The Journal of Neuroscience, March 6, 2013 • 33(10):4213–4215 • 4213

Journal Club

Editor’s Note: These short, critical reviews of recent papers in the Journal, written exclusively by graduate students or postdoctoral fellows, are intended to summarize the important findings of the paper and provide additional insight and commentary. For more information on the format and purpose of the Journal Club, please see http://www.jneurosci.org/site/misc/ifa_features.xhtml.

Where Is the Anterior Temporal and What Does It Do?

Michael F. Bonner1 and Amy R. Price1,2 1Department of Neurology and 2Neuroscience Graduate Group, University of Pennsylvania, Philadelphia, Pennsylvania 19104 Review of Peelen and Caramazza

The anterior (ATL) is these findings have implicated a large re- represented (Simmons and Barsalou, thought to be critical for semantic – gion of the ATL in . 2003). This work has implications for our our knowledge of objects, people, words, One might expect that complementary understanding of how ATL structures dif- and facts. However, there is substantial functional neuroimaging studies would ferentially contribute to semantic memory, disagreement over the precise role of the provide a more fine-grained picture of and how the semantic system is shaped ATL in semantic memory, and there is ATL function. Unfortunately, the evi- by the modalities of the information it considerable variability in the anatomic dence from functional neuroimaging has processes. findings that link the ATL with semantic not clearly pointed to the ATL as a critical In their study, subjects viewed images processing. The inconsistent findings across region for conceptual knowledge. Indeed, of objects. Half were objects that are typi- studies may be related to the diverse ana- fMRI examinations of semantic memory cally found in a kitchen (e.g., pepper mill), tomic structures within the ATL and their have implicated a broad cortical network and half in a garage (e.g., staple gun). Half differential contribution to distinct modali- in conceptual processing, with sensory were objects that are typically acted upon ties of semantic information (e.g., visual, au- and motor regions contributing to repre- by rotating (e.g., pepper mill), and half ditory, affective). sentations of the perceptual and action by squeezing (e.g., staple gun). Subjects Much of the evidence implicating the features of concepts (Martin, 2007), and performed a one-back task, indicating ATL in semantic memory has come from inferior parietal and posterolateral tem- whether they saw two objects in a row that neuropsychology. In particular, patients poral regions supporting multimodal shared a particular conceptual feature with semantic , a neurodegen- conceptual information (Binder and De- (shared location for half of the trials, erative disease affecting the ATL, exhibit a sai, 2011). Inconsistent ATL activation shared action for the other half). The au- profound deficit in semantic knowledge can partly be explained by imaging arti- thors reasoned that if a cortical region with a relative sparing of most other cog- facts near the sinuses in fMRI, which may encodes abstract conceptual features, it nitive domains (Warrington, 1975; Pat- degrade signal detection in the ATL. should have a similar neural response to terson et al., 2007). These patients have Nonetheless, the variability of the evi- objects with a shared conceptual feature, little trouble performing episodic recall dence to date has left many unanswered even if those objects are different in all tasks, visual perceptual tasks, or numeri- questions about the functions of ATL other respects. They tested this hypothesis cal tasks, and can even retain complex be- structures. using multivoxel pattern analysis to look haviors like performing novel musical pieces—yet they have a striking impair- In a recent report in the Journal of Neu- for regions where objects with a shared ment on nearly all assessments of seman- roscience, Peelen and Caramazza (2012) conceptual feature had highly similar pat- tic knowledge. Given the broad anatomic provide fMRI evidence that a region of terns of fMRI responses. This was per- extent of disease in , the ventral temporal pole encodes infor- formed in regions of interest along the mation about the abstract conceptual ventral temporal lobe, and in whole- properties of objects. Their findings searchlight analyses. complement neuropsychological studies These analyses revealed a graded effect Received Jan. 4, 2013; revised Jan. 16, 2013; accepted Jan. 21, 2013. of the ATL and fit within a broad theoret- along the posterior–anterior axis of the This work was supported by National Institutes of Health Grant AG017586. We thank Murray Grossman and Jonathan Peelle for helpful ical framework of and mem- ventral temporal lobe, with the highest comments on this manuscript. ory, in which the ventral visual stream degree of similarity in bilateral ventral CorrespondenceshouldbeaddressedtoMichaelF.Bonner,Department processes a hierarchy of increasingly com- temporal pole (this anatomic region is of Neurology—2 Gibson, University of Pennsylvania, 3400 Spruce Street, plex information, culminating in anterior illustrated in Fig. 1). A similar effect was Philadelphia, PA 19104. E-mail: [email protected]. DOI:10.1523/JNEUROSCI.0041-13.2013 regions of the temporal lobe where ab- observed for both the location and action Copyright©2013theauthors 0270-6474/13/304213-03$15.00/0 stract conceptual properties of objects are features of concepts. These effects were 4214 • J. Neurosci., March 6, 2013 • 33(10):4213–4215 Bonner and Price • Journal Club

concepts, sound concepts) through all their experiment, their results may nonethe- modes of reception and expression (e.g., less be consistent with this account. Their language, photos, sounds). However, the task assessed knowledge of how individual ATL is a large, heterogeneous region con- objects are used and where they are located. taining numerous cytoarchitectonically This task may have relied on highly specific distinct cortical areas (Brodmann, 1909; semantic information in the most anterior superior temporal Blaizot et al., 2010) with a structured pat- regions of the temporal lobe, whereas a less tern of differential anatomic connectivity specific semantic task, like lexical decision, temporal pole (Kondo et al., 2003). It is unlikely that might rely more on the fusiform or parahip- these anatomically distinct regions are pocampal gyri. functionally equivalent. Furthermore, the Finally, it is important to consider the visualcomplex non-visual basic spatial resolution of functional neuroimag- modality of the semantic content that was Figure 1. Structures of the anterior temporal lobe. The ing techniques is more than adequate for probed in Peelen and Caramazza’s (2012) axes indicate theoretical gradients of differential semantic identifying the differential contribution of study. Their experiment examined the processes. Along one axis, ventral structures contribute more these regions to semantic processing. conceptual processing of objects, a cate- totheprocessingofvisualinformation(e.g.,objectconcepts), One plausible account for the ana- gory of concepts with strongly associated and dorsolateral structures contribute more to the processing tomic variability in previous studies is visual features. They reasoned that their of non-visual information (e.g., abstract concepts, auditory that there is some degree of modality- ATL findings reflect an extension of the concepts).Alongthesecondaxis,posteriorstructurescontrib- ute more to basic object representations, and anterior struc- specificity in the functions of the ATL (Yi processing hierarchy in the ventral visual tures contribute more to the representation of complex et al., 2007), and more broadly through- stream. However, this interpretation dif- conceptual associations. out the temporal lobe (Bonner and Gross- fers from the hypothesis that the ATL is an man, 2012). In other words, semantic amodal semantic hub (Patterson et al., processing in the ATL may not be fully 2007). From the amodal perspective, it is present regardless of whether the task amodal. Indeed, the results of a recent proposed that the ATL represents all cat- probed location or action knowledge, meta-analysis are consistent with this per- egories of concepts, including those with suggesting that the responses were not spective, demonstrating that visual object weak visual feature associations, such as purely task-driven. By comparison, often recruits ventral ATL abstract concepts like “truth” and audi- similarity across the stimuli was encoded structures, while linguistic and auditory tory concepts like “thunder.” Yet, criti- in early visual processing regions of processing tend to recruit lateral ATL cally, most studies of ATL function have temporal-occipital cortex. Altogether, structures (Visser et al., 2010). This meta- focused on the representation of object these findings illustrate a model of hierar- analysis focused on the modality of the concepts, which depend heavily on visual chical processing along the ventral visual materials (e.g., pictures vs feature knowledge. As with Peelen and stream, with lower-level visual-perceptual words), and did not directly address ques- Caramazza’s (2012) study, these investi- features processed in temporal-occipital tions about the modality of the semantic gations do not address the question of cortex, and higher-level conceptual asso- information being processed (e.g., visual whether subregions of the ATL differen- ciations processed in more anterior ven- concepts, action concepts, sounds con- tially contribute to representations of vi- tral temporal regions. cepts). Nonetheless, the findings indicate sual and non-visual concepts. Indeed, the Given the anatomic variability of pre- that there is some degree of modality- strong focus on object concepts in the lit- vious ATL findings, it is worthwhile con- specificity in the semantic functions of the erature may have partly biased findings sidering the precise anatomy of Peelen ATL. In this , it is relevant that the to more ventral temporal regions. It will and Caramazza’s (2012) results in light of stimuli used by Peelen and Caramazza be important to know in future studies specific hypotheses about ATL structures. (2012) depended on . whether there is some degree of modality- Until recently, the most prominent ac- This may explain in part why their effects specificity in the semantic processes of the counts of ATL function focused largely on were located in the ventral rather than lat- ATL, with ventral temporal regions con- the temporal pole, the most anterior por- eral ATL, as the ventral ATL appears to tributing disproportionately to visual tion of the ATL (Patterson et al., 2007). process predominantly visual informa- concepts. But more recent work has placed an in- tion (Visser et al., 2010). A careful consideration of the ana- creasing emphasis on ventral and lateral We can also consider Peelen and Cara- tomic subregions of the ATL can inform temporal regions that are posterior to the mazza’s (2012) findings in relation to the our understanding of the neural basis of temporal pole, including the fusiform and posterior–anterior axis of the ATL. It is semantic memory. We suggest that Peelen parahippocampal gyri (Mion et al., 2010) worth noting that their strongest effects and Caramazza’s (2012) findings on ob- and the middle and superior temporal were in the most anterior region of the ject concepts in vision fit within a more gyri (Visser et al., 2010). All of these struc- ATL—the temporal pole—and not in re- fine-grained anatomic hypothesis that pro- tures of the ATL are illustrated in Figure 1. gions more posterior to this, like the fusi- poses modality-specific functions within the Across many studies, these various ana- form or parahippocampal gyri (Mion et substructures of the ATL. In particular, this tomic structures have been referred to as al., 2010). One relevant hypothesis pro- hypothesis relates visual semantic represen- the “anterior temporal lobe” and ascribed poses a posterior-to-anterior gradient in tations to ventral regions of the temporal a similar function: serving as an amodal the specificity of semantic processing in pole and the fusiform and parahippocampal hub that integrates the information associ- the ATL, with more fine-grained semantic gyri. This interpretation contrasts with a ated with a concept (Patterson et al., 2007). information relying more on anterior re- purely amodal account of ATL function, The term amodal suggests that the ATL rep- gions (Tyler et al., 2004). Though Peelen and it raises important questions about the resents all modalities of conceptual infor- and Caramazza (2012) did not examine functional specificity of ATL substructures. mation (e.g., object concepts, action multiple levels of semantic specificity in Future studies can address these issues by Bonner and Price • Journal Club J. Neurosci., March 6, 2013 • 33(10):4213–4215 • 4215 considering where neuroimaging findings tionslehre der Großhirnrinde in ihren Prin- temporal cortex. J Neurosci 32:15728–15736. fall within the ventral–lateral and posterior– zipien dargestellt auf Grund des Zellenbaues. CrossRef Medline anterior axes of the ATL, and how the anat- Leipzig: Barth. Simmons WK, Barsalou LW (2003) The Kondo H, Saleem KS, Price JL (2003) Differen- omy of these findings relates to the similarity-in-topography principle: recon- tial connections of the temporal pole with the ciling theories of conceptual deficits. Cogn modality-specific attributes of experimental orbital and medial prefrontal networks in ma- Neuropsychol 20:451–486. CrossRef Medline stimuli. caque monkeys. J Comp Neurol 465:499–523. Tyler LK, Stamatakis EA, Bright P, Acres K, Ab- CrossRef Medline dallah S, Rodd JM, Moss HE (2004) Process- References Martin A (2007) The representation of object ing objects at different levels of specificity. Binder JR, Desai RH (2011) The neurobiology concepts in the brain. Annu Rev Psychol 58: J Cogn Neurosci 16:351–362. CrossRef of semantic memory. Trends in Cognitive Sci- 25–45. CrossRef Medline Medline ences 15:527–536. CrossRef Medline Mion M, Patterson K, Acosta-Cabronero J, Pen- Visser M, Jefferies E, Lambon Ralph MA (2010) Blaizot X, Mansilla F, Insausti AM, Constans JM, gas G, Izquierdo-Garcia D, Hong YT, Fryer Semantic processing in the anterior temporal Salinas-Alama´n A, Pro´-Sistiaga P, Mohedano- TD, Williams GB, Hodges JR, Nestor PJ Moriano A, Insausti R (2010) The human (2010) What the left and right anterior fusi- lobes: a meta-analysis of the functional neuro- parahippocampal region: I. Temporal pole cy- form gyri tell us about semantic memory. imaging literature. J Cogn Neurosci 22: toarchitectonic and MRI correlation. Cereb Brain 133:3256–3268. CrossRef Medline 1083–1094. Medline Cortex 20:2198–2212. CrossRef Medline Patterson K, Nestor PJ, Rogers TT (2007) Where Warrington EK (1975) The selective impair- Bonner MF, Grossman M (2012) Gray matter do you know what you know? The represen- ment of semantic memory. Q J Exp Psychol density of auditory association cortex relates to tation of semantic knowledge in the human 27:635–657. CrossRef Medline knowledge of sound concepts in primary pro- brain. Nat Rev Neurosci 8:976–987. CrossRef Yi HA, Moore P, Grossman M (2007) Reversal gressive . J Neurosci 32:7986–7991. Medline of the concreteness effect for verbs in semantic CrossRef Medline Peelen MV, Caramazza A (2012) Conceptual dementia. Neuropsychology 21:9–19. Brodmann K (1909) Vergleichende Lokalisa- object representations in human anterior CrossRef Medline