The sorter was activated from phase two onwards. During training phases, mice were sorted into the operant compartment for between 2 and 9 sessions per day (median 5 sessions) and the operant module was occupied for nearly 14 h per day with mice performing sessions. Mice, especially at the beginning, tended to crowd together in the sorter, which led to the sorting procedure being aborted. Thus, for each successful visit to the operant compartment, a mouse needed an average of 5 entries to the sorter. A mouse completed 122 trials on average per day (98–167 trials). The minimum of 50 trials that were required for each of the training phases three and four were completed within a single day by all mice except one. Odor Pair Discrimination Acquisition and Reversal

Analysis of mouse performance during subsequent odor discriminations indicated that animals not only successfully learnt to discriminate each odor pair but also formed a learning set, as the median number of errors between the first and forth odor pair discrimination acquisition halved. Very few studies have investigated olfactory learning set formation in mice to date; while one study reported a 70% decrease in errors between the first and second discrimination ( Larson et al., 2003), two other studies observed effect sizes similar to those reported here ( Larson and Sieprawska, 2002; Patel and Larson, 2009).In this study, we used an automated olfactory task and a home cage with a group of ID-chipped mice. We then connected the two compartments by an RFID-based animal sorter. This sorter allowed continuous testing throughout the 24-h period by giving mice individual access to the test compartment automatically. Experimental sessions were thus self-initiated and voluntary as well as free from the interference of cagemates. As shown previously, rodents readily adapt to the animal sorting process with minimal sorter training ( Winter and Schaefers, 2011; Rivalan et al., 2017). The animal study was reviewed and approved by the Animal Welfare Officer of Humboldt University. Author Contributions A further step along is choosing your own criteria for sorting, for example when grouping a collection of pebbles. No. 2 son was inspired by a trip to the mineral galleries at the Natural History Museum to create his own display table in his bedroom. But how to organise them? By colour? By size? By material? Over the next few days he changed the categories several times but the choices were always logical. And not a pom pom or coloured bowl in sight. Best sorting toys for preschoolers The sequence of odor pairs and the initial S+ odor werepseudo-randomly assigned to the mice for counterbalancing. We firstcreated a 4 × 4 Latin square for all odor pairs across thenumber of discriminations and then replicated this Latin square withcontingencies reversed between S+ and S-. Therefore, if an animal had S+ anisole during the third initial acquisition, another animal had S+ eugenol during the third initial acquisition (from the anisole/eugenol pair). As we had 12 subjects, we needed two additional random sequences and their counter-balanced sequences. As one mouse that did not learn was excluded from the analysis, the data shown are for 11 mice. Both the initial acquisition and reversal stages ended when performance reached the criterion of 85% correct responses in 20 consecutive trials. The experimental switch to the next stage (reversal or next odor pair) occurred within ongoing sessions. We implemented this performance-based stage switching in the experimental control software so that it occurred automatically. Otherwise, as commonly done, a mouse could have advanced to the next experimental stage only on the next experimental day. This would have significantly extended the duration of the whole experiment. Also, maintaining training after the criterion is reached could lead to overtraining which may impact later training stages. The experiment ended for a mouse when it had completed all eight stages of odor discrimination learning and reversal. After finishing the experiment, the mouse stayed in the system and was re-started on its discrimination series until all the other mice had completed the experiment. Odors

Children can sort by more sophisticated criteria by this age. A 100-piece puzzle becomes a brilliant sorting activity once you realise that it’s much more efficient to categorise the pieces into groups such as edges, corners, middle pieces as well as grouping according to which part of the picture they correspond to, e.g. for a landscape, you might collect all the sky pieces in one pile and the land in another. We would like to thank Alexej Schatz for software development, Peter Spende for engineering support, Francesco Bagorda and Katja Frei for technical assistance. We are grateful to Vladislav Nachev, Lucille Alonso, Andreas Schaefer, Andrew Erskine and Cristina Marin for providing valuable comments and suggestions on the manuscript. Supplementary Material Another cognitive function of interest here is cognitive flexibility, which underlies performance in reversal learning. Reversal learning is a special case of discrimination learning in which reward contingencies between a previously rewarded stimulus and an unrewarded stimulus are switched. As reversal learning requires adapting to new reward contingencies and inhibiting pre-potent responses to previously rewarded stimuli, it is considered to be a measure of cognitive flexibility. Deficits in reversal learning are observed in many neurological and psychiatric conditions, such as schizophrenia ( Waltz and Gold, 2007; Schlagenhauf et al., 2014), dementia ( Freedman and Oscar-Berman, 1989; Rahman et al., 1999), depression ( Robinson et al., 2012) and addiction ( Ersche et al., 2011). This association with neurological and psychiatric conditions makes reversal learning in rodents a translationally relevant task.To provide an overview of session distributions and stage progressions for each individual animal in the automated system, we plotted the duration of stages and start times of each session against the timeline of the experiment ( Supplementary Figure 2). Furthermore, we visually investigated how the time of day was correlated with the number of trials performed within the session and with performance during the sessions ( Supplementary Figure 3). Although performance during the sessions did not correlate with the time of day, it appears that during certain time bins within the dark cycle, mice performed more trials within sessions. However, we found no visible correlation between number of trials performed in a session and performance during the session ( Supplementary Figure 4). Discussion