Mice extrapolate temporal information based on previously learned spatiotemporal mappings: an asymmetrical case

Abstract

One of the computational affordances of isomorphic magnitude representations is the extrapolation of temporal information based on previously experienced spatiotemporal pairings. We initially trained mice on the association of two intervals (10 s and 30 s) with two hoppers (H2 and H4, counterbalanced) in a five-choice nose-poke box with the following setup. One of the three novel hoppers (H1) neighbored H2 only, the other novel hopper (H5) neighbored H4 only, and the third novel hopper (H3) neighbored H2 and H4 (H1Novel -> H2Trained -> H3Novel -> H4Trained -> H5Novel). During test trials, one of the five hoppers was illuminated. We estimated the trial time at which the anticipatory response rate was maximal (peak time) separately for each hopper. Mice extrapolated temporal information only in a forward fashion; the peak time for H5 was longer than that for H4. Mice did not extrapolate temporal information backward; the timed response curves in H1 and H3 were closely similar to those in H2. Thus, our findings suggest that mice can extrapolate temporal information, but also indicate that the computations underlying this process are directionally constrained. We discuss the possible reasons behind asymmetrical extrapolation.