Purpose Frequent sounds and frequent words are both acquired at an earlier age and are produced by children more accurately. Recent research suggests that frequency is not always a facilitative concept, however. Interactions between input frequency in perception and practice frequency in production may limit or inhibit growth. In this study, we consider how a range of input frequencies affect production accuracy and referent identification. Method Thirty-three typically developing 3- and 4-year-olds participated in a novel word-learning task. In the initial test block, participants heard nonwords 1, 3, 6, or 10 times—produced either by a single talker or by multiple talkers—and then produced them immediately. In a posttest, participants heard all nonwords just once and then produced them. Referent identification was probed in between the test and posttest. Results Production accuracy was most clearly facilitated by an input frequency of 3 during the test block. Input frequency interacted with production practice, and the facilitative effect of input frequency did not carry over to the posttest. Talker variability did not affect accuracy, regardless of input frequency. The referent identification results did not favor talker variability or a particular input frequency value, but participants were able to learn the words at better than chance levels. Conclusions The results confirm that the input can be facilitative, but input frequency and production practice interact in ways that limit input-based learning, and more input is not always better. Future research on this interaction may allow clinicians to optimize various types of frequency commonly used during therapy.