This study investigates the effectiveness of four types of grammar instruction and the extent to which students' learning style affects the learning outcomes of these instruction types. Our focus is on a complex grammar structure, viz. English conditionals. A total of fourteen Dutch classes with senior secondary school students aged 15-17 and their ten teachers participated in the study. Teachers and their classes were randomly distributed among implicit, incidental, inductive and deductive treatment groups and a control group. A pretest-posttest design, including a grammaticality judgement test and a semi-free writing test, was used to study the effectiveness of the treatment groups for students with a learning style focused on either learning from active experimentation or from reflective observation. Results of a multilevel covariance analysis indicate that explicit-inductive instruction effectively raises students' performance concerning complex grammatical sentences and it does so more effectively than incidental instruction but no more than other forms of grammar instruction. Post hoc comparisons reveal that these outcomes hold for all students, irrespective of their learning style.