| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
School of Psychology, University of St. Andrews, Fife, Scotland, United Kingdom
Submitted 23 December 2004; accepted in final form 2 March 2005
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Ikemoto and Panksepp 1999; Parkinson et al. 2000; Robbins and Everitt 1996, 2002; Salamone and Correa 2002; Wise 1982). Findings from neurophysiological studies seem consistent with this "outcome-prediction " hypothesis, with demonstrations of single-nucleus accumbens neuronal responses to outcome delivery and to a diverse range of actions and stimuli that predict an upcoming outcome (Bowman et al. 1996; Carelli and Ijames 2001; Carelli et al. 2000; Chang et al. 1998; Cromwell and Schultz 2003; Hassani et al. 2001; Hollerman et al. 1998; Nicola et al. 2004a; Setlow et al. 2003; Shidara et al. 1998; Wilson and Bowman 2004).
Alternatively, it has been hypothesized that nucleus accumbens neurons respond to salient stimuli, irrespective of their outcome-predictive properties, to facilitate a subsequent behavioral switch by the organism (Bakshi and Kelley 1991a,b; Cools 1980; Evenden and Carli 1985; Evenden and Robbins 1983a,b; Horvitz 2002; Oades 1985; Reading and Dunnett 1991; Reading et al. 1991; Redgrave et al. 1999a,b; Robbins and Koob 1980; Robbins and Sahakian 1983; van den Bos and Cools 2003). Thus it is possible that single-nucleus accumbens neural responses to outcome or outcome-predictive stimuli are in fact responses to salient stimuli to cause a switch in the organism's subsequent behavioral sequence.
We aimed to test whether nucleus accumbens neurons respond to process outcome prediction and/or a switch in the rat's subsequent behavior by recording the activity from single neurons in the nucleus accumbens of thirsty rats responding within a modified go/no-go task. Rats were trained to make responses (bar-presses or spigot-licks) that were followed by the presentation of conditioned stimuli signaling the availability of either a rewarding (sweet liquid) or aversive outcome (bitter liquid). Rats could either make a "go " response to trigger the outcome delivery or they could withhold their response to avoid outcome delivery. Depending on the initial response (pressing vs. licking) and the subsequent "go/no-go " response, there was either a switch or "no-switch " in the rat's behavior after reward/aversive-predictive conditioned stimuli. This design allowed us to answer the following questions: 1) Do neurons respond primarily in anticipation of the upcoming outcome, to the switching of the animal's behavior irrespective of the upcoming outcome, or to a combination of outcome-prediction and switching information? 2) Is the valence of neural response differential between upcoming outcome types and/or to the presence versus absence of a behavioral switch?
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Eleven Listar Hooded adult male rats (Harlan UK), weighing 411 g (±59 g, 95% CI) when training began, were housed in quadruplets on a 12-h light/12-h dark light cycle. During experimental procedures rats were placed on a regime of restricted water access with free access to water available from 4:00 to 5:00 PM each weekday and from Friday 4:00 PM until Sunday afternoon. The rats' body weights were not allowed to dip below 85% of their free-drinking weight. After surgery rats were housed singly. The "Handbook of Laboratory Animal Management and Welfare" (Wolfensohn and Lloyd 1998) was followed and all procedures conformed to the United Kingdom 1986 Animals (Scientific Procedures) Act.
Apparatus
BEHAVIOR. Rats were trained in sound-attenuated testing chambers (34 x 29 x 25 cm; Med Associates, St. Albans, VT) fitted with video cameras (Santec smart vision, model VCA 5156, Sanyo Video Vertrieb GmbH, Ahrensburg, Germany). Located on the left wall of each chamber were a retractable lever (left side), drinking spigot (center), houselight (top center), and piezoelectric buzzer (behind spigot; model EW-223A, Med Associates). Liquids were delivered through the drinking spigot at a rate of 0.05 ml/s by 2 computer-controlled syringe pumps (model PHM-100, Med Associates) through 50-ml glass syringes (Rocket, London, UK) with stainless steel plungers to ensure repeatable flow rates. One of these syringes dispensed 0.25% wt/vol sodium saccharin solution whereas the other dispensed 0.2% wt/vol quinine hydrochloride solution. These solutions were delivered through separate lines of Teflon tubing to avoid mixing. Solutions were delivered at precise times with reliable flow rates because the stiff syringes, plungers, and tubing prevented pressure waves produced by the pumps from being attenuated.
NEUROPHYSIOLOGY.
Electrode arrays containing a movable bundle of four 50-µm stainless steel microwires coated in Teflon (tip impedance 0.5–1.5 M
) were used. Differential activity from 2 pairs of wires was amplified, filtered, and then processed by a CED 1401 data-acquisition system (Cambridge Electronic Design, Cambridge, UK). Although the rate at which data were sampled on the CED system was 20 kHz, the resolution for time-stamping behavioral events was limited to that of the MED-PC system (2 ms). All neurophysiology apparatus and surgical, histological, and spike-sorting techniques were identical to those described previously in Wilson and Bowman (2004).
Procedures
During the development of the behavioral task, the first group of rats (n = 7) received procedures that were slightly different from those described below, that is, initially lower concentrations of quinine, different lengths of time-outs, and different amounts of training per stage. However, the final testing stage was identical between the 2 groups. Rats were advanced to each stage in the training when response levels reached an asymptote. In cases in which responding ceased at a given stage, rats were either moved back to an earlier stage for retraining or advanced to the subsequent stage when appropriate.
TRAINING STAGE 1: LICKING RESPONSES TO EARN REWARDING AND AVERSIVE OUTCOMES.
A) Rats were initially trained for a single 30-min session using the following procedure: when the animal first licked the drinking spigot there was a variable delay of 0.1, 0.2, 0.4, or 0.8 s. Rats were subsequently presented with the reward-predictive–conditioned stimulus, lasting 0.5 s. The rats were divided into 2 groups of conditioned-stimulus modality with one group (n = 7) receiving a tone reward-predictive–conditioned stimulus using the piezoelectric buzzer, and the other group (n = 4) receiving a light reward-predictive–conditioned stimulus using the houselight. After the first presentation of the conditioned stimulus (conditioned stimulus-1) there was a 1-s delay, followed by a second presentation of the same reward-predictive–conditioned stimulus (conditioned stimulus-2). The double presentation of the same conditioned stimulus gave the rats a time window to prepare their subsequent "go/no-go " response. If a lick was made within the next 2 s the animal received 0.1 ml saccharin solution reward lasting 2 s, along with the continued presentation of conditioned stimulus-2. At the end of the lick bout (defined as an interlick interval 
300 ms) there was a period of 4 s, unsignaled to the rat, before the beginning of the next trial. If no lick was made within 2 s of conditioned stimulus-2 then an error was recorded, the conditioned stimulus turned off, and there was time-out period of 2.5 s, unsignaled to the rat, before starting the next trial.
B) In a second 30-min session, the rats were required to lick on a variable ratio-3 schedule (random selection of 1–5 licks per trial) to initiate the processes outlined above for obtaining saccharin reward.
C) Rats were then trained over 2–5 daily 30-min sessions, as outlined above except the reward-predictive–conditioned stimulus was replaced by an aversive-predictive–conditioned stimulus (a tone or a light, but not the same as the reward-predictive stimulus) on 1 out of 3 trials (pseudo-randomly determined on a trial-to-trial basis). On these trials, aversive quinine solution (0.1 ml) was delivered as the outcome.
TRAINING STAGE 2: PRESSING RESPONSES TO EARN REWARDING AND AVERSIVE OUTCOMES. A) The next stage replaced the operant licking responses with bar-pressing responses. Over 7 daily 30-min sessions rats learned to perform one bar-press to receive reward-predictive conditioned stimuli (no aversive trials at this stage) and then reward (initially the 2-s time window that allowed the rat to lick for reward after the onset of the conditioned stimulus-2, was increased to 10 s).
B) One 30-min session was then given where rats were required to press on a variable ratio-3 schedule on each trial instead of a single press.
C) Finally, during two 30-min sessions rats pressed on a variable ratio-3 schedule for either saccharin or quinine outcomes (see training stage 1C).
TRAINING STAGE 3: PRESSING AND LICKING RESPONSES TO EARN REWARDING AND AVERSIVE OUTCOMES. A) Rats were trained for approximately 3 wk to press or lick on a variable ratio-3 schedule for saccharin or quinine outcomes. The operant response required by the rat was the same for 10 consecutive trials, which constituted a block of trials. The first block of trials was randomly assigned as pressing or licking. The block type was then sequentially alternated and signaled to the rat by protrusion or withdrawal of the bar, respectively. Figure 1 illustrates this final testing stage of the modified go/no-go task.
|
After behavioral training, an electrode array was permanently implanted onto the skull of each rat with the electrode targeted stereotaxically at the nucleus accumbens (+1.7 mm anterior and +1.5 mm lateral from bregma; –6.0 mm ventral to skull surface).
Neurophysiological recording
Rats were given 5–7 days to recover from surgery. We then recorded successfully from 8 rats while they behaved during the modified go/no-go task. Neural recording lasted about 4 wk.
Histology
After neurophysiological recording rats were killed by overdose with 0.7 ml Dolethal (200 mg/l pentobarbitone sodium BP; Univet, Oxford, UK) and perfused intracardially with 0.1% phosphate buffer saline followed by a fixative (4% paraformaldehyde in 0.1 M phosphate buffer). The paths of electrode tracts were mapped onto standardized sections of the brain (Paxinos and Watson 1997).
Data analysis
BEHAVIOR. We restricted our behavioral analysis to the testing sessions within which activity from classified nucleus accumbens neurons was sampled (n = 41 sessions). We independently performed repeated-measures ANOVA on the average percentage of trials within which rats made "go " responses, and the average lick rates during outcome delivery, respectively, over 2 within-subject factors, Operant response type (licking vs. pressing) and Outcome type (rewarding vs. aversive).
NEUROPHYSIOLOGY.
Spike sorting.
Spikes were resorted off-line in Spike2 by performing principal components analysis on every waveform in the data set. When identical neurons were recorded over consecutive testing days (as identified by visual inspection of the waveform shape/duration, interspike interval histogram, average firing rate, and event-related activity) we used data only from the session within which the rat responded maximally.
Windows for spike counts.
Using Spike2, we constructed histograms, rasters, and spike counts for each neuron and for the average response of the population of neurons. These were generated relative to time windows around both conditioned stimulus-1 and -2, the quinine and saccharin delivery, and the lick and press onsets. There appeared to be a consistent pattern of neural responses that occurred phasically after presentations of conditioned stimulus-1 and -2, as well as throughout reward delivery. We restricted our analysis to quantify responses only to the presentation of conditioned stimulus-1, and to the aversive and rewarding outcomes, because neurons showing visible responses to the conditioned stimulus-2 also showed responses at conditioned stimulus-1, which were usually of greater magnitude. We calculated the average firing rate (Hz) of each neuron within 3 time windows after the onset of conditioned stimulus-1 [0–100 ("baseline"), 100–200, and 200–300 ms (to capture any late responses)]. We also compared the average firing rate around a baseline window around the trial onset (–1 to 1 s trial onset) to firing during the 2 s of outcome delivery.
Assignment of trial types.
We were able to analyze the effect of outcome-prediction on neural responses because about 66% of trials per session had conditioned stimuli predicting a rewarding outcome ("CS1 sweet, " Fig. 1) and about 33% conditioned stimuli predicting an aversive outcome ("CS1 bitter, " Fig. 1). Additionally, we were able to examine the effects of the neural responses on the rat's subsequent switching behavior because in each trial the rats made either a subsequent behavioral switch or no-switch (see Fig. 2). Trials were defined as containing a behavioral switch either when the rat switched from operant bar-pressing to spigot-licking after conditioned stimulus presentation (Fig. 2A) or when there was a switch from operant spigot-licking to avoidance of spigot-licking (Fig. 2B). Conversely, trials were defined as containing no behavioral switch when operant spigot-licking was continued throughout conditioned stimulus presentation. (Fig. 2C). It should be noted that one trial type was excluded from analysis (Fig. 2D) because it was unclear whether the rat made a behavioral switch.
|
0.05) and a significant pairwise comparison (P 0.05) between 2 of the 3 epoch time windows (0–100 vs. 100–200 ms, 0–100 vs. 200–300 ms, 100–200 vs. 200–300 ms) after conditioned stimuli predictive of the rewarding ("CS1 sweet," Fig. 1) or aversive ("CS1 bitter, " Fig. 1) outcome. Neurons were classified as exhibiting a switching response (switch or no switch) to the conditioned stimulus when there was a significant Epoch x Switching type interaction (P 0.05) and a significant pairwise comparison (P 0.05) between 2 of the 3 epoch time windows after conditioned stimuli that caused the rat to switch (Fig. 2, A and B) or "not-switch " (Fig. 2C). Neurons were classified as exhibiting an outcome-switching response (reward-switch, reward-no switch, aversive-switch, aversive-no switch) to the conditioned stimulus when there was a significant Epoch x Outcome type x Switching type interaction (P 0.05) and a significant pairwise comparison (P 0.05) between 2 of the 3 epoch time windows after conditioned stimuli that caused the rat to switch or "not-switch " and were predictive of one type of outcome. It was possible that a neuron could satisfy more than one of these criteria and be classified as having more than one response. When repeated-measures ANOVA was performed, the Hunyh–Feldt correction was used to decrease the effect of heterogeneity of variance. When multiple pairwise comparisons were made, the Sidak test was performed to adjust for multiple comparisons.
Mixed-design repeated-measures ANOVAs with pairwise comparisons were also performed on spike frequency (Hz) across each trial per neuron over a baseline (–1 to +1 s trial onset) and outcome delivery time windows (repeated-measures factor, Epoch) comparing aversive versus rewarding outcomes (between-group factor Outcome type). This baseline time window was used to allow us to keep the baseline and response time windows of equal length. Neurons were classified as exhibiting an outcome response if there was a significant Epoch x Outcome type interaction effect (P 0.05) and a significant pairwise comparison between aversive and rewarding (P 0.05) outcome types. Details of additional analyses are presented in the appropriate figure legends and were performed using Microsoft Excel 2000 and SPSS 10.0 for Windows. Rasters and histograms were presented using Spike 2.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
As shown in Fig. 3, rats found saccharin delivery rewarding, as indicated by the high lick rates, and quinine delivery aversive, as indicated by the very low lick rates [it has also been demonstrated previously that oral injections of quinine in rats causes aversive facial reactions (Grill and Norgren 1978)]. We wanted rats to learn the associations between conditioned stimuli and the upcoming outcome. This appears to have been the case because rats changed their behavioral response after presentation of conditioned stimuli to avoid quinine and consume saccharin (see Fig. 4). Finally, we sought to identify different trial types on the basis of switching/not-switching of the rat's behavioral response after conditioned stimulus presentation. Analysis of the population licking responses indicates that there was continued licking from pre- to postconditioned stimulus presentation in "no-switch " trials (see Fig. 5A) and a switch from licking to not-licking (see Fig. 5B) or pressing to licking (see Fig. 5C) in "switch " trials.
|
|
|
HISTOLOGY. From the 8 subjects that were successfully tested neurophysiologically, 7 had electrode tracks within the nucleus accumbens that included portions of both the core and the shell over a large anterior–posterior range (see Fig. 6). However, because we could not determine the location of each recorded neuron, we did not perform separate analyses on core versus shell nucleus accumbens neurons.
|
|
|
|
|
|
Additionally, we compared the magnitude of outcome-prediction and behavioral switching effects. We found that across all responsive neurons (n = 28) the magnitude of the outcome-prediction effect [median eta2 (
2) = 0.051] was about 2.5 times that of the effect of behavioral switching (median 2 = 0.020), and about 3.5 times that of an interaction effect between outcome-prediction and behavioral switching (median 2 = 0.014).
MOST NEURAL RESPONSES WERE TO REWARD-RELATED RATHER THAN AVERSIVE-RELATED STIMULI. We next wanted to establish whether there was a bias in the type of outcome predicted by nucleus accumbens neural responses because the behavioral switching hypothesis suggests that reward-related and aversive-related stimuli would evoke equivalent activity. As shown in Table 1 the majority of outcome-predicting responses (19/24) and outcome-switch responses (8/11) were in anticipation of rewarding rather than aversive outcomes. All reward outcome-predicting responses were excitatory (although one reward-no switch response was inhibitory), and were usually short (<200 ms) phasic bursts (see example in Fig. 8). In contrast, aversive outcome-predicting responses were excitatory (2/5) or inhibitory (3/5; see Fig. 10 for an example response pattern). Typically, when neurons responded to aversive-predictive–conditioned stimuli they also responded to the reward-predictive–conditioned stimuli (4/5 neurons). In most of these cases (3/4) the neurons responded with differential valence to the reward-predictive (excitation) versus aversive-predictive (inhibition) conditioned stimuli (see Fig. 10). When we considered the average response of all neurons (n = 82; see Fig. 11) we found that the response was largely modulated by the reward-predictive properties of conditioned stimuli. However, in switch versus no-switch conditions the excitatory response to the reward-predictive–conditioned stimulus was enhanced and the response to the aversive-predictive–conditioned stimulus reversed in sign.
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
All neural responses to conditioned stimuli encoded outcome-predicting information
Our data are consistent with many previous reports that single neurons in the nucleus accumbens respond to conditioned stimuli predictive of positive or negative outcomes, as well as to the outcomes themselves (Bowman et al. 1996; Carelli et al. 2000; Chang et al. 1998; Cromwell and Schultz 2003; Hassani et al. 2001; Hollerman et al. 1998; Nicola et al. 2004a; Peoples and West 1996; Setlow et al. 2003; Shidara et al. 1998; Tremblay et al. 1998; Wilson and Bowman 2004). Moreover, our data replicate previous reports of a bias in the valence of responses to reward-predictive–conditioned stimuli toward excitations versus inhibitions (Carelli and Ijames 2001; Hollerman et al. 1998; Nicola et al. 2004a; Wilson and Bowman 2004), a bias in the valence of responses to rewarding outcome delivery toward inhibition versus excitation (Chang et al. 1998; Nicola et al. 2004b; Peoples and West 1996; Taha and Fields 2005; Wilson and Bowman 2004), and a bias toward the proportion of neurons responding to reward-predictive versus aversive-predictive–conditioned stimuli (Williams et al. 1993).
However, it has recently been reported that nucleus accumbens neurons responded more to aversive-predictive than reward-predictive–conditioned stimuli in rats performing a go/no-go task for sucrose/quinine, and in cats within a nonoperant environment (Setlow et al. 2003; Yanagimoto and Maeda 2003). This suggests that nucleus accumbens neurons are not inherently biased to process reward-related information. To explain the differences in reward versus aversive responses among studies we considered variations in the spatial distribution of sampled recording sites because there is evidence of a rostrocaudal gradient in the shell of the nucleus accumbens for processing appetitive versus aversive reactions (Reynolds and Berridge 2002, 2003). However, this is unlikely to explain the low numbers of aversive responses seen in our study because our electrode tracts often extended into caudal areas of the shell. Moreover, we found responses to aversive-predictive–conditioned stimuli at both anterior (+1.7 mm from bregma) and posterior (+0.7 mm from bregma) sites. In contrast, the recording sites of Setlow et al. were less caudal to ours and were confined to the core subregion of the nucleus accumbens.
A more plausible explanation for the presence of a bias toward reward-related responses in our study is that rats were exposed to more reward versus aversive trials throughout training and testing. Consequently, the conditioned stimulus was more predictive of the outcome in reward versus aversive trials. Indeed, stronger reward versus aversive conditioning was reflected in rats' behavior because there were fewer "incorrect no-go " responses in reward trials than "incorrect go " responses in aversive trials (see Fig. 4).
Reward outcome-predicting responses did not correlate with conditioned lick responses
Although outcome-predicting responses did not correlate with the rat's subsequent switching behavior they might have triggered conditioned lick responses. We found limited evidence against this hypothesis because the average population response of reward outcome-predicting responses did not differentiate between trials within which reward-predictive conditioned stimuli were followed by a conditioned lick response versus trials when they were not. Indeed, independent analyses revealed significant neural responses to conditioned stimulus-1 in both the conditioned lick response and no-conditioned lick response conditions, with a trend of greater excitation in the no-conditioned lick response versus conditioned response condition. Additionally, we found no correlation between the magnitude of reward outcome-predicting response and the vigor with which the rat licked after conditioned stimulus-1 presentation. However, we recognize that the number of trials in which the reward-predictive–conditioned stimulus failed to evoke a conditioned lick response is low. Therefore although the neural response to the reward-predictive–conditioned stimulus did not differentiate between trials in which a conditioned lick response occurred versus when it did not, the statistical power available from this sample is low. Moreover, we cannot exclude the possibility that nucleus accumbens activity triggers motor responses in a probabilistic way depending on the pattern of activity in other neural structures.
Recently, single neurons in the ventral striatum were recorded during learning within a go/no-go task (Setlow et al. 2003). Setlow et al. (2003) divided their neuronal population into neurons that started to make discriminative responses between appetitive and aversive stimuli before acquisition of the discrimination versus neurons that showed differential neuronal activity only after acquisition. Characteristics of the former type of neuron included a lack of modulated activity in the presence of a subsequent conditioned response and a relatively high baseline-firing rate. It seems that our sample was akin to that of the first type of neuron described by Setlow et al. Conversely, other workers (e.g., Schultz) seem to sample more of the second type (motor-related) of neurons (Hassani et al. 2001; Hollerman et al. 1998; Tremblay et al. 1998). Possible explanations for these differences include neuronal sampling biases stemming from differences in neurophysiology equipment, species differences, and/or neuronal sampling during different stages of learning between studies.
Indeed, there are previous reports that nucleus accumbens (Bowman et al. 1996; Carelli and Ijames 2001; Hassani et al. 2001; Hollerman et al. 1998; Setlow et al. 2003; Shidara et al. 1998; Wilson and Bowman 2004), midbrain dopaminergic (Schultz 1998), ventral pallidal (Tindell et al. 2004), and basolateral amygdala (Schoenbaum et al. 1999) neurons seem to respond to the motivational but not motor aspects of stimuli. In this regard, the typical response pattern seen in our neurons (excitations to conditioned stimulus-1 and -2 with greater magnitude to conditioned stimulus-1) are similar to response patterns of dopamine neurons in the macaque [excitations to 2 different, consecutively presented conditioned stimuli, with greater magnitude of response to the conditioned stimulus presented earliest within the trial (Schultz et al. 1993)] and ventral pallidal neurons in the rat (Tindell et al. 2004). Here, we have provided additional evidence that many nucleus accumbens neural responses to conditioned stimuli do not correlate with subsequent behavioral switching.
A proportion of neural responses encoded outcome-predictive and behavioral switching information
Our task design allowed us to analyze neural responses to conditioned stimuli that triggered a switch in the rat's subsequent behavior, involving a reallocation of behavioral resources, as previously defined by Redgrave et al. (1999b). Thus in one type of "switch " trial rats stopped bar-pressing and reallocated their behavioral resources to spigot-licking, whereas in the other they stopped spigot-licking and reallocated their behavioral resources to avoid the spigot (typically rats moved back or turned away from the spigot). These trials were in contrast to "no-switch" trials within which there was no change in the allocation of behavioral resources because the rats continually licked the spigot from pre- to postconditioned stimulus presentation.
We found no neuron within our sample responded exclusively during switching of the rat's upcoming behavior. However, nearly 1/3 of neural responses evoked by the conditioned stimulus were determined by an interaction of the anticipated outcome and behavioral switching, as was the activity of all neurons when considered as a population. These responses are interpreted as signaling a switch or "no-switch " in the rat's subsequent behavior for a particular outcome. Akin to reward outcome-predicting responses, it is unlikely that these responses triggered conditioned lick responses because both reward-switch and reward-no switch responses were followed by a conditioned lick response. The pattern in the valence and number of "outcome-switch " (4 excitations + 1 inhibition = 5 responses) versus "outcome-no switch " (4 excitations + 2 inhibitions = 6 responses) demonstrated that approximately equal numbers of neurons encoded outcome-switching versus outcome-no switching information. It should be noted that the median magnitude of effect of the outcome-switching interaction, across all responsive neurons, was 3.5 times less than that of the outcome-prediction effect.
These data corroborate with modulations in cue-directed behavioral switching after cell-body lesions of the nucleus accumbens (Bowman and Brown 1998; Reading and Dunnett 1991; Reading et al. 1991), neurochemical lesions of dopaminergic inputs to the nucleus accumbens (Evenden and Carli 1985; Robbins and Koob 1980), and psychopharmacological modulation of dopamine neurotransmission within the nucleus accumbens (Bakshi and Kelley 1991a,b; Cools 1980; Evenden and Robbins 1983a,b; Oades 1985; Robbins and Sahakian 1983; van den Bos and Cools 2003; Yun et al. 2004). Additionally, our data demonstrate that neural responses in the nucleus accumbens can occur in both the presence and absence of a subsequent behavioral switch. Furthermore, it is possible that nucleus accumbens neurons encode switching information only when they also encode outcome-predicting information. Indeed, it seems from our data that behavioral switching has a more limited role than outcome-prediction in nucleus accumbens processing because no neuron solely encoded behavioral switching information, there were fewer outcome-switching than outcome-prediction neurons, and outcome-prediction had the greatest magnitude of effect on neural responses to conditioned stimuli.
In summary, over 2/3 of conditioned stimulus-responsive nucleus accumbens neurons in our sample responded to the outcome-predictive properties of conditioned stimuli and did not seem to facilitate behavioral switching. However, nearly 1/3 of neural responses were outcome-predictive with additional modulation by the presence or absence of a subsequent behavioral switch. These data suggest a primary functional role for the nucleus accumbens in encoding outcome-predicting information (Cardinal and Everitt 2004; O'Doherty et al. 2004; Parkinson et al. 2000; Robbins and Everitt 2002; Schultz et al. 2003; Setlow et al. 2002, 2003) and a more limited role in behavioral switching (Evenden and Carli 1985; Reading and Dunnett 1991; Reading et al. 1991; Robbins and Koob 1980; Robbins and Sahakian 1983; van den Bos and Cools 2003).
|
GRANTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: D.I.G. Wilson, School of Psychology, University of St. Andrews, St. Mary's Quadrangle, South Street, St. Andrews, Fife, Scotland KY16 9JP, U.K. (E-mail: digw{at}st-and.ac.uk)
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Bakshi VP and Kelley AE. Dopaminergic regulation of feeding behavior: II. Differential effects of amphetamine microinfusion into three striatal subregions. Psychobiology 19: 233–242, 1991b.
Berridge KC and Robinson TE. What is the role of dopamine in reward: hedonic impact, reward learning, or incentive salience? Brain Res Brain Res Rev 28: 309–369, 1998.[CrossRef][Medline]
Bowman EM, Aigner TG, and Richmond BJ. Neural signals in the monkey ventral striatum related to motivation for juice and cocaine rewards. J Neurophysiol 75: 1061–1073, 1996.
Bowman EM and Brown VJ. Effects of excitotoxic lesions of the rat ventral striatum on the perception of reward cost. Exp Brain Res 123: 439–448, 1998.[CrossRef][ISI][Medline]
Cardinal RN and Everitt BJ. Neural and psychological mechanisms underlying appetitive learning: links to drug addiction. Curr Opin Neurobiol 14: 156–162, 2004.[CrossRef][ISI][Medline]
Carelli RM and Ijames SG. Selective activation of accumbens neurons by cocaine-associated stimuli during a water/cocaine multiple schedule. Brain Res 907: 156–161, 2001.[CrossRef][ISI][Medline]
Carelli RM, Ijames SG, and Crumling AJ. Evidence that separate neural circuits in the nucleus accumbens encode cocaine versus "natural" (water and food) reward. J Neurosci 20: 4255–4266, 2000.
Chang JY, Janak PH, and Woodward DJ. Comparison of mesocorticolimbic neuronal responses during cocaine and heroin self-administration in freely moving rats. J Neurosci 18: 3098–3115, 1998.
Cools AR. Role of the neostriatal dopaminergic activity in sequencing and selecting behavioural strategies: facilitation of processes involved in selecting the best strategy in a stressful situation. Behav Brain Res 1: 361–378, 1980.[ISI][Medline]
Cromwell HC and Schultz W. Effects of expectations for different reward magnitudes on neuronal activity in primate striatum. J Neurophysiol 89: 2823–2838, 2003.
Evenden JL and Carli M. The effects of 6-hydroxydopamine lesions of the nucleus accumbens and caudate nucleus of rats on feeding in a novel environment. Behav Brain Res 15: 63–70, 1985.[CrossRef][ISI][Medline]
Evenden JL and Robbins TW. Dissociable effects of d-amphetamine, chlordiazepoxide and alpha-flupenthixol on choice and rate measures of reinforcement in the rat. Psychopharmacology (Berl) 79: 180–186, 1983a.[CrossRef][Medline]
Evenden JL and Robbins TW. Increased response switching, perseveration and perseverative switching following d-amphetamine in the rat. Psychopharmacology (Berl) 80: 67–73, 1983b.[CrossRef][Medline]
Grill HJ and Norgren R. The taste reactivity test. I. Mimetic responses to gustatory stimuli in neurologically normal rats. Brain Res 143: 263–279, 1978.[CrossRef][ISI][Medline]
Hassani OK, Cromwell HC, and Schultz W. Influence of expectation of different rewards on behavior-related neuronal activity in the striatum. J Neurophysiol 85: 2477–2489, 2001.
Hollerman JR, Tremblay L, and Schultz W. Influence of reward expectation on behavior-related neuronal activity in primate striatum. J Neurophysiol 80: 947–963, 1998.
7 ms; y-axis = number of action potentials; x-axis = time between consecutive spikes (s); bin size = 1 ms; n = 13,289 spikes].
Hz) on trials within which the reward- or aversive-predictive conditioned stimulus caused a switch or no-switch in the rat's subsequent behavior from all neurons (n = 82; ±95% CI). Repeated-measures ANOVA revealed that the population conditioned stimulus-1 response was significantly influenced by the upcoming outcome alone [F(2,739) = 13.452, P < 0.001, 
