Introduction

The anterior part of the inferotemporal cortex, which corresponds to the cytoarchitectural area TE, is the final purely visual stage along the occipitotemporal or ventral visual pathway. It is thought to be essential for visual object recognition, since a bilateral lesion of TE produces severe and selective deficits in discrimination and recognition of visual images of objects (Blum et al., 1950; Chow, 1951; Mishkin, 1954; Mishkin and Pribram, 1954) [for a review see Gross (Gross, 1994)]. Correspondingly, cells in TE respond selectively to the complex features of object images (Gross et al., 1972; Desimone et al., 1984; Tanaka et al., 1991; Fujita et al., 1992; Kobatake and Tanaka, 1994; Ito et al., 1994, 1995; Sheinberg and Logothetis, 1997).

TE can be divided into a dorsal and ventral part (Table 1). Brodmann divided the inferotemporal region, which corresponds to TE, into areas 20 (ventral) and 21 (dorsal) (Brodmann, 1909). The border between these two regions is located at the anterior middle temporal sulcus. It was later found that the ventral and dorsal parts of TE have differential connections with the amygdala (Iwai and Yukie, 1987; Iwai et al., 1987; Cheng et al., 1997), striatum (Cheng et al., 1997), superior temporal sulcus (Saleem et al., 2000), perirhinal and entorhinal cortices (Saleem and Tanaka, 1996) and hippocampal formation (Iwai and Yukie, 1988; Yukie and Iwai, 1988; Saleem and Hashikawa, 1998). Based on these ventral versus dorsal differences as well as the previously reported posterior versus anterior difference in the afferent connections from the prelunate gyrus (Shiwa, 1987; Morel and Bullier, 1990; Yukie et al., 1990), TE has been divided into four sub-regions; TEpv (posterio-ventral), TEpd (posterio- dorsal), TEav (anterio-ventral) and TEad (anterio-dorsal). Saleem and Tanaka (Saleem and Tanaka, 1996) found that the border between TEav and TEad described by Yukie et al. (Yukie et al., 1990) corresponds to the cytoarchitectural border between TE1 and TE2 described by Seltzer and Pandya (Seltzer and Pandya, 1978) at the anterior–posterior levels where the anterior middle temporal sulcus is found. They proposed using the cyto- architectural TE1/TE2 border as the TEav/TEad border. This border is located around the lip or at some depth in the lateral bank of the anterior middle temporal sulcus, depending on the anterior–posterior level. It has also been shown that TEav and TEad receive visual inputs through separate pathways in parallel (Martin-Elkins and Horel, 1992; Yukie et al., 1992; Saleem et al., 2000). Felleman and Van Essen divided TE into CITv, CITd, AITv and AITd (Felleman and Van Essen, 1991). The anterior– posterior level of CITv and CITd corresponds to that of TEpd and TEpv and that of AITv and AITd to that of TEad and TEav. However, the border between AITd and AITv is located close to the ventral lip of the superior temporal sulcus. Thus, AITv overlaps both TEad and TEav.

The stimulus selectivity of TEav cells is largely unknown, while the selectivity of TEad cells has been extensively studied (Desimone et al., 1984; Tanaka et al., 1991; Fujita et al., 1992; Kobatake and Tanaka, 1994; Ito et al. 1994, 1995). Most previous studies of responses of TEav cells mainly focused on response aspects other than stimulus selectivity (Miyashita, 1988; Miller et al, 1991). Moreover, only one study has compared TEav and TEad. Baylis and co-workers (Baylis et al., 1987) recorded a large number of cells from a wide extent of the inferotemporal cortex and the superior temporal sulcus in monkeys performing a fixation task and compared response properties among the different cytoarchitectural areas defined according to Seltzer and Pandya (Seltzer and Pandya, 1978). However, Baylis et al. were only able to activate 36% of cells recorded from TE1 and TE2 (Baylis et al., 1987). Thus, the first aim of the present study was to compare the response properties of cells in TEad and TEav.

A time course analysis of visually evoked responses may help identify the carrier of information and provide an insight into the mechanisms by which the stimulus selectivity of responses arises. There are two contradictory views regarding the carrier of information in the responses of inferotemporal cells. One view emphasizes the importance of temporal modulation over the period of stimulus presentation (several hundred milli- seconds). It has been shown that temporal patterns of responses recorded from TEad cells were stimulus dependent and that the temporal patterns carried a significantly larger amount of information about the stimulus than the information derived from the averaged firing rate (Optican and Richmond, 1987; Richmond et al., 1987). The other view emphasizes the richness of information contained in the early phase of responses. It has been shown that the average firing rate within 150 ms after stimulus onset represents most of the information carried by the whole response, including the temporal modulation (Tovee et al., 1993). Consistent with this finding, electroenchepalogram recordings from the prefrontal cortex showed that 150 ms is enough time to judge whether there is an animal in a complex scene (Thorpe et al., 1996; Fabre-Thorpe et al., 1998). The second aim of the present study was to examine the temporal patterns of responses in TEav and TEad cells.

We routinely used a large number (100) of visual stimuli in the present study to quantitatively analyze the response properties of TEav and TEad cells. We presented all 100 stimuli to all the recorded cells. We believe that the use of a large number of stimuli is essential in studying stimulus selectivity, because it has been reported that most TE cells respond to specific stimuli (Tanaka et al., 1991; Fujita et al., 1992; Kobatake and Tanaka, 1994; Ito et al., 1994, 1995; Sheinberg and Logothetis, 1997). An incorrect conclusion might be drawn if the stimulus set did not include the optimal or sub-optimal stimuli for individual cells. We used awake, behaving animals in this study because we found it difficult to activate cells in TEav in anesthetized animals (H. Tamura and K. Tanaka, unpublished observation). Some of the results have been reported previously in abstract form (Tamura and Tanaka, 1996, 1997).

Materials and Methods

Cell responses were recorded from two macaque monkeys (Macaca fuscata) while they performed a visual fixation task. The monkeys weighed 5.4 and 7.4 kg. All procedures were performed in accordance with the Guiding Principles for the Care and Use of Animals in the Field of Physiological Science (Japanese Physiological Society).

A lever press by the monkey initiated a trial. A small white spot (0.15° in diameter) was presented on a CRT display (PCKM153; NEC, Tokyo, Japan), which was placed at 57 cm from the eye. The monkey was required to fixate the spot within 500 ms of spot onset and continue fixating for a period varying randomly between 2.5 and 7 s. The size of the fixation window was set to 4 × 4°. At the end of each trial the color of the fixation spot changed slightly and the monkey had to release the lever within 500 ms. Correct performance on a trial was rewarded with a drop of juice. A 1.5 s inter-trial interval was imposed before the next trial could begin.

Preparation

The behavioral task was first trained without head fixation or eye position measurement. After the monkeys learned the task they were prepared for cell recordings by aseptic surgery. The surgery commenced with injection of atropine sulfate (0.1 mg/kg i.m.). Anesthesia was induced by i.m. injection of ketamine hydrochloride (12 mg/kg), maintained by i.p. injection of sodium pentobarbital (35 mg/kg) and supplemented by additional injections of sodium pentobarbital (9 mg/kg each). Three blocks for head restraint and a recording chamber were implanted in the dorsal surface of the skull. After a week of recovery, training was resumed, but this time the eye position was measured. Cell recording was started when the monkey had succeeded in completing the task with continuous eye fixation.

Recording

Extracellular single cell recordings were made from anterior infero- temporal cortex with tungsten electrodes (FHC, ME) while the monkeys were performing the fixation task. The electrode was advanced with an Evart type manipulator (MO95; Narishige, Tokyo, Japan) from the dorsal surface of the brain through a guide cannula, which was inserted to ~15 mm above the recording sites. We determined the position of the dura at the ventral surface of the brain by detecting the characteristic noise in the recorded voltage when the electrode tip contacted the dura and limited the studies to the gray matter at the ventral surface of the brain (Fig. 1). Except for a few cells at the ventral lip, no cells in the superior temporal sulcus were studied. We did not select cells according to their responsiveness, but always studied the first well-isolated cells (with spikes >4 times as large as the baseline noise) at the intended depth. Successively studied cells were spaced at 100 μm intervals or greater. In some penetrations the initial part of the gray matter was ignored so that different cortical layers would be evenly sampled.

The timing of action potentials and lever press and release by the monkey were recorded with 1 ms resolution and stored on a personal computer (PC9821; NEC). Peristimulus time histograms were displayed on-line. The eye position was monitored every 16 ms by an infrared method. An infrared laser beam was projected onto the cornea of the right eye and the reflected image was captured by a charge coupled device camera. The eye position was then determined by an image processor (Percept Scope; Hamamatsu Photonics, Hamamatsu, Japan). At the end of the day's recording we took X-ray images of the electrode within the brain to determine the electrode position relative to the bony landmarks.

Data Analysis

The statistical significance of responses was determined in two steps. First, cells were examined for stimulus selectivity. Response magnitude in individual trials was calculated as the mean firing rate during the 400 ms response period (with the window position shifted by 60 ms from the stimulus presentation period to compensate for the response latency) minus the spontaneous firing rate during the 200 ms period immediately prior to stimulus onset. Its dependency on the stimulus was assessed by one-way ANOVA (P < 0.01). Second, the significance of individual responses was examined for the cells selected in the first step by comparing the distribution of firing rates during the 400 ms response window in individual trials with that of the spontaneous firing rates during the 200 ms period immediately prior to stimulus onset in individual trials [P < 0.01 by the Kolmogorov–Smirnov (K-S) test]. A cell was further analyzed if its responses were stimulus selective and at least one of the 100 visual stimuli evoked significant responses in it. The first step was necessary to avoid false detection of a responsive cell due to multiple statistical examination of 100 responses in individual cells.

To determine the latency of response onset and peak response, the spike trains of individual trials were first converted into 1 ms resolution spike density functions by convoluting the spike trains with a Gaussian kernel (σ = 5 ms). Next, we determined the level required to significantly exceed activity within the 200 ms spontaneous period (P = 0.01 by K-S test). The onset of the response was designated the first bin that exceeded this level following stimulus onset and was followed by five consecutive bins also exceeding this level. The peak response was designated the peak with the maximum magnitude in the spike density function and the time from stimulus onset to response peak was measured as peak latency. When multiple peaks of the same amplitude were found, the first peak was taken as the response peak. Decay time was defined as the latency between stimulus onset and the first bin in the spike density function at which the firing rate returned to the spontaneous firing range after the response peak.

Unless otherwise noted, the magnitude of responses is computed as the mean firing rate during the whole 400 ms response window (with the window position shifted by 60 ms from the stimulus presentation period) minus the mean spontaneous firing rate during the 200 ms period immediately preceding stimulus onset. The height of the response peak beyond the spontaneous firing level was also used as another measure of response magnitude when we examined the correlation between response time course parameters and response magnitude.

We defined the tuning width of individual cell excitatory responses using two measures. The first was the number of stimuli that evoked responses >25% of the maximal cell response (TW25). TW25 was measured to the first digit by interpolating responses just above and below 25% of the maximal response. We also counted the number of effective stimuli that evoked significant (P < 0.01) excitatory responses in individual cells (TWP<0.01). TWP<0.01 was measured as an integer.

$\mathit{I}(\mathit{s},\mathit{r})\ {=}\ {{\sum}_{\mathit{s}\ {=}\ 1}^{\mathit{S}}}{{\sum}_{\mathit{r}\ {=}\ 1}^{\mathit{R}}}\ \mathit{P}(\mathit{s},\mathit{r})log{\{}{[}\mathit{P}(\mathit{s},\mathit{r}){]}/{[}\mathit{P}(\mathit{s})\mathit{P}(\mathit{r}){]}{\}}$
where s represents the stimulus type and r the response type. S and R are the number of possible stimuli and possible responses, respectively. r and R depend on the decoding method. We used the decoding method formulated by Optican and Richmond (Optican and Richmond, 1987). The discrete distribution of responses (spikes/bin) to a stimulus s in individual trials was converted to a continuous probability density function P(rs) by convoluting the discrete events with the Gaussian function specified by the Parzen estimate (Parzen, 1962). The range of the original distribution was analyzed with 1/3 spike/bin resolution. P(r) was obtained by summing P(rs) for all stimuli. P(s,r) was given by Bayes' law as
$\mathit{P}(\mathit{s},\mathit{r})\ {=}\ \mathit{P}(\mathit{s})\mathit{P}(\mathit{r}{\mid}\mathit{s})$
P(s) was identical for all the stimuli and 1/100 in our case.

It may be worth mentioning that the information and selectivity indices are different measures. When a cell responds to many stimuli with slightly different amplitudes the selectivity is low while the amount of information carried by the responses is large. Conversely, when a cell responds to only a limited number of stimuli the selectivity is high while the amount of information is small.

The correlation between two different measures taken for a given cell or cell response (for example the onset latency of the maximal response versus the stimulus selectivity) was examined by calculating the Pearson's correlation coefficient. However, where both of the two measures correlated with a common third measure (for example the magnitude of the maximum response), the net correlation between the first two measures was examined by calculating the partial correlation coefficient given by the formula:

$\mathit{r}_{\mathit{xy\ {\cdot}\ z}}\ {=}\ (\mathit{r}_{\mathit{xy}}\ {\mbox{--}}\ \mathit{r}_{\mathit{yz}}\mathit{r}_{\mathit{zx}})/{[}{\surd}(1\ {\mbox{--}}\ \mathit{r}_{\mathit{yz}}^{2})\ {\surd}(1\ {\mbox{--}}\ \mathit{r}_{\mathit{zx}}^{2}){]}$
where rxy • z represents the partial correlation coefficient between x and y controlling the effects of z and rxy, ryz and rzx represent the correlation coefficients between x and y, y and z and z and x, respectively.

Histology

After completion of the recordings the monkeys were deeply anesthetized and perfused in the heart. The brain was removed and cut into 50 μm thick sections for reconstruction of electrode penetration. The penetrations were reconstructed using the damage trace of the electrode penetrations and the X-ray images taken at the end of each individual penetration. During recording we had determined the location of transition from gray matter to white matter and the position of the dura at the ventral surface of the brain (see above). We determined the vertical positions of recorded cells along the penetrations using these positions together with the readings of the electrode manipulator.

The sampling of cells covered a posterior–anterior extent from 13 to 21 mm anterior to the ear bar, where the anterior middle temporal sulcus is located (Fig. 1). Therefore, the recording sites corresponded to area TEa. The lateral–medial extent covered from the ventral lip of the superior temporal sulcus to around the midpoint between the anterior middle temporal sulcus and the rhinal sulcus (Fig. 1). We followed Saleem and Tanaka for determination of the borders between TEad, TEav and the perirhinal cortex (Saleem and Tanaka, 1996). Several cells were excluded from the analysis because the recordings were taken at the border between TEav and TEad.

Cells in the Perirhinal Cortex

In addition to cells recorded from TEav and TEad, we recorded 32 cells from the perirhinal cortex. Only 12 cells (38%) were stimulus selective (P < 0.01 by ANOVA) and showed significant responses (P < 0.01 by K-S test) to at least one of the 100 stimuli. The proportion of such cells in perirhinal cortex was significantly smaller than in TEav (P < 0.001 by χ2 test) and TEad (P < 0.001 by χ2 test). The magnitude of the maximal excitatory response of individual perirhinal cortex cells (14.4 ± 16.9 spikes/s) tended to be smaller than that of TEav (P = 0.032 by U-test) and TEad cells (P < 0.001 by U-test), while the spontaneous firing rate of perirhinal cells (7.5 ± 11.7 spikes/s) did not differ significantly from that of TEav (P = 0.239 by U-test) or TEad cells (P = 0.093 by U-test). Since we found that perirhinal cells were not very active under the present experimental conditions (fixation task), we made no further recordings from these cells and excluded the 32 cells from the analyses in this paper.

Results

Time Course of Responses

We will first describe the stimulus dependence of the response time course commonly found in TEav and TEad cells and then compare various aspects of the responses between TEav and TEad.

Stimulus Selectivity

$Transience\ {=}\ (initial\ {\mbox{--}}\ late)/(initial\ {+}\ late)$
where ‘initial’ and ‘late’ represent the averaged firing rate (spikes/s) above the spontaneous firing level during the 110–150 and 330–370 ms time windows, respectively. Transience varied between –1 and 1 among responses, but the distribution had a single peak at ~0.4 with a mean ± SD of 0.26 ± 0.46 (median 0.33; Fig. 5, bottom). This means that a majority of excitatory responses had both initial and late components, with the initial components bigger than the late components.

The tuning of excitatory response stimulus selectivity tended to be broader and the amount of transmitted information larger in cells with greater maximal excitatory responses. This conclusion is based on the results that TW25, TWP<0.01 and the information content were significantly correlated with the maximal responses in individual cells (r = 0.40 for TW25, r = 0.72 for TWP<0.01 and r = 0.46 for information, all P < 0.001). They were also significantly correlated with the maximal peak response in individual cells (r = 0.48, P < 0.001 for TW25; r = 0.62, P < 0.001 for TWP<0.01; r = 0.34, P = 0.022 for information).

Magnitude and Variance

Of the 213 cells examined in this study, 113 were recorded from TEav and 100 from TEad. Eighty-two of the 113 TEav cells (73%) and 87 of the 100 TEad cells (87%) were stimulus selective (P < 0.01 by ANOVA) and showed significant responses (P < 0.01 by K-S test) to at least one of the 100 stimuli. The proportion of such responsive cells was significantly smaller in TEav than in TEad (P = 0.009 by χ2 test). The proportion of cells with significant excitatory responses was significantly smaller in TEav (73/113) than in TEad (83/100, P = 0.002), while the proportion of cells with significant inhibitory responses did not differ significantly between TEav (30/113) and TEad (20/100, P = 0.797). Thus, the difference in the proportion of significant responses between the two areas was mostly due to the difference in the proportion of significant excitatory responses.

The magnitude of the maximal excitatory response of individual TEav cells tended to be smaller than that of individual TEad cells, while the spontaneous firing rate averaged over 100 responses and the magnitude of the maximal inhibitory response did not differ significantly between TEav and TEad cells. The magnitude of the maximal excitatory response, measured as the mean firing rate in the 400 ms response window minus the spontaneous firing level for individual cells, ranged from 2.5 to 141.0 spikes/s. The means ± SD were 22.5 ± 16.9 spikes/s (median 17.9) in TEav and 35.0 ± 27.9 spikes/s (median 27.0) in TEad. The two distributions differed significantly (P = 0.002 by U-test). On the other hand, the means ± SD of the averaged spontaneous firing rate were 7.2 ± 7.5 spikes/s (median 5.0) in TEav and 7.6 ± 6.9 spikes/s (median 6.0) in TEad. The two distributions were not significantly different (P = 0.275 by U-test). The magnitude of the maximal inhibitory response, measured as the spontaneous firing level minus the mean firing rate in the 400 ms response window, ranged from 4.0 to 28.4 spikes/s. The means ± SD were 10.8 ± 5.9 spikes/s (median 9.0) in TEav and 12.2 ± 6.4 spikes/s (median 11.5) in TEad. The two distributions were not significantly different (P = 0.208 by U-test).

There was no significant difference between TEav and TEad in the trial-by-trial variance of excitatory responses to specific stimuli. For each cell the trial-by-trial variances of responses to the 100 stimuli were plotted against their means on a log10 scale and the slope of the regression line was calculated. The variance was calculated for 52 TEav and 61 TEad cells. The remaining 30 TEav and 26 TEad cells had some responses smaller than 0, which could not be plotted on a log10 scale. The variances and means were significantly (P < 0.05) correlated in 51 of 52 TEav cells and 60 of 61 TEad cells. The slope of the regression line represents the magnitude of the variance in comparison to the mean. The larger the slope, the more variable the cell responses were. The slope was 1.01 ± 0.31 for TEav and 0.96 ± 0.34 for TEad, with no significant difference between the two distribuions (P = 0.375 by U-test).

Stimulus Preference

To compare the overall effectiveness of individual stimuli in activating TEav cells versus TEad cells, the number of cells in which a stimulus evoked statistically significant (P < 0.01) excitatory responses was counted separately for each stimulus. The counts were determined separately for 73 TEav and 83 TEad cells that showed significant excitatory responses to at least one stimulus and normalized to the total number of cells with significant excitatory responses in each area. We refer to these normalized values as ‘overall effectiveness’. A stimulus effective for all the responsive cells would thus have an overall effectiveness of 1. An overall effectiveness of 0 means that the stimulus failed to elicit excitatory responses in any responsive cell.

$Colorfulness\ {=}\ {\{}{\Sigma}{\surd}{[}(red\ {\mbox{--}}\ mean)^{2}\ {+}\ (green\ {\mbox{--}}\ mean)^{2}\ {+}\ {\ }{\ }{\ }(blue\ {\mbox{--}}\ mean)^{2}{]}{\}}\ /25\ 600$

$mean\ {=}\ (red\ {+}\ green\ {+}\ blue)/3$
‘Red’, ‘green’ and ‘blue’ are the intensity of the red, green and blue components of the color of individual pixels in the stimulus images, which took a value between 0 and 255, and ‘mean’ represents the mean luminance level of each pixel. A summation was taken over all pixels (25 600 = 160 × 160). A Colorfulness value of 0 indicates that the image was composed of pixels identical in hue to the gray background. The greater Colorfulness, the more colorful the image was. In our stimulus set Colorfulness ranged from 0 (the black and white and gray images) to 68 (stimulus 6, the image of a flower; Fig. 13).

Sharpness of Stimulus Selectivity

Changes in stimulus selectivity were observed throughout the response time course in both TEav and TEad cells. To compare the degrees of stimulus selectivity sharpening between TEav and TEad we compared the distributions of the differences between the initial and late parts of responses for TW25, TWP<0.01 and amount of information. There were no significant differences between the two areas (P > 0.05 by U-test).

Onset Latency

In both TEav and TEad cells a negative correlation was found between onset latency and peak height of responses within individual cells and between peak latency and peak height. A positive correlation was observed between decay time and peak height. The incidence of a negative correlation between onset latency and peak height of responses did not differ significantly between TEav (5/12 cells) and TEad (21/30 cells) (P = 0.088 by χ2 test). Likewise, the incidence of a negative correlation between peak latency and peak height of responses did not differ significantly between TEav (4/12 cells) and TEad (13/30 cells) (P = 0.551). Finally, the incidence of a positive correlation between decay time and peak height of responses did not differ significantly between TEav (5/11 cells) and TEad (15/28 cells) (P = 0.648).

Discussion

There are several previous studies of the stimulus selectivity of TE cells in anesthetized monkeys. By comparing the present results obtained in alert, fixating monkeys with those obtained in anesthetized monkeys we can deduce the effects of anesthesia on TE cell response. Compared with the previous results in anesthetized preparations, the proportion of responsive cells in alert, fixating monkeys was larger and the magnitudes of their maximal responses were greater. The proportion of unresponsive cells in TEad was only 16% in the present study, compared with 35% in TEad and TEpd in anesthetized monkeys (Tanaka et al., 1991). The mean of the maximal responses of responsive TEad cells was 40.5 spikes/s in the present study, whereas for the maximal responses of responsive cells in TEad and TEpd it was 17.5 spikes/s in Tanaka et al. (Tanaka et al., 1991) and 17.5 spikes/s in Kobatake et al. (Kobatake et al., 1994).

In contrast, the tunings of stimulus selectivity obtained in the present study were not very different from the tunings obtained in previous studies with anesthetized monkeys. Responses >25% of the maximal in individual cells were evoked only by one fifth of the 100 stimuli in TEad cells in the present study. Because most of the stimuli used in the present study were object images and contained multiple features, the tuning would have been sharper if we had measured the sharpness in the feature domain.

Time Course of Responses

We found that the response time course of cells in TEav and TEad depended on the stimulus. Different stimuli evoked responses with different time courses even in the same TE cells. Such stimulus-dependent temporal modulation of TE cell response was first reported by Richmond et al. (Richmond et al., 1987). In addition to confirming this previous finding, we found that a principal factor in the variation of the response time course is the difference between responses that have initial transient components only ~130 ms after stimulus onset and responses that have late components after 240 ms from stimulus onset in addition to the initial transient components. Thus, the late components are more selective than the initial components. However, the function of the later responses is not obvious. By detecting the no-go potential (Gemba and Sasaki, 1989; Sasaki et al., 1993) in the electroencephalogram of the human frontal cortex, Thorpe et al. found that perceptual judgment for the presence/absence of any animal in a complex scene ends within 150 ms after stimulus onset (Thorpe et al., 1996). Thus, the later responses may be used to recover the 3-dimensional reality of the object or to form memory traces in neuronal circuits. The sparser representation of object images in the later parts of responses has the advantage of increasing memory capacity (Field, 1994; Foldiak and Young, 1995).

Another principal factor in the response time course variation was the difference in onset and peak latencies, as reported by Richmond (Richmond et al., 1987; Liu and Richmond, 2000). The time for the firing rate to decay from the initial peak to the spontaneous firing level also varied. We found that stronger responses tended to have shorter onset and peak latencies and longer decay time from the initial peak to the spontaneous firing level. The rapid visual discrimination of complex stimuli found by Thorpe et al. (Thorpe et al., 1996) may be achieved by detecting the difference in onset and peak latencies, rather than by comparing the magnitudes of response peaks. The coding of information by response latency has also been discussed theoretically (Hopfield, 1995; Korner et al., 1999).

There are several other previous studies relevant to the stimulus-dependent time course of TE cells. Oram and Perrett analyzed the time course of cell responses in the upper bank of the superior temporal sulcus to various views of the human head (Oram and Perrett, 1992). They found that while onset latency was almost identical among responses to different head views, the stronger responses to good views had a longer decay than the weaker responses to bad views. Although the results with onset latency are not consistent with the present study, the decay time results agree. The difference in the onset latency results may be due to the difference in visual stimuli and/or differences in recording sites.

Tovee et al. and Gershon et al. calculated the cumulative information content and found that most of the information about the stimulus existed in the activity during the first 200 ms period after stimulus onset (Tovee et al., 1993; Gershon et al., 1998). Gochin et al. analyzed the time course of ensemble performance, which is the probability of successfully classifying stimuli based on the activity of a population of cells, and found a rapid increase in ensemble performance after stimulus onset (Gochin et al., 1994). Consistent with these previous findings, we found a rapid increase in information content after stimulus onset in the present study.

Sugase et al. have recently reported that the information content grew faster for global categories (human faces/monkey faces/simple shapes) than for fine categories (human or monkey facial expressions and identity) after stimulus onset (Sugase et al., 1999). The mean difference between the two peaks was 51 ms. The peaks of global category information coincided with the discharge peaks and the peaks of fine category information were located in the decay phase, where the discharge rate decayed from the initial peak to the spontaneous firing level or the maintained firing level. Thus, this finding of Sugase et al. (Sugase et al., 1999) may correspond to the finding by Oram and Perrett (Oram and Perrett, 1992) and ours that the speed of decay from the initial discharge peak varies depending on the stimulus. We assume that the rising phase of the initial discharge coincided between responses to faces with different expressions and identities, while some of the responses decayed faster than others and the variation in the discharge rate was greatest during the decay phase, thereby giving the most information about expressions and identities.

The difference in response onset latency and the negative correlation between onset latency and response magnitude can be explained by the difference in temporal summation requirements for action potential generation between optimal and less optimal stimuli. The optimal stimuli evoke strong excitatory synaptic potentials, which immediately trigger action potentials, while less optimal stimuli evoke weaker synaptic potentials for which temporal summation is needed to trigger action potentials. Such latency differentiation may occur not only in the cell being recorded, but also in cells located in its afferent pathway. It is also possible that weaker inputs evoked by less optimal stimuli need to be supported by additional excitatory inputs conveyed through lateral interaction to trigger action potentials, while strong inputs evoked by optimal stimuli trigger action potentials by themselves. The inputs conveyed through lateral interaction come later than the direct feed-forward afferent inputs.

We can suggest two possible mechanisms for the later sharpening of stimulus selectivity. One possibility is that there is a correlation between selectivity and onset latency in the inputs from lower stages. This supposes that less selective inputs are conducted more quickly while more selective inputs are conducted more slowly. Then, the initial responses evoked by the fast inputs will be less selective than those evoked by the slow inputs. We do not think this is plausible, because while we predict, on the same supposition, that TE cells with longer onset latencies are more selective, we did not find any significant net correlation between either TW25 or TWP<0.01 and minimum onset latency of individual cells. A second possibility is mutual inhibition between TE cells. A relatively large number of TE cells are first activated, but weaker responses disappear through mutual inhibition and only strong responses make it to the later response phase. Indeed, we found that the start of the inhibitory response coincided with decay of the excitatory response. Excitatory interactions between TE cells may also contribute to selection. Cells with excitatory connections to other activated cells would remain active beyond the initial phase. Fujita and Fujita showed extensive horizontal connections in TE (Fujita and Fujita, 1996). Mutual interactions between or within columns can explain both the differences in the decay speed from the initial peak and the magnitudes of late response components.

Functional Differences between TEav and TEad

Although we did not examine the responses of cells in the superior temporal sulcus in the present study, previous studies have found some unique properties in cells of the superior temporal sulcus [reviewed by Allison et al. (Allison et al., 2000)]. Baylis et al. found that cells selectively responding to faces were more numerous in areas TPO, TEa and TEm than in other sub-regions of the inferotemporal cortex (Baylis et al., 1987). Janssen et al. found that cells sensitive to disparity gradients were more numerous in the lower bank of the superior temporal sulcus than in TEad (Janssen et al., 2000). The present results, together with the previous results described above, suggest that the inferotemporal cortex is composed of several functionally distinct sub-divisions.

The tuning width defined at 25% of the maximal response (TW25) was not significantly different between cells in TEav and TEad, while the number of effective stimuli that evoked significant responses in individual cells (TWP<0.01) was significantly larger in TEad than in TEav and the information transmitted by responses of single cells was significantly greater in TEad than in TEav. TWP<0.01 and the amount of information are dependent on the trial-by-trial response variance to given stimuli as well as the shape of the tunings. TW25 might also be influenced by the trial-by-trial variance: responses with real magnitudes just below 25% may happen to exceed 25% due to the variance and responses just above 25% may happen to be smaller than 25%. However, if the number of repetitions is large enough compared with the variance, this influence will be minimal. We compared TW25 measured over the averaged responses of 10 repetitions, with TW25 measured on half of the data (five repetitions) for 54 cells for which we had completed 10 repetitions of the 100 stimuli. There was a strong correlation between the two sets of TW25 measures (r = 0.98) and the slope of the regression line was close to 1 (slope = 0.97). Therefore, we believe that TW25 was the most direct measure of the tuning shape and concluded that stimulus selectivity did not differ significantly between TEav and TEad cells.

However, because most of the stimuli used in the present study were object images, we can not deduce the features critical to selectivity. Horel suggested that TEad held information about fine stimulus features (Horel, 1994a). Although Horel (Horel, 1994a,b) did not mention it, a possibility that follows from this notion is that the global features of stimuli are processed in TEav. It is still possible, despite the present results, that the stimulus selectivity of TEav and TEad cells is determined by different stimulus domains.

The probability of a real complex object image evoking significant responses was significantly larger than the two- dimensional geometrical shapes in both TEav and TEad cells. We have previously shown that most cells in TEad respond to moderately complex features but not to whole object images (Tanaka et al., 1991; Fujita et al., 1992; Kobatake et al., 1994). The superiority of object images to geometrical shapes for TEad cells can be easily understood by considering that complex object images contain many features. The superiority of object images to geometrical shapes was more prominent for TEav than TEad cells. The above explanation for TEad cells may also hold for TEav cells, but it is also possible that TEav cells respond preferentially to whole object images rather than to moderately complex features.

A superiority of colorful object images to less colorful ones was found in TEav but not in TEad. This may appear inconsistent with the two lines of previous findings. One is the presence of color-selective cells in TEad (Desimone et al., 1984; Tanaka et al., 1991; Komatsu et al., 1992). The second is the effects of lesioning and inactivation of TEad on color discrimination tasks. Horel found that cooling TEad degraded simple red–green and blue–yellow discrimination (Horel, 1994b) and Buckley et al. found that a lesion of TEad impaired discrimination of small differences in hue or color saturation (Buckley et al., 1997). However, Horel and Buckley et al. did not examine the effects of cooling or lesioning TEav and there are no previous studies on the color selectivity of TEav cells. The color selectivity of TEav cells, especially with complex images, may be more prominent than in TEad cells. Indeed, there is an abstract (Sato and Kawamura, 1990) reporting that color-sensitive cells were more frequently observed in the ventral region than in the dorsal region.

The minimum onset latencies of individual cells were not significantly different between TEav and TEad. This result is consistent with previous anatomical studies and supports the idea that TEav and TEad are located at the same hierarchical levels of different serial pathways rather than at serial stages along a single pathway. Martin-Elkins and Horel showed that the anterior inferior temporal gyrus between the anterior middle temporal sulcus and the rhinal sulcus (which included a part of TEav) received strong projections from the posterior regions located along the occipitotemporal sulcus on the ventral surface of the brain (Martin-Elkins and Horel, 1992). The anterior inferior temporal gyrus does not receive projections from the posterior regions on the lateral surface of the brain, from which TEad receives strong projections (Desimone et al., 1980; Shiwa, 1987; Webster et al., 1991; Saleem et al., 1993). Based on these data, it has been suggested that TEav and TEad receive their visual afferent inputs through different pathways in parallel (Martin-Elkins and Horel, 1992; Yukie et al., 1992). We have recently confirmed this with more localized dye injections into TEav and TEad (Saleem et al., 2000).

Notes

We wish to thank Dr Hidekazu Kaneko for his contribution to the data analysis. This work was supported by RIKEN Brain Science Institute and the Agency of Industrial Science and Technology.

Address correspondence to Hiroshi Tamura, Department of Bio- physical Engineering, Graduate School of Engineering Science, Osaka University, Toyonaka, Osaka 560-8531, Japan.

Table 1

Correspondence between different sub-division systems of the anterior inferotemporal cortex

Brodmann Selzer and Pandya Felleman and Van Essen
The correspondences are not perfect.
TEav Area 20 TE1 AITv
TEad Area 21 TE2 AITv, AITd
Brodmann Selzer and Pandya Felleman and Van Essen
The correspondences are not perfect.
TEav Area 20 TE1 AITv
TEad Area 21 TE2 AITv, AITd
Figure 1.

The recording sites of 113 cells in TEav (open circles) and 100 cells in TEad (filled circles). (Top) Lateral views of monkey brains. Vertical lines indicate the antero-posterior extent of recordings. A13, A15, A17, A19 and A21 indicate that the frontal sections are located at 13, 15, 17, 19 and 21 mm anterior to the ear bar position, respectively. The arrows indicate the border between the TEav and perirhinal cortex (area 36). amts, anterior middle temporal sulcus; rs, rhinal sulcus; sts, superior temporal sulcus; a, amygdala; h; hippocampus.

Figure 1.

The recording sites of 113 cells in TEav (open circles) and 100 cells in TEad (filled circles). (Top) Lateral views of monkey brains. Vertical lines indicate the antero-posterior extent of recordings. A13, A15, A17, A19 and A21 indicate that the frontal sections are located at 13, 15, 17, 19 and 21 mm anterior to the ear bar position, respectively. The arrows indicate the border between the TEav and perirhinal cortex (area 36). amts, anterior middle temporal sulcus; rs, rhinal sulcus; sts, superior temporal sulcus; a, amygdala; h; hippocampus.

Figure 2.

The 100 stimulus images used in this study.

Figure 2.

The 100 stimulus images used in this study.

Figure 4.

Peristimulus time histograms for a second TEav cell. The full scale of the y-axis is 150 spikes/s. Other conventions are as in Figure 3.

Peristimulus time histograms for a second TEav cell. The full scale of the y-axis is 150 spikes/s. Other conventions are as in Figure 3.

Figure 5.

Averaged peristimulus time histograms of all significant excitatory responses (top) and the distribution of the Transience (bottom). The black bar on the horizontal axis in the top figure indicates the stimulus presentation period. The window for the initial part of responses was set from 110 to 150 ms after stimulus onset (initial peak) and for the later part of the responses it was set from 330 to 370 ms after stimulus onset (late peak). The window for the response trough was set from 230 to 270 ms after stimulus onset. Transience was defined as:

$Transience\ {=}\ (initial\ {\mbox{--}}\ late)/(initial\ {+}\ late)$
where ‘initial’ and ‘late’ represent the averaged firing rate (spikes/s) during the initial peak and late peak windows, respectively, above the spontaneous firing rate.

Figure 5.

Averaged peristimulus time histograms of all significant excitatory responses (top) and the distribution of the Transience (bottom). The black bar on the horizontal axis in the top figure indicates the stimulus presentation period. The window for the initial part of responses was set from 110 to 150 ms after stimulus onset (initial peak) and for the later part of the responses it was set from 330 to 370 ms after stimulus onset (late peak). The window for the response trough was set from 230 to 270 ms after stimulus onset. Transience was defined as:

$Transience\ {=}\ (initial\ {\mbox{--}}\ late)/(initial\ {+}\ late)$
where ‘initial’ and ‘late’ represent the averaged firing rate (spikes/s) during the initial peak and late peak windows, respectively, above the spontaneous firing rate.

Figure 6.

Averaged time course of the tuning width at 25% of the maximum response (TW25, top) and the number of effective stimuli that evoked significant (P < 0.01) responses in individual cells (TWP<0.01, middle) and the amount of transmitted information (bottom). Filled and hatched bars indicate the windows for initial and late peaks, respectively.

Figure 6.

Averaged time course of the tuning width at 25% of the maximum response (TW25, top) and the number of effective stimuli that evoked significant (P < 0.01) responses in individual cells (TWP<0.01, middle) and the amount of transmitted information (bottom). Filled and hatched bars indicate the windows for initial and late peaks, respectively.

Figure 7.

A comparison of the stimulus selectivity between the initial and late responses of all responsive cells. (Top) TW25; (middle) TWP<0.01; (bottom) the amount of transmitted information between the initial and late peaks. Points represent single cells.

Figure 7.

A comparison of the stimulus selectivity between the initial and late responses of all responsive cells. (Top) TW25; (middle) TWP<0.01; (bottom) the amount of transmitted information between the initial and late peaks. Points represent single cells.

Figure 8.

Averaged peristimulus time histograms of all significant inhibitory responses (thick line) compared with that of all significant excitatory responses (thin line). The black bar on the horizontal axis indicates the stimulus presentation period.

Figure 8.

Averaged peristimulus time histograms of all significant inhibitory responses (thick line) compared with that of all significant excitatory responses (thin line). The black bar on the horizontal axis indicates the stimulus presentation period.

Figure 9.

The distributions of onset latency (top), peak latency (middle) and decay time (bottom) among all significant (P < 0.01) excitatory responses. The decay time is the time from stimulus onset to the bin at which the firing rate returned to the spontaneous firing level after the initial peak (see text for the exact definition).

Figure 9.

The distributions of onset latency (top), peak latency (middle) and decay time (bottom) among all significant (P < 0.01) excitatory responses. The decay time is the time from stimulus onset to the bin at which the firing rate returned to the spontaneous firing level after the initial peak (see text for the exact definition).

Figure 10.

Scatter plots of onset latency (top), peak latency (middle) and decay time (bottom) against the peak height for all significant excitatory responses.

Figure 10.

Scatter plots of onset latency (top), peak latency (middle) and decay time (bottom) against the peak height for all significant excitatory responses.

Figure 11.

Correlation of onset latency, peak latency and decay time with peak response height in a single TEad cell. (Top) Spike density functions of three responses with different peak heights (maximum, 60% and 40% of maximum). Solid arrows indicate the onset of responses, solid double arrows the response peaks and open arrows the decay from the peak to the spontaneous level. (Lower three) Scatter plots of onset latency, peak latency and decay time against peak height of individual responses. The linear regression lines are superimposed.

Figure 11.

Correlation of onset latency, peak latency and decay time with peak response height in a single TEad cell. (Top) Spike density functions of three responses with different peak heights (maximum, 60% and 40% of maximum). Solid arrows indicate the onset of responses, solid double arrows the response peaks and open arrows the decay from the peak to the spontaneous level. (Lower three) Scatter plots of onset latency, peak latency and decay time against peak height of individual responses. The linear regression lines are superimposed.

Figure 12.

Distributions of correlation coefficients between onset latency and peak response height in individual cells (top), between peak latency and peak height (middle) and between decay time and peak height (bottom). Filled regions indicate cells with significant (P < 0.05) correlation.

Figure 12.

Distributions of correlation coefficients between onset latency and peak response height in individual cells (top), between peak latency and peak height (middle) and between decay time and peak height (bottom). Filled regions indicate cells with significant (P < 0.05) correlation.

Figure 13.

Overall effectiveness of the stimuli in activating cells in TEav (thick line) and TEad (broken line) with an index for colorfulness of stimuli (Colorfulness, thin line). The stimuli are aligned along the horizontal axis in the same order as in Figure 2. The overall effectiveness is defined for each stimulus as the ratio of the number of cells in which the stimulus evoked significant (P < 0.01) responses to the total number of visually responsive cells. See text for the definition of Colorfulness.

Overall effectiveness of the stimuli in activating cells in TEav (thick line) and TEad (broken line) with an index for colorfulness of stimuli (Colorfulness, thin line). The stimuli are aligned along the horizontal axis in the same order as in Figure 2. The overall effectiveness is defined for each stimulus as the ratio of the number of cells in which the stimulus evoked significant (P < 0.01) responses to the total number of visually responsive cells. See text for the definition of Colorfulness.

Figure 14.

Averaged stimulus preference of TEav cells. The stimuli were aligned in order of overall effectiveness in TEav from left to right and then from top to bottom in the upper part and along the horizontal axis in the bottom graph. The overall effectiveness (gray line) and Colorfulness (black line) of individual stimuli are plotted on the vertical axis at the bottom. See text for the definition of Colorfulness.

Figure 14.

Averaged stimulus preference of TEav cells. The stimuli were aligned in order of overall effectiveness in TEav from left to right and then from top to bottom in the upper part and along the horizontal axis in the bottom graph. The overall effectiveness (gray line) and Colorfulness (black line) of individual stimuli are plotted on the vertical axis at the bottom. See text for the definition of Colorfulness.

Figure 15.

Comparison of stimulus selectivity between TEav and TEad cells. The graphs at the top represent the original (left) and normalized magnitudes of responses (right) averaged for responses of the same rank order in individual cells for TEav cells (thick line) and TEad cells (thin line). The histograms in the middle and bottom show the distributions of tuning width defined at 25% of the maximal response (TW25, left), the number of effective stimuli that evoked significant responses in individual cells (TWP<0.01, center) and the amount of information transmitted (right) for cells recorded from TEav (middle) and TEad (bottom).

Figure 15.

Comparison of stimulus selectivity between TEav and TEad cells. The graphs at the top represent the original (left) and normalized magnitudes of responses (right) averaged for responses of the same rank order in individual cells for TEav cells (thick line) and TEad cells (thin line). The histograms in the middle and bottom show the distributions of tuning width defined at 25% of the maximal response (TW25, left), the number of effective stimuli that evoked significant responses in individual cells (TWP<0.01, center) and the amount of information transmitted (right) for cells recorded from TEav (middle) and TEad (bottom).

Figure 16.

The distributions of minimum onset and peak latencies in individual cells (upper) and the distributions of onset and peak latencies of all significant excitatory responses (bottom) in TEav and TEad.

Figure 16.

The distributions of minimum onset and peak latencies in individual cells (upper) and the distributions of onset and peak latencies of all significant excitatory responses (bottom) in TEav and TEad.

References

Allison T, Puce A, McCarthy G (
2000
) Social perception from visual cues: role of the STS region.
Trends Cogn Sci

4
:
267
–278.
Baylis GC, Rolls ET, Leonard CM (
1987
) Functional sub-divisions of the temporal lobe neocortex.
J Neurosci

7
:
330
–342.
Brodmann K (1909) Vergleichende lokalisationslehre der grosshirnrinde: in ihren Prinzipien dargestellt anf Grund des Zellenbaus. Leipzig: Verlag von Johann Ambrosius Barth.
Blum JS, Chow KL, Pribram KH (
1950
) A behavioral analysis of the organization of the parieto-temporo-preoccipital cortex.
J Comp Neurol

93
:
53
–100.
Buckley MJ, Gaffan D, Murray EA (
1997
) Functional double dissociation between two inferior temporal cortical areas: perirhinal cortex versus middle temporal gyrus.
J Neurophysiol

77
:
587
–598.
Cheng K, Saleem KS, Tanaka K (
1997
) Organization of corticostriatal and corticoamygdalar projections arising from the anterior inferotemporal area TE of the macaque monkey: a Phaseolus vulgaris leuco- aggulutinin study.
J Neurosci

15
:
7902
–7925.
Chow KL (
1951
) Effects of partial extirpations of the posterior association cortex on visually mediated behavior.
Comp Psychol Monogr

20
:
187
–217.
Desimone R, Fleming J, Gross CG (
1980
) Prestriate afferents to inferior temporal cortex: an HRP study.
Brain Res

184
:
41
–55.
Desimone R, Albright TD, Gross CG, Bruce C (
1984
) Stimulus-selective properties of inferior temporal neurons in the macaque.
J Neurosci

4
:
2051
–2062.
Fabre-Thorpe M, Richard G, Thorpe SJ (
1998
) Rapid categorization of natural images by rhesus monkeys.
NeuroReport

9
:
303
–308.
Felleman DJ, Van Essen DC (
1991
) Distributed hierarchical processing in the primate cerebral cortex.
Cereb Cortex

1
:
1
–47.
Field DJ (
1994
) What is the goal of sensory coding?
Neural Comput

6
:
559
–601.
Foldiak P, Young MP (1995) Sparse coding in the primate cortex. In: The handbook of brain theory and neural networks (Arbib MA, ed.), pp. 895–898. Cambridge, MA: MIT Press.
Fujita I, Fujita T (
1996
) Intrinsic connections in the macaque inferior temporal cortex.
J Comp Neurol

368
:
467
–486.
Fujita I, Tanaka K, Ito M, Cheng K (
1992
) Columns for visual features of objects in monkey inferotemporal cortex.
Nature

360
:
343
–346.
Gemba H, Sasaki K (
1989
) Potential related to no-go reaction of go/no-go hand movement task with color discrimination in human.
Neurosci Lett

101
:
263
–268.
Gershon ED, Matthew CW, Latham PE, Richmond BJ (
1998
) Coding strategies in monkey V1 and inferior temporal cortices.
J Neurophysiol

79
:
1135
–1144.
Gochin PM, Colombo M, Dorfman A, Gerstein GL, Gross CG (
1994
) Neural ensemble coding in inferior temporal cortex.
J Neurophysiol

71
:
2325
–2337.
Gross CG (
1994
) How inferior temporal cortex became a visual area.
Cereb Cortex

5
:
455
–469.
Gross CG, Rocha-Miranda CE, Bender DB (
1972
) Visual properties of neurons in inferotemporal cortex of the macaque.
J Neurophysiol

35
:
96
–111.
Hoppfield JJ (
1995
) Pattern recognition computation using action potential timing for stimulus representation.
Nature

376
:
33
–36.
Horel JA (
1994
) Local and global perception examined by reversible suppression of temporal cortex with cold.
Behav Brain Res

65
:
157
–164.
Horel JA (
1994
) Retrieval of color and form during suppression of temporal cortex with cold.
Behav Brain Res

65
:
165
–172.
Ito M, Fujita I, Tamura H, Tanaka K (
1994
) Processing of contrast polarity of visual images in inferotemporal cortex of the macaque monkey.
Cereb Cortex

5
:
499
–508.
Ito M, Tamura H, Fujita I, Tanaka K (
1995
) Size and position invariance of neuronal responses in monkey inferotemporal cortex.
J Neurophysiol

73
:
218
–226.
Iwai E, Yukie M (
1987
) Amygdalofugal and amygdalopetal connections with modality-specific visual cortical areas in macaques (Macaca fuscata, M. mulatta, and M. fascicularis).
J Comp Neurol

261
:
362
–387.
Iwai E, Yukie M (
1988
) A direct projection from hippocampal field CA1 to ventral area TE of inferotemporal cortex in the monkey.
Brain Res

444
:
397
–401.
Iwai E, Yukie M, Suyama H, Shirakawa S (
1987
) Amygdalar connections with middle and inferior temporal gyri of the monkey.
Neurosci Lett

83
:
25
–29.
Janssen P, Vogels R, Orban GA (
2000
) Selectivity for 3D shape that reveals distinct areas within macaque inferior temporal cortex.
Science

288
:
2054
–2056,
Kobatake E, Tanaka K (
1994
) Neuronal selectivities to complex object features in the ventral visual pathway of the macaque cerebral cortex.
J Neurophysiol

71
:
856
–867.
Komatsu H, Ideura Y, Kaji S, Yamane S (
1992
) Color selectivity of neurons in the inferior temporal cortex of the awake macaque monkey.
J Neurosci

12
:
408
–424.
Korner E, Gewaltig M-O, Korner U, Richter A, Rodemann T (
1999
) A model of computation in neocortical architecture.
Neural Networks

12
:
989
–1005.
Liu Z, Richmond BJ (
2000
) Response differences in monkey TE and perirhinal cortex: stimulus association related to reward schedules.
J Neurophysiol

83
:
1677
–1692.
Martin-Elkins CL, Horel JA (
1992
) Cortical afferents to behaviorally defined regions of the inferior temporal and parahippocampal gyri as demonstrated by WGA-HRP.
J Comp Neurol

321
:
177
–192.
Miller EK, Lin L, Desimone R (
1991
) A neural mechanism for working and recognition memory in inferior temporal cortex.
Science

254
:
1377
–1379.
Mishkin M (
1954
) Visual discrimination performance following partial ablations of the temporal lobe: II. Ventral surface vs. hippocampus.
J Comp Physiol Psychol

47
:
187
–193.
Mishkin M, Pribram KH (
1954
) Visual discrimination performance following partial ablations of the temporal lobe: I. Ventral vs. lateral.
J Comp Physiol Psychol

47
:
14
–20.
Miyashita Y (
1988
) Neuronal correlate of visual associative long-term memory in the primate temporal cortex.
Nature

335
:
817
–820.
Morel A, Bullier J (
1990
) Anatomical segregation of two cortical visual pathways in the macaque monkey.
Vis Neurosci

4
:
555
–578.
Optican LM, Richmond BJ (
1987
) Temporal encoding of two-dimensional patterns by single units in primate inferior temporal cortex. III. Information theoretic analysis.
J Neurophysiol

57
:
162
–178.
Oram MW, Perrett DI (
1992
) Time course of neural responses discriminating different views of the face and head.
J Neurophysiol

68
:
70
–84.
Parzen E (
1962
) On estimation of a probability density function and mode.
Ann Math Statist

33
:
1065
–1076.
Richmond BJ, Optican LM, Podell M, Spitzer H (
1987
) Temporal encoding of two-dimensional patterns by single units in primate inferior temporal cortex. I. Response characteristics.
J Neurophysiol

57
:
132
–146.
Saleem KS, Hashikawa T (
1998
) Connections of anterior inferotemporal area TE and perirhinal cortex with the hippocampal formation in the macaque monkey.
Soc Neurosci Abstr

24
:
898
.
Saleem KS, Tanaka K (
1996
) Divergent projections from the anterior inferotemporal area TE to the perirhinal and entorhinal cortices in the macaque monkey.
J Neurosci

16
:
4757
–4775.
Saleem KS, Tanaka K, Rockland KS (
1993
) Specific and columnar projection from area TEO to TE in the macaque inferior temporal cortex.
Cereb Cortex

3
:
454
–464.
Saleem KS, Suzuki W, Tanaka K, Hashikawa T (
2000
) Connections between anterior inferotemporal cortex and superior temporal sulcus region in the macaque monkey.
J Neurosci

20
:
5083
–5101.
Sasaki K, Gemba H, Nambu A, Matsuzaki R (
1993
) No-go activity in the frontal association cortex of human subjects.
Neurosci Res

18
:
249
–252.
Sato T, Kawamura T (
1990
) Distribution of cells in area TE selectively responsive to visual patterns or color spots during matching two visual stimuli at different locations.
Jpn J Physiol

40
:
S238
.
Seltzer B, Pandya DN (
1978
) Afferent cortical connections and architectonics of the superior temporal sulcus and surrounding cortex in the rhesus monkey.
Brain Res

149
:
1
–24.
Sheinberg DL, Logothetis NK (
1997
) The role of temporal cortical areas in perceptual organization.

94
:
3408
–3413
Shiwa T (
1987
) Corticocortical projections to the monkey temporal lobe with particular reference to the visual processing pathways.
Arch Ital Biol

125
:
139
–154.
Sugase Y, Yamane S, Ueno S, Kawano K (
1999
) Global and fine information coded by single neurons in the inferotemporal visual cortex.
Nature

400
:
869
–873.
Tamura H, Tanaka K. (
1996
) Comparison of visual responses of cells in ventral and dorsal parts of inferotemporal cortex.
Soc Neurosci Abstr

22
:
1615
.
Tamura H, Tanaka K (
1997
) Recurrent circuits contribute to the formation of stimulus selectivity in the inferotemporal cortex.
Soc Neurosci Abstr

23
:
2063
.
Tanaka K, Saito H-A, Fukada Y, Moriya M (
1991
) Coding visual images of objects in the inferotemporal cortex of the macaque monkey.
J Neurophysiol

66
:
170
–189.
Thorpe S, Fize D, Marlot C (
1996
) Speed of processing in the human visual system.
Nature

381
:
520
–522.
Tovee MJ, Rolls ET, Treves A, Bellis RP (
1993
) Information encoding and the responses of single neurons in the primate temporal visual cortex.
J Neurophysiol

70
:
640
–654.
Yukie M, Iwai E (
1988
) Direct projections from the ventral TE area of the inferotemporal cortex to hippocampal field CA1 in the monkey.
Neurosci Lett

88
:
6
–10.
Yukie M, Takeuchi T, Hasegawa Y, Iwai E (1990) Differential connectivity of inferotemporal area TE with the amygdala and the hippocampus in the monkey. In: Vision memory and the temporal lobe (Iwai E, Mishkin M, eds), pp. 129–135. New York: Elsevier.
Yukie M, Hikosaka K, Iwai E (
1992
) Organization of cortical visual projections to the dorsal and ventral parts of area TE of the inferotemporal cortex in macaques.
Soc Neurosci Abstr

18
:
294
.
Webster MJ, Ungerleider LG, Bachevalier J (
1991
) Connections of inferior temporal areas TE and TEO with medial temporal-lobe structures in infant and adult monkeys.
J Neurosci

11
:
1095
–1116.