ABSTRACT. Since the pioneering attempts to obtain quantitative soft-bottom sediment samples at the beginning of this century, many different devices for sampling soft bottoms have been presented. Most techniques, however, yield semi-quantitative samples only. For studies dealing with certain aspects of microlayering, freezing in situ seems to be a good method. In more general studies, quantitative samples appear to be best obtained using instruments based on the coring principle. Improper handling or faulty design of the conng device, however, results in unrepresentative samples. Major factors in b~a s are (1) loss of surficial sediments. ( 2 ) redistribution, resuspension and loss of enclosed sediment, ( 3 ) core shortening, and ( 4 ) repenetratlon. Frictional drag and consequent smearing against the coring tube is also a potential source of error General design features improving the reliability of sampling are a supporting stand and, when dealing with open corers, a n unimpeded water flow through the sample compartment. Box corers appear best suited for sampling certain macrofauna and sediments of coherent and silty/sandy character. Adequately designed and properly operated tube corers are the most versatile instruments at present. A general concern, however, is that coring tubes enter and penetrate the sediment carefully and slowly. Under favourable conditions, a n open-barrel gravity corer can b e used for sampling the unshortened upper layer, in combination with a larger box corer and subsequent piston-coring technique for sampling the underlying layers. When sampling under less favourable conditions, or when extreme care is required, the corer must be based on a structure with a supporting stand, which permits the device to settle on the bottom before the coring tubes enter the sediment. Mounting an underwater video camera as a standard on such devices is recommended since lt enables visual inspection and recording of each sampling. Other tested means and promising improvements in instrument design to improve sampling reliability are also discussed.