This paper reports on an experiment investigating the acquisition of Spanish, a language that has a gender feature for nouns and gender agreement for determiners and adjectives, by speakers of a first language (L1) that also has gender (French), as well as an L1 that does not (English). Number (present in all three languages) is also investigated. Subjects were adult learners of Spanish, at three levels of proficiency, as well as a control group of native speakers. Oral production data were elicited. Subjects were also tested on an interpretation task, in which the selection of pictures corresponding to particular sentences depends on number and gender contrasts. The results from both tasks show significant effects for proficiency; low proficiency groups differ significantly from native speakers, but advanced and intermediate groups do not. There were no significant effects for L1 or for prior exposure to another second language with gender. The findings are discussed in the context of two different theories as to the possibility of parameter resetting in nonnative acquisition, namely, the failed functional features hypothesis and the full transfer full access hypothesis. The results are consistent with the latter hypothesis.