Carbon dioxide production by heterotrophic respiration in soils is a major component of the carbon budget of terrestrial ecosystems. Despite recent development in micrometeorological techniques, nearly all soil-surface CO2 flux (FCO2) measurements are currently made using chamber techniques. The objective of this study was to compare flow-through (FT) and non flow-through (NFT) non steady-state chambers for the measurements of FCO2 in a loamy and a clay soils cropped corn or barley. For each comparison, deployment time was 3 min for the FT chambers but varied from 18 to 60 min for the NFT chambers. On the loamy soil, slopes of the “NFT vs FT” relationships indicated that FCO2 estimated using NFT chambers were 9 to 20% greater than when using FT chambers and that the coefficient of determinations decreased with NFT deployment time (r2 = 0.88 for 18 min, 0.79 for 24 min and 0.49 for 60 min). Compared to the loamy soil, relationships between the two chamber types on the clay soil had poorer fit (0.28< r2< 0.57) and FCO2 estimates were smaller using NFT than FT chambers (slopes < 1). FCO2 on the loamy soil obtained using 18- and 24-min deployments of NFT chambers were similar to FT estimates but poor agreement between the two chamber types was found for all deployment durations on the clay. These results were unexpected because non steady-state chambers theoretically perform better on soils with lower macroporosity. Possible reasons for variable chamber estimates on the clay include soil disturbance during the installation of chambers and presence of shrinkage cracks. Effects of other factors, such as soil water content, will be discussed.