One thing that is certain about human language is that it continuously changes. A point that is almost too obvious even to mention is that however a language changes through time, it must remain a human language. In other words, the changes cannot be random in nature. An important idea put forth by some linguists is that the changes are predetermined in the sense that diachronic variation within a particular language directly refl ects variations found across contemporary languages; diachrony is synchrony when it comes to variations that can occur. In a clear expression of this idea, Lightfoot (1979, viii) observes that the formulation of “a possible grammar will provide the upper limits to the way in which a given grammar may change historically, insofar as it cannot change into something which is not a possible grammar.” In a similar vein, Joseph (1980, 346), in addressing the loss of the infi nitive form in Greek, notes that “[u]niversal constraints which hold in synchronic grammars are used to explain the direction taken by certain changes in syntax.”