The chapter is a printing convention before it is anything else. Manuscripts were divided for practical considerations -- a scribe's stint, the capacity of a gathering, the necessity of locating a passage again -- and early printed books inherited the divisions along with much else. Nothing about the divisions was expressive. They designated where a unit of production terminated.
What happened subsequently is the interesting part. A convention that exists for one reason will, if it is sufficiently consistent, acquire a meaning, because readers learn to expect something at the place where it falls. A chapter break came to mean a pause; a pause came to mean that time might have passed, or that attention was about to be redirected elsewhere; and once readers had formed that expectation, writers could use it. The break stopped being a place where the text was interrupted and became a place where the writer could accomplish something that necessitated an interruption.
The effects available at a break are ones no sentence can produce, because they are dependent on the reader stopping. A chapter can end on a statement, and the reader then has time to determine its consequences before the subsequent chapter confirms or repudiates them. It can end at the moment of a decision and open after the decision has been implemented, so that the act is reported and not shown. It can end twice on the same occurrence witnessed by different individuals, and the second narration derives its force from the gap. All of these are structural, and all of them presuppose a reader who interprets the white space as instruction rather than as accident.
The point generalises beyond the chapter. Verse lines commenced as a constraint of oral composition and became the site of enjambment. That effect is available only because a line end is expected to coincide with a pause and can be made not to. The paragraph originated as a mark in the margin and became a unit of argument. In each case a feature that recorded how the thing was constructed was absorbed into how it signifies, and the absorption is not decadence or accident but the ordinary way expressive resources are acquired.
It has a practical corollary for anyone writing in a new form. The conventions of a medium in its first decades are almost invariably production artefacts, and the ones that will ultimately matter cannot be identified beforehand, for what makes a feature expressive is not anything about the feature but the fact that readers have come to expect it. The resources of a form are fashioned by its audience, gradually, and the writer's contribution is to notice what has already transpired. This is why the most productive question to ask of a new medium is not what it can accomplish that its predecessors cannot, but which of its accidents its audience has begun to expect -- a question that can be answered only late, and only by observation.
