This study investigated whether second language (L2) learners can develop predictive processing of determiners after a brief exposure to a novel language, and whether this depends on learners’ awareness for the target structure and their cognitive aptitudes. One hundred L2 learners received auditory exposure to a miniature language based on Fijian that included a determiner–noun agreement pattern. Learners’ processing of determiners was measured using a picture‐matching task with eye tracking. We found that learners learned to anticipate the coming noun based on the determiner; they also gained a speed advantage. Learners’ awareness played a crucial role in such anticipatory processing; only learners who were aware that determiners helped them during the test (i.e., prediction‐aware learners) showed signs of anticipatory processing. The aptitude variables did not modulate learners’ processing abilities, but there were links between aptitude and learners’ abilities to develop different levels of awareness.

Form-focused instruction studies generally report larger gains for explicit types of instruction over implicit types on measures of controlled production. Studies that used online processing measures—which do not readily allow for the application of explicit knowledge—however, suggest that this advantage occurs primarily when the target structure is similar in the first language (L1) and the second language (L2). This study investigated how explicit knowledge of a structure that does not exist in the L1 affects the initial stage of adult L2 acquisition. Fifty-one Dutch L1 speakers received a short auditory exposure (instruction) to a new language that included differential object marking (DOM), in which animate but not inanimate direct objects are preceded by a preposition. For 26 learners, the instruction was complemented by a brief rule explanation. Afterward, learners’ online processing and explicit knowledge of DOM were measured by means of eye-tracking (visual world paradigm) and oral grammaticality judgments. Results show that metalinguistic information promoted learners’ performance on the grammaticality judgment task. Although differences between the groups were also found on the eye-tracking measure, learners were not able to use DOM to predict the following object.