The ability to adapt behavior in a changing environment is a hallmark of intelligent systems. From adjusting our driving speed to match road conditions to responding to a last-minute change of plans, mental flexibility underpins much of our day-to-day functioning.

To perform optimally, an animal must continuously monitor its own behavior and adjust it according to circumstances. A region of the brain called the caudate nucleus is thought to contribute to this process by keeping track of the relation between an action and its outcomes, but it is not clear how it monitors cognitive aspects of ongoing behavior.

Yanike and Ferrera have clarified this process by recording electrical activity from the caudate nucleus in two monkeys as they categorized visual stimuli. The monkeys viewed a moving stimulus and classified it as ‘fast’ or ‘slow’ relative to a reference speed that varied from trial to trial. The monkeys were trained to use two different references speeds and were told which reference speed to use at the start of each trial. They used an eye movement to indicate their decision.

Most neurons within the caudate nucleus responded after the monkey had made a decision, suggesting that these neurons might be involved in evaluating the decision that had just been made. The response of the neurons depended on the stimulus speed, and also on the category (fast or slow) in which the stimulus belonged. This observation indicates that the caudate nucleus tracked the context (reference speed) as well as the stimulus speed.

Yanike and Ferrera also showed that the response of the entire population of caudate neurons could be decoded to reveal both the speed of the stimulus and whether the monkey had categorized it as fast or slow. This shows that after a decision has been made, neurons continue to signal both the stimulus and the context in which that stimulus was presented. Such ‘post-decision’ monitoring is important for anticipating the outcome of the decision. Overall the results suggest that the caudate nucleus helps animals to adapt their behavior to rapidly changing circumstances by supporting decision-making that takes context into account.

We adapt to different situations by sorting information into behaviorally meaningful categories that can change rapidly across contexts. This categorization process allows us to interact with such environments. Eager to get to an important meeting, we drive fast, but the actual speed depends on the speed limits along the way. In this situation, categories are delineated by a category boundary (i.e., speed limit) that varies with different environments. Flexible categorization requires continuous monitoring of the changes in the environment to ensure that classification reflects current task demands, even after categories are well learned. How the brain monitors such complex cognitive behavior is not well understood. The basal ganglia, particularly the caudate nucleus, are known to be essential for flexible behavior (Wise et al., 1996; Barnes et al., 2005) and together with the interconnected prefrontal and parietal cortices play an important role in categorization (Poldrack et al., 1999; Poldrack et al., 2001; Seger and Cincotta, 2005; Freedman and Assad, 2006; Seger, 2008; Antzoulatos and Miller, 2011; Ashby and Maddox, 2011; Mendez et al., 2011; Merchant et al., 2011). Many psychiatric and neurological disorders that compromise the caudate nucleus are characterized by impairment in cognitive flexibility (Knowlton et al., 1996; Shohamy et al., 2004; Montoya et al., 2006). Human and animal studies have provided extensive evidence for the critical role of this structure in category learning (Seger and Miller, 2010; Antzoulatos and Miller, 2011). Neurobiological models of categorization have suggested that the caudate nucleus, together with interconnected cortical and subcortical structures, contributes to the maintenance and switching of rules that guide categorization (Maddox and Ashby, 2004; Ashby and Ennis, 2006).

Despite the established role of the caudate nucleus in flexible behavior by linking actions and outcomes (Hikosaka et al., 2000; Yin et al., 2005; Graybiel, 2008; Balleine and O'Doherty, 2010), its role in the monitoring of such behavior is less clear. Recent studies have highlighted the strong contribution of the caudate to post-action evaluation by monitoring behavioral performance based on reward information (Lau and Glimcher, 2007; Ding and Gold, 2010; Thorn et al., 2010; Kim et al., 2013). The general idea is that the caudate detects a mismatch between expected and actual outcomes and these prediction error signals tend to alert an organism about the overall level of behavioral performance. Some studies have shown that post-action neuronal activity maintained memory traces of specific actions, possibly linking them to outcomes (Lau and Glimcher, 2007; Kim et al., 2013). Others have found that post-action activity represented outcomes independent of specific actions (Ding and Gold, 2010). Monitoring signals are sensitive to behavioral context change (Hikosaka and Isoda, 2010) and can also incorporate uncertainty estimates (Badre, 2012; Kepecs and Mainen, 2012) known to modulate caudate activity (Ding and Gold, 2012; Yanike and Ferrera, 2014). Thus, a systematic evaluation of caudate monitoring signals is important for understanding its role in behavioral flexibility, when changes in context and outcome are frequent.

To study what aspects of cognitive flexible behavior are monitored in the caudate nucleus we recorded neuronal activity during a categorization task in which decision criteria changed rapidly across trials. We evaluated post-decision signals both at the level of individual neurons and their population activity. We found that while individual neurons were highly context specific, their population activity across ensembles of neurons provided an accurate and separable read-out of sensory and cognitive aspects of ongoing behavior.

We trained two monkeys on a speed categorization task in which they categorized stimulus speed depending on the position of two different category boundaries (Figure 1A). Each boundary divided a set of random dot stimuli of 8 different speeds (2, 4, …, 16 deg/s) into ‘fast’ and ‘slow’ categories. On each trial, one of the boundary speeds (5 or 13 deg/s), selected randomly, was indicated by a visual cue at the beginning of the trial, prior to the motion stimulus. Therefore, depending on the visual cue, a particular speed of dot movement would switch from belonging to the fast or slow category. Two choice targets (red/green), the locations of which were randomized from trial to trial, were presented adjacent to the motion stimulus. Monkeys made saccades to the red target if they categorized the stimulus as ‘slow’ and to the green target if they categorized the stimulus as ‘fast’.

Figure 1

Download asset Open asset

Both monkeys flexibly adjusted their categorization behavior depending on the position of the category boundary. We sorted behavioral data across animals and trials by the boundary position. We obtained two shifted psychometric functions (Figure 1B), suggesting that animals classified the stimuli using two different internal estimates of the boundary position. To estimate the internal representation of each boundary we found the stimulus speed for which the animal was equally likely to classify the stimulus as ‘fast’ or ‘slow’ (i.e., the point of subjective equality (PSE), Figure 1B). The per-session PSE's formed two non-overlapping distributions (Figure 1C) with means (slow: 6.9 ± 0.1; fast: 11.3 ± 0.12) shifted inward relative to the actual boundary speeds (5 and 13, respectively; Figure 1C). This implies that the monkeys' categorical decisions were less accurate for stimulus speeds near the boundaries, possibly due to uncertainty about the boundary position or its internal estimate. Consistently, the greater categorization uncertainty at the boundaries, expressed as an error rate, correlated with a higher rate of trials that the animal aborted without making a choice (broke fixation) (Figure 1D). Both the error rate and the rate of fixation breaks peaked for speeds 6 and 12, suggesting that subjects had greater difficulty categorizing these near boundary stimuli.

Individual neurons monitored specific context

We analyzed a total of 155 presumed projection neurons from the associative caudate nucleus (Figure 2A) in two monkeys performing the speed categorization task (Monkey C: n = 91; Monkey F: n = 64). For each neuron only trials in the neuron's response field were included. The data from the two monkeys were combined as they were qualitatively similar. We found that the majority of caudate neurons were responsive after, and not prior to, the animal making a decision. Therefore we focused on two post-decision periods of the task, one right after the decision (‘post-saccade’, 0–400 ms after saccade onset, median saccade onset 309 ms) and another when the correctness of the decision was revealed at 800 ms after the decision onset (‘reward’, 0–600 ms after reward onset). We found that many neurons (71 out of 155, 46%, Figure 2B, top) showed significantly different activity between the fixation period and at least one of the post-decision periods of the task (bootstrap test, p < 0.01). Next, we asked whether caudate neurons represented category-related information regardless of the properties of the visual stimuli or their stimulus selectivity. We identified 38 out of 71 neurons that were sensitive to the category context (i.e., boundary position) during either one or both post-decision periods (Figure 2B, bottom). The firing rates of these neurons were significantly different between the two boundary positions for at least one speed either during the post-saccade (‘psacc’, n = 31) and/or reward (‘rwd’, n = 23) periods of the task (Figure 3, boostrap test, p < 0.05 Bonferroni corrected). Figure 4 shows example neurons with activity during the post-saccade (A) and reward (B) periods with a significantly different average response to one out of 8 stimuli, corresponding to speed 12 (see Figure 4—figure supplement 1 for other examples), for which they responded with either significantly higher (Figure 4A) or lower (Figure 4B) firing rate on trials with the fast compared to slow boundary positions (bootstrap test, p < 0.0001). These neurons responded similarly to most other stimuli, irrespective of the boundary position (Figure 4A,B, bottom). The majority of these post-decision neurons (psacc: 19/31, 61%; rwd: 15/23, 65%) discriminated significantly only one out of 8 stimuli and many of them (psacc: 22/31, 71%; rwd: 11/23, 48%) had significantly different neuronal activity for stimuli near the category boundaries, speeds 6 and 12 (Figure 4E,F, bottom).

Figure 2

Download asset Open asset

Figure 3

Download asset Open asset

Figure 4 with 1 supplement see all

Download asset Open asset

To quantify the sensitivity to the boundary position, we calculated a category index for each neuron (CI; see ‘Materials and methods’) separately for each period of the task. The CI measures how well neuronal responses discriminate the same stimuli when presented with different boundaries. The index ranges between 0 and 1, with values close to 0 indicating weak discrimination and values close to 1 indicating strong differentiation between the two categories. Similarly to the example neurons, the CIs between the stimuli close to the boundaries (speeds 6 and 12) were different for most individual neurons in both periods of the task (Figure 4C,D). When averaged across neurons, the CIs for the two most difficult stimuli were the highest during both periods of the task (Figure 4E,F, top). The ability of individual neurons to multiplex their function according to the relevant boundary position is inconsistent with the known sensitivity of caudate neurons to the rate of reward coding, which varied similarly with two boundaries (Cromwell and Schultz, 2003; Hassani et al., 2001). These results show that individual neurons were limited in their ability to monitor behavior across multiple boundaries. However, as a population, the neuronal activity of these neurons could differentiate stimuli near both category boundaries.

The neuronal population read-out provided a global monitoring signal

Next we tested whether combined signals across pools of individual neurons could provide a reliable population code (Pouget et al., 2003) to monitor behaviors across multiple boundaries. We predicted on each trial the stimulus speed and category boundary position (for a total of 16 conditions: 8 × 2) based on the neuronal activity of populations of independently recorded neurons (see ‘Materials and methods’). To evaluate the prediction accuracy, we used the proportion of correct estimates for either speed or boundary position. We found that on each trial, the population activity of ensembles of caudate neurons provided a reliable read-out of the two signals: the identity (speed) of the previously categorized stimulus and the context (boundary position) in which the stimulus was categorized with above chance accuracy during the post-saccade (Figure 5A, B; n = 31) and reward (Figure 5C,D; n = 23) periods of the task. To account for variable levels of the subjects' performance, we obtained the prediction accuracy for the speed and boundary position separately for correct only trials and for all trials (correct and incorrect) separately. On average, the prediction accuracies for the two signals were significantly better on correct trials compared to all trials for the post-saccade (Figure 5A,B; 1-way ANOVA (trial type), p < 0.0001) and reward (Figure 5C,D; p < 0.0001) populations. We also evaluated decoding accuracy as a function of the number of neurons (Figure 5—figure supplement 1). The decoding performance converged towards an asymptote faster for the boundary position (∼5 neurons) compared to the speed (∼10–15 neurons) in each neuronal population. This suggests a more redundant neuronal code for the boundary position.

Figure 5 with 1 supplement see all

Download asset Open asset

To further quantify the relationship between the accuracy of the population read-out and subjects' behavioral performance for each stimulus, we determined the prediction accuracy of the stimulus speed by splitting trials into slow and fast boundary positions. We found that the accuracy of the population read-out for the identity of the previously categorized stimuli varied with the categorization difficulty at each boundary. This was shown by a significant interaction between the speed and boundary position (Figure 6A,C; 2-way ANOVA (speed, boundary), p < 0.001) in each neuronal population. To correlate subjects' categorization performance and the reliability of the population representation, we averaged the prediction accuracy between the slow and fast boundary positions across all speeds and plotted it as a function of the distance to the boundary for post-saccade (Figure 6B) and reward (Figure 6D) populations. The average prediction accuracy was the highest for the near boundary stimuli on the right, coinciding with the highest behavioral error rate (Figure 1B,D), then decreased with the distance to the boundary. The population read out of the previous stimulus speed was strongly correlated with subjects' behavioral performance, expressed as an error rate, averaged across two boundaries for each speed (Figure 1B), similarly in both neuronal populations (Figure 6B,D; psacc: R² = 0.85; rwd: R² = 0.76). Thus, the population was best at maintaining the sensory representation of stimuli, for which the subjects were prone to make categorization errors.

Figure 6

Download asset Open asset

We then compared the predication accuracy between trials when subjects categorized the same stimuli correctly and incorrectly for a subset of stimuli, speeds 6 and 12, with sufficient number of incorrect trials (see ‘Materials and methods’). In both neuronal populations, the incorrect categorical judgments were followed by the failure to represent the identity of the preceding stimulus (Figure 6A,C). In contrast, correct categorization of the same stimuli, either near or far from the boundary, were followed by a reliable but different read-out of stimulus speed based on the population activity (2-way ANOVA [trial type, boundary], p < 0.001). These results suggest that outcome uncertainty near the boundaries can act as a gating mechanism to enhance representation of the boundary stimuli to potentially overcome decision uncertainty.

The monitoring signal reflected behavioral relevance

We reasoned that if this neuronal code is cognitive and reflects categorization process, instead of reward processing, then even irrelevant information which could be pertinent for categorization should be processed in a task meaningful way. We asked if caudate neurons encoded category-irrelevant features of otherwise relevant stimuli and whether the post-decision evaluation of the irrelevant information reflected categorization difficulty. We tested this hypothesis by computing how well we could predict discrimination accuracy between two opposite directions of dot motions for each stimulus separately for the slow and fast boundaries based on neuronal activity (see ‘Materials and methods’). We found that even those features of stimuli (i.e., dots direction) that were not linked to category identity were selectively suppressed when stimuli were near the boundary compared to when they were far from the boundary in both neuronal populations (see ‘Materials and methods’; Figure 7A,B; 3-way ANOVA (direction, speed, boundary), p < 0.01). Similarly to single unit studies (Hussar and Pasternak, 2009), the post-decision population read-out in the caudate strongly reflected behavioral relevance of stimuli during categorization.

Figure 7 with 1 supplement see all

Download asset Open asset

We found that caudate neurons monitored flexible categorization behavior by providing distinct representation of decision-pertinent variables, from stimulus specific features to more cognitive signals reflecting internal estimates of decision variables. The monitoring signal in the caudate was not only accurate at representing ongoing behavior, but it also interpreted behavioral consequences in light of changing demands of the task. The accuracy of the monitoring signal was inversely related to behavioral proficiency, suggesting that the caudate nucleus allocates more efficient coding for changing and/or uncertain information critical for behavioral flexibility. These findings provide direct evidence for the role of the caudate nucleus in online monitoring of complex cognitive behavior, possibly allowing for contextually specific decisions to adapt to rapidly changing context (Daw et al., 2006; Hikosaka and Isoda, 2010; Pearson and Platt, 2013).

It is well-established that the caudate nucleus supports cognitive function, including categorization, by utilizing reward or motivation information. However, how cognitive and reward signals interact to support complex behavior is still unclear. In this report, we showed that the monitoring signals in the caudate nucleus reflecting cognitive variables were shaped by reward modulation, as more accurate representation correlated with the most variable reward outcome. We found that individual neurons were highly selective at linking sensory and reward information. In other words, the neurons did not generalize across sensory inputs to code reward-related information. However, their population response provided a reliable and separable representation of both cognitive and motivational signals on a trial-by-trial basis. The reported monitoring signals in the caudate can reflect the outcome prediction and/or the rate of reward coding (Hollerman et al., 1998; Hassani et al., 2001; Cromwell and Schultz, 2003). The motivational signals potentially shape categorical representation in the caudate to monitor context specific decisions to adapt to rapidly changing conditions.

Our findings are consistent with the growing evidence for the strong contribution of the caudate nucleus to post-decision monitoring and evaluation (Lau and Glimcher, 2007; Ding and Gold, 2010; Thorn et al., 2010). Traditional views postulate that the striatum, which includes the caudate nucleus, contributes to behavioral evaluation through updates based on differences between the observed and expected outcomes through reinforcement learning (Daw et al., 2006; Williams and Eskandar, 2006). Recent views suggest that such updates can also incorporate uncertainty estimates, similar to the prefrontal cortex (Badre, 2012; Kepecs et al., 2008; Kepecs and Mainen, 2012). Caudate neuronal activity is sensitive to both stimulus (Ding and Gold, 2012) and outcome uncertainty (Yanike and Ferrera, 2014), which can potentially shape more efficient coding of changing or uncertain information.

Our findings that the caudate nucleus provides online monitoring of flexible behavior are consistent with a recent study showing that individual caudate neurons coded values of visual objects in a flexible manner (Kim and Hikosaka, 2013). Specifically, neurons in the head of the caudate nucleus represented changes in recent value of objects while monkeys made saccades to visual objects with different values and inactivation of this structure disrupted their behavioral preference for high value objects. While we also recorded in the anterior part of the caudate nucleus (see ‘Materials and methods’), our experimental paradigm was different. During the categorization task, monkeys had to maintained representation of both stable (i.e., always far from boundaries) and changing (i.e., near the boundaries) values of dot stimuli to perform successfully, well after the initial learning had occurred. Therefore, our findings are complementary to the aforementioned study because we showed that the caudate nucleus provides abstract cognitive monitoring signals beyond flexible reward associations. Taken together, these results suggest that the anterior caudate plays an important role in the change detection network (Isoda and Hikosaka, 2011; Pearson and Platt, 2013) contributing to a rapid adjustment of behavior.

Our findings that the representation of category-relevant information was enhanced while category-irrelevant information was suppressed near the boundaries are consistent with the possible role of the caudate nucleus in the working memory updating. Computational models (O'Reilly and Frank, 2006) have suggested that the striatum contributes to a selective gating of information flow into the working memory in the prefrontal cortex. The present findings suggest that such selective gating in the caudate nucleus can occur via a neuronal population code, which allows for independent read-out of all task variables by downstream structures. In our experimental paradigm the direction of the dots motion was irrelevant for the speed categorization and the animals were never explicitly required to utilize that information. Yet, the accuracy of the population read-out for the dots motion direction in the caudate was suppressed near the boundary compared to that far from the boundary. These results suggest that the caudate nucleus sorts information by its relevance to the task at hand, possibly automatically, to potentially affect ongoing behavior (Badre, 2012).

Finally, our findings suggest that unrevealing the nature of neuronal population code can greatly benefit our understanding of the neural basis of complex cognitive behavior. We found that individual neurons provided a context-specific code, while their population read-out covered all aspects of behavior and multiple separable signals could be reliably extracted. Previous studies on visual categorization have shown that representation of multiple categories in the prefrontal cortex at the level of single neurons was either distributed or sparse depending on how much the visual categories overlapped (Cromer et al., 2010; Roy et al., 2010). How these differences translate into population codes remains to be studied. The results of our experiment show that the caudate neuronal population code can reliably represent all aspects of cognitive complexity during flexible behavior.

1. 1. Antzoulatos EG
   2. Miller EK

   (2011) Differences between eural activity in prefrontal cortex and striatum during learning of novel abstract categories

   Neuron 71:243–249.

   https://doi.org/10.1016/j.neuron.2011.05.040

   • Google Scholar
2. 1. Ashby FG
   2. Ennis JM

   (2006) The role of the basal ganglia in category learning

   The Psychology of Learning and Motivation 46:1–36.

   https://doi.org/10.1016/S0079-7421(06)46001-1

   • Google Scholar
3. 1. Ashby FG
   2. Maddox WT

   (2011) Human category learning 2.0

   Annals of the New York Academy of Sciences 1224:147–161.

   https://doi.org/10.1111/j.1749-6632.2010.05874.x

   • Google Scholar
4. 1. Badre D

   (2012) Opening the gate to working memory

   Proceedings of the National Academy of Sciences of USA 109:19878–19879.

   https://doi.org/10.1073/pnas.1216902109

   • Google Scholar
5. 1. Barnes TD
   2. Kubota Y
   3. Hu D
   4. Jin DZ
   5. Graybiel AM

   (2005) Activity of striatal neurons reflects dynamic encoding and recoding of procedural memories

   Nature 437:1158–1161.

   https://doi.org/10.1038/nature04053

   • Google Scholar
6. 1. Balleine BW
   2. O'Doherty JP

   (2010) Human and rodent homologies in action control: corticostriatal determinants of goal-directed and habitual action

   Neuropsychopharmacology 35:48–69.

   https://doi.org/10.1038/npp.2009.131

   • Google Scholar
7. 1. Cromer JA
   2. Roy JE
   3. Miller EK

   (2010) Representation of multiple, independent categories in the primate prefrontal cortex

   Neuron 66:796–807.

   https://doi.org/10.1016/j.neuron.2010.05.005

   • Google Scholar
8. 1. Cromwell HC
   2. Schultz W

   (2003) Effects of expectations for different reward magnitudes on neuronal activity in primate striatum

   Journal of Neurophysiology 89:2823–2838.

   https://doi.org/10.1152/jn.01014.2002

   • Google Scholar
9. 1. Daw ND
   2. O'Doherty JP
   3. Dayan P
   4. Seymour B
   5. Dolan RJ

   (2006) Cortical substrates for exploratory decisions in humans

   Nature 441:876–879.

   https://doi.org/10.1038/nature04766

   • Google Scholar
10. Book

    1. Duda RO
    2. Hart PE
    3. Stork DG

    (2001)

    Pattern classification (2nd edition)

    Wiley.

    • Google Scholar
11. 1. Ding L
    2. Gold JI

    (2010) Caudate encodes multiple computations for perceptual decisions

    The Journal of Neuroscience 30:15747–15759.

    https://doi.org/10.1523/JNEUROSCI.2894-10.2010

    • Google Scholar
12. 1. Ding L
    2. Gold JI

    (2012) Separate, causal roles of the caudate in saccadic choice and execution in a perceptual decision task

    Neuron 75:865–874.

    https://doi.org/10.1016/j.neuron.2012.07.021

    • Google Scholar
13. 1. Ferrera VP
    2. Yanike M
    3. Cassanello C

    (2009) Frontal eye field neurons signal changes in decision criteria

    Nature Neuroscience 12:1458–1462.

    https://doi.org/10.1038/nn.2434

    • Google Scholar
14. 1. Freedman DJ
    2. Assad JA

    (2006) Experience-dependent representation of visual categories in parietal cortex

    Nature 443:85–88.

    https://doi.org/10.1038/nature05078

    • Google Scholar
15. 1. Graf AB
    2. Kohn A
    3. Jazayeri M
    4. Movshon JA

    (2011) Decoding the activity of neuronal populations in macaque primary visual cortex

    Nature Neuroscience 14:239–245.

    https://doi.org/10.1038/nn.2733

    • Google Scholar
16. 1. Graybiel AM

    (2008) Habits, rituals, and the evaluative brain

    Annual Review of Neuroscience 31:359–387.

    https://doi.org/10.1146/annurev.neuro.29.051605.112851

    • Google Scholar
17. 1. Goldstone R

    (1994) Influences of categorization on perceptual discrimination

    Journal of Experimental Psychology. General 123:178–200.

    https://doi.org/10.1037/0096-3445.123.2.178

    • Google Scholar
18. 1. Hassani OK
    2. Cromwell HC
    3. Schultz W

    (2001)

    Influence of expectation of different rewards on behavior-related neuronal activity in the striatum

    Journal of Neurophysiology 85:2477–2489.

    • Google Scholar
19. 1. Hikosaka O
    2. Isoda M

    (2010) Switching from automatic to controlled behavior: cortico-basal ganglia mechanisms

    Trends in Cognitive Sciences 14:154–161.

    https://doi.org/10.1016/j.tics.2010.01.006

    • Google Scholar
20. 1. Hikosaka O
    2. Takikawa Y
    3. Kawagoe R

    (2000)

    Role of the basal ganglia in the control of purposive saccadic eye movements

    Physiological Reviews 80:953–978.

    • Google Scholar
21. 1. Hollerman JR
    2. Tremblay L
    3. Schultz W

    (1998)

    Influence of reward expection on behavior-related neuronal activity in primate striatum

    Journal of Neurophysiology 80:947–963.

    • Google Scholar
22. 1. Hussar CR
    2. Pasternak T

    (2009) Flexibility of sensory representations in prefrontal cortex depends on cell type

    Neuron 64:730–743.

    https://doi.org/10.1016/j.neuron.2009.11.018

    • Google Scholar
23. 1. Isoda M
    2. Hikosaka O

    (2011) Cortico-basal ganglia mechanisms for overcoming innate, habitual and motivational behaviors

    The European Journal of Neuroscience 33:2058–2069.

    https://doi.org/10.1111/j.1460-9568.2011.07698.x

    • Google Scholar
24. 1. Jazayeri M
    2. Movshon JA

    (2006) Optimal representation of sensory information by neural populations

    Nature Neuroscience 9:690–696.

    https://doi.org/10.1038/nn1691

    • Google Scholar
25. 1. Kepecs A
    2. Mainen ZF

    (2012) A computational framework for the study of confidence in humans and animals

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 367:1322–1337.

    https://doi.org/10.1098/rstb.2012.0037

    • Google Scholar
26. 1. Kepecs A
    2. Uchida N
    3. Zariwala HA
    4. Mainen ZF

    (2008) Neural correlates, computation and behavioural impact of decision confidence

    Nature 455:227–231.

    https://doi.org/10.1038/nature07200

    • Google Scholar
27. 1. Kim HF
    2. Hikosaka O

    (2013) Distinct basal ganglia circuits controlling behaviors guided by flexible and stable values

    Neuron 79:1001–1010.

    https://doi.org/10.1016/j.neuron.2013.06.044

    • Google Scholar
28. 1. Kim H
    2. Lee D
    3. Jung MW

    (2013) Signals for previous goal choice persist in the dorsomedial, but not dorsolateral striatum of rats

    The Journal of Neuroscience 33:52–63.

    https://doi.org/10.1523/JNEUROSCI.2422-12.2013

    • Google Scholar
29. 1. Knowlton BJ
    2. Mangels JA
    3. Squire LR

    (1996) A neostriatal habit learning system in humans

    Science 273:1399–1402.

    https://doi.org/10.1126/science.273.5280.1399

    • Google Scholar
30. 1. Lau B
    2. Glimcher PW

    (2007) Action and outcome encoding in the primate caudate nucleus

    The Journal of Neuroscience 27:14502–14514.

    https://doi.org/10.1523/JNEUROSCI.3060-07.2007

    • Google Scholar
31. 1. Maddox WT
    2. Ashby FG

    (2004) Dissociating explicit and procedural-learning based systems of perceptual category learning

    Behavioural Processes 66:309–332.

    https://doi.org/10.1016/j.beproc.2004.03.011

    • Google Scholar
32. 1. Mendez JC
    2. Prado L
    3. Mendoza G
    4. Merchant H

    (2011) Temporal and spatial categorization in human and non-human primates

    Frontiers in Integrative Neuroscience 5:50.

    https://doi.org/10.3389/fnint.2011.00050

    • Google Scholar
33. 1. Merchant H
    2. Crowe DA
    3. Robertson MS
    4. Fortes AF
    5. Georgopoulos AP

    (2011) Top-down spatial categorization signal from prefrontal to posterior parietal cortex in the primate

    Frontiers in Integrative Neuroscience 5:69.

    https://doi.org/10.3389/fnsys.2011.00069

    • Google Scholar
34. 1. Montoya A
    2. Price BH
    3. Menear M
    4. Lepage M

    (2006)

    Brain imaging and cognitive dysfunctions in Huntington's disease

    Journal of Psychiatry & Neuroscience 31:21–29.

    • Google Scholar
35. 1. O'Reilly RC
    2. Frank MJ

    (2006) Making working memory work: a computational model of learning in the prefrontal cortex and basal ganglia

    Neural Computation 18:283–328.

    https://doi.org/10.1162/089976606775093909

    • Google Scholar
36. 1. Pearson JM
    2. Platt ML

    (2013) Change detection, multiple controllers, and dynamic environments: insights from the brain

    Journal of the Experimental Analysis of Behavior 99:74–84.

    https://doi.org/10.1002/jeab.5

    • Google Scholar
37. 1. Pouget A
    2. Dayan P
    3. Zemel RS

    (2003) Inference and computation with population codes

    Annual Review of Neuroscience 26:381–410.

    https://doi.org/10.1146/annurev.neuro.26.041002.131112

    • Google Scholar
38. 1. Poldrack RA
    2. Clark J
    3. Paré-Blagoev EJ
    4. Shohamy D
    5. Creso Moyano J
    6. Myers C
    7. Gluck M

    (2001) Interactive memory systems in the human brain

    Nature 414:546–550.

    https://doi.org/10.1038/35107080

    • Google Scholar
39. 1. Poldrack RA
    2. Prabhakaran V
    3. Seger CA
    4. Gabrieli JDE

    (1999) Striatal activation during acquisition of a cognitive skill

    Neuropsychology 13:564–574.

    https://doi.org/10.1037/0894-4105.13.4.564

    • Google Scholar
40. 1. Roy JE
    2. Riesenhuber M
    3. Poggio T
    4. Miller EK

    (2010) Prefrontal cortex activity during flexible categorization

    The Journal of Neuroscience 30:8519–8528.

    https://doi.org/10.1523/JNEUROSCI.4837-09.2010

    • Google Scholar
41. 1. Sanger TD

    (1996)

    Probability density estimation for the interpretation of neural population codes

    Journal of Neurophysiology 76:2790–2793.

    • Google Scholar
42. 1. Seger CA

    (2008) How do the basal ganglia contribute to categorization? Their roles in generalization, response selection, and learning via feedback

    Neuroscience and Biobehavioral Reviews 32:265–278.

    https://doi.org/10.1016/j.neubiorev.2007.07.010

    • Google Scholar
43. 1. Seger CA
    2. Cincotta CM

    (2005) The roles of the caudate nucleus in human classification learning

    The Journal of Neuroscience 25:2941–2951.

    https://doi.org/10.1523/JNEUROSCI.3401-04.2005

    • Google Scholar
44. 1. Seger CA
    2. Miller EK

    (2010) Category learning in the brain

    Annual Review of Neuroscience 33:203–219.

    https://doi.org/10.1146/annurev.neuro.051508.135546

    • Google Scholar
45. 1. Shohamy D
    2. Myers CE
    3. Grossman S
    4. Sage J
    5. Gluck MA
    6. Poldrack RA

    (2004) Cortico-striatal contributions to feedback-based learning: converging data from neuroimaging and neuropsychology

    Brain 127:851–859.

    https://doi.org/10.1093/brain/awh100

    • Google Scholar
46. 1. Thorn CA
    2. Atallah H
    3. Howe M
    4. Graybiel AM

    (2010) Differential dynamics of activity changes in dorsolateral and dorsomedial striatal loops during learning

    Neuron 66:781–795.

    https://doi.org/10.1016/j.neuron.2010.04.036

    • Google Scholar
47. 1. Williams ZM
    2. Eskandar EN

    (2006) Selective enhancement of associative learning by microstimulation of the anterior caudate

    Nature Neuroscience 9:562–568.

    https://doi.org/10.1038/nn1662

    • Google Scholar
48. 1. Wise SP
    2. Murray EA
    3. Gerfen CR

    (1996) The frontal cortex basal ganglia system in primates

    Critical Reviews in Neurobiology 10:317–356.

    https://doi.org/10.1615/CritRevNeurobiol.v10.i3-4.30

    • Google Scholar
49. 1. Yanike M
    2. Ferrera VP

    (2014) Representation of outcome risk and action in the anterior caudate nucleus

    The Journal of Neuroscience 34:3279–3290.

    https://doi.org/10.1523/JNEUROSCI.3818-13.2014

    • Google Scholar
50. 1. Yin HH
    2. Ostlund SB
    3. Knowlton BJ
    4. Balleine BW

    (2005) The role of the dorsomedial striatum in instrumental conditioning

    The European Journal of Neuroscience 22:513–523.

    https://doi.org/10.1111/j.1460-9568.2005.04218.x

    • Google Scholar

Author details

1. Marianna Yanike

   Department of Neuroscience, Columbia University, New York, United States

   Contribution

   MY, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   marianna@vis.caltech.edu

   Competing interests

   The authors declare that no competing interests exist.

2. Vincent P Ferrera

   1. Department of Neuroscience, Columbia University, New York, United States
   2. Department of Psychiatry, Columbia University, New York, United States

   Contribution

   VPF, Conception and design, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

• 754

  views

• 64

  downloads

• 8

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.