This article identifies empirical evidence (Dao, 2007; in preparation) conflicting with Processability Theory's (PT) prediction that in acquisition of English as a second language (ESL), plural-marking emerges first in bare nouns and only later in numeric expressions. Specifically, it presents results from Dao's (2007) cross-sectional study of ESL in 36 Vietnamese learners, which was designed to test PT's predictions that inflections emerge in lexical contexts before agreement in phrasal contexts, but found that emergence occurred in the reverse order. The article explores whether Dao's findings invalidate PT's crosslinguistic principles or whether there is a problem in applying these to language-specific empirical contexts. The exploration reveals weaknesses in the description of PT's principles, as these are based on implicit assumptions, which may be invalid in specific first language / second language (L1/L2) typological contexts and thus lead to incorrect predictions. The findings are explained by reference to L1 transfer represented in the framework of one of PT's feeder theories: Levelt's (1989) Theory of Speaking as modelled in Weaver++ (Levelt et al., 1999). Our L1 transfer account is in line with PT's Developmentally Moderated Transfer Hypothesis.