I'm guessing his concern is visually looking at a histogram and gauging its normality vs doing so with a q-q plot, which will likely show deviations from normality much better.