Results for optimizing an array of conducting polymer gas sensors for sensing one of five analytes in the presence of up to four interferents are presented. The optimized array consists of subarrays of homogeneous (like) sensors contributing to a larger heterogeneous array of up to ten points (unlike sensors) in multidimensional sensor space. The optimization techniques presented here are linear, since the polymer sensors in their useful (low concentration) operating range exhibit linear and additive response characteristics. The optimization of these arrays produces maximum separability between analytes, demonstrating the trade-off between the addition of both information and variability induced by increasing the size of the heterogeneous array. Optimization results for sensing acetone, hexane, THF, toluene, and ethanol in the presence of interferents result in array sizes that are significantly less than the maximum available number of sensors (ten in the heterogeneous partition of the array). This result adds fuel to the argument that fewer sensors are better; the argument for more sensors, however, is also made in the context of the electronic nose systems where significant chemical diversity is required. Homogeneous subarrays of up to four elements each improve the separability of analytes in these optimized heterogeneous arrays by over 10% and also effectively flag broken or unhealthy sensors in a manner that is independent of analyte and concentration.