What is the evidence of the typological cycle theory?

Question

There is a theory that languages move from one morphological typology (isolating, inflected, agglutinating) to the next in a usual, predictable cycle.

What evidence is proposed to support this theory?

As background, I am also familiar with natural sciences and usually things move to the most stable state (macroscopically. for example, things go down, hydrogen atoms fuse to become helium atoms, etc) and do not naturally go back without external force. I am wondering why this isn't the case for languages and why languages go in cycle instead.

Answer 1

The reason the analogy with the physical sciences breaks down, I think, is that languages are transmitted through learning and, during the course of that, there can be perceptual distortions, production errors, processing mistakes, reanalysis, etc.

None of isolating, inflected, agglutinating are any more stable than the others so there's no natural state for a language to settle on. Instead, language changes put pressure on other parts of the language to change, either through analogy or a need to disambiguate things which have been conflated.

As to the evidence, processes like grammaticalization and periphrasis can been seen in a lot of languages; and those that have a long enough written record bear testimony to going back and forth—perhaps not always in a full cycle but certainly oscillating between alternatives, neither of which is any more stable than the other.

Answer 2

The major piece of evidence is the mere existence of agglutinating and inflecting languages.

If there were only evolutionary processes that make formerly inflecting or agglutinating languages isolating , there would be no inflecting or agglutinating languages (given the time language is around, all languages would have reached the isolating stage and stay there forever. Also, Pidgins and Creoles are typically isolating, so they do not provide a source of inflecting languages).

It is of course theoretically possible that an isolating language becomes directly inflecting without and agglutinating intermediate step, but afaik this transition has not been observed yet.

It is, of course, somewhat unsatisfactory to have only a "non-constructive" argument for the typological cycle, seeing all proposed transitions in action (and not seeing the transitions the other way round) would be more satisfactory.