The inherently variable nature of language presents a particular challenge for the study of language development and language maintenance. Years of research have shown that linguistic variation is systematic and probabilistic. However, this view of variation in research on language development and heritage language maintenance is rare. Some approaches to heritage language development in particular overlook variation and attribute inter-individual variability to incomplete acquisition processes and reduced input. Alternatively, this dissertation considers linguistic variation to be a central component in our understanding of the processes and outcomes of language acquisition. Presented in this work are three studies of the variable properties of differential object marking (DOM) in different populations of Spanish speakers. The first study contributes to research on language variation in language development by comparing patterns of DOM in monolingual Mexican children and their caregivers. Corpus-based analyses reveal important similarities in the linguistic factors that constrain DOM in child and caregiver speech. Moreover, children younger than three years of age demonstrate knowledge of only one constraint--animacy--on DOM, while the children between three and five years old show knowledge of both constraints examined--i.e., animacy and specificity. The distribution of variable contexts in the caregivers' speech may explain why children acquire the animacy constraint before specificity. The remaining experimental studies focus on heritage speakers of Spanish--specifically, adult Spanish-English bilinguals who grew up in an English-dominant community in the United States. In production, heritage speakers demonstrate some variable patterns of DOM comparable to those of Spanish-dominant bilinguals. However, usage-based patterns seem to affect DOM usage in the heritage speakers differently from the comparison group. Importantly, these usage-based patterns have roots in the historical development of DOM in Spanish and do not indicate any direct influence from English. Finally, an exploratory study of heritage speakers' comprehension of DOM in spoken language reveals that their sensitivity to the presence or absence of DOM depends on direct object animacy. Morphologically related perceptual factors are also found to influence this sensitivity. Ultimately, this dissertation supports a nuanced view of the role of variation across the lifespan. This perspective yields a more comprehensive approach to the study of morphosyntax in monolingual acquisition and heritage language maintenance.