<p>Research into the effects of contextual diversity on lexical processing has flourished in the past 20 years, encompassing different tasks, populations, and languages, and informing influential theories of word learning. This review provides a comprehensive synthesis of the field. Eighty-six articles (145 experiments) composed of three distinct study types (behavioural [<i>N</i> = 111], computational modelling [<i>N</i> = 20], and corpus validations [<i>N</i> = 14]) met preregistered inclusion criteria. Across experiments, the terminology used for different diversity metrics has been inconsistently applied. We classify all metrics into four categories (count-based, computational, composite, unspecified) to standardise comparisons. Four key findings emerge from this review: Experiments that assessed the impact of diversity on <i>word-form</i> processing (<i>N</i> = 85) show a consistent high-diversity advantage, possibly because high-diversity words are more likely to be ‘needed’ in the future. Effects of diversity on <i>word-meaning</i> processing (<i>N</i> = 13) were more mixed, showing both low- and high-diversity benefits. We attribute these inconsistencies to varying task demands. Specifically, we conclude that selecting highly precise semantic information can be challenging for words that occur in variable contexts. Computational modelling studies indicate that diversity metrics that quantify the distinctiveness of contexts in which words occur better predict behaviour than simple context counts. Corpus validations show that diversity effects are consistent across languages. This review confirms that diversity in linguistic experience is a key organizational principle of the lexicon but indicates that current theories lack specificity when describing the underlying mechanisms. We make specific recommendations for future research within a structured research cycle.</p>