Reading the Room When There Is No Room

The common view is that a video call is an impoverished version of a meeting: you get the face and lose the rest. That description is wrong in a way that matters.

What video removes: the body, the hands at rest, the orientation of the torso, physical distance, and the timing that makes turn-taking feel natural. What it adds: network latency, compression, a frame the other person chose, and your own face in the corner of the screen.

The trouble is that the added items look like behaviour. A three-hundred-millisecond network delay is indistinguishable from a hesitation. And hesitation, read commercially, means something.

The three artefacts you will read as signals

Latency. Every pause on a call contains a component that is network. On a poor connection that approaches half a second — precisely the order of magnitude of the pause you would interpret as discomfort. There is no way to separate them by watching. Only by comparing against the same person's baseline on the same call, which is the only way to hold the network condition constant.

Gaze direction. Nobody can look at the camera and at the other person's face simultaneously. Anyone looking at the image of their counterpart — the natural behaviour — appears to be looking down. The person holding perfect eye contact is staring at a lens, not at you. The signal you think you are reading is hardware geometry. Gaze aversion already meant less than people claim; on video it means nothing at all.

Framing. The other person chose what appears. Tense shoulders, hands, foot movement — all of it may simply be outside the frame. Absence of a signal is not a signal of absence.

What survives

Three channels remain valid, and every one of them requires a baseline:

1. Pace and volume of speech. These survive compression. Someone who spoke in long blocks and starts answering in short sentences has changed state, and that is not a network artefact.
2. Relative latency. As with written communication, the network delays everything equally. If a person answers quickly on numbers and slowly on commitment, the difference between the two is real, because the connection is the same in both cases.
3. Who drops out of the conversation. On a multi-person call, who stops contributing and when is the richest information available — and it is more visible than in person, because silence cannot be disguised by physical presence.

What to actually do

Not lighting and framing advice, which exists everywhere. Advice on producing information:

• Run the four-minute baseline anyway. It matters more on video than in person, because you are calibrating the network at the same time as the person. The light opening questions measure both.
• Ask for the camera in the invitation, not mid-call. A meeting without video removes everything. But "could you turn your camera on?" halfway through is an imposition that changes the behaviour you wanted to observe.
• Turn off your own image. Watching your own face during a conversation raises self-consciousness and fatigue; prolonged self-observation is one of the mechanisms proposed for videoconference exhaustion.¹
• Do not fill the silence. The urge is stronger on video, because a pause looks like a technical fault. That is exactly when the information arrives. We covered why in Silence Is the Cheapest Question.
• For decisions that matter, change medium. A voice call without video frequently produces more information than video: it removes the visual artefacts and leaves pace and relative latency intact.

The uncomfortable conclusion

Video is worse than in person for reading people and better than text. That leaves it in an awkward place: good enough that you believe you are reading someone, poor enough that you are frequently wrong.

Treat a video call as text with a soundtrack: use it for content, not for inference. When the inference matters, change the channel.