Knowing what to learn in a cluttered environment is fundamental to all aspects of development. But how do infants know what to learn when faced with uncertain input? We know that both social and non-social cues can shift infants’ attention. How do these attention cues mediate learning during infancy? I will present three eye-tracking studies demonstrating that infants’ ability to learn about structures in their environment (i.e., predicting the appearance of audiovisual events and forming expectations about co-occurring features) is dependent on the presence and nature of attention cues. By 8 months of age, infants learn these events better with social cues (e.g., eye gaze, infant-directed speech, expressions of interest) than with non-social cues (e.g., flashing lights) or without any attentional cueing. Importantly, when presented with multiple events to learn and cued by a face to one specific event, infants learned the cued event rather than the non-cued event. The third study trained infants to learn from unfamiliar attention cues (i.e., salient flashing lights). I will also present a neuro-computational model showing that attentional filter size can determine whether learning from attention cues occurs in this paradigm, suggesting that social cues can optimize this filter. This research provides compelling evidence for how attention cues shape infants’ learning in the typical cluttered environment, informing investigations on optimal strategies for learning to learn among distractions.