The N170 component and its links to configural face processing: A rapid neural adaptation study

Abstract

A neural adaptation paradigm where adaptor and test stimuli were presented in rapid succession was employed to investigate links between the face-sensitive N170 component and configural face processing. In Experiment 1, schematic adaptor stimuli preceded naturalistic images of upright faces, inverted faces, or isolated eyes. Relative to a baseline condition with schematic house adaptors, upright and inverted schematic faces adapted the N170 to subsequent naturalistic faces, demonstrating that this component is associated with neural processes involved in the analysis of first-order relational face configuration. In Experiment 2, two-tone Mooney faces adapted the N170 to naturalistic faces relative to a baseline condition with Mooney houses, suggesting links between the N170 and holistic face processing. Results demonstrate that the N170 component does not exclusively reflect the detection and analysis of individual face parts, but also the processing of first-order configural and global gestalt features of faces. They also show that neural adaptation procedures can be used to identify the neural mechanisms that are responsible for category-specific ERP components such as the N170.

Introduction

Humans are highly skilled in face recognition, and this ability has been attributed to the configural processing of faces. Configural processing is based on relations between component parts of faces, and is often contrasted with featural or part-based processing, which focuses on local information such as individual facial features and their properties (e.g., Maurer et al., 2002, McKone and Yovel, 2009). It has been suggested that configural face perception includes several levels of processing (Maurer et al., 2002; see also Diamond and Carey, 1986, Tanaka and Farah, 1993, Rossion and Gauthier, 2002): Initially, generic first-order relational information (two eyes above nose, nose above mouth) is retrieved, which is then combined into a holistic gestalt-like representation. Finally, second-order relational information (i.e., spatial distances between facial features) is processed, which distinguishes individual faces and forms the basis of face recognition.

Evidence for face-specificity in visual processing comes from event-related potential (ERP) and magnetoencephalographic (MEG) studies. The N170 component and its MEG counterpart (M170) reflect face-selective electromagnetic responses in occipitotemporal areas that are triggered 150–190 ms after stimulus onset (e.g., Bentin et al., 1996, Eimer, 2000a, Halgren et al., 2000). Because of their face-sensitivity, these components are regarded as early manifestations of face-selective cortical processing. Several studies have shown that the N170 is not affected by the familiarity of faces (e.g., Bentin and Deouell, 2000, Eimer, 2000a), indicating that this component is associated with perceptual face processing stages that precede the identification of individual faces. The question to what degree N170/M170 components are linked to configural (as opposed to part-based) face processing is still a matter of debate. For example, the fact that the N170 is larger in response to eyes presented in isolation than to full faces has led to the suggestion that this component is primarily linked to the processing of individual face components, and in particular the eyes (e.g., Bentin et al., 1996). However, subsequent studies have provided substantial evidence that the N170 is also sensitive to configural face processing. It has been repeatedly demonstrated that N170 amplitudes and latencies are strongly modulated by face inversion, with enhanced and delayed N170 components to inverted as compared to upright faces (e.g., Eimer, 2000b, Rossion et al., 2000, Itier et al., 2007). Adverse effects of stimulus inversion on face perception (e.g., Yin, 1969) are generally regarded as a hallmark of configural face processing. The fact that the N170 is highly sensitive to face inversion therefore strongly suggests that the underlying neural processes are not just involved in the part-based processing of face components, but also in the analysis of configural face properties.

Further positive evidence for a link between the N170 component and configural face processing comes from ERP studies that have measured the N170 in response to schematic face stimuli and two-tone Mooney faces. Schematic face stimuli (such as Smiley faces) retain the overall spatial configuration of faces, but no parts that could be individually recognized as face components. Therefore, schematic faces are assumed to primarily engage configural processing. With respect to the classification proposed by Maurer et al. (2002), schematic faces allow the retrieval of first-order relational information, and may also be processed at the subsequent holistic level (see also Latinus and Taylor, 2006). Mooney faces (Mooney, 1957) usually contain no individually recognizable facial features and, unlike schematic faces, no prototypical first-order configural arrangement of their component parts. To be recognized as faces, their global structure that is provided by shape-from-shading information needs to be processed. It has therefore been argued that the first face-specific processing stage activated by Mooney faces is holistic (Latinus and Taylor, 2005, Latinus and Taylor, 2006). If the N170 component is sensitive to configural face processing, it should not just be triggered by naturalistic face images, but also in response to schematic faces, or even by two-tone Mooney images of faces. This has indeed been demonstrated in previous ERP studies. Schematic faces were found to trigger an N170 that did not differ from the N170 in response to naturalistic faces, in line with the hypothesis that this component is associated with the processing of configural face information (Sagiv and Bentin, 2001, Latinus and Taylor, 2006). Inversion resulted in a delayed N170 for schematic faces, and either in a reduction (Sagiv and Bentin, 2001) or no effect (Latinus and Taylor, 2006) for N170 amplitude. There is also some evidence that Mooney faces elicit face-specific N170 components, suggesting that in the absence of first-order relational information, the N170 is sensitive to holistic face processing. Latinus and Taylor, 2005, Latinus and Taylor, 2006 observed larger N170 components for Mooney faces as compared to scrambled non-face Mooney control stimuli. Inversion did not affect the latency of the N170 to Mooney faces, and an attenuation of N170 amplitude for inverted Mooney faces was only found in one of these two studies (Latinus and Taylor, 2005). Furthermore, George et al. (2005) demonstrated that N170 amplitudes to inverted Mooney faces were larger on trials where these stimuli were identified as faces.

In summary, these previous ERP studies have obtained initial evidence that the face-sensitive N170 component does not just reflect a neural response to individual face parts, but is also associated with configural face processing. However, this conclusion has recently been called into question by the results of an MEG study by Harris and Nakayama (2008), where a rapid neural adaptation procedure (first developed by Jeffreys, 1996) was used to investigate links between the M170 component and configural versus part-based face processing. Neural adaptation (or repetition suppression) paradigms provide a new and potentially powerful technique to determine the functional properties of ERP components and their links to underlying neural processing modules. Neural adaptation effects are observed when the activity level of single neurons or neural populations in response to test stimuli is reduced because these stimuli were preceded by physically identical or categorically equivalent adaptor stimuli. Adaptation paradigms have been widely used in fMRI research to study the response profile of domain-specific brain regions (e.g., Henson, 2003), and were more recently also employed in EEG and MEG investigations of the N170/M170 component. N170 adaptation was demonstrated by Jacques and Rossion, 2004, Jacques and Rossion, 2006 by showing that N170 amplitudes to laterally presented faces were reduced when another face was already present at fixation, relative to trials where a central non-face control stimulus was present (see also Kovacs et al., 2006, for similar observations). Whereas these studies used N170 adaptation to demonstrate the general face-selectivity of this component, a more recent study from our group (Eimer et al., 2010) has shown that N170 adaptation effects reflect the activation of face-selective neurons by full faces as well as by individual face parts, such as the eyes.

Harris and Nakayama (2008) measured the M170 component to test faces that were preceded by full faces, scrambled faces, or isolated face parts as adaptors, and observed equally strong M170 adaptation effects for all three adaptor categories. Critically, they found no M170 adaptation effects at all on trials where adaptors were schematic line drawings of face configurations. This pattern of adaptation effects suggests that the M170 is linked to the detection of individual face parts, but does not reflect the processing of face-specific first-order relational properties. According to Harris and Nakayama (2008), the M170 reflects the activity of a structural encoding mechanism that is responsible for the part-based analysis of face components, and should therefore not be interpreted as a marker of configural face processing.

The aim of the present study was to use rapid neural adaptation procedures to study the links between the N170 and configural face processing by measuring N170 adaptation effects produced by schematic face adaptors (Experiment 1) or Mooney face adaptors (Experiment 2) on the N170 elicited by photographic images of faces. In both experiments, adaptor stimuli (S1) and test stimuli (S2) were presented for 200 ms, and were separated by a 200 ms interstimulus interval (Fig. 1). Participants monitored these stimulus sequences in order to detect and respond to infrequent target stimuli defined by the presence of a red outline shape that was aligned with the outer contours of an adaptor or test stimulus. Analogous to the MEG study by Harris and Nakayama (2008), Experiment 1 investigated whether schematic line drawings of face configurations produce N170 adaptation effects in response to naturalistic test face stimuli. Naturalistic face stimuli were preceded by different types of schematic adaptors (Fig. 1, top). Upright schematic faces showed the prototypical face configuration. Schematic houses, albeit visually similar, did not. The presence of N170 adaptation effects (i.e., reduced N170 amplitudes) for upright naturalistic face stimuli preceded by upright schematic faces, relative to trials with schematic house adaptors, would demonstrate links between the N170 and configural face processing. As naturalistic inverted faces and isolated eyes also trigger robust N170 components (Bentin et al., 1996), which are subject to adaptation when preceded by naturalistic face stimuli (Eimer et al., 2010), these two types of test stimuli were included in addition to naturalistic upright faces. Furthermore, schematic inverted faces and schematic eyes were included as adaptors in Experiment 1. This was done to investigate whether these non-canonical stimuli would still elicit some first-order relational configural processing (as suggested by Latinus and Taylor, 2006, for inverted schematic faces), resulting in N170 adaptation effects for naturalistic test face stimuli.

In Experiment 2, two-tone Mooney stimuli were presented as adaptors, and upright naturalistic faces served as test stimuli. As discussed above, Mooney faces are assumed to engage holistic stages of configural face processing, but not the first-order relational stage (Latinus and Taylor, 2006). Adaptors were upright and inverted Mooney faces as well as Mooney houses (Fig. 1, bottom). The presence of N170 adaptation effects in response to upright naturalistic test faces preceded by upright Mooney faces, as compared to trials with Mooney house adaptors would provide additional evidence for an association between the N170 component and configural (holistic) face processing, and against the hypothesis that this component is exclusively linked to a part-based analysis of faces. Inverted Mooney faces were included as adaptor stimuli in order to test whether and to what degree these stimuli would also activate configural-holistic face processing, and therefore produce an adaptation of the N170 to subsequent naturalistic test faces.