Face Recognition

Although the R-ATL group had a similar number of face recognition hits as controls and the L-ATL group, they did commit an increased number of false-positive recognition errors. This finding is consistent with previous reports of an increased rate of false recognition and misidentification of unfamiliar faces (Rapcsak et al., 1994, 1996) and difficulties in discriminating famous faces from foils (Ellis et al., 1991; Seidenberg et al., 2002) following right hemisphere damage. False alarms were evident particularly in the initial learning trials but were no longer observed after trial 4. Overall, the R-ATL group demonstrated a moderate impairment in face recognition characterized by a difficulty in establishing a reliable face recognition unit. However, they overcame this difficulty with multiple exposures. In contrast, two other studies have shown that right TLE participants are impaired in unfamiliar face recognition after multiple trial presentations (Beardsworth & Zaidel, 1994; Dade & Jones-Gotman, 2001).

The current study findings may reflect the beneficial contribution of the accompanying biographical vignettes on face recognition performance. Polster and Rapcsak (1996) described a prosopagnosic patient with a right hemisphere lesion (RJ) who showed different levels of unfamiliar face learning across four encoding conditions. RJ was more successful in acquiring unfamiliar faces when he rated them for personality traits or listened to an accompanying biographical vignette than when he rated them on physical features or identified the most distinctive feature. Similarly, the vignettes used in this study may serve to elicit “deep” encoding operations that help to create identity-specific semantic codes that in turn assist in the creation of structural face codes. Furthermore, RJ was unable to generalize the ability to acquire unfamiliar faces when alternate views were tested. In the current study, the same face photograph was repeatedly used over the seven learning trials and it is quite possible that the use of alternate views of the target unfamiliar faces would prove more challenging for the right ATL participants.

Acquisition of Biographical Information

Although both the R-ATL and L-ATL groups showed poorer face-biographical vignette acquisition than controls, the extent of impairment was considerably more severe for the L-ATL group. The impairment in recall of biographical information was evident for trials in which accurate face recognition occurred, and face recognition hits did not differ between the three groups (see Table 3). The L-ATL group demonstrated a global impairment relative to controls in the acquisition of all three biographical components of the presented vignettes—occupation, city, and person name—despite normal levels of face recognition. In addition, the L-ATL group recalled significantly less occupation and person name information than the R-ATL group. In contrast, the R-ATL group performed similar to controls (across trials) in the recall of occupation information, but was impaired for the recall of both city and person name information.

The global impairment in the recall of biographical information observed for the L-ATL group differs from previous findings regarding recognition and identification of famous faces. Several investigators have shown that left hemisphere lesions result in a specific impairment in the retrieval of proper names, with normal levels of retrieval of other semantic identification information, such as occupation or reason for fame (Flude et al., 1989; Fukatsu et al., 1999; Hanley, 1995; Harris & Kay, 1995; Hittmair-Delazer et al., 1994; Reinkemeier et al., 1997; Seidenberg et al., 2002; Semenza & Zettin, 1989; Shallice & Kartsounis, 1993).

A possible explanation for the contrasting findings observed here is that the current study employed a new learning (episodic memory) paradigm, rather than examination of retrieval from a previously acquired knowledge base (semantic memory). Consistent with this hypothesis, a subset of L-ATL subjects (n = 7) who had previously served as participants in a protocol examining famous face recognition and identification demonstrated accurate knowledge of biographical information (i.e., occupation or reason for fame), but not the person's name for famous faces (Seidenberg et al., 2002). Thus, retrieval of occupation information for already known familiar faces was relatively preserved in the L-ATL group compared to the ability to form new associative face-occupation relationships.

In the context of a global impairment in the acquisition of biographical information, the L-ATL group demonstrated a very marked difficulty in associating a person's name to a new face. Indeed, they exhibited an essentially flat learning curve for person name recall over the seven trials (see Figure 2c). Their performance also remained significantly below both the controls and the R-ATL group during the recognition memory probe for person's name. Thus, the left temporal lobe appears to play a critical role both in the acquisition and association of a person's name to a face, as well as in the retrieval of previously acquired names of familiar people (Damasio et al., 1996).

The R-ATL group was also impaired relative to controls in learning names and city information associated with unfamiliar faces. A deficit in person identity knowledge has been previously reported for famous faces in patients with right temporal lobe damage (Ellis et al., 1989; Evans et al., 1995; Hanley et al., 1990; Seidenberg et al., 2002). A role for the right temporal lobe in person identity has also been suggested in recent functional neuroimaging studies for recognition of famous faces (Leveroni et al., 2000), and for the acquisition of biographical information (names and occupation) to unfamiliar faces (Tsukiura et al., 2002). Although the R-ATL group was impaired in the acquisition of biographical information compared to the controls, their impairment was not as severe as the L-ATL group (see Table 3). The L-ATL group was impaired relative to both controls and the R-ATL group on the recognition memory probe for both city and person name information. In contrast, the R-ATL group performed similar to controls on the recognition memory probe for biographical information. However, recognition performance for the control group was at ceiling, which may have obscured possible group differences. Nevertheless, it is evident that the difficulty in the acquisition of person identity information is more pronounced following L-ATL than R-ATL.

The TLE participants examined in the current study underwent a tailored resection based on findings from intracranial electrocorticography. Thus, the resection estimates provided in the Methods section provide a general reference point for the typical surgery procedure conducted. More detailed information concerning the extent of resection in the lateral and mesial regions is not available. Nevertheless, it should be noted that that the typical resection was larger for participants undergoing R-ATL as compared to L-ATL, and the current study findings implicate a greater impairment in acquiring biographical information for the L-ATL group. In addition, the findings point to a dissociation in the nature of impairment evident between L-ATL and R-ATL groups for unfamiliar face acquisition and acquisition of biographical information, and a dissociation in the extent of impairment in learning different types of biographical information (names, occupation, city). Thus, the pattern of results is difficult to reconcile with a simple extent of resection explanation.

Recall of Biographical Information

The present study points to the considerable difficulty in learning to associate names to unfamiliar faces, and is consistent with previous reports in neurologically normal (Cohen, 1990; Craigie & Hanley, 1997; McCluney & Krauter, 1997; McWeeny et al., 1987) and brain-damaged participants (Fukatsu et al., 1999; Lucchelli & De Renzi, 1992; Milders, 1998). For all three groups, recall of name information was significantly lower than recall of city information, which was in turn poorer than recall of occupation information. This was also evident even when credit was given for correct recall of just the first or last name rather than the full name. A serial order learning effect (i.e., primacy or recency) cannot account for the pattern of observed recall findings as each vignette used a standard presentation order for the biographical information in which person name was presented first, followed by occupation and city name (see Appendix). Other factors which could affect recall levels for the different types of biographical information such as category set size and frequency of occurrence were not controlled in this study, and warrant investigation in subsequent studies.

The ubiquitous nature of name learning difficulty has been the subject of considerable interest within the context of current models of face processing ability (Bruce et al., 1994; Burton & Bruce, 1992; Cohen, 1990; Stanhope & Cohen, 1993). In the present study, recall of city names was intermediate between occupation and name recall for all three groups. City names are also proper names and the question arises as to why city names are more easily acquired than person names. City names do have characteristics that are similar to person names. They are both arbitrary verbal labels that uniquely describe a place/person. However, city names may not truly be meaningless in the sense that they may have “borrowed semanticity” (Cohen & Faulkner, 1986) that links them into the semantic network. For example, the city Seattle may produce conceptual associations (e.g., rain, coffee, Space Needle) that allow for integration into a semantic network. Additionally, city names are not truly unique in that they are also a property that describes many other individuals in the person identity system, (i.e., several person identity nodes may share them). That is, while there is only one geographical area named Seattle, there are many individuals who share Seattle as a biographical property. It is also possible to adjectivize a city in reference to a set of characteristics. In this sense, a city name can serve as an associative cue to a more general set of semantic features.

In addition, name recall (person or city) rarely occurred without the accompanying recall of occupation information. Indeed, person name recall occurred most frequently (for all three groups) when both occupation and city name were also recalled. These findings are consistent with serial/hierarchical proposals of person identification processes described for retrieval of familiar faces (Bruce & Young, 1986; Hay et al., 1991). The current study results extend this perspective to the process of acquiring new face person identity representations.

Face-Semantic Network Acquisition and Cognitive Functioning

The performance of the ATL groups on the standard clinical neuropsychological measures showed a pattern of material-specific effects following unilateral temporal lobectomy (Bell & Davies, 1998; Hermann et al., 1995, 1999; Milner, 1968; Seidenberg et al., 1998). The R-ATL group was impaired relative to controls on face discrimination, recognition memory for faces, and visual memory. Conversely, the L-ATL group was impaired on measures of confrontation naming, recognition memory for words, and verbal episodic memory. The L-ATL group was also impaired on visual memory, a finding frequently observed and hypothesized to be related to the verbalizability of the stimuli (Bell & Davies, 1998; Hermann et al., 1992). The performance of the ATL groups on the unfamiliar face/biographical vignette-learning task was also generally consistent with material specific effects. The R-ATL group was impaired on face learning and discrimination ability and the L-ATL group was more impaired on learning the biographical verbal information. Furthermore, unfamiliar face recognition performance was significantly correlated with indices of face discrimination ability and face recognition memory, suggesting that both perceptual and mnemonic abilities are related to the formation of a face recognition unit. In contrast, learning of biographical information was strongly related to indices of associative verbal and nonverbal memory. Thus, it appears that general associative learning ability is critically linked to the acquisition of new face-semantic identity representations.

Anterograde associative learning ability is strongly linked to the functional integrity of the mesial temporal lobe region (Eichenbaum et al., 1994; Squire, 1992), and previous findings of pre-post ATL have confirmed that associative learning ability is more negatively affected following left-sided resections compared to right-sided resections (Seidenberg et al., 1998). However, one cannot rule out the potential contribution of regions adjacent to the medial temporal lobe as the typical resection also removes a portion of the anterior temporal lobe as well. Recent findings from both human lesion studies and functional neuroimaging investigations have implicated the anterior temporal lobe in familiar face identification processes including naming famous faces (Gorno-Tempini et al., 1998; Leveroni et al., 2000; Tranel et al., 1997). Particular attention has focused on the possible role of the left temporal pole and left middle temporal gyrus in the recognition and naming of familiar faces. Furthermore, the left temporal polar region has been linked to retrieval of names for both unique familiar persons and unique familiar landmarks (Grabowski et al., 2001). These studies address the retrieval of already known concepts (faces and places). Of interest, a recent fMRI study examining activation patterns for newly learned faces and biographical information implicated the involvement of the left temporal lobe in the retrieval of people's names irrespective of whether they were famous or newly learned. In addition, the right anterior temporal lobe showed increased activity during retrieval of people's names from newly learned faces compared to well known famous faces (Tsukiura et al., 2002). These findings correspond with the current study findings showing impairment of both the L-ATL and R-ATL groups (compared to controls) in the acquisition of names for newly acquired faces.

Taken together, the current findings are consistent with the viewpoint that the acquisition of new face semantic networks is related to two distinct cognitive operations—face processing and associative learning. Laterality of temporal lobe lesion (as reflected by ATL) appears to produce a dissociation, which parallels this distinction. The R-ATL group was more impaired for unfamiliar face learning while the L-ATL group was more impaired for the acquisition of biographical information. Furthermore, the acquisition of person-identity information appears to be strongly linked to both verbal and nonverbal associative learning ability (Hittmair-Delazer et al., 1994; Shallice & Kartsounis, 1993). Accordingly, the pronounced impairment evident for the L-ATL group in acquiring new face-related biographical representations may be viewed as a consequence of their substantial impairment in both verbal and nonverbal associative learning capabilities. These findings also serve to highlight important similarities and differences between the effect of lateralized temporal lobe lesions on retrieval of (already known) information concerning famous faces and the acquisition of new face biographical associations.