MicroRNAs (miRNAs) regulate gene expression by inhibiting translation of target mRNAs through pairing with miRNA recognition elements (MREs), usually in 3′ UTRs. Because pairing is imperfect, identification of bona fide mRNA targets presents a challenge. Most target recognition algorithms strongly emphasize pairing between nucleotides 2-8 of the miRNA (the 'seed' sequence) and the mRNA but adjacent sequences and the local context of the 3′ UTR also affect targeting. Here, we show that dispatched 2 is a target of miR-214. In zebrafish, dispatched 2 is expressed in the telencephalon and ventral hindbrain and is essential for normal zebrafish development. Regulation of dispatched 2 by miR-214 is via pairing with three, noncanonical, weak MREs. By comparing the repression capacity of GFP reporters containing different dispatched 2 sequences, we found that a combination of weak sites, which lack canonical seed pairing, can effectively target an mRNA for silencing. This finding underscores the challenge that prediction algorithms face and emphasizes the need to experimentally validate predicted MREs.