The paper discusses the importance of features and exponents in analyzing bilingual and multilingual grammars, emphasizing the need for a clear distinction between underlying syntactic features and their morphophonological realizations (exponents). It argues that an exoskeletal approach to grammar, which separates features from their exponents, offers a powerful framework for analyzing data from bilingual and multilingual speakers. This approach subsumes the Feature Reassembly mechanism, which was developed by Donna Lardiere. The paper presents three case studies demonstrating how the exoskeletal approach can be applied without additional constraints or mechanisms. The paper begins by addressing two foundational questions in bilingual grammar research: whether bilingual grammars can be modeled systematically and whether architectural changes are necessary compared to monolingual grammars. It argues that bilingual grammars are just as systematic and constrained as monolingual ones, and that no architectural changes are needed. Instead, a theoretical framework that distinguishes syntax from morphology, where morphology occurs after syntax, is sufficient to handle grammatical structures in bilingual populations. The paper then discusses the theoretical framework of exoskeletal models of grammar, which emphasize the separation of syn-sem features from their exponents. It highlights the core aspects of Feature Reassembly and exoskeletal models, and argues that exoskeletal models provide all the necessary tools for Feature Reassembly. The paper presents three case studies illustrating how exoskeletal models can account for phenomena where Feature Reassembly has typically been used. These case studies include the L2 acquisition of number marking in Korean and Indonesian, the L2 acquisition of Chinese imperfective markers, and definiteness in Norwegian heritage language. The paper concludes that exoskeletal models can capture the necessary mechanisms without the need for a separate theory like Feature Reassembly.The paper discusses the importance of features and exponents in analyzing bilingual and multilingual grammars, emphasizing the need for a clear distinction between underlying syntactic features and their morphophonological realizations (exponents). It argues that an exoskeletal approach to grammar, which separates features from their exponents, offers a powerful framework for analyzing data from bilingual and multilingual speakers. This approach subsumes the Feature Reassembly mechanism, which was developed by Donna Lardiere. The paper presents three case studies demonstrating how the exoskeletal approach can be applied without additional constraints or mechanisms. The paper begins by addressing two foundational questions in bilingual grammar research: whether bilingual grammars can be modeled systematically and whether architectural changes are necessary compared to monolingual grammars. It argues that bilingual grammars are just as systematic and constrained as monolingual ones, and that no architectural changes are needed. Instead, a theoretical framework that distinguishes syntax from morphology, where morphology occurs after syntax, is sufficient to handle grammatical structures in bilingual populations. The paper then discusses the theoretical framework of exoskeletal models of grammar, which emphasize the separation of syn-sem features from their exponents. It highlights the core aspects of Feature Reassembly and exoskeletal models, and argues that exoskeletal models provide all the necessary tools for Feature Reassembly. The paper presents three case studies illustrating how exoskeletal models can account for phenomena where Feature Reassembly has typically been used. These case studies include the L2 acquisition of number marking in Korean and Indonesian, the L2 acquisition of Chinese imperfective markers, and definiteness in Norwegian heritage language. The paper concludes that exoskeletal models can capture the necessary mechanisms without the need for a separate theory like Feature Reassembly.