T-maze

The question

Time to stretch our legs. How does one left-or-right choice reveal memory? You're watching Morris Water Maze Software Guide, over at morriswatermazesoftware.com.

Meet the T-maze

The T-maze is a start arm leading to a junction with two goal arms. In spontaneous alternation, animals naturally tend to visit the arm they didn't visit last time, which depends on remembering the previous choice. Rewarded versions test learning of a rule or a place.

How it runs

How do you run one? First, place the animal in the start arm and let it choose a goal arm. Next, for spontaneous alternation, return it to the start and let it choose again; alternation means it chose the other arm. Then, for rewarded alternation, bait the arm opposite the one visited on a forced sample run. Finally, run several trials per session, with a short, consistent delay between sample and choice.

Strengths and limits

What's it good for? Very quick, with little training for the spontaneous version. Clear binary outcome. Delay length can be adjusted to load working memory. Now, the small print. Chance is 50 percent, so you need enough trials for statistical power. Side biases must be checked. Odour trails can guide the choice unless the maze is cleaned.

What to measure

This is where tracking software earns its keep. Percentage alternation: the core memory score. Choice latency: hesitation at the junction. Side bias: the proportion of left versus right choices. Head turns at the junction: vicarious trial and error, a sign of deliberation. A mouse that alternates 60 percent of the time looks close to chance, but if it hesitates at the junction on every trial, the latency data tell you it's working at it.

The takeaway

If you remember only one thing, make it this: choice latency and its neighbours tell you how the task was done, not just how fast. Everything we covered is written up below.