City-to-city spread of speech forms

Why does a new pronunciation reach a distant city before the village next door?

The intuitive picture of a spreading pronunciation is a stain on cloth. It starts somewhere, wets the fibres next to it, and creeps outward at a steady rate, so that if you mapped it year by year you would get widening rings. Distance from the origin would predict everything.

Real dialect maps almost never look like that. A feature that starts in a capital turns up in a large city three hundred kilometres away while the market town forty kilometres out still does not have it, and the villages between the two cities acquire it last of all — sometimes decades later, sometimes never. The rings are broken, and the breaks are not random: they line up with population. Something other than distance is doing the steering.

Ask what a pronunciation actually needs in order to move. Not a road — a conversation. A speaker acquires a variant by hearing it repeatedly from people they talk to and have some reason to sound like. So the map we should be drawing is not a map of physical space but a map of who talks to whom, and how much.

Now ask who talks to whom. Where do the residents of a large city travel for work, business, university, family? Overwhelmingly to other large cities. Where do the residents of a village between them travel? To their nearest town. The rail line, the motorway, the flight, the labour market and the marriage market all connect the big nodes to each other directly, hopping over what lies between. If contact is the medium, the network is hierarchical, and a hierarchical network delivers to distant hubs before it delivers to near leaves.

That intuition has a rough quantitative form, borrowed by sociolinguists from geography: a gravity model. The interaction between two places is taken to rise with the product of their populations and fall with the square of the distance between them. Peter Trudgill applied a version of this to dialect diffusion in the 1970s, adding a term for prior linguistic similarity — a place adopts more readily from a variety already close to its own. Put a capital of eight million and a distant city of a million into that formula and the numerator can easily overwhelm the distance penalty that a village of four hundred could never overcome.

But quantity of contact is not the whole of it, or every innovation would spread and none would be resisted. Ask the second question: why would a speaker take the variant up? Usually because it carries some social value — it sounds urban, modern, unmarked, like the people whose lives look like the life the speaker wants. That is what "prestige form" names, and it is a directional term: prestige flows down the settlement hierarchy far more readily than up. A capital pronunciation arriving in a provincial city reads as cosmopolitan; the provincial pronunciation arriving in the capital reads as provincial. The gradient in contact and the gradient in evaluation point the same way, and they multiply.

The classic illustration is the uvular r that spread out of Paris from around the seventeenth century onwards: it appears in Copenhagen, Berlin, Bergen and other northern European cities well before it appears in the countryside surrounding any of them, and in several regions the rural hinterland never took it at all. City to city, then downward.

Two honest caveats. First, the gravity model is a description that fits reasonably well, not a mechanism that has been measured directly — the exponents are fitted, and critics point out it can be tuned to match almost any map after the fact. Second, the flow is not always downward: linguists document counterhierarchical diffusion too, where a rural or low-prestige form spreads upward and outward because it carries covert prestige — toughness, authenticity, local belonging. The hierarchy biases the flow; it does not dictate it.

It helps to name the two competing pictures. Wave or contagion diffusion is the stain on cloth: spread by adjacency, and it does genuinely happen — for features that move through everyday local interaction across a settled countryside, the rings are real. Hierarchical or cascade diffusion is the city-to-city pattern. They are not rivals so much as two regimes, and which one a given feature follows depends on whose network is carrying it.

A misconception worth heading off: the mass media are usually not the vehicle here. Broadcast exposure is enormous and largely one-directional, and the evidence is that speakers do not acquire phonological systems from a screen the way they acquire them from people they interact with. Media may spread individual words and catchphrases; the structured, gradual, socially graded change described here tracks face-to-face contact networks, which is why commuting patterns predict it better than viewing figures.

• How well gravity models actually predict, and the critique that they are fitted after the fact rather than derived.
• Covert prestige and counterhierarchical spread: which kinds of features travel upward, and why.
• What happens to diffusion when the network changes shape — new rail links, remote work, or online communities of practice.

Every term the collection defines is gathered in the glossary.