Is it possible to predict language changes?

Question

The comparative method is used to reconstruct unattested languages from the attested ones. By comparing different sounds for the same words in various sister languages, it is possible to infer some phonological rules. The internal reconstruction method, on the other hand, compares variant forms within a single language to discover how they evolved from a past regular form.

Would it be possible to combine these methods to "look forward", so to speak? For example, comparing several dialects of a language with how they were fifty years ago and then extrapolate this data to predict how these dialects will change in the near future? Of course, some degree of uncertainty would have to be allowed.

Answer 1

It's tough. On the one hand, in phonetic changes like the Northern Cities Shift, it is possible to record successive generations of speakers and measure acoustically how their vowel production (in this case) slowly moves through vowel space.

It gets trickier for change that is categorical in nature (rather than continuous). Johnson's dissertation tracks a known merger (between cot and caught) along the boundary, and identifies certain families in non-merged territory where children appear to suddenly merge these sounds. Predicting entirely new changes is a tricky business – it is possible to say what changes are likely to happen in a language, but their realization depends on many factors, some of them completely random. In general, I'd say that we don't yet have the ability to predict a novel categorical change in a specific time window (50 years, like your question).

To give one concrete example, English has been "likely" to develop negative concord (NC) for centuries as the next step in the Jespersen Cycle. Optional NC is nearly ubiquitous in low-prestige vernacular speech, but it has not entered the standard language in a unified way, due (in all likelihood) to prescriptive norms against it.

Answer 2

I think it all depends on how you define "near" future. To make things concrete, we have a very large corpus of spoken English from Philadelphia, where dates-of-birth of our speakers range from 1888 to 1991. Looking at the historical trends across 100 years, I'd feel comfortable making predictions about how the language will look in 3 to 4 years from now, but I think what things will look like 10 to 15 years from now is totally up in the air.

For instance, in the 1970s, it was reported that the vowel /aw/ (as in house and down) was fronting and raising. This fronting and raising pattern looks pretty robust in data collected all the way through the 1980s, into the early 1990s. If you had asked me in 1995 what /aw/ would sound like in 2020, the best guess I could have made based on the available data would be that it would continue shooting up to be something like [iw].

However, starting in the mid 1990s, the data starts to turn around, and it looks like /aw/ is taking a nosedive towards a lower, backer pronunciation. As such, I'd guess that in 2015, /aw/ is going to be a little lower and backer than it is now, but I don't want to bet any money on what'll be going on in 2030.

---

In more general terms, there are very few language changes which are guaranteed to happen. As @Aaron said, English has been ripe for wide-spread negative concord for centuries, but it hasn't happened yet. All languages which allow voiced, word-final consonants are ripe for developing word-final devoicing, but it's only going to happen in a small subset of them.

The issue is that at any given point in time, there are a lot of language changes which are likely, and for all language changes, it seems obvious why they occurred, post-hoc. However, which language changes are actually going to occur next is up to a lot of historical and social chance.

The whole thing is very comparable to trying to figure out what kind of genetic evolution is going to occur next. As always, some are more likely than others, but non are guaranteed to occur.

Answer 3

Language is a complex system used simultaneously by a large number of people. Changes typically begin on a very limited area and spread to eventually encompass the entire language. (Or not. That's a simple version of how dialects come into being.) Whether they succeed and spread or fail and die out, depends on how people (individuals) talk to one another.

In other words, changes in language spread not entirely unlike diseases. It is usually relatively easy to find out how and where a pandemic started but it is very difficult to foresee one. There are many places in the world where the hygienic conditions are horrible but somehow a pandemic does not break out. (That is similar to Aaron's example of English having been ripe for developing NC for centuries.) And at the same time, there are places where hygienic conditions are not bad at all, and yet a disease can from time to time manage to spread very far.

Another analogy would be road or air traffic. When an accident happens, we are usually able to reconstruct how it happened. But regrettably, we are no good at predicting where and when one is going to happen, so we can't prevent it.