The Journal of Neurophysiology Vol. 80 No. 6 December 1998, pp. 3148-3162
Copyright ©1998 by the American Physiological Society

Large-Scale Functional Connectivity in Associative Learning: Interrelations of the Rat Auditory, Visual, and Limbic Systems

A. R. Mcintosh¹ and F. Gonzalez-Lima²

¹ Rotman Research Institute of Baycrest Centre and Department of Psychology, University of Toronto, Toronto, Ontario M6A 2E1, Canada; and ² Department of Psychology and Institute for Neuroscience, University of Texas, Austin, Texas, 78712

	ABSTRACT

Abstract Introduction Methods Results Discussion References

McIntosh, A. R. and F. Gonzalez-Lima. Large-scale functional connectivity in associative learning: interrelations of the rat auditory, visual, and limbic systems. J. Neurophysiol. 80: 3148-3162, 1998. Functional relations between specialized parts of the brain may be important determinants of learned behaviors. To study this, we examined the interrelations of the auditory system with several extraauditory structures in two groups of rats having different behavioral histories. Both groups were trained to associate a tone conditional stimulus (CS) with an aversive unconditional stimulus (US). For one group, a light presented with the tone predicted the absence of the US (group TL). In the other group, the light was a neutral stimulus (group TL⁰). Fluorodeoxyglucose (FDG) incorporation was measured in the presence of the tone-light compound. Because the tone-light compound was physically identical for both groups, neural differences between groups reflected differences in the learned associative properties of the stimuli. Covariances of FDG uptake in the auditory system and extraauditory structures were examined using partial least squares. Three strong covariance or functional connectivity patterns were identified. The first pattern mainly reflected similarities between groups, with strong interrelations between the subcortical auditory system and the thalamocortical visual system, cerebellum, deep cerebellar nuclei, and midline thalamus. This pattern of interactions may represent part of a common circuit for relaying the associative value of the tone CS to the cerebellum and the midline thalamus. The external nucleus of the inferior colliculus and medial division of the medial geniculate nucleus were associated more strongly with this pattern for group TL, which was interpreted as representing the change of the associative value of the tone by the light, mediated through extraauditory influences on these two regions. A second pattern involved midbrain auditory regions, superior colliculus, zona incerta, and subiculum and was stronger for group TL⁰. The relations between midbrain structures may represent the excitatory conditioned response (CR) evoked by the tone in this group. The final pattern was strongest in group TL and involved interrelations of the thalamocortical auditory system with hippocampus, basolateral amygdala, and hypothalamus. This pattern may represent the learned inhibition of the CR to the tone in the presence of the light. These findings are consistent with behavioral studies suggesting that at least two types of associations are formed during associative learning. One is the sensory relation of the stimuli and another is the relation between the CS and the affective components of the US. These behavioral associations are mapped to the patterns of functional connectivity between auditory and extraauditory regions.

	INTRODUCTION

Abstract Introduction Methods Results Discussion References

One approach to the study of the neural basis of learning and memory involves the exploration of how multiple brain regions interact in different learned behaviors (Gonzalez-Lima and McIntosh 1994; McIntosh and Gonzalez-Lima 1994a). Rather than focusing on particular neural structures, the approach emphasizes how the changing relationships among many regions and/or systems lead to the appropriate learned behavior (John and Schwartz 1978). It assumes that learning and memory are ubiquitous properties of neural tissue and that the involvement of an area in learning and memory depends on the specific requirements for behavioral change (Wolpaw and Lee 1989).

Our main emphasis has focused on the ability of the central auditory system to code both the physical parameters of a given stimulus and its learned behavioral significance. For example, across several studies we have shown that the determination of stimulus significance can involve auditory regions as peripheral as the cochlear nuclei (Gonzalez-Lima and McIntosh 1994). These findings agree with several electrophysiological studies in rodents (Edeline et al. 1990; Weinberger et al. 1990) and monkeys (Recanzone et al. 1992) and in neuroimaging studies of humans (Molchan et al. 1994; Schreurs et al. 1997). Furthermore, the classic division of the central auditory system into lemniscal and extralemniscal pathways (Weinberger and Diamond 1987) seems to hold for the determination of stimulus significance with the extralemniscal pathway becoming more engaged when the stimuli come to acquire some relevance.

Learning-related changes manifest not only as changes in activity but also in interactivity. Indeed, the change in the operation of parallel pathways is best appreciated by examining the influences auditory system regions have on each other. We observed such learning-related changes in the auditory system when comparing interactions elicited by presentation of a tone trained as a conditioned excitor (a conditioned stimulus that predicts the unconditioned stimulus) versus the same tone trained as a conditioned inhibitor (a conditioned stimulus that predicts the absence or withholding of the unconditioned stimulus) (McIntosh and Gonzalez-Lima 1993). Here the most striking differences in neural interactions were in the extralemniscal brain stem auditory system at the level of the dorsal cochlear nuclei, suggesting that the behavioral relevance of a stimulus impacts on the auditory system at the very earliest stages of processing.

The interactions between parallel auditory pathways also appear to change with learning. This was especially evident in the case where the behavioral relevance of an auditory stimulus depended on a visual stimulus (McIntosh and Gonzalez-Lima 1995). Two groups of rats received pairings of a tone (conditioned excitor: T⁺) with a mild footshock. Group TL was trained in a Pavlovian conditioned inhibition paradigm (T⁺/TL) where the tone-light compound signaled the absence of footshock, making the light an inhibitor (L). Group TL⁰ was trained with the tone as the excitor and the light as a "neutral" stimulus in that it predicted the absence of footshock on only 50% of trials. During fluorodeoxyglucose (FDG) incorporation, both groups were presented with the TL compound. The functional interactions in the auditory system were assessed with anatomically-based structural equation modeling using the covariances of FDG activity. Group differences in interactions between the two pathways were noted mainly at the level of the inferior colliculus (IC) and medial geniculate, possibly reflecting the unique extraauditory anatomic relation of these regions. Ascending and descending influences from the IC, particularly the extralemniscal component, were stronger for group TL. Altered converging effects on auditory cortex (AC) from the two parallel paths were noted from the ventral division of medial geniculate (MGV), which is considered a lemniscal structure, and the medial division of the medial geniculate nucleus (MGM), which is considered extralemniscal. Effects from MGM on AC were positive and from MGV were negative for group TL⁰ and reversed for group TL. These data emphasize that auditory system operations are modified by learning, even in the face of physically identical stimuli.

Because, in the aforementioned study, the relevance of the auditory stimulus depended on a visual event, it is possible that auditory system interactions with the visual system and with brain regions that receive information from more than one modality may have showed a learning-related change. The goal of the present paper was to focus on extraauditory regions to determine whether the interactions of these regions with the auditory system depended on the behavioral relevance of a stimulus event. Given that the behavioral task involved integration among auditory, visual, and somatosensory stimuli, one hypothesis is that the extraauditory regions having the anatomic capacity to integrate information from these sensory modalities should be engaged preferentially. Anatomic studies of rat neocortex have identified some possible candidates in the perilimbic cortical areas (Paperna and Malach 1991). These include perirhinal, entorhinal, insular, medial frontal, and retrosplenial cortices. All of these areas are related anatomically to two or more of the above sensory modalities. Subcortical candidates are numerous. The superior and inferior colliculi are prime candidates for mediation of auditory-visual interactions, especially considering the multimodal sensory maps in the deeper layers of the superior colliculus (Huffman and Henson 1990). Finally, it has been shown that the amygdala is an important mediator in fear conditioning (LeDoux 1993). Given the potential for auditory interactions via anatomic projections from the MGM (LeDoux et al. 1984) and cortical influences from auditory and visual cortices, there is also potential for interactions involving the amygdala.

The measurement of neural interactions in neuroimaging has proceeded using two methods. The first method seeks to explore the interactions in terms of covariance patterns between two or more brain regions. The second technique incorporates additional information, such as anatomic connections, to quantify explicitly the effect one brain region has on another. These two approaches are known as functional and effective connectivity, respectively. Both terms were introduced in the context of electrophysiological recordings from multiple cells (Aertsen et al. 1989; Gerstein et al. 1978). More recently, they have been used in reference to neuroimaging data (Friston 1994). In most of our previous examinations of neural interactions, we have focused on effective connectivity as quantified through structural equation modeling (McIntosh and Gonzalez-Lima 1991, 1994b). In the present application, we used measures of functional connectivity to assess large-scale relations of the auditory system extraauditory regions. Such an examination would be cumbersome with structural equation models because the anatomic complexity needed to link all regions would make the results intractable.

	METHODS

Abstract Introduction Methods Results Discussion References

Details of the experimental protocol have been published in our report on the interactions within the auditory system (McIntosh and Gonzalez-Lima 1995). The behavioral relevance of auditory and visual stimuli was manipulated using a Pavlovian conditioned inhibition paradigm (Domjan and Burkhard 1982; Rescorla 1975). Two groups of rats received pairings of a low-frequency FM tone (1-2 kHz, 65 dB SPL; stimulus T) with a mild footshock (tone conditioned excitor: T⁺). Conditioned inhibition (T⁺/TL) was trained in group TL (n = 6) where the tone-light compound signaled the absence of footshock, making the light the inhibitor (L, a flashing white light). Group TL⁰ (n = 7) was trained with the tone as the excitor and the light as a "neutral" stimulus (T⁺/TL⁰). A summary of the training protocol is presented in Table 1. Both groups received an identical number of tone, light, and footshock presentations. The difference between the two groups was in the arrangement of the stimulus contingencies between conditions.

Rats were trained to drink in an operant chamber before conditioning began. The day after baseline training was completed, a single pretraining session was conducted for each group. In this session, a high-frequency FM tone (10-20 kHz, 65 dB SPL; stimulus C) was trained as a weak excitor. During the pretraining session, the high-frequency tone was presented 75 times and paired with the footshock 3 times [C  unconditional stimulus (US); reinforced probability 0.04]. This made the C stimulus a weak excitor. The C stimulus was used as a comparator stimulus in the subsequent probe sessions to assess conditioned inhibition. The light (L) also was presented, but not reinforced, during pretraining to help counterbalance the total number of presentations. The pretraining was conducted in a different context than the subsequent excitatory and inhibitory conditioning to minimize the possibility of latent inhibition developing (habituation) to the light (Miller and Schachtman 1985).

For 2 days after pretraining, phase 1 training was conducted. For both groups, training sessions consisted of three paired presentations of the low-frequency tone and US (T  US, or T⁺). The tone was presented for 15 s and immediately followed by the footshock (US). The first trial began ~1 min after the rat was placed into the chamber, and the intertrial interval averaged 4 min. On the third day, phase 2 training began. For group TL, sessions consisted of two reinforced trials of the low tone (T  US) and four unreinforced trials where the low tone and the flashing light were presented simultaneously (TL compound). The order of presentation of the T  US trials and TL trials was randomized each day. For group TL⁰, three types of trials were conducted, T  US, TL, and TL  US (where the compound was reinforced). The combination of reinforced and unreinforced compound stimulus trials was used to ensure that the light did not acquire properties consistent with a strong conditioned inhibitor in group TL⁰. Training continued daily for 14 consecutive days.

After phase 2 training, probe trials were conducted to determine if subjects in both group TL and group TL⁰ showed behavior consistent with conditioned excitation and if subjects in group TL showed conditioned inhibition. This probe session consisted of two nonreinforced presentations each of the T, L, and the C stimuli, as well as two presentations of the TL and TC compounds. The TC compound was presented for the first time during the probe trials. The presentation of the C stimulus, both alone and in compound with the T stimulus, was to verify that the conditioned inhibition developed specifically to the L stimulus and that the behavioral effects observed were not simply due to the presence of another stimulus with the T stimulus during compound presentations. Conditioning was evaluated with two behavioral criteria: suppression of drinking and latency to start drinking after termination of the test stimulus (Miller and Schachtman 1985). Suppression was measured using suppression ratios (Annau and Kamin 1961) where a ratio of zero indicates complete suppression of drinking and a ratio of 0.5 indicates no suppression. Latency was measured 5 min after stimulus presentation. The expectation for group TL⁰, where no conditioned inhibition was trained, was that there would be little change in the response to the conditioned excitor (T) when presented in compound with either L or C stimuli. On the day before FDG injection, all subjects were given another session of phase 2 training.

For the FDG uptake session, animals received an intraperitoneal injection of 18 µCi/100 g body wt of [¹⁴C(U)]2-fluoro-2-deoxy-glucose (300-360 mCi/mM spec. activity; American Radiolabeled Chemicals) in 0.2 ml sterile saline and were immediately placed in the operant chamber where stimulus presentations began. All rats were presented with the TL compound during the 60-min FDG-uptake period. The TL compound was given in a 5-s on, 1-s off cycle to optimize the uptake of FDG evoked by the stimuli. To assess the neural effects of the TL compound, it was not reinforced during the FDG-uptake period. The effect of the change in conditional stimulus (CS) duration (from 15 to 5 s) was assessed previously with another group of animals, and no detectable behavioral differences were observed when CS duration was modified (McIntosh and Gonzalez-Lima 1994a). The average time to extinction of suppressive behavior during FDG uptake was 45 min (standard error = 1.57). It has been estimated that peak uptake of the tracer occurs in the first 5 min postinjection, and most of the remaining tracer is trapped by 30 min postinjection, and unincorporated FDG is cleared in the final 30 min (Gonzalez-Lima 1992; Sokoloff et al. 1977). This suggests that the majority of the resulting FDG label in the present study reflects initial responses to the tone-light compound, long before behavioral extinction is observed.

On completion of the 60-min test period, the animal was removed from the chamber and quickly decapitated. The brain was processed for autoradiography, and after development of autoradiographs, the sections were counterstained as described elsewhere (Gonzalez-Lima 1992). Quantification of FDG incorporation was performed using JAVA image analysis software (version 1.4, Jandel Scientific). Areas chosen for image analysis were based on prior findings (McIntosh and Gonzalez-Lima 1993, 1994a). Three adjacent brain sections were chosen for the analysis of each region of interest (ROI). In the present experiment, 63 ROIs were sampled (Table 2). Eleven of these were classified as auditory system regions. The ¹⁴C values from each brain area were divided by the average ¹⁴C value for all regions measured for each animal (whole-brain ratio). There was no systematic difference in the whole-brain means between the groups, validating this adjustment. All subsequent references to activity values are expressed in terms of the ¹⁴C labeling of the ROI relative to the mean of all ROIs measured.

Three observations for each structure for each subject were used to compute the interregional correlations for the analysis of functional connectivity. This was done to ensure that rank of the covariance matrix used for the analysis of functional connectivity (described in the next section) was sufficient to account for the interactions of the 11 auditory regions. Because the number of observations resulted from multiple measures from the same subject, the resulting correlations were inflated because of this extra source of variance. But, as the source of this variance was known, it was removed using a regression procedure (Pedhazur 1982). Regional activity measures were regressed on two predictor variables that contrast the within-subjects repeated measure, namely the measures of the same structure. Regressing out the between-section variance rendered the observations mathematically independent for the purposes of computing interregional correlations. Thus the resulting correlations were corrected for the extra source of variance, and the matrix to be analyzed was of sufficient rank for analysis (McIntosh and Gonzalez-Lima 1991, 1993, 1994a, 1995). Although the procedure does not affect the pattern of covariances (McIntosh and Gonzalez-Lima 1994b), it does not substitute for a larger sample size.

Partial least-squares analysis of functional connectivity

Partial least squares (PLS) is a multivariate tool that can be used to describe the relation between a one set of measures, like experimental design or behavioral measures, and a large set of dependent measures, in this case brain activity. It has been used extensively for one-dimensional images from spectrographs, as in chemometrics or remote sensing or toxicology and behavioral teratology (e.g., Heise et al. 1989; Hellberg et al. 1986; Streissguth et al. 1993). PLS recently has been adapted for neuroimaging analysis (McIntosh et al. 1996); here we describe modifications of the method to address issues of functional connectivity (McIntosh et al. 1997). The PLS procedure used here is understood most easily as an extension of a univariate interregional correlation analysis used by Horwitz et al. (1993) and has strong similiarities to the cross-hemispheric functional connectivity analysis described by Friston (1994). The specific purpose of the present analysis was to identify dominant patterns of auditory-extraauditory functional connections and determine which of these patterns were different between the two groups.

A highly idealized graphic description of the PLS procedure used to analyze interregional correlation changes is presented in Fig. 1 (a mathematical description and a small numerical example are presented in the appendix). In Fig. 1A, activity from a single ROI from the auditory system is correlated with the activity from the rest of the ROI data set (referred to here as extraauditory ROIs for simplicity) within two groups (Fig. 1A, left). This produces a correlational map of extraauditory ROIs that are correlated with the auditory ROI for each group (Fig. 1A, right). The correlation procedure usually stops here, and the maps are compared between tasks to ascertain any experimental changes in the patterns.

View larger version (45K): [in this window] [in a new window]   FIG. 1. Graphic representation of the partial least-squares analysis used to assess functional connectivity patterns. A: region of interest (ROI) from the auditory system (auditory ROI) is correlated with several extraauditory ROIs (displayed on a midsagittal rat brain schematic) within 2 groups, resulting in 1 correlation map per group. , strong negative correlation; , strong positive correlation. Correlations of the auditory ROI with a medial prefrontal region are similar between groups, whereas extraauditory ROIs at posterior neocortex and pons are different. B: correlation maps are stacked into 1 large cross-block covariance matrix and decomposed with singular value decomposition resulting in 2 latent variable (LV) pairs. Each pair consists of the weights or saliences for the auditory ROIs for each group and for extraauditory ROIs. Auditory saliences on LV1 are the same for both groups, suggesting a common pattern of correlations, or functional connections, with the salient extraauditory regionmedial prefrontal cortex. For LV2, the auditory saliences are different between groups, indicating that the salient extraauditory regions, posterior neocortex and pons, show different correlations with the auditory region between groups.

The PLS extends the correlation analysis by combining correlation maps into a single matrix referred to as the cross-block covariance (correlation) matrix, where the blocks are the auditory and extraauditory ROIs. This matrix is analyzed with singular value decomposition (SVD) providing sets of mutually orthogonal latent variable (LV) pairs. Each LV extracted through the SVD accounts for progressively less of the summed squared cross-block covariance (SSCC), a rough index of importance. A single value is calculated for each LV, which is the covariance between the auditory and extraauditory LVs in the pair, and indexes the proportion of SSCC accounted for.

The LVs can represent common correlation patterns or correlation patterns that show group differences. The first element of the LV pair contains the numerical weights for the auditory ROI, the auditory saliences, from each group and the pattern of weights indicates whether the LV represents a commonality or difference of correlations between groups. The second element of the LV pair contains the weights for the extraauditory ROIs, the extraauditory saliences, and their variation across the brain shows the pattern of ROIs that show either the commonalities or differences in their correlations with the auditory ROIs salient on the LV. For example, the first LV in Fig. 1B represents the commonalities between groups because the auditory salience is the same between groups. The auditory salience on the second LV depicts a difference between the two groups with the weight being positive for one group and negative for the other, and the pattern of extraauditory saliences locates the differences in auditory-extraauditory correlations between groups.

To summarize, each LV contains two components: saliences for the extraauditory ROIs and saliences for the auditory ROI indicating how the groups relate to the extraauditory pattern of saliencesi.e., is it a common pattern or a group difference? The PLS analysis also can be extended to include more than one auditory ROI, as in the present application.

In the present application, the 11 auditory ROIs were correlated with 52 extraauditory ROIs within each of the two groups. For a given LV, auditory saliences that were similar for both groups would indicate a common pattern of functional connectivity with the extraauditory regions salient on that LV. Conversely, if saliences across auditory regions differed between groups, this would indicate a pattern of functional connectivity with extraauditory regions that distinguished groups.

Statistical evaluation

UNIVARIATE TESTS. Regional FDG uptake was evaluated for group differences using the regression approach to the analysis of variance (Pedhazur 1992). The probability values for the regression analysis are based on the evaluation of the squared correlation of this regression and are derived from a permutation test of group assignment (Edgington 1980).

MULTIVARIATE TESTS. With 11 auditory regions for two groups, there are 22 LVs that are computed through SVD. Obviously, not all LVs that are extracted through SVD are meaningful, either statistically or theoretically. In other applications of PLS to imaging data, permutation tests have been used to assign a measure of statistical significance to the LV structure (Cabeza et al. 1997; McIntosh et al. 1996; Nyberg et al. 1996). There are two statistical questions that can be answered about the present data set using permutation tests. The first is whether the patterns represented on the LVs are on the whole significant. The second, assuming the pattern is significant, is whether there are group differences on the pattern.

The first question was addressed by evaluating the singular values from the SVD because they are the covariances of the latent variables. The computed probability represented the number of times, out of 5,000 random permutations of the auditory-extraauditory pairings across subjects, a singular value greater than or equal to the original singular value for each LV was obtained. This permutation also served to assess individual saliences, giving a threshold to decide which regions were contributing the most of the LV pattern. Because the saliences and singular values were calculated in a single mathematical step, there was no need to correct for multiple comparisons across the ROIs.

The assessment of group effects was done by evaluating the difference of auditory saliences for the two groups on an particular LV. Auditory saliences would differ the most on LVs where there was a strong group effect. Referring back to Fig. 1, the group difference would be small for LV1 because there is no group effect on the LV, but differences would be higher for LV2. Correction for multiple comparisons was afforded by a Bonferroni correction of the statistical threshold for comparison of the 11 auditory regions on each LV, which yielded a threshold of significance at P < 0.005 (0.05/11 with rounding).

THEORETICAL CONSTRAINTS. Statistical tests use a mathematical criteria to aid the investigator in trying to decide which, out of a large number of potential findings, are the most reliable. However, few statistical tests explicitly combine the mathematical criteria with theoretical or biological constraints. These additional constraints require intervention from the investigator. The merits of using theory as the final judge of statistical outcomes have been debated in several disciplines (e.g., Freedman 1987). For PLS, this issue is no less important. Because PLS is dealing with relations across a large number of dimensions, it is plausible that some of the structure identified represents statistical noise and may not be biologically relevant despite attaining statistical significance (the reverse will also be true in some instances). Because of the small sample size in the present study and the inherent redundancy of the nervous system, determining whether the saliences cluster into known anatomic or functional groupings was used to add another constraining dimension to the interpretation of the PLS results.

	RESULTS

Abstract Introduction Methods Results Discussion References

Behavior

Suppression ratios and latency measurements taken during the probe session after presentation of each of the stimuli demonstrated that the light had acquired conditioned inhibitory properties for group TL but not for group TL⁰. In group TL, the mean suppression ratio (±SE) for the T alone was 0.03 ± 0.02 and for the TL compound was 0.35 ± 0.03. The difference between these was significant [t(6) = 5.31, P < 0.01]. Further, in group TL the mean suppression ratio for the compound of the two tones (TC) was the same as for the tone alone (M = 0.03 ± 0.03), further suggesting that the light was acting as a conditioned inhibitor. The latency to drink, for group TL, after presentation of T was significantly greater compared with the latency after the TL compound (log-latency, T, M = 1.74 ± 0.15; TL, M = 1.26 ± 0.15; z = 2.52, P < 0.01; Wilcoxon matched-pairs signed-ranks test). In group TL⁰, this was not the case. The suppression ratios for the T, TL, and TC compound did not differ from each other (T, M = 0.06 ± 0.05; TL, M = 0.08 ± 0.05; TC, M = 0.00 ± 0.00). Latency measures showed a ceiling effect for all stimuli in group TL⁰ and were therefore equal.

Regional means

Mean ratio-adjusted values, along with SDs, from the 63 regions for both groups are presented in Table 1. When evaluated statistically at the conventional level of P < 0.05, differences in the auditory system FDG incorporation were observed at the dorsal and ventral divisions of the medial geniculate only, although probability for the dorsal division was exactly 0.05 evaluated over 5,000 permutations. Both areas showed lower means in group TL compared with group TL⁰. Extraauditory regions showed higher mean values for group TL in the medial frontal cortex, nucleus accumbens, and ventromedial hypothalamus and reduced mean FDG incorporation for the anterior pretectal nucleus. If the criterion for significance was to be adjusted for the number of independent comparisons (in this case 63), none of these regional differences would remain significant.

Partial least-squares analysis

Nine of the 22 LVs were significant by permutation tests at P < 0.05, and Table 3 lists the singular values for the 9 LVs with percentage and cumulative percentage SSCC accounted for. Three LVs were dominant from the PLS analysis and were considered for interpretation. Together these three accounted for 79% of the SSCC. Although the subsequent LVs were significant, only one or two auditory regions were salient within a group, and there was little obvious regionalization for the extraauditory LVs beyond the third. The functional connectivity patterns first will be described as they pertain to overall statistical reliability and spatial distribution, then in terms of auditory system involvement that was most closely linked to group differences.

LV1 (SSCC = 41%). This latent variable reflected a similar pattern of auditory-extraauditory connectivity in both groups, with a somewhat stronger pattern in group TL. For both groups, auditory regions from the cochlear nuclei to the midbrain colliculus were salient, but only in group TL was the auditory thalamic areas also salient. Figure 2 presents the saliences on LV1 for all the auditory (top) and extraauditory regions (bottom) with an indication of significance. All saliences are presented, rather than just those significant by permutations, so an appreciation for the full pattern of relations can be gained. For Fig. 2, and subsequent figures, the saliences were rescaled through multiplication by their singular value (Streissguth et al. 1993). The saliences for all auditory regions, with the exception of cortex, were negative for both groups, indicating a negative correlation with the extraauditory pattern. Fewer auditory regions met statistical significance in group TL⁰.

View larger version (39K): [in this window] [in a new window]   FIG. 2. LV1 from the analysis of functional connectivity between auditory and extraauditory regions. Top: graph contains the saliences for auditory regions for the 2 groups (legend: top right). Bottom: saliences for extraauditory regions. - - -, approximate cutoff for statistical significance of the saliences as assessed through permutation tests (P = 0.05). Direction and magnitude of most auditory saliences are similar for both groups, suggesting a common pattern of correlations with the pattern of extraauditory region saliences (bottom). Region abbreviations for areas are defined in Table 2.

Extraauditory regions showed two distinct clustering of saliences. The most prominent clustering comprised the thalamocortical components of the visual system [dorsal lateral geniculate nucleus (LGNd), ventral lateral geniculate n. (LGNv), and lateral posterior n. (LPN), medial occipital cortex 2 (OC2M), occipital cortex 1, lateral occipital cortex 2 (OC2L)] and anterior pretectal n. (APT), and parts of posterior limbic cortices [retrosplenial cortex (RS), perirhinal cortex, and entorhinal cortex (ENTO)]. Only the thalamic regions and OC2L were significant by permutation tests. The saliences for these areas were all positive, suggesting that for both groups, these areas were correlated negatively with the auditory system, though somewhat stronger in group TL. For example, the correlation of LGNd and ventral nucleus of the lateral lemniscus (LL) was 0.81 for group TL and 0.71 for group TL⁰. Another area of note showing a similar pattern of correlations was the paratenial n. in the thalamus. Negative saliences were quite focal and restricted to the deep cerebellar nuclei (DCBL), cerebellar hemisphere, and vestibular n. Correlations of these regions were strongest with DCN and VCN for both groups, though stronger for group TL. All of these areas showed positive correlations with the auditory regions.

LV2 (SSCC = 22%). Saliences for the subcortical auditory system showed opposite pattern for the two groups with only the saliences for group TL⁰ reaching statistical significance (Fig. 3). Midbrain and brain stem components of the auditory system were salient especially DCN and external nucleus of the inferior colliculus (ICE). The dominant positive saliences for LV2 were in the superficial layer of the superior colliculus (SGS), deep layer of the superior colliculus (SGM), and midbrain reticular formation (MRF). Dominant on the negative saliences were the subiculum (SUB) and zona incerta (ZI). All of these extraauditory regions were more strongly correlated with auditory regions in group TL⁰, save for SUB, which showed correlations of equal magnitude but different sign between groups.

View larger version (39K): [in this window] [in a new window]   FIG. 3. LV2 from the analysis of functional connectivity between auditory and extraauditory regions. Direction and magnitude of auditory saliences are opposite for the groups indicative of an opposite pattern of correlations with the pattern of extraauditory region saliences (bottom).

LV3 (SSCC) = 16%). The auditory system regions loading on this LV were thalamocortical (Fig. 4), and only the auditory areas from group TL were significant. Positive saliences were noted in lateral frontal cortex and lateral preoptic area/horizontal diagonal band, but these areas were not reliable by permutation tests. Negative saliences were observed in the hippocampus proper in the CA subfields, lateral hypothalamus, ventromedial hypothalamus (VMH), and basolateral amygdala, all of which were significant.

View larger version (34K): [in this window] [in a new window]   FIG. 4. LV3 from the analysis of functional connectivity between auditory and extraauditory regions. Auditory saliences are strongest for group TL, suggesting the pattern of extraauditory correlations is strongest in this group.

EXPERIMENTAL EFFECT ON LV PATTERN. In Fig. 5, the patterns of functional connectivity between the auditory and extraauditory regions are presented on rat brain schematics. The highlighted auditory regions in the figure are those where the differences in saliences between groups were significant by permutation tests.

View larger version (55K): [in this window] [in a new window]   FIG. 5. Schematic representation of the areas with the strongest contributions to the pattern of functional connectivity across the 3 latent variables. Areas highlighted in white showed positive saliences and those in black showed negative saliences. A: 2 auditory regions indicated by an asterisk (ICE and MGM) showed significant group differences in the pattern of covariance with the extraauditory regions highlighted in the schematic. B: DCN of the auditory system showed significantly higher covariance patterns with the highlighted extraauditory areas in group TL0. C: auditory cortex (AC) showed a significantly different pattern of covariances with the highlighted extraauditory areas between groups. Although the figure highlights particular regions, the collective contribution of all regions is emphasized in the analysis.

LV1 is shown in Fig. 5A where only the MGM and ICE were different between groups. For both areas, the relations of MGM and ICE were stronger in group TL, although there was also a change in sign for MGM. To help illustrate this difference, the MGM was correlated with the DCBL at 0.48 for group TL and 0.28 in group TL⁰; the ICE showed a correlation with the DCBL of 0.71 for group TL and 0.02 for group TL⁰. (We emphasize that the PLS is optimized formally to use the entire pattern of correlations. The univariate correlations serve only as an illustration of the pattern that contributed to the LV and not as a formal test statistic.)

The summary for LV2 is presented in Fig. 5B. Although the saliences for most auditory regions showed opposite signs between groups, only the DCN was statistically significant after Bonferroni correction. To aid in interpretation, in group TL, the correlation of DCN with SGM was 0.55, whereas for group TL⁰, the correlation was 0.60; the DCN-SUB correlation was 0.75 for group TL and 0.51 for group TL⁰.

For LV3, shown in Fig. 5C, auditory cortex was the only region that showed a significant difference between groups. The extraauditory areas that were salient on this pattern showed negative correlations with auditory cortex in group TL and weak positive correlations in group TL⁰. For example, the correlation of AC and VMH was 0.66 for group TL and 0.37 for group TL⁰.

	DISCUSSION

Abstract Introduction Methods Results Discussion References

Simple associative learning, such as was studied in the present paper, requires the integration of information across or within sensory modalities. We propose that this integration occurs at the level of large-scale interactions between neural systems specialized for processing the particular sensory modalities. The present study was designed to test this hypothesis and begin to describe these interacting systems. To that end, functional connections of the rat auditory system with several extraauditory regions were assessed in two groups. An important point is that for both groups the physical properties of the stimuli were identical, but the behavioral relevance differed, therefore differences in the pattern of functional connectivity may reflect the unique behavioral history of the group. In the next section, we consider the patterns of functional connectivity from the perspective of interacting systems.

Patterns of functional connectivity

For the present paper, we focused on the functional relations between the auditory system and other brain regions. Because the task involved the modification of the associative properties of an auditory stimulus, a reasonable starting point for the investigation would be the auditory system. Even within this narrow focus, we have shown a rich set of interrelations that bridge across several other neural systems (e.g., visual, cerebellar, limbic), adding to the idea that learning and memory operations engage several brain areas and that the relations between those areas are the important determinants of behavior. The three patterns of functional connectivity are orthogonal within the confines of PLS, but it is likely that within the confines of the brain the three patterns combine and interact to affect the response of the organism to the conditioned stimuli.

Because behavioral studies have been able to reliably distinguish between different processes in learning (e.g., Konorski 1967; Wagner and Brandon 1989), it seems reasonable that these processes should be served by different neural systems or by changing interactions between the same neural systems. For example, one behavioral distinction is between the sensory or perceptual relation of the CS and US, and the association of the CS and emotive or affective aspects of the US (McNish et al. 1997; Wagner and Brandon 1989). In the present paper, we interpret the patterns of functional connectivity as they may relate to the sensory-sensory and sensory-affective associative dimensions.

LV1. As an internal validation of the use of PLS for exploration of functional connectivity, the pattern observed on LV1 is reassuring. For both groups, there was a consistent relation between auditory and visual system regions, which would be anticipated given that auditory and visual stimuli were presented to both groups. Most of the auditory regions showed similar negative correlations between auditory and visual areas in the two groups. Because the involvement of these regions did not depend on the experimental manipulation, it is possible that these functional connections are present whenever visual and auditory stimuli are presented together. Whether this negative relation is a reflection of an attentional mechanism would require further experiments, but such an inverse relation of activity between auditory and visual systems has been observed in earlier FDG studies using MRF stimulation (Gonzalez-Lima 1989) and in neuroimaging studies of selective attention in humans (Haxby et al. 1994).

The involvement of auditory regions ICE and MGM in the pattern of extraauditory connectivity differed between groups. Both have extensive connections outside the auditory system (Aitken et al. 1978; LeDoux et al. 1984) that allow for extraauditory effects to be conferred to other auditory regions through the ICE and MGM. Auditory system structural equation models for the two groups showed that both areas had clear distinguishing interactions (McIntosh and Gonzalez-Lima 1995). ICE showed stronger ascending and descending influences in group TL, whereas the MGM showed a change in the sign of its interactions with AC between the two groups. Considered in light of the present results, the group differences in the interactions of these areas within the auditory system may represent the detection and transmission of the extraauditory effects that signaled the change in the tone's behavioral significance.

For the extraauditory pattern, the APT may have been recruited because of the nature of the visual stimulus. The pretectal area has been suggested to mediate brightness discrimination (Legg 1988), and the activation of this area in classical conditioning of visual stimuli has been noted (Gonzalez-Lima et al. 1987). Because the visual stimulus used was a flashing diffuse light, the involvement of the APT seems reasonable. The same logic certainly would hold for the LGN and probably the occipital cortex regions. Although they did not exceed the significance threshold, the cortical pattern of extraauditory saliences encompassed all parts of the rat visual cortex and extended into limbic cortices (RS and ENTO). The involvement of these cortical areas is consistent with other behavioral studies (e.g., Gabriel 1990; McIntosh and Gonzalez-Lima 1994a) and has strong anatomic foundation (Paperna and Malach 1991), but the present findings should be regarded as providing only tentative support given the statistical outcome.

The involvement of the DCBL nuclei in the pattern of functional connectivity is consistent with the hypothesis that this area is involved intimately in associative behaviors (Lavond et al. 1993). We have suggested previously that the DCBL may mediate the interactions between the auditory brain stem, particularly the DCN and the other forebrain regions (McIntosh and Gonzalez-Lima 1994a). DCN and DCBL were connected functionally in both groups, but there was no group difference in this association. Instead, differences were observed more rostrally in the interactions between DCBL and the ICE and MGM. The ICE has well-characterized projections to the cerebellum (Huffman and Henson 1990), and there are connections between DCBL and the medial geniculate nucleus (Carpenter 1960). The interrelations of the ICE, MGM, and DCBL may reflect the modification of the tone CS in the presence of the light in group TL. This is particularly noteworthy because all three regions receive multimodal sensory information and have the potential for cross-modal integration. The weaker functional connections between these areas in group TL⁰ may be expected because the presence of the light had no consistent bearing on the tone CS.

In summary, LV1 appears mainly to reflect a common neural system for auditory conditioning that involves DCN, DCBL, and the midline thalamus and is similar to one we presented previously (McIntosh and Gonzalez-Lima 1994a). The commonality between groups may reflect the same learned CS-US sensory association (T+), independent of the differential conditioned responding. Recruitment of additional areas (APT, OC2L, OC2M, LGNd, LGNv, LPN) likely reflects the presence of the visual stimulus. However, the involvement of ICE and MGM was stronger for group TL compared with group TL⁰. Given the unique extraauditory anatomic connections of these areas, we interpret this aspect of LV1 as representing the change of the associative value of the tone by the light, mediated through these extraauditory influences.

LV2. The pattern for LV2 showed opposite functional connectivity patterns for TL⁰ and TL and was strongest for TL⁰. The auditory system involvement was restricted to subcortical structures and was particularly strong for the DCN. Extraauditory regions engaged were in the midbrain (SGS, SGM, and MRF) and subiculum. The spatial clustering of saliences in the extraauditory pattern implies a greater representation in the midbrain extending to RN and VTA and subcortically to the subiculum. Midbrain involvement also included a cluster focused on ZI with added recruitment of subthalamic n. and lateral habenula. The addition of the superior colliculus to this pattern is interesting on two fronts. First, the SGS is a primary retinorecipient site in the rat and sends projection to the SGM. Visual, somatic, and auditory maps have been found in SGM, which receives auditory inputs from ICE (Huffman and Henson 1990). Both colliculi are connected with the MRF, which was also salient on LV2, so part of the functional connectivity may reflect the interactions between these areas that depend on the defensive CR evoked by the CS in group TL⁰. These auditory-extraauditory interrelations would be expected to differ for group TL because the presence of the light inhibited the defensive CR.

The involvement of ZI in the pattern for LV2 could be related to two dimensions of the experiment: drinking behavior and orienting movements. Several investigators using pharmacological, anatomic and behavioral studies have linked ZI to drinking behavior (e.g., Tonelli and Chiaraviglio 1995). FDG investigations into drinking behavior have shown greater FDG uptake in ZI in rats that consumed water relative to a satiated control group (Gonzalez-Lima et al. 1993). In the present study, rats were trained to drink in the operant chamber and suppression of drinking behavior was an index of conditioning. The difficulty with an interpretation of ZI involvement reflecting drinking behavior is that rats were not drinking during the FDG uptake period. The anatomic relation of ZI to the superior colliculus and somatosensory cortex leads to another possible interpretation emphasizing orienting behavior (Nicolelis et al. 1995). The opposite sign for the saliences of SGM and ZI in the LV2 pattern is interesting because the connections from ZI to the SGM are predominantly through GABAergic cells, implying a postsynaptic inhibitory influence (Kim et al. 1992). The midbrain regions identified on LV2 also have been proposed to comprise part of a mesencephalic motor system, which contributes to the locomotor component of adaptive behaviors resulting from limbic forebrain (possibly the subiculum) operations (Mogenson and Yang 1991; Mogenson et al. 1985). For group TL⁰ rats, stronger functional connections of ZI, SGM, and subiculum with auditory structures may be related to the conditioned suppression of ongoing motor behavior, given the auditory signal from ICE. For group TL, these functional connections were weak.

The reason for the recruitment of subicular cortex to this pattern, without other related limbic structures, may be anticipated from the unique involvement of this region in the limbic-motor integration proposed by Mogenson (1987). The subiculum had auditory system correlations of equal magnitude in both groups, but the sign of the correlations differed. We have observed this opposing pattern of correlations in other limbic regions when a conditioned stimulus had opposite associative meanings (McIntosh and Gonzalez-Lima 1994a). Of further interest, given the pattern of limbic connectivity in LV3 discussed below, is that chemical lesion work has noted significant differences in the behavioral deficits elicited by subicular versus hippocampal damage (Jarrard 1983). Taken together, these observations add to increasing data that demonstrate the potential for different parts of the hippocampal formation to be related to somewhat different components of learned behaviors (Rosen et al. 1992).

LV3. The final LV pattern of functional connectivity fits into several notions of the neural basis of auditory fear conditioning. Dominant on the auditory side were the thalamocortical components of the system. The three divisions of the medial geniculate nucleus (MGD, MGV, and MGM) were salient and somewhat stronger in group TL, and auditory cortex showed oppositely signed saliences between groups. Several investigators have suggested that the medial geniculate is a central region for auditory fear conditioning, especially given the anatomic relations with key extraauditory regions (LeDoux et al. 1984). Although some of the work has emphasized the MGM, other investigators have also noted functional plasticity in the MGV and MGD (Edeline and Weinberger 1991).

Functional plasticity in auditory cortex also has been well documented. Field-potential and single-cell electrical recordings in rodent auditory cortex have noted reliable learning-related changes in receptive fields (Bakin and Weinberger 1990; Weinberger and Diamond 1987). Other work has extended this to suggest that auditory cortex shows learning-related responses across different memory tasks (Sakuri 1994). While some lesion studies have suggested the auditory cortex and thalamus may have an equal role in simple excitatory conditioning (Romanski and LeDoux 1992), the present result of opposite relations of the AC with extraauditory structures between groups suggest a more prominent role for the cortex when the relevance of the auditory signal is modified by a signal from another sensory modality. Stated more generally, auditory cortex functional interactions may be particularly strong when the relevance of an auditory signal is contextually dependent, such as in discrimination learning or compound stimulus conditioning.

Extraauditory saliences were strongest for the hypothalamus, the basolateral amygdala, and the hippocampus. There is a long history implicating the relation of the hypothalamus in motivation (Olds 1962), and lateral hypothalamus activity seems to be most related to positively reinforced behaviors (Kozhedub et al. 1997; Nakamura and Ono 1986). Lesions of the amygdala disrupt the expression of conditioned fear (LeDoux 1993), and electrical recordings show changes in firing patterns that follow the acquisition of conditioned fear responses (Maren et al. 1991). The hippocampus has shown electrophysiological activityand, interestingly, interactivityrelated to the acquisition of associative behavior (Deadwyler et al. 1996). Particularly important is the speculation that the hippocampus is involved partly in the identification of contiguity between spatial and temporal events (Larochem et al. 1987). In this light, we interpret the stronger functional connections between the thalamocortical auditory system and hippocampus for group TL to reflect detection of the change of the associative value of the tone in the presence of the light. Such connections would not be there in group TL⁰ because the associative value of the tone did not differ in the presence of the light. At the same time, the affective value of the tone also would have differed in group TL, which may be carried through the functional interactions with the hypothalamus and amygdala. The change in affective value for the tone would not be expected in group TL⁰. The third LV may therefore represent the inhibition of the affective value of the tone CS when presented in compound with the light conditioned inhibitor.

Conclusion

Based on neuroanatomy, an hypothesis was put forth in the introduction proposing that the extraauditory regions having the anatomic capacity to integrate information from visual, auditory, and somatosensory modalities should show strong patterns of interactions. While this may seem obvious at some level, the examination of interregional interrelations confirmed this hypothesis and extended it by identifying three strong patterns of large-scale functional connectivity that were related to different dimensions of the learned associative behavior. LV1 reflects the common sensory associations between the two group, with the added difference in the auditory-visual association for group TL mediated by auditory regions with extensive extraauditory connections. LV2 represents the interactions among midbrain regions supporting the defensive CR and was most salient in group TL⁰. The final pattern we considered, LV3, was interpreted as the identifying neural interactions supporting the change in the associative and affective value of the tone CS in the presence of the light-conditioned inhibitor. Together, these data support the contention that learning results from the interactions among different brain regions depending on the stimuli, the process, and behavioral response.

	ACKNOWLEDGEMENTS

  The authors thank Dr. N. J. Lobaugh and two anonymous referees for valuable comments in the development of this paper.

  This work was supported by National Institute of Mental Health Grant MH-43353 and National Science Foundation Grant IBN9222075 to F. Gonzalez-Lima and partly by Natural Sciences and Engineering Council of Canada Grant OGP017034 and Medical Research Council of Canada Grant MT-13623 to A. R. McIntosh.

	APPENDIX

Mathematical description of partial least-squares analysis

For the PLS analysis of functional connectivity, the data from k groups were each partitioned into an N_k × A auditory (M_a) and a N_k × B extraauditory (M_e) matrices, where N_k is the number of observations within group k, and A and B are the number of auditory and extraauditory regions, respectively. Each data matrix is z-score transformed so that the operation yields a matrix Y_K that is an A × B nonsymmetric correlation matrix of each auditory region with each extraauditory region (superscript T represents a matrix transpose).

The correlation matrices from each group Y₁ and Y₂ then are stacked into a single matrix Y, having C = 2(A) rows and B columns. Y then is subjected to a singular value decomposition (SVD) where From the decomposition, U is an C × B matrix containing the extraauditory saliences, V is an C × C matrix of auditory saliences, and S is a diagonal matrix of the C nonzero singular values. The first A rows of matrix U are the auditory saliences for group 1 and the next A rows are the saliences for group 2.

Worked example

Consider a simple correlation matrix Y where the row for group 1 contains the correlation of an "auditory" ROI with two "extraauditory" ROIs (E and F) for that group, and the row for group 2 contains the correlation vector for group 2. The correlation of E with the auditory ROI is roughly equal between the two groups while the correlation of F differs in sign.

An SVD performed on matrix Y yields Auditory saliences (matrix U from previous section) Extraauditory saliences (matrix V) The singular values for the two latent variables are 1.20 and 0.56.

Because the auditory saliences on LV1 are similar between groups, the LV represents the similarities of correlations between auditory and extraauditory ROIs, and LV2 represents the differences between groups, in this case a difference in sign. The extraauditory saliences identify which extraauditory areas show these relations. Region E is most salient on LV1, the similarities of correlations, and region F is salient on LV2, the difference of correlations.

The PLS analysis is equivalent to a fully parameterized multivariate linear model containing terms for the grand mean and terms for the deviations from that mean. In a previous paper (McIntosh et al. 1996), it was stated that the analysis of correlation patterns such as in matrix Y also could be carried out by computing the sum and difference of rows for groups 1 and 2 followed by SVD because of the solution is invariant to orthonormal transformations. To illustrate, matrix Y was recomputed as the sum of groups 1 and 2 and difference of groups 1 and 2. Decomposition of this matrix yields Auditory saliences Extraauditory saliences (matrix V) The singular values for LV1 and LV2 are 1.700 and 0.798, respectively.

Note that the extraauditory saliences are identical to those from the analysis of the original correlation vectors. The interpretation of the LV structure would be identical for both analyses. With a small number of ROIs, one could construct the matrix Y to contain the sum (grand mean) and the difference (deviation) of correlation vectors, but this becomes rather cumbersome when several ROIs are used. The invariance of the SVD solution to orthogonal rotation obviates such preprocessing steps.

	FOOTNOTES

  Address for reprint requests: A. R. McIntosh, Rotman Research Institute of Baycrest Centre, 3560 Bathurst St., Toronto, Ontario M6A 2E1, Canada.

  Received 30 March 1998; accepted in final form 31 August 1998.

	REFERENCES

Abstract Introduction Methods Results Discussion References

• AERTSEN, A. M. H., GERSTEIN, G. L., HABIB, M. K., PALM, G. Dynamics of neuronal firing correlation: modulation of "effective connectivity." J. Neurophysiol. 61: 900-917, 1989.[Abstract/Free Full Text]
• AITKEN, L. M., DICKHASU, H., SCHULT, W., ZIMMERMANN, M. External nucleus of the inferior colliculus: auditory and spinal somatosensory afferents and their interactions. J. Neurophysiol. 412: 837-846, 1978.
• ANNAU, Z., KAMIN, L. J. The conditioned emotional response as a function of the intensity of the U.S. J. Comp. Physiol. Psychol. 54: 428-432, 1961.[Medline]
• BAKIN, J. S., WEINBERGER, N. M. Classical conditioning induces CS-specific receptive field plasticity in the auditory cortex of the guinea pig. Brain Res. 536: 271-286, 1990.[Medline]
• CABEZA, R., GRADY, C. L., NYBERG, L., MCINTOSH, A. R., TULVING, E., KAPUR, S., JENNING, J. M., HOULE, S., CRAIK, F.I.M. Age-related differences in neural activity during memory encoding and retrieval: a positron emission tomography study. J. Neurosci. 17: 391-400, 1997.[Abstract/Free Full Text]
• CARPENTER, M. B. Experimental anatomical-physiological studies of vestibular nerve and cerebellar connections. In: Neural Mechanisms of the Auditory and Vestibular Systems, edited by G. Rasmussen, and W. Windle Springfield, IL: Thomas, 1960, p. 279-323.
• CHAN-PALAY, V. In: Cerebellar Dentate Nucleus: Organization, Cytology and Transmitters. Berlin: Springer-Verlag, 1977, 548 pp.
• DAVIS, M. The role of the amygdala in fear and anxiety. Annu. Rev. Neurosci. 15: 352-375, 1992.
• DEADWYLER, S. A., BUNN, T., HAMPSON, R. E. Hippocampal ensemble activity during spatial delayed-nonmatch-to-sample performance in rats. J. Neurosci. 16: 354-72, 1996.[Abstract/Free Full Text]
• DOMJAN, M. AND BURKHARD, B. The Principles of Learning and Behavior. Belmont, CA: Brooks/Cole, 1982, 389 pp.
• EDELINE, J.-M., NEUENSCHWANDER-EL, M., DUTRIEUX, G. Frequency-specific cellular changes in the auditory system during acquisition and reversal of discriminative conditioning. Psychobiology 18: 382-393, 1990.
• EDELINE, J.-M., WEINBERGER, N. M. Thalamic short-term plasticity in the auditory system: associative retuning of receptive fields in the ventral medial geniculate body. Behav. Neurosci. 105: 618-639, 1991.[Medline]
• EDELMAN, G. M. Group selection and phasic reentrant signaling: a theory of higher brain function. In: The Mindful Brain, edited by F. O. Schmitt, G. M. Edelman, and V. B. Mountcastle Cambridge, MA: MIT Press, 1987, p. 51-100.
• EDGINGTON, E. S. Randomization Tests. New York: Marcel Dekker.
• FREEDMAN, D. A. As others see us: a case study in path analysis. J. Educ. Stat. 12: 101-128, 1987.
• FRISTON, K. J. Functional and effective connectivity: a synthesis. Human Brain Map. 2: 56-78, 1994.
• FRISTON, K., FRITH, C., FRACOWIAK, R. Time-dependent changes in effective connectivity measured with PET. Human Brain Map. 1: 69-79, 1993.
• GABRIEL, M. Functions of anterior and posterior cingulate cortex during avoidance learning in rabbits. Prog. Brain Res. 85: 467-483, 1990.[Medline]
• GERSTEIN, G. L., PERKEL, D. H., SUBRAMANIAN, K. N. Identification of functionally related neural assemblies. Brain Res. 140: 43-62, 1978.[Medline]
• GONZALEZ-LIMA, F. Functional brain circuitry related to arousal and learning in rats. In: Visuomotor Coordination: Amphibians, Comparisons, Models and Robots, edited by J.-P. Ewert, and M. A. Arbib New York: Plenum Press, 1989, p. 729-765.
• GONZALEZ-LIMA, F. Brain imaging of auditory learning functions in rats: studies with fluordeoxyglucose autoradiography and cytochrome oxidase histochemistry. In: Advances in Metabolic Mapping Techniques for Brain Imaging of Behavioral and Learning Functions, edited by F. Gonzalez-Lima, T. Finkenstädt, and H. Scheich Dordrecht: Kluwer Academic Publishers, 1992, NATO ASI Series D: vol. 68, p. 39-109.
• GONZALEZ-LIMA, F., HELMSTETTER, F. J., AGUDO Functional mapping of the rat brain during drinking behavior: a fluorodeoxyglucose study. Physiol. Behav. 54: 605-612, 1993.[Medline]
• GONZALEZ-LIMA, F., MCINTOSH, A. R. Neural network interactions related to auditory learning analyzed with structural equation modeling. Human Brain Map. 2: 23-44, 1994.
• GONZALEZ-LIMA, F., RUSSELL, I. S., GONZALEZ-LIMA, E. Localization of neural substrates for visual memory demonstrated with 2-deoxyglucose. Neurosci. Suppl. 22: s423, 1987.
• GONZALEZ-LIMA, F., SCHEICH, H. Neural substrates for tone-conditioned bradycardia demonstrated with 2-deoxyglucose. I. Activation of auditory nuclei. Behav. Brain Res. 14: 213-233, 1984.[Medline]
• GONZALEZ-LIMA, F., SCHEICH, H. Neural substrates for tone-conditioned bradycardia demonstrated with 2-deoxyglucose. II. Auditory cortex plasticity. Behav. Brain Res. 20: 281-293, 1986.[Medline]
• GONZALEZ-LIMA, F., SCHEICH, H. Classical conditioning of tone-signaled bradycardia modifies 2-deoxyglucose uptake patterns in cortex, thalamus, habenula, caudate-putamen and hippocampal formation. Brain Res. 363: 239-256, 1986b.[Medline]
• HAXBY, J. V., HORWITZ, B., UNGERLEIDER, L. G., MAISOG, J. M., PIETRINI, P., GRADY, C. L. The functional organization of human extrastriate cortex: a PET study of selective attention to faces and locations. J. Neurosci. 14: 6336-6353, 1994.[Abstract]
• HEISE, H. M., MARBACH, R., JANATSCH, G., KRUSE-JARRES, J. D. Multivariate determination of glucose in whole blood by attenuated total reflection infrared spectroscopy. Annal. Chem. 61: 2009-20015, 1989.
• HELLBERG, S., SJOSTROM, M., WOLD, S. The prediction of bradykinin potentiating potency of pentapeptides. An example of a peptide quantitative structure-activity relationship. Acta Chem. Scand. [B] 40: 135-140, 1986.[Medline]
• HEBB, D. O. Drives and the C. N. S. (conceptual nervous system). Psychol. Rev. 62: 243-254, 1955.[Medline]
• HORWITZ, B. Simulating functional interactions in the brain: a model for examining correlations between regional cerebral metabolic rates. Int. J. Biomed. Comput. 26: 149-170, 1990.[Medline]
• HORWITZ, B., MAISOG, J., KIRSCHNER, P., MENTIS, M., FRISTON, K., MCINTOSH, A. R. A computerized system for determining pixel-by-pixel correlations of functional activity measured by positron emission tomography (PET). Soc. Neurosci. Abstr. 19: 1604, 1993.
• HORWITZ, B., SONCRANT, T. T., HAXBY, J. V. Covariance analysis of functional interactions in the brain using metabolic and blood flow data. In: Advances in Metabolic Mapping Techniques for Brain Imaging of Behavioral and Learning Functions, edited by F. Gonzalez-Lima, T. Finkenstädt, and H. Scheich Dordrecht: Kluwer Academic Publishers, 1992, NATO ASI Series D: vol. 68, p. 189-217.
• HUFFMAN, R. F., HENSON, O. W. The descending auditory pathway and acousticomotor systems: connections with the inferior colliculus. Brain Res. Rev. 15: 295-323, 1990.[Medline]
• HULL, C. L. In: Principles of Behavior. New York: Appleton-Century-Crofts, 1943, 422 pp.
• JARRARD, L. E. Selective hippocampal lesions and behavior: effects of kanic acid lesions on performance of place and cue tasks. Behav. Neurosci. 97: 873-889, 1983.[Medline]
• JOHN, E. R. High nervous system functions: brain functions and learning. Annu. Rev. Physiol. 23: 451-484, 1961.[Medline]
• JOHN, E. R. AND SCHWARTZ, E. L. The neurophysiology of information processing and cognition. Annu. Rev. Psychol. 29: 1n29, 1978.
• KIM, U., GREGORY, E., HALL, W. C. Pathway from the zona incerta to the superior colliculus in the rat. J. Comp. Neurol. 321: 555-575, 1992.[Medline]
• KONORSKI, J. In: Integrative Activity of the Brain. Chicago: University of Chicago Press, 1967, 531 pp.
• KOZHEDUB, R. G., ZAICHENKO, M. I., RAIGORODSKII, YU V., PAVLIK, V. D., YAKUPOVA, L. P., AND MIKHAILOVA, N. G. Features of the coordinated activity functionally identified neurons in the hypothalamus in different motivational-emotional states. Neurosci. Behav. Physiol. 27: 137-144, 1997.[Medline]
• LAROCHEM, S., NEUENSCHWANDER-EL, M., EDELINE, J. M., DUTRIEUX Hippocampal associative cellular responses: dissociation with behavioral responses revealed by a transfer-of-control technique. Behav. Neural. Biol. 47: 356-368, 1987.[Medline]
• LASHLEY, K. S. Integrative functions of the cerebral cortex. Physiol. Rev. 13: 1-42, 1933.[Free Full Text]
• LAVOND, D. G., KIM, J. J., THOMPSON, R. F. Mammalian brain substrates of aversive classical conditioning. Annu. Rev. Psychol. 44: 317-342, 1993.[Medline]
• LEDOUX, J. E. Emotional memory systems in the brain. Behav. Brain Res. 58: 69-79, 1993.[Medline]
• LEDOUX, J. E., SAKAGUCHI, A., REIS, D. J. Subcortical efferent projections of the medial geniculate nucleus mediate emotional responses to acoustic stimuli. J. Neurosci. 4: 683-698, 1984.[Abstract]
• LEGG, C. R. The pretectum and visual discrimination in the rat. Behav. Brain Res. 29: 27-34, 1988.[Medline]
• MAREN, S., POREMBA, A., GABRIEL, M. Basolateral amygdala multi-unit neuronal correlates of discriminative avoidance learning in rabbits. Brain Res. 549: 311-316, 1991.[Medline]
• MCINTOSH, A. R., BOOKSTEIN, F. L., HAXBY, J. V., GRADY, C. L. Spatial pattern analysis of functional brain images using partial least squares. Neuroimage 3: 143-157, 1996.[Medline]
• MCINTOSH, A. R., GONZALEZ-LIMA, F. Structural modeling of functional neural pathways mapped with 2-deoxyglucose: effects of acoustic startle habituation on the auditory system. Brain Res. 547: 295-302, 1991.[Medline]
• MCINTOSH, A. R., GONZALEZ-LIMA, F. Network analysis of functional auditory pathways mapped with fluorodeoxyglucose: associative effects of a tone conditioned as a Pavlovian excitor or inhibitor. Brain Res. 627: 129-140, 1993.[Medline]
• MCINTOSH, A. R., GONZALEZ-LIMA, F. Network interactions among libric cortices, basal forebrain and cerebellum differentiate a tone conditioned as a pavlovian excitor or inhibitor: fluorodeoxyglucose mapping and covariance structural modeling. J. Neurophysiol. 72: 1717-1733, 1994a.[Abstract/Free Full Text]
• MCINTOSH, A. R., GONZALEZ-LIMA, F. Structural equation modeling and its application to network analysis of functional brain imaging. Human Brain Map. 2: 2-22, 1994b.
• MCINTOSH, A. R., GONZALEZ-LIMA, F. Functional network interactions between parallel auditory pathways during Pavolvian conditioned inhibition. Brain Res. 683: 228-241, 1995.[Medline]
• MCINTOSH, A. R., NYBERG, L., BOOKSTEIN, F. L., TULVING, E. T. Differential functional connectivity of prefrontal and medial temporal cortices during episodic memory retrieval. Human Brain Map. 5: 323-327, 1997.
• MCNISH, K. A., BETTS, S. L., BRANDON, S. E., WAGNER, A. R. Divergence of conditioned eyeblink and conditioned fear in backward Pavlovian training. Anim. Learn. Behav. 25: 43-52, 1997.
• MESULAM, M. M. Large-scale neurocognitive networks and distributed processing for attention, language and memory. Ann. Neurol. 28: 597-613, 1990.[Medline]
• MILLER, R. M., SCHACHTMAN, T. R. Conditioning context as an associative baseline: implictions for response generation and nature of conditioned inhibition. In: Information Processing in Animals: Conditioned Inhibition, edited by R. R. Miller, and N. E. Spear Hillsdale, NJ: Erlbaum Associates, 1985, p. 51-88.
• MOGENSON, G. J. Limbic-motor integration. Prog. Psychophysiol. Physiol. Psychol. 12: 117-170, 1987.
• MOGENSON, G. J., SWANSON, L. W., WU, M. Evidence that projections from substantia innominata to zona incerta and mesencephalic locomotor region contribute to locomotor activity. Brain Res. 334: 65-76, 1985.[Medline]
• MOGENSON, G. J., YANG, C. R. The contribution of basal forebrain to limbic-motor integration and the mediation of motivation to action. Adv. Exp. Med. Biol. 295: 267-290, 1991.[Medline]
• MOLCHAN, S. E., SUNDERLAND, T., MCINTOSH, A. R., HERSCOVITCH, P., SCHREURS, B. G. A functional anatomical study of associative learning in humans. Proc. Nat. Acad. Sci. USA 91: 8122-8126, 1994.[Abstract/Free Full Text]
• NAKAMURA, K., ONO, T. Lateral hypothalamus neuron involvement in integration of natural and artificial rewards and cue signals. J. Neurophysiol. 55: 163-181, 1986.[Abstract/Free Full Text]
• NICOLELIS, M. A., CHAPIN, J. K., LIN, R. C. Development of direct GABAergic projections from the zona incerta to the somatosensory cortex of the rat. Neuroscience 65: 609-631, 1995.[Medline]
• NYBERG, L., MCINTOSH, A. R., CABEZA, R. E., HABIB, R., HOULE, S., TULVING, E. General and specific brain regions involved in encoding and retrieval of events: what, where and when. Proc. Nat. Acad. Sci. USA 93: 11280-11285, 1996.[Abstract/Free Full Text]
• OLD, J. Hypothalamic substrates of reward. Physiol. Rev. 42: 554-604, 1962.[Free Full Text]
• PAPERNA, T., MALACH, R. Patterns of sensory intermodality relationships in the cerebral cortex of the rat. J. Comp. Neurol. 308: 432-456, 1991.[Medline]
• PAXINOS, G., WATSON, C. In: The Rat Brain in Stereotaxic Coordinates, 2nd ed.. New York: Academic Press, 1986.
• PEDHAZUR, E. J. In: Multiple Regression in Behavioral Research: Explanation and Prediction, 2nd ed.. New York: Holt, Reinhart and Winston, 1982, 822 pp.
• RECANZONE, G. H., SCHREINER, C. E., MERZENICH, M. M. Plasticity in the frequency representation of primary auditory cortex following discrimination training in adult owl monkeys. J. Neurosci. 13: 87-103, 1992.[Abstract]
• RESCORLA, R. A. Pavolvian excitatory and inhibitory conditioning. In: Handbook of Learning and Cognitive Processes: Conditioning and Behavior Theory, edited by and W. K. Estes Hillsdale, NJ: Erlbaum Associates, 1975, vol. 2, p. 7-35.
• ROMANSKI, L. M., LEDOUX, J. E. Equipotentiality of thalamo-amygdala and thalamo-cortico-amygdala circuits in auditory fear conditioning. J. Neurosci. 12: 4501-4509, 1992.[Abstract]
• ROSEN, J. B., HITCHCOCK, J. M., MISERENDINO, M.J.D., FALLS, W. A., CAMPEAU, S., DAVIS, M. Lesions of perirhinal cortex but not frontal, medial prefrontal, visual, or insular cortex block fear-potentiated startle using a visual conditioned stimulus. J. Neurosci. 12: 4624-4633, 1992.[Abstract]
• SAKURI, Y. Involvement of auditory cortical and hippocampal neurons in auditory working memory and reference memory in the rat. J. Neurosci. 14: 2602-2623, 1994.
• SCHREURS, B. G., MCINTOSH, A. R., BAHRO, M., HERSCOVITCH, P., SUNDERLAND, T., MOLCHAN, S. E. Lateralization and behavioral correlation of changes in regional cerebral blood flow with classical conditioning of the human eyeblink response. J. Neurophysiol. 77: 2153-2163, 1997.[Abstract/Free Full Text]
• SOKOLOFF, L., REIVICH, M., KENNEDY, C., DES ROSIERS, M. H., PATLAK, C. S., PETTIGREW, K. D., SAKURADA, O., SHINOHARA, M. The [14-C]deoxyglucose. method for the measurement of local cerebral glucose utilization: theory, procedure and normal values in the conscious and anesthetized albino rat. J. Neurochem. 28: 897-916, 1977.[Medline]
• STREISSGUTH, P., BOOKSTEIN, F. L., SAMPSON, P. D., BARR, H. M. In: The Enduring Effects of Prenatal Alcohol Exposure on Child Development: Birth Through 7 Years, a Partial Least Squares Solution. Ann Arbor, MI: Univ. of Michigan Press, 1993.
• TONELLI, L., CHIARAVIGLIO, E. Dopaminergic neurons in the zona incerta modulates ingestive behavior in rats. Physiol. Behav. 58: 725-729, 1995.[Medline]
• VAUDANO, E., LEGG, C. R. Cerebellar connections of the ventral lateral geniculate nucleus in the rat. Anat. Embryol. 186: 583-588, 1992.[Medline]
• WANG, X. F., WOODY, C. F., CHIZHEVSKY, V., GRUEN, E., LANDEIRA-FERNANDEZ, J. The dentate nucleus is a short-latency relay of a primary auditory transmission pathway. Neuroreport 2: 361-364, 1991.[Medline]
• WAGNER, A. R., BRANDON, S. E. Evolution of a structured connectionist model of Pavlovian conditioning (AESOP). In: Contemporary Learning Theories: Pavlovian Conditioning and the Status of Learning Theory, edited by S. B. Klein, and R. R. Mower Hillsdale, NJ: Erlbaum, 1989, p. 149-189.
• WEINBERG, R. J., RUSTIONI, A. A cuncecochlear pathway in the rat. Neuroscience 20: 209-219, 1987.[Medline]
• WEINBERGER, N. M., ASHE, J. H., METHERATE, R., MCKENNA, T. M., DIAMOND, D. M., BAKIN, J. Returning of auditory cortex by learning: a preliminary model of receptive field plasticity. Concepts Neurosci. 1: 91-132, 1990.
• WEINBERGER, N. M., DIAMOND, D. M. Physiological plasticity in auditory cortex: rapid induction by learning. Prog. Neurobiol. 29: 1-55, 1987.[Medline]
• WELSH, J. P., HARVEY, J. A. Cerebellar lesions and the nictitating membrane reflex: performance deficits of the conditioned and unconditioned response. J. Neurosci. 9: 299-311, 1989.[Abstract]
• WOLPAW, J. R., LEE, C. L. Memory traces in primate spinal cord produced by operant conditioning of H-reflex. J. Neurophysiol. 61: 563-572, 1989.[Abstract/Free Full Text]
• ZOLA-MORGAN, S., SQUIRE, L. R., AMARAL, D. G., SUZUKI, W. A. Lesions of perirhinal and parahippocampal cortex that spare the amygdala and hippocampal formation produce severe memory impairment. J. Neurosci. 9: 4355-4370, 1989.[Abstract]

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