Although word learning unfolds over days, weeks, and months, individual naming events are over in a matter of seconds. To benefit from a naming event, children must at least hear the label and see the referent. We tested 1-, 2- , 3-, and 4-year old children in a naturalistic word learning task with two conditions: one that taxed both speech processing and rapid gaze-following, and one in which a social cue-to-reference was available for an extended time. The development of word-learning in the extended condition paralleled the development of speech processing, but learning in the brief condition lagged behind. However, learning from both the brief and extended cues was predicted by individual differences in speech processing and cue-following together. Thus, even through the 4th year, real-time processing of social and linguistic information are a critical bottleneck for word learning.