This paper analyzes the relationship between wages and productivity during the early years of an employment relationship. Data from the Employment Opportunity Pilot Project show that worker productivity grows substantially during the first two years on the job, with most of the growth in productivity occurring at the very start of the job. Correcting for measurement error and the fact that expected productivity beyond the start of the job may be folded into the starting wage if wage revisions are not instantaneous, one finds that variation in productivity is only partially reflected in wages. Not only is productivity growth stemming from human capital accumulation while on the job only partially reflected in wage growth, but starting productivity differences for workers in the same job — in large part driven by differences in relevant experience - are only partially reflected in starting wage differences. Our empirical findings can be explained by a simple model of employer — worker cost sharing in which (a) the cost to a worker of locating and moving to a new job increases with the worker's stock of human capital and (b) equity norms prevent employers from paying senior workers lower wages than junior workers who are no more productive.