Age-related changes in processing faces from detection to identification: ERP evidence

Sharon Daniel; Shlomo Bentin

doi:10.1016/j.neurobiolaging.2010.09.001

. Author manuscript; available in PMC: 2013 Jan 1.

Published in final edited form as: Neurobiol Aging. 2010 Oct 18;33(1):206.e1–206.e28. doi: 10.1016/j.neurobiolaging.2010.09.001

Age-related changes in processing faces from detection to identification: ERP evidence

Sharon Daniel ¹, Shlomo Bentin ²

PMCID: PMC3025306 NIHMSID: NIHMS237943 PMID: 20961658

Abstract

We examined the ability of people 70 to 90 years old to apply global, configural and featural face-processing strategies. In addition we investigated age-related changes in the ability to categorize faces at basic, subordinate and individual levels. Using the N170 potential as index of early face processing and the P300 component as index of categorical decision making and effort, we found significant age-related perceptual changes which slowed and somewhat impaired face processing. Specifically, older participants had problems integrating face features into global structures, demonstrating enhanced dependence on distal global information. They did not apply configural computations by default while processing faces which suggests that, unless identification is required, they process faces only at a basic level. These perceptual changes could be the cause for slower and less accurate subordinate categorization, particularly when it is based on details. At the neural levels face processing was not right-lateralized, reflecting excessive involvement of the left hemisphere in perception leading to a more general reduction of inter-hemispheric asymmetry. In addition we found excessive but nonselective activation of frontal regions adding support to the view that executive control and particularly inhibition of irrelevant input are reduced in the elderly.

Evidence from human (Bian and Andersen, 2008; Fiorentini, et al., 1996; Habak and Faubert, 2000) and monkey research (e.g. Wang, et al., 2005; Zhang, et al., 2008) indicates that, though not dramatically (e.g. Norman, et al., 2008), the efficiency of visual perception declines with age even in healthy individuals. Within this framework age-associated changes in face recognition have been the focus of many studies using performance (e.g. Bartlett and Fulton, 1991; Boutet and Faubert, 2006; Habak, et al., 2008) as well as EEG (e.g. Pfutze, et al., 2002) and hemodynamic neuroimaging (e.g. Grady, 2002; Grady, et al., 2007) methods. Most of these studies revealed reduced face recognition ability in the elderly (e.g. Lott, et al., 2005; Searcy, et al., 1999), even more so than it is evident for object recognition (Boutet and Faubert, 2006). Notwithstanding the amply documented age-related impairment in memory functions, which obviously accounts at least in part for impaired face recognition and identification (e.g. Bartlett and Fulton, 1991; D'Argembeau and Van der Linden, 2004; Fulton and Bartlett, 1991), several studies explored the possibility that perceptual factors might also cause face recognition difficulty in the normally aging population (Lott, et al., 2005). Some of these studies associated this impairment to the deterioration of optical functions such as acuity (Tejeria, et al., 2002) and contrast sensitivity (Owsley, et al., 1981; for broader reviews see Sekular and Sekular, 2000; Spear, 1993). Yet, there is also evidence that age-related changes in low-level vision can hardly be the single (or even the main) account for face recognition impairments (Anstey, et al., 2002). Indeed, other studies suggested that face recognition difficulties might be caused by age-related impairments in higher-level perceptual processes, such as the inability to quickly transform perceptual representations into familiar templates (Habak, et al., 2008), reduced neural specialization for faces (Park, et al., 2004; Payer, et al., 2006), impaired categorical perception (Kiffel, et al., 2005), changes in face-scanning strategies (Firestone, et al., 2007), impaired spatiotemporal integration (Del Viva and Agostini, 2007; Norton, et al., 2009), reduced sensory processing speed (Salthouse, 1996; Salthouse, 2000), reduced attention/working memory capacity (Bopp and Verhaeghen, 2005; Lange and Verhaeghen, 2009), augmented distractibility (Healey, et al., 2008; Machado, et al., 2009), impaired pre-frontal inhibitory control (Chao and Knight, 1997; for reviews see Grady, 2008; Reuter-Lorenz, 2002), and reduced cognitive flexibility as evidenced by a general reduction in task switching performance (Reimers and Maylor, 2005). Given this wealth of studies suggesting that impaired face recognition might result from perceptual impairments, it is surprising that only a few studies attempted to directly investigate possible age-related changes in the perceptual processes involved in face recognition.

Four types of information can be extracted from faces and (arguably) involved in different stages of face processing. Since some ambiguity exists regarding their definition we will introduce each type as relevant for the present study. One type of information, which we termed global (Bentin, et al., 2006) refers to the general structure of the face that is sufficient to categorize it at a basic level. It includes the typically symmetric placement of the two eyes on both sides of a vertical axis above the nose and the mouth, all included in an elliptical contour. Ignoring the contour, Maurer and colleagues (2002) defined the face-characteristic global structure only on the basis of the typical arrangement of its inner components and coined this configuration as first-order relations. A second type of information extracted from faces results from what has been labeled holistic processing. Following others (e.g. Tanaka and Farah, 1993), by holistic processing we refer to the integration of the inner components and the face contour leading to the perception of the face as a unitary gestalt. This information seems to be essential during the individuation of the face, as best demonstrated in the composite effect, which shows a prevalence of the whole face structure over its constituent parts (Young, et al., 1987) (but see Richler, et al., 2008, for a different interpretation). The third type of information extracted from faces is featural. Resulting from analysis at a local level (Bentin, et al., 2006) this information refers to the shape of face components such as the roundness of the eyes, the thickness of the eye-brows, the curvature of the lips, etc., and is used in face identification as well as for subordinate level categorization (particularly by race Zhao & Bentin, unpublished). Finally, the fourth type of information processed during face recognition is labeled configural. These are the spatial metric computations relating the inner components one to the other and to the face contour. In Maurer et al.'s (2002) terms these are the second-order relations between the face components that, by many accounts, are most important for the individualization of a face and its identification (e.g. Rhodes, et al., 1993).

With only a notable exception (Boutet and Faubert, 2006), there are no studies that systematically investigated the ability of older participants to apply the perceptual processes necessary for extracting the above described types of information while porcessing faces. Moreover, even in that study the authors addressed only a part of the face-relevant processes. They found that older participants were sensitive to the face-inversion effect (FIE) as much as younger participants. Since the FIE allegedly reflects the reliance of the observer on configural information for face recognition (e.g. Freire, et al., 2000; Searcy and Bartlett, 1996), Boutet and Faubert concluded that the ability to extract configural information is not affected by aging. Similarly, finding an advantage in identifying face components in the context of the whole face than in isolation (cf., Tanaka and Farah, 1993) in older participants, the authors concluded that the ability to apply holistic processing is also spared in the elderly. However, the latter conclusion should be tempered by the fact that the older participants in that study did not show the composite effect, which is also a measure of holistic processing (Young, et al., 1987). The latter deficiency might, in fact, reflect a more general age-related decline in contour integration ability (Roudaia, et al., 2008).

As far as we know there are no studies testing perception of either global or local (features) information in faces in the elderly, but we may infer these abilities from studies that tested aging-related effects on global and local processing of other stimuli, such as hierarchical letters (Navon, 1977).¹ Some studies found an age-related shift from global to local processing precedence (Lux et al., 2008, Oken et al., 1999), and interpreted the relative faster decline of global processing to a narrowed attentional field (cf., Kosslyn et al., 1999). Other studies, however, found preserved global precedence in the normal elderly albeit compared with younger participants, the older participants were overall slower, less accurate (Bruyer and Scailquin, 2000; Bruyer, et al., 2003; Georgiou-Karistianis, et al., 2006 (Experiment 1)), and could not easily switch from one processing level to the other (Georgiou-Karistianis, et al., 2006 (Experiment 2)).

Hence, on the basis of the current literature we might provisionally conclude that face recognition difficulties encountered in the elderly cannot be explained by age-related changes in the ability to use face-specific perceptual strategies and extract the relevant information for face recognition. However, such a conclusion might be premature. First, as evident in the above review not all the perceptual processes that are relevant to face recognition have been properly and systematically tested. Second, although studies showed that face inversion as well as other manipulations affecting holistic and global processing influence the performance of older participants in face recognition tasks like that of younger ones, in most those studies performance was overall slower and less accurate in the older than in the younger groups. Importantly, similar age-related effects were not found (when tested) for the recognition of other objects (Boutet and Faubert, 2006). Given this ambiguity, additional studies seem to be necessary for shedding more light on age-related changes in perceptual processing that may account for specific impairments in face recognition. In particular, it is important to explore face processing online in order to distinguish between face recognition processes, the completion of which might be particularly delayed or impaired, and processes that are not affected by age. Whereas reaction time (RT) and accuracy measures cannot provide an online and continuous measure for face processing, contemporary developments in electrophysiological neural manifestations of face processing demonstrated that event-related potentials (ERPs) could complement the RT studies in this respect.

Scalp-recorded ERPs are time continuous fluctuations of averaged EEG activity, time locked (directly or indirectly) to the onset of an event, such as the appearance of a stimulus in the observer's perceptual field. Ample research identified a series of potentials distinguished by peak latency and scalp distribution and putatively associated these ERP components with different cognitive mechanisms. Among those, relevant to the current study are the N170 and the P300.

The N170 is a negative ERP component normally peaking between 150 and 180 ms from stimulus onset with considerable higher amplitudes in response to faces than to any other stimulus categories and, therefore, has been associated with early perceptual processes elicited by the detection of faces in the visual field (Bentin, et al., 1996; for a recent review see Rossion and Jacques, 2008). This categorical difference between faces and non-face stimuli, labeled the N170-effect, is most conspicuous at posterior-temporal recording sites where its absolute amplitude is maximal. The lateral distribution distinguishes this face-sensitive neural response from the non-specific N1, which peaks at more medial sites in response to any visual stimulus, sensitive to spatial attention but not to the stimulus content (e.g., Clark & Hillyard, 1996).. Finally the N170 is typically larger over right than left hemisphere sites, which goes along with ample neuropsychological evidence for higher involvement of the right hemisphere in face processing.

P300 is a generic name for a variety of relatively late positive components with a centro-parietal or centro-frontal midline distribution (Donchin, 1981; Polich, 2007).² While initially discovered in response to task-relevant oddball (infrequent) stimuli (Sutton, et al., 1965), and found sensitive to the subjective probability assigned to the occurrence of the eliciting event (Duncan-Johnson and Donchin, 1977), many theories about the cognitive mechanism(s) manifested by this neural event have been proposed (Donchin, 1987; Donchin and Coles, 1998; Verleger, 1988; Verleger, et al., 2005). Notwithstanding those controversies, it is unanimously agreed that the P300 latency reflects the length of stimulus evaluation processes when a two-choice RT is required (e.g. Kutas, et al., 1977; McCarthy and Donchin, 1981) and that, other factors being kept constant, its amplitude is largely determined by the stimulus relevance (Gray, et al., 2004), by the amount of attention allocated to the stimulus (Kok, 2001) and by the task complexity (Johnson, 1986).

In Experiment 1 we compared the N170-effect in elderly and younger groups of healthy participants while manipulating stimulus conditions to assess the sensitivity of each group to global, configural and featural processing. In Experiment 2 both the N170 and the P300 potentials were used to compare the efficiency of basic-level categorization of faces (distinguishing faces from cars), subordinate categorization (distinguishing between male and female faces) and individual face categorization (distinguishing between familiar and unfamiliar faces).

General Methods

Participants

Two groups of 16 participants with no history of psychiatric or neurological disorders were defined by age and tested in both experiments. The younger group included undergraduate students (10 females) with an age range between 19 and 33 years (Mean = 24.3). They were all healthy individuals that received either course credit or money payment for their participation. The older included 16 healthy individuals (10 females) with an age range between 70 and 90 years (Mean = 77.1). They live an independent life in the community and were paid for participation.

To ensure normal cognitive abilities they were tested with a Mini Mental State Examination (MMSE - Tombaugh and McIntyre, 1992) and only people who scored 27/30 and up were included in the study. Contrast sensitivity of all participants was tested using the Functional Acuity Contrast Test (Vision Sciences Research Corporation). Both groups were within normal ranges (corrected for age), there was no difference between the younger and older participants at low-spatial frequency (1.5 cycle/degree), while at high frequencies the contrast sensitivity was higher in the younger than in the older participants. Importantly, however, the analysis of the P1 component³ (which is modulated primarily by low-level vision factors) revealed no difference between the groups either for amplitude or latency. If anything the P1 tended to peak higher and earlier in the older than in the younger groups (4.2 μV (108 ms) and 2.5 μV (110 ms) for the older and younger groups, respectively; F(1,22) = 3.6, MSe = 1218.6, P = 0.07, and F(1,22) < 1.0, for amplitude and latency, respectively). Visual acuity was tested for each participant using a Snellen chart (Rosenbaum Pocket Vision Screener) and it was normal or corrected to normal in the older as well as the younger group (equally distributed between 20/20, 20/25 and 20/30). The experiment was approved by the Hebrew University ethical committee and all participants signed an informed consent form.

EEG recording

The EEG analog signals were recorded continuously (from DC up to a low-pass filter set at 100 Hz) by 64 Ag-AgCl pin-type active electrodes mounted on an elastic cap according to the extended 10–20 system (Biosemi™). Two additional electrodes were placed at the right and left mastoids and one on the tip of the nose. All electrodes were referenced during recording to a common-mode signal (CMS) electrode between POz and PO3 and were subsequently re-referenced digitally (see data processing below). Eye movements, as well as blinks, were monitored using bipolar horizontal and vertical EOG derivations via two pairs of electrodes, one pair attached to the external canthi, and the other to the infraorbital and supraorbital regions of the right eye. Both EEG an EOG were digitally amplified and sampled at 256 Hz using a Biosemi Active II system.

Data pre-processing

The data were processed using Brain Vision Analyzer software (Brain Products). Raw EEG data was initially high-pass filtered at 0.5 Hz (24 dB) with a notch filter at 50 Hz and re-referenced off-line to the tip of the nose. Eye movements were corrected using an ICA procedure (Jung, et al., 2000a; Jung, et al., 2000b) and remaining artifacts exceeding transitions of ±100 μV in amplitude during a period of 50 ms at the sites included in the analyses (see below) were rejected. Artifact-free data were then segmented into epochs ranging from −100 ms before to 600 ms after stimulus onset for the N170 analysis and from −100 ms to 1000 ms for the P300 analysis (see Experiment 2).

Experiment 1

The goal of this experiment was to assess possible age-related differences in perceptual processing of faces and its time course as expressed by modulations of the N170-effect during a relatively non demanding, target monitoring task. Notwithstanding current controversies about the particular face-processing mechanism reflected by the N170 (e.g. Carmel and Bentin, 2002; Gauthier and Curby, 2005; Rossion, et al., 2002) or its precise neural origin (Itier and Taylor, 2004), there is no serious argument that this potential is influenced primarily by the stimulus categorical content (mostly faces) rather than its low-level visual quality (see discussion in Rossion and Jacques, 2008). Based on previous studies and particularly on the fact that the amplitude of the N170 is higher for faces than other objects, in the present study we assumed that higher N170 amplitude reflects a better correspondence between the stimulus presented and the sensitivity of the perceptual mechanism involved in categorical distinction of faces. To this end, the N170 is a good candidate to explore possible aging effects on neural activity associated with categorical perception of faces in the visual system.

Previous studies demonstrated that for young participants the N170 is usually larger over the right than the left hemisphere (Bentin, et al., 1996; Rossion, et al., 2003), which is in line with ample evidence for the right-hemisphere superiority in face-processing (Levy, et al., 1972; Yovel, et al., 2008). Suggesting a special sensitivity to face features, the N170 is more robust when the inner components are presented outside of the face contour (Zion-Golumbic and Bentin, 2007) and larger (as well as somewhat delayed) in response to isolated eyes than full faces (Bentin et al., 1996; Itier, et al., 2007). Moreover, similarly large N170 potentials are elicited by intact and scrambled faces when the components are recognizable (Zion-Golumbic and Bentin, 2007), which suggests that the global face structure is not a necessary condition for activating this neural mechanism (albeit it is probably a sufficient condition). The sufficiency of a face global structure is supported by significant N170-effects in response to correctly configured schematic faces⁴ (Sagiv and Bentin, 2001) but not when the inner components of the schematic faces are scrambled (Bentin and Golland, 2002). Finally, the N170 amplitude is consistently delayed (and usually enhanced) by face inversion (Rossion and Gauthier, 2002) and, more importantly, the difference between the N170 elicited by inverted faces and non-face stimuli is significantly reduced. Whereas the enhancement of the N170 amplitude in response to inverted relative to upright faces is difficult to interpret, the delay of the N170 peak and particularly the reduction of the N170-effect might reflect the fact that face-specific computations are not easily applied to inverted faces (for a discussion see Rossion, et al., 2000). Assuming that the N170 manifests the activation of a mechanism apt to extract diagnostic features from faces (Bentin, et al., 1999), the significant reduction of the N170-effect in response to inverted faces suggests that inverted faces are processed like other objects. In other words, inverted faces do not benefit from perceptual processes tuned to extract face-specific diagnostic information that allows the distinction between individual faces. This might explain why inverted faces are not easily identified.

The few previous studies that compared the N170 in different age-groups reported larger N170 potentials in older than younger groups particularly when the faces were unattended (de Fockert, et al., 2009). Moreover, the N170-effect in the elderly is fairly robust, albeit not lateralized to the right hemisphere (Chaby, et al., 2001; Gao, et al., 2009; Pfutze, et al., 2002) and delayed relative to younger controls (Chaby, et al., 2003). Interestingly, the effect of face-inversion on the N170 potential, which is very reliable in the younger population, is absent in older participants (Gao, et al., 2009). Along with the above line of reasoning, the absence of the FIE on the N170 might indicate that elderly people do not automatically apply configural processes while seeing a face. Based on this outcome Gao and colleagues suggested that face identification is not the default level of processing in the elderly.

In addition to their scarcity, none of the existent studies compared the face-perception processes in younger and older participants systematically. To address this absence, in the present experiment, in addition to exploring the N170-effect in the two age groups by comparing ERPs elicited by upright faces to those elicited by wrist-watch panels, we also compared the sensitivity of the perceptual mechanism reflected by the N170 to global face analysis, to face features, and to configural processing. Global processing and the sensitivity to first-order relations were assessed comparing the response to scrambled faces with normally structured faces. The observed absence of difference between these two types of face stimuli in young participants might not replicate itself in elderly participants who might depend more than the young ones on the global face structure. If so, the N170 elicited by scrambled faces in the elderly group might be reduced relatively to the normally configured faces. Featural analysis was assessed comparing the ERPs elicited by isolated components with those elicited by full upright faces. If the aging perceptual system loses some of its ability to process stimulus details, it is possible that when the inner face components are seen outside of a face contour, the stimulus is not immediately categorized as a face. In that case, in contrast to the pattern which has been observed in younger participants, the N170 elicited by inner components outside the face contour should be smaller than that elicited by full faces and this reduction should be more conspicuous when the configuration of the isolated inner components is also scrambled. Finally, the face inversion effect was used to assess the sensitivity to configural processing. Since configural computations obviously require detailed analysis of the inner face components, we assume that elderly participants are less likely to apply this perceptual strategy than younger participants. Consequently, face inversion should not change the processing of the face much in the elderly group, which should result in similar N170 components to upright and inverted faces.

Methods

Stimuli

The initial stimulus set consisted of 80 photographs of different faces, 80 photographs of different watches (W) and 120 photographs of different flowers. The faces were edited to form five face conditions, with 80 stimuli in each condition: Regularly configured full faces (RF), Inverted Faces (IF), Scrambled Faces (SF), that is faces in which the spatial location of the inner components was scrambled, Inner Components (IC), that is, normally configured eyes nose and mouth without the face contour), and Scrambled Inner Components (SIC), that is, two isolated eyes, a nose and a mouth presented in random configuration and without the face contour (see examples of the stimuli in Figure 1).

All stimuli (including watches and flowers) were equated for luminance, so that, on the average, they were matched for luminance and brightness. The stimuli were presented at fixation and, seen from a distance of approximately 70 cm occupied 8.1° × 11.3° in the visual field (10 cm × 14 cm).

Task, Design and Procedure

Like in many previous N170 studies in our laboratory, the task was oddball target monitoring in which stimuli from different experimental categories were presented one after another and participants were requested to press a button each time a flower appeared on the screen. This procedure ensured that all stimulus categories of interest in this study were equally task-relevant; that is, they were all distracters that did not require any action or additional processing after classification. The 600 stimuli were fully randomized and presented in 6 blocks of 100 stimuli each, with a short break between blocks for refreshment.

The experiment was run in an acoustically treated and electrically isolated chamber. Following the mounting of the electrode-cap the participants sat in a comfortable reclining chair and the monitor was raised at their eyes level. Each trial began with a 500 ms long fixation mark, followed by a stimulus exposed for 200 ms. The inter-trial intervals varied between 1400 ms and 1600 ms.

N170 analysis and statistical evaluation

ERPs resulted from averaging the segmented trials separately in each condition. The averaged waveforms were smoothed by applying a low-pass filter of 17 Hz (24 dB) and were baseline-corrected by subtracting the mean amplitude during an epoch starting 100 ms before stimulus onset. The characteristic scalp distribution of the N170 in each condition was estimated by spherical spline interpolations with 4 levels. Based on previous studies and on scrutiny of the present N170 distribution, the statistical analysis was restricted to posterior lateral regions including sites P8, PO8, and P10 over the right hemisphere and the homologue sites over the left (Figure 2). For each subject the peak of the N170 was initially determined at these sites (based on the filtered waveforms) as the most negative peak between 130 and 230 ms. Subsequent visual scrutiny ensured that the most negative values represented real peaks rather than end points of the epoch. The N170 amplitudes at the 3 sites in each region of interest were averaged to determine the latency of the N170 component over each hemisphere. Then we measured again the amplitude of the N170 potential at these latencies separately for each recording site within the two regions of interest. Four major effects were compared within and across groups: (a) Basic sensitivity to faces as a distinct category was assessed comparing the N170 elicited by upright faces (RF) with that elicited by watches (W); (b) Configural processing was assessed by comparing the N170 elicited by upright faces (RF) with that elicited by inverted faces (IF); (c) Global processing (first order) was assessed by comparing the N170 elicited by upright faces (RF) with that elicited by faces with scrambled inner components (SF); (d) Feature processing effects were assessed comparing the N170 elicited by upright faces (RF) with that elicited by inner components presented in a normal configuration (IC) as well as scrambled inner components outside the face contour (SIC).

The 64 recording sites. The N170 was analyzed at the dark-gray locations. The P300 was analyzed at the light-gray locations. The selection of sites was based on previous research in our laboratory as well as many previous published studies.

In addition, since across stimulus types the absolute amplitudes of the N170 were different in different participants and different in the old and young groups, a normalization procedure was necessary to allow a proper comparison of the effect sizes (see also Gao, et al., 2009).⁵ To this end, a normalized N170-effect on the amplitudes was calculated as the ratio (RF − W)/(RF + W). Since in some participants the amplitudes elicited by watches were positive, in order to avoid ratios larger than 1 (which would be meaningless) we elevated the baseline by subtracting the maximum positive amplitude observed across watches and faces and participants from all individual ERPs. Hence the normalized N170-effect ranged between 1 (when the W=0) and −1 (when RF=0). The same procedure was used to calculate the face-inversion effect as (IF−RF)/(IF+RF), the global effect as (RF−SF)/(RF+SF), and features processing effect as (IC−RF)/(IC+RF) and as (IC−SIC)/(IC +SIC).

Mixed-model ANOVAs were used to analyze the N170 amplitude and latency. The between-subjects factor was Group (young, elderly) and the within-subjects repeated factors were Hemisphere and a Stimulus-type factor reflecting the analyzed effect (see results). The analysis of the N170 amplitude also included the Site (P7/8, PO7/8, P9/10) as an additional factor. For factors with more than two levels, the degrees of freedom were corrected for non-sphericity using the Greenhouse-Geisser correction (for simplicity, the uncorrected degrees of freedom and G-G Epsilon (G-GE) are presented).

Results

Both young and older participants were at ceiling in detecting the target category (flowers) and, therefore, we did not analyze performance in this experiment.

Robust N170 components were elicited in both age groups and all experimental conditions. These data were analyzed separately for each a priori-defined effect.

Sensitivity to faces as a distinct category - Amplitude

The data are presented in Figure 3. A Group (old, young) × Stimulus type (face, watch) × Hemisphere (left, right) × Site (P7/8, PO7/8, P9/10) ANOVA showed that all 4 main effects were significant. The N170 amplitude in the elderly group (−10.3 μV) was larger than in the young group (−7.0 μV), F(1,30) = 5.4, MSe = 205.0, p < .05; faces elicited larger N170 (−9.2 μV) than watches (−8.1 μV), F(1,30) = 24.9, MSe = 17.2, p < .001; the amplitudes were larger at right (−9.2 μV) than at left hemisphere sites (−8.1 μV), F(1,30) = 4.6, MSe = 28.2, p < .05; and the N170 amplitude differed across sites F(2,60) = 3.2, MSe = 14.5, p < .5. Post hoc contrasts showed that the Site effect was explained by bigger amplitudes at P9/10 (8.9 μV) and PO7/8 (−8.9 μV) than at P7/8 (−8.0 μV). Importantly, all the within-subject main effects were qualified by interactions with Group [F(1,30) = 4.9, MSe= 17.2, p < .05, F(1,30) = 5.1, MSe = 28.2, p < .05; F(2,60) =4.0, MSe = 14.5, p < .025, for Stimulus type, Hemisphere, and Site, respectively]. These interactions were further explored by separate analyses for each age group. These analyses revealed that: (a) In both groups faces elicited larger N170 than watches, but the difference was larger in the elderly group [3.0 μV; F(1, 15) = 30.2, MSe = 14.7 p < .001] than in the younger group [1.17 μV; F(1, 15) = 3.4, MSe = 19.6, p = .08]; (b) In the younger group the N170 was larger at right hemisphere (−8.1 μV) than at left hemisphere sites [−5.8 μV; F(1,15) = 12. 6, MSe = 21.7, p < .005], but there was no inter-hemispheric difference in the older group [−10.4 μV and −10.3 μV, for the left and right hemispheres, respectively; F(1,15) < 1.0]; (c) The main effect of Site was significant in the older group, F(2,30) = 5.3, MSe = 14.0, p < .025, but not in the younger group, F(2,30) = 2.0, MSe = 15.4 p = .15; (d) Finally t-tests revealed that although the N170 amplitudes were larger for elderly than for younger participants in response to both faces and watches the difference was significant for faces (−4.3 μV; t(30) = 2.621, p < .025) but only tended to be so for watches (−2.4 μV; t(30) = 1. 78,3 p = .085).

ERPs elicited at right and left posterior temporal sites in the elderly and younger groups. (A) The N170 at P9 and P10 elicited by upright faces and watches. (B) The Current Source Density (CSD) seen from the back and scalp voltage distribution of the N170 seen from above. Both distributions and CSD were calculated at the peak latency of the N170 at P10 and P9 in each group Notice the more symmetrical sources and larger amplitudes in the elderly relative to the younger groups. (C) The normalized N170 effect (face – watches, see text). Notice that the difference between face and watches is actually slightly larger in the elderly group.

The analysis of the normalized N170 effect showed that this effect was numerically larger in the older than in the younger group albeit not significantly so [0.09 μV and 0.04 μV for older and younger participants, respectively; F(1,30) = 3.1, MSe = 0.33, p = .088]. There was no significant hemisphere difference [F(1,30) = 1.1]. Across groups, the N170-effect was larger at P9/10 (0.09 μV) and P7/8 (0.06 μV) than at the more medial sites PO7/8 [0.05 μV; F(2, 60) = 7.0, MSe = 0.08, p < .0025].

Sensitivity to faces as a distinct category – Latency

Across the other conditions the N170 peaked later for the older group (181 ms) than for the younger group [165 ms; F(1,30) = 8.0, MSe = 1063, p < .01]. Across groups and hemispheres the N170 latency elicited by faces was shorter (166 ms) than that elicited by watches [179 ms; F (1,30) = 38.6, MSe = 143, p < .001]. Importantly, there was no interaction between stimulus type and group, [F(1, 30) < 1], indicating that the stimulus type effect was similar for young and elderly subjects. Across stimuli and groups the N170 recorded at left hemisphere sites (173 ms) was similar to that recorded at the right hemisphere sites (172 ms) [F(1,30) < 1.0] and the Hemisphere effect did not interact with Stimulus type [F(1,30) < 1.0]. There were no significant interactions.

Summary of the face sensitivity effects

The present results suggest that the basic ability to detect faces as revealed by the N170 is similar for older and younger participants, albeit it takes longer in the older group. For both groups faces elicited higher amplitudes and shorter latencies than watches. In line with previous studies, across stimuli the N170 amplitudes were larger in the elderly group than in the younger group (albeit the difference was significant only for faces). Whereas the N170 amplitudes were higher at right than left hemisphere sites for young participants, no hemispheric differences were found for the elderly participants. In particular, it is important to notice that in addition to being significant in both groups the normalized N170 effect and its distribution across sites and hemispheres were similar for older and younger participants.

Configural processing - Amplitude

These data are presented in Figure 4. Despite the observed overall amplitude differences of the N170 between old (−11.1 μV) and young (−9.7 μV), this difference failed to reach significance, though a tendency is apparent [F(1, 30) = 3.1, MSe, = 273.8, p = .09]. Across groups and sites the N170 elicited by up-right faces was smaller (−9.7 μV) than by inverted faces [−11.0 μV; F (1.30) = 12.3, MSe = 13.8, p < .001 and, importantly, the face inversion effect significantly interacted with group [F(1,30) = 13.1, MSe = 13.8, p <.001]. Separate ANOVAs for each group revealed that inversion enhanced the amplitude of the N170 significantly in the young group [−7.5μV and −10.3 μV for upright and inverted faces, respectively; F(1,15) = 20.8, MSe = 16.8, p < .001], but not in the old group [−11.8 μV and −11.9 μV for upright and inverted faces, respectively; F(1, 15) < 1].

ERPs elicited by inverted faces and upright faces in the two age groups. (A) The N170 components. Notice the slight delay of the N170 peak in response to inverted faces, the conspicuous amplitude differences between the two conditions in the young group and the absence of such a difference in the elderly group. (B) The Current Source Density (CSD) seen from the back and scalp voltage distribution of the N170 in response to *inverted* faces seen from above. Notice that, as for upright faces, the distribution of the response to inverted faces in the elderly group is symmetrical across hemispheres. (C) The normalized face-inversion effect (inverted – upright, see text) in the two age groups. Notice the absence of a face inversion effect in the elderly group.

Across stimuli and groups the N170 amplitude was larger at right hemisphere (−11.1 μV) than left hemisphere sites [−9.7 μV; F(1,30) = 4.6, MSe = 41.3, p <.05]. The laterality effect interacted with Stimulus type [F(1,30) = 5.4, MSe = 2.9, p < .05], and, as in the face detection analysis, the inter-hemispheric difference interacted with group [F(1,30) = 4.1, MSe = 41.3, p = .05]. Separate ANOVAs for each age-group revealed that across face orientation the N170 elicited in the elderly group was similar at left hemisphere (−11.8 μV) and right hemisphere sites (−11.9 μV); F(1, 15) < 1]. In contrast, for the younger participants, the N170 was larger at right hemisphere (−10.3 μV) than at left hemisphere sites [−7.5 μV); F(1, 15) = 11.4, MSe = 31.7, p < .005].

The N170 was different at the different sites [F(2,60) = 3.5, MSe = 25.2, p < .05]. Post hoc contrasts showed that it was significantly larger at P9/P10 (−11.0 μV) and PO7/8 (−10.5 μV) than at the more medial sites P7/P8 (−9.6 μV; p < .005). There was also an interaction between Site and Group [F(2, 60) = 3.7, MSe = 25.2, p < .05]. For the older group the largest amplitude was at PO7PO8 sites (−12.7μV), while for the younger group the largest amplitude was at P9/P10 (−10.2 μV). For the older group the lowest amplitude was at P7/P8 (−10.9 μV), while for the young group it was at PO7/PO8 (−8.1 μV) sites. There were no other significant interactions.

The analysis of the normalized inversion effect showed that it was significantly higher for younger participants (.07 μV) than for older participants [−0.01 μV; F(1,30) = 10.7, MSe = .031, p < .005]. No other main effects or interactions were found.

Configural processing - Latency

Across stimulus orientation, the N170 peak was delayed in older people (180 ms) relative to the younger participants [163 ms; F(1,30) = 8.1, MSe = 1028, p < .01] and, across groups, the N170 peak was delayed by face inversion [177 ms and 166 ms for inverted and upright faces, respectively; F(1,30) = 126.8, MSe = 28.5, p < .001]. There was no interaction between the face inversion effect and group [F(1, 30) = 1.0], indicating that the stimulus type effect was similar for young and elderly subjects.

Across stimuli and groups the latency of the N170 at left hemisphere sites (172 ms) was not significantly later than at right hemisphere sites [171 Ms; F(1,30) < 1.0] and there were no significant interactions

Summary of the configural processing effects

Like shown in previous studies, face inversion increased the amplitude and delayed the latency of the N170. However, this effect differed between young and old participants: Whereas face inversion significantly delayed the N170 peak similarly for both age groups, the amplitude of the N170 was augmented by face inversion in the younger group but not in the older group. Indeed, the normalized N170 effect on amplitudes was significant only for the younger group and, obviously, higher than the normalized face-inversion effect in the older group (Figure 4).

Global processing effects- amplitude

As evident in Figure 5, across scrambling conditions the amplitude of the N170 was larger for the elderly group (−11.5 μV) than for the younger group [−7.9 μV; F(1, 30) = 5.7, MSe = 213, p < .025]. Although as a main effect the N170 was not affected by scrambling the inner components in the face [F(1,30) < 1.0], a significant Stimulus type × Group interaction [F(1,30) = 5.1, MSe = 11.0, p < .05] revealed that this manipulation influenced the two groups differently. Separate ANOVA for each group showed that destroying the global structure reduced the N170 amplitude in the elderly group [−11.9 μV and −11.1 μV, for normally configured and scrambled faces, respectively; F(1,15) = 3.5, MSe = 7.5, p = .07], but slightly increased it in the younger group [−7.5 μV and −8.3 μV, for normally configured and scrambled faces, respectively; F(1,15) = 2.1, MSe = 14.5, p = .170].

ERP elicited by scrambled faces compared with the upright, intact faces in the two groups. (A) The N170 components. Notice the slightly larger N170 amplitudes in response to intact relative to scrambled faces in the elderly group and the inverse tendency in the younger group. (B) The Current Source Density (CSD) seen from the back and scalp voltage distribution of the N170 in response to *scrambled* faces seen from above. Notice that, as for intact faces, the distribution of the N170 to scrambled faces in the elderly group is symmetrical across hemispheres. (C) The effect of destroying the global shape of the faces (scrambled – intact, see text). Notice the opposite trends in the young and elderly groups.

As in the previous comparisons, the N170 amplitude was larger over right hemisphere than left hemisphere sites in the younger group [−9.0 μV and −6.9 μV, respectively, F(1,15) = 9.8, MSe = 22.8, p < .01], while no such asymmetry was found in the elderly group [11.5 μV for both hemispheres; F(1,15) < 1.0]. In addition, there was a small effect of Site for both groups [F(2,30) = 3.5, MSe = 21.5, p < .05 and F(2,30) = 3.1, MSe = 16.0, p = 0.6, for the younger and older groups, respectively. As before, for the younger participants the N170 was largest at the most lateral sites P9/P10 (−8.9 μV) and smallest at the more medial sites PO7/PO8 (−7.3 μV); in contrast, for the elderly, the largest N170 amplitudes were recorded at the more medial sites (−12.3 μV). Neither the Hemisphere nor the Site effects interacted with Stimulus Type for any of the age-groups.

Global processing - latency

The N170 latency was longer in the elderly group (177 ms) than in the younger group [164 ms; F(1,30) = 4.9, MSe = 1076, p < .05] and, across groups, it was longer for scrambled (174 ms) than normally configured faces [166 ms; F(1,30) = 28.8, MSe = 73.1, p < .001]. A significant Stimulus type × Group interaction followed by separate ANOVA in each group showed that the N170 latency was significantly delayed by scrambling the face in both groups [F(1,15) =15.0, MSe = 16.3, p < .005 and F(1,15) =18.7, MSe = 130.7, p < .001, for the older and younger groups, respectively], but the delay was larger in the younger group (12 ms) than in the older group (3.9 ms). Across groups there was no effect of Hemisphere [F(1,3) < 1.0], and no Group × Hemisphere interaction [F(1,30) = 1.4, MSe = 181, p = .240]. Also, there was no interaction between Hemisphere and Stimulus type [F(1,30) <1.0], and no second order interaction (F(1,30) < 1.0].

Summary of the global processing effects

The disruption of the global face structure reduced the N170 amplitude in the elderly group but had no effect on the younger group. The latency of the N170 was delayed by scrambling the inner components within the face in both age groups, but more so in the younger one. There were no inter-hemispheric differences.

Feature processing - amplitude

The N170 elicited by inner face components and the normalized feature-processing effects are presented in Figure 6. Evidently, separating the features from the full global structure had a different effect in the younger and older groups. This pattern was analyzed by a mixed model ANOVA as above except that the Stimulus Type factor included three levels, RF, IC, and SIC. This ANOVA revealed that the main effect of Group was not significant [F(1,30) = 1.5, MSe = 325.6, p = .233], but it significantly interacted with all other main effects [F(2,60) = 18.5, MSe = 20.0, p <.001, F(1,30) = 4.9, MSe = 47.9, p < .05 and F(2,60) = 3.6, MSe = 30.1, p < .05, for Stimulus Type, Hemisphere and Site, respectively]. These interactions were explored post hoc by separate ANOVAs within each group.

(A) ERPs elicited by normally configured and scrambled inner components compared with full (upright) faces. (B) The effect of excluding the contour from the faces (inner components – full faces, see text). (C) The effect of presenting spatially scrambled inner components (scrambled inner components – full faces). Notice the absence of an effect of eliminating the face contour from faces (indeed, the slightly reduced N170 amplitudes for inner components in the elderly group and the robust dissociation between the effects of presenting scrambled inner components to the two groups.

In the younger group, the Stimulus type effect on the N170 amplitude was significant [F(2,30) = 14.8, MSe = 20.9, p < .001], reflecting that both inner components (−10.2 μV) and scrambled inner components (−10.1 μV) were larger than full faces (−7.5 μV; for both comparisons p < .005), and there was no effect of scrambling (p = 1.0). The main effect of Stimulus type was significant also in the elderly group [F(2,30) = 10.4, MSe = 21.0, p < .0025]. However, in contrast to the younger group, in the elderly group the N170 elicited by inner components was the same as that elicited by full faces (−11.8 μV), and scrambling of the components significantly reduced the amplitude [−9.6 μV; p < .025]. As in previous analyses, in the younger group the N170 was larger at right hemisphere (−10.6 μV) than at left hemisphere sites [−8.0 μV; F(1,15) = 14.0, MSe = 35.5, p < .0025]. In contrast there was no inter-hemispheric asymmetry in the elderly group [−11.1 μV and −11.2 μV for the left and right hemispheres, respectively; F(1,15) < 1.0]. Finally, as previously, the Site effect was significant in both groups [F(2,30) = 3.8. MSe = 38.7, p < .05 and F(2,30) = 3.4, MSe = 22.6, p <.05 for the younger and older, groups, respectively], reflecting , however, a different distribution. Again, replicating the results of the previous comparisons in the younger group the N170 was largest at P9/10 (−10.4 μV; p < .05) than at either P7/8 or PO7/8 (−8.7 μV at both sites). In the older, group, however, the largest N170 amplitude was more medially, at PO7/8 (−11.9 μV) while there was no difference between the lateral sites P9/10 and P7/8 (−10.3 μV and −11.1 μV, respectively). None of the interactions were significant either as second order or as first order in the separate ANOVAs.

Feature processing - latency

As in all previous analyses the N170 peaked later in the elderly group (186 ms) than in the younger group [171 ms; F(1,30) = 5.7, MSe = 1883, p < .025]. There was a main effect of Stimulus type (F(2,60) = 82.6, MSe = 112.2, p < 001]. Which revealed that the latency of the N170 elicited by isolated inner components (182 ms) peaked later that that elicited by full faces (166 ms), and scrambling the inner components delayed the N170 peak even more (189 ms). However, in contrast to the amplitude, there was no Stimulus type × Group interaction [F(2,60) < 1.0]. The main effect of hemisphere was not significant (F(1,30) < 1.0). None of the other interactions were significant (all F values < 1.0).

Summary of the features processing effects

The effects of isolating and scrambling the inner components had conspicuously different patterns in elderly and younger groups. For younger participants the isolation of inner components yielded a larger N170 peak than full faces without further sensitivity to their spatial organization. In contrast, for the older participants the N170 amplitude elicited by isolated components was similar to that elicited by full faces if their spatial organization was normal and reduced if the isolate components were scrambled. The latency of the N170 was delayed by isolating the components in both groups, but only in the younger group this delay was reduced when the spatial organization of inner components was normal.

Discussion

The present experiment examined age-related differences in the application of face-characteristic perceptual strategies and their time course as revealed by the N170 component. The N170 was robust for the older as well as the younger participants, in fact larger for the former than the latter. The N170 peaks were delayed in the elderly for both faces and objects but, nevertheless, the N170-effect was similarly conspicuous in both groups. This pattern suggests that although the speed of visual processing is probably reduced by aging, the ability of the perceptual system to distinguish faces from other visual stimuli is preserved. To this end, the present data do not support an overall loss of neural specialization in the ventral extrastriate system as suggested by fMRI studies (Park, et al., 2004). Face inversion delayed the N170 peak similarly across groups but the older participants were apparently less sensitive than younger ones to face inversion, as suggested by the absence of the FIE effect on the N170 amplitude both when it was measured directly or when it was individually adjusted and normalized. To the extent that the FIE reflects, on the one hand, the impaired ability to apply configural computations to inverted faces and, on the other hand, the imposed focus on details, the absence of the N170 FIE effect in the elderly suggests that the neural mechanism which putatively sustains configural processing is not as efficient in the elderly. In other words, we interpret that the absence of the FIE on the N170 as evidence that older participants were less likely to apply configural computations even when faces were presented upright. Note that this interpretation does not preclude the possibility that detailed processing of face features is also impaired (see below). In fact, it stands to reason that the ability to process individual features is a pre-requisite of the ability to compute their relative spatial locations. Hence, it is possible that the apparently reduced tendency of elderly people to apply configural computations while processing faces is a result of a more basic impairment in processing details.

In contrast to the FIE, older participants were more sensitive than the younger ones to the global structure of the face, as disrupting this structure reduced the N170 in the older but not in the younger group. Yet, this manipulation delayed the N170 peak similarly in both groups. An enhanced reliance on global shapes might also reflect problems with processing details. Indeed, the current results revealed that older participants are particularly impaired in detecting faces when the physiognomic value of the stimuli is revealed only by components in the absence of the global face structure. This impairment is indicated by the considerable reduction of the N170 amplitude in the elderly group when the contour of the face was eliminated, and even more so, when the normal configuration of the isolated inner components was also disrupted. Note that presenting face components outside of the face contour augmented the N170 in the younger group, both when the components kept their normal face-like configuration and when they were spatially scrambled. The increase in the N170 amplitude in response to isolated face components has been amply documented in young participants (Itier, et al., 2007; Itier, et al., 2006; Zion-Golumbic and Bentin, 2007). This pattern suggests that young observers discover the physiognomic value of face components independently of their first-order relations. In contrast, it seems that in the elderly first-order relations (i.e., the global structure of the face) are more essential for detecting a face and for triggering face-characteristic perception mechanisms. It is interesting to consider this trend in light of the apparently increased involvement of the left hemisphere in processing faces by the elderly participants: As evident in Figure 3, the symmetrical distribution of the N170 across hemispheres in the elderly was not caused by a reduction of the amplitudes at RH sites but rather an enhancement of the amplitudes at LH sites (relative to the younger group). Given the traditional view that assigns detailed analysis to the left hemisphere and global analysis to the right (e.g. Bradshaw and Nettleton, 1983, Chapter 9; see also Hubner and Volberg, 2005; Martinez, et al., 1997; Weissman and Woldorff, 2005), the apparently greater involvement of the LH in processing faces is incompatible with our finding that elderly people rely more on global than local strategies while processing faces. It is also incompatible with studies that found aging-induced difficulties to attend to or process local elements (Pesce, et al., 2005).

On the other hand, the greater activation of the LH is compatible with studies that found local precedence in the elderly while processing Navon letters (Lux et al, 2008; Oken et al., 1999). A possible account for the seeming contradiction between the excessive reliance on the global shape while processing faces and local-processing precedence while discriminating between Navon hierarchical letters, is that aging entails a reduced cooperation between the hemispheres. Hence the verbal value of letters involves predominantly left hemisphere local processing strategies, while the right hemisphere lateralization for faces involves predominantly global processing strategies, without the ability to efficiently switch between strategies or compensate for a particular hemisphere-bound weakness. Although this hypothesis is post hoc and obviously calls for research aimed to testing inter-hemispheric transfer in the elderly, it is partly supported by studies that showed difficulties to switch between levels of processing in the elderly (Georgiou-Karistianis, et al., 2006). It also resonates findings showing reduced white-matter connectivity in the right hemisphere (Thomas et al., 2008). To this end, it is possible that the enhancement of the N170 amplitude at LH sites and, consequently, its symmetrical distribution, might not be related to face processing; in fact, a similar pattern has been also found while recording simple pattern-reversal evoked potentials (De Sanctis, et al., 2008).

To summarize, notwithstanding the above caveat, the symmetrical distribution of the N170 across the hemispheres in the elderly, which contrasts with the typical right hemisphere lateralization of the N170 evident in the younger group of participants, resonates with previous studies demonstrating reduced hemispheric asymmetry in older people (Cabeza, 2002; Dolcos, et al., 2002; Gao, et al., 2009; Pfutze, et al., 2002). The enhanced involvement of the left hemisphere in face processing goes along with the age-related right-hemisphere decline hypothesis (Bentin, et al., 1978; Brown and Jaffe, 1975; Cherry, et al., 2005). In addition, the results of Experiment 1 support previous studies indicating an overall ageing-related slowing of perceptual speed, which was found in performance (e.g. Salthouse, 1996; Salthouse, 2000) as well as in electrophysiological measurements (De Sanctis, et al., 2008; Rousselet, et al., 2009; Tobimatsu, 1995). The presently discovered age-related changes in face processing could partly account for the observed reduction in face recognition efficiency. In Experiment 2 we explored how these perceptual changes affect the categorization of faces at different levels.

Experiment 2

Following the pattern of age-related differences in the application of face-characteristic perceptual strategies observed in Experiment 1, the goal of the present experiment was to investigate if and how these differences affect the neural/electrophysiological manifestations that are associated with the categorization of faces at basic, subordinate and individual levels. To reach this goal, ERPs were recorded while the older and the younger participants were requested to distinguish between faces and cars (basic level categorization), between female and male faces (subordinate level categorization) and between famous and unfamiliar faces (individual level categorization). We focused on the P300 as well as on the N170 ERP component.

As mentioned in the general introduction, the latency of the P300 reflects the stimulus evaluation time (Kutas, et al., 1977) and it is independent of the participant response time, which in the elderly might be tainted by motor sluggishness (Duncan-Johnson, 1981; McCarthy and Donchin, 1981). The amplitude of the P300 is modulated by many factors, such as the probability of a stimulus from a particular category to occur (for a review see Donchin and Coles, 1988), mental workload, resource allocation and processing capacity (Donchin, 1987; Isreal, et al., 1980; for a review see Kok, 2001). A methodological but important factor that also affects the P300 amplitude is the inter-trial latency variability (“latency jitter”). Note, however, that the latency jitter is not without interest since it may correlate with task complexity and the ability to sustain attention. Indeed, it has been suggested that the latency jitter increases with age (McDowell, et al., 2003). Therefore, whereas the N170 reflects relative early perceptual processes, the P300 should unveil possible age-related changes in higher-level categorization of faces as well as decision making processes.

There are no previous studies investigating possible age effects on face-categorization at different levels. However, a large number of studies compared the P300 component elicited in younger and older groups and in most of these studies the participants made categorical decisions. The great majority of these studies documented normal ageing-related delayed latencies in a variety of auditory, visual and somato-sensory oddball detection paradigms (Brown, et al., 1983; Goodin, et al., 1978; Picton, et al., 1984; for a meta-analysis of 32 studies see Polich, 1996). The effects of aging on the P300 amplitude are not as consistent. Whereas some studies found lower amplitudes in older than younger groups (Pfefferbaum and Ford, 1988; Polich, 1997; Walhovd, et al., 2008), other studies did not find any amplitude modulations (Pfefferbaum, et al., 1984). Still other studies found reduced P300 amplitudes only in low-performing participants (Lorenzo-Lopez, et al., 2007). Some of this ambiguity might be/ accounted for by the level of cognitive performance of the older participants (see also Walhovd and Fjell, 2003). Moreover, additional confusion might stem from age-related changes in the scalp distribution of the P300 that have been consistently found. Most of these studies described a more pronounced and faster decline of the P300 amplitude recorded at posterior relative to frontal sites (Fjell and Walhovd, 2001), or even enhancement of the P300 amplitude at frontal sites in the elderly (Friedman, 2003). This pattern was explained as a non-selective involvement of frontal mechanisms when it should be typically inhibited (Friedman, et al., 1997; Lorenzo-Lopez, et al., 2008). Other studies, however, showed a reversed pattern with a more pronounced P300 reduction frontally (Anderer, et al., 2003).

The equivocal pattern of age-related changes in the amplitude of P300 is consistent with the multitude of factors that affect this measure, as well as with findings showing that this component combines the activity of several intracranial sources, each putatively associated with a slightly different cognitive system (Anderer, et al., 2003; Lorenzo-Lopez, et al., 2008). Nevertheless, regardless of age, the amplitude of the P300 should be reduced as a function of task complexity. Therefore, we predicted that for both younger and older participants the amplitude of the P300 would decrease from the basic level (face-car) to the subordinate level (male-female) and would be smallest in the individual-level categorization task (famous-unfamiliar). We also predicted that the latency of the P300 would increase in parallel with the decrease in amplitude and that the latency delay would be more pronounced in the older than the younger participants.

Methods

Stimuli

The stimulus set consisted of 500 photographs of different faces and 100 photographs of cars (photographed from an angle of 45°). The faces consisted of 200 unfamiliar male faces, 200 unfamiliar female faces, and 100 famous faces (48 males). The fame of the face was determined in a pilot survey which included participants of various ages rated the familiarity of 400 putatively famous faces downloaded from public internet websites. The famous faces selected for this experiment were recognized by at least 90% of the participants and consisted of political figures and TV personalities. Importantly, when debriefed at the end of the experiment, the elderly as well as the young participants recognized the names of at least 85% of the famous people used in this study, proving existence of semantic knowledge about them. All faces presented in the gender categorization task were without hair and edited to exclude any external visual cues except the inner face components. The use of cropped stimuli forced the participants to distinguish between male and female faces only on the basis of face physiognomy. The stimuli were presented in a range of gray tones and equated for contrast and luminance (see examples in Figure 7). All stimuli were presented at fixation and seen from a distance of 70 ms each face occupied 8.1° × 11.3° in the visual field (10 cm × 14 cm).

Task, Design and Procedure

Three categorization tasks were designed to tap three levels of categorization. Each task presented two categories with 100 exemplars in each. The 200 stimuli were fully randomized and presented one at a time. An alternative response key was a priori mapped to the each category and the participant was instructed to select and press a key in each trial according to the category to which the exemplar belonged. At the basic level categorization the distinction was between (unfamiliar) faces and cars. The subordinate level categorization was between (unfamiliar) male and female faces. The individual level categorization was between famous faces (politicians, movie stars and TV personalities) and unfamiliar faces. The categorization tasks were presented in fixed order from basic to individual level. The reason for using a fixed order was to avoid subordinate categorization carry-over strategies that could have been automatically imposed on higher-level categorization, albeit not required. Note that the inverse effects are prevented by the task.

Each trial began with a 500 ms fixation mark, followed by a stimulus exposed for 200 ms. The SOA to the subsequent trial varied between 1600 ms and 1800 ms, during which the response was expected. Stimuli within level-of-categorization task were fully randomized and a short break was allowed after 100 stimuli. The participants were instructed to consider both speed and accuracy, but accuracy was emphasized more. All other procedures were identical to Experiment 1.