A phoria is not a misalignment. It's a prediction.

Optometrists are taught to measure phoria the way you'd measure a deviation from a fixed standard. Dissociate the eyes, record the number, compare it to Sheard's or Percival's criterion, decide whether the fusional system has enough reserve to keep covering for it. The phoria is treated as if it's the problem, and fusional vergence is the compensation for it. If the compensation is adequate, leave it alone. If it isn't, that's when you intervene — with prism, with lenses, with therapy — to shrink the deviation or grow the fusional reserves. Either way, the phoria itself is treated as the thing that's wrong.

That framing assumes the resting position of the eyes is a kind of leftover — what's left when you strip away binocular control and the spatial processing system. But that's not exactly what the literature describing the science of how 'vergence' works, actually shows.

Schor's model of vergence control splits the system into a fast, reflexive component driven by disparity, and a slow, adaptive component that sets the baseline around which the fast system operates. That baseline isn't a neutral position, it's a maintained neural bias — it is ongoing innervation to the extraocular muscles that is continuously recalibrated by recent visual behavior. Rosenfield's 1997 review is blunt about how badly this gets underestimated clinically: under truly open-loop conditions, the eyes rest at a vergence angle of roughly 23 degrees, not the fraction of a prism diopter we're used to calling a phoria. What we measure clinically isn't raw tonic innervation. It's a composite — tonic vergence, plus whatever slow fusional adaptation happened during the last few minutes of near work, plus residual fast vergence contribution at the moment of testing. The measured number is only a snapshot of an actively managed operating point, not a stable anatomical fact.

Owens and Leibowitz's work on dark vergence and dark focus makes the same point from a different angle. When you strip away the sensory information the system normally uses — put someone in the dark, taking away both contrast disparity cues — the eyes don't drift to the parallel "ortho" posture, and accommodation doesn't relax to viewing optical infinity. Both settle at an individually characteristic intermediate position. That's the tell. If phorias were simply where the eyes end up when binocular control fails, you'd expect it to fail toward some universal default — optical infinity, zero convergence angle. But it doesn't. It falls back to something specific to that person, which is exactly what you'd expect from a system that maintains its own prior settings and defaults to them when incoming information is insufficient to justify anything else.

There's also direct evidence that vergence doesn't just react — it anticipates. Alvarez's 2010 fMRI work compared vergence responses to predictable-versus-random disparity steps and found that predictable demand recruited the frontal and supplementary eye fields differently than unpredictable demand did, with faster, anticipatory movements to match. The system was pre-positioning ahead of a stimulus that it expected to see. That's not a reflex correcting an error after the fact. That's a neurological controller acting on an internal model of where the eyes are about to need to go.

Put those three findings together and the picture changes. A phoria isn't the deviation left behind when fusion is dissociated. It's a readout of the operating point the brain has already built — its cheapest, most-primed position given that person's accommodative and vergence history — and fusional vergence is the mechanism recruited to override that prediction when task demand disagrees with it. Sheard's criterion isn't really asking "is the error big enough to fix." It's asking whether the override system has enough spare capacity to comfortably out-vote the prediction on demand. That's a meaningfully different clinical question, even if the math on the exam form looks identical.

This is where the plus-lens question gets interesting, because the standard explanation for the classic dictum, "don't push plus because it makes exo worse" has always bothered me. The story is that plus relaxes accommodation, which relaxes accommodative convergence via the AC/A link, which lets the exophoria increase — as if the lens is actively degrading something that was otherwise fine. But the response AC/A literature tells a more specific story: the amount of accommodative convergence lost when you add plus tracks how far the person's actual accommodative response was sitting from their resting state, not simply the dioptric power of the lens. Rosenfield and Ciuffreda's 1996 work found the accommodative vergence crosslink is driven specifically by the phasic — the actively engaged — component of accommodation, not the tonic baseline.

Here's my read on that: in someone displaying exophoria, some of that phasic accommodation may be doing double duty — not just clearing the target, but generating extra accommodative convergence that's propping the eyes closer to alignment than their resting bias would otherwise put them. Add in plus sphere power (or reduce minus sphere power if the person is myopic), and you remove the piece of accommodation that isn't dioptrically necessary. What you're left with looks like "worse" exophoria. However: it may not be new misalignment. It may be the 'resting posture' becoming visible once the phasic system stops covering for it. Low plus isn't necessarily pushing the eyes away from where they belong — it may be letting them settle toward where they were already primed to go, and taking away the accommodative effort that was quietly shifting it away. Said in an easier way, the plus power is putting the image where the eyes want to naturally sit; it's not making anything worse.

If that idea is correct, it reframes what we're deciding when we prescribe plus for a high-AC/A esophore versus withholding it from a low-AC/A exophore. We're not applying the same tool to opposite problems. We're deciding, in each case, whether removing phasic accommodative effort brings the system closer to comfortable equilibrium or strips away compensation the fusional system was relying on. Same mechanism, opposite consequence, depending on which direction that person's prediction already points.

Take someone who's spent years at a chronic, close working distance — dense near work, hour after hour. An esophoria in that person isn't automatically a problem to shrink. It may be the cheapest configuration for keeping the brain working as intended — the eyes having converged to a resting point toward where the brain wants to perform work at, so it doesn't have to keep spending fast vergence effort getting there and holding on. Treating that esophoria as a defect to correct back toward some population-average zero value assumes the goal should be a number on a chart rather than a match to how that person's visual system is actually designed to work.

Now take the person who hates reading. The instinct is to look at their exophoria and blame that — insufficient convergence and inadequate reserves, it makes sense their avoidance of near-point tasks. But it's worth asking the question from the other direction. Maybe that exophoria isn't a barrier standing between them and reading? Maybe that's what an efficient system looks like in someone whose spatial processing has been built around distance and movement — sport, open space, tracking things seen away from the body — where a divergent resting bias is the cheap, primed position, and near convergence is the expensive, effortful task. The exophoria isn't malfunctioning in this instance, as their brains are optimized for a different job than the one we're currently testing it against.

Neither of those is a diagnosis I can hand you a citation for — they're an interpretative leap. But it's the leap the tonic vergence and dark-vergence data actually invite: if the resting position is built by, and biased toward, sustained visual behavior, then reading a phoria without knowing what that person's visual system has spent years being built to do is reading half the measurement. The chart data tells you where their eyes rest, but it doesn't tell you why that's the cheapest place for this brain to rest — and that second question is the one that decides whether you're looking at dysfunction or at fit.