Children on the autism spectrum often demonstrate atypical joint attention, leading some researchers to consider joint attention deficits a core feature of the autism spectrum. Structured measures, such as the Early Social Communication Scales (ESCS), are commonly used to provide a metric of joint attention. To explore the assessment of joint attention in multiple contexts, we implemented an alternative system for coding joint attention behaviors. We compared initiation of joint attention (IJA) and response to joint attention (RJA) behaviors coded from naturalistic examiner-child play samples with similar IJA and RJA behaviors elicited within the structured ESCS protocol. Participants were 20 toddlers on the autism spectrum. Levels of IJA and RJA within the two assessment contexts were significantly and positively correlated, providing support for the use of naturalistic sampling of joint attention skills as a viable alternative, or supplement, to structured measures. (Contains 6 tables.)