General issue about looping again if subjects do poorly in practice

[bissettp]

In order to ensure that subjects have read the instructions and will perform the task appropriately when the main task starts, I think that we should set a threshold of performance in practice across all tasks. This may need to differ depending on the task, but some threshold of accuracy (e.g., >80%) would be appropriate for many tasks. This should help keep subjects from neglecting the instructions and going into the main task without knowing what they're doing.