We report on a program of spectroscopic observations of gravitationally lensed QSOs with multiple images. We seek to establish whether microlensing is occurring in each QSO image using only single-epoch observations. We calculate flux ratios for the cores of emission lines in image pairs to set a baseline for no microlensing. The offset of the continuum flux ratios relative to this baseline yields the microlensing magnification free from extinction, as extinction affects the continuum and the lines equally. When we find chromatic microlensing, we attempt to constrain the size of the QSO accretion disk. SDSSJ1004+4112 and HE1104-1805 show chromatic microlensing with amplitudes 0.2 < |Δm| < 0.6 and 0.2 < |Δm| < 0.4 mag, respectively. Modeling the accretion disk with a Gaussian source (Ivpropexp (- R 2/2r 2 s )) of size rs vpropλ p and using magnification maps to simulate microlensing, we find rs (λ3363) = 7 ± 3 lt-day(18.1 ± 7.8 × 1015 cm) and p = 1.1 ± 0.4 for SDSS1004+4112, and rs (λ3363) = 6 ± 2 lt-day(15.5 ± 5.2 × 1015 cm) and p = 0.7 ± 0.1 for HE1104-1805. For SDSSJ1029+2623, we find strong chromaticity of ~0.4 mag in the continuum flux ratio, which probably arises from microlensing, although not all the available data fit within this explanation. For Q0957+561, we measure B - A magnitude differences of 0.4 mag, much greater than the ~0.05 mag amplitude usually inferred from light-curve variability. It may substantially modify the current interpretations of microlensing in this system, likely favoring the hypothesis of smaller sources and/or larger microdeflectors. For HS0818+1227, our data yield possible evidence of microlensing.