Fig 1.
Behavioral protocols used in Experiments 1 to 3.
(A) Drawing code used to represent the radial maze. (B) Classic LIWM and HIWM training (see also Materials and methods) performed in Experiment 1, illustrated with one session (four trials). The pair of arms used for each trial is different for the LIWM training while it is identical for HIWM training. Training was spaced, meaning that two sessions were separated by 24h with a total of 10 sessions. The sequence of arms presentation shown in (B) is only informative and does not represent the sequence used each day. This sequence is different every day and pseudo-randomly determined by the experimenter. (C) In Experiment 2, the two tasks used were the same as in A), except that training was massed (mLIWM and mHIWM): each session was separated by 10 min, with 10 sessions on day 1 and 10 other sessions on day 2. (D) In Experiment 3, we used a spaced training as in A), but the arms used during the choice phase were no longer adjacent but formed a 90° angle (LIWM90 and HIWM90), as illustrated in the example.
Fig 2.
Performance drop in the final sessions of HIWM training reveals accumulation of PI across days, and hence long-lasting WM.
(A) Evolution of the percentage of correct choices across blocks of two sessions, for rats trained in the LIWM or the HIWM task. ANOVA revealed a significant Group effect (*** p < 0.001), due to the drop of performance in the HIWM group across days revealed by the split-by Group analysis. (B) Percentage of correct choices made by trial for the first and the last 5 days of training, for the two groups. For Days 1 to 5 ANOVA revealed no Group effect but a significant Trial effect with a drop of performance on trial 4 for the HIWM group (Sidak's post-hoc test, * p < 0.05), indicative of short-term PI. On the contrary for days 6 to 10 there was a significant Group effect, due to lower performance in the HIWM group from the very first trial (Sidak's post-hoc test, ** p < 0.01). This reduction of performance in the HIWM task could not be attributable to short-term PI, but to accumulation of PI across days indicating that WM was not entirely reset at the end of each training and could partly last at least 24h.
Fig 3.
Massed training prevents performance drop in the HIWM group but long-term proactive interference appears after a 24 hour-interruption that allows consolidation of the content of WM into a more durable form of memory.
Percentage of correct choices across blocks of two sessions with a massed training of 10 sessions by day during two days. ANOVA revealed a significant Group effect only on day 2 (*** p < 0.001) due to lower performance in the HIWM group. A split-by Group analysis of performance on day 2 showed a significant improvement in HIWM trained animals leading to similar level of performance between the two groups on final block 10.
Fig 4.
Reducing the need for pattern separation prevents performance drop in the HIWM group.
Percentage of correct choices across blocks of two sessions with a spaced training in the two tasks during which the angle between two choice arms was 90° to reduce the need for pattern separation relative to the same tasks with 45° spaced-arms as used in experiment 1. ANOVA revealed no Group effect (NS: non-significant).