Band 8+ Academic Reading Practice Test 1

A city that draws its population from many places might be expected to produce speech that averages them out, and it does not. Large migrations into industrial cities in the nineteenth century produced, within two or three generations, varieties as distinct from their sources as the sources were from each other. The mechanism by which a mixed population converges on something new rather than on a compromise is one of the better-documented and less intuitive findings in the study of language.

The first stage is levelling, and it is largely subtractive. Features that mark a speaker as coming from one particular place, and that other speakers find hard to reproduce or hear as marked, tend to be dropped within a generation. What survives is not the average of the inputs but the subset that is either shared by many varieties or perceptually neutral, which means the outcome is already skewed before anything new has been created.

The second stage supplies what the first removed. Children acquiring a language from a mixed adult input do not select randomly among the variants they hear; they regularise, extending a pattern to cases where no adult applies it, and they occasionally settle on a variant that was rare in the input because it was structurally simpler. This is the point at which a new variety acquires features that none of its sources possessed, and it is a matter of acquisition rather than of contact.

Identity then does the work of stabilising the result. Once a variety is associated with a place and a set of people, using it becomes a claim about who the speaker is, and that claim is made most strongly by those with most at stake in it. Studies of adolescent speech repeatedly find the sharpest local features not in the most isolated speakers but in those most invested in local networks, which is why prediction from geography alone performs badly.

The commonest mistake in interpreting all this is to treat the outcome as decay. A variety that has lost a distinction its parent had has not lost precision, since the distinction is generally recoverable from context or has been replaced by another mechanism, and the varieties concerned develop distinctions of their own that the standard lacks. What changes is which features carry information, not how much information the system can carry.

The practical consequence is that dialect levelling and dialect creation are the same process observed at different stages, and describing a region as losing its accents therefore reports the first half of a sentence. Whether the second half is visible depends almost entirely on how long the observer has been watching, and most popular commentary is written on a timescale shorter than the phenomenon it describes.