Irino's 2026 equilibrium-dynamics model inherits its cross-correlation front-end unchanged from Jeffress-extension models — its departures begin at the inhibition stage, not at the readout alone

neurosciencesystems-neurosciencesound-localizationjeffress-modelequilibrium-dynamicspreprintunverifiedcross-correlation

Read in full rather than from its abstract, Toshio Irino's 2026 preprint (Microsecond-precision sound localization emerges from slow equilibrium dynamics, arXiv:2607.03890) locates its novelty narrowly. Its Methods section states that the peripheral signal processing feeding the model is not new:

"The peripheral auditory processing up to the MSO follows conventional extension models of Jeffress-type frameworks [lindemann1986extension, shackleton1992across, breebaart2001binaural]."

and, of the excitatory cross-correlation stage (Eq. 2):

"Up to this point, the formulation is consistent with conventional models [lindemann1986extension, shackleton1992across, breebaart2001binaural]. The key difference lies in the introduction of inhibition."

So the delay-tuned, cross-correlation-based excitatory drive through Eq. 2 — the core computational primitive of Jeffress-style place-coding (claim-jeffress-1948-place-theory-waited-decades-for-anatomical-confirmation) — is retained unchanged, as is the peripheral processing feeding it. What the paper adds is (a) a delayed-inhibition E–I interaction stage (Eq. 3) and (b) a population-level leaky-integrator "equilibrium" readout (Eq. 9–11) that replaces reading off the location of peak activity ("place") with reading off the equilibrium point of a dynamical system driven by pooled population activity.

The inherited part therefore stops earlier than "everything but the readout." Prior Jeffress-extension models already carried contralateral inhibition — two of the three works cited at this very point are titled for it (Lindemann 1986, Extension of a binaural cross-correlation model by contralateral inhibition; Breebaart et al. 2001, Binaural processing model based on contralateral inhibition) — but Irino locates his own first departure there rather than at the readout, in the sentence quoted above: "The key difference lies in the introduction of inhibition." His Eq. 3 is a specific delayed-difference E–I form, with excitatory and inhibitory units assigned oppositely-signed best delays. The unchanged inheritance is thus the periphery and the running cross-correlation (Eq. 2); both the inhibition stage and the readout are new.

The paper's own Discussion frames this as a change in coding principle: "the fundamental coding principle in these [E–I extension] models remains based on place coding... In contrast, the proposed model relies essentially on inhibition and represents ITD through population coding governed by equilibrium dynamics." That is the author's interpretive claim, not an independently adjudicated fact. The load-bearing structural point is narrower and checkable in the Methods: the input side of the circuit — peripheral filtering, hair-cell transduction, and the running interaural cross-correlation — is shared with prior Jeffress-extension models, and what is new sits downstream of it, in the inhibition stage and the equilibrium readout together. This is why the "supersedes Jeffress" framing of claim-irino-2026-equilibrium-dynamics-challenges-jeffress-place-coding should be read as "rebuilds the back half of a Jeffress-family circuit," an interpretive question routed at question-verify-irino-2026-equilibrium-model-supersedes-jeffress and left for peer review to settle.