nucleation

THEORY-909: As reviewed by Gerstner et al. (2018) from five papers, mostly in slices and one laboratory per region, pairing pre- and postsynaptic activity changed the synapse when a neuromodulator or plateau potential followed within about 1 s in striatum, within 5 s (noradrenaline, LTP after pre-before-post pairing) or 2.5 s (serotonin, LTD after post-before-pre pairing) in visual and prefrontal cortex, and within about 2 s in CA1, and not at longer delays; the reviewers read this as a decaying synaptic eligibility trace and as experimental support for three-factor learning rules on behavioural time scales

2 min read

Properties7
StatusActive
Version2
Tagsneuroscience, learning-and-conditioning
SourceLIT-1494 — Eligibility Traces and Plasticity on Behavioral Time Scales: Experimental Support of NeoHebbian Three-Factor Learning Rules
Bears onQUESTION-311 — Do synaptic eligibility traces on the scale of seconds carry learning from delayed outcomes in behaving animals, so that interfering with the trace between action and outcome abolishes learning from the delayed outcome but not from immediate ones?
Cited as “corroborated by” byCLAIM-443 — A learned prior over which processes govern, P(g | z, H) ∝ exp(βQ(g, z) + b_H(g, z)), its bias b_H trained by reward-modulated Hebbian updates to the relations by which one process recruits or endorses another, is a candidate mechanism for durable authority and for amortized governance; but persistence is not amortization, which is shown only when earlier work replaces later computation, communication or recruitment, and a prior over authority differs from a prior over actions only when what is appropriate changes
Cited as “extends” byTHEORY-1178 — In slices of striatum, visual and prefrontal cortex and hippocampus, Hebbian co-activation leaves a synaptic eligibility trace, lasting about 1 s in striatum, 3 to 10 s in cortex and about 2 s in CA1, that a later neuromodulator or plateau potential converts into a weight change and whose sign it can set: experimental support for three-factor learning rules on behavioural time scales

Graph View

Source

Gerstner, Lehmann, Liakoni, Corneil & Brea, LIT-1494, read in full in NOTE-1130 from the version of record: section 3 (pp. 7–11) and Figure 2, reviewing Yagishita et al. (2014), He et al. (2015), Brzosko et al. (2015, 2017) and Bittner et al. (2017).

What was actually shown

Each experiment paired presynaptic with postsynaptic activity and gave the third signal at a controlled delay.

  • Striatum. Dopamine during or up to 1 s after the pairing enlarged spines; at 4 s it did not. The trace decays “on a time scale of 1 s” (p. 7).
  • Cortex. Noradrenaline within 5 s of pre-before-post pairing gave LTP, at 10 s not; serotonin within 2.5 s of post-before-pre pairing gave LTD, at 5 s not (p. 9). The reviewers read this as the third factor setting the direction (p. 4); as reported, the direction also followed the pairing order, so the two are not separated here.
  • Hippocampus. A plateau potential within about 2 s of CA3 input potentiated it by about 200% in five pairings (p. 10). Dopamine within about a minute turned an LTD protocol into LTP (p. 10), which the reviewers think may be consolidation.

What could have come out otherwise: a delayed modulator could have had no effect at any delay, as the classical Hebbian picture expects, or the window could have been the hour of synaptic tagging. In striatum and cortex it was seconds.

What this does not say

  • That these traces carry learning in behaving animals. Almost all of it is slices with imposed protocols.
  • That eligibility traces are the brain’s solution. The reviewers: “there is nothing—apart from conceptual elegance—to favor eligibility traces over alternative neuronal mechanisms” (p. 12), and the results “do not exclude that other mechanisms work in parallel”.
  • That the third factor is reward. It was a neuromodulator or a plateau potential delivered by the experimenter. Whether in life it signals reward, surprise or novelty is open.
  • Anything above the synapse. Nothing here is about processes recruiting processes.

Connections

  • CLAIM-443. Supports the biological plausibility of the claim’s schematic three-factor rule at the synapse, and narrows its “three-factor rules are established”: recently supported in slices on the scale of seconds, established only for hour-scale consolidation. It also shows that a rule meant to learn from delayed outcomes needs a decaying trace, which the claim’s rule lacks.

Backlinks