Random, unpredictable, unstructured stimuli are an everyday part of life. Yet despite this breadth of experience and sophisticated statistical learning mechanisms, humans report patterning even in stimuli that are paradigmatically random. In two experiments, participants evaluated structured and random environments presented in a common statistical learning paradigm, the Serial Reaction Time task. I presented random and specifically nonrandom sequences to humans (Experiments 1 and 2) and rhesus macaques (Macaca mulatta, Experiment 2) to explore the seemingly antagonistic relationship between explicit, intuitive beliefs about these sequences and implicit statistical learning of sequence properties. Sequence predictability and experience with a given sequence type significantly predicted reaction times only weakly and inconsistently across the two experiments. Accordingly, participant choices scarcely deviated from chance, and in those rare cases they deviated from chance largely without directionality, and were not significantly predicted by either sequence predictability or experience with a given sequence type.