Hippocampal “Time Cells” Bridge the Gap in Memory for Discontiguous Events

MacDonald, Christopher J.; Lepage, Kyle Q.; Eden, Uri T.; Eichenbaum, Howard

doi:10.1016/j.neuron.2011.07.012

Cited by 1,040 publications

(1,129 citation statements)

References 68 publications

Supporting

Mentioning

1,050

Contrasting

Unclassified

Order By: Relevance

“…This sparse coverage approach differs from most other timing models in the spectral timing family (e.g., Grossberg & Schmajuk, 1989;Machado, 1997; but see Buhusi & Schmajuk, 1999), improves upon earlier versions of the MS TD model that used many more MSs per stimulus Ludvig et al, 2009), and stills one of the main criticisms of this class of models, that they suffer from the "infinitude of the possible" (Gallistel & King, 2009). In addition, as further support for such an approach to learning and timing, a spectrum of MS-like traces have recently been found during temporally structured tasks in the basal ganglia (Jin, Fujii, & Graybiel, 2009) and hippocampus (MacDonald, Lepage, Eden, & Eichenbaum, 2011).…”

Section: Discussionmentioning

confidence: 85%

Evaluating the TD model of classical conditioning

2012

View full text Add to dashboard Cite

The temporal-difference (TD) algorithm from reinforcement learning provides a simple method for incrementally learning predictions of upcoming events. Applied to classical conditioning, TD models suppose that animals learn a real-time prediction of the unconditioned stimulus (US) on the basis of all available conditioned stimuli (CSs). In the TD model, similar to other error-correction models, learning is driven by prediction errors-the difference between the change in US prediction and the actual US. With the TD model, however, learning occurs continuously from moment to moment and is not artificially constrained to occur in trials. Accordingly, a key feature of any TD model is the assumption about the representation of a CS on a moment-to-moment basis. Here, we evaluate the performance of the TD model with a heretofore unexplored range of classical conditioning tasks. To do so, we consider three stimulus representations that vary in their degree of temporal generalization and evaluate how the representation influences the performance of the TD model on these conditioning tasks.Keywords Associative learning . Classical conditioning . Timing . Reinforcement learning Classical conditioning is the process of learning to predict the future. The temporal-difference (TD) algorithm is an incremental method for learning predictions about impending outcomes that has been used widely, under the label of reinforcement learning, in artificial intelligence and robotics for real-time learning (Sutton & Barto, 1998). In this article, we evaluate a computational model of classical conditioning based on this TD algorithm. As applied to classical conditioning, the TD model supposes that animals use the conditioned stimulus (CS) to predict in real time the upcoming unconditioned stimuli (US) (Sutton & Barto, 1990). The TD model of conditioning has become the leading explanation for conditioning in neuroscience, due to the correspondence between the phasic firing of dopamine neurons and the reward-prediction error that drives learning in the model (Schultz, Dayan, & Montague, 1997; for reviews, see Ludvig, Bellemare, & Pearson, 2011;Maia, 2009;Niv, 2009;Schultz, 2006).The TD model can be viewed as an extension of the Rescorla-Wagner (RW) learning model, with two additional twists (Rescorla & Wagner, 1972). First, the TD model makes real-time predictions at each moment in a trial, thereby allowing the model to potentially deal with intratrial effects, such as the effects of stimulus timing on learning and the timing of responses within a trial. Second, the TD algorithm uses a slightly different learning rule with important implications. As will be detailed below, at each time step, the TD algorithm compares the current prediction about future US occurrences with the US predictions generated on the last time step. This temporal difference in US prediction is compared with any actual US received; if the latter two quantities differ, a prediction error is generated. This prediction error is then used to alter the associative ...

show abstract

Section: Discussionmentioning

confidence: 85%

Evaluating the TD model of classical conditioning

2012

View full text Add to dashboard Cite

show abstract

“…Increased preferred theta-phase variability could arise through a rate-phase transformation 19 and a reduced excitatory drive in VR due to a lack of repeatedly paired sensory and motor cues, as described below. The underlying network mechanism could thus generate motif-like activity under a variety of conditions, including hippocampal place cells from normal subjects 21,31,33 and transgenic mice with taupathy 40 , entorhinal cortical grid cells 38 , episode or time cells during wheel or treadmill running 22,23 , neural activity during rapid eye movement sleep 41 and neural activity during free recall in humans 42 .…”

Section: Discussionmentioning

confidence: 99%

“…Hence, to understand the mechanisms of the hippocampal spatial rate and temporal codes, it is important to determine whether the two can be dissociated during spatial exploration. In addition, dorsal hippocampal neurons are typically active for sustained periods lasting more than 1 second 1,2 , even under a variety of conditions [22][23][24][25] , and this sustained nature of activity has received little attention.…”

mentioning

confidence: 99%

Impaired spatial selectivity and intact phase precession in two-dimensional virtual reality

et al. 2014

View full text Add to dashboard Cite

During real-world (RW) exploration, rodent hippocampal activity shows robust spatial selectivity, which is hypothesized to be governed largely by distal visual cues, although other sensory-motor cues also contribute. Indeed, hippocampal spatial selectivity is weak in primate and human studies that use only visual cues. To determine the contribution of distal visual cues only, we measured hippocampal activity from body-fixed rodents exploring a two-dimensional virtual reality (VR). Compared to that in RW, spatial selectivity was markedly reduced during random foraging and goal-directed tasks in VR. Instead we found small but significant selectivity to distance traveled. Despite impaired spatial selectivity in VR, most spikes occurred within ~2-s-long hippocampal motifs in both RW and VR that had similar structure, including phase precession within motif fields. Selectivity to space and distance traveled were greatly enhanced in VR tasks with stereotypical trajectories. Thus, distal visual cues alone are insufficient to generate a robust hippocampal rate code for space but are sufficient for a temporal code. npg © 2015 Nature America, Inc. All rights reserved.1 2 2 VOLUME 18 | NUMBER 1 | JANUARY 2015 nature neurOSCIenCe a r t I C l e S same physical location in space can be approached from multiple different directions at different speeds. We thus investigated the contribution of distal visual cues only in determining selectivity in such an experimental setup. RESULTS Nature of spatial selectivity of hippocampal responsesWe measured hippocampal activity during a two-dimensional randomforaging task in RW and VR [35][36][37] with similar distal visual cues (Fig. 1a). In VR, the rats were body-fixed, i.e., their bodies were held in place, with a harness on a floating ball, allowing for head movements but precluding full-body turns, thus minimizing vestibular cues (Online Methods) 21,36 . Rats quickly learned to avoid the virtual edges entirely on the basis of visual cues 36 and spent a similar amount of time away from the edges and in the center of the platform (Fig. 1a) compared to in RW. We measured the activity of 1,066 and 1,238 principal neurons in RW and VR, respectively, in the dorsal CA1 of four rats under a variety of conditions (Online Methods). Neurons fired vigorously in restricted regions of space in RW, as expected (Fig. 1b,c, Supplementary Fig. 1a and Supplementary Video 1) 1 . In contrast, the neurons showed little spatial selectivity in VR during random foraging (Fig. 1b,d and Supplementary Fig. 1b).Across the ensemble, neurons had moderately reduced (25%) mean firing rates but greatly reduced (68%) peak firing rates in VR ( Fig. 2a and Supplementary Fig. 2a). Neurons in VR also had greatly reduced spatial information content (75%), stability (59%), sparsity (42%) and coherence (40%) (Fig. 2b-d and Supplementary Fig. 2b,c) compared to spatially localized, stable and sparse RW rate maps (Fig. 2c). Although the mean firing rate was inversely correlated with information content (Supplementary Fig. 2d)...

show abstract

“…The hippocampus is thought to be involved in navigation and memory in both spatial and temporal space [71][72][73][74][75]. In addition, both hippocampal 'time cells' that integrate episodic information across events [19,76] and a hippocampal neuronal coding mechanism that represents the recency of an experience over extended intervals [77] have been reported. These constructs demonstrate the importance of the hippocampus to interval timing per se; however, from an evolutionary standpoint, it would be hard to believe that each brain area works in isolation.…”

Section: Discussion (A) Dorsal and Ventral Hippocampal Lesions Differmentioning

confidence: 99%

Comparison of interval timing behaviour in mice following dorsal or ventral hippocampal lesions with mice having δ -opioid receptor gene deletion

Yin

Meck

2014

Phil. Trans. R. Soc. B

View full text Add to dashboard Cite

Mice with cytotoxic lesions of the dorsal hippocampus (DH) underestimated 15 s and 45 s target durations in a bi-peak procedure as evidenced by proportional leftward shifts of the peak functions that emerged during training as a result of decreases in both ‘start’ and ‘stop’ times. In contrast, mice with lesions of the ventral hippocampus (VH) displayed rightward shifts that were immediately present and were largely limited to increases in the ‘stop’ time for the 45 s target duration. Moreover, the effects of the DH lesions were congruent with the scalar property of interval timing in that the 15 s and 45 s functions superimposed when plotted on a relative timescale, whereas the effects of the VH lesions violated the scalar property. Mice with DH lesions also showed enhanced reversal learning in comparison to control and VH lesioned mice. These results are compared with the timing distortions observed in mice lacking δ -opioid receptors (Oprd1 −/− ) which were similar to mice with DH lesions. Taken together, these results suggest a balance between hippocampal–striatal interactions for interval timing and demonstrate possible functional dissociations along the septotemporal axis of the hippocampus in terms of motivation, timed response thresholds and encoding in temporal memory.

show abstract

Hippocampal “Time Cells” Bridge the Gap in Memory for Discontiguous Events

Cited by 1,040 publications

References 68 publications

Evaluating the TD model of classical conditioning

Evaluating the TD model of classical conditioning

Impaired spatial selectivity and intact phase precession in two-dimensional virtual reality

Comparison of interval timing behaviour in mice following dorsal or ventral hippocampal lesions with mice having δ -opioid receptor gene deletion

Contact Info

Product

Resources

About