How does attention optimize our visual system for the task at hand? Two mechanisms have been proposed for how attention improves signal processing: gain and tuning. To distinguish between these two mechanisms we use the equivalent-noise paradigm, which measures performance as a function of external noise. In the present study we explored how spatial and feature-based attention affect performance by assessing their threshold-vs-noise (TvN) curves with regard to the signature behavioral effects of gain and tuning. Furthermore, we link our psychophysical results to neurophysiology by implementing a simple, biologically-plausible model to show that attention affects the gain and tuning of population responses differentially, depending on the type of attention being deployed: Whereas spatial attention operates by boosting the gain of the population response, feature-based attention operates by both boosting the gain and sharpening the tuning of the population response.