Time for change
In order to get an angle on deep-time historical relationships between languages, beyond what can be fathomed by the Comparative Method, and in order to model and thereby understand the evolution of typological diversity, attention is increasingly being paid to the question of time-stability of lexical as well as of grammatical traits. Proceeding by inference rather than through longitudinal study, crosslinguistic distributions have been interpreted as revealing how stable or unstable particular traits are. Despite all methodological sophistication, the conclusions that have been reached about time-stability in this indirect way are alarmingly contradictory. As a corrective, I suggest that this research programme be reoriented and that time-stability be studied directly, namely diachronically. Within this general context, the particular issue addressed here is the tempo of change: traits will appear relatively time-stable, not only if they are wholly resistant to change, but also if the tempo of changes affecting them is slow. When this matter is addressed at all, the literature again is remarkably contradictory: uniformitarians would assert that the tempo of change is uniform and diversitarians that it can randomly be rapid or slow. A particular development, the grammaticalisation of a local adposition ‘at’ from a noun ‘dwelling, home’, will be examined in detail here with the aim of determining the length of time this kind of change takes and of comparing its tempo across several languages where it has occurred. Relevant instances are French <i>chez</i> ‘at’ from Late Latin <i>casa/chiés</i>; Swedish, Danish, Norwegian <i>hos</i> ‘at’ from Old Norse <i>hus</i>; Icelandic and Faroese <i>hjá</i> ‘at, next to, by, with; of’ from Old Norse <i>hión</i> ‘family, household’; and late Pāli <i>gē</i> ‘at; of’ from Prakritic Indo-Aryan <i>geha</i> (with the postposition subsequently turned into a suffix in Sinhalese and Maldivian). All four occurrences have indeed taken about the same length of time to reach completion: approximately 400 years, or some 16 generations, 16 cycles of acquisition. I conclude that grammaticalisation of this kind is very slow, and ceteris paribus proceeds at a uniform tempo. I suggest that the most significant factors that can prolong change are that a change is a whole cascade of individual reanalyses rather than elementary and that it diffuses through speech communities slowly rather than rapidly.