Writers talk about scenes constantly and rarely define them, which leaves a beginner assuming a scene is whatever happens in one place before the setting changes. That definition is close to useless. A scene is a unit of story with three properties, and sequences that lack them feel inert regardless of how much occurs inside them.
The three are a boundary, a question and a change. Nearly every complaint that a section feels slow, or that nothing is happening, comes down to one of the three being absent, and diagnosing which one is missing is far quicker than rewriting on instinct.
The boundary
A scene needs a beginning and an end that a reader can feel, even if no chapter break marks them. Boundaries come from location, time, participants or a shift in what the story is attending to. Without one, sequences run together and the reader loses the sense of progress that comes from finishing units.
Boundaries are also where control over pace lives. A story made of many short scenes reads quickly; one made of few long ones reads deliberately. Neither is better, but drifting between them without noticing produces a rhythm nobody chose, which readers experience as unevenness.
The question
Every scene should raise something the reader wants resolved. Will she say it. Does the other one already know. Is this going to go wrong. The question does not need to be dramatic and it very rarely needs to be stated; it needs to exist, because it is what makes a reader continue rather than merely keep reading.
Scenes without questions are usually scenes where the writer knew what had to happen and simply staged it. That is a common and understandable failure, and the fix is often as small as withholding one piece of information that was previously given away in the first line.
The change
Something must be different at the end. A decision, a piece of knowledge, a shift in the balance between two people. It can be tiny. What it cannot be is nothing, and the test is to state the change in one sentence, which is the same test recommended in the piece on writing two-handers.
Scenes that fail this test are usually doing something else useful, like establishing a place or a mood, and the honest answer is to fold that work into a scene that also changes something. Standalone atmosphere scenes are expensive, and readers skim them more often than writers like to believe.
Enter late, leave early
The oldest advice about scene construction remains the most reliable. Begin after the situation has already started, and end before everything is settled. This removes the preamble that writers use to warm up and the resolution that a reader is happy to supply themselves.
It is an editing operation rather than a writing one, which makes it easy to apply in a second pass. Cut the first paragraph and the last, then check whether anything essential was lost. In a surprising proportion of cases nothing was, and the scene reads considerably better.
Scene questions should chain
A sequence of scenes each asking the same question produces the repetition problem described in the piece on pacing. Each scene's change should create the next scene's question, so the chain pulls a reader forward rather than presenting a series of similar situations.
This is also how a story about a relationship avoids stalling. If the same uncertainty persists across five scenes, the relationship is not developing, and no amount of variation in setting or activity will conceal that from a reader who is following it.
Testing a scene in three questions
Ask: where does this begin and end, what does the reader want to know, and what is different afterwards. Three answers, each one sentence. A scene that cannot supply all three has a structural problem that revision at the sentence level will not fix.
The staging side of the same discipline is covered in the blocking piece, and characters here remain fictional and unmistakably adult throughout.