We consider the effects of convection on the Balmer line profiles (H
α and H
β) of A, F, and G stars. The standard mixing-length theory (MLT) atlas9 models of Kurucz (1993, CD_ROM 13), with and without overshooting, are compared to atlas9 models based on the turbulent convection theory proposed by Canuto & Mazzitelli (
1991ApJ...370..295C,
1992ApJ...389..724C) and implemented by Kupka (1996, ASP Conf. Proc. 108, 73), and the improved version of this model proposed by Canuto et al. (
1996ApJ...473..550C) also implemented by Kupka. The Balmer line profiles are a useful tool in investigating convection because they are very sensitive to the parameters of convection used in the stellar atmosphere codes. The H
α and H
β lines are formed at different depths in the atmosphere. The H
α line is formed just above the convection zone. The H
β line, however, is partially formed inside the convection zone. We have calculated the T
eff of observed stars by fitting Balmer line profiles to synthetic spectra and compared this to: (i) the T
eff of the fundamental stars; (ii) the T
eff of stars determined by the Infra-Red Flux Method and (iii) the T
eff determined by Geneva photometry for the stars in the Hyades cluster. We find that the results from the H
α and H
β lines are different, as expected, due to the differing levels of formation. The tests are inconclusive between three of the four models; MLT with no overshooting, CM and CGM models, which all give results in reasonable agreement with fundamental values. The results indicate that for the MLT theory with no overshooting it is necessary to set the mixing length parameter α equal to 0.5 for stars with T
eff≤6000K or T
eff≥7000K. However for stars with 6000K≤T
eff≤7000K the required value for the parameter is α≥1.25. Models with overshooting are found to be clearly discrepant, consistent with the results with uvby photometry by Smalley & Kupka (
1997A&A...328..349S).