hen you remember something, something in you has changed since you learned it. The question of what changed, and where, is old enough to have a nineteenth-century name: the engram, Richard Semon's word for the trace an experience leaves. It has been declared unfindable, found, and argued over again.

the question

Is a memory stored in particular cells, so that you could point to it? In the strengths of the connections between them? Inside the cells, in molecules? Or nowhere in particular, as a pattern spread across a great deal of tissue?

the classic finding

Karl Lashley spent three decades cutting rats' cortex and testing what they had learned. The loss tracked how much he removed, not where, and his 1950 summary (Lashley, K. S., 1950) all but gave up the search. Donald Hebb answered him with the cell assembly (Hebb, D. O., 1949): memories live in groups of cells bound by strengthened synapses, spread out, so no single cut removes them. The mechanism Hebb guessed at turned up as long-term potentiation (Bliss & Lømo, 1973). Meanwhile Brenda Milner's patient H.M. showed that one region, the hippocampus, was needed to lay down new memories of events (Scoville & Milner, 1957).

what held up, and what did not

In 2012 a team tagged the cells in a mouse's hippocampus that were active while it learned to fear one box, and later switched just those cells on with light in a different box. The mouse froze (Liu et al., 2012). The same method planted a false fear, of a place where nothing bad had happened (Ramirez et al., 2013), and recovered memories from mice made amnesic by a drug (Ryan et al., 2015), which suggests the drug had blocked retrieval rather than erased the trace. Josselyn and colleagues call this finding the engram; their later review takes it into imagination (Josselyn & Tonegawa, 2020). Memories also become changeable again when recalled (Nader, Schafe & Le Doux, 2000), so a trace is not a fixed object.

The synaptic story has dissenters. RNA taken from trained sea slugs seemed to transfer a simple learned response to untrained ones (Bédécarrats et al., 2018), a result that has not been widely accepted. Gallistel argues that numbers a brain must remember, such as durations and distances, need a store that can be written and read like a register, and that synapses are the wrong kind of thing [paraphrase, not a quotation].

what it looks like in machines

In a trained network, what was learned is spread across millions of weights, much as Lashley found and Hebb proposed. Asking where a model "keeps" a fact is the engram question again, and it gets the same mixed answers: sometimes a small set of units can be edited to change one fact, sometimes not. See interpretability and hallucination.

what is still open

Whether engram cells are the memory or only an index to it; whether the content lives in synapses, in cells or in both; and how a trace survives the constant turnover of the molecules that make it.

in the tour: the room called "The assembly and the engram" has a placard called "Lit, and reawakened" ···

Reading list

1. Lashley, K. S. (1950). "In search of the engram." Symposia of the Society for Experimental Biology, vol. 4, Physiological Mechanisms in Animal Behaviour. Cambridge University Press.
2. Hebb, D. O. (1949). The Organization of Behavior: A Neuropsychological Theory. New York: Wiley.
3. Bliss & Lømo (1973). Long‐lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. The Journal of Physiology. doi:10.1113/jphysiol.1973.sp010273
4. Scoville & Milner (1957). Loss of recent memory after bilateral hippocampal lesions. Journal of Neurology, Neurosurgery & Psychiatry. doi:10.1136/jnnp.20.1.11
5. Liu et al. (2012). Optogenetic stimulation of a hippocampal engram activates fear memory recall. Nature. doi:10.1038/nature11028
6. Ramirez et al. (2013). Creating a False Memory in the Hippocampus. Science. doi:10.1126/science.1239073
7. Ryan et al. (2015). Engram cells retain memory under retrograde amnesia. Science. doi:10.1126/science.aaa5542
8. Josselyn, Köhler & Frankland (2015). Finding the engram. Nature Reviews Neuroscience. doi:10.1038/nrn4000
9. Josselyn & Tonegawa (2020). Memory engrams: Recalling the past and imagining the future. Science. doi:10.1126/science.aaw4325
10. Nader, Schafe & Le Doux (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature. doi:10.1038/35021052
11. Bédécarrats et al. (2018). RNA from trained Aplysia can induce an epigenetic engram for long-term sensitization in untrained Aplysia. eneuro. doi:10.1523/eneuro.0038-18.2018
12. Gallistel (2021). The physical basis of memory. Cognition. doi:10.1016/j.cognition.2020.104533

see also

	Karl Lashley
	Donald Hebb
	timeline: memory
	annotated: the nervous system