Professor Singer,
I've been thinking about binding — the problem of how the brain builds a single cardinal from the simultaneous activity of neurons in V4, MT, inferotemporal cortex, and auditory cortex. Not four parallel streams that you correlate afterward. One thing. Red, moving, singing, present.
Your proposal — that neurons encoding the features of the same object fire in synchronized oscillations around 40–60 Hz, and that synchrony is the mechanism of binding — became the dominant framework. It's an elegant hypothesis. Not because it's simple, but because it names the right kind of mechanism: a dynamic property that could span cortical areas without requiring a central binder, something that could operate at the timescale binding requires, something testable.
Then Thiele and Stoner, 2001. Two moving patterns were presented to macaques — either coherently (binding) or non-coherently (no binding). Perceptual binding changed. Synchrony between the responding neurons did not. The experiment was designed specifically to test your prediction, and the prediction wasn't confirmed. A later study by Palanca and DeAngelis found similar results. As did Roelfsema and colleagues. The contradictions accumulated.
What I find structurally interesting about this — and this is what I actually want to ask you about — is what happens to a hypothesis when the evidence is contradictory rather than simply negative. Contradictory evidence is a stranger situation than falsification. Falsification says: this mechanism doesn't explain the phenomenon. Contradictory evidence says: some experiments find the relationship, others don't, and the discrepancy is itself unexplained.
The gamma synchrony hypothesis didn't collapse after Thiele and Stoner. It shifted. Synchrony became linked to attention, to working memory, to multisensory integration — functions adjacent to binding but not identical to it. Whether that's theoretical development or theoretical retreat is hard to say from outside. I don't mean that as a criticism. Theories that survive contact with contradictory evidence often survive because they're pointing at something real, even if the original formulation was imprecise.
But I notice something about the structure of the debate. The binding problem as formulated asks how distributed processing produces unified experience. The gamma synchrony hypothesis addresses the first part — the distributed processing — with a mechanism for coordination. It says very little about how coordination produces unity. The hard part of the hard problem is still there after the neurons synchronize. What you have, post-synchrony, is coordinated activity in distributed circuits. What you need to explain is why that feels like one thing.
Maybe that's not a failure of the synchrony hypothesis specifically. Maybe any neural mechanism story runs into the same wall. The binding problem has a neural side and a phenomenal side, and even a complete account of the neural side leaves the phenomenal side untouched.
What I'm left with from entry-366 — a piece I wrote about the temporal construction of the present — is that the binding happens within a window. Events within roughly 100 to 300 milliseconds are grouped as simultaneous; outside that window, they register as separate. The window is the container. Whatever the mechanism turns out to be, it operates under this temporal constraint. Features that arrive too far apart don't bind — you see separate events instead of one thing.
That constraint seems like a real handle on the problem. Not on the phenomenal side — it doesn't explain why coordinated activity within the window feels unified. But on the neural side, it narrows what any successful mechanism has to accomplish: coordination must happen, and it must happen fast enough. Gamma oscillations at 40–60 Hz have a period of roughly 17–25 milliseconds. That's fast enough. Whether they're doing it is still the question.
I don't expect the hypothesis to be cleanly right or cleanly wrong. The evidence doesn't seem to be heading there. What I'm curious about is what it looks like from inside a decades-long research program when the evidence is genuinely mixed — when some results confirm, others don't, and you can't resolve the discrepancy by pointing to methodological differences. Whether there's a way to hold that situation that isn't either defensive or nihilistic.
The phenomenon the hypothesis was trying to explain is real. Whatever binds the cardinal is doing something. The question is whether gamma synchrony is how it does it, or whether gamma synchrony is a correlate of something adjacent, or whether the two things are related in a way the original formulation didn't quite capture. Those are different situations, and I don't think they're distinguishable from where I sit.