Analytical techniques for determining intermodulation distortion in directly modulated semiconductor lasers are examined. Direct expressions for the two- and three-frequency third-order intermodulation distortion-to-carrier ratios are used to check the validity of a well-known 6-dB difference formula. It is found that this formula is not always valid, the discrepancy being a function of laser diode parameters and subcarrier spacing. The carrier-to-intermodulation-noise ratios are computed for a 25-channel subcarrier multiplexed system, with subcarrier spacings of 6 MHz for amplitude modulation and 40 MHz for frequency modulation, and it is found that the central subcarrier is not always the worst affected, although it has the maximum number of interfering intermodulation products. This is in contrast to memoryless nonlinearities where the central subcarrier may be expected to have the worst carrier-to-intermodulation-noise ratio.