An empirical analysis is conducted on lane changing behavior using a trajectory data set collected at a weaving section in Grenoble (France). A detailed literature review shows that strong empirical understanding of the weaving mechanisms is still lacking. The goal is to investigate the lane changing behavior at a microscopic level. Data collection distinguished lane changes from the main road and those towards the main road. Descriptive empirical analysis examines the positions of the lane changes and the corresponding accepted gaps. Results show that under heavy congested traffic conditions the weaving vehicles tend to change lane as soon as possible after the start of the weaving section. When the traffic conditions are fluid, the weaving vehicles coming from the main road tend to change lane earlier than the weaving vehicles coming from the auxiliary lane. Moreover, weaving vehicles coming from the auxiliary lane accept smaller gaps and headways than the weaving vehicles coming from the main road. Our findings are questioning some results of previous works using micro-simulation to study weaving sections. As the weaving vehicles change lane at the beginning of the studied weaving section, our findings ask whether the length of the weaving section is a key variable to estimate its capacity. Our findings raise also some questions about the relevance of the gap acceptance theory to model lane changes at weaving sections. But further research are needed to asses these hypothesis.