1. Introduction
Individuals differ greatly in their face memory/recognition abilities (e.g., Wilhelm et al., 2010). For example, whereas super-recognizers (Russell et al., 2009) remember a vast number of faces, individuals with developmental or acquired prosopagnosia may have a hard time recognizing even the faces of their family members (Behrmann & Avidan, 2005). Although the general outlines of the functional neuroanatomy of face processing are understood (Haxby et al., 2000), research on the neural correlates of individual differences in face processing is still scarce. In the present study, we explored whether the amplitude of the N250 component in the event-related potential (ERP) is related to individual differences in face recognition and whether this relationship depends on the distinctiveness of the faces to be memorized.
In their influential face processing model describing its functional neuroanatomy, Haxby et al. (2000) distinguished between a core system and an extended system of facial information processing. The core system includes the fusiform face area (FFA), the occipital face area (OFA), and the posterior superior temporal sulcus. This core system is held to process facial features, face identity, and facial expressions, respectively. The extended system includes brain regions that encode biographical and semantic information gleaned from the face. In the ERP, various components in response to faces have been described. Best known is the N170, an occipito-temporal negativity peaking around 170 ms after stimulus onset, which is larger to faces than to most common objects and has therefore been interpreted as reflecting structural encoding of faces. Following the N170 and peaking between 200 to 300 ms after stimulus onset a repetition effect can be observed, consisting in a temporo-occipital negativity to repeated relative to non-repeated famous or personally familiar faces. This N250r has been related to a transient activation of facial representations for recognition (Schweinberger & Burton, 2003). Immediately thereafter, starting around 300 ms, a second repetition effect for faces has been described (e.g., Schweinberger et al., 1995). This has been related to the activation of multimodal biographical and semantic person representation (for a review see Schweinberger & Neumann, 2016).
Tanaka et al. (2006) investigated the N250 component in the context of long-term repetition. A specific face was designated as a target and had to be distinguished from several non-target faces. The target face was repeatedly shown multiple times but widely spaced and interspersed between other (non-target) faces. The N250 amplitude to the target face increased from the first to the second half of the experimental session, indicative of the build-up of face representations. The findings of Tanaka et al. (2006) were replicated by Sommer et al. (2021). Gosling and Eimer (2011) reported that the N250 can distinguish already familiar faces from unfamiliar faces. Since the N250 increases after repeated presentations of a face also when different images are used across presentations (Andrews et al., 2017; Kaufmann et al., 2009), the evidence suggests that the N250 reflects facial representations at a certain level of abstraction (Wiese et al., 2021). The neural generators of the N250 and N250r have been reported to be in or near the fusiform gyrus (Kaufmann et al., 2009; Schweinberger et al., 2002; Schweinberger et al., 2007).
For assessing the neural correlates of face processing-related abilities, knowledge about the psychometric structure of these abilities at the behavioral level becomes crucial. Fairly comprehensive psychometric work based on multivariate test batteries applied to large participant samples and analyzed by confirmatory factor analyses has shown that face processing-related abilities are robustly separable from general intelligence and are multi-faceted (Hildebrandt et al., 2010; Wilhelm et al., 2010). In difficult (accuracy) tasks, face perception and face memory abilities can be dissociated, but in easy (speed) tasks, there is only a single ability of face recognition speed. The separation of face perception and face memory has been also applied to the single-task Cambridge face memory test (Duchaine & Nakayama, 2006) versus the Cambridge face perception test (Duchaine et al., 2007).
Given the structure of face processing-related abilities derived from psychometric work, research on the neural correlates of individual differences in domain specific face processing abilities has recorded ERPs in independent testing sessions for EEG recordings versus psychometric measurements. Thus, in two independent large samples, Herzmann et al. (2010) and Kaltwasser et al. (2014) showed that both the accuracy of face perception and memory (in difficult tasks) and the speed of face processing (in easy tasks) are negatively associated with the latency of the N170 component and with the amplitude of the N250r. In other words, the shorter the N170 latency of an individual and the larger the priming effect in the N250r, the better are the mentioned face processing abilities. The negative association between the N170 latency and the accuracy of face perception and memory was further confirmed by Nowparast Rostami et al. (2017). These ERP studies were concerned with explicit memory tasks where individuals attempted to intentionally memorize and recognize faces during the EEG session as well as in most of the tasks administered in the psychometric test session.
Several studies with somewhat heterogeneous approaches and results were designed to test associations between face processing performance and fMRI data. Furl et al. (2011) compared a group of participants with developmental prosopagnosia and normal controls (total n = 35). Performance differences in a test battery of face identification were mirrored in the face selectivity of the fusiform gyrus during an incidental task; there was no relationship with repetition suppression (often likened to priming effects) in fMRI. Huang et al. (2014) used a localizer task in fMRI with passive viewing in two large samples. In both samples, the face selectivity of fMRI activation in both FFA and OFA were positively correlated with the accuracy of recognizing previously learned faces (controlling for the recognition performance for objects). Li et al. (2017) showed that face selectivity in the right and left FFA correlated with two different aspects of holistic face perception. Ramot et al. (2019) investigated resting state connectivity and related it to performance in the Cambridge Face Memory Test. They found no relationship for the connectivity with the face network but with connectivity between the face network and hubs outside of this network.
Altogether, ERP and fMRI studies show brain-behavior relationships. For ERPs, the most consistent findings concern the N170 and N250r. Considering that both the N170 and the N250r have been suggested to be generated in the fusiform gyrus, these findings are largely in line with the reports that fusiform face selectivity in fMRI is predictive of individual differences in face memory/recognition.
An important variable in face processing tasks is the typicality or distinctiveness of a face relative to the average of the faces experienced by a person. This average face is viewed as a prototype in a multidimensional face space in which any known or perceived individual face can be located (e.g., Leopold et al., 2001; Valentine, 1991; Valentine et al., 2016). If a face is located close to the prototype it is “typical” and if it is far away from the prototype it is referred to as an untypical or distinctive face. Distinctive faces are usually easier to perceive and recognize than typical faces (e.g., Bartlett et al., 1984; Sommer et al., 1995; Valentine & Endo, 1992).