This study investigates whether two different means of learning the word sequence "despite the fact (that)" lead to differential outcomes of the acquisition of the grammatical structure imbedded in the word string which is, due to L1 transfer, especially problematic for Chinese learners of English. The participants of the study were randomly divided into two groups. While one group was taught the grammatical structure under investigation through explicit instruction, the other group was required to commit the word string to memorization through repetition or recitation without being given any explicit instruction. It turns out, following the results of subsequent tests, that the recitation group significantly outperformed the instruction group in terms of procedural knowledge whereas the instruction group outperformed the recitation group on declarative knowledge. The inference that can be drawn from the study, among other things, is that the memorization of the whole word string may facilitate the acquisition of the imbedded structure, especially on the procedural level, whereas explicit instruction can only exert visible influence on the awareness of knowledge. One explanation for the result is that "despite the fact (that)" was memorized as an unanalysed whole, i.e. chunk. It is suggested that formal criteria for grouping phrases be introduced in language teaching programmes. (Contains 4 tables.)