Why you must not ignore retrieval processes when thinking about memory:

If one only thinks of memory in terms of memory traces or engrams or objects, and only thinks of memory tests as varying in sensitivity to the strength or firmness or size of these memories, then a major part of the story is missed. The information that is retrieved, or expressed, during a memory test can vary depending on the technique used to retrieve the information.

“The key process of memory is retrieval.”
-Endel Tulving

Memory researchers long assumed that the level of processing of a stimulus has a systematic effect on how the memory of the stimulus is later recalled. They believed that intermediate processing will always result in stronger memories than shallow processing, and deep processing will always result in stronger memories than intermediate processing, regardless of retrieval conditions. However, these long-held assumptions about memory are nullified when the retrieval condition is manipulated between a standard recognition test and a word identification test involving a briefly (50 ms) flashed cue. Although encoding conditions remained the same, the method of retrieval became the “key process” in this experiment.

Another common assumption about memory overturned by retrieval manipulation is the theory that recognition is always more accurate and fruitful than recall. However, research by Tulving and Thompson in 1973 showed that even the recognition of self-generated words could be worse than recall. To do so requires controlling the retrieval conditions. When the retrieval conditions match the encoding conditions that surrounded the self-generation procedure, the counter-intuitive failure of recognition of self-generated words presents itself.

State dependent conditions during retrieval and encoding can effect a memory test. If you learn material while drinking alcohol, you will benefit from being in a similar state during retrieval. The same applies for various types of drug intoxication. Also, if you learned words while they were blurry, you will benefit in a recognition test if the test words are blurry, too. There are a lot of fun state dependent retrieval effects.

Retrieval conditions can also be a matter of life and death. Even in situations of police lineups, where hits and false alarms can have large consequences and encoding conditions can be from powerful, highly emotional situations, retrieval can be the decisive factor. Just by warning a viewer of a lineup that the culprit might not be present in the display can decrease mistakes by 40%. This sort of cue is only a change in the retrieval condition. Limiting the viewer to one potential culprit at a time by using a sequential lineup can cut mistaken identifications in half. Again, this manipulation only involves the retrieval condition.

As Tulving proposed, retrieval conditions do in fact play a “key” role in the outcome of various forms of memory tasks, and many times regardless of encoding conditions.