CONTEXTComplex rice systems (CRS) are polycultures in which rice is grown together with one or more plant or animal species using combined methods of practices from indigenous knowledge and modern science in augmenting ecological processes to foster ecosystem services in rice agroecosystems. However, their implementation faces challenges due to farmers' knowledge gaps, high capital outlay and labour shortages. OBJECTIVEThis study aims to link types of rice farming and farmer perceptions to facilitate recommendations to scale up CRS. METHODSWe constructed a farm typology based on a survey of 111 farm households and aggregated cognitive maps (ACMs) based on fuzzy cognitive maps developed in focus group discussions in Malang and Lamongan, East Java Province, Indonesia. RESULTS AND CONCLUSIONSThe farm typology classified farm households into three types: (a) small farms with high inputs of agrochemicals (SH, n = 29) which were all identified in Malang; (b) medium-size farms with high input intensity of agrochemicals (MH, n = 43), distributed across Malang and Lamongan; and (c) medium-size farms with low inputs of agrochemicals (ML, n = 39), all detected in Lamongan. ACMs revealed the differences in farmer group's perceptions on farm input-output relations, agroecosystems processes and CRS component interactions. SH and MH farmers prioritised economic benefits over ecosystem services and crop-livestock components. SH farmers were locked into high-input and high-output systems, while MH farmers failed to obtain high output despite high input use. Intervention schemes could be developed such as thorough training on economics for SH farmers and efficient use and on-farm production of fertilisers for MH farmers. Meanwhile, with the majority of ML farmers having adopted CRS, their priorities were more evenly distributed across all components, implying their more holistic understanding of the interacted components. However, technical assistance on agronomic, post-harvest management and marketing were still considered to be needed to ensure a long-term sustainability of CRS practices by ML farmers. SIGNIFICANCECombining farm typology and fuzzy cognitive mapping could facilitate a comprehensive understanding of farming system types, their characteristics and farmer perceptions as a key to formulate appropriate interventions for a specific farm type. Therefore, these methodological approaches could provide guidance to accelerate transitions toward complex agroecosystems.