This paper studies to what extent job characteristics such as physical and cognitive demands, use of technologies, responsibility, difficulty, stress, peer pressure, and relations with co-workers are related to full or partial retirement. We study employment transitions and retirement expectations of older workers by exploiting the wealth of information about individuals older than age 50 in the Health and Retirement Study (HRS), and characteristics of different occupations provided by the Occupation Information Network (O*NET) database. Controlling for basic demographics, wages, benefits, health, cognitive ability, personality, and other personal characteristics, we find strong and statistically significant relationships between labor force transitions and job characteristics. These relationships are typically more pronounced and more precisely estimated when we use objective job attributes taken from the O*NET than when we use self-reported job characteristics taken from the HRS, but self-reported characteristics are more strongly related to moves from full-time to part-time employment. Using expected retirement age or subjective probabilities of working full-time at older ages gives similar results to using actual labor force transitions as the dependent variable. The estimated effects of job characteristics are again stronger and more robust to alternative specifications when measures of job attributes are taken from the O*NET than from the HRS. Our findings suggest that nonmonetary job characteristics are important determinants of labor supply decisions at older ages, but our analysis is still preliminary in its attempt to uncover causal relationships: Unobservable individual characteristics responsible for sorting into specific occupations may also shape retirement decisions.