There are many ways in which people have attempted to explain human memory. William James (1890) made a distinction between "primary" and "secondary" memory. Events occurring in the immediate past (few moments) require no recollection and hence could be described as primary. In contrast he described secondary memory as a person"s awareness of a previous state of mind after it drops off (at least once) from consciousness. Hermann Ebbinghaus (1885) had a different definition, which involved three different ways of retrieving information. The three kinds of retrievals are "intentional", "unintentional", and "unconscious retrieval". The first two kinds are commonly referred to as "explicit memory", which, again, includes two sub-types; declarative memory and procedural memory. The unconscious memory type, on the other hand, is referred to as "implicit memory" in modern cognitive science. Studies on memory have captured the richness involved in it. Studies on the relation between learning and memory, and the reconstructive and reproductive aspects, among others, have resulted in the general acceptance that human memories are multifaceted. Besides being reproductive and reconstructive, they could be intentional and unintentional as well as explicit or implicit, and could differ in direction and scope.

The diversity of memory forms warrants different methods for its study, and a common goal in memory research involves the attempt to arrive at generalizations about remembering, for instance "the forgetting curve", "the power law of practice", and "the encoding specificity principle".

The methods warranted for memory studies are diverse, and include techniques to assess memory for facts, skills, implicit memory, and memory"s manifestation at the level of neural structures. In the study of memory for facts, free-recall and recognition tasks are two major experimental techniques. In Free recall tasks subjects try to recall a items from a pre-memorized list. The serial position effect is a major finding, where subjects recall items from the initial and the most recent locations, and items in the center are poorly represented. The recognition tasks include a study phase and a test phase. In the study phase subjects are presented with a list of items, and in the test phase retention is assessed by presenting a list which includes the study items as well as some distracter items. The recall and recognition tasks make assessments of a kind of memory called episodic memory (memory of specific episodes known by their time and place of acquisition). Semantic memory (a person"s knowledge of the world) are assessed through a different set of experiments. Sequence verification tasks is one such kind, where the time taken to make a correct judgement is studied to understand memory structure.

Procedural memory is studied in terms of performance rather than in terms of explicit recollection. Subjects get repeatedly exposed to a set of stimuli while performing a specific task in such experiments, where priming and skilled performance is measured. Priming tasks typically involve an inspection phase and a testing phase. During the inspection phase subjects read a sequence of stimuli. In the testing phase the impact of the priming on experiments like word-stem completion and cued-recall is measured. One important finding relates to the evidence of implicit memory in amnesic patients in word-stem completion tasks, thus leading to the hypothesis that explicit and implicit memory are based on different memory systems. Measuring practice effects is another important set of experiments in this domain. In essence, it involves measuring changes in the speed in which people perform certain task. Modern PET Imaging has led to further studies of the effect of learning and practice on performance. Neuroimaging methods include a combination of brain scanning technique and computer technology to measure brain"s transient activity, where for instance, PET scans inject non-toxic radioactive substances to detect changes in blood-flow pattern in the brain.

Several memory models have been propounded to explain the various memory phenomena. Some are local in character where as a few are more global. Atkinson and Shiffrin"s (1968) two-store memory model is an instance of global model where there are two main slots; one for short term memory and one for long term memory. Another memory model was based on the neural network, arising out of many theorist"s concern that memory models should reflect parallel processing phenomena in cognitive processes. Such models have been successfully used to simulate word recognition and other cognitive functions, for instance McClelland and Rumelhart"s (1981) model of word recognition.

Since memory plays a fundamental role in cognition, and it is limited in capacity, other phenomena including attention, pattern recognition and problem solving (which depend heavily on memory) have attracted the attention of cognitive scientists.