While there is growing evidence that pronunciation teaching can work, there is a need to establish what it is that makes it work. The study reported here tested for the effect of two particular factors: socially constructed metalanguage (SCM) and critical listening (CL). SCM is a term proposed for metalanguage developed by students working together with the teacher using already understood first language (L1) concepts to help in the formation of target language phonological concepts. CL is based on listening and contrasting to learn phonological categories and their boundaries. The study involved four groups of six high-intermediate level adult students. Each group received 45-50 minutes of instruction on pronouncing syllable codas. The instruction was carefully scripted for each group to create four conditions; SCM+/CL+, SCM+/CL-, SCM-/CL+, and SCM-/CL-. The results showed significant immediate effects for SCM on speech production and for CL on speech perception. The SCM+/CL+ lesson was replicated for four members of the SCM-/CL- group, who then made similar gains to those achieved by the first group to receive SCM+/CL+ instruction. These results add to the theoretical debate on what makes pronunciation instruction effective and have practical implications for the classroom. (Contains 4 figures and 6 tables.)