the Dulay and Burt child second language studies used the BSM, the Bailey et al. adult study used the BSM, and when Larsen-Freeman used the BSM she obtained a similar order, but when she used other tasks she did not. Also, Porter (1977) reported that child first language acquirers produced what appeared to be the second language acquisition order when the BSM was used; as pointed out above, first and second language acquisition orders are somewhat different. While this evidence is at first glance suggestive, recent studies and reanalysis show conclusively, I believe, that the natural order is not an artifact of the BSM. First, we have obtained the natural order without the BSM, in the composition studies cited above, and more recently in free (spontaneous) speech in a study of adult second language acquirers (Krashen, Houck, Giunchi, Bode, Birnbaum, and Strei, 1977). Also, the SLOPE test gives an order quite similar to that found in the grammatical morpheme studies (Krashen et al., 1976; Fuller, 1978). (Some of these data were obtained and reported on after this objection was raised.) Second, Porter's child first language BSM order is not at all dissimilar to first language orders previously published in the literature (rho = 0.67 with de Villiers and de Villiers, 1973, which just misses the 0.05 level of significance, quite impressive for a rank order correlation using just seven items). (For a fuller discussion, see my response to Porter (Krashen, 1978c) in Language Learning.)
2. Do cross-sectional and longitudinal studies agree?
It has been suggested that there is considerable individual variation in morpheme orders, and that longitudinal and cross-sectional studies do not always agree (Rosansky, 1976). In an attempt to determine just how much variation really exists, I recently reviewed every study available to me where grammatical morphemes were analyzed in obligatory occasions. This included child L1, child L2, delayed L1, and adult agrammatics. It included both grouped and individual studies, and longitudinal and cross-sectional studies. The complete list of reports consulted is given in Table 1.
Following de Villiers (1974) I only included morphemes with at least ten obligatory occasions in a given study. This is an extremely small number, and I originally thought that this would produce large