![[Home]](http://www.bayleshanks.com/cartoonbayle.png)
\documentclass{article}
%% NOTE: before running EasyLatex? on this, remove comments within [[[...?]]] and this comment
\oddsidemargin -0.5in \evensidemargin -0.5in
\textwidth 7.4in
\topmargin -.3in \textheight 10in %% \footheight 0.1in
\usepackage{amsmath} %% pkg required by EasyLatex? \usepackage{amsfonts} %% pkg required by EasyLatex? \usepackage{amssymb} %% pkg required by EasyLatex? \usepackage{psfrag} %% pkg required by EasyLatex? \usepackage[dvips]{graphicx} %% pkg required by EasyLatex? \begin{document}
Bayle Shanks
In this paper I'll review the literature on the involvement of the hippocampus in temporal sequence learning. [[[Besides_being_interesting_in_its_own_right,_it_may_help_decide_the_question_of_whether_animals_have_episodic_memory_(since_organization_into_temporal_sequences_is_a_primary_characteristic_of_episodic_memory),_and_may_serve_as_a_fundamental_computational_primitive_for_other_processes.?]]]
[[[A_range_of_evidence_shows_that_the_hippocampus_is_involved_in_explicit_memory_for_sequences_after_a_single_exposure?]]]
Multiple studies have shown that rats with hippocampal lesions have difficulty remembering sequences of spatial locations. [[[Rats_who_traverse_the_arms_of_a_maze_and_who_are_later_asked_to_indicate_which_arm_came_earlier_in_a_forced_choice_have_trouble_after_removal_of_their_hippocampus.?]]]However, since the hippocampus has been hypothesized to play a special role in spatial memory, sequence memory studies using other sensory modalities (such as odor) are the most convincing, so we'll focus on those.
At least two studies[[[(Kesner_et_al_'02),_(Fortin_et_al_'02?]]] have shown that hippocampal lesions impair a rat's ability to learn which odor came first in sequences of odors. One of these studies (Fortin et al '02) also provided a single dissociation between familiarity and memory of sequential order; that is, hippocampal rats were able to remember which odors were in the sequence, but had deficits in remembering which ones came before others.
[[[ There are other converging lines of evidence, however. In (Agster et al '02) rats were trained to distinguish two sequences of odors. The study showed that this task was also hippocampal-dependent. ]]]
In addition to those two, (Agster et al '02) showed that hippocampal rats had deficits keeping track of which sequence of odors was currently transpiring. Rats were trained on two different odor sequences, both of which contained the same overlapping elements at one point. Then one of the sequences was presented, and just after the overlapping elements, the rats were made to indicate which of the two sequences they were in the middle of. That is, the rats were supposed to figure out which sequence was happening before the overlapping elements, and then to display this knowledge after the presentation of the overlapping elements.
[[[In_(Agster_et_al_'02),_the_hippocampus's_role_could_be_explained_by_a_general_role_in_maintaining_context_,_rather_than_a_specific_function_in_the_memory_of_temporal_sequences._TODO?]]]
(Lee and Wilson '02) made rats run through the place fields of hippocampal CA1 place cells in a fixed order. This caused the place cells to fire in a fixed order. During sleep, the cells replayed this sequence. A number of other studies have also found reply of temporal sequences in the hippocampus during sleep. [[[However,_(Lee_and_Wilson_'02),_and_a_number_of_other_studies,_find_that_memories_of_sequential_spatial_paths_are_replayed_during_slow_wave_sleep._This_is_just_the_sort_of_representation_that_one_might_expect_to_find_if_the_hippocampus_were_involved_in_temporal_sequence_memory,_and_so_serves_as_an_important_converging_line_of_evidence.?]]]
Note, however, that the brain definitely contains other mechanisms for temporal sequence learning outside of the hippocampus. The basal ganglia, the cerebellum, and the prefrontal cortex have all been implicated in sequence learning. For instance, both human hippocampal amnesiacs and hippocampal rats are able to learn sequences using implicit/procedural memory on the Serial Reaction Time Task (Christie and Dalrymple-Alford '04[Christie and Dalrymple-Alford '04]).
Because this study is one of the few to attempt to show that the hippocampus is not just doing something which has some passing relationship to temporal sequence learning tasks, but rather is essentially doing temporal sequence learning, it deserves a closer look.
The study assays sequence memory by testing pairs of odors and making the rat indicate which odor came first in the sequence (the ordering test). Instead of memorizing a sequence explicitly, however, rats might perform this task by comparing the strength of their memory traces from the prior odor presentations (the more recent item would be the one which the rats "remember more"). The study attempts to eliminate this possibility by showing that the hippocampal rats are not deficient in a different task which is thought to rely on evaluation of memory trace strength.
This other task tests if rats can distinguish odors which were present in the sequence from odors which were not there at all (the familiarity task). The result is that they can. In other words, the rats have trouble telling which stimuli are more recent, but they do not have trouble distinguishing novel stimuli from familiar stimuli. This suggests that damage to the hippocampus does not impair memory trace strength. Therefore, the normal rats were comparing the ordering of stimuli using some other mechanism, presumably sequence memory.
However, single dissociations can be misleading. There is the possibility that both tasks in fact rely on memory trace strength comparison, and that no sequence representation is learned in either task. In this case, maybe hippocampal animals have damage to memory trace strength machinery (rather than to specialized sequence learning machinery), and the ordering task is just a more sensitive assay for this damage.
The study attempts to mitigate this danger by comparing probes in the two tasks in which control rats' performance was equal (i.e. controls made a mistake about 1 in 5 times on both tasks). However, equal error rates cannot guarantee that the ordering task does not somehow make more demands on memory trace strength machinery. For example, the ordering task could demand a finer granularity of temporal precision (see also the "temporal reasoning" hypothesis, below).
A double dissociation (in this case, the discovery of an experimental condition in which the familiarity task was impaired without affecting performance on the ordering task) would have made for a stronger argument that the deficit in the ordering task was due to problems with sequence-specific memory, rather than to problems with memory trace strength. So, while this study is an important clue, it does not by itself resolve the question of whether the hippocampus encodes or stores memories of temporal sequences.
Evidence of hippocampal involvement in temporal sequence learning tasks is not sufficient for us to conclude that there is machinery specialized to temporal sequences within the hippocampus. It is possible that the hippocampus is actually fulfilling a more general function which would explain its involvment in temporal sequence learning as a special case. Four such postulated functions are episodic memory, declarative memory, relational inference, and temporal reasoning.
Episodic memory: Episodic memory is thought to be organized as temporal sequences of events. Hence, the episodic memory system must contain some mechanism for temporal sequence memory. Human hippocampal amnesiacs have implied that the hippocampus is involved with episodic memory.
Declarative/explicit memory: Many sequence recall tasks address explicit memory. It is thought that the hippocampus is involved with explicit memory. If the hippocampus performed some general function needed for all explicit memory, its removal would impair sequence recall in many tasks.
Relational inference: If you learn that $A < B < C$, you can reason that $A < C$. This is an example of __transitive inference__, which is one kind of reasoning about ordering relations. [[[Formally, ordering relations don't have to be a linear sequence; for instance, if "$<$" represents "is child of", then "you $<$ parents" and "sister $<$ parents", but neither "you $<$ sister" nor "sister $<$ you". ]]]Perhaps the hippocampus implements a generalized mechanism for storing or reasoning with ordering relations. As a special case, this would allow the hippocampus to assist with storage and recall of temporal sequences (in which the relation is "came earlier than"), even if it had no machinery specialized for linear temporal sequences.
At least three imaging studies show hippocampal activation during transitive inference tasks, although one does not. There was a study which purported to show that hippocampal rats were deficient in a transitive inference task, but now it looks like the task was flawed and the rats weren't actually doing transitive inference.
Temporal reasoning: Another hypothesis is that the hippocampus contains some general machinery for perceiving, remembering, or reasoning about time. Hippocampal rats were deficient in a task which required them to distinguish between a 2-second and 8-second stimulus presentation[[[(Jackson_et_al_'98)?]]]. [[[Hippocampal_damage_in_humans_causes_problems_with_the_memory_of_temporal_distance_between_events.?]]]Also, subjects with hippocampal damage were deficient in trace conditioning[[[Clark_and_Squire_'98?]]]. Perhaps the sequence learning deficits are due to an inability to remember the time at which something happened, or an inability to make fine discriminations between remembered times.
In addition to experimental evidence, there are a number of computational models which hypothesize that a primary function of the hippocampus is temporal sequence learning. Theorists have postulated that the involvement of the hippocampus in many of the kinds of tasks which it seems to participate in could be explained if the hippocampus was involved in sequence memory and sequential pattern completion[[[Levy?]]].
Many of the models are based on CA3, although the experimental evidence doesn't definitively distinguish between CA3 and CA1. Many of the models are intimately related to the phenomenon of hippocampal theta phase precession.
One of characteristic oscillations, or rhythms, of the hippocampus is the theta oscillation (4-12 hz). It seems to be imposed on the hippocampus by the medial septum-diagonal band of Broca[[[Buzsaki_'02?]]]. It is seen during exploratory activity and REM sleep. It can be evoked and its phase can be reset by the presentation of sensory stimuli[[[Tesche_and_Karhu_PNAS_'97?]]].
Theta phase precession is a phenomenon relating the theta rhythm, the firing of individual neurons, and place fields. A hippocampal CA1 place cell's firing rate markedly increases when the animal is within a certain spatial area in the context of a certain task. This area is called the "place field" of the place cell. Consider the phase of each spike of the place cell in relation to the hippocampal theta oscillation. Theta phase precession is a phenomenon where the phase of neural firing moves earlier each theta cycle as the animal moves through the place field.
Theorists have postulated various explanation for the theta rhythm and theta phase precession in conjunction with models of hippocampal sequence learning. Functions of these phenomena have been related to:
* holding many memories in mind at once
* rehersing nearby events in a remembered sequence
* a "program counter" which determines which step of a learning algorithm is currently being executed.
* a carrier frequency for neural coding based on spike phase
I don't have space to go into all of the models, but I'll briefly discuss one, which I'll call the "theta multiplexed model" (Jensen and Lisman '96). It proposes that the hippocampus is storing temporal sequences of stimuli, and constantly rehersing the nearby elements in each sequence as elements in the sequence are being experienced. This model relates the theta and gamma (20-60 hz) hippocampal oscillations. It notes that there are about 7 gamma cycles within each theta cycle. It postulates that this constitutes a multiplexed short-term memory buffer; each of the seven gamma subcycles corresponds to one memory. So, during each theta cycle, the brain runs through each of these stored memories. These memories represent a stored temporal sequence of spatial locations, and they include the current location.
The model demonstrates how a spike phase code lends itself to storage and retrieval of stimuli, and predicts that phase precession would be observed as one moves through the sequence.
[[[Hippocampal lesions only prevents subjects from learning sequences when the sequence recall task involves explicit memory ((http://www.jneurosci.org/cgi/content/abstract/24/5/1034 Christie and Dalrymple-Alford '04)), (Gazzaniga et al '02), although see (http://dx.doi.org/10.1038/sj.mp.4001424 Schendan et al '03: "Sequence? What sequence?")\footnote{although some imaging studies do show hippocampal involvement in this task nevertheless}]]]
[[[TODO: "refs" AND "references"]]]
\newpage \emph{Please note that we were only permitted to have 5 citations in the above paper}
While preparing this paper, I also created a preliminary bibliography of subjects related to the topic, with over 125 citations divided into over 25 cross-linked category listings. In hopes that it may be useful for future researchers, I made it available online as part of a larger database; the entry point is at:
http://purl.net/net/neurowiki/TemporalSequenceLearningAndTheHippocampus
Kara L. Agster, Norbert J. Fortin, and Howard Eichenbaum. The Hippocampus and Disambiguation of Overlapping Sequences. J. Neurosci. 22: 5760-5768; doi:20026559. (2002)
Michael A. Christie, and John C. Dalrymple-Alford. A New Rat Model of the Human Serial Reaction Time Task: Contrasting Effects of Caudate and Hippocampal Lesions. J. Neurosci. 24: 1034-1039; doi:10.1523/JNEUROSCI?.3340-03.2004 (2004)
Fortin NJ, Agster KL, Eichenbaum HB. Critical role of the hippocampus in memory for sequences of events. Nat Neurosci. 2002 May;5(5):458-62.
Jensen, O. and Lisman, J. E. Hippocampal CA3 region predicts memory sequences: accounting for the phase precession of place cells. Learning and Memory 3, 279-287 (1996)
Lee AK, Wilson MA. Memory of sequential experience in the hippocampus during slow wave sleep. Neuron. 2002 Dec 19;36(6):1183-94. (2002) \end{document}
