This paper documents nominal wage stickiness using an original quarterly firm-level dataset. We use the ACEMO survey, which reports the base wage for up to 12 employee categories in French firms over the period 1998 to 2005, and obtain the following main results. First, the quarterly frequency of wage change is around 35 percent. Second, there is some downward rigidity in the base wage. Third, wage changes are mainly synchronized within firms but to a large extent staggered across firms. Fourth, standard Calvo or Taylor schemes fail to match micro wage adjustment patterns, but fixed duration "Taylor-like" wage contracts are observed for a minority of firms. Based on a two-thresholds sample selection model, we perform an econometric analysis of wage changes. Our results suggest that the timing of wage adjustments is not state-dependent, and are consistent with existence of predetermined of wage changes. They also suggest that both backward- and forward-looking behavior is relevant in wage setting.