Animals can pursue a goal while handling distractions and delays, starting from a variety of initial conditions, and adapting their responses in the face of unexpected obstacles. To guide such flexible goal-directed behavior, there needs to be a representation of the goal itself that is robust to these circumstances, which can provide top-down instruction for selecting appropriate actions. The prefrontal cortex (PFC) plays a central role in flexible goal-directed behavior (Miller and Cohen, 2001; Fuster, 2015), and one of its primary functions is thought to be the maintenance of information relevant for achieving the current goal (i.e. working memory) (Fuster and Alexander, 1971; Funahashi et al., 1989; Miller et al., 1996; Rainer et al., 1998; Romo et al., 1999; Wang, 1999; Erlich et al., 2011; Wimmer et al., 2014; Inagaki et al., 2019; Wu et al., 2020). Therefore, the PFC is a prime candidate area for containing a representation of the current goal itself.

Spatial working memory tasks are well-suited for investigating such representations. First, the current goal, a particular spatial location, can be clearly specified and is ethologically relevant for many species (O'Keefe and Nadel, 1978). Second, all of the aspects of flexibility mentioned above can be incorporated naturally in a form that many animals, including rodents, can solve. However, rodent spatial working memory experiments employed to date with recording in PFC or other brain areas (Wood et al., 2000; Frank et al., 2000; Baeg et al., 2003; Fujisawa et al., 2008; Gill et al., 2011; Harvey et al., 2012; Pfeiffer and Foster, 2013; Wikenheiser and Redish, 2015; Ito et al., 2015; Spellman et al., 2015; Kim et al., 2016; Guise and Shapiro, 2017; Bolkan et al., 2017) have not combined all of these elements of flexibility in a single task. As a result, potential working memory representations of the current goal have been difficult to dissociate from behavioral or sensory correlates. For instance, classic T-mazes have a single start and single route to each of the two goals. Thus, differential neural activity before reaching the T junction could represent the goals themselves, the two sets of stereotyped actions used to reach each goal as well as any associated sensory correlates (e.g. looking left versus right), or the plan to ‘go left’ or ‘go right’. Furthermore, any goal-specific activity could differ if the animal were to start from another point, and therefore not be usable under different initial conditions. Here, we devised a novel spatial working memory task incorporating multiple aspects of behavioral flexibility, allowing a search for a ‘pure’ representation of the goal itself by disambiguating goal-related activity from other correlates.

In our task, rats needed to remember one of three goal locations in each trial. The current goal was encoded during a ‘sample phase’, in which animals were guided with light cues to one goal where they received a small reward. Rats had to remember this location until they needed to navigate to that goal again in the ‘test phase’ – starting from one of three different locations and via one of three routes in the absence of lighted cues – to receive a large reward. The different routes were implemented by using an elevated maze design with bridges that could be raised (open) and lowered (blocked). The available route was only revealed after a 3 s (3.2 s in one animal) ‘fixation’ period (hereafter referred to as the ‘delay period’) during which animals had to hold their nose in one of the three test phase ‘start’ location ports. The correct start location port was assigned pseudorandomly in each trial. To find the correct start port between each trial’s sample and test phase, animals had to poke their nose into different ports until they found the one that elicited a tone when poked, which indicated the correct choice (Figure 1A,B, Video 1). The goal sample phase route and test phase route were also pseudorandomly assigned in each trial (see Materials and methods). This design requires animals to update the currently relevant goal every trial (working memory task) and pushes them to remember the goal itself instead of memorizing a specific behavioral sequence or planning a particular motor action to reach the goal. Thus, the goal location is the key variable to retain, which must then be used flexibly to solve the task: navigating to that goal from any location by any route. Furthermore, having more than two goals excludes a strategy of navigating to one goal by default and only remembering when the goal is the other one – that is having three goals promotes the use of working memory representations of each goal itself.

Figure 1

Download asset Open asset

Video 1

Download asset

Animals reached high levels of performance (Figure 1C,D, mean performance over all rats: 77.37, 95% confidence interval [CI]: [72.46, 81.97]). To examine whether animals indeed remembered the goal location itself instead of particular routes, we compared the performance in trials where the route (bridge) animals had to take in the test phase was the same or different from the outbound and/or return bridge in the sample phase. The performance was comparable across routes for all animals, indicating that rats remembered goal locations instead of routes (Figure 1D, left). In trials where the goal was adjacent to (i.e. 60 degrees in either direction from) the available route, rats most often took the shortest route (88%, 81%, and 95% of trials for each animal, respectively), supporting the idea that animals used a spatial map instead of a fixed route plus recognition strategy. Furthermore, the performance was comparable and above chance for all goal locations, implying that rats remembered each goal location rather than relying on a strategy of remembering only a subset of them (Figure 1D, center left). Rats spent several seconds in each task phase and generally ran faster on test outbound runs than sample outbound runs (Figure 1D, center right). The search time to find the correct start location was on average slightly higher in the trials where the subsequent choice was incorrect, but overlapped with search times in correct trials (Figure 1D, right). During the test phase, rats rarely stopped at the end of the available bridge after crossing it on their way to the goal in a manner that might reflect vicarious trial and error behavior at the choice point (Johnson and Redish, 2007) (see Materials and methods for details).

A Neuropixels probe (Jun et al., 2017a) was chronically implanted in medial prefrontal cortex (mPFC), previously shown to be required in variants of simpler rodent spatial working memory tasks (Spellman et al., 2015; Guise and Shapiro, 2017). One hundred to two hundred units were simultaneously recorded across subareas including anterior cingulate, prelimbic (PL), infralimbic, and dorsal peduncular cortices (Figure 1B) (n = 3 animals: 182, 131, and 98 cells in the three main sessions analyzed, one from each animal; in addition, for two animals, one novel rotation experiment session each, consisting of 152 and 186 cells, was used for the analysis of relearning). We primarily focused our analyses on the delay period since, during that time, behavior is well-controlled, and animals must remember a given goal while being in different defined locations and facing different defined directions, as well as not knowing the required future motor action. Analyses were applied to all putative principal cells with stable firing rates throughout the session (97, 84, and 68 cells in the three main sessions, and 84 and 105 cells in the additional relearning sessions; see Materials and methods) and pooled across subareas (results including all cells or split by subarea were similar and are provided in the supplementary figures as indicated).

We searched for representations of the current goal encoded in terms of the major forms of delay period activity previously found in other working memory tasks, spatial or non-spatial, involving recordings from the PFC or elsewhere in primates and rodents: activity reflecting the representation at the time of encoding the sample item (Funahashi et al., 1989; Miller et al., 1996; Rainer et al., 1998; Romo et al., 1999; Wu et al., 2020), elevated/suppressed activity in single cells (Fuster and Alexander, 1971; Funahashi et al., 1989; Miller et al., 1996; Rainer et al., 1998; Romo et al., 1999; Kim et al., 2016; Inagaki et al., 2019), or sequential firing patterns across multiple cells that tile the delay period (Baeg et al., 2003; Fujisawa et al., 2008; Pastalkova et al., 2008; Harvey et al., 2012; Ito et al., 2015), oscillatory phase-dependent firing (Siegel et al., 2009; Watrous et al., 2018), and elevated/suppressed covariances in firing among pairs of neurons (Barbosa et al., 2020).

We first tested whether memory of the current goal could be maintained during the delay period by a firing rate pattern across cells (population vector, PV) similar to the one when the animal was at the goal itself during the sample phase. To begin with, the PV at each goal during the sample phase was distinguishable and stable over time (Figure 2A, Figure 2—figure supplement 1A), including across sample and test phases (Figure 2—figure supplement 1B). In addition to distinguishing which goal the animal was at in both sample and test phases, the task phase could also simultaneously be almost perfectly decoded (Figure 2—figure supplement 1B, right), likely because the animal received less reward in the sample phase. Having established the presence of a stable spatial representation at each goal, we correlated the overall activity during the delay period with the sample phase PV at each goal and asked whether it was more correlated with the currently remembered goal’s PV. This was not the case (Figure 2B), as also seen in a different task (Guise and Shapiro, 2017). To test whether the remembered goal was represented as transient increases in correlation or a switching between current and other goals with the relevant (current) one being overrepresented (Kelemen and Fenton, 2010), we attempted to predict the goal based on correlation scores resulting from sliding a window of variable width across the delay period. While the animal’s current (i.e. start) location was readily decodable (as expected), the remembered goal was not, using a wide range of time bins (Figure 2C) or activity restricted to individual subareas (Figure 2—figure supplement 2) or to cells with significant spatial selectivity at goals (Figure 3—figure supplement 1; note that individual cells in any subarea could show spatial selectivity at goals, as well as at starts, at both, or neither).

Figure 2 with 2 supplements see all

Download asset Open asset

If the remembered goal is not maintained by activity directly related to activity at the goal itself, it could be (1) transformed into a different, but goal-specific, pattern, potentially dependent on the start location, (2) encoded in egocentric coordinates (i.e. the direction relative to the current start location; Sarel et al., 2017) instead of in terms of the absolute (allocentric) goal location, (3) represented by a sequential, instead of tonic, activity pattern, and/or (4) reflected in the phase of spike times or the short timescale interactions between pairs of neurons (Barbosa et al., 2020). We initially tested if any single-cell activity in 100-ms- to full-delay-period-sized time bins showed consistent firing differences for allocentric, start-dependent, or egocentric goal location (Figure 3—figure supplement 2). We found significant differences for the current (i.e. start) location, but not allocentric or egocentric goal location. There was also no evidence of start-dependent encoding of goals, that is a unique code for the nine start-goal pairs.

We then tested whether the remembered goal was encoded with a sequential activity pattern across multiple cells that may not be detectable at the single-cell level (Figure 3). We employed several classification methods at multiple time resolutions (Figure 3C). Note that, for this analysis, potential activity patterns were always referenced to the delay period onset, as previously seen for memory-related sequences that tile the delay period (Fujisawa et al., 2008; Pastalkova et al., 2008; Harvey et al., 2012). Results were consistent across methods and time resolutions: current location could readily be decoded with any classification method for time bin resolutions between 100 ms to the full delay period duration. In contrast, no method could successfully classify the remembered goal in allocentric or egocentric coordinates at any time resolution (Figure 3C), including when using all cells regardless of firing rate stability (Figure 3—figure supplement 3).

Figure 3 with 4 supplements see all

Download asset Open asset

We considered whether in our well-trained animals the representation of the remembered goal might be less prominent than during learning, and therefore, if goal representations may be more easily identified when the animal is learning (Liu et al., 2014; Maggi et al., 2018). While we did not record activity during the long training period, we performed a behavioral manipulation experiment after the task had been learned in which animals (n = 2 sessions from two animals) had to relearn the task in an altered configuration. Specifically, we rotated the maze by 60 degrees ~ 1/3 of the way through the recording session, so that the reference frame was changed, with the goal and start positions now in between their previous positions (Figure 4A). Animals had never seen this configuration before or seen a rotation of the maze. Consistent with the idea that the animals must adapt to the new configuration and learn to apply previously internalized rules, performance dropped dramatically then gradually improved over the remainder of the session (Figure 4A). However, during this relearning, when the cognitive demand might be higher and thus elicit a more prominent representation of critical task variables, population analysis still could not decode the currently remembered goal during the delay period. In contrast, spatial representation of the start locations was again clearly represented even as the animal adapted to the new configuration (Figure 4B,C).

Figure 4

Download asset Open asset

We next tested whether information might be stored in spike timing relative to local field potential (LFP) oscillations (Siegel et al., 2009; Watrous et al., 2018). We explored three frequency bands (2.5–5 Hz, 5–12 Hz, and 15–30 Hz), identified based on their elevated power in the delay period (Figure 5A, left). First, we calculated each cell’s goal-specific phase preference in the delay period. We compared the distribution of phase preference magnitudes to one where goal labels were shuffled. The distributions were not different, neither when including all stable cells, nor when selecting only cells that showed significant overall phase locking, suggesting the remembered goal does not affect overall phase preference (Figure 5A, Figure 5—figure supplement 1A,B). Second, we asked whether spike counts at specific phases might differ in a goal-specific manner, either when using all stable cells or only those that were significantly phase locked to at least one of the goals, but they did not (Figure 5A, Figure 5—figure supplement 1C). We also explored the possibility of a recently described form of ‘activity-silent’ memory, in which working memory is expressed in the spiking synchrony between pairs of neurons while stimulus information is not decodable from firing rates (Barbosa et al., 2020). However, neither for the pairs of neurons exhibiting excitatory interactions nor for the pairs of neurons exhibiting inhibitory interactions did covariances differ between trials associated with one goal versus the others (Figure 5B, Figure 5—figure supplement 2, see Materials and methods). Together, these results suggest that previously described forms of working memory maintenance are not responsible for storing the current goal in our task in which this information must be employed flexibly.

Figure 5 with 2 supplements see all

Download asset Open asset

To test whether goal representations are present in other periods than the (nosepoke ‘fixation’) delay period, we analyzed data in a time-resolved fashion aligned to multiple time points during trial progression. Specifically, we tested how well the current goal could be decoded with respect to key behavioral reference time points: (1) when the animal arrives at the goal during the sample phase, (2) when the animal returns to the center during the transition between sample and test phases, (3) when the route becomes available, (4) at the choice point when the animal enters the outer ring of the maze in the test phase, and (5) when the animal arrives at the goal during the test phase. Importantly, we performed these analyses not only using the neural data but also, separately, using the position tracking data of the two LEDs on the animal’s head. We found that it was possible to decode the currently remembered goal location from the neural data at various time points. However, the remembered goal could also be decoded from the behavioral tracking data alone with a very similar time course. Thus, it is likely that this neural representation of the currently remembered goal at these times is due to the animal’s position, orientation, posture, or other behavioral features (Figure 3—figure supplement 4). Crucially, the decoding performance based on tracking data was at chance levels throughout the nose poke delay period (Figure 3—figure supplement 4, third row, right, period from −3 to 0 s), providing direct evidence that our task design reduces the presence of confounding behavioral variables during that period.

If the mPFC does not encode memory of the goal in this task, what task-relevant processes might it support? We tested whether other task features could be decoded from mPFC activity. First, we analyzed if the activity at the goal in the sample phase (presumably during encoding) differs when there is an error in the subsequent test phase or not, but this was not the case. In contrast, after the animal had made an incorrect choice in the test phase, activity at the goal was markedly different, presumably due to the lack of reward; however, correlations between the PVs of activity at the goal among correct trials and, separately, among error trials was comparable (Figure 6A). Furthermore, mPFC delay period activity did not indicate an upcoming or past error (Figure 6B), further corroborating that mPFC might not directly store current memory content. We then checked if mPFC distinguished the two task phases not only at the goal (Figure 2—figure supplement 1B), which could be due to the amount of reward but when animals returned to the center. Before reaching the center, and at the center, task phase was not decodable (Figure 6C) (note the decodability afterwards could arise from cue or behavior differences in the two phases that were not present earlier). Lastly, we compared the population activity while rats engaged in different behaviors at different locations. Within each group of behaviors (i.e. waiting at a start nose port, crossing a bridge/route, consuming reward at a goal), mPFC displayed spatial selectivity (e.g. it differentiated the three bridges). Furthermore, we found that this spatial selectivity was embedded within a larger organization of activity in which these distinct, task-relevant groups were clearly separable from each other (Figure 6C, Figure 6—figure supplement 1).

Figure 6 with 1 supplement see all

Download asset Open asset

Previous work has shown sensory stimulus-specific delay period activity (delay activity) independent of motor plans (Romo et al., 1999; Wu et al., 2020) or resistant to distractor stimuli (Miller et al., 1996), and start-independent spatial (Brown et al., 2016; Guise and Shapiro, 2017; Watrous et al., 2018) or route-independent visuospatial (Saito et al., 2005) goal-specific delay activity in the PFC. A pair of studies (using a single start location; Spellman et al., 2015; Bolkan et al., 2017) showed no evidence of goal-selective delay activity in rodent PL mPFC. In one of these studies, motor planning was prevented, but not in the other (i.e. standard T-maze), and a study similar to the latter one (Kim et al., 2016) found goal-selective delay activity. Another study (Lara and Wallis, 2014) found no evidence of visual stimulus-selective delay activity in primate PFC. However, this study used color as the relevant stimulus dimension and found little evidence of color-selective activity even during stimulus presentation. Since PFC neurons have been shown in other cases to encode sample stimulus color (Buschman et al., 2012), this suggests that encoding of the stimulus during the sample period may be a prerequisite for observing stimulus-selective activity in the delay period. Spatial information is strongly represented in PFC in primates (Funahashi et al., 1989; Rainer et al., 1998; Saito et al., 2005; Lara and Wallis, 2014) and essentially all rodent studies, including here (Figure 2A, Figure 2—figure supplement 1, Figure 3—figure supplement 1); yet, we found no spatial goal-selective delay activity. Furthermore, our ability to decode the current start position and also the goal at various time points during the task (Figure 3—figure supplement 4), and the relatively high number of cells recorded simultaneously in our study, suggests that our data set was large enough to have detected an effect of sizes previously reported in the literature for simpler tasks.

In contrast to previous studies, we combined all elements of flexibility in one task – distractions, different start locations, and different unpredictable routes, as well as more than two goals. We found no evidence of goal-selective delay activity in fully trained animals or during relearning in any of the major forms previously documented over a wide range of parameters and across large numbers of simultaneously recorded cells in multiple mPFC subareas. Thus, these representations are unlikely to serve as general working memory correlates that can be employed under conditions of high behavioral flexibility, such as those often encountered in the real world. Whether animals in simpler tasks use such canonical representations of working memory to solve those tasks, employ an alternative strategy, or rely on an as-yet-undiscovered pattern of activity remains an open question. Our work therefore stresses the importance of employing more cognitively challenging tasks that allow dissociation between correlates of high-level cognitive variables and other task-related variables. For instance, the potential confounds resulting from such task-related variables are seen in the strikingly similar time course of goal decoding using either neural or behavioral tracking data outside of the well-controlled delay period (Figure 3—figure supplement 4). The search for a pure, flexible working memory correlate could focus both on other brain areas and on exploring as-yet-unobserved activity patterns or alternative memory mechanisms involving the mPFC. Such mechanisms could, for instance, be related to short-term synaptic plasticity (Mongillo et al., 2008), but differ from the previously reported covariance patterns (Barbosa et al., 2020) investigated here. Finally, our results suggest a role for mPFC in working memory tasks by representing task structure in terms of groups of behaviorally related elements (Jung et al., 1998; Yu et al., 2018; Kaefer et al., 2020), consistent with findings that the PFC forms long-term memories of learned stimulus categories (Freedman et al., 2001).

The data that support the main findings of this study are available at https://github.com/LeeA-Lab/Boehm_Lee_MSMGMR (copy archived at https://archive.softwareheritage.org/swh:1:rev:55d28ccf0459c33e6009d4cd66edb37e9b7870c4/).

1. 1. Baeg EH
   2. Kim YB
   3. Huh K
   4. Mook-Jung I
   5. Kim HT
   6. Jung MW

   (2003) Dynamics of population code for working memory in the prefrontal cortex

   Neuron 40:177–188.

   https://doi.org/10.1016/S0896-6273(03)00597-X

   • PubMed
   • Google Scholar
2. 1. Barbosa J
   2. Stein H
   3. Martinez RL
   4. Galan-Gadea A
   5. Li S
   6. Dalmau J
   7. Adam KCS
   8. Valls-Solé J
   9. Constantinidis C
   10. Compte A

   (2020) Interplay between persistent activity and activity-silent dynamics in the prefrontal cortex underlies serial biases in working memory

   Nature Neuroscience 23:1016–1024.

   https://doi.org/10.1038/s41593-020-0644-4

   • PubMed
   • Google Scholar
3. 1. Bolkan SS
   2. Stujenske JM
   3. Parnaudeau S
   4. Spellman TJ
   5. Rauffenbart C
   6. Abbas AI
   7. Harris AZ
   8. Gordon JA
   9. Kellendonk C

   (2017) Thalamic projections sustain prefrontal activity during working memory maintenance

   Nature Neuroscience 20:987–996.

   https://doi.org/10.1038/nn.4568

   • PubMed
   • Google Scholar
4. 1. Brown TI
   2. Carr VA
   3. LaRocque KF
   4. Favila SE
   5. Gordon AM
   6. Bowles B
   7. Bailenson JN
   8. Wagner AD

   (2016) Prospective representation of navigational goals in the human Hippocampus

   Science 352:1323–1326.

   https://doi.org/10.1126/science.aaf0784

   • PubMed
   • Google Scholar
5. 1. Buschman TJ
   2. Denovellis EL
   3. Diogo C
   4. Bullock D
   5. Miller EK

   (2012) Synchronous oscillatory neural ensembles for rules in the prefrontal cortex

   Neuron 76:838–846.

   https://doi.org/10.1016/j.neuron.2012.09.029

   • PubMed
   • Google Scholar
6. 1. Erlich JC
   2. Bialek M
   3. Brody CD

   (2011) A cortical substrate for memory-guided orienting in the rat

   Neuron 72:330–343.

   https://doi.org/10.1016/j.neuron.2011.07.010

   • PubMed
   • Google Scholar
7. 1. Frank LM
   2. Brown EN
   3. Wilson M

   (2000) Trajectory encoding in the Hippocampus and entorhinal cortex

   Neuron 27:169–178.

   https://doi.org/10.1016/S0896-6273(00)00018-0

   • PubMed
   • Google Scholar
8. 1. Freedman DJ
   2. Riesenhuber M
   3. Poggio T
   4. Miller EK

   (2001) Categorical representation of visual stimuli in the primate prefrontal cortex

   Science 291:312–316.

   https://doi.org/10.1126/science.291.5502.312

   • PubMed
   • Google Scholar
9. 1. Fujisawa S
   2. Amarasingham A
   3. Harrison MT
   4. Buzsáki G

   (2008) Behavior-dependent short-term assembly dynamics in the medial prefrontal cortex

   Nature Neuroscience 11:823–833.

   https://doi.org/10.1038/nn.2134

   • PubMed
   • Google Scholar
10. 1. Funahashi S
    2. Bruce CJ
    3. Goldman-Rakic PS

    (1989) Mnemonic coding of visual space in the monkey's dorsolateral prefrontal cortex

    Journal of Neurophysiology 61:331–349.

    https://doi.org/10.1152/jn.1989.61.2.331

    • PubMed
    • Google Scholar
11. Book

    1. Fuster JM

    (2015)

    The Prefrontal Cortex

    Academic Press.

    • Google Scholar
12. 1. Fuster JM
    2. Alexander GE

    (1971) Neuron activity related to short-term memory

    Science 173:652–654.

    https://doi.org/10.1126/science.173.3997.652

    • PubMed
    • Google Scholar
13. 1. Gill PR
    2. Mizumori SJ
    3. Smith DM

    (2011) Hippocampal episode fields develop with learning

    Hippocampus 21:1240–1249.

    https://doi.org/10.1002/hipo.20832

    • PubMed
    • Google Scholar
14. 1. Guise KG
    2. Shapiro ML

    (2017) Medial prefrontal cortex reduces memory interference by modifying hippocampal encoding

    Neuron 94:183–192.

    https://doi.org/10.1016/j.neuron.2017.03.011

    • PubMed
    • Google Scholar
15. 1. Harvey CD
    2. Coen P
    3. Tank DW

    (2012) Choice-specific sequences in parietal cortex during a virtual-navigation decision task

    Nature 484:62–68.

    https://doi.org/10.1038/nature10918

    • PubMed
    • Google Scholar
16. 1. Inagaki HK
    2. Fontolan L
    3. Romani S
    4. Svoboda K

    (2019) Discrete attractor dynamics underlies persistent activity in the frontal cortex

    Nature 566:212–217.

    https://doi.org/10.1038/s41586-019-0919-7

    • PubMed
    • Google Scholar
17. 1. Ito HT
    2. Zhang SJ
    3. Witter MP
    4. Moser EI
    5. Moser MB

    (2015) A prefrontal-thalamo-hippocampal circuit for goal-directed spatial navigation

    Nature 522:50–55.

    https://doi.org/10.1038/nature14396

    • PubMed
    • Google Scholar
18. 1. Johnson A
    2. Redish AD

    (2007) Neural ensembles in CA3 transiently encode paths forward of the animal at a decision point

    Journal of Neuroscience 27:12176–12189.

    https://doi.org/10.1523/JNEUROSCI.3761-07.2007

    • PubMed
    • Google Scholar
19. 1. Jun JJ
    2. Steinmetz NA
    3. Siegle JH
    4. Denman DJ
    5. Bauza M
    6. Barbarits B
    7. Lee AK
    8. Anastassiou CA
    9. Andrei A
    10. Aydın Ç
    11. Barbic M
    12. Blanche TJ
    13. Bonin V
    14. Couto J
    15. Dutta B
    16. Gratiy SL
    17. Gutnisky DA
    18. Häusser M
    19. Karsh B
    20. Ledochowitsch P
    21. Lopez CM
    22. Mitelut C
    23. Musa S
    24. Okun M
    25. Pachitariu M
    26. Putzeys J
    27. Rich PD
    28. Rossant C
    29. Sun WL
    30. Svoboda K
    31. Carandini M
    32. Harris KD
    33. Koch C
    34. O'Keefe J
    35. Harris TD

    (2017a) Fully integrated silicon probes for high-density recording of neural activity

    Nature 551:232–236.

    https://doi.org/10.1038/nature24636

    • PubMed
    • Google Scholar
20. Preprint

    1. Jun JJ
    2. Mitelut C
    3. Lai C
    4. Gratiy SL
    5. Anastassiou CA
    6. Harris TD

    (2017b) Real-time spike sorting platform for high-density extracellular probes with ground-truth validation and drift correction

    bioRxiv.

    https://doi.org/10.1101/101030

    • Google Scholar
21. 1. Jung MW
    2. Qin Y
    3. McNaughton BL
    4. Barnes CA

    (1998) Firing characteristics of deep layer neurons in prefrontal cortex in rats performing spatial working memory tasks

    Cerebral Cortex 8:437–450.

    https://doi.org/10.1093/cercor/8.5.437

    • PubMed
    • Google Scholar
22. 1. Kaefer K
    2. Nardin M
    3. Blahna K
    4. Csicsvari J

    (2020) Replay of behavioral sequences in the medial prefrontal cortex during rule switching

    Neuron 106:154–165.

    https://doi.org/10.1016/j.neuron.2020.01.015

    • PubMed
    • Google Scholar
23. 1. Kelemen E
    2. Fenton AA

    (2010) Dynamic grouping of hippocampal neural activity during cognitive control of two spatial frames

    PLOS Biology 8:e1000403.

    https://doi.org/10.1371/journal.pbio.1000403

    • PubMed
    • Google Scholar
24. 1. Kim D
    2. Jeong H
    3. Lee J
    4. Ghim JW
    5. Her ES
    6. Lee SH
    7. Jung MW

    (2016) Distinct roles of parvalbumin- and Somatostatin-Expressing interneurons in working memory

    Neuron 92:902–915.

    https://doi.org/10.1016/j.neuron.2016.09.023

    • PubMed
    • Google Scholar
25. 1. Lara AH
    2. Wallis JD

    (2014) Executive control processes underlying multi-item working memory

    Nature Neuroscience 17:876–883.

    https://doi.org/10.1038/nn.3702

    • PubMed
    • Google Scholar
26. 1. Liu D
    2. Gu X
    3. Zhu J
    4. Zhang X
    5. Han Z
    6. Yan W
    7. Cheng Q
    8. Hao J
    9. Fan H
    10. Hou R
    11. Chen Z
    12. Chen Y
    13. Li CT

    (2014) Medial prefrontal activity during delay period contributes to learning of a working memory task

    Science 346:458–463.

    https://doi.org/10.1126/science.1256573

    • PubMed
    • Google Scholar
27. 1. Maaten L
    2. Hinton G

    (2008)

    Visualizing data using t-SNE

    Journal of Machine Learning Research 9:2579–2605.

    • Google Scholar
28. 1. Maggi S
    2. Peyrache A
    3. Humphries MD

    (2018) An ensemble code in medial prefrontal cortex links prior events to outcomes during learning

    Nature Communications 9:1–12.

    https://doi.org/10.1038/s41467-018-04638-2

    • Google Scholar
29. 1. Miller EK
    2. Erickson CA
    3. Desimone R

    (1996) Neural mechanisms of visual working memory in prefrontal cortex of the macaque

    The Journal of Neuroscience 16:5154–5167.

    https://doi.org/10.1523/JNEUROSCI.16-16-05154.1996

    • PubMed
    • Google Scholar
30. 1. Miller EK
    2. Cohen JD

    (2001) An integrative theory of prefrontal cortex function

    Annual Review of Neuroscience 24:167–202.

    https://doi.org/10.1146/annurev.neuro.24.1.167

    • PubMed
    • Google Scholar
31. 1. Mongillo G
    2. Barak O
    3. Tsodyks M

    (2008) Synaptic theory of working memory

    Science 319:1543–1546.

    https://doi.org/10.1126/science.1150769

    • PubMed
    • Google Scholar
32. Book

    1. O'Keefe J
    2. Nadel L

    (1978)

    The Hippocampus as a Cognitive Map

    Oxford University Press.

    • Google Scholar
33. 1. Pastalkova E
    2. Itskov V
    3. Amarasingham A
    4. Buzsáki G

    (2008) Internally generated cell assembly sequences in the rat Hippocampus

    Science 321:1322–1327.

    https://doi.org/10.1126/science.1159775

    • PubMed
    • Google Scholar
34. 1. Pedregosa F
    2. Varoquaux G
    3. Gramfort A
    4. Michel V
    5. Thirion B
    6. Grisel O
    7. Blondel M
    8. Prettenhofer P
    9. Weiss R
    10. Dubourg V
    11. Vanderplas J
    12. Passos A
    13. Cournapeau D
    14. Brucher M
    15. Perrot M
    16. Duchesnay E

    (2011)

    Scikit-learn: machine learning in Python

    Journal of Machine Learning Research 12:2825–2830.

    • Google Scholar
35. 1. Pfeiffer BE
    2. Foster DJ

    (2013) Hippocampal place-cell sequences depict future paths to remembered goals

    Nature 497:74–79.

    https://doi.org/10.1038/nature12112

    • PubMed
    • Google Scholar
36. 1. Rainer G
    2. Asaad WF
    3. Miller EK

    (1998) Selective representation of relevant information by neurons in the primate prefrontal cortex

    Nature 393:577–579.

    https://doi.org/10.1038/31235

    • PubMed
    • Google Scholar
37. 1. Romo R
    2. Brody CD
    3. Hernández A
    4. Lemus L

    (1999) Neuronal correlates of parametric working memory in the prefrontal cortex

    Nature 399:470–473.

    https://doi.org/10.1038/20939

    • PubMed
    • Google Scholar
38. 1. Saito N
    2. Mushiake H
    3. Sakamoto K
    4. Itoyama Y
    5. Tanji J

    (2005) Representation of immediate and final behavioral goals in the monkey prefrontal cortex during an instructed delay period

    Cerebral Cortex 15:1535–1546.

    https://doi.org/10.1093/cercor/bhi032

    • PubMed
    • Google Scholar
39. 1. Sarel A
    2. Finkelstein A
    3. Las L
    4. Ulanovsky N

    (2017) Vectorial representation of spatial goals in the Hippocampus of bats

    Science 355:176–180.

    https://doi.org/10.1126/science.aak9589

    • PubMed
    • Google Scholar
40. 1. Siegel M
    2. Warden MR
    3. Miller EK

    (2009) Phase-dependent neuronal coding of objects in short-term memory

    PNAS 106:21341–21346.

    https://doi.org/10.1073/pnas.0908193106

    • PubMed
    • Google Scholar
41. 1. Spellman T
    2. Rigotti M
    3. Ahmari SE
    4. Fusi S
    5. Gogos JA
    6. Gordon JA

    (2015) Hippocampal-prefrontal input supports spatial encoding in working memory

    Nature 522:309–314.

    https://doi.org/10.1038/nature14445

    • PubMed
    • Google Scholar
42. 1. Wang XJ

    (1999) Synaptic basis of cortical persistent activity: the importance of NMDA receptors to working memory

    The Journal of Neuroscience 19:9587–9603.

    https://doi.org/10.1523/JNEUROSCI.19-21-09587.1999

    • PubMed
    • Google Scholar
43. 1. Watrous AJ
    2. Miller J
    3. Qasim SE
    4. Fried I
    5. Jacobs J

    (2018) Phase-tuned neuronal firing encodes human contextual representations for navigational goals

    eLife 7:e32554.

    https://doi.org/10.7554/eLife.32554

    • PubMed
    • Google Scholar
44. 1. Wikenheiser AM
    2. Redish AD

    (2015) Hippocampal theta sequences reflect current goals

    Nature Neuroscience 18:289–294.

    https://doi.org/10.1038/nn.3909

    • PubMed
    • Google Scholar
45. 1. Wimmer K
    2. Nykamp DQ
    3. Constantinidis C
    4. Compte A

    (2014) Bump attractor dynamics in prefrontal cortex explains behavioral precision in spatial working memory

    Nature Neuroscience 17:431–439.

    https://doi.org/10.1038/nn.3645

    • PubMed
    • Google Scholar
46. 1. Wood ER
    2. Dudchenko PA
    3. Robitsek RJ
    4. Eichenbaum H

    (2000) Hippocampal neurons encode information about different types of memory episodes occurring in the same location

    Neuron 27:623–633.

    https://doi.org/10.1016/S0896-6273(00)00071-4

    • PubMed
    • Google Scholar
47. 1. Wu Z
    2. Litwin-Kumar A
    3. Shamash P
    4. Taylor A
    5. Axel R
    6. Shadlen MN

    (2020) Context-Dependent decision making in a premotor circuit

    Neuron 106:316–328.

    https://doi.org/10.1016/j.neuron.2020.01.034

    • PubMed
    • Google Scholar
48. 1. Yu JY
    2. Liu DF
    3. Loback A
    4. Grossrubatscher I
    5. Frank LM

    (2018) Specific hippocampal representations are linked to generalized cortical representations in memory

    Nature Communications 9:2209.

    https://doi.org/10.1038/s41467-018-04498-w

    • PubMed
    • Google Scholar

Author details

1. Claudia Böhm

   Howard Hughes Medical Institute, Janelia Research Campus, Ashburn, United States

   Contribution

   Conceptualization, Visualization, Writing - original draft, Writing - review and editing, performed experiments and analyzed data

   For correspondence

   boehmc@janelia.hhmi.org

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-9802-1162

2. Albert K Lee

   Howard Hughes Medical Institute, Janelia Research Campus, Ashburn, United States

   Contribution

   Conceptualization, Supervision, Funding acquisition, Writing - review and editing

   For correspondence

   leea@janelia.hhmi.org

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-4332-8332

• 2,343

  views

• 333

  downloads

• 20

  citations

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