Better understanding of the role of learning in technical progress is important for the development of innovation theory and technology policy. This paper presents a comparative analysis of the effect of learning and technical change in electricity generation technologies. We use simultaneous two-factor learning and diffusion models to estimate the effect of learning by doing and learning by research on technical progress for a range of technologies in four stages of development. We find learning patters broadly in line with the perceived view of technical progress. The results generally show higher learning by research than learning by doing rates. Moreover, we do not find any development stage where learning by doing is stronger than learning by research. We show that simple learning by doing curves overstate the effect of learning in particular for newer technologies. Finally, we find little substitution potential between learning by doing and research for most technologies.