Answers are grounded in this paper and cite their sources.
A memory can last for decades, yet the proteins of a synapse are replaced within days. Crick (1984) proposed that the molecules that store a memory restore one another, so that the pattern survives its parts, and several molecular switches of this kind have since been modelled. Here we ask a question that does not depend on the mechanism: what is the longest time that any switch made of N molecules can remember? We consider all switches in which the number of active molecules changes one molecule at a time, with an activation rate per inactive molecule between a spontaneous floor ε and a catalytic ceiling k, and a loss rate per active molecule between the turnover rate μ and an erasure ceiling λ. (i) We prove that, for every readout threshold, both the potentiated and the unpotentiated state are retained longest, in the sense of stochastic order, by one and the same switch: a threshold rule that restores at full speed when at least m molecules are active and erases at full speed otherwise. (ii) We give its mean retention times in closed form, as sums of binomial coefficients, and evaluate them in exact rational arithmetic. (iii) When the two contrasts κ=k/μ and χ=λ/ε are equal to c and λ=μ, the longest possible memory T^*(N) satisfies N⁻¹(√ c/2)N≤μ T^*(N)≤((1+c)/√ c)N for even N; numerically, each added molecule multiplies it by about (1+c)/(2√ c). (iv) With molecules that last 30 hours, no switch of this class holds both states for a mean of a century with fewer than 20 molecules if $c=10$, 12 if $c=30$, 8 if $c=100$ and 4 if $c=1000$, and these numbers are attained; thirty molecules suffice as soon as c≥7, while a single dimer with $c=100$ lasts 50 molecular lifetimes. (v) The retention times of the ideal switch have an increasing failure rate, so that a mean of a hundred centuries guarantees a century with probability above $0.99$, at the price of two to ten more molecules. (vi) Smooth cooperativity is far from the limit: a switch with Hill exponent 2 needs 187 molecules where the ideal one needs 12. In this class, then, a lifetime of memory requires few molecules but a sharp threshold. These are limits for one-variable switches with constant rates, not a model of a particular synapse. This record contains the preprint (PDF, PCI template), its LaTeX source, and the Python code and results file that reproduce every theorem, table and figure (code under the MIT license).
No takes yet. Share an insight, caveat, or question.
Luca Eliseo Pavesi (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: