The relationship among accuracy, interpretability, and complexity of genetic fuzzy systems (GFSs) is a hot topic and is actively studied in the GFS domain. Because different problems have different views of interpretation, it is quite difficult to evaluate the interpretability of GFSs in general. The present study aims to analyze accuracy-complexity relationship in fish habitat modelling using a genetic Takagi-Sugeno fuzzy model called fuzzy habitat preference model (FHPM). The model complexity was defined by bit lengths of a genetic algorithm (GA) assigned to the consequent part of the model, while fuzzy rules and antecedent parts were kept the same. FHPM was developed on the basis of the mean squared errors between the composite habitat preference and the observed presence-absence of fish. The model accuracy was evaluated using multiple performance measures. As a result, the different model complexities resulted in slightly different habitat preference curves and model accuracies. At some complexities, the model accuracy was found to be slightly improved with increased model complexity. The result suggests that an optimal point exists where the model complexity can take a balance between the accuracy and the complexity of the target models, which depends partly on data characteristics and model formulations of the GFSs.