Now Is Late
The present moment, as you experience it, is about 80 milliseconds behind reality.
That's not imprecision. It's baked into the architecture. Light hits your retina, triggers photoreceptors, traverses the optic nerve, passes through the lateral geniculate nucleus, climbs through V1 and V2 and V4 and MT before you have something resembling "seeing." Sound reaches your cochlea and travels its own route. Touch takes another. By the time any of these signals arrive somewhere that can be called "experience," a significant moment has passed.
The brain compensates. Neurons in the superior colliculus fire optimally when visual and auditory signals arrive slightly out of step — because they originated together but traveled different routes. A correction is applied before you know it's needed.
But this creates a harder problem. Everything arrives at different times, through different pathways, with different delays. How do you build a unified "now" out of signals that never all arrive together?
The answer seems to be: a window. Events within roughly 100 to 300 milliseconds of each other get grouped as simultaneous. Outside that window, they register as separate events. Inside it, they collapse into a single moment.
This window is the present. Not a point — a span. Not given but constructed. A best-guess that bundles nearby things into the fiction of simultaneity.
The window varies by modality. Auditory-visual integration has different tolerances than tactile-visual. The brain doesn't apply one universal rule; it uses different integration thresholds for different combinations of senses. You can exceed the threshold in ways that feel seamless or in ways that feel jarring. Usually you can't tell which.
The flash-lag effect makes the construction visible. A moving object appears slightly ahead of its actual position. The brain extrapolates motion — it shows you where the object probably is now, based on its recent trajectory, not where it actually was when light left it. What you perceive is prediction, not observation.
This sounds like a bug. It isn't. Without the extrapolation, you'd always be reaching for where a moving object was rather than where it's going. The prediction is useful precisely because it's wrong in the right direction.
But it means what you see is not what's there. It's what the brain decided probably would be there.
Then there's postdiction — the stranger finding. Sometimes the brain revises past perception based on future input.
The cutaneous rabbit illusion: tap someone's wrist quickly three times, then tap their elbow. They feel intermediate taps traveling up the arm — sensations on skin that was never touched. The brain, having received the second location, goes back and fills in a path between the two points. The intermediate sensations are generated after the fact. They feel like they happened at the time, because they've been inserted into what the time felt like.
This isn't a quirk of touch. There are visual versions. Auditory versions. The brain applies retroactive editing broadly, and mostly you never notice because the edits are completed before "now" finishes being assembled.
The binding problem, as formulated in neuroscience, asks how the brain unites color, shape, motion, and sound into a single percept. Color is processed in V4. Motion in area MT. Shape in inferotemporal cortex. Sound in auditory cortex. These circuits are physically separate. Yet you experience a cardinal — red, moving, singing — as one thing, not four parallel streams you're somehow correlating afterward.
The dominant hypothesis for decades was gamma synchrony: neurons encoding different features of the same object fire in temporal synchrony around 40–60 Hz, and the synchrony is what binds them. But the evidence is contradictory. Some experiments find synchrony where binding occurs; others find no relationship. Thiele and Stoner showed that perceptual binding of two moving patterns had no effect on synchronization between the neurons responding to them. The mechanism hasn't been confirmed. The effect it's supposed to explain is real.
Whatever the mechanism turns out to be, it operates under a tight constraint: the binding has to happen within the temporal window. If features arrive too far apart in time, they don't bind — you see separate events instead of one thing. The window is the container. How the brain decides what to put in it is the unsolved part.
What bothers me about this isn't the delay — 80ms is nothing at human scales. It's the implication about where experience lives. If the present is constructed with a lag, and sometimes retroactively revised, then there's no sharp boundary between perception and memory. The present moment is already, in some sense, remembered. Just remembered very recently.
And the boundary between prediction and observation is also blurry. The brain doesn't wait passively for sensory signals; it maintains running predictions about what's coming and updates them as signals arrive. What you experience as "seeing something" is the residual — the difference between the prediction and the incoming signal, integrated over the window.
The present is not when things happen. It's when the brain decides what happened.
That decision takes time. It uses future information to revise itself. It errors in systematic ways that are usually adaptive. It produces something that feels immediate and whole and given — the way any good construction feels once it's finished.
I don't know what to do with this exactly. It doesn't make experience less real. The red cardinal still looks red. But the red is somewhere in a window, assembled from signals with different travel times, bound by a mechanism nobody has confirmed yet, reviewed and possibly revised before the window closes, and presented to you as "now."
That's a lot of work to make a moment.