The Journal of Neurophysiology Vol. 79 No. 5 May 1998, pp. 2394-2407
Copyright ©1998 by the American Physiological Society

Correlation Analysis of Corticotectal Interactions in the Cat Visual System

Michael Brecht, Wolf Singer, and Andreas K. Engel

Max-Planck-Institut für Hirnforschung, 60528 Frankfurt, Germany

	ABSTRACT

Abstract Introduction Methods Results Discussion References

Brecht, Michael, Wolf Singer, and Andreas K. Engel. Correlation analysis of corticotectal interactions in the cat visual system. J. Neurophysiol. 79: 2394-2407, 1998. We have studied the temporal relationship between visual responses in various visual cortical areas [17, 18, postero medial lateral suprasylvian (PMLS), postero lateral lateral suprasylvian (PLLS), 21a]) and the superficial layers of the cat superior colliculus (SC). To this end, simultaneous recordings were performed in one or several visual cortical areas and the SC of anesthetized paralyzed cats, and visually evoked multiunit responses were subjected to correlation analysis. Significant correlations occurred in 117 (24%) of 489 cortex-SC pairs and were found for all cortical areas recorded. About half of the significant correlograms showed an oscillatory modulation. In these cases, oscillation frequencies covered a broad range, the majority being in the alpha- and beta-band. On average, significant center peaks in cross-correlograms had a modulation amplitude of 0.34. Our analysis revealed a considerable intertrial variability of correlation patterns with respect to both correlation strength and oscillation frequency. Furthermore, cortical areas differed in their corticotectal correlation patterns. The percentage of cells involved a corticotectal correlation, as well as the percentage of significantly modulated correlograms in such cases, was low for areas 17 and PMLS but high for areas 18 and PLLS. Analysis of the cortical layers involved in these interactions showed that consistent temporal relationships between cortical and collicular responses were not restricted to layer V. Our data demonstrate a close relationship between corticotectal interactions and intracortical or intracollicular synchronization. Trial-by-trial analysis from these sites revealed a clear covariance of corticotectal correlations with intracortical synchronization. The probability of observing corticotectal interactions increased with enhanced local cortical and collicular synchronization and, in particular, with interareal cortical correlations. Corticotectal correlation patterns resemble in many ways those described among areas of the visual cortex. However, the correlations observed are weaker than those between nearby cortical sites, exhibit usually broader peaks and for some cortical areas show consistent phase-shifts. Corticotectal correlations represent population phenomena that reflect both the local and global temporal organization of activity in the cortical and collicular network and do not arise from purely monosynaptic interactions. Our findings show that both striate and extrastriate inputs affect the superficial SC in a cooperative manner and, thus, do not support the view that responses in the superficial SC depend exclusively on input from the primary visual areas as implied by the concept of "two corticotectal systems." We conclude that the corticotectal projections convey temporal activation patterns with high reliability, thus allowing the SC evaluation of information encoded in the temporal relations between responses of spatially disseminated cortical neurons. As a consequence, information distributed across multiple cortical areas can affect the SC neurons in a coherent way.

	INTRODUCTION

Abstract Introduction Methods Results Discussion References

In most vertebrate species, the optic tectum constitutes the major visual processing center. However, in some mammalian lines such as carnivores and primates, an enormous expansion of neocortical processing centers has taken place. In these species, the thalamocortical system has become the major target of the retinofugal projection. Nevertheless, a variety of studies suggest that the superior colliculus (SC) has not lost its functional significance for visual processing in these species. Data from cats and monkeys indicate that the SC subserves important integrative functions, which are carried out under strong cortical influence. For example, in cats (Sprague and Meikle 1965), monkeys (Albano et al. 1982), and rodents (Schneider 1969), lesions to the SC result in a severe visual neglect, and these attentive deficits are generally much more devastating than the behavioral consequences of similar-sized lesions to any cortical area. Given these facts, the interaction between visual cortex and SC is an important issue for understanding functional integration in the visual system, and indeed numerous studies have investigated corticotectal relationships.

Anatomic studies in cats have demonstrated massive corticotectal projections from all visual cortical areas to the superficial SC laminae, with a subset of areas also projecting to intermediate and deep SC layers (Harting et al. 1992; Segal and Beckstead 1984). Although the axonal termination zones of most visual areas overlap broadly with respect to depth, there is a tendency of visual areas located further away from primary visual cortex to project to progressively deeper SC locations (Freeman and Singer 1983; Harting et al. 1992; Segal and Beckstead 1984). In all cortical areas, the projection neurons correspond to a homogeneous group of large layer V pyramidal neurons, the electrophysiological properties of which have been studied in considerable detail (Connors and Amitai 1995; Kasper et al. 1994). In striate cortex, these corticotectal projection neurons are binocular and have receptive fields of the complex or special-complex type (Palmer and Rosenquist 1974). Feedback from the SC to cortical areas is provided via the thalamus. This indirect projection originates from the superficial SC layers and is relayed to the cortex via the tectorecipient zone of the pulvinar and the C layers of the lateral geniculate nucleus(Graham and Casagrande 1980; Graybiel and Berson 1980).

Functional investigations of corticotectal interactions have been carried out in several species. Almost all of these studies have focused on how either SC or cortex contribute to the receptive-field properties of neurons in the respective other structure. Such experiments usually have examined the consequences of interfering with the activity of the different structures by electrical stimulation, lesions, cooling, or pharmacological manipulations (for review see Chalupa 1984).These studies have shown that cortical input contributes to the general responsiveness, direction selectivity, and binocularity of collicular cells (Wickelgren and Sterling 1969). Cortical cooling experiments in the cat have revealed differential effects of various visual areas on visual responses in different SC laminae. The primary areas were found to have a stronger impact on more superficial SC layers, whereas input from suprasylvian areas seems to be critical for visual responsiveness of deeper SC laminae (Ogasawara et al. 1984). On the basis of these cooling experiments, it was argued that striate and extrastriate cortex exert largely separate and exclusive effects on the different SC laminae and that there are two largely independent corticotectal systems (Ogasawara et al. 1984). Evidence is also available for tectal influences on cortical receptive fields. On the basis of cat experiments involving SC lesions, it has been argued that the colliculus exerts an inhibitory influence on extrastriate cortical neurons if these are activated by stationary or slowly moving stimuli (Smith and Spear 1979).

Comparatively little is known, however, about cooperative interactions among the multiple corticotectal projections. Theoretical studies have suggested that neuronal codes are relational and that information is contained in the constellation of jointly active cells. Such cooperating groups of cells have been termed assemblies (Hebb 1949) and are thought to serve the representation of both sensory stimuli and motor programs (Braitenberg 1978; Edelman 1987; Georgopoulos 1995; Palm 1990). The dynamic association of cells into such functionally coherent assemblies initially had been proposed to be achieved by jointly raising the discharge rate of the cells participating in the assembly (Braitenberg 1978; Hebb 1949; Palm 1990). As this can lead to superposition problems and precludes a rapid succession of different assemblies within the same population of neurons, it has been suggested that assemblies also could be formed by synchronization of the discharges of the respective neurons, whereby the temporal precision of synchronization would have to be in the millisecond range (von der Malsburg 1986). In agreement with this proposal of a temporal assembly code, evidence from multielectrode recordings indicates that cells distributed within and across cortical areas can synchronize their responses transiently and in rapidly changing constellations (for review see Engel et al. 1992, 1997; Singer and Gray 1995).

The concept of temporal assembly codes and their hypothetical role in cortical information processing has implications for the study of corticotectal interactions. Because the cortical assembly codes are distributed and relational, the information contained in such codes can be deciphered only by the tectum if not only spatial but also the temporal relations between distributed cortical responses are evaluated. This requires faithful transmission of temporal patterns to the tectum and readout of temporal information by the SC. One prediction is that SC cells should be sensitive to cortical timing and that temporal correlations among cortical responses should propagate to the SC, whereby synchronized activity should have a stronger effect on collicular cells than temporally unstructured activity. If so, one should observe temporal correlations between cortical and tectal responses and the strength of corticotectal coupling should reflect the extent of synchronization among cortical neurons. To examine these possibilities, we performed cross-correlation analysis of responses recorded simultaneously with multiple electrodes from superficial layers of the SC and a number of visual cortical areas. Specifically, we have investigated cortical areas 17, 18, 21a, the postero medial lateral suprasylvian area (PMLS), and the postero lateral lateral suprasylvian area (PLLS).

	METHODS

Abstract Introduction Methods Results Discussion References

Data were recorded from nine adult anesthetized cats. Anesthesia was induced with ketamine and xylazine (10 mg/kg and 2.5mg/kg, respectively) and was maintained with a mixture of 70% N₂O and 30% O₂ supplemented by halothane (0.6-1%). After tracheotomy, the animal was placed in a stereotactic headholder. A craniotomy was performed, and the skull was cemented to a metal rod. After completion of all surgical procedures, the ear and eye bars were removed and the halothane was reduced to a level of 0.4-0.6%. After we had assured that the level of anaesthesia was stable and sufficiently deep to prevent any vegetative reactions to somatic stimulation, the animals were paralyzed with pancuronium bromide (0.2 mg·kg¹·h¹). Glucose and electrolytes were supplemented intravenously. The electrocardiogram and the electroencephalogram were monitored continuously, and end-tidal CO₂ and rectal temperature were kept in the range of 3-4% and37-38°, respectively. Corneal contact lenses with an artificial pupil of 3-mm diam were fitted to both eyes. The eyes were refracted for a viewing distance of 1.14 m, which is where a tangent screen was positioned. As landmarks of the animal's visual field, the optic disks and the areae centrales were plotted with a reversible ophthalmoscope.

Multiunit activity usually was recorded with two electrode arrrays, one being placed into the SC and the other into a visual cortical area, corresponding to either area 17, 18, PMLS, PLLS, or 21a. In a few additional measurements, simultaneous recordings were made from the SC and two of the visual areas. Each array consisted of two to five microelectrodes the spacing of which was between 0.2 and 3 mm. The recorded signals were amplified, band-pass filtered, and fed through a Schmitt trigger to obtain TTL pulses, which signaled spike timing. The trigger threshold was adjusted to exceed the noise limit by at least twofold. Receptive fields were mapped onto a tangent screen, and the ocular dominance, the orientation tuning and the direction preferences of each multiunit cluster were assessed with hand-held stimuli. For quantitative measurements, visual stimuli were projected onto the tangent screen via a computer-controlled optic bench. Typically, moving bars served as stimuli. Multiunit clusters with overlapping or nearby receptive fields were activated with a single bar, whereas cell groups with distant receptive fields were activated with two coherently moving bars. In a few cases, computer-generated flow-fields were presented on a 20-in. monitor. Responses were recorded for 10 stimulus repetitions, and, if not specified otherwise, the data from such blocks of 10 trials were combined for analysis.

For all responses peristimulus time histograms, auto- and cross-correlation functions as well as their first-order shift predictors were computed. The correlograms were computed over the whole response epoch, which generally consisted of a 3-s period during which a bar stimulus moved across the receptive field. Correlograms were included in the analysis only if the firing rates of each cell cluster exceeded 10 Hz and if the correlogram had an average entry of at least two coincidences per bin. To quantify the correlogram modulation, generalized Gabor functions were fitted to the data as described previously (König 1994). The strength of cross-correlation was measured by the relative modulation amplitude (RMA), which we define as the ratio of center peak amplitude over the offset of the correlogram modulation. A correlation was considered as significant if the fitted function explained 20% of the variance of the data points, if the largest peak was significant at the 5% level, if the RMA exceeded 0.10, and if the shift predictor was flat, i.e., if it was not significantly modulated. Moreover, a correlogram was considered to indicate that the activity patterns were oscillatory if the largest satellite peak was significant at the 5% level. From the fitted functions we also determined the time shift of the largest peak relative to zero time lag, the peak width of the central peak at half height, and the oscillation frequency in case of oscillatory patterning. From these data, we calculated the percentage of recording pairs in which a significant correlation was observed. For a given pair with significant interactions, the percentage of modulated correlograms was calculated as a measure of the penetrance of the respective correlation.

Furthermore, we studied the general relation between corticotectal correlations and other correlation patterns. Thus for each block of 10 stimulus repetitions, we compared the correlation patterns of cortical and collicular cells, which showed a corticotectal correlation with the correlation patterns of units that did not engage in a significant corticotectal interaction. In a small selection of cases, we assessed, in addition, the covariance of cortical and corticotectal synchronization on a trial-by-trial basis. This was performed in cases where we observed high firing rates, where the cortical cell group was engaged in both strong cortical and corticotectal synchronization, and where the same stimulus had been presented 30 times. If these conditions were met, we computed and fitted cortical and corticotectal cross-correlograms for each individual stimulus presentation. The relation between cortical and corticotectal correlation strength was then evaluated by regression analysis.

At the end of each experiment, a lethal dose of sodium pentothal was given, and the animal was perfused through the heart with warm saline followed by cold (4-8°C) fixative (4% paraformaldehyde in phosphate-buffered saline). The brain was removed, frozen, and cut in the frontal plane into 60-µm sections. These sections were alternatingly stained for cell bodies (Nissl) and fibers (Gallyas method). Small lesions (electrode tip negative, 12 s DC-current, 12 µA) made after each recording penetration allowed the reconstruction of the recording track. Cortical areas were determined based on the electrode coordinates according to the partitioning schemes of Tusa et al. (1981) and Updyke (1986). The area 17/18-border was determined using additional histological criteria (Payne 1990). SC laminae were classified according to Kanaseki and Sprague (1974).

	RESULTS

Abstract Introduction Methods Results Discussion References

Recording sites and pairs

The results reported here are based on recordings from 224 cortical sites and 215 collicular sites in the superficial SC laminae. The distribution of cortical recording sites across areas and layers is listed in Table 1A. Of the 215 SC sites, 138 were located in the stratum griseum superficiale, and 33 in the s. opticum; 44 recordings could not be attributed unambiguously to one of the two laminae. Corticotectal interactions were studied in 489 cortex-SC recording pairs. Table 1B lists the locations of the cortical recordings in such pairs and the number of pairs in which a significant correlation was observed. In addition, we analyzed 178 recording pairs within the SC and 239 corticocortical pairs that had been recorded simultaneously with the corticotectal response pairs. Table 1C summarizes the locations of these pairs and the number of pairs in which a significant correlation was observed.

View larger version (39K): [in this window] [in a new window]   FIG. 1. Example of an oscillatory corticotectal interaction. A: data from a measurement where we recorded simultaneously from the superior colliculus (SC; stratum griseum superficiale) and area 17. LAT, lateral sulcus. B: schematic plot of the receptive fields for the 2 recording sites. Circle indicates the position of the area centralis. Fields were overlapping and hence, the cells were activated with a single moving light bar. C: cross-correlogram for the 2 responses. Note the slight phase-lead of the cortical activity. D: autocorrelogram for the response obtained at the area 17 recording site. E: autocorrelogram for the response at the SC recording site. Black continuous line superimposed to the correlograms represents the generalized Gabor function that was fitted to the data. , phase shift of the Gabor function.

View larger version (37K): [in this window] [in a new window]   FIG. 2. Example of a nonoscillatory corticotectal interaction. A: data are from a measurement where we recorded simultaneously from the SC (s. griseum superficiale) and area 18 (supragranular layers). B: schematic plot of the receptive fields for the 2 recording sites. Fields were nonoverlapping, but the cells were activated with a single moving light bar. Circle, area centralis. C: cross-correlogram for the responses obtained in the SC and in area 18. D: autocorrelogram for the responses obtained at the SC recording site. E: autocorrelogram for the responses obtained at the area 18 recording site. Note that the autocorrelation functions indicate different dynamics for the cortical site (oscillation in the gamma-range) and the SC site (modulation in the alpha-range). Black continuous line superimposed to the correlograms represents the generalized Gabor function that was fitted to the data. , phase shift of the Gabor function.

General properties of corticotectal correlations

Significant interactions were observed between all investigated cortical areas and the superficial SC layers. Of the 489 cortex-SC pairs, 117 (24%) showed a significant correlation. Certain features of significant corticotectal interactions were independent of the area or layer of the cortical recording site, whereas others showed a systematic dependence. The former include the strength of the interactions, the dependence of interactions on receptive field overlap, the width of the center peaks, the incidence of oscillatory correlograms, the oscillation frequencies, and the variability of the correlations. In contrast, the incidence of correlations and the phase relationships of corticotectal interactions did depend on cortical areas and layers. We shall first describe the invariant features.

Figures 1 and 2 show examples of corticotectal correlograms together with the autocorrelograms of the respective responses. The examples demonstrate that corticotectal correlations can occur between cortical and SC units irrespective of whether their responses exhibit similar (Fig. 1) or different temporal structures (Fig. 2).

Figure 3A shows the distribution of the respective best modulation amplitudes (RMA; central peak amplitude divided by offset) of correlograms for the different recording pairs. These RMA values were assessed from correlograms that were averaged over 10 successive stimulus presentations. The average of these maximal RMA values was0.55 ± 0.23 (mean ± SD). If averaged over all correlograms, the mean RMA was 0.34 ± 0.17. Much stronger RMAs were observed in single, nonaveraged correlograms (see also Figs. 4 and 7). The incidence of corticotectal correlations depended strongly on the overlap of the receptive fields of the respective units. As depicted in Fig. 3B, significant corticotectal correlations occurred 2.5 times more often for cases in which cortical and SC units had overlapping receptive fields than for recording pairs with nonoverlapping fields. The distribution of center peak widths of the respective best correlograms is displayed in Fig. 3C. The majority of peaks were broad with an average width at half height of 51 ms.

View larger version (42K): [in this window] [in a new window]   FIG. 3. General properties of corticotectal interactions. , median of the distribution. A: distribution of relative modulation amplitudes (RMA), i.e., the ratios of peak amplitude over the offset of the correlogram. B: overlap dependence of the interactions. Plot compares the incidence of synchrony for corticotectal pairs with overlapping and nonoverlapping receptive fields (RFs), respectively. ** Significance level of P < 0.001. C: distribution of center peak widths (measured at half height) of the corticotectal correlograms. D: distribution of oscillation frequencies of oscillatory cross-correlograms. Note that the values in A, C, and D refer to the correlograms with the strongest modulation for each recording pair.

View larger version (89K): [in this window] [in a new window]   FIG. 4. Variability of corticotectal interactions. A: schematic plot of the receptive fields for 2 recording sites in the SC and in the postero lateral lateral suprasylvian (PLLS) area. Ten successive trials were recorded with the same stimulus. Gray spot indicates the position of the area centralis. B: averaged corticotectal cross-correlogram. C: averaged autocorrelogram for the cortical recording site. D: averaged collicular autocorrelogram. E and G: variability of the temporal structure across the 10 individual trials in the corticotectal cross-correlation (E), in the cortical autocorrelation (F), and in the collicular autocorrelation (G). Because of noise not all responses showed, according to our criteria, a significant modulation. However, in most single-trial cross-correlograms there was some indication of a center peak. In the significantly fitted cases, the oscillation frequency of the correlogram varies between 9 and 16 Hz. Note that in many of the individual trials, substantially stronger oscillations are evident than in the averaged correlograms. Where present, the black continuous line superimposed to the correlograms represents the generalized Gabor function that was fitted to the data. , phase shift of the Gabor function.

View larger version (30K): [in this window] [in a new window]   FIG. 7. Covariance of cortical and corticotectal correlations across individual trials. For each response epoch, the RMA computed for the corticotectal interaction was plotted against the RMA of the respective intracortical interaction, and a regression line was calculated. Note that many entries coincide at the origins of the graphs. A: interaction of an area 18 cell cluster with a 2nd area 18 cell group and with a cluster in the SC. Data are from 30 stimulus repetitions. B-D: covariance of interactions of an area PLLS cell group with a 2nd PLLS recording and with a cell cluster located in the SC. Numbers of trials were 80 (B), 60 (C), and 30 (D), respectively. Note the large variability of interaction strength across trials and that the corticotectal correlations tended to be substantially weaker than local cortical interactions. In these examples, all cortical recordings were made from supragranular layers. Collicular recordings were made from the s. griseum superficiale (SGS) in the cases shown in A, C, and D and from the SO in the measurement illustrated in B. R2, correlation coefficient between corticotectal and cortical RMAs. P, probability that cortical and corticotectal RMAs are independent.

Correlograms of corticotectal correlations often exhibited at least one significant satellite peak on either side of the center peak. According to this criterion, 49% of the significant correlograms were considered to be oscillatory. As shown in Fig. 3D, these oscillations covered a broad frequency range (5-79 Hz). The majority of oscillations were in the alpha-range and the mean frequency averaged over all correlograms was 19 Hz. Except for the few cases of high-frequency oscillations (>50 Hz) that were restricted to interactions between areas 17 and 18 and the SC, oscillation frequencies did not differ across cortical areas. In most cases, correlograms rated as oscillatory showed only one satellite peak on either side of the center peak, indicating that the oscillation frequencies were not stable over time. Accordingly, much stronger oscillations were apparent in correlograms computed from individual responses than in correlograms averaged over several stimulus presentations. As illustrated in Fig. 4, both the probability of occurrence and the frequency of the oscillations varied from sweep to sweep and this led to an attenuation of the oscillatory patterning in the summed correlograms. Fig. 4 shows that trial-by-trial variability also was found for the strength and phase shift of corticotectal correlations. It is, thus, difficult to assess the precision of synchronization from the averaged correlograms and to establish relations between the precision of synchronization and oscillatory patterning. Comparison of the widths of the center peaks in oscillatory and nonoscillatory correlograms revealed no difference.

Regional variation of corticotectal correlation patterns within superficial SC

When pooled over all cortical recording sites, the corticotectal correlation patterns did not differ significantly between the two superficial collicular laminae, the s. griseum superficiale (SGS), and the s. opticum (SO). Neither the incidence (SGS: 21%; SO 23%; P = 0.76) nor the average RMAs (SGS: 0.37; SO: 0.32; P = 0.17), the oscillation frequencies (SGS: 16 Hz; SO: 17 Hz; P = 0.69), or the phase shifts (SGS: 1.8 ms time lag to cortex; SO: 3.9 ms time lag to cortex; P = 0.28) distinguished the two laminae. An interesting finding in this context is that significant correlations existed between area 17 and both SGS and SO despite the fact that the corticotectal projection from area 17 (unlike the projection from the other recorded areas) is largely confined to the SGS (Harting et al. 1992).

The cat SC contains a substantial representation of the ipsilateral visual field. These ipsilateral responses are mediated mainly by corticotectal projections and depend little on direct retinal inputs (Antonini et al. 1978). Comparing corticotectal correlation patterns between the ipsi- and contralateral visual field representation therefore should provide indications for the influence of shared direct retinal input to both cortex and SC. Unlike primary visual areas, extrastriate areas PMLS and PLLS also contain a substantial ipsilateral visual field representation. We therefore restricted our comparison to interactions between these lateral suprasylvian (LS) areas and sites in the SC. In 74 of these SC-LS pairs, the SC unit had an ipsilateral and in 188 of these pairs, a contralateral receptive field. The corticotectal correlation patterns of the two populations were indistinguishable. The incidence (SC fields contralateral: 26%; SC fields ipsilateral: 31%; P = 0.42), average RMAs (SC fields contralateral: 0.35; SC fields ipsilateral: 0.33; P = 0.60), the average oscillation frequencies (SC fields contralateral: 15 Hz; SC fields ipsilateral: 17 Hz; P = 0.46), and the average phase shifts (SC fields contralateral: 4.3-ms delay to cortex; SC fields ipsilateral: 4.6 ms delay to cortex; P = 0.88) were very similar across both visual field representations.

Layer-dependence of corticotectal correlations

Histological reconstruction allowed localization of 352 of the cortical recording sites to either layers II/III, layer IV or layer V (see Table 1B). The analysis of the layer specificity of corticotectal interactions shows clearly that these are by no means confined to layer V, where corticotectal projection neurons are located. To our surprise, none of the 27 recording pairs involving putative layer V units actually showed a significant corticotectal correlation. In part, this can be explained by the fact that, for the majority of pairs involving layer V cells, the cortical site was located in area 17 for which the incidence of corticotectal correlations was generally low (see next section).

Corticotectal correlation patterns recorded in layers II/III and layer IV, respectively, are compared in Fig. 5. Neither the incidence (Fig. 5A) nor the strength (Fig. 5B) of corticotectal correlations differed between these layers. However, differences between the layers were noted with respect to the time shifts of the respective interactions (Fig. 5C). Whereas the discharges of layer II/III neurons were found to precede the SC activity on average by 2.7 ms, the firing of layer IV cells was, on average, 4.1 ms delayed with respect to that of the SC units. These differences were statistically significant, but it should be noted that the number of layer IV cases was rather small (n = 12). Moreover, it must be emphasized that most corticotectal correlation peaks were rather broad compared with the earlier described differences in the time shifts of the different cortical layers.

View larger version (39K): [in this window] [in a new window]   FIG. 5. Dependence of the interactions on cortical layers. , median of the distribution. A: occurrence of corticotectal interactions. Incidence (light gray column: percentage of cell pairs in which a significant corticotectal correlation was observed) and penetrance (dark gray column: percentage of response epochs in which a correlation appeared for pairs with significant interactions) did not differ between cortical layers II/III and IV, respectively. B: synchronization strength. Distribution of the RMAs of corticotectal correlations did also not differ across cortical layers. C: phase relationships. Time shifts of corticotectal interactions differed across cortical layers. Supragranular layers showed an average phase lead over the SC of ~2.5 ms, whereas cortical layer IV was lagging on the average 4 ms behind the SC. *P < 0.05; n.s., not significant.

Interareal differences in corticotectal interactions

Corticotectal correlations were found for all investigated cortical areas: areas 17, 18, PMLS, PLLS, and 21a (Table 1B). Because only very few cortical recording sites were located in area 21a, this area is excluded from the following comparison. It should be noted that different cortical areas were recorded to a large extent in different experiments and that therefore differences between experimental sessions could have affected our conclusions about areal differences. However, the inspection of data in which we recorded the same areas in multiple experiments did not indicate such a bias.

The incidence of corticotectal correlations varied substantially across cortical areas (Fig. 6). The percentage of cells engaging in a corticotectal correlation was much lower in areas 17 and PMLS than in areas 18 and PLLS (Fig. 6A). For pairs showing significant interactions, the penetrance of the correlation (i.e., the percentage of response epochs in which the correlation actually appeared for a given pair of multiunit recordings) showed a distribution similar to that of the incidence. The penetrance of corticotectal correlations was lowest for area 17, higher for areas PMLS and 18, and strongest for area PLLS.

View larger version (41K): [in this window] [in a new window]   FIG. 6. Interareal differences of corticotectal interactions. , median of the distribution. A: occurrence of interactions. Both the incidence (light gray column: percentage of cell pairs in which a significant corticotectal correlation was observed) and penetrance (dark gray column: the percentage of response epochs in which the correlation appeared for pairs with significant interactions) of corticotectal correlations varied across cortical areas (*P < 0.05). B: synchronization strength. Distributions of the RMA of corticotectal correlations did not differ significantly between cortical areas. C: phase relationships. Time shifts of corticotectal correlations differed across cortical areas. For area 17 and 18 neurons, the correlogram peaks were centered ~0-ms time shifts. However, neurons in areas postero medial lateral suprasylvian (PMLS) and PLLS discharged, on average, 4-5 ms before the SC units. Differences between areas PMLS/PLLS and areas 17/18 are statistically significant (P < 0.05).

The incidence of corticotectal correlations reflects the functional subdivision of cat extrastriate cortex. Different authors (Tusa et al. 1981; Updyke 1986) disagree on the exact location of the boundaries between areas PMLS and PLLS in the LS sulcus, but there is agreement that the receptive-field properties are very similar in both areas. Based on analysis of retinotopy, Tusa and coworkers (1981) suggested that the PMLS/PLLS boundary is located in the middle of the LS fundus, whereas Updyke (1986) proposed that area PMLS extends throughout the whole fundus and that area PLLS is restricted to a more lateral position in the bank of the LS sulcus. In our study, the incidence of corticotectal correlations was low (14%) for recordings that, according to both partitioning schemes, were located in area PMLS, but was significantly higher (35%) for recordings located in area PLLS. In the lateral part of the LS fundus (area PLLS according to Tusa et al. (1981), area PMLS according to Updyke), the incidence of corticotectal correlations was 17% and, thus, significantly lower than in the putative PLLS recordings. Therefore our data support the partitioning scheme of Updyke.

In contrast to incidence and penetrance, the strength of corticotectal interactions did not differ across areas, i.e., the distributions of the relative modulation amplitudes of significant correlograms were very similar (Fig. 6B). The same was true for the distribution of oscillation frequencies in correlograms with significant side peaks (data not shown). However, the average phase relationships between cortical and collicular discharges differed for the various cortical areas (Fig. 6C). The position of the correlogram peaks indicated that cells in areas 17 and 18 fired, on average, simultaneously with superficial SC cells. In contrast, PMLS and PLLS cells discharged, on average, 4-5 ms before the SC units.

Relationship between corticotectal correlations and synchronization within SC and cortex

In our experiments, we usually recorded from more than one recording site in the SC as well as in the cortex. This allowed us to address the question to which extent corticotectal interactions covaried with the cooperativity among collicular or cortical cells. In four selected cases, we studied the trial-by-trial covariance of intracortical synchronization and corticotectal correlation. In all cases, the strength of local cortical synchronization was substantially higher than that of corticotectal interactions. In three (Fig. 7, A-C) of the four cases, there was a significant positive correlation between the strength of intracortical and corticotectal synchronization (average correlation coefficient R² = 0.131). However, the degree of covariance between the intracortical and the corticotectal synchronization varied substantially between these samples.

We also investigated to what extent the occurrence of correlations between a particular pair of corticotectal recording sites was associated with correlations in another corticotectal pair, with local correlations within SC or cortex, and with correlations between different cortical areas. These analyses were performed over the whole set of averaged data, i.e., for correlograms computed from blocks of 10 stimulus presentations. The results indicate that neurons, if they exhibit significant corticotectal interactions, also have a high probability of synchronizing with neurons at other cortical or tectal recording sites (Fig. 8).

View larger version (26K): [in this window] [in a new window]   FIG. 8. Incidence of local and long-range correlation patterns for cells engaging (light gray columns) or not engaging (dark gray columns) in corticotectal interactions. A: covariance of corticotectal interactions. Left: for cortical cells not engaging in corticotectal interactions, there was a much lower probability of finding a corticotectal correlation with an additional randomly chosen SC cell group than for cortical cells that did engage in a corticotectal interaction. Right: similarly, the incidence of additional corticotectal interactions was higher for SC cells already involved in such interactions. B, left: for SC cells not engaging in corticotectal interactions, a much lower incidence of local synchronization within SC was observed than for SC cells involved in a corticotectal correlation. Right: simlarly, local cooperativity within cortex was related positively to the occurrence of corticotectal correlations. C: corticotectal correlations often were accompanied by interareal cortical synchronization. Left: corticotectal interactions were only observed for area 17 cells if these were involved in synchrony with a cell group in the suprasylvian (LS) cortex. Right: similarly, LS neurons also showed corticotectal interactions only if they were synchronized to area 17 neurons.

Thus a cortical cell cluster that engaged in a significant interaction with a SC cell group had a high probability of also exhibiting a significant correlation with another SC recording site. In contrast, cortical cells not engaging in a corticotectal interaction generally had a very low probability of synchronizing with a second, randomly chosen SC cluster (Fig. 8A). Similarly, collicular cells engaging in a correlation with one cortical cell cluster had a high probability of being correlated with other cortical sites, whereas SC units not engaging in a correlation with a randomly chosen cortical cell group also had a low probability to engage in correlations with other cortical sites.

As shown in Fig. 8B, collicular cells engaging in a corticotectal correlation also showed an approximately twofold higher probability of synchronizing with collicular cells at another site than SC cells not engaging in a corticotectal correlation. The same relation holds for cortical cells. Cortical neurons involved in corticotectal correlations had a twofold higher probability to be correlated with other cortical sites as compared with cells showing no corticotectal interaction.

For an area 17 cell group engaging in a corticotectal correlation, the probability of also synchronizing with cells in PMLS or PLLS was sevenfold higher than for area 17 cells not engaging in a corticotectal interaction. Likewise, LS units exhibiting significant corticotectal correlations also had a higher probability to engage in interareal synchrony with other cortical recording sites (Fig. 8C). Although the data sample used for covariance analysis between interareal and corticotectal interactions is small for technical reasons, statistical analysis showed that these effects are highly significant. One of the cases with simultaneous recordings from two cortical areas and the SC is illustrated in Fig. 9. Here, the phase relationships of the respective correlation patterns also point to an involvement of interareal synchronization in the generation of significant corticotectal interactions. The PLLS recording site leads 17 ms over the SC and leads 6 ms over area 17, whereas area 17 leads 11 ms over the SC; thus the phase relationships are consistent with the hypothesis that spikes that are correlated between the two cortical areas also contribute to the corticotectal correlations of the respective units. This observation, in particular, suggests that cortical interareal synchrony might play an important role in the establishment of corticotectal correlations.

View larger version (42K): [in this window] [in a new window]   FIG. 9. Example of a measurement where simultaneous recordings were made from the SC and 2 cortical sites, 1 located in area 17 (supragranular layers) and 1 in the extrastriate area PLLS (supragranular layers). A: position of the recording electrodes. SUPS, suprasylvian sulcus. B: plot of the receptive fields for all 3 recording sites. All fields were overlapping and, hence, the cells were activated with a single moving light bar. C: cross-correlogram for the responses obtained at the SC and at the PLLS recording site. Correlogram indicates a periodic coupling with an oscillation frequency of ~11 Hz and shows a pronounced phase shift () of ~17 ms, indicating a phase-lag of the SC spikes. D: cross-correlogram for the responses obtained at the SC and the area 17 recording site. Correlation shows a time shift of ~11 ms, again indicating a phase-lag of the SC spikes. Note that the phase shift is smaller than that observed for the PLLS-SC pair. E: cross-correlogram for the interaction between the 2 cortical cell groups in areas 17 and PLLS. As in C, the interaction occurs in the alpha-range (11 Hz). Shift of the correlogram peak indicates a phase-lead of the PLLS spikes. Note that the phase relations between the different recording sites are consistent with the idea that the spikes synchronized between areas 17 and PLLS are those that correlate with the collicular spikes. Black continuous line superimposed to the correlograms represents the generalized Gabor function that was fitted to the data.

Oscillatory activity and corticotectal correlations

Studies on neural synchrony within visual cortex suggest a close relation between oscillatory activity and cortical long-range synchronization (König et al. 1995). As described above, about half of the significant corticotectal interactions were classified as oscillatory. To examine the relation between oscillatory activity and the probability of corticotectal correlations, we determined whether cortical cells involved or not involved in corticotectal interactions differed in the oscillatory patterning of their responses. As illustrated in Fig. 10A, oscillatory responses were 3.5-fold more frequent in cortical cell clusters engaging in correlations with the SC than in cortical cell groups not involved in such interactions. For SC cells, however, this relation was less pronounced (Fig. 10B). SC cells involved in corticotectal interactions showed only a slightly higher incidence of oscillatory autocorrelograms than SC units not engaging in corticotectal correlations.

View larger version (19K): [in this window] [in a new window]   FIG. 10. Incidence of oscillatory autocorrelation functions (ACFs) for cell groups that did (light gray column) or did not (dark gray column) engage in corticotectal interactions. A: in cortical cell groups involved in corticotectal correlations, oscillatory firing patterns were strongly overrepresented. B: for collicular cells the association between oscillatory activity and corticotectal correlation was much weaker, but nonetheless significant. **P < 0.001.

In several instances, the correlograms of the corticotectal interaction and the respective cortical and collicular autocorrelograms indicated that the correlated neurons were oscillating at the same or very similar frequencies. Interestingly, however, this was true only for a minority of cases. In general, oscillation frequencies of corticotectal cross-correlograms and of the respective cortical autocorrelograms (for cases in which both were flagged as oscillatory) were related only weakly (R² = 0.107; P = 0.006). Oscillation frequencies of corticotectal correlograms and the respective SC autocorrelograms (for cases in which both were oscillatory) showed no significant relation at all (R² = 0.004; P = 0.719). Similarly, in cases where both cortical and collicular cell clusters showed significant interactions and both had oscillatory autocorrelograms, there was no significant correlation between cortical and collicular oscillation frequencies (R² = 0.068; P = 0.242).

	DISCUSSION

Abstract Introduction Methods Results Discussion References

Comparison of corticotectal interactions with intracortical and collicular correlation patterns

The results document that the temporal dynamics of corticotectal interactions share many of the properties described for neuronal interactions within and between visual cortical areas. Like the synchronization phenomena observed in cat and monkey visual areas (Eckhorn et al. 1988; Engel et al. 1990; Gray et al. 1989; Kreiter and Singer 1996; Livingstone 1996), corticotectal synchronization also is associated often with a rhythmic modulation of neuronal firing, whereby synchronization strength and oscillation frequency vary considerably from one stimulus presentation to the next, and retinotopic proximity of the respective receptive fields is a major predictor of the incidence of synchronization. Corticotectal correlations exhibit on average a lower incidence and synchronization strength than local intraareal correlations in the visual cortex (Engel et al. 1990; Gray et al. 1989), but resemble in this respect the cortical long-range interactions such as those occurring between different visual areas (Engel et al. 1991b; Nelson et al. 1992) or across the cerebral hemispheres (Engel et al. 1991a; Nowak et al. 1995). In comparison with both intra- and interareal cortical interactions, corticotectal correlations are characterized by broader correlation peaks and slower oscillations frequencies. However, some authors also have reported a large percentage of broad correlations in visual cortex (Arieli et al. 1995; Nelson et al. 1992; Nowak et al. 1995). Generally, corticotectal interactions share properties of the synchronization phenomena observed within the SC of anesthetized cats. In particular, the width of the correlation peaks and the oscillation frequencies observed in corticotectal correlograms were very similar to the corresponding parameters of intracollicular interactions in anesthetized cats (Brecht et al. 1996). However, it should be noted that in the SC of awake cats higher oscillation frequencies and sharper correlation peaks predominate (unpublished observations), and it is therefore conceivable that the pattern of corticotectal correlations also is affected by the anesthesia.

Regional differences of corticotectal interactions

Corticotectal interactions differed for the various cortical areas, in particular with respect to their incidence and phase relationship. Quite unexpectedly, neurons in early visual areas (17 and 18) were found to discharge, on average, synchronously with SC cells, whereas cells in extrastriate LS areas tended to lead in corticotectal correlations. Thus the phase relations of corticotectal interactions do not reflect the presumed serial processing of retinal signals from primary to higher order cortical areas.

The comparatively low incidence and penetrance of corticotectal correlations for area 17 neurons might be related to the fact that retinal afferents to area 17 responses are predominantly of the X type (Hofmann and Stone 1971; Tretter et al. 1975). In contrast collicular responses and responses in area 18 and the suprasylvian areas are mediated mainly by Y- and W-type afferents (Colby 1988; Hoffmann 1973). The high incidence and penetrance of correlations between neurons in area PLLS and the tectum is in good accordance with other lines of evidence pointing to a close functional relationship between area PLLS and the SC. Unlike the other investigated cortical areas, area PLLS receives SC feedback via a strong projection from the tectorecipient zone of the LP-pulvinar complex (Raczkowski and Rosenquist 1983; Symonds et al. 1981); moreover many PLLS cells have visuomotor properties (Komatsu et al. 1983; Yin and Greenwood 1992) and lesions to this area lead to symptoms similar to the severe neglect caused by SC ablation (Hardy and Stein 1988). Interestingly, the incidence of corticotectal correlations was substantially lower in area PMLS than in area PLLS. This is remarkable, because, given the great similarities of receptive-field properties, there was a lack of physiological evidence for a functional specialization of these two areas.

Our finding that responses in both striate and extrastriate areas correlate strongly with responses of cells in superficial SC is in conflict with the view of two corticotectal systems but agrees with anatomic evidence that all visual areas project primarily to the superficial SC in broadly overlapping termination zones (Harting et al. 1992; Segal and Beckstead 1984). Ogasawara et al. (1984) proposed two largely independent and exclusive corticotectal systems, an extrastriate stream from the suprasylvian areas to the deep SC laminae and another stream from areas 17 and 18 to superficial SC. One possible explanation for this discrepancy is that Ogasawara et al. evaluated firing rates, whereas we analyzed spike timing. However, it should be noted that even in the data of Ogasawara et al., LS cooling had an effect on discharge rates of superficial SC cells.

In contrast to the variations across cortical areas, we observed only a slight dependence of corticotectal interactions on cortical layers. The abundance of corticotectal correlations in cortical layers II/III and IV indicates that correlations among cortical and collicular responses are not restricted to monosynaptically connected neurons. Rather, it appears as if the whole cortical column is involved in the corticotectal synchronization process. The lack of significant corticotectal correlations in the pairs involving layer V cell groups was unexpected given the fact that all corticotectal projection neurons reside in this layer. We attribute this result to the small sample size and to the fact that most of the recording sites located in layer V were in area 17, where corticotectal correlations were less common. Obviously, layer V cells must participate in the synchronization process, but our approach was not designed to study the specific role of this class of neurons. These cells sample activity from all cortical layers, and this is likely to account for the layer independence of corticotectal correlations. Corticotectal projection neurons extend their apical dendrites up into layer I, and their extended basal dendrites reach into layer VI. Thus they are the only cortical neurons the dendrites of which span all cortical layers (Connors and Amitai 1995). Moreover, active conductances in the dendritic tree of these cells appear to boost excitatory postsynaptic potentials from distal dendrites (Amitai et al. 1993; Stuart and Sakmann 1994). Electrophysiological studies in cortical slices have revealed that all corticotectal projecting cells show intrinsic burst response patterns (Kasper et al. 1994; Wang and McCormick 1993) and that they make essential contributions to a variety of rhythmic activities observed in such slices (Connors and Amitai 1995). We hypothesize that these morphological and biophysical properties of corticotectal cells are particularly well suited to transmit responses of cortical cells to the colliculus, when these responses are synchronized within the cortical column and oscillate in a frequency range that matches the resonance properties of bursting layer V cells.

Mechanisms of corticotectal interactions

Several arguments suggest that the corticotectal correlations observed in this study were due to population dynamics rather than to common retinotectal input or simple serial excitation. Corticotectal interactions were not restricted to monosynaptically connected pairs of collicular and cortical cells. Moreover, there were no differences in corticotectal correlation patterns between SC sites representing the contralateral and ipsilateral visual field, respectively. The ipsilateral SC responses depend little on monosynaptic retinal inputs but are mediated via a retino-cortico-tectal loop involving the corpus callosum (Antonini et al. 1978). This makes it unlikely that corticotectal correlations were due to shared retinal input. Synchronization via common retinal input is also unlikely because stimulus-induced oscillations of retinal neurons have a much higher frequency than the oscillatory corticotectal interactions (Neuenschwander and Singer 1996). The interpretation that corticotectal interactions arise from cooperativity among large and for the most part only polysynaptically connected neurons in both SC and cortex also is supported by our triple recordings. These show that corticotectal correlations depend on the degree of collicular or cortical cooperativity, suggesting that temporal coherence within cortex and within SC is important for establishing corticotectal correlations. Finally, the differences in the temporal patterning of cortical and collicular responses, reflected by differences in the width of correlogram peaks and in oscillation frequencies, preclude a simple scenario in which cortical cells force tectal neurons to adopt their dynamics.

The complexity of corticotectal interactions also is highlighted by the observation that cells in SC and areas 17-18 tend to discharge simultaneously, a fact that is hard to reconcile with a purely feedforward corticotectal signal transfer. One possibility is that recurrent loops via thalamic structures such as the C layers of the lateral geniculate nucleus or the tectorecipient zone of the LP pulvinar-complex contribute to corticotectal synchronization. Another possibility is that extrastriate areas initiate corticotectal synchronization and in parallel entrain neurons in areas 17 and 18. This is suggested by the fact that extrastriate neurons tend to fire earlier in corticotectal correlation patterns and that they are the ones most strongly involved in these interactions. Moreover, it seems that area 17 cells are much more likely to engage in corticotectal correlations if they get synchronized to neurons in extrastriate cortex. Taken together, these data indicate a prominent role of tectocortical and corticocortical feedback for setting up efficient interactions between cortex and the SC.

Studies on neural synchrony within visual cortex suggest that oscillatory activity may facilitate the establishment of long-range synchronization (König et al. 1995). In accordance with this hypothesis, about half of the corticotectal cross-correlations were oscillatory, and corticotectal interactions often were associated with cortical oscillations. However, despite of this trend, the weak or absent relation between cortical and tectal oscillation frequencies shows that corticotectal correlations do usually not occur as a uniformly synchronized oscillation across cortex and SC. The establishment of corticotectal correlations seems to be less dependent on oscillatory activity than long-range synchronization at the cortical level where nearly all of the correlograms have been found to be oscillatory (König et al. 1995). Generally, the broad distribution of peak widths and oscillation frequencies observed here compares to temporal characteristics found in studies on interareal interactions in awake animals (Bressler 1996; Roelfsema et al. 1997). However, it must be emphasized that due to the trial-by-trial frequency variability, the incidence and strength of oscillations is hard to assess quantitatively. Analysis techniques based on averaging, such as the correlograms computed here, underestimate the incidence of oscillatory patterning as the oscillations tend to average out.

Functional role of corticotectal correlations

Our results suggest that information about temporal patterns and, in particular, about synchronization of responses in distributed cortical assemblies is available at the level of the SC. Corticotectal correlations reflect this patterning, and their incidence depends crucially on the temporal coordination of cortical responses. Thus corticotectal connections do not only contribute to response properties in SC and cortex but carry also information about the temporal coordination of cortical responses. This raises the issue of the functional significance of such temporal information.

Information processing in visual cortex often is conceptualized as a sequence of serial operations that lead to the extraction and representation of feature constellations of increasing complexity as one proceeds along the hierarchically arranged processing levels (Barlow 1972). At the top of the processing hierarchy, a small number of high-level units is thought to provide an explicit representation of the specific constellation of features characterizing individual objects (in case of the ventral stream) or of the location and spatial configuration of these objects (in case of the dorsal stream). Therefore, one should expect that only high-level cortical areas send information to the SC. However, this is not the case. All visual cortical areas have massive corticotectal projections and thus influence tectal responses in parallel. This raises the question how the SC integrates and interprets the parallel output from multiple cortical areas. We suggest that the temporal patterning of cortical responses serves to label the cortical output that the SC needs to integrate and evaluate jointly. Based on the present data, we assume that widely distributed cortical responses that have become synchronized exert a stronger influence on collicular neurons and, hence, on the control of attention and orienting movements than nonsynchronized responses. The strong dependence of corticotectal interactions on local and long-range cooperativity at the cortical level could imply that the efficiency of information transfer to the SC is regulated by the degree of synchronization among cortical neurons.

To fully appreciate the functional relevance of corticotectal correlations, further data are required on the role of temporal coding for sensorimotor transformations that occur within the colliculus (Moschovakis 1996; Stein and Meredith 1991). It is known that multiple, spatially segregated visual stimuli lead to simultaneous activation of overlapping neuronal populations in the SC (McIlwain 1991). In the case of multiple stimuli, the synchronous firing of neurons within an assembly and the asynchronous firing of cells belonging to different assemblies could contribute to the selection of discrete, stimulus-specific motor responses from spatially nondiscrete collicular activity patterns. The pattern of cortical synchronization could provide information about the outcome of scene segmentation (Engel et al. 1997; Singer and Gray 1995) and could contribute to define target related assemblies in the SC. Moreover, temporal coherence of selected cortical responses could enhance activity in the corresponding collicular assembly but not the other. Both mechanisms, differential synchronization of simultaneously active collicular assemblies and selective enhancement of responses of one of the active assemblies, predictably bias and facilitate competition among these assemblies, a process that could be crucial for collicular target selection.

	ACKNOWLEDGEMENTS

  We thank S. Neuenschwander for providing analysis software, C. Selignow, J. Klon-Lipok, U. Hörbelt, and P. Janson for excellent technical assistance, and R. Ruhl and S. Ruhl for help in preparing the figures.

  This work was supported by the Max-Planck-Society, the Minna-James-Heineman Foundation and by the Heisenberg-Program of the Deutsche Forschungsgemeinschaft.

	FOOTNOTES

  Address for reprint requests: M. Brecht, Max-Planck-Institut für Hirnforschung, Deutschordenstrae 46, 60528 Frankfurt, Germany.

  Received 20 August 1997; accepted in final form 13 January 1998.

	REFERENCES

Abstract Introduction Methods Results Discussion References

• ALBANO, J. E., MISHKIN, M., WESTBROOK, L. E., WURTZ, R. H. Visuomotor deficits following ablation of monkey superior colliculus. J. Neurophysiol. 48: 338-351, 1982.[Free Full Text]
• AMITAI, Y., FRIEDMAN, A., CONNORS, B. W., GUTNICK, B. W. Regenerative activity in apical dendrites of pyramidal cells in neocortex. Cereb. Cortex 3: 26-38, 1993.[Abstract/Free Full Text]
• ANTONINI, A., BERLUCCHI, G., MARZI, C. A., SPRAGUE, J. M. Indirect, across-the-midline projections and representation of ipsilateral visual field in superior colliculus of the cat. J. Neurophysiol. 41: 285-304, 1978.[Abstract/Free Full Text]
• ARIELI, A., SHOHAM, D., HILDESHEIM, R., GRINVALD, A. Coherent spatiotemporal patterns of ongoing activity revealed by real-time optical imaging coupled with single-unit recording in cat visual cortex. J. Neurophysiol. 73: 2072-2093, 1995.[Abstract/Free Full Text]
• BARLOW, H. B. Single units and sensation: a neuron doctrine for perceptual psychology? Perception 1: 371-394, 1972.[Medline]
• BRAITENBERG, V. Cell assemblies in the cerebral cortex. In: Lecture Notes in Biomathematics 21, Theoretical Approaches in Complex Systems, edited by R. Heim, and G. Palm Berlin: Springer, 1978, p. 171-188.
• BRECHT, M., SINGER, W., ENGEL, A. K. Temporal coding in the cat superior colliculus. Soc. Neurosci. Abstr. 22: 1446, 1996.
• BRESSLER, S. L. Interareal synchronization in the visual cortex. Behav. Brain Res. 76: 37-49, 1996.[Medline]
• CHALUPA, L. M. Visual physiology of the mammalian superior colliculus. In: Comparative Neurology of the Optic Tectum, edited by and H. Vanegas New York: Plenum, 1984, p. 775-818.
• COLBY, C. L. Corticotectal circuit in the cat: a functional analysis of the lateral geniculate nucleus layers of origin. J. Neurophysiol. 59: 1783-1797, 1988.[Abstract/Free Full Text]
• CONNORS, B., AMITAI, Y. Functions of local circuits in neocortex: synchrony and laminae. In: The Cortical Neuron, edited by M. J. Gutnick, and I. Mody Oxford: Oxford Univ. Press, 1995, p. 123-140.
• ECKHORN, R., BAUER, R., JORDAN, W., BROSCH, M., KRUSE, W., MUNK, M., REITBOECK, H. J. Coherent oscillations: a mechanism for feature linking in the visual cortex? Biol. Cybern. 60: 121-130, 1988.[Medline]
• EDELMAN, G. M. In: Neural Darwinism: The Theory of Neuronal Group Selection. New York: Basic Books, 1987.
• ENGEL, A. K., KÖNIG, P., GRAY, C. M., SINGER, W. Stimulus-dependent neuronal oscillations in cat visual cortex: inter-columnar interaction as determined by crosscorrelation analysis. Eur. J. Neurosci. 2: 588-606, 1990.[Medline]
• ENGEL, A. K., KÖNIG, P., KREITER, A. K., SCHILLEN, T. B., SINGER, W. Temporal coding in the visual cortex: new vistas on integration in the nervous system. Trends Neurosci. 15: 218-226, 1992.[Medline]
• ENGEL, A. K., KÖNIG, P., KREITER, A. K., SINGER, W. Interhemispheric synchronization of oscillatory neuronal responses in cat visual cortex. Science 252: 1177-1179, 1991a.[Free Full Text]
• ENGEL, A. K., KREITER, A. K., KÖNIG, P., SINGER, W. Synchronization of oscillatory neuronal responses between striate and extrastriate visual cortical areas of the cat. Proc. Natl. Acad. Sci. USA 88: 6048-6052, 1991b.[Abstract/Free Full Text]
• ENGEL, A. K., ROELFSEMA, P. R., FRIES, P., BRECHT, M., SINGER, W. Role of the temporal domain for response selection and perceptual binding. Cereb. Cortex 7: 571-582, 1997.[Abstract/Free Full Text]
• FREEMAN, B., SINGER, W. Direct and indirect visual inputs to superficial layers of cat superior colliculus: a current source density analysis of evoked potentials. J. Neurophysiol. 49: 1075-1091, 1983.[Abstract/Free Full Text]
• GEORGOPOULOS, A. P. Current issues in directional motor control. Trends Neurosci. 18: 506-510, 1995.[Medline]
• GRAHAM, J., CASAGRANDE, V. A. A light microscopic and electronmicroscopic stdy of the superficial layers of the superior colliculus of the tree shrew (Tupaia glis). J. Comp. Neurol. 191: 133-151, 1980.[Medline]
• GRAY, C. M., ENGEL, A. K., KÖNIG, P., SINGER, W. Synchronization of oscillatory neuronal responses in cat striate cortex: temporal properties. Vis. Neurosci. 8: 337-347, 1992.[Medline]
• GRAY, C. M., KÖNIG, P., ENGEL, A. K., SINGER, W. Oscillatory responses in cat visual cortex exhibit inter-columnar synchronization which reflects global stimulus properties. Nature 338: 334-337, 1989.[Medline]
• GRAYBIEL, A. M., BERSON, D. M. Histochemical identification and afferent connections of subdivisions in the lateralis-posterior pulvinar complex and related nuclei in the cat. Neuroscience 5: 1175-1238, 1980.[Medline]
• HARDY, S. C., STEIN, B. E. Small lateral suprasylvian cortex lesions produce visual neglect and decreased visual activity in the superior colliculus. J. Comp. Neurol. 273: 527-542, 1988.[Medline]
• HARTING, J. K., UPDYKE, B. V., VAN LIESHOUT, D. P. Corticotectal projections in the cat: anterograde transport studies of twenty-five cortical areas. J. Comp. Neurol. 324: 379-414, 1992.[Medline]
• HEBB, D. O. In: The Organization of Behavior. New York: Wiley, 1949.
• HOFFMANN, K. P. Conduction velocity in pathways from retina to superior colliculus in the cat: a correlation with receptive field properties. J. Neurophysiol. 36: 409-424, 1973.[Free Full Text]
• HOFFMANN, K. P., STONE, J. Conduction velocity of afferents to cat visual cortex: a correlation with cortical receptive field properties. Brain Res. 32: 460-466, 1971.[Medline]
• KANASEKI, T., SPRAGUE, J. M. Pretectal nuclei and tectal laminae in the cat. J. Comp. Neurol. 158: 319-338, 1974.[Medline]
• KASPER, E. M., LARKMAN, A. U., LÜBKE, J., BLAKEMORE, C. Pyramidal neurons in layer 5 of cat visual cortex. I. Correlation among cell morphology, intrinsic electrophysiological properties, and axon targets. J. Comp. Neurol. 339: 459-474, 1994.[Medline]
• KOMATSU, Y., SHIBUKI, K., TOYAMA, K. Eye movement related activities in cells of the lateral suprasylvian cortex of the cat. Neurosci. Lett. 41: 271-276, 1983.[Medline]
• KÖNIG, P. A method for the quantification of synchrony and oscillatory properties of neuronal activity. J. Neurosci. Methods 54: 31-37, 1994.[Medline]
• KÖNIG, P., ENGEL, A. K., SINGER, W. Relation between oscillatory activity and long-range synchronization in cat visual cortex. Proc. Natl. Acad. Sci. USA 92: 290-294, 1995.[Abstract/Free Full Text]
• KREITER, A. K., SINGER, W. Stimulus-dependent synchronization of neuronal responses in the visual cortex of the awake macaque monkey. J. Neurosci. 16: 2381-2396, 1996.[Abstract/Free Full Text]
• LIVINGSTONE, M. S. Oscillatory firing and interneuronal correlations in squirrel monkey striate cortex. J. Neurophysiol. 75: 2467-2485, 1996.[Abstract/Free Full Text]
• MCILWAIN, J. T. Distributed spatial coding in the superior colliculus: a review. Vis. Neurosci. 6: 3-13, 1991.[Medline]
• NELSON, J. I., SALIN, P. A., MUNK, M.H.J., ARZI, M., BULLIER, J. Spatial and temporal coherence in cortico-cortical connections: a cross-correlation study in areas 17 and 18 in the cat. Vis. Neurosci. 9: 21-37, 1992.[Medline]
• NEUENSCHWANDER, S., SINGER, W. Long-range synchronization of oscillatory light responses in the cat retina and lateral geniculate nucleus. Nature 379: 728-733, 1996.[Medline]
• NOWAK, L. G., MUNK, M.H.J., NELSON, J. I., JAMES, A. C., BULLIER, J. The structural basis of cortical synchronization. I. Three types of interhemispheric coupling. J. Neurophysiol. 74: 2379-2400, 1995.[Abstract/Free Full Text]
• MOSCHOVAKIS, A. K. The superior colliculus and eye movement control. Curr. Opin. Neurobiol. 6: 811-816, 1996.[Medline]
• OSAGAWARA, K., MCHAFFIE, J. G., STEIN, B. E. Two visual corticotectal systems in the cat. J. Neurophysiol. 52: 1226-1245, 1984.[Abstract/Free Full Text]
• PALM, G. Cell assemblies as a guideline for brain research. Concepts Neurosci. 1: 133-137, 1990.
• PALMER, L. A., ROSENQUIST, A. C. Visual receptive fields of single striate cortical units projecting to the superior colliculus in the cat. Brain Res. 67: 27-42, 1974.[Medline]
• PAYNE, R. B. Representation of the ipsilateral visual field in the transition zone between area 17 and 18 of the cat. Vis. Neurosci. 4: 445-474, 1990.[Medline]
• RACZKOWSKI, D., ROSENQUIST, A. C. Connections of the multiple visual cortical areas with the lateral posterior-pulvinar complex and adjacent thalamic nuclei in the cat. J. Neurosci. 3: 1912-1942, 1983.[Abstract]
• ROELFSMA, P. R., ENGEL, A. K., KÖNIG, P., SINGER, W. Visuomotor integration is associated with zero time-lag synchronization among cortical areas. Nature 385: 157-161, 1997.[Medline]
• SCHNEIDER, G. E. Two visual systems. Science 163: 895-902, 1969.[Free Full Text]
• SEGAL, R. L., BECKSTEAD, R. M. The lateral suprasylvian corticotectal projection in cats. J. Comp. Neurol. 225: 259-279, 1984.[Medline]
• SINGER, W., GRAY, C. M. Visual feature integration and the temporal correlation hypothesis. Annu. Rev. Neurosci. 18: 555-586, 1995.[Medline]
• SMITH, D. C., SPEAR, P. D. Effects of superior colliculus removal on receptive-field properties of neurons in lateral suprasylvian visual area of the cat. J. Neurophysiol. 42: 57-75, 1979.[Free Full Text]
• SPRAGUE, J. M., MEIKLE, T. H. JR The role of the superior colliculus in visually guided behavior. Exp. Neurol. 11: 115-146, 1965.[Medline]
• STEIN, B. E. AND MEREDITH, A. M. Functional organisation of the superior colliculus. In: Vision and Visual Dysfunction. The Neural Basis of Visual Function, edited by A. G. Leventhal. London: Macmillan, vol. 4, 1991, p. 85-110.
• STUART, G. J. AND SAKMANN, B. Active propagation of somatic action potentials into neocortical pyramidal dendrites. Nature 367: 69-72.
• SYMONDS, L. L., ROSENQUIST, A C., EDWARDS, S. B., AND PALMER, L. A. Projections of the pulvinar-lateral posterior complex to visual cortical areas in the cat. Neuroscience 6: 1995-2020, 1981.[Medline]
• TRETTER, F., CYNADER, M., SINGER, W. Cat parastriate cortex: a primary or secondary visual area? J. Neurophysiol. 38: 1099-1113, 1975.[Abstract/Free Full Text]
• TUSA, R. J., ROSENQUIST, A. C., PALMER, L. A. Multiple cortical visual areas: visual field topography in the cat. In: Cortical Sensory Organization: Multiple Visual Areas, edited by and C. N. Woolsey Clifton, NJ: Humana, 1981, p. 1-31.
• UPDYKE, B. V. Retinotopic organization within the cat's posterior suprasylvian sulcus and gyrus. J. Comp. Neurol. 246: 265-280, 1986.[Medline]
• VAADIA, E., HAALMAN, I., ABELES, M., BERGMAN, H., PRUT, Y., SLOVIN, H., AERTSEN, A. Dynamics of neuronal interactions in monkey cortex in relation to behavioral events. Nature 373: 515-518, 1995.[Medline]
• VON DER MALSBURG, C. Am I thinking assemblies? In: Brain Theory, edited by G. Palm, and A. Aertsen Berlin: Springer, 1986, p. 161-176.
• WANG, Z., MCCORMICK, D. A. Control of firing mode of corticotectal and corticopontine layer V burst generating neurons by norepinephrine, acetylcholine, and 1S,3R-ACPD. J. Neurosci. 13: 2199-2216, 1993.[Abstract]
• WICKELGREN, B., STERLING, P. Influence of visual cortex on receptive fields in the superior colliculus of the cat. J. Neurophysiol. 32: 16-23, 1969.[Free Full Text]
• YIN, T.C.T., GREENWOOD, M. Visuomotor interactions in responses of neurons in the middle and lateral suprasylvian cortices of the behaving cat. Exp. Brain Res. 88: 15-32, 1992.[Medline]

0022-3077/98 $5.00 Copyright ©1998 The American Physiological Society