The encoding specificity principle is the general principle that matching the encoding contexts of information at recall assists in the retrieval of episodic memories. It provides a framework for understanding how the conditions present while encoding information relate to memory and recall of that information.

It was introduced by Thomson and Tulving who suggested that contextual information is encoded with memories which affect the retrieval process. When a person uses information stored in their memory it is necessary that the information is accessible. The accessibility is governed by retrieval cues, these cues are dependent on the encoding pattern; the specific encoding pattern may vary from instance to instance, even if nominally the item is the same, as encoding depends on the context. This conclusion was drawn from a recognition-memory task. A series of psychological experiments were undertaken in the 1970s which continued this work and further showed that context affects our ability to recall information.

The context may refer to the context in which the information was encoded, the physical location or surroundings, as well as the mental or physical state of the individual at the time of encoding. This principle plays a significant role in both the concept of context-dependent memory and the concept of state-dependent memory.

Examples of the use of the encoding specificity principle include; studying in the same room as an exam is taken and the recall of information when intoxicated being easier when intoxicated again. 

Ebbinghaus, a pioneer of research into memory, noted that associations between items aids recall of information thus the internal context of a list matters. This is because we look for any connection that helps us combine items into meaningful units. This started a lot of research into lists of to-be-remembered (tbr) words, and cues that helped them. In 1968 Tulving and Osler made participants memorise a list of 24 tbr words in the absence or presence of cue words. The cue words facilitated recall when present in the input and output of memorising and recalling the words. They concluded that specific retrieval cues can aid recall if the information of their relation to the tbr words is stored at the same time as the words on the list. Tulving and Thomson studied the effect of the change in context of the tbr by adding, deleting and replacing context words. This resulted in a reduction in the level of recognition performance when the context changed, even though the available information remained context. This led to the encoding specificity principle.

Semantics do not always play a role in encoding specificity; memory, rather, depends upon the context at encoding and retrieval. Early research has shown that semantically related cues should be effective in retrieving a word provided the semantic cue was encoded along with the target word. If the semantically related word is not present at the time of encoding, it will not be efficient at cuing recall for the target word.

In a laboratory study, a subject presented with an unrelated word pair is able to recall a target word with much more accuracy when prompted with the unrelated word it was matched with at the time of encoding, than if presented with a semantically related word that was not available during the time of encoding. During a recall task, people benefit equally from a weakly related cue word as from a strongly related cue word, provided the weakly related word was present at encoding.

Regardless of semantic relatedness of the paired words, participants more effectively recalled target words that had been primed when prompted for recall. Many of the following experiments employed a method modeled off of Thomson and Tulving's. All, however, had slight variations which allowed the researchers to discover their own individual findings. The following table shows the importance of priming through word pairs to achieve enhanced recall of words encoded together.