ABSTRACT
Predictive IL data It is time to provide some detailed CA and EA data to gain a deeper understanding of the contrasts discussed so far in this volume. Our continuing aim is to inspect their 'value', in the Saussurian sense discussed above, as potential units of IL and IL learning. The point of classical CA is to predict 'errors' that would (or could) be made by an L2 learner of one NL specifically attempting to master another NL, called the TL. Even with the difficulties of the enterprise of error prediction as discussed above, the success of that research enterprise is that it (especially through empirical calls by Lado) inspired researchers actually to go out and look at learner speech with specific predictions in mind. Classical CA, EA and bilingualism studies gave those predictions a framework, provided empirical training to SLA scholars and helped in no small measure to bring us to IL and the flow of studies we currently see in SLA. In particular, CA and EA data lead to predictive IL data in terms of possible interlingual identifications.
