1. An enduring change in behaviour, or in the behavioural potential, that results from the individual's behavioural experience.
2. The retention over time of *learned information.
3. The retention over time of experience-dependent *internal representations, or of the capacity to reactivate or reconstruct such representations.
Memory, the mother of all the muses (Hesiod 8C BC,
"mnemonics), resides in many types of "systems, including inanimates (Moravec 1988; Weng et al. 2001).
We will refer here solely to memory in biological organisms with a nervous system. As in the case of "learning, the term 'behavioural experience' in definition 1 refers to the wide gamut of sensory, motor, emotional, and cognitive events that take place throughout a lifetime. Other types of experience, such as the unfolding of a rigid "developmental programme independent of the environment, or disease, injury and poisoning, are excluded. The term 'enduring' means longer than the perceptual or the cognitive present, which is a fraction of a second to a few seconds at most ("phase). Another type of definition refers to memory in terms of 'information' (definition 2). It is noteworthy that brains and nervous systems are not at all mentioned in the behavioural or informational types of definitions. This lack of reference to the biological substrate of behaviour is the combined heritage of multiple schools in psychology, information, and system theory (reviewed in Bower and Hilgard 1981; Baddeley 1997). Over the years, only rarely were complaints expressed in the literature that 'few dictionaries contain any reference to memory as a feature of a physical system' (Young 1987). Definition 3 refers to memory as retention over time of experience-dependent modifications in "internal representations, where internal representations are neuronally encoded models of the world that could guide behaviour (Dudai 1989). The focus on the representational properties of neural systems is meant to guide the identification of the relevance to memory of findings at multiple "levels of analysis ("reduction). This attitude fits the multilevel nature of research in the neurosciences better than the attitude reflected in the behavioural and/or the informational definitions (Dudai 1989,1992; Crick 1994).
The view taken here is that all memories, regardless of the species and the task, are biological "internal representations. Similarly to 'learning', the difference among types of memory is in the ecological function, computational theory, "algorithms, and neural implementation of the internal representation. The pursuit of internal representations and their modification by experience is thus tagged as the gist of memory research. Internal representations are assumed to be encoded in the spatiotemporal states of neural circuits ("cell assembly, "homunculus). In contemporary neurobiology, quite a few research programmes that claim to target memory concentrate solely on the molecular and the isolated single-cell level. Such programmes address cellular information storage, or neuronal "plasticity, not 'memory'. This of course should not all belittle the huge excitement, signal importance, and immense usefulness of the molecular and cellular approach; it only means that molecular and cellular neurobiologists shall aim at integrating their programs with higher level approaches if they wish to focus on memory. And vice versa: system neurobiologists should tune to molecular and cellular biology, and exploit its tools, if they wish to understand the nuts and bolts of biological information-storage machinery in the brain.
'Memory' could be subjected to multiple "taxonomies, based on different "dimensions, such as duration, associativity, "conscious awareness, and behavioural function (e.g. Augustine 400; de Biran 1804; James 1890; Bergson 1908; Hebb 1949; Ryle 1949; Tolman 1949; Milner et al. 1998). Such taxonomies are exemplified under "learning above. The rationale and evidence for selected classifications of memory are detailed in the relevant entries in this book. At this point, it is useful to consider briefly only a few general properties of memory.
1. By definition, not all internal representations that guide behaviour are memories. Internal representations could also be "a priori, innate constructs, encoded by the genes and established by developmental programmes even in the absence of learning. "Percepts are also internal representations but not memory. Memories are only those species of internal representations that result from learning. At the same time, it is important to note that the concept of 'learning' used in this book is comprehensive, and includes not only physiological or behavioural experience but also rearrangements of internal representations that yield new knowledge. Therefore, in brains, probably also in very simple brains and ganglia, even the innate internal representations are expected to undergo a process of change with experience and become bona fide memory.
2. In contrast with what is connoted by popular "metaphors, memories, being spatiotemporal activity states of the nervous system, are unlikely to be stored over time as such. Rather, they are probably reactivated or reconstructed each time anew, to regain their meaning, content-wise, only in "retrieval (Tulving 1991). Retrieval is any "stimulus-induced or spontaneous activation of the representation, whether accompanied by behavioural "performance or not. So what is it that is stored over time? This profound issue is discussed under "persistence; suffice it to note here only two basic types of scenarios. One, that "recognition or "recall is the activation of a dedicated circuit, probably identical to the one activated in training. One could expect to encounter this scenario in simple circuits that subserve simple behaviours, such a reflexes ("Aplysia). Another possibility is that what is stored is only a compressed 'core memory', or index, that permits the reactivation or reconstruction of a full-blown representation. The activated circuit could be different from that employed in learning or previous retrieval. Further, retrieval itself could induce use-dependent alterations in the circuit, including re-"consolidation of the modified-anew trace. Add to this the "association of the "context at the time of retrieval with the retrieved "engram, and the result is the potential "falsification of memory.
3. A balance is expected in the evolution of biological memory between universal processes and components, that subserve many types of memory systems, and particular processes and components, that subserve only a particular type of memory system. Whereas economy in evolution could favour universals, the need to endow the system with new capabilities, as well as opportunism in evolution, could favour the addition of particular molecular, cellular, or system "algorithms and devices. The point to note is that even if two learned behaviours seem to share the phenomenology, or two memory circuits seem to share algorithms, such as the Hebbian, and hardware devices, such as the "glutamate N-methyl-o-aspartate receptor, a closer look may unveil many particulars among the apparent universals. In any case, as memory refers to so many different types of internal representations, neuronal systems, functions, and behaviours, as in the case of learning, it is unlikely to expect a master solution to the mechanisms of memory.
Selected associations: Engram, Internal representation, Learning, Palimpsest, Plasticity
Was this article helpful?