| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1Research Group for Brain and Cognitive Sciences, School of Medicine, Shaheed Beheshti University, Tehran, Iran; 2School of Cognitive Sciences, Institute for Studies in Theoretical Physics and Mathematics, Niavaran, Tehran, Iran; and 3Cognitive Brain Mapping Laboratory, RIKEN Brain Science Institute, Saitama, Japan
Submitted 21 May 2004; accepted in final form 10 April 2005
 |
ABSTRACT |
|---|
|
|
|
INTRODUCTION |
|---|
|
). One of the most prominent modulations in the time course of responses exists in their onset latencies (Gawne 2000; Gawne et al. 1996; Oram et al. 2002; Reich et al. 2001; Tamura and Tanaka 2001; Tovee et al. 1993). Latencies of responses in cells of the primary visual cortex (V1) and inferior temporal cortex depend on the stimulus contrast, while the orientation of bar stimuli only changes the magnitudes of responses (Gawne 2000; Gawne et al. 1996; Oram et al. 2002; Reich et al. 2001). However, aside from stimulus contrast, it is not known what kinds of features in stimulus images determine the onset latency of responses. Cells in the inferior temporal cortex are selectively activated by complex visual stimuli (Desimone et al. 1984; Gross 1973; Logothetis and Sheinberg 1996; Perrett et al. 1982; Tanaka 1996). Differences in the onset latency of responses may be used to code complex object features in the inferior temporal cortex. Eifuku et al. (2004) found changes in onset latency of responses to faces with changes in the viewing angle, whereas consistent onset latencies were reported for responses to different face views in another study (Oram and Perrett 1992). To examine the possibility that response onset latencies carry information about complex object images, we examined responses of neurons in the inferior temporal cortex to >1,000 object images in alert fixating monkeys. |
|
METHODS |
|---|
|
Recordings and stimuli
In an aseptic surgery, a recording chamber was stereotaxically placed on the dorsal surface of the skull, on the left side for one monkey and on the right side for the other monkey. After recovery, extracellular single-cell recordings were made from the anterior inferior temporal cortex with tungsten electrodes (FHC) while the monkey performed a fixation task. The electrode was advanced with an Evart-type manipulator (Narishige, Japan) from the dorsal surface of the brain through a stainless steel guide tube inserted into the brain down to 10–15 mm above the recording sites. Recordings were made on an evenly spaced grid, regardless of the stimulus selectivity of cells, with 1-mm intervals between penetrations over a wide region, which spanned from the medial lip of the anterior middle temporal sulcus to the fundus of the superior temporal sulcus, and from anterior 15–20 in one monkey and anterior 13–20 in the other monkey. The recording positions were determined stereotaxically referring to the magnetic resonance images acquired before the surgery, and the gray and white matter transitions determined during electrode advancement (Tamura and Tanaka 2001). The action potentials from a single neuron were isolated in real time by a template matching algorithm (Worgotter et al. 1986). A conventional amplitude-based window discriminator was used in parallel to avoid contamination of activities from other neurons.
The monkey had to fixate the eyes, with a precision of ±2°, on a 0.5° circular fixation spot presented at the center of the display. The eye position was measured by an infra-red system (i_rec, http://staff.aist.go.jp/k.matsuda/eye/), which allowed a precision of 1° in measurement of eye position. Limiting the analysis to trials in which the eye position stayed within ±1° from the center of the fixation spot did not change the results. The presentation of the stimulus sequence started when the monkey had maintained eye fixation for 300 ms. Drops of juice were provided to the monkey every 1.5–2 s during the fixation. The stimulus presentation was terminated after a continuous fixation for 6.7 s or when the monkey broke eye fixation.
Stimuli, which were mainly full-color photographs of objects, were presented in sequence with a 105-ms presentation time per image without an interstimulus interval in pseudo-random order. Neurons in inferior temporal cortex preserve their stimulus selectivity in such rapid serial presentations, even when presentation times are as short as 14–28 ms (Keysers et al. 2001). In some cells (n = 27), stimuli werepresented in both an ordinary way (245-ms presentation time with a 245-ms blank period between successive stimuli) and by the rapid-serial-presentation method without blank periods. Nine of the cells showed selective responses to faces, and they showed comparable latency differences between responses to human and animal faces with the two presentation methods [10.1 ± 10.0 vs. 7.3 ± 8.5 (SD) ms with and without blank periods, respectively; P = 0.4, paired Wilcoxon]. Stimuli were presented on a gray background of 15 cd/m2 with their centers located at the center of the display.
To determine the luminosity contrast and total luminance of stimuli, we first determined the gamma functions (Brainard et al. 2002) of the three color channels of the monitor (red, green, and blue) by nonlinearly fitting the intensity of light photometrically measured at 16 evenly spaced values covering the whole range of intensity. The validity of the gamma functions and the assumption of channel independence were then confirmed by an extensive photometrical measurement for each of the three channels and their combinations. The luminance of each pixel of a stimulus was then calculated by introducing red, green, and blue values of the pixel into the respective gamma functions and summing the three output values. The average luminance of the stimulus was determined by averaging the pixel luminance over the stimulus region. The luminance contrast was defined by the Michaelson formula of (L1 – L2)/(L1 + L2). For the figure/ground contrast, L1 and L2 were the average luminance of the stimulus and the luminance of background, respectively. For the within-figure contrast, L1 and L2 were the maximum and minimum pixel luminance, respectively, within the stimulus. In the stimulus set, human and animal faces had statistically comparable figure/ground and within-figure contrasts (P > 0.3). The total luminance (total energy) of each stimulus was also calculated by integrating the pixel luminance over the stimulus. The total luminance of the human faces was significantly larger than that of the animal faces (P = 0.02), but this difference in total luminance could not explain a main part of the latency difference (Fig. 4B) as will be described in RESULTS. We didn’t examine the power spectrum of spatial frequency components of stimuli because there is no simple way to test statistical differences between two groups of stimuli.
|
The stimulus set was introduced to the monkeys a week before the recording started. All non-primate animal faces were novel to the monkeys at the time of introduction. All human faces but four were also novel to the monkeys. The four familiar face images were taken from the experimenters and care persons. There were no significant differences in response latency between the familiar and other faces (the mean onset latency averaged across the 136 face cells was 104 ± 23.6 ms for the familiar faces and 103.3 ± 13.0 ms for the other faces; P = 0.33, paired Wilcoxon). Images of the rhesus monkey faces were taken from the monkeys being kept in the same room as the experimental monkeys. Neuronal activities were recorded on 120 days in the first monkey and on 71 days in the second monkey. The difference in response latency between human and animal faces did not change during the course of recordings [Spearman correlation coefficient was 0.13 (P = 0.23) in the 1st monkey and 0.15 (P = 0.32) in the 2nd monkey].
Data analyses
The spontaneous activity was measured in the 200-ms period immediately before the sequential stimulus presentation was initiated. Presentations in which fixation breaks occurred within 300 ms after stimulus onset were excluded. The last two presentations in completed sequence presentations were also excluded. To exclude the contamination of lasting excitation from the previous presentation, we excluded presentations in which neural activity in the first 50 ms immediately after the stimulus onset was >2 SD above the mean of the spontaneous activity. The SD of spontaneous activity was calculated for trial-by-trial variations of spontaneous firing. This last condition excluded on average 15% of stimulus presentations. It has been shown that the backward masking has minimal effect on the initial part of neuronal responses with stimulus onset intervals >60 or 80 ms (Kovács et al. 1995; Rolls and Tovee 1994).
The significance of responses to individual stimuli was determined by comparing the magnitudes of responses, measured by the averaged firing rate within a 140-ms window starting at 71 ms after the stimulus onset, with those of the spontaneous firing rate (P < 0.05, Kolmogorov-Smirnov test). This relatively large window covered the variation in response latency and durations across stimuli and across cells. The latency-adjusted magnitude of responses was calculated by the averaged firing rate within a 100-ms window starting at the onset of response determined as described in the following text. The onset latency of responses was determined for responses of individual cells to individual stimuli. The exact size of the window was not critical for the main results in this paper. Changes in the window size within a range from 60 to 140 ms did not change the main results.
Cells that showed a significant tendency to respond to a larger proportion of face stimuli than non-face stimuli were selected for further analyses. This tendency was quantified by the odds ratio (OR) (Agresti 1996; Somes and O’Brien 1988). The ratio of the number of effective stimulus members that evoked significant responses (P < 0.05, Kolmogorov-Smirnov test) to that of ineffective members (odds) was calculated for faces and for non-face objects, and the OR was defined by the ratio of the odds for faces to those for non-face objects. Thus OR is a ratio between ratios. An OR significantly larger than 1 means that the odds of effective stimuli were significantly larger in faces as compared with non-face objects. The significance level was set at P = 0.01.
The response onset was determined from a peristimulus time histogram (PSTH), which was constructed for each stimulus with 1-ms bins and smoothed with a Gaussian kernel (SD, 3 ms). The first 1-ms bin of 20 consecutive bins that was 2.58 SD. above the mean of the spontaneous activity defined the response onset. The SD of the spontaneous activity, used here, was calculated across bins in a PSTH made for spontaneous activity in the 200-ms period and smoothed by the same Gaussian kernel.
To separate the dependence of response latency on face group (human vs. animal faces) from the dependence on more primitive features of stimulus images, we used a linear model. The model is expressed by a regression equation: dev(S) = w0 + w1 I(S) + w2 G(S) +g, where dev(S) is the deviation of stimulus response latency from the mean face response latency of the cell, I(S) is the stimulus luminance or contrast depending on the hypothesis, and G(S) is the face group (0 for human faces and 1 for animal faces). wi are the coefficients, which were determined by the least square fitting. The noise term g is a Gaussian distribution with a mean of 0. To test the significance of dependence on the stimulus group, we evaluated the null hypothesis H0:w2 = 0 (F test) (Draper and Smith 1998). The value of w2 in this model also gives us an estimation of latency difference between human and animal faces independent from effects of differences in the contrast or total luminance between human and animal faces.
The reliability of categorical discrimination of stimuli based on the firing of a neuron can be quantified by Receiver Operating Characteristics (ROC) analysis (Hilgers 1991; Metz 1986; Thompson and Zucchini 1989). To examine the time course of the reliability, we applied ROC analysis to activities of individual cells within a 25-ms window and slid the window in 1-ms steps. For the discrimination between human and animal faces, responses to all individual presentations of all human and animal faces were included in the analysis. Here the horizontal axis of ROC plots the proportion of responses larger than threshold within all the responses to an animal face presentation, and the vertical axis plots the proportion within all the responses to a human face presentation. The threshold was swept over the whole range of response magnitudes to draw an ROC curve, and the area below the curve gave an estimate of the discrimination reliability (ROC value). For discrimination between faces and non-face objects, responses to all individual presentations of all faces and non-face objects were included. The ROC value depends on the size of the time window in which spikes are counted. The window should be big enough to reduce intertrial variance, while not so big so as to smooth away the differences in the time courses of responses to the two groups of stimuli. We selected 25 ms because we found it a good compromise between these two factors in most cells. Values in the paper show mean ± SD, unless otherwise mentioned.
|
|
RESULTS |
|---|
|
|
These 136 cells tended to respond with shorter onset latencies to human faces than to animal faces. Figures 2 and 3 show the results from one example cell and the whole population, respectively. The cell exemplified in Fig. 2 showed significant responses to 94% of human faces and 83% of animal faces. The magnitudes of the responses to human and animal faces largely overlapped with each other (Fig. 2B, right; P = 0.1, Wilcoxon), whereas the onset latencies of the responses showed a clear difference (Fig. 2B, left; P < 10–7, Wilcoxon). The difference between the mean latencies was 24.8 ms.
|
|
Figure 3B shows a corresponding plot for the means of response magnitudes. The points are distributed along the diagonal line, although the paired Wilcoxon test showed slightly but significantly larger responses to human versus animal faces in the cell population (P < 0.01; differences were 1.5 ± 4.5 spikes/s with median of 0.8 spikes/s). The onset latency of individual cells’ responses tended to vary more among animal faces than among human faces (P < 10–8, paired Wilcoxon; Fig. 3C). There were no such differences between variances among human and animal faces for the magnitudes of individual cells’ responses (P = 0.9, paired Wilcoxon; Fig. 3D).
A slight negative statistical correlation was previously observed between the magnitudes and onset latencies of responses in inferior temporal cells (Tamura and Tanaka 2001). The differences in latency means and those in magnitude means between responses to human and animal faces were also negatively correlated (Pearson r, –0.2, P < 0.05). To test the possibility that the differences in onset latencies (Fig. 3A) originated in the differences in magnitudes of responses (Fig. 3B), we performed two analyses. First, we excluded 40 cells in which human faces evoked significantly larger responses than animal faces (P < 0.05, 1-tailed t-test) and compared the mean onset latencies between responses to human and animal faces for the remaining 96 cells. There was a significant difference in this cell population (P < 10–12, paired Wilcoxon; differences were 13.5 ± 12.9 ms with median of 13.3 ms). Moreover, 52 cells that showed numerically larger mean responses to animal faces than to human faces (cells plotted below the diagonal line in Fig. 3B) showed significantly shorter onset latencies to human faces than to animal faces (P < 10–7, differences were 11.6 ± 11.1 ms with median of 10.0 ms). The latency differences in the 83 cells that showed statistically comparable magnitudes of responses to human and animal faces were 14.4 ± 13.3 ms, with median of 13.8 ms, which may be a good estimate for the genuine latency difference free from the secondary effects originating in differences in response magnitude.
Second, we used only the responses having overlapping magnitudes between human and animal faces in individual cells to calculate mean latencies, and we compared the mean latencies between human and animal faces in the 136 cells. By removing the nonoverlapping responses, the statistical difference between the magnitudes of mean responses to human faces and those of mean responses to animal faces disappeared (P > 0.05, paired Wilcoxon). Nonetheless, means of response onset latencies for human faces were significantly shorter than means for animal faces (P < 10–17; differences, 15.1 ± 13.9 ms with median of 15.7 ms). These results showed that the presence of shorter latencies for responses to human faces compared with animal faces was largely independent from the difference in the magnitudes of responses.
One may suspect that the systematic difference in mean onset latencies for human and animal faces was due to greater heterogeneity of animal faces. Faces of some animals evoked early-onset responses, whereas others evoked late responses. The larger variance in onset latencies of individual cells among animal faces compared with that among human faces (Fig. 3C) may support this notion. However, even when we focused on the 53 cells (39% of the 136 cells) that showed no difference in variance of onset latencies between human and animal faces (P > 0.3, variance ratio test), the mean onset latencies for human faces were significantly shorter than those for animal faces (P < 10–7, paired Wilcoxon, differences were 13.0 ± 12.0 ms). Thus the difference in variance could not be the only cause of the systematic difference in mean onset latencies between human and animal faces.
The luminance contrast of face stimuli could not have contributed to the latency difference, because the figure/ground and within-figure contrasts were comparable across the human and animal face stimuli (P > 0.3, t-test). Nevertheless, to further examine the relation between stimulus contrast and response latency, we calculated the mean relative latency, averaged across cells, for each of the face stimuli. For each face stimulus, the onset latency of a significant response was subtracted by the mean latency of the cell to face stimuli and then averaged across cells that showed significant responses to the stimulus. Figure 4A plots this averaged relative latency against the figure/ground contrast of the stimulus. Each dot represents a stimulus. It is obvious that human faces (black circles) were largely separated from animal faces (gray triangles) in the vertical direction at any contrast range. The regression analysis (see METHODS) revealed that the difference in face group (human vs. animal faces) resulted in a 14.4-ms latency difference (significantly larger than 0, P < 10–8), whereas there was no significant effect of stimulus contrast (P = 0.1). When the within-figure contrast was used as a factor, instead of the figure-ground contrast, very similar results were obtained. The difference in face group manifested a 15.1-ms (significantly larger than 0, P < 10–8) difference in mean latency, whereas there was no significant effect of the within-figure contrast (P = 0.4).
The total luminance (total energy) of the human faces was significantly greater than that of the animal faces (P = 0.02). However, as can be seen in Fig. 4B, the averaged relative latencies of responses to human faces were larger than those of responses to animal faces even when the comparison was made between stimuli with comparable total luminance. The regression analysis showed a significant effect of face group, which accounted for a 14.2-ms latency difference (significantly larger than 0, P < 10–8). Thus although the total luminance had a marginally significant negative effect on response latency (P = 0.05), the majority of the latency difference between responses to human and animal faces was independent from the effect of total luminance.
When the regression analysis was applied to the latency of responses in individual cells, 43% (59 cells) of the 136 face cells showed significant (P < 0.05) effects of face group on onset latency, whereas only 7% (9 cells) and 6% (8 cells) showed significant effects of the figure-ground and within-figure contrast, respectively. Only 8% (11 cells) showed significant effects of the total luminance. These latter numbers of cells deviated from the number expected by chance from the P value (6.8) with no (9, P = 0.15; 8, P = 0.24) or marginal significance (11, P = 0.05). These results show that onset latency was much more dependent on the face group than on the contrast or total luminance for single cells. However, we might have observed larger effects of contrast and total luminance if the stimuli covered wider ranges of contrast and total luminance (Oram et al. 2002).
Because there were more profile faces in the animal face group (
26/43) than in the human face group (6/49), we calculated mean latencies of responses to human profile faces for each cell. Means of onset latencies of responses to human profile faces were significantly (P < 10–8, paired Wilcoxon) shorter than the means to animal faces (differences were 10.0 ± 18.4 ms). Thus it is not likely that the differences in onset latencies between human and animal faces were due to the difference in the proportion of profile faces. The size of the stimuli, measured by the largest dimension of the head, was very similar between human and animal faces (139 ± 25 vs. 136 ± 30 pixels, respectively, P = 0.8). Moreover, there was no significant correlation between means of onset latencies for individual human faces averaged across cells and the size of the human faces quantified by the distance between the upper lip and the nasal root (the midpoint between the 2 eyes; P = 0.7, Spearman) or between averaged latencies for individual animal faces and the size of the animal faces determined by the largest dimension (P = 0.12, Spearman).
ROC analyses
Onset latencies of responses to effective human faces were significantly shorter (P < 0.01, 1-tailed t-test) than those to effective animal faces in 38% of individual cells (52/136), whereas no cell showed significantly shorter latencies to animal faces. A similar trend was found for most of the remaining cells: 79% (66/84) of cells without significant statistical difference had mean human face latencies shorter than mean animal face latencies. To examine how reliably the instantaneous firing of a single cell could discriminate human faces from animal faces, we applied an ROC analysis (see METHODS) to the activity of cells within a 25-ms window and moved the window in 1-ms steps. This analysis compared activities in individual presentations of all the stimulus members belonging to the two groups and provides information about how reliably an ideal observer, who is only observing the instantaneous firing rate in one trial but knowing distributions of firing rates in all presentations, can determine whether the presented stimulus belongs to one group or the other. ROC values 0.5 indicate chance performance, and values of 1 and 0 indicate error-free classification (1 indicates human faces and 0 animal faces here). Most of the cells (110/136 cells) showed peaks with ROC values significantly (P < 0.05) larger than 0.5 for human versus animal faces (Fig. 5A). The mean time when the ROC value first exceeded the significance level was 98.7 ± 28.9 ms after stimulus onset. This value corresponds well to the mean onset latencies to human faces, suggesting that the initial spikes to human face presentations were already sufficiently informative for the categorical discrimination of human faces from animal faces. The mean time of the peak in the ROC time courses was 118.4 ± 26.1 ms after stimulus onset. Two-thirds of the cells (84/136 cells) showed troughs significantly smaller than 0.5, and most of them (70 cells) showed both significant peaks and significant troughs. For all 70 cells, the troughs followed the peaks. The mean time of the bottoms in troughs was 168.8 ± 26.6 ms after stimulus onset. The biphasic shapes can also be seen in the averaged ROC time courses averaged over the 136 cells (Fig. 5B, pink). Because the stimulus presentation period was fixed in the present study, we could not determine whether a longer stimulus presentation would result in a change in the time of the trough in human versus animal face ROC time courses. However, responses previously obtained with longer stimulus presentations usually consisted of initial peaks followed by plateaus (though not entirely flat) of reduced magnitudes (Sugase et al. 1999; Tamura and Tanaka 2001). Therefore it is likely that the relative magnitudes of firing rates to human and animal faces reverse, which results in a change of the ROC value from a peak to a trough (crossing the chance ROC value of 0.5), in the initial parts of responses regardless of the stimulus presentation period. The ROC time courses calculated for faces as a whole versus non-face objects (Fig. 5, A, left, and B, green) showed slower time courses with a peak at the time roughly corresponding to the middle between the peak and the bottom of the human versus animal face ROC curve.
|
|
In later recordings, we added 16 macaque monkey faces to the image set. In the 63 cells that responded to a significantly larger proportion of faces (including human, monkey, and non-primate animal faces) than non-face objects, means of onset latencies of responses to the monkey faces were significantly shorter than the means for non-primate animal faces (P = 0.01, paired Wilcoxon) but not different from the means for human faces (P > 0.3).
Latency of responses to faces in non-face cells and that of responses to non-face stimuli in face cells
Distinction of inferior temporal neurons into face cells and non-face cells was a matter of degree rather than categorical distinction. Face cells responded to some non-face stimuli and non-face cells were driven by some faces. Also, the OR for the discrimination of faces versus non-face objects was continuously distributed over cells with a unimodal distribution. The criterion used in this study (OR significantly larger than 1) provides a lenient selection of the cells with relative preference to faces. To examine whether the latency difference between responses to human and animal faces observed in these selected cells was common to occasional responses to human and animal faces in non-face cells, we accumulated the relative onset latency of significant responses to individual human and animal faces for the two groups of cells. Because non-face cells usually showed significant responses only to a few human faces and a few animal faces, a paired comparison between the two mean latencies of individual cells could not be applied to this group of cells. The relative latency was obtained by subtracting the mean onset latency of the cell averaged over all the effective stimuli (face and non-face objects).
Figure 7 shows the distribution of relative onset latency for the 136 face cells (A and B) and 365 non-face cells (C and D). In the group of face cells, distributions of relative latencies for human and animal faces had mean values of –7.5 and +7.5 ms, respectively, which are consistent with the main results. Although the mean relative latency in the non-face cell group was still negative (–3.8 ms) for human faces and positive (+2.8 ms) for animal faces, the difference between the two means was smaller. In addition to the attenuation of mean latency difference, the variances of relative latency were significantly larger in the non-face cell group than those in the face cell group (P < 10–8, variance ratio test). Therefore while the latency difference between face classes was distributed over cells in area TE, face cells were more efficient in discriminating human from animal faces based on onset latency of responses.
|
|
|
DISCUSSION |
|---|
|
We here refer to only human and macaque faces by "primate faces" because faces of other primates were not included in the present study. A difference of a similar magnitude has previously been found in peak latencies of evoked potentials in human subjects. Human faces evoked potentials 10–14 ms earlier than animal faces (Carmel and Bentin 2002; McCarthy et al. 1999). Slightly smaller but significant difference (6 ms) was also found for faces in scenes (Rousselet et al. 2004).
The latency difference between face groups could not be caused by differences in the exact eye position on the stimuli. Due to the short presentation time (105 ms) and the lack of blank intervals in the present experiments, any saccade triggered by a presented stimulus had to land on the subsequent stimulus. Because the order of stimulus presentation was randomized, such saccades, even if present, could not result in systematic differences in the exact position of eye fixation between stimuli.
Our finding that human faces evoked responses as rapidly as did monkey faces in the inferior temporal cortex of monkeys may seem inconsistent with the known species advantage. Humans can discriminate the identity of human faces better than that of monkey faces, and the reverse is true for monkeys (Pascalis and Bachevalier 1998). We should note here that the neuronal activity was recorded in laboratory monkeys in the present study for which human faces are as salient as monkey faces. It is reasonable to expect that human faces are treated in a similar way as monkey faces in the visual system of laboratory monkeys, for which human faces are as important as monkey faces.
In model neurons, as the stimulus becomes more effective, the magnitude of response increases and its latency decreases, while the change in magnitude saturates sooner than that in latency (reviewed in Koch 1999). Actually, as the contrast of a bar stimulus increases, V1 cells decrease the onset latency of their responses after the magnitude of the response saturates (Gawne et al. 1996; Reich et all. 2001). Such dissociation between changes in magnitude and latency may also occur as stimuli become more effective in other aspects, although it cannot be a general principle because changes occur only in the magnitude and not latency in V1 cells when the orientation of bars changes (Gawne et al. 1996). It is possible that non-primate faces, although weaker stimuli than primate faces for monkey-face cells, may still have been sufficiently effective so as to fall on the saturated portion of the response magnitude curve, and thus produce almost the same firing rates as primate faces. At the same time, those non-primate faces, being weaker stimuli, produced longer latencies.
One possible way the brain might use these latency differences is to enhance or suppress a particular kind of information based on time-dependent modulation. For example, primate faces contain information useful for social interactions, whereas non-primate animal faces may have to be processed in terms of predator versus prey relationships rather than in the social domain. Given the latency differences, each of these different kinds of information may be selectively enhanced or suppressed by time-dependent modulation. Here we need to assume that there are modulatory, or gating, inputs that fire during particular time windows and that different modulatory inputs are activated in different behavioral contexts. Different kinds of information coded in different time windows may also be sorted to different downstream sites by using similar mechanisms. A time-dependent gating has been shown in the signal flow from area MT to preoculomotor circuitry, although the size of time gate was larger than that considered here (Seidemann et al. 1998). Latency differences may also be used to increase the capacity of coding as has been repeatedly discussed for differences in the time pattern of responses in general and for latency differences in particular (Eckhorn and Popel 1974; Optican and Richmond 1987; Rieke et al. 1997; Thorpe 1990). Both the enhancing/suppressing and coding require knowing the time that has passed since the stimulus onset. Neural oscillations phase-locked to stimulus onsets, discharges of stimulus-nonselective cells, and neuronal discharges related with previous saccades may be used as reference points in time (Reich et al. 2001; VanRullen and Thorpe 2002).
|
|
GRANTS |
|---|
|
|
|
ACKNOWLEDGMENTS |
|---|
|
Present address of R. Kiani: Program in Neurobiology and Behavior, University of Washington, Seattle, WA 98195-7270.
|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: H. Esteky, Research Group for Brain and Cognitive Sciences, School of Medicine, Shaheed Beheshti University, P.O. Box 19835-181, Tehran, Iran (E-mail: esteky{at}ipm.ir)
|
|
REFERENCES |
|---|
|
Brainard DH, Pelli DG, and Robson T. Display characterization. In: Encyclopedia of Imaging Science and Technology, edited by Hornak J. New York: Wiley, 2002, p. 172–188.
Carmel D and Bentin S. Domain specificity versus expertise: factors influencing distinct processing of faces. Cognition 83: 1–29, 2002.[CrossRef][Web of Science][Medline]
Desimone R, Albright TD, Gross CG, and Bruce C. Stimulus-selective properties of inferior temporal neurons in the macaque. J Neurosci 4: 2051–2062, 1984.[Abstract]
Draper NR and Smith H. Applied Regression Analysis. New York: Wiley, 1998.
Eckhorn R and Popel B. Rigorous and extended application of information theory to the afferent visual system of the cat. I. Basic concepts. Kybernetik 16: 191–200, 1974.[CrossRef][Web of Science][Medline]
Eifuku S, De Souza WC, Tamura R, Nishijo H, and Ono T. Neuronal correlates of face identification in the monkey anterior temporal cortical areas. J Neurophysiol 91: 358–371, 2004.
Gawne TJ. The simultaneous coding of orientation and contrast in the responses of V1 complex cells. Exp Brain Res 133: 293–302, 2000.[CrossRef][Web of Science][Medline]
Gawne TJ, Kjaer TW, and Richmond BJ. Latency: another potential code for feature binding in striate cortex. J Neurophysiol 76: 1356–1360, 1996.
Gross CG. Visual functions of inferotemporal cortex. In: Handbook of Sensory Physiology, edited by Autrum H, Jung R, Lowenstein W, Mckay D, and Teuber HL. Berlin: Springer, 1973, vol VII/3B, p. 451–482.
Hilgers RA. Distribution-free confidence bounds for ROC curves. Methods Inf Med 30: 96–101, 1991.[Web of Science][Medline]
Keysers C, Xiao DK, Foldiak P, and Perrett DI. The speed of sight. J Cogn Neurosci 13: 90–101, 2001.[CrossRef][Web of Science][Medline]
Koch C.Biophysics of Computation: Information Processing in Single Neurons. New York: Oxford, 1999.
Kovács G, Vogels R, and Orban GA. Cortical correlate of pattern backward masking. Proc Natl Acad Sci USA 92: 5587–5591, 1995.
Logothetis NK and Sheinberg DL. Visual object recognition. Annu Rev Neurosci 19: 577–621, 1996.[CrossRef][Web of Science][Medline]
McCarthy G, Puce A, Belger A, and Allison T. Electrophysiological studies of human face perception. II. Response properties of face-specific potentials generated in occipitotemporal cortex. Cereb Cortex 9: 431–444, 1999.
Metz CE. Statistical analysis of ROC data in evaluating the diagnostic performance. In: Multiple Regression Analysis: Applications in The Health Sciences, edited by Herbert DE and Myers RH. New York: Am. Inst. of Physics, 1986, p. 365–384.
Optican LM and Richmond BJ. Temporal encoding of two-dimensional patterns by single units in primate inferior temporal cortex. III. Information theoretic analysis. J Neurophysiol 57: 162–178, 1987.
Oram MW and Perrett DI. Time course of neural responses discriminating different view of the face and head. J Neurophysiol 68: 70–84, 1992.
Oram MW, Xiao D, Dritschel B, and Payne KR. The temporal resolution of neural codes: does response latency have a unique role? Philos Trans R Soc Lond B Biol Sci 357: 987–1001, 2002.
Pascalis O and Bachevalier J. Face recognition in primates: a cross-species study. Behav Process 43: 87–96, 1998.[CrossRef]
Perrett DT, Rolls ET, and Caan W. Visual neurons responsive to faces in the monkey temporal cortex. Exp Brain Res 47: 329–342, 1982.[Web of Science][Medline]
Reich DS, Mechler F, and Victor JD. Temporal coding of contrast in primary visual cortex: when, what, and why. J Neurophysiol 85: 1039–1050, 2001.
Richmond BJ, Optican LM, Podell M, and Spitzer H. Temporal encoding of two-dimensional patterns by single units in primate inferior temporal cortex. I. Response characteristics. J Neurophysiol 57: 132–146, 1987.
Rieke F, Warland D, de Ruyter van Steveninck RR, and Bialek W. Spikes: Exploring the Neural Code. Cambridge, MA: MIT Press, 1997.
Rolls ET and Tovee MJ. Processing speed in the cerebral cortex and the neurophysiology of visual masking. Proc R Soc Lond B Biol Sci 257: 9–15, 1994.[Medline]
Rousselet G, Macé MJ, and Fabre-Thorpe M. Animal and human faces in natural scenes: how specific to human faces is the N170 ERP component? J Vis 4: 13–21, 2004.[CrossRef][Web of Science][Medline]
Seidemann E, Zohary E, and Newsome WT. Temporal gating of neural signals during performance of a visual discrimination task. Nature 394: 72–75, 1998.[CrossRef][Medline]
Somes GW and O’Brien KF. Odds ratio estimators. In: Encyclopedia of Statistical Sciences, edited by Kotz L and Johnson NL. New York: Wiley, 1988, vol. 6, p. 407–410.
Sugase Y, Yamane S, Ueno S, and Kawano K. Global and fine information coded by single neurons in the temporal visual cortex. Nature 400: 869–873, 1999.[CrossRef][Medline]
Tamura H and Tanaka K. Visual response properties of cells in the ventral and dorsal parts of the macaque inferotemporal cortex. Cereb Cortex 11: 384–399, 2001.
Tanaka K. Inferotemporal cortex and object vision. Annu Rev Neurosci 19: 109–139, 1996.[CrossRef][Web of Science][Medline]
Thompson ML and Zucchini W. On the statistical analysis of ROC curves. Stat Med 8: 1277–1290, 1989.[Web of Science][Medline]
Thorpe SJ.Spike arrival times: a highly efficient coding scheme for neural networks. In: Parallel Processing in Neural Systems, edited by Eckmiller R, Hartman G, and Hauske G. Amsterdam: Elsevier, 1990, pp. 91–94.
Tovee MJ, Rolls ET, Treves A, and Bellis RP. Information encoding and the responses of single neurons in the primate temporal visual cortex. J Neurophysiol 70: 640–654, 1993.
VanRullen R and Thorpe SJ. Surfing a spike wave down the ventral stream. Vision Res 42: 2593–2615, 2002.[CrossRef][Web of Science][Medline]
Worgotter F, Daunicht WJ, and Eckmiller R. An on-line spike form discriminator for extracellular recordings based on an analog correlation technique. J Neurosci Methods 17: 141–151, 1986.[CrossRef][Web of Science][Medline]
This article has been cited by other articles:
|
|
 |
|
  T. Masquelier, E. Hugues, G. Deco, and S. J. Thorpe Oscillations, Phase-of-Firing Coding, and Spike Timing-Dependent Plasticity: An Efficient Learning Scheme J. Neurosci., October 28, 2009; 29(43): 13484 - 13493. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Mirpour and H. Esteky State-Dependent Effects of Stimulus Presentation Duration on the Temporal Dynamics of Neural Responses in the Inferotemporal Cortex of Macaque Monkeys J Neurophysiol, September 1, 2009; 102(3): 1790 - 1800. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. P. Sripati and C. R. Olson Representing the Forest before the Trees: A Global Advantage Effect in Monkey Inferotemporal Cortex J. Neurosci., June 17, 2009; 29(24): 7788 - 7796. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. Mormann, S. Kornblith, R. Q. Quiroga, A. Kraskov, M. Cerf, I. Fried, and C. Koch Latency and Selectivity of Single Neurons Indicate Hierarchical Processing in the Human Medial Temporal Lobe J. Neurosci., September 3, 2008; 28(36): 8865 - 8872. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. A. Orban Higher Order Visual Processing in Macaque Extrastriate Cortex Physiol Rev, January 1, 2008; 88(1): 59 - 89. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Q. Quiroga, L. Reddy, C. Koch, and I. Fried Decoding Visual Inputs From Multiple Neurons in the Human Temporal Lobe J Neurophysiol, October 1, 2007; 98(4): 1997 - 2007. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Kiani, H. Esteky, K. Mirpour, and K. Tanaka Object Category Structure in Response Patterns of Neuronal Population in Monkey Inferior Temporal Cortex J Neurophysiol, June 1, 2007; 97(6): 4296 - 4309. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. De Baene, E. Premereur, and R. Vogels Properties of Shape Tuning of Macaque Inferior Temporal Neurons Examined Using Rapid Serial Visual Presentation J Neurophysiol, April 1, 2007; 97(4): 2900 - 2916. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. H. Lemon and D. V. Smith Influence of response variability on the coding performance of central gustatory neurons. J. Neurosci., July 12, 2006; 26(28): 7433 - 7443. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
TOP
ABSTRACT
INTRODUCTION
