The current study investigated the impact of recasts together with form‐focused instruction (FFI) on the development of second language speech perception and production of English /ɹ/ by Japanese learners. Forty‐five learners were randomly assigned to three groups—FFI recasts, FFI only, and Control—and exposed to four hours of communicatively oriented lessons. Whereas many FFI activities including explicit instruction were embedded into the treatment in order for the experimental groups to notice and practice /ɹ/ in a meaningful discourse, an instructor provided recasts only to the FFI‐recast group in response to their mispronunciation of /ɹ/. Perception was measured using a two‐alternative forced choice identification task, while pronunciation performance was elicited using controlled and spontaneous production tests and assessed by 10 naïve native‐speaking listeners. According to the statistical comparisons, whereas the FFI‐only group attained perception and production improvement particularly under trained lexical conditions, the FFI‐recast group demonstrated similar but generalizable gains both in trained and untrained lexical contexts. The results indicate that (a) FFI itself impacts various domains of L2 speech learning processes (perception, controlled, and spontaneous production) and (b) recasts promote learners’ attentional shift away from lexical units as a whole to phonetic aspects of second language speech (i.e., vocabulary to sound learning).