to PINs [15]). In the model of face recognition by Burton Bruce, and colleagues [13], [14], [15], PINs are the level of representation at which information from different stimulus types (e.g. faces, voices, names) initially converges into a common representation of a person’s identity. The idea that long-term priming involves a change in the links between representations—rather than a change in activation levels of the representations themselves—is also widespread in word and object recognition [79], [80]. Accordingly, the processing of a target face that is primed across long intervals may not change at levels up to and including the FRU. Rather, it is held that PIN activation reaches the threshold earlier for a primed than an unprimed face as a result of the more efficient FRU-to-PIN links. It may be speculated that, whereas changes in the activation of representations that involve the synchronized mass action of neuron assemblies may be readily observable by scalp-recorded ERPs, changes in neural connections may not in themselves be observable by ERPs, other than in terms of their effect on subsequent neural representations.² Thus, with ERPs it should be possible to detect changes in face processing at the level of both FRUs and PINs, but it may not be possible to directly detect changes at the FRU-to-PIN links.

In two separate experiments, we investigated long-term repetition priming in the recognition of famous people by recording ERPs and RTs. In Experiment 1, participants performed speeded two-choice responses depending on whether a stimulus showed a familiar or an unfamiliar face. Some of the faces were unprimed, whereas others had been shown in an earlier priming phase using either the same or a different portrait of that person. This design allows for a separation of priming effects that are caused by the repetition of an identical image from effects that are caused by the repetition of a face regardless of the precise image used. An additional purpose of Experiment 1 was to compare repetition effects for familiar faces with those for unfamiliar faces. Unfamiliar faces presumably have no stored representations (FRUs), at least not at first presentation. Therefore, any repetition effects that are specific to familiar but not unfamiliar faces must arise at processes that are only available for familiar faces (e.g. FRUs, PINs, access to semantic information). Conversely, any repetition effects that are seen for both familiar and unfamiliar faces must arise at processes that are common to familiar and unfamiliar faces (e.g. perceptual encoding of a face). Experiment 2 employed an analogous person familiarity task but used names as stimuli, again with some names having been unprimed and others having been shown in an earlier priming phase using the same or a different font type.

If ERP correlates of long-term repetition priming reflect changes only at the level of PINs and subsequent semantic levels of processing, one would predict that these effects should be seen at a longer latency, i.e. beyond 300 ms, when compared to transient repetition effects that presumably reflect changes in FRU activation levels. Alternatively, if long-term ERP repetition effects reflect differences in the activation of recognition units in a similar way as short-term repetition effects do [54], [66], we would expect to observe a stimulus-specific temporal N250r modulation for repeated faces and names similar to these previous studies.

As a final step, we will therefore compare repetition effects for faces and names across experiments. Divergent repetition effects should be seen for faces and names to the extent that these effects directly reflect differences at stimulus-specific perceptual levels of processing, i.e. in FRU or NRU activation. Divergent ERP repetition effects for faces and names should also be seen to the extent that these effects directly reflect a strengthening of stimulus-specific (either FRU–PIN or NRU–PIN) links. It is also noteworthy that research into the N400 ERP component has suggested topographical differences between responses to pictures of objects and words [29], [48]. These findings suggest that the neural generators for the effects with different domains of stimuli are non-identical, implying that semantic knowledge might not be stored in a domain-independent manner (for a recent review, see [44]). For present purposes, findings of an N400-like ERP repetition effect which differs topographically between faces and names might, therefore, give rise to the possibility that semantic knowledge for people may be stored in a domain-specific manner. This type of finding would be in some contrast to current models of face recognition [13], [14], [15] which suggest that knowledge for people is accessed via PINs in a domain-independent manner. Thus, if ERP correlates of priming reflected changes only at the level of domain-independent PINs (even if these were ultimately a result of a strengthening of either FRU–PIN or NRU–PIN links, respectively), they should be comparable for faces and names.