the fact that as least some of the corrections were aimed at morphology. We were struck by the fact that natural orders, at least in our laboratory, were turning up everywhere, and we replicated one of the few studies that reported to give an unnatural order, Larsen-Freeman's "writing" task. The replication was successful; this task again produced a clearly unnatural order, one that was nearly identical to Larsen-Freeman's reported order. While both tasks seem to imply a focus on form (self-corrections and the writing task), there are crucial differences. In their self corrections, the composition students were not focused on a particular item or on a specific rule, and apparently did the task on the basis of their "feel" for grammaticality. Their corrections were apparently motivated by their desire to communicate. In Larsen-Freeman's writing task, which included sentences like:

Last year he (work) ____________ in a factory

where the subject only had to correctly inflect the verb, there might have been a greater tendency to call forth conscious rules for specific items. In other words, Larsen-Freeman's task was what is called a discrete-point test, while ours was not. This leads to the interesting hypothesis that it takes a discrete-point test to bring out conscious learning, while anything less does not, at least for subjects who have had a chance to do a meaningful amount of natural acquisition. Thus, we might expect a discrete-point test to yield an unnatural order and an integrative test to yield a natural order. So far this is what we have found. (In a previous paper (Krashen, 1976a) I suggested that integrative tests were not necessarily acquisition tests. In those days, I had only considered the time element as the necessary condition for Monitor use. This was clearly inadequate, and I thank Heidi Dulay and Marina Burt for pointing out that having time is not sufficient to ensure Monitor use, and John Oller for pointing out that my conclusions on integrative testing were premature, suggestions that are certainly confirmed by these current data.)

Janet Keyfetz Fuller's dissertation (Fuller, 1978) also speaks to this point. Fuller administered the SLOPE test to adult ESL subjects in both written and oral versions. She tested the effect of first language (Indo-European versus non-Indo-European), second versus third language acquisition, and order of presentation. In general, there were no

54