Tillage systems can influence weed seed viability and the distribution with depth of weed seeds in soil. Weed seed bank composition was determined at two soil depths (0 to 10 and 10 to 20 cm) in three tillage systems [mouldboard plough (MP), shallow tillage (ST), and direct drilling (DD)] established for 14 years on a sandy loam (Podzol) in Atlantic Canada, under a soybean (Glycine max L. Merr.) - barley (Hordeum vulgare L.) rotation. The size and composition of the viable soil seed bank were evaluated using the seedling germination method. The diversity of weed species was slightly lower for MP (17 species) compared to the ST (21 species) and DD treatments (22 species). For the total soil depth sampled (0 to 20 cm), weed seed population was significantly greater under DD (56 weeds m^-2) and ST (66 weeds m^-2), compared to MP (25 weeds m^-2). Comparison of the 0 to 10 with the 10 to 20 cm soil depth showed a relatively uniform weed seed distribution for the MP treatment, while a greater proportion of weed seeds was found at the lower soil depth for DD and ST. This distribution tended to be weed species dependent. Soil texture and weed seed characteristics were considered to have a critical impact on the total weed seed bank size, specifically for the 10 to 20 cm soil depth. Overall, the weed bank size was relatively small indicating that adoption of conservation tillage practices for sandy loams in Atlantic Canada should not cause a major change in weed community and weed populations, or present a need for significant changes in weed control management.