Early theorists divided memory into short- and long-term stores (STS and LTS, respectively) (Atkinson & Shiffrin, 1968) based on evidence of seemingly different properties. For example, short-term memory was theorized to be affected by recency and long-term memory by primacy effects. Later models refined this modal model to produce a working memory model (Baddeley & Hitch, 1974), although more recent theories favour a single memory system.

Although a huge body of research has focussed on the distinction between memory structures such as STS and LTS, the more fundamental issue is how memory operates. Several factors operating at the encoding stage have been proposed to influence whether material will be retrieved. The levels of processing in the model suggest that the depth of processing (in particular, processing of meaning) influences how well the material is remembered. The phenomenon of context-dependent memory demonstrates that the cues available at encoding aid retrieval; it is thought that this is because they are stored along with the material being encoded.

What causes us to forget has been examined, and there is relatively little evidence that memory is discarded per se. Although decay may occur and memory loss may result from what McGeoch called ‘deterioration of the organic correlate’, we can explain most (possibly all) of forgetting in terms of the effects of interference or the matching of retrieval cues to what was originally encoded. If memories compete for recall – the essence of most interference theories – then, learning new information inevitably carries a cost. Forgetting thus occurs because it is necessary to keep memory efficient (e.g. see Anderson & Schooler, 2000).

Memory is unreliable. It is a constructive process in which we interpret what has been experienced in light of our expectations, which in turn are based on existing mental representations (schemata) about the world. Memory has also been demonstrated to be influenced by later events that interfere with earlier memory traces. Psychological investigation of such misinformation effects has been influential in our understanding of the accuracy of memory and has made a significant contribution to real-world issues such as eyewitness testimony and the phenomenon of false memory.

Anderson and Schooler (2000) suggest that it is puzzling, in evolutionary terms, why intention to learn has so little effect on what we remember. But this finding is less puzzling if we think of memory in terms of transfer-appropriate processing. People are probably only rarely aware that they’ll need to know something in future (or what retrieval cues will be available when they try to remember it). A logical consequence of transfer-appropriate processing is that it is probably unwise to think of memory as separate from other aspects of cognition. The processing involved in interacting with and thinking about the world, from perception through to reasoning or problem-solving, probably produces memory as a by-product (e.g. Crowder, 1993; Lansdale, 2005; Payne & Baguley, 2006). Of course, it is somewhat ironic to end a chapter on memory by noting that memory shouldn’t just be considered in isolation from the rest of cognition – or indeed the rest of psychology.