NEUROPSYCHOLOGICAL FINDINGS IN THE ANTERIOR TEMPORAL LOBES

Proponents of the hub account have almost universally asserted that the semantic hub is located in the ATLs, primarily on the bases of neuropsychological data from semantic dementia patients.Footnote ¹ Semantic dementia, a subtype of frontotemporal dementia, is a degenerative disease associated with atrophic temporal lobes and broad semantic memory impairments in the presence of relatively intact functioning in other cognitive domains (Patterson et al., Reference Patterson, Nestor and Rogers2007). Semantic dementia patients’ conceptual defects are more highly correlated with pathology in the lateral surface of the ATLs than medial temporal cortex (Levy, Bayley, & Squire, Reference Levy, Bayley and Squire2004). This finding is consistent with evidence that repetitive Transcranial Magnetic Stimulation (rTMS) over the lateral ATLs in healthy subjects temporarily slows object naming response speed (Pobric, Jefferies, & Lambon-Ralph, Reference Pobric, Jefferies and Lambon-Ralph2007).

There are reasons to be cautious, however, when interpreting both the rTMS and semantic dementia data. First, we know of no evidence that bombarding the anterior temporal region with 10 min of continuous stimulation before experimental tasks, as was done by Pobric and colleagues, produces a “virtual lesion” that is specific to the ATLs. To the contrary, this procedure alters neural activity in regions remote from the site of stimulation (e.g., Rounis et al., Reference Rounis, Stephan, Lee, Siebner, Pesenti and Friston2006). And with respect to semantic dementia patients, pathology is clearly not restricted to the anterior temporal lobes. Pathology often extends into the amygdala (Noppeney et al., Reference Noppeney, Patterson, Tyler, Moss, Stamatakis and Bright2007), and the frontal lobes (Brambati et al., Reference Brambati, Rankin, Narvid, Seeley, Dean and Rosen2009), and semantic memory impairments are equally correlated with pathology in the ATLs and posterior regions of the fusiform gyrus (Williams, Nestor, & Hodges, Reference Williams, Nestor and Hodges2005). Furthermore, anterior temporal resection for intractable epilepsy does not invariably lead, and in fact rarely leads, to significant domain-general semantic memory impairments (Drane et al., Reference Drane, Ojemann, Aylward, Ojemann, Johnson and Silbergeld2008), although it is almost invariably associated with episodic memory deficits.Footnote ²

In contrast to the global semantic impairments of semantic dementia patients, focal ATL damage has been associated with more selective semantic memory deficits. For example, anterior temporal resection can result in circumscribed impairments in recognizing and naming familiar/famous people, implicating the ATLs in the storage of person-specific information (Fukatsu, Fujii, Tsukiura, Yamadori, & Otsuki, Reference Fukatsu, Fujii, Tsukiura, Yamadori and Otsuki1999; Glosser, Salvucci, & Chiaravalloti, Reference Glosser, Salvucci and Chiaravalloti2003; Tsukiura et al., Reference Tsukiura, Namiki, Fujii and Iijima2003). Moreover, these patients also occasionally exhibit impaired recognition and naming of familiar/famous landmarks (Drane et al., Reference Drane, Ojemann, Aylward, Ojemann, Johnson and Silbergeld2008; Tranel, Reference Tranel2006), suggesting that the ATLs may store semantic information about unique entities specifically, rather than semantic memory generally. Finally, the left ATL may be important for the retrieval of proper names for people and landmarks (Tranel, Reference Tranel2006), and so serve as a convergence region linking conceptual knowledge about unique entities and entity-associated word (e.g., names), both of which are stored outside the ATLs (Damasio et al., Reference Damasio, Tranel, Grabowski, Adolphs and Damasio2004).

Proponents of ATL hub accounts generally offer two responses to the absence of gross semantic memory deficits following ATL resection and the relative specificity of unique entity deficits. First, they note that resected tissue is by definition unhealthy and that functional reorganization both pre- and postsurgery may mean the brains of long-term epilepsy patients are less than ideal models for normal brain–behavior relationships (Powell et al., Reference Powell, Parker, Alexander, Symms, Boulby and Wheeler-Kingshott2007). Second, unlike semantic dementia, resections are unilateral, thus leaving a healthy, functioning hemisphere. This argument finds support in the observation that global semantic memory deficits are more pronounced when pathology is bilateral (Noppeney et al., Reference Noppeney, Patterson, Tyler, Moss, Stamatakis and Bright2007). Given these criticisms, functional neuroimaging data from healthy individuals engaged in conceptual tasks may be particularly useful.

FUNCTIONAL NEUROIMAGING FINDINGS IN THE ANTERIOR TEMPORAL LOBES

The vast majority of positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) studies of conceptual processing do not report activation in the ATLs. On the contrary, these studies largely implicate posterior ventral and lateral temporal cortices, as well as left inferior frontal cortex (Thompson-Schill, Reference Thompson-Schill2003). When ATL activations are reported in functional neuroimaging studies, it is usually when subjects are engaged in processing unique entities or performing social conceptual processing tasks.

Consistent with the unique entity account of ATL function, ATL activity has often been associated with viewing famous and familiar faces (Damasio et al., Reference Damasio, Tranel, Grabowski, Adolphs and Damasio2004; Grabowski et al., Reference Grabowski, Damasio, Tranel, Ponto, Hichwa and Damasio2001; Gorno-Tempini et al., Reference Gorno-Tempini, Price, Josephs, Vandenberghe, Cappa and Kapur1998; Leveroni et al., Reference Leveroni, Seidenberg, Mayer, Mead, Binder and Rao2000; Nakamura et al., Reference Nakamura, Kawashima, Sato, Nakamura, Sugiura and Kato2000; Pourtois, Schwartz, Segheir, Lazeyras, & Vuilluemier, Reference Pourtois, Schwartz, Segheir, Lazeyras and Vuilluemier2005; Rotshtein, Henson, Treves, Driver, & Dolan, Reference Rotshtein, Henson, Treves, Driver and Dolan2005; Sergent & Signoret, Reference Sergent and Signoret1992; Sugiura et al., Reference Sugiura, Kawashima, Nakamura, Sato, Nakamura and Kato2001; Tsukiura, Namiki, Fujii, & Iijima, Reference Tsukiura, Namiki, Fujii and Iijima2003). The unique entity account gains yet more support from two PET studies reporting ATL activation when subjects viewed unique landmarks relative to unfamiliar locations (Grabowski et al., Reference Grabowski, Damasio, Tranel, Ponto, Hichwa and Damasio2001; Nakamura et al., Reference Nakamura, Kawashima, Sato, Nakamura, Sugiura and Kato2000). There are at least two reasons to be cautious, however, about extending the famous/familiar face findings to the representation of unique entities generally, rather than social knowledge specifically. First, Kriegeskorte, Formisano, Sorger, and Goebel (Reference Kriegeskorte, Formisano, Sorger and Goebel2007) using high-resolution fMRI and multivariate information-mapping analyses identified a region in the right inferior ATL carrying information that reliably discriminated among individual faces, but failed to observe ATL regions that discriminated among individual houses. Second, ATL activation does not require referencing information about specific, unique individuals. For example, ATL activation has often been reported when subjects infer the mental states and emotions of generic others in theory of mind tasks (for review, see Olson, Plotzker, & Ezzyat, Reference Olson, Plotzker and Ezzyat2007). Similarly, activity in the left ATL has been reported to correlate with deceit detection (Grezes, Frith, & Passingham, Reference Grezes, Frith and Passingham2004), and the right ATL has been found to activate when subjects view photographs depicting scenes with moral connotations (e.g., physical assaults) versus similar nonmoral control scenes (Moll et al., Reference Moll, de Oliveira-Souza, Eslinger, Bramati, Mourao-Miranda and Andreiuolo2002). More recently, Zahn and colleagues used both fMRI and PET imaging to show that activation of the right superior ATL was related to the representation of social abstract concepts such as courage or generosity (Zahn et al., Reference Zahn, Moll, Krueger, Huey, Garrido and Grafman2007, Reference Zahn, Moll, Iyengar, Huey, Tierney and Krueger2009). Indeed, stimuli need not even be explicitly social to engage the ATLs. For example, the ATLs activate when subjects infer animacy and intentionality in the movements of abstract geometric shapes (Castelli, Happe, Frith, & Frith, Reference Castelli, Happe, Frith and Frith2000; Schultz et al., Reference Schultz, Grelotti, Klin, Kleinman, Van der Gaag and Marois2003). Taken together, these findings suggest that the ATLs might be engaged in social conceptual processing generally, rather than the representation of unique entities per se.

The inconsistency of ATL activations outside of social conceptual processing tasks is also curious in light of the semantic hub account. If the ATL is the hub of the human conceptual system, one might expect it to be consistently activated by conceptual processing tasks. Semantic hub proponents offer at least two explanations for this inconsistency. First, they point out that fMRI is currently the predominant imaging modality, and its image signal quality over the ATLs is relatively poor due to magnetic susceptibility artifacts (Devlin et al., Reference Devlin, Russell, Davis, Price, Wilson and Moss2000). To address this concern, hub proponents point to PET studies demonstrating ATL activations during a variety of conceptual processing tasks (Crinion, Lamon Ralph, Warburton, Howard, & Wise, Reference Crinion, Lambon Ralph, Warburton, Howard and Wise2003; Davis & Johnsrude, Reference Davis and Johnsrude2003; Devlin et al., Reference Devlin, Russell, Davis, Price, Wilson and Moss2000; Rogers et al., Reference Rogers, Hocking, Noppeney, Mechelli, Gorno-Tempini and Patterson2006; Vandenberghe, Nobre, & Price, Reference Vandenberghe, Nobre and Price2002; Vandenberghe, Price, Wise, Josephs, & Frackowiak, Reference Vandenberghe, Price, Wise, Josephs and Frackowiak1996). Second, hub proponents have asserted that many imaging studies of conceptual processing fail to observe ATL activation because they require subjects to process stimuli at a level of specificity that is too general (Rogers et al., Reference Rogers, Hocking, Noppeney, Mechelli, Gorno-Tempini and Patterson2006; Tyler et al., Reference Tyler, Stamatakis, Bright, Acres, Abdallah and Rodd2004). As described earlier, semantic hub accounts claim that the ATL stores information about similarity relations among concepts. Rogers and colleagues (Reference Rogers, Hocking, Noppeney, Mechelli, Gorno-Tempini and Patterson2006) hypothesized that the ATL should thus be more strongly activated by subordinate-level classification judgments of highly related concepts (e.g., differentiating among two different kinds of birds) than more general classifications (e.g., differentiating among animals and artifacts generally). In support of this hypothesis, they observed greater bilateral ATL activation when subjects made specific/subordinate-level classification judgments (blue jay vs. robin) than when they made basic-level (bird vs. dog) and superodinate (animal vs. vehicle) classification judgments. Even so, some studies do find ATL activation with basic-level concepts (e.g., Crinion et al. Reference Crinion, Lambon Ralph, Warburton, Howard and Wise2003; Devlin et al. Reference Devlin, Russell, Davis, Price, Wilson and Moss2000; Vandenberghe et al. Reference Vandenberghe, Price, Wise, Josephs and Frackowiak1996).

Rogers et al. (Reference Rogers, Hocking, Noppeney, Mechelli, Gorno-Tempini and Patterson2006) unfortunately did not directly compare ATL activation for social and nonsocial stimuli, so it is difficult to know how their findings relate to the other theories of ATL functioning. In fact, none of the imaging studies that support a general semantic hub account of the ATL directly compare social and nonsocial stimuli.

CONCLUSIONS AND RECOMMENDATIONS

Although the domain-general semantic hub account of ATL functioning has gained much attention recently, the neuropsychological and imaging data supporting it remain sparse and not without complications. In contrast, social conceptual processing, and particularly knowledge of famous and familiar individuals, has been consistently implicated in the deficits of ATL lesion patients and activation patterns of healthy adults undergoing functional neuroimaging. More work remains, however, if we are to more precisely understand the ATL’s information content. We offer the following recommendations as a guide to future studies of the ATL’s contribution to conceptual knowledge representation.

First, much could be learned from studies directly comparing the domain-general semantic hub account with the unique entity and social conceptual processing accounts. In general, evidence for these three accounts has evolved in parallel literatures, with little effort to directly test these accounts against each other. In fact, at present only one study has directly compared the domain-general semantic hub and social conceptual processing accounts, with the results strongly supporting the latter (Simmons, Reddish, Bellgowan, & Martin, in press).

Second, although there is much to implicate the ATLs in the processing of social conceptual knowledge generally, there also remain good reasons to think that the region is dedicated to the storage of information about unique entities. In light of this, it will be important for future research to directly test these two accounts. It is also worth noting, however, that the accounts are not necessarily mutually exclusive. A potentially integrative account could be that the ATLs store information about unique entities, with a relative bias toward social unique entities. This makes sense in light of the fact that only two of the three possible patterns of deficits have been reported among patients with famous/familiar person- and landmark-knowledge deficits. As described earlier, patients most commonly present with intact landmark-knowledge, but deficient person-knowledge. Less frequently they present with deficits for both categories. There have been no reports, however, of intact person-knowledge in the presence of landmark deficits.

A more speculative possibility is that distinct subregions exist in the anterior temporal lobes for person- and place-unique entities, thus recapitulating the topographic distinction observed in posterior ventral temporal cortex, with place information represented medially in the parahippocampal cortex and face information represented laterally in the fusiform gyrus (see Grill-Spector, Reference Grill-Spector2003, for review). Importantly, such micro-organization in ATL functional anatomy would be almost impossible to isolate in lesion patients, smoothed away in lower-resolution PET data, and obscured in fMRI by signal artifacts that are worse over anteromedial regions. In light of these limitations, perhaps the only means currently available to identify functional micro-organization in the ATLs is direct cortical stimulation and/or recordings.

Finally, although more neuropsychological patient data are always welcome, theories of ATL function may benefit from research using new functional imaging methodologies to map the ATL’s information-selectivity and functional connectivity. For example, multi-voxel pattern analyses are somewhat less susceptible to signal degradation in the ATL than traditional univariate fMRI analyses (Mur, Bandettini, & Kriegeskorte, Reference Mur, Bandettini and Kriegeskorte2009) and may be a powerful tool for determining whether population codes across the region carry information discriminating among social- and nonsocial stimuli. In addition, knowing the regions to which the ATL structures are most strongly connected should help clarify its role within the wider conceptual system. This calls for better characterizations of the ATL’s functional connectivity profile, and how its connectivity is altered during various task contexts.