This paper argues that, on one hand, there are compelling theoretical reasons to believe that interlanguage (IL) grammars are both systematically and randomly variable, and that the relationship between the two types of variation is a complex one. At any one stage of IL development, some structures may be systematically variable, but at the same time the existence of forms in free variation creates conditions conducive to systematic variation "setting in" at a later stage of development. On the other hand, the paper argues, there is no incontestable empirical evidence for free variation and concludes that the existence of free variation is based more on speculation than on empirical grounds. (Contains 14 references.) (JL)