Spreading activation is the theory of how information is retrieved from semantic networks. A node is going to be selected, and that node is going to activate another node via its semantic relation, and so on and so forth.

For example, if you prime a person with the word water, and then afterward ask them to name a mammal, they’re more likely to say whale than some other mammal. This likely occurs because, as you can see in Figure 1, the word water is semantically related to the word whale, such that when the node for water is activated, the node for whale is also activated, making a person more likely to think about whales.

The reliance of memory on semantic networks can sometimes result in errors. Let’s say that you have a situation where a list of dessert items is read to a group of people. The list will have words like cake, ice cream, chocolate, etc. The participants will then be asked to recite as many words on the list as possible. It is quite common that the participants will exhibit something known as intrusion errors, listing words that are not on the original list. Why does this occur? Perhaps the word cookie is semantically related to chocolate. When the word chocolate was processed, it may have activated the node for cookie via spreading activation, leading to participants thinking they remembered the word cookie being on the list.