In a recent meta-analysis, Bowles and Torres (forthcoming) found that instruction has a medium to large positive effect on heritage bilingual children’s learning outcomes in the heritage language. However, this is based on only two studies (Wright, Taylor & McCarthur, 2000; Cuza, Miller, Pasquarella & Chen, 2017), and it remains unknown how instruction alters heritage bilingual children’s morphosyntactic development, especially for vulnerable structures in adult heritage bilingual acquisition (e.g., Silva-Corvalán, 1994; Benmamoun, Polinsky & Montrul, 2013; Montrul, 2016; Polinsky & Scontras, 2019). To address this issue, the current study adopts a complexity theory perspective to language development (e.g., Larsen-Freeman & Cameron, 2008; Larsen-Freeman, 2017), which endorses a usage-based perspective in which language is viewed as a dynamic and adaptive system. Based on complexity theory, then, a critical question that this study addresses is to what extent does the role of instruction contribute to heritage bilingual children’s adaptation of their linguistic resources. In this longitudinal study, we tested 108 heritage bilingual children of Spanish who were in either 2nd or 4th grade at Time 1 of the study, and then tested again a year later for Time 2. Each child completed the Bilingual English Spanish Assessment Middle Extension or BESAME test (Peña, Bedore, Gutiérrez-Clellen, Iglesias & Goldstein, n.d.) at both Time 1 and Time 2. This study reports on the Spanish morphosyntax subset of the BESAME, which consisted of the completion of cloze set phrases and sentence repetition tests that included a number of morphosyntactic structures in Spanish (e.g., noun-adjective agreement, aspect, subjunctive among others). Through the Bilingual Input-Output Survey or BIOS (Peña, Gutiérrez-Clellen, Iglesias, Goldstein & Bedore, 2014), teachers reported on the number of hours a week that students received Spanish input or instruction as well as their language production in Spanish and English. The main results were the following: (1) Based on the scores of the cloze set phrases and sentence repetition tests, a cluster analysis yielded three groups of learners (n = 90), who differed primarily on their grade and weekly hours of instruction in Spanish; (2) In comparison to the control group (n = 18), instruction overall had a positive effect on morphosyntactic development, but the magnitude of such effects varied across the three groups of learners and time of testing; (3) a more detailed analysis of four vulnerable morphosyntactic structures in heritage language acquisition revealed that the groups of learners mostly maintained or progressed in their Spanish morphosyntactic development, whereas the control group demonstrated more patterns of attrition of these structures. The findings will be discussed in light of Bowles and Torres’ (forthcoming) meta-analysis and the theoretical underpinnings of complexity theory.