A computational model of a multispectral imaging system was constructed and used to simulate the recovery of the spectral reflectance of surfaces imaged by the system. The model allows parameters such as number of sensors, sensor spectral sensitivity and quantization noise to be evaluated in terms of their effect on the accuracy of recovery. A set of 1269 Munsell surface reflectance factors were used to test the model. The recovery process employs a linear system whereby spectral reflectance functions are represented by a small number of basis functions. The results show that increasing the number of sensors in the system or increasing the number of basis functions in the linear model does not necessarily increase recovery performance. However, in general, error does monotonically decrease with increasing sensor number when the number of basis functions used in the linear model is allowed to vary independently of sensor number. These performance aspects of the system are closely correlated with the condition number of the solution matrix.