Daily-life contact affects the own-age bias and neural correlates of face memory in elderly participants

Abstract

Previous studies revealed consistently enhanced recognition memory for same- as compared to other-age faces (own-age bias, OAB) in young adults, but inconsistent results in elderly participants. To resolve these discrepancies, we examined recognition memory and event-related potentials (ERPs) for young and old faces in young participants and two elderly groups, who either reported high or low degrees of daily contact with elderly relative to younger persons. As expected, young adults showed more accurate memory for young versus old faces. While no OAB was found in old/low contact participants, old/high contact participants were more accurate with old versus young faces. ERPs in young adults revealed a parietal old/new effect from 500 to 800 ms (hits>correct rejections) for young but not old faces. Whereas no old/new effect was seen in the old/low contact group, the old/high contact participants exhibited a prominent reversed old/new effect (hits<correct rejections) for old faces. These results suggest that contact may account for earlier discrepant results with regard to the OAB in elderly participants. A behavioral OAB in elderly participants may depend on high degrees of contact towards old people. The finding of ERP old/new effects specific to own-age faces suggests enhanced recollection of study phase detail in young participants, whereas it may reflect increased engagement in processes aiming at compensating for a deficit in recollection in elderly participants.

Introduction

Although humans are often considered to be experts in face processing, recognition memory for different classes of faces differs substantially. For instance, humans more accurately remember faces from their own as compared to another ethnic group (see e.g., Malpass and Kravitz, 1969, Meissner and Brigham, 2001), a phenomenon known as the own-race bias. Analogously, several studies suggest better memory for faces of the viewer's own age group (e.g., Bartlett and Leslie, 1986, Fulton and Bartlett, 1991, Harrison and Hole, 2009).

Different theoretical accounts have been proposed to explain the own-race bias. Importantly, as these accounts all understand this effect as but one example of a more fundamental phenomenon, they all can be similarly applied to explain the own-age bias (OAB, see e.g., Hugenberg et al., 2010, Valentine and Endo, 1992). First, it has been proposed that increased contact and/or perceptual expertise with faces of one's own group results in better recognition memory (see e.g., Chiroro and Valentine, 1995, Furl et al., 2002, Tanaka and Pierce, 2009). Accordingly, a relatively small degree of lifetime experience with other-race faces may be responsible for less accurate representations. Thus, it has been suggested that the face representational system (often referred to as multidimensional face space) is optimized to discriminate between individual own- but not other-race faces, due to limited perceptual expertise with this latter face category (Byatt and Rhodes, 2004, Furl et al., 2002, Papesh and Goldinger, 2010, Valentine, 1991, Valentine and Endo, 1992). In addition, decreased expertise or contact may result in less efficient processing of other-race faces. More specifically, several studies demonstrated less efficient holistic processing (i.e., the integration of individual facial features into a Gestalt-like representation) and second-order configural processing (i.e., the analysis of metric distances between facial features) for other-race relative to own-race faces, which in turn is suggested to result in less accurate memory (configural/holistic processing account; see Hancock and Rhodes, 2008, Michel et al., 2006, Rhodes et al., 1989, Rhodes et al., 2006). Second, socio-cognitive accounts suggest that the categorization of a face as belonging to the viewer's social “in-group” or “out-group” drives the occurrence of memory biases. Although group membership and contact may often be confounded, experimental research has been taken to suggest that even when the degree of contact is kept constant, assignment of a specific face to an observer’s in-group yields better memory compared to when the same face is assigned to an out-group (Bernstein et al., 2007, Hugenberg et al., 2010, Shriver et al., 2008). Accordingly, categorizing other-age faces as belonging to an age-related “out-group” (and not contact or perceptual expertise) may drive the own-age bias.

Importantly, these types of accounts make different predictions with respect to the occurrence of an OAB in elderly participants. First, if lifetime perceptual expertise or contact over long periods of time (i.e., in the range of decades) is crucial, one would predict an OAB in young, but not in elderly participants (Fulton & Bartlett, 1991). According to this view, young participants likely have preferential contact to, and thus more expertise with young faces. By contrast, elderly participants have been young in the course of their lifetime, and thus have accumulated substantial expertise with different age groups. Second, amount or quality of recent daily-life contact, rather than lifetime expertise, may be crucial. If so, an OAB should occur in both young and elderly participants alike, provided that both groups exhibit preferential contact towards own-age people, which is in line with a very recent study (He, Ebner, & Johnson, 2011). Similarly, recent research suggests an OAB in terms of less accurate memory for children's compared to young adult faces in young adult observers (Harrison and Hole, 2009, Hills and Lewis, 2011), although the interpretation of those results may be complicated by developmental factors influencing face encoding in the first two decades of life (see e.g., Crookes & McKone, 2009). Third, while socio-cognitive accounts also predict an OAB in both young and elderly groups, provided that both categorize own-age faces as belonging to a social “in-group”, this OAB should be independent of contact and/or expertise.

Empirically, discrepant results were reported with respect to an OAB in elderly participants. While some studies did not detect this effect (Bartlett and Leslie, 1986, Fulton and Bartlett, 1991, Mason, 1986), others reported a significant OAB in the elderly. Specifically, Bäckman (1991) reported more accurate memory for old as compared to young adult faces in participants aged between 63 and 70 years. In this very same experiment, however, two older groups (aged 76 and 85 years, respectively) did not show an OAB. More recently, Anastasi and Rhodes (2005) reported superior memory for old as compared to children's faces in elderly participants. Finally, two additional studies with young and old adult faces also reported corresponding effects (Anastasi and Rhodes, 2006, Lamont et al., 2005).

Independent of facial age, cognitive aging may substantially affect face memory (see e.g., Bäckman, 1991, Bartlett et al., 1989). Critically, lower levels of overall performance pose a methodological problem for studies examining the OAB, because its absence in elderly participants may be related to near-floor performance. As a possible example, Bäckman (1991) reported both decreasing overall performance across three consecutive age groups, and a significant OAB in the youngest group of elderly participants only. One solution for studies investigating the OAB in young and elderly participants may thus be to utilize a measure that is corrected for overall performance differences between groups.

While measures of performance can only capture the end-product of a number of sub-processes, event-related potentials (ERPs) provide more detailed information about neural mechanisms underlying these sub-processes, with precise time resolution. In the context of studies on face memory biases, ERPs have provided important information about the mechanisms underlying the own-race bias (Herzmann et al., 2011, Lucas et al., 2011, Stahl et al., 2010) and the OAB (Wiese, 2012, Wiese et al., 2008; see also Ebner, He, Fichtenholtz, McCarthy, & Johnson, 2011). In general, this research has provided evidence for modulations by face group (i) of perceptual processing (indicated by amplitude effects on early occipito-temporal N170 and P2 components), (ii) of the access to facial representations for repeated faces (indicated by enhanced occipito-temporal N250 responses), and (iii) for a modulation of episodic memory retrieval (indicated by later centro-parietal ERP old/new effects). In the present study we assessed these ERP responses in order to understand the mechanisms mediating any memory differences between young and old faces in two groups of elderly participants, who differed with respect to relative contact towards young and old people.

The earliest ERP component typically observed in face processing studies is the P1, a positive peak maximal at occipital channels between 80 and 120 ms following the presentation of visual stimuli in general. P1 amplitude and latency are sensitive to physical stimulus characteristics, such as brightness, contrast, and spatial frequency information (see e.g., Luck, 2005). Studies on pattern-reversal visually evoked potentials demonstrated smaller and delayed P1 responses in elderly participants (for a review, see Spear, 1993), which is particularly prominent for high spatial frequencies (Celesia et al., 1987, Sokol et al., 1981, Spear, 1993). Given its functional characteristics, P1 is presumably not related to the OAB, but might depict effects of facial age, as old relative to young faces contain more high frequency information due to less smooth skin texture and wrinkles.

Compared to a range of other object categories, faces elicit a prominent occipito-temporal negative wave around 170 ms (N170; see e.g., Bentin et al., 1996, Eimer, 2000a, Itier and Taylor, 2004, Rossion and Jacques, 2008), which is often thought to reflect those early perceptual processing stages referred to as structural encoding in cognitive models of face perception (such as the model by Bruce & Young, 1986; see Bentin & Deouell, 2000) or the detection of a face-like pattern (Amihai et al., 2011, Schweinberger and Burton, 2003). N170 is increased and delayed for inverted faces (see e.g., Bentin et al., 1996, Eimer, 2000b, Itier and Taylor, 2002, Rossion et al., 2000, Wiese et al., 2009), possibly reflecting the resulting difficulty to detect a face-like pattern. Previous studies suggested the possibility of differences in structural encoding of old versus young faces, in terms of larger N170 amplitudes for old faces (Wiese et al., 2008), with similar findings for other- versus own-race faces (see e.g., Caharel et al., 2011, Herrmann et al., 2007, Walker et al., 2008). Other studies suggest altered structural encoding in old versus young participants, in terms of delayed (Gazzaley et al., 2008, Rousselet et al., 2009, Wiese et al., 2008) and/or increased N170 responses (de Fockert et al., 2009, Gao et al., 2009, Rousselet et al., 2009, Wiese et al., 2008) in elderly relative to young adult participants. However, an interaction of face and participant age has not been observed in the N170, since both young and elderly participants demonstrated larger amplitudes for old versus young faces (Wiese et al., 2008).

Subsequent face-sensitive ERPs include an occipito-temporal P2 and a following posterior temporal N250. Of relevance, P2 has been found to be significantly more positive for own- as compared to other-race faces (Stahl et al., 2008, Stahl et al., 2010), as well as for young as compared to old faces (Wiese et al., 2008), an effect, which was more prominent in young relative to elderly participants and may be related to the perceived typicality of a face. In addition, N250 has been demonstrated to reflect face learning, with larger negativity for learned as compared to new faces (Kaufmann et al., 2009, Tanaka et al., 2006). In the two directly relevant previous studies (Wiese, 2012, Wiese et al., 2008), a larger right occipito-temporal N250 effect, with more negative amplitudes for learned as compared to new faces, was observed for young but not old faces in young participants. This finding was suggested to reflect easier access to structural facial representations for own-age faces in this group.

A prominent ERP phenomenon in episodic recognition memory experiments is the so-called old/new effect, which refers to the finding that correctly recognized learned items elicit more positive amplitudes as compared to correctly rejected new items (Rugg & Coles, 1995). Several previous studies, mostly using verbal material, reported two separate effects. While an early frontal old/new effect has been interpreted to either reflect fluency-based episodic memory (i.e., familiarity; see Curran, 2000) or conceptual priming (Paller, Voss, & Boehm, 2007), a later left-parietal effect is commonly assumed to reflect conscious recollection (Rugg & Curran, 2007), a process that involves remembering details of the study episode. At some variance with these results, for face stimuli a more widespread and frontally distributed neural correlate of recollection has been reported (MacKenzie & Donaldson, 2009). Moreover, while a number of studies did not detect amplitude differences of the later parietal old/new effect in young versus elderly participants (Friedman, 2000, Mark and Rugg, 1998, Trott et al., 1997), several other studies described this old/new effect in elderly participants to be reduced or absent (Nessler et al., 2007, Swick and Knight, 1997, Wolk et al., 2009). As discussed by Wolk et al., 2009, this discrepancy may at least partly be attributed to task effects: While those studies, which explicitly asked participants to remember details of the study episode (e.g., via source memory tasks or remember/know judgments) did detect similar parietal old/new effects in the elderly, those studies, which tested simple item recognition memory did not. Finally, several studies reported reversed old/new effects in elderly participants, with more positive amplitudes for correct rejections as compared to hits (Duarte et al., 2006, Friedman et al., 2010, Li et al., 2004, Swick et al., 2006), in conditions in which young participants demonstrated normal old/new effects (for studies reporting a reversed polarity of the old/new effect in young participants from approximately 600 ms on, see e.g., Johansson and Mecklinger, 2003, Herron, 2007). With respect to the OAB, a left centro-parietal old/new effect in young participants was more pronounced for own- as compared to other-age faces (Wiese et al., 2008), thus reflecting an ERP correlate of the observed behavioral memory bias. Additionally, the old/new effect was smaller in the elderly group (which is in line with the interpretation of Wolk et al. (2009) described above, since study phase details were not directly task-relevant), and did not differ with respect to face age.

Based on the literature discussed above, we aimed at testing the effect of daily-life contact towards young and old people on the OAB in a relatively large group of elderly participants. For that purpose, the elderly group was divided into two subgroups via a median split on a self-report measure of relative contact towards old versus young people. A group of young adult participants was additionally tested. Note that we did not divide the young participants group into contact-based subgroups, because in contrast to findings in elderly participants (i) our previous research suggested a consistent OAB in young participants (Wiese, 2012, Wiese et al., 2008), and (ii) relative contact is homogenously biased towards young people in this group. Crucially, to ensure that potentially reduced or absent age biases in elderly participants were not related to low overall performance levels, and thus to potential floor effects, memory biases towards young or old faces were additionally corrected for individual performance level.

The following predictions were derived from the different theoretical accounts described above: If lifetime perceptual expertise is crucial for the occurrence of the OAB, neither of the two old groups should show a corresponding effect (since both subgroups have been young in the course of their lifetime and should have accumulated experience with all adult age groups). By contrast, a contact-based account, emphasizing the importance of more recent daily-life interactions, would predict an OAB for those elderly participants only who report relatively more contact towards own-age as compared to other-age people. Finally, assuming that both elderly subgroups perceive young faces as belonging to a social “out-group”, socio-cognitive accounts (e.g., Hugenberg et al., 2010) would predict an OAB in both groups, independent of self-reported contact.

In ERPs, we expected to replicate earlier results, namely a smaller N170 and larger P2 for young faces in young participants and in those elderly participants who do not show a behavioral OAB (Wiese et al., 2008). In addition, since the configural/holistic processing variant of the perceptual expertise account assumes more in-depth analysis of own-age faces at the level of early perceptual processing, the investigation of N170 and P2 in different groups of elderly participants allowed an examination of this account. More precisely, if either one or both elderly groups demonstrated a memory advantage for own-age faces, and if N170 and/or P2 reflected processing stages relevant for the OAB, a reversed effect, namely a smaller N170 and larger P2 for old relative to young faces may be detected in those groups. In addition, larger effects of stimulus repetition (N250) and recollection-based episodic memory (parietal old/new effect) for own-age faces were expected in our young participants. Similarly, if one (or both) of the elderly groups demonstrated a behavioral OAB, a larger N250 repetition effect for own-age faces would be expected for this group, since this component likely reflects processes of implicit memory, commonly assumed to be relatively spared by cognitive aging (Balota, Dolan, & Duchek, 2000). By contrast, the late old/new effect presumably reflects conscious recollection, usually assumed to be reduced in the elderly (Anderson, Craik, & Tulving, 2000). Accordingly, we would not necessarily predict that a behavioral OAB in old participants is accompanied by larger late old/new effects for own-age faces.