The inferotemporal cortex (IT) is thought to be critical for visual object discrimination and recognition, because lesions of IT in the monkey induce selective impairment in learning tasks which require visual object discrimination or recognition.  The ultimate goal of the present study is to understand the neuronal mechanisms of the object discrimination and recognition in IT.  Indivisual neurons in the anterior part of IT selectively responded to visual stimuli which have a particular complex feature: a complex shape or a combination of texture or color with shape (Tanaka et al. 1991).  Simultaneous recordings from multiple neurons with a single electrode demonstrated that adjacent neurons had similar stimulus selectivity.  (Fujita et al. Soc. Neurosci. Abstr. 1990).      To assess the spatial arrangement of neurons with similar stimulus preference, we made two kinds of penetrations in IT, one directed vertically and the other directed tangentially to the cortical surface, and we examined stimulus selectivities of neurons successively recorded along the penetrations.
   Extracellular recordings were made in the anterior part of IT in anesthetized and immobilized Japanese monkeys (Macaca fuscata).  1. The critical feature for the activation was determined for a neuron located at a middle depth:  tens of 3-D objects and various paper cutouts were presented to the monkey for the initial survey to find stimuli which activated the neuron; and then the stimulus feature critical (necessary and enough) for the activation was determined by a procedure involving step by step simplification of the image of effective objects using a computer graphic system.  2. A test set of stimuli was made including optimal, suboptimal and ineffective stimuli for the neuron.  3. Neurons were sampled at 200 micrometer step along the penetration and their selectivity was examined with the same set of stimuli.
    In the vertical penetration, neurons recorded over a distance of 0.6 ? 2.1 mm responded to common but limited number of stimuli in the text set, i.e., the optimal and suboptimal stimuli for the first neuron.  In 6 out of 8 penetrations, the span of neurons with similar selectivity covered a most part of the gray matter of the cortex.  We thus suggest that neurons with similar stimulus selectivity are vertically distributed from the surface to the white matter across the cortical layers.  Frequently, the selectivity was slightly different among neurons along a penetration, as found in the previous simultaneous-recording experiments.
     In the tangential penetrations, neurons with related stimulus selectivity were found within a span of 0.2 ? 0.7 mm. Neurons outside the span did not respond to the stimuli in the text set, but they are not necessarily “unresponsive cells”. We assumed that they might respond if we presented a completely different kind of stimuli.  We thus suggest that neurons with similar stimulus selectivity are localized within a small region which is vertically elongated.
    The present study demonstrated that the neurons with similar stimulus selectivity form a column.  We suggest that the anterior IT forms a modular structure composed of hundreds of columns.  Critical feature for neuronal activation differs from module to module, but the individual critical feature was not complex enough to represent the concept of a natural object seen in the “nature”.  Activation of one module indicates the presence of a particular partial feature in the image.  Concept of an object would be described by combined activations of a set of modules.  On the other hand, within a module, stimulus selectivity of neurons were well overlapped but the details of stimulus selectivity was slightly different from cell to cell.  Precise difference of each partial feature in the image would be described by a pattern of neural activities of neurons in a module.