Estimating Rats Working Memory Capacity in an Object

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 The 14 th Annual Meeting of the Comparative Cognition Society Melbourne Beach, Florida, March, 2007

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

How much information can be stored and for how long? Open radial maze tasks confounded with direction cues n 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) n n Assess retention interval for one or two objects Non-spatial DMTS, DNMTS versions difficult to acquire.

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 animal’s working memory capacity

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 n Only food wells covered with objects baited with sunflower seeds n Pool of 60 objects Trials partitioned into study and test segments n Study segment: n object-covered wells, each baited with one seed n Test segment: one or more ‘new’ objects replace ‘old’ objects - baited with 6 seeds n I-min inter-segment-interval

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. n 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.

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.

Example for one rat

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.

Estimation of Number of Old Objects Recognized Phase 2: Six objects / Phase 3: Nine objects n Rat 2 B between 1 and 4 / 0 and 4 n Rat 2 D between 2 and 4 / 6 and 8 n Rat 3 A between 0 and 3 / none n Rat 3 B between 1 and 4 / 2 and 4 n Rat 3 C between 0 and 3 / 0 and 3 n Rat 3 D between 0 and 3 / 0 and 3

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.

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.

Example for one rat

Re-estimation of Number of Recognized Old Objects Phase 3: 1 st / 2 nd / 3 rd jackpot n n n 2 B: 2 D: 3 A: 3 B: 3 C: 3 D: 6 -8/3– 7/0– 4 7 -9/7– 9/6– 8 6 - 8 / 2 – 6 / none 7 -8/3– 7/1 -4 5– 8/2– 6/0 -3 7 -9/3– 6/0 -3

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

Thank You !