If we only abandon highly simplistic "metaphors of memory storage, such as 'cabinet files' or 'computer disks', 'persistence' (per- + sistere, Latin for 'to stand') becomes a prominent "enigma of memory research. As memory is the retention of acquired information over time, persistence is clearly its central attribute. Generally speaking, there is a family of'persistence problems'. The "classic philosophical problem (definition 2 above) is also known in metaphysics as 'The Ship of Theseus': the ship of the mythical Greek hero was placed on display in Athens, and with time, parts of it were replaced, one by one, till none of the original remained. Is this still the same ship? (Plutarch, Theseus 1-2C AD/1914b; Kim and Sosa 1999). Metaphysics notwithstanding, this discussion will focus primarily on the more pragmatic 'persistence problems', which relate directly to memory (definition 3a,b). First, how come that in spite of the notorious frailness of the individual components of the biological material in which the trace is registered, the "engram endures, sometimes for a lifetime? This is 'the endurance issue'. Second, how do items in memory persist over those periods in which they are not expressed? This is 'the dormancy issue'. Both issues seem to call for some engineering solutions. They also invoke, however, interesting conceptual and methodological issues. For example, although the 'dormancy issue' can be satisfied rather easily, its probable solution leads to the paradoxical conclusion that specific items in memory are not at all 'stored' in the common sense of the term.
Let's turn to the 'endurance issue' first. Organisms are epitomes of the pre-Socratic saying 'everything flows' ('Panta Rei', Guthrie 1962). Their constituents never experience a dull moment. Metabolic instability contributes to "plasticity and adaptation, hence to survival. Stability and completion, probably contrary to intuition, could promote atrophy: 'How long/Do works endure? As long/As they are not completed/Since as long as they demand effort/They do not decay' (Brecht 1929-33). In the mature organism, cells are constantly born and die (Alberts et al. 1994). Proteins within cells turnover within minutes to weeks, their actual life expectancy being determined by their type, location and history (Varshavsky 1992; Shi et al. 1996; Krupnick and Benovic 1998; Huh and Wentold 1999; Xu and Salpeter 1999). Furthermore, it is now evident that cells are born even in tissues that were traditionally considered to be stable throughout adult life; the brain is no exception (e.g. Kirn et al. 1994; Eriksson et al. 1998; Gould et al. 1999b). This situation raises the following questions.
1. The molecular *level. If an experience-dependent change is embodied in modified "synaptic components (e.g. "Aplysia, "long-term potentiation), such as enzymes, "ion channels, and "receptors, which have a limited life span—how does the change outlast the limited life span of the proteins, and becomes immune to the consequences of molecular turnover? Without such resistance to the effect of turnover, no long-term memory would be possible. We should also worry about the persistence of post-translational modifications in protein molecules ("protein synthesis), because these modifications are unlikely to survive immediate remodification by enzymes in vivo (e.g. Shuster et al. 1985). Multiple mechanistic solutions have been proposed to account for the immunity of use-dependent neuronal changes to molecular turnover (Crick 1984b; Lisman 1985; Goelet et al. 1986; Buxbaum and Dudai 1989; Dudai 1989; Chain et al. 1999; Lisman and Fallon 1999). These proposals include molecular positive feedback loops, that, once activated, are shifted into a new stable state, and regenerate the molecular change again and again ("protein kinase); modulation of gene expression, that results in a new, stable expression pattern ("CREB, "immediate early genes, "protein synthesis); or a combination of the above.
2. The synaptic to circuit levels. If traces are subserved by synaptic connections, but these connections are continuously remodelled in vivo, how is the trace preserved (e.g. Bailey and Kandel 1993; Kleim et al. 1997)? At least in complex circuits, the problem is solvable by assuming distributed codes, in which no single node in the net is exclusive in representing a significant chunk of the message ("cell assembly, "homunculus, "model). The idea is thus that at any given moment, the trace is retained by a sufficiently large chunk of the circuit, so that it can tolerate elimination of part of the nodes. This property is termed 'graceful degradation'.
Similarly, if newly-born neurons are incorporated into functional circuits in the adult brain (Kirn et al. 1994; Eriksson et al. 1998; Gould et al. 1999b; though see Rakic 2002), how does the perturbed circuitry sustain the old memory? Again, the conceptual difficulty is ameliorated by assuming a distributed code as above. Note that by explaining how hardware turnover does not undermine the persistence of the trace, we do not solve the metaphysical 'Ship of Theseus' identity problem. But at least we can explain how copies of the ship become available. That this is a partial answer is perfectly OK, because neuroscientists should definitely relegate some problems exclusively to philosophers.
And what about the 'dormancy issue'? We have defined '"memor/ as the retention over time of experience-dependent "internal representations, and noted that representations are expected to be encoded in the spatiotemporal activity patterns of neuronal circuits. If this is the case, how is the memory retained in our brain when the representation is not actualized? Note that for the sake of argument, it does not really matter whether the representation is activated only in explicit "retrieval of the item in memory, which may occur very rarely, or also, probably more frequently, in the course of hypothetical, implicit 'house-keeping' routines in the brain.
The 'dormancy issue' can be rather easily resolved if we only switch the level of analysis and recall the difference between '"memory and neuronal and synap-tic 'storage'. What is retained over time is not the actual internal representation, but rather the capacity to generate it. The information is stored as 'hardware' alterations in the circuit that is capable of expressing that specific representation. For a memory to be retrieved, certain "cues are required to engage the circuit and generate the relevant activity pattern anew. In other words, memories are not retained 'as is', but reconstructed; what persists after learning is the change in the system that leads to their reconstruction in a certain way but not another; retrieval is not merely an expression of memory, but rather a condition for its mere existence. We can also conclude that the study of synapses and individual nerve cells is expected to generate information on the generic mechanisms of storage, whereas the study of active circuits suits better the search for the fingerprints of specific instances of memory.
Selected associations: Consolidation, Engram, False memory, Metaphor, Plasticity
Was this article helpful?