This paper studies "duration dependence" -- the probability that a longer unemployment spell reduces a worker's job opportunities -- by sending fi…ctitious resumes to real job postings across the U.S. We randomly manipulate the length of the current unemployment spell on each …fictitious resume, and we track callbacks for interviews for all submissions. Preliminary results from the ongoing …field experiment suggest that callbacks are strongly decreasing in unemployment duration. We develop a theoretical framework and show that it is possible to discern among leading models of duration dependence, based on employer screening, human capital depreciation and ranking, by exploring how duration dependence varies with labor market tightness. We …find suggestive evidence that duration dependence tends to be stronger in labor markets with lower unemployment rates, suggesting that employer screening plays a relatively more important role in generating duration dependence.