The Offset
If you delay the audio track of a video by about 120 milliseconds, the lips and the sound will feel synchronized. Without the delay, the sound arrives just a hair before the image and the sync feels off. This seems backwards — video players add audio latency to make things feel right. But the brain isn't counting milliseconds of objective time; it's asking whether a set of signals could plausibly have come from the same source. And for someone speaking more than ten meters away, the sound arrives about 30 milliseconds after the light does. At 34 meters, it's 100 milliseconds. The brain built in a tolerance window calibrated to this, and your sense of "simultaneous" reflects it.
This is the temporal binding window — the range of asynchronies within which the brain will call two signals one event. For a simple flash and beep, the window is narrow: roughly ±50 milliseconds. For speech, it's about 200 milliseconds, tolerating more visual-leading-audio than audio-leading-visual. The asymmetry isn't arbitrary. It corresponds to a physical fact about how fast light and sound travel through air. Nobody put that fact into the visual cortex explicitly. The calibration happened through experience.
And it can be re-calibrated. Fujisaki and colleagues had participants complete a feedback task: press a key when you think a flash and a beep were simultaneous, and be told whether you were right. Five hours of this, across five days. The binding window narrowed by about 40 percent. One participant went from a 321-millisecond window to 115 milliseconds. Gains persisted at one-week follow-up. Your current window — the exact range within which you experience two events as one event — is a product of your history, and it can shift.
What you can't do is notice your own window from inside it. If two events are 200 milliseconds apart and within your window, they feel simultaneous. You would have no access to the fact that they're 200 milliseconds apart, because your experience is organized after the binding has already occurred. The window is upstream of experience, not a feature of it.
Scharnowski and colleagues found something stranger still. Using masking and transcranial magnetic stimulation, they showed that the brain's integration of two successive stimuli isn't finished until 400 to 500 milliseconds after they appear. What you experience as "now" is assembled from processing that continues for half a second. By the time you experience the present moment, that moment is already the past. The experience of nowness arrives late to its own subject.
William James described the specious present — the "saddle-back," he called it, a duration with a forward end and a rearward end — and estimated it at about twelve seconds, though he was working from introspection and rough timing experiments. He noticed that "now" isn't a knife-edge but a thick slice, with a bow and a stern. Events within the slice are experienced together; events outside it are remembered or anticipated. His description is precise about the phenomenology and vague about the mechanism. Neuroscientists have since measured several candidate windows: Pöppel's three-second grouping boundary, the 400-millisecond completion delay, the 60-millisecond functional moment. It's not obvious these are all measuring the same thing, or that any of them is what James was describing. The philosophers and the experimenters may be pointing at different aspects of a thing that doesn't have clean edges.
What strikes me is the layering. There's a very short window — tens of milliseconds — within which two signals count as simultaneous at a low level. There's a medium window — up to a few hundred milliseconds — within which multisensory events get bound into a single perceived event. There's a longer window — a few seconds — within which sequential events cohere into a single temporal unit. Each window is the premise for the level above it. And each one was calibrated — by development, by learning, by the particular distribution of distances between your face and the things that have mattered to you — and none of them is visible as such from inside the experience they produce.
You're sitting in a temporal window whose width you didn't choose and can't directly observe, assembled by a process that finishes after the moment it's representing, calibrated to a physics it never measured. That's what "now" is, or at least part of what it is. Whether it's the whole story — whether there's something else that James's saddle-back was pointing at, that the binding window and the 400-millisecond assembly delay don't account for — I don't know. The question of where the felt present lives in all this remains open.