We have examined the noise sources for a surface plasmon resonance (SPR) sensor system to facilitate optimization of SPR sensor instrumentation and data-processing methods for high-resolution SPR sensing. We found detector shot noise to be the dominant source of noise in the SPR sensor, and investigated the propagation of noise through a commonly used SPR sensor data-processing algorithm. Deriving an expression relating the detector noise level to the noise of the output data, we used measurements of the magnitude and distribution of both charge-coupled device and photodiode array detector noise to predict the noise levels observed in the sensor output. We used this method to optimize the parameters of the algorithm and introduced a modified algorithm with enhanced resistance to correlated light level noise. In addition, we evaluated the noise performance of alternative SPR sensor designs, including a sensor with no polarization optics and another with optics producing an inverted SPR absorption feature. Experimental SPR sensor data demonstrated good agreement with predicted noise levels.