The knowledge needed for development of new products isbecoming increasingly complex and distributed. There is aneed for firms to develop their capabilities for knowledgeintegration. Although prior knowledge-based literaturehas pointed to the importance of knowledge integrationfor competitive advantage, our understanding of how theintegration of different types of knowledge affects newproduct development (NPD) performance is still limited.We quantitatively examine the effect of the integrationof different types of knowledge on performance in NPD.Knowledge integration refers to the process of combiningspecialized but complementary knowledge. In this article weseek to study the effects of knowledge integration on NPDperformance. Previous studies point to the positive effectsof knowledge integration on NPD performance, but theydo not distinguish between different types of knowledgein examining these effects. This article therefore seeks tocontribute to this emerging literature by explicitly studyingthe integration of different types of knowledge and the effectsthat such integration have on NPD performance. We drawupon a classification of knowledge suggested by Ullman (1997)in discussing what types of knowledge that is particularlypertinent in engineering practices. That is, what knowledgeengineers draw upon in conducting design and developmentwork. We address the types of domain-specific knowledge,procedural knowledge, and general knowledge integration.Three hypotheses suggesting that the capabilities for integrationof each type of knowledge respectively affect NPD performancepositively are tested. A fourth hypothesis suggests that there arecomplementarity effects between integration of the three typesof knowledge upon NPD performance. Data was collectedduring 2009 from a sampling frame of 355 medium-sizedSwedish manufacturing firms in four industries. We received193 valid answers, i.e. a 54% response rate) Hypotheses weretested with standard OLS regression together with EFA andCFA analysis. The results provide support for the first threehypotheses, while the fourth one was rejected. This impliesthat capabilities to integrate domain-specific, proceduraland general knowledge are all independently affecting NPDperformance positively, but no combined effect above andbeyond the individual variables. This indicates that oneknowledge type can be integrated without an absolute need tointegrate two types or all three types.