While confounding factors typically jeopardize the possibility to use observational data to measure peer effects, field experiments offer the possibility to obtain clean evidence. In this paper we measure the output of four randomly selected groups of individuals who were asked to fill letters in envelopes, with a remuneration completely independent of output. For two of these groups the output of peers was exogenously manipulated (low or high) by making individuals aware of the number of letters previously produced by artificial colleagues. In the third group individuals were set up to work one in front of the other, while the fourth group gave the baseline output for independent not manipulated work. Our first finding is that effort of the less productive workers reacts in a sizeable and statistically significant way to peer pressure. Second, there is strong evidence of peer effects when individuals work in pairs. Third, these peer effects work in the direction of making the least productive individuals work harder, thereby increasing overall productivity.