Spatial modulation (SM) is a low complexity, highly spectral efficient multiple-input -multiple-output scheme that has been proposed in the literature. The authors apply adaptive modulation to conventional SM in order to maximise the average throughput of the scheme. For this to be possible, the equivalent received signal-to-noise ratio (SNR) needs to be defined and is done so via two proposed approaches: using the first-order statistics SNR and using the average statistics SNR. Average theoretical bit-error-rate (BER) bounds are derived for both of these SNR approaches. Also, adaptive M-ary quadrature amplitude spatial modulation switching levels are determined to maximise the throughput while meeting the average target BER. The Monte Carlo simulation results successfully validate the derived theoretical BER bounds and also prove that the average throughput is improved in comparison to conventional SM. The two proposed definitions for the equivalent received SNR are confirmed to yield comparable BER and throughput performances via simulations, implying that the equivalent received SNR can be defined using either approach.