If This Theory Is Right, What Should Happen?
A theory becomes useful at the moment it predicts something you could check. Following one prediction from claim to test shows what a result can and cannot say.
Browse the directoryHold a short list of letters in mind and try to do something else at the same time. Does it matter what the something else is? The working memory model says it does. It proposes separate systems for verbal and for visual and spatial material, plus a system that directs attention. From that, a prediction follows: a second task that draws on the same system as the letters should interfere more than one that draws on a different system.
Here is an invented test. People memorize six letters and, while holding them, either read numbers aloud (verbal) or track the positions of dots on a grid (spatial). The two tasks have been checked beforehand to be equally hard. If the reading task disrupts memory for the letters more than the tracking task does, the result fits the model.
The equal difficulty matters. If one task were simply harder, it could disrupt memory through general effort alone, which would leave the comparison ambiguous about separate systems. Even a good fit leaves room for another account to predict the same result, so the model gains support without becoming the only candidate.
This theory says what components a system has, and it has changed. Baddeley and Hitch proposed the original multicomponent version in 1974, and in 2000 Baddeley added an episodic buffer to answer limits of that version. The accounts below emphasize processes that change behavior, and they predict in a different way.
Learning theories focus on other relationships. In operant conditioning, an action's consequences affect how likely it is to recur. In classical conditioning, a cue that has signaled an event can come to elicit a response. A comparison of learning predictions can vary whether an event follows a cue or depends on an action. Both processes may contribute in the same setting.
Social learning theory makes a different kind of prediction. A child watches a sibling take a cookie and get scolded. The child may acquire how to take a cookie and also what happens next, and still never do it, because the scolding lowers the wish to try. Acquiring a response and performing it can come apart. Here is a proposed test, not a reported study. After watching a new action, give observers a chance to show it under comparable motivation and consequences, alongside a group who did not watch. If the watchers reproduce it more often, even though nobody was rewarded during the watching, that points to acquisition separate from performance.
Across the three, the strongest tests were the ones where rival accounts expected different results.