We use Caenorhabditis elegans to test proposed general rules for microRNA (miRNA)-target interactions. We show that G.U base pairing is tolerated in the 'seed' region of the lsy-6 miRNA interaction with its in vivo target cog-1, and that 6- to 8-base-pair perfect seed pairing is not a generally reliable predictor for an interaction of lsy-6 with a 3' untranslated region (UTR). Rather, lsy-6 can functionally interact with its target site only in specific 3' UTR contexts. Our findings illustrate the difficulty of establishing generalizable rules of miRNA-target interactions.