Face Recognition Turned Upside-Down

cell-type diversity. One particularly intriguing observamore carefully). How difficult would it be to recognize tion is that LIM-HD expression is dynamic (Tsuchida et your friends? Would inverted hors d’oeuvres pose a al., 1994; Appel et al., 1995; Sharma et al., 1998), raising challenge? If you selected an upside-down cheese blintz the possibility that precise temporal regulation of LIMwithout pause but took a moment longer to recognize HD expression expands the code. Strategies to drive a coworker, you would be in good company. Psychodifferent combinations of stable LIM-HD expression in physical studies have shown that inversion of visually specific motor neuron subclasses will provide tests of presented stimuli disproportionately impairs face recogthis idea. Nevertheless, expression of known LIM-HD nition compared to recognition of other objects. This proteins cannot account for the full range of motor neufinding has been interpreted as evidence that face perron diversity. Thus, it will be important to learn how LIMception is subserved by computational processes that HD proteins interact with other factors, such as ETS differ, at some point, from the larger realm of object proteins, that further define motor neuron subtypes (Lin recognition (Valentine, 1988). In this issue, Haxby and et al., 1998). Still another exciting area of research will colleagues (1999) report the results of a functional neube investigation of LIM-HD regulatory targets to learn if roimaging experiment designed to elucidate the neurothese proteins control expression of factors important computational correlates of the inversion effect. for sensing axon guidance cues. In particular, it will be Behavioral studies of stimulus inversion dovetail nicely interesting to learn if regulatory targets of homologous with neuropsychological studies of object recognition. LIM-HD factors are conserved between flies and verteFollowing brain lesions in the posterior, ventral portion brates. Perhaps, we will find that the axon pathways of the neocortex, patients occasionally demonstrate isotracked by motor neurons expressing similar LIM-HD lated impairments in their ability to recognize certain combinations in flies and vertebrates are marked by types of objects. Well documented are cases of prososimilar signals, suggesting that the factors that guide pagnosia, in which the patient cannot recognize faces axons to their targets have been maintained through but can recognize other objects, and object agnosia, in evolution as pathfinding cassettes. which the patient can recognize faces but not general objects. These perceptual deficits seem to result from damage to “specialized” cortical areas that are necesBruce Appel sary for the perception of the given stimulus class (i.e., Department of Molecular Biology faces versus general objects) (Farah, 1990). InterestVanderbilt University ingly, prosopagnosic patients handle inverted face stimNashville, Tennessee 37232 uli with the same (or even better) facility as do normal controls (Farah et al., 1995), in contrast to an agnosic Selected Reading patient who was found to be severely impaired at the recognition and manipulation of inverted faces (MoscoAppel, B., Korzh, V., Glasgow, E., Thor, S., Edlund, T., Dawid, I.B., and Eisen, J.S. (1995). Development 121, 4117–4125. vitch et al., 1997). This pattern suggests that, when preDawid, I.B., Breen, J.J., and Toyama, R. (1998). Trends Genet. 14, sented in an upright orientation, faces are preferentially 156–162. processed by a “face-specific” cortical system. When, Landgraf, M., Bossing, T., Technau, G.M., and Bate, M. (1997). J. however, the face is inverted, the stimulus is handled Neurosci. 17, 9642–9655. by more general “object” recognition systems. Lin, J.H., Saito, T., Anderson, D.J., Lance-Jones, C., Jessell, T.M., Haxby and colleagues used functional magnetic resoand Arber, S. (1998). Cell 95, 393–407. nance imaging (fMRI) to define in each of several subPfaff, S.L., Mendelsohn, M., Stewart, C.L., Edlund, T., and Jessell, jects cortical areas that respond more to faces or to T.M. (1996). Cell 84, 309–320. objects (in this case, houses). These regions were asSharma, K., Sheng, H.Z., Lettieri, K., Li, H., Karavanov, A., Potter, sumed to correspond to the “face” and “object” cortical S., Westphal, H., and Pfaff, S.L. (1998). Cell 95, 817–828. sites that, when damaged, result in prosopagnosia and Shawlot, W., and Behringer, R.R. (1995). Nature 374, 425–430. general object agnosia. The critical question then posed Thor, S., and Thomas, J.B. (1997). Neuron 18, 397–409. was: will the presentation of inverted faces, as compared Thor, S., Andersson, S.G.E., Tomlinson, S., and Thomas, J.B. (1999). to upright faces, shift the site of active neural processing Nature, in press. from the “face” region to the “object” region? An affirmaTsuchida, T., Ensini, M., Morton, S.B., Baldassare, M., Edlund, T., tive answer would, first, confirm that inverted faces call Jessell, T.M., and Pfaff, S.L. (1994). Cell 79, 957–970.