Estimating Rats

1
Estimating Rats Working Memory Capacity in an
Object Recognition Foraging Task

• Jerome Cohen, Xue Han, Anca Matei, Vara
  Parameswaran
• Department of Psychology, University of Windsor
• Myron Hlynka
• Department of Mathematics and Statistics,
  University of Windsor

The14th Annual Meeting of the Comparative
Cognition Society Melbourne Beach, Florida,
March, 2007
2
Difficulties in assessing memory in non-verbal
animals

• DMTS, DNMTS, radial maze, food cache recovery
  errors not always due to loss in working memory
  (Thorpe, Jacova, Wilkie, 2004 Wilkie, Willson,
  Carr, 1999)
• Overestimation (Cole Chappell-Stephenson, 2004)
  or underestimation (Cohen et al. 2004) of spatial
  location working memory in radial maze task

3
How much information can be stored and for how
long?

• Open radial maze tasks confounded with direction
  cues
• Rats may retain how far away arms are from a
  given direction sampled rather than specific arm
  locations.
• Object recognition tasks DMTS, DNMTS, novel
  object recognition (see Mumby, 2001)
• Assess retention interval for one or two objects
• Non-spatial DMTS, DNMTS versions difficult to
  acquire.

4
Five important characteristics of a good working
memory task for rats

• Object recognition task should be easy to acquire
• Working memory should be biologically more
  relevant to the animal than search response
  algorithms
• Reward new object recovery so as to prevent
  premature search termination
• Large pool of objects to prevent proactive
  interference from object repetition over
  successive trials
• Assess individual animals working memory
  capacity

5
Our Object Location Recognition Task (version 2
of Cohen et al.2006)

• Subjects Six Long-Evans male hooded rats
• Apparatus Large enclosed square foraging area
  with 25 (5 x 5) covered food wells
• Only food wells covered with objects baited with
  sunflower seeds
• Pool of 60 objects
• Trials partitioned into study and test segments
• Study segment n object-covered wells, each
  baited with one seed
• Test segment one or more new objects replace
  old objects - baited with 6 seeds
• I-min inter-segment-interval

6
(No Transcript)
7
Three Phases each 48 trials

• Phase 1 trial One new object in test segment
  replaces one of the three old objects from
  previous study segment.
• Object locations randomized over trials but not
  within a trial
• Phase 2 trial Two new objects in test segment
  replace two of the six old objects from previous
  study segment
• Phase 3 trial Three new objects in test segment
  replaces three of nine old objects from previous
  study segment.

8
(No Transcript)
9
(No Transcript)
10
Dependent Variable and Measure of Estimated No.
of Old Objects Retained

• Number of choices to find all jackpot objects
• Observed distribution of cumulative proportion of
  trials rat finds all jackpots in each phase.
• Compare observed distributions with theoretical
  cumulative probability distributions for finding
  all jackpots based on no memory (chance) to
  perfect working memory of all old objects from
  the study segment (K-S tests)
• Estimate of recognized old objects based on 95
  confidence interval around the observed
  distribution.

11
(No Transcript)
12
Example for one rat
13
Summary of Results

• Phase 1 No rat shows above chance performance in
  finding one jackpot out of three objects.
• Phase 2 All rats show above chance performance
  in finding two jackpots out of six objects.
• Phase 3 Five rats show above chance performance
  in finding three jackpots out of nine objects.

14
Estimation of Number of Old Objects Recognized

• Phase 2 Six objects / Phase 3 Nine objects
• Rat 2B between 1 and 4 / 0 and 4
• Rat 2D between 2 and 4 / 6 and 8
• Rat 3A between 0 and 3 / none
• Rat 3B between 1 and 4 / 2 and 4
• Rat 3C between 0 and 3 / 0 and 3
• Rat 3D between 0 and 3 / 0 and 3

15
Conclusions

• Increasing set or patch size promotes working
  memory processes relative to other search
  strategies.
• Our rats seem to be able to recognize about 2 or
  3 old objects.

16
Does recognition of old objects change as more
jackpots are found?

• Compare observed distributions for finding all
  three jackpots with those for finding two
  jackpots or only one jackpot from nine objects.
• Generate new theoretical distributions for number
  of choices to find the first, second, or third
  jackpot based on no memory to perfect memory for
  old objects.

17
Example for one rat
18
Re-estimation of Number of Recognized Old Objects

• Phase 3 1st / 2nd / 3rd jackpot
• 2B 6 - 8 / 3 7 / 0 4
• 2D 7 - 9 / 7 9 / 6 8
• 3A 6 - 8 / 2 6 / none
• 3B 7 - 8 / 3 7 / 1 - 4
• 3C 5 8 / 2 6 / 0 - 3
• 3D 7 - 9 / 3 6 / 0 - 3

19
Conclusion and Question

• As rat finds more jackpots, its performance for
  recognizing old objects declines
• Is this effect due to a loss in rats working
  memory capacity or switching to other search
  strategies?

20
Thank You !