Declarative Memory: Lecture 10 Flashcards
“Place” vs “Response”
Imagine rats having to choose between two arms of a maze, on of which contains a reward.
The “place” strategy is using a cog. map to determine the location of food (if rotated, the rat will turn the opposite way in the maze to go in the same compass direction).
The “response” strategy is developing a habit of turning to e.g. the right (if rotated, rat will still turn right).
These are cognitivist and behaviourist explanations, respectively.
Place strategy is used if salient extra-maze cues are present (declarative memory). Response strategy is used when cues are absent or rats are over-trained (procedural memory)
T maze experiment. Areas of brain involved in two strategies.
Rats trained for 1 week in maze. Then rotated and tested.
Then trained for a further week using the original position. Then rotated and tested again.
Rats initially adopt place strategy. After the second round of training, adopted response strategy.
Initially guided by cog. map. Overtraining –> response “habit” developing.
Also found brain circuits involved: lidocaine injected into hippocampus (essentially a reversible lesion) - far fewer rats used “place” strategy on first test (striatum injection had no effect). Injection into striatum - far fewer used “response” strategy on second test.
So place = hippocampus, response = striatum
Modified T Maze
Top of the T joined to bottom at either end. Rats must alternate between left and right turns to get reward.
Recorded from place cells. Found that some cells with fields on the “stem” fired according to the direction that was required to get reward. I.e. some fired most when the turn was going to be left, others when it was going to be right. They always fired in the same place.
Other cells fired as “typical” place cells.
DNMTS task with rats
Cup filled with sand scented with 1 of 9 odours and placed at 1 of 9 locations. Reward place in sand when odour did not match that of the previous trial. Firing of hippocampal neurons recorded - could be correlated with position (spatial) and odour (non-spatial) information.
65% of cells fired in association with one or more task variable. 1/3 of these = spatial, another 1/3 = non-spatial. Others fired to particular odour or to match/mismatch condition. Some cells fired to particular cup position.
Most cells fired to combo of cup position, match/mismatch and start of approach to cup.
Suggests that hippocampus combines elements of spatial, temporal and object cues into episodic memories.
Relational memory
Associations BETWEEN memories.
Trained rats using pairs of odourised sands. One of the two odours would signal food for that pair.
A>B>C>D>E
Rats tested for transitive inference. Required for e.g. B vs D but NOT for A vs E (because A is always reward and E is never rewarded). Showed transitive inference is present in rats.
Then lesioned (either parahippocampal region to remove inputs to hippocampus or fornix to remove outputs). Both impaired transitivity (original pairs and A/E pair were normal).
“Memory Space” hypotheses
Eichenbaum suggests that:
1) The hippocampal network represents episodic memories as a sequence of events AND places. Activity of each cell encodes small segment of the episode, including information about relevant stimuli, behaviour and position.
2) Memories may overlap at common points e.g. entering a particular spatial location
Hippocampus can encode all aspects of an episode (what, where, when), represented by firing neurons.
Encoding memory sequences (2 types of network)
1) Autoassociative network: neurons send projections only back onto themselves. Works for single items (each set of neurons can represent one element) but can’t encode sequences
2) Heteroassociative network: axons connect to different cells, not back on themselves. Information can be recalled in the order it was learned. Allows for sequences.
Pattern separation vs pattern completion
If we have two similar inputs e.g. a black cat and a panther, we want them to have separate representations. Dentate gyrus is responsible for this. Allows us to discriminate between highly similar stimuli.
A full memory can also be evoked from a partial set of inputs (pattern completion). CA3 is responsible for this.
Role of gamma and theta
Theta and gamma frequency oscillations which occur in the brain can interact with each other. It has been suggested that dual oscillations form a code for representing multiple items in an ordered manner.
Each memory is represented by a group of neurons firing, associated with a single gamma cycle/wave. Memories are activated in order e.g. A–>G, each associated with a single gamma wave, all of which ride on a single theta wave. This sequence is repeated with every theta wave.
In the hippocampus, 7 +/-2 gamma waves are seen with each theta, which is the same range for number of items that can be kept in working memory.
Place cells and theta
As rat runs along linear track, place cell firing increases into the place field. At first (as they enter the place field), they fire at the start of each theta phase. Then firing gets progressively earlier with each theta cycle (fires at the trough when maximally active/at the centre of the place field and even earlier as they exit the place field). This is theta phase precession.
This is a temporal code whereby time of firing relative to the theta phase identifies location (position of the animal is encoded by the phase of theta at which the place cell fires)
Place cell populations
By looking at the activity across all place cells simultaneously we see that the history of the animal’s movement is encoded by the sequential firing pattern across cells.
i.e. mean firing of place cells for particular place field creates a curve. Curves for different place fields overlap.
This means that when at the centre of a given place field, there is maximal firing of place cells for that given field (at the trough of theta) but also firing at different phases of theta for places that have previously been visited AND places that are about to be visited.
This organized, phase-specific firing pattern thereby provides a running record of where the animal was, is and will be. This pattern of activity = episodic memory
phase precession = a substrate for memory encoding
Sternberg Paradigm
Set of digits memorised –> delay –> presented with probe digit –> in the set, y/n?
Mean reaction time increases linearly with memory load.
Explained by a serial scanning mechanism - each item in the set compared with the probe, one by one.
Reaction time increases by 38ms per additional digit, which is the same as the time between gamma cycles. Suggests that scanning is clocked by gamma oscillations - one item is compared per gamma cycle.
Theta paces the gamma scans.
Theta during Sternberg task
Theta can be seen in EEG and MEG recordings (gamma cannot).
MEG recording from 61 cortical locations (humans). Given Sternberg task (1, 3, 5 or 7 digits) or control (sequence of crosses).
Found that theta increases in amplitude with task engagement (i.e. Sternberg rather than sensory control task). Theta amp also increases with number of items to be remembered, but this is saturated once the limit for working memory is reached (7 items).
Supports the role of theta in working memory.
