The notion that "stimuli produce enduring physical changes in the brain, and that these changes are the basis for memory, has been with us since early times (e.g. see Plato's etched wax tablets of memory, Theaete-tus; "metaphor). About 100 years ago, a German scholar, Richard Semon, termed the material record engraved by a stimulus in living tissue as the 'engram' (Semon 1904). The etymological roots of the term are Greek, and it means 'something converted into writing'. Semon had in mind a rather general theory of experience-dependent records in living organisms, which included not only neural but also "developmental and genetic memory in all types of tissue. In his book, The mneme, Semon suggested two mnemic 'laws'. The first is the 'law of engraphy'. It states that 'all simultaneous excitations within an organism form a coherent simultaneous excitation complex which acts engraphically; that is, it leaves behind a connected engram-complex constituting a coherent unit/ (ibid., p. 273). The second is the 'law of ecphory'. 'Ecphory' according to Semon is the process that 'awaken(s) the mnemic trace or engram out of its latent state into one of the manifested activity'. The etymological roots of 'ecphor/ are also Greek, and it means 'to be made known'. The law of ecphory states that 'the partial recurrence of the excitation complex, which left behind the engram complex, acts ecphorically on this simultaneous engram complex, whether the recurrence be in the form if an original or of a mnemic excitation.' (ibid., p. 274). Elements of Semon's writings are nowadays echoed in the discussions of "cell assemblies, "models of neural networks, and "retrieval. Unfortunately, Semon and his book were almost forgotten (Schacter 1982).
Most of the popularity of 'engram' stems from a noted paper by Lashley (1950), entitled 'In search of the Engram'. Though seemingly an epitome of the crosstalk of scholarly ideas (Semon's chapter V in The mneme is entitled 'The Localisation of Engrams'), typically of Semon's fate, Lashley did not cite Semon even once in this paper.1 Lashley aimed at identifying 'habits of the conditioned reflex type' in the brain. Following the path of Flourens, Franz, and others (Gomulicki 1953; Herrenstein and Boring 1965; Brazier 1988), Lashley used the "methodology of inference of function from dysfunction. He inflicted anatomical lesions on various parts of the "rat or "monkey brain, and tested the effect of the intervention on brightness discrimination and
"maze learning (Lashley 1929, 1950). After many years of research, Lashley came to the conclusion that no cortical area, except the relevant primary sensory areas, is obligatory for learning and memory. He summarized his findings in two principles: (a) the equipotentiality principle, which states that cortical areas are equipoten-tial for learning, and can generally substitute for each other in learning, and (b) the mass action principle, which states that the reduction in learning and memory "performance is roughly proportional to the amount of tissue destroyed, rather to its position. Other conclusions were also drawn from the data, e.g. that the effect of a lesion is proportional in magnitude to the complexity of the task. 'This series of experiments', concluded Lashley, 'has yielded a good bit of information about what and where the memory trace is not... I sometimes feel, in reviewing the evidence on the localization of the memory trace, that the necessary conclusion is that learning just is not possible ... Nevertheless, in spite of such evidence against it, learning does sometimes occur' (Lashley 1950).
Lashley's conclusions about the engram were already criticized at the time of their publication (Hunter 1930). The main objections were, first, that the behavioural tasks were not well defined in terms of the sensory inputs and the behavioural strategies required for successful performance, and therefore a "subject with partial brain damage could still succeed in the task by using the undamaged areas; second, the lesions were not delicate enough to differentiate the contribution of distinct functional divisions in the brain. Disappointed with the search for the engram, Lashley came to favour field theories of brain function, which regarded sensations and memories as patterns of excitations being reduplicated throughout the cortical surface like spreading waves in a liquid surface (Lashley 1950). This view was compatible with the Gestalt theories.2 In the later part of his career, Lashley put this idea to experimental test. He placed pieces of gold on the visual cortex of the monkey. The metal was expected to short-circuit the cortex and therefore disrupt visual perception. This did not happen (Lashley et al. 1951). Similar conclusions were reached at about the same time by other investigators. Not surprisingly, Lashley himself remarked that he had destroyed all theories of behaviour, including his own (Hebb 1959). Lashley's seminal contribution to the study of learning was not, however, in his ad-hoc experimental conclusions. Rather, it was in the methodology and the concepts. On the methodological side, he was a pioneer in combining careful lesion techniques with behavioural analysis. On the conceptual side, he came to argue that dedicated effector-affector connections could not be considered as building blocks of high brain function. This notion was influential in shaping later distributed models of brain and memory, for example, those of Lashley's student, Hebb ("algorithm, "cell assembly, "classic).
What is the current status of the search for the engram, a century after Semon and half a century after Lashley? The picture is not simple. At this stage, it could be useful to resort to the basic issues.
1. The existence of the engram. Is there an engram? The question where the engram is should not be confounded with the question whether an engram is at all necessary. The answer in a nutshell is definitely yes, there must be some type of engram to ensure "persistence of memory. Whether the term should apply to the entire circuit that encodes the item, or only to those use-dependent changes in the circuit, is already a matter of convention or taste. Even if we consider certain types of memory as selective reactivation ("retrieval) of endogenous spatiotemporal patterns of neuronal activity ("a priori, "cell assembly), use-dependent alterations must leave a material tag at one "level or another of the neural tissue. This tag, however, might be minute, and in some cases even hide as modification in an index of internal representations in a brain area far remote from the area in which the retrieved representation resides.
2. The locale of the engram. The question is hence not whether there is an engram to search for, but rather whether the engram is localized to a single location, whether this location is stable over time, and whether the combination of locale and stability permits us to put our hands on it. This becomes a quantitative rather than a qualitative issue, interwoven with methodological and technical difficulties. In a nutshell, it is likely that the engram of modified reflexes in "simple systems will be found in a dedicated pathway, whereas that of complex memories in humans will be highly distributed ("cell assembly). Further, some types of engrams may shift from one locale to another; this, for example, seems to occur routinely in the "consolidation of "declarative memories, in which with time the engram becomes independent of the "hippocampus.
3. The differentiation of function in the engram. In spite of the expected multiplicity of locations, and the potential uncertainty in the localization of complex engrams over time, discrete anatomical parts of the brain are clearly implicated by the available data in specific types of memory. For example, it is now evident that the primate mediotemporal lobe subserves "declarative memory, the corticostriatal system "skill memory, and the prefrontal "cortex "working memory.
Candidate engrams have also been identified in classical conditioning in mammals, in lower vertebrates ("bird-song), and in invertebrates ("Aplysia, "Drosophila). In most cases, the question of what are the precise functions of the identified regions in retaining the trace over time remains, however, unresolved. First, the fact that a circumscribed brain region is necessary for a certain type of memory does not entail that this region is sufficient to encode the memory, even more so exclusive in doing so ("criterion). Second, some areas in the candidate engram might be required for only some "phases of memory but not for other phases. And third, the engram may have a 'core', responsible for the persistence over time of the essential attributes of the item, and auxiliary components that either encode additional "dimensions of the item, or are recruited only in the encoding or the expression of the memory. All these issues resurface in the renewed attempts to identify engrams by means of "functional neuroimaging.
4. The transformation of the engram. In addition to the differential function of identified brain areas and pathways in a particular experimental protocol, the relative contribution of distinct areas and pathways to the engram could change dramatically when the "acquisition or retrieval parameters of the task are altered. This is the case even in relatively simple forms of learning, for example, in classical conditioning, when a shift is made from delay to trace conditioning, or from elemental to "contextual conditioning (Thompson and Kim 1996; Nader and LeDoux 1997; Desmedt et al. 1998). This again calls for the depiction of engrams as multicomponent "systems, the individual components of which are recruited and expressed in a permutable manner depending on the specific task demands.
5. The level of the engram. Traditionally, the search for the engram refers to an anatomical expedition. But nowadays it is also cellular and molecular. Problems similar to those concerning the neuroanatomical search haunt the molecular and cellular search. Synapses and neurons modified by experience should carry some type or another of a physical record of the experience (Dudai and Morris 2000). But the change could be minute, distributed, and alternate over time from one site to another. Again, in the molecular like in the circuit level, there might be a core engram and additional tiers of mechanisms that encode additional stimulus dimensions (Berman and Dudai 2001).
In conclusion, whatever the level of analysis, to expect the search for the engram in the mammalian brain to end up in the identification of a single locale is naive. And even if a candidate area is identified as important in maintaining the engram, the problem becomes that of both resolution and "reduction: Where is it within that area? In a dedicated circuit? In identified neurons? In synapses? To what size of circuit do we hope to nail it down? What answer will finally satisfy our urge to localize things in the world? And if we keep reducing the area, or the number of units in the area— where will the transition be from an engram, to a fragment of that engram, or to a non-engram?3 These problems could apply to simple brains as well (Glanzman 1995).
And as to term 'engram' per se, many experimenters use it, others think that it carries the risk of spilling into purple prose, and prefer the term 'trace'. Most authors, however, simply use 'trace' and 'engram' interchangeably (e.g. Thompson and Kim 1996).
Selected associations: Criterion, Memory, Metaphor, Persistence, Phrenology
'He was, however, well aware of Semon's writings (Lashley 1933). 2For what the Gestalt was, see *insight.
3This is another example of the classical sorites paradox, which is explained under *insight; see also 'reduction.
Was this article helpful?