Classic studies in second language (L2) learning offer little evidence for the validity of the notion of universals in L2 learning. The present study investigates this notion in data collected from 15 observational studies of 40 L2 learners who acquired the L2 naturally, that is, they were not taught the language in any formal ways. Interpretation of both input data and production data (e.g., L2 auxiliary development, auxiliary development with negation, auxiliary inversion in question formation, and input frequency and question production) suggests that communication is the principal goal of such learners, and that consequently those parts of the language system which are not important to communication are learned slowly. However, if a structure is extremely frequent in the input data, the learner will produce it. Effects of frequency are modified in a number of ways. If a form has low semantic power, or if it requires changes in word order, it will be learned late, as will structures having a variety of forms. Variations in strategies and speed of L2 acquisition reflect psychological differences, and must be understood before a definitive model of universals in L2 learning can be developed. (DB)