Deviations from simple lognormality of size frequency distributions (SFD) are being increasingly interpreted in terms of depositional processes. In doing so, it is tacitly assumed that the sedimentary volume sampled is homogeneous with respect to SFD (i.e. if the sample volume were subdivided and analyzed, truncation points of cumulative SFD would not vary appreciably). However, sediment is commonly laminated and each lamina might carry size frequency data unlike that of the bulk sample. For this reason the variability of SFD on a microlevel are of interest. Accordingly, soft sediment peels were obtained from several environments and five to eight adjacent laminae in each were microsampled vertically. Size was determined in a manner suitable to produce size data comparable to sieve data. Significant variation exists in mean size and shape of the SFD at microlevels. Most intralamina frequency data are coarse-tail truncated and unimodal, although a few coarse-tail truncated bimodal distributions were observed. Boundaries between laminae were often marked by sharp grain size increases. These observations suggest that commonly observed characteristics of SFD (truncation points, subpopulations, etc.) cannot have a simple genetic explanation inasmuch as these features vary considerably at microlevels. Cumulative frequency distributions constructed from sublamina samples, which are neither normal nor lognormal, can commonly be described by three intersecting lines on probability paper. The fact that a uniform distribution or two overlapping normal distributions will produce the same pattern makes it difficult to interpret such data in terms of traction, saltation, and suspension processes. Commonly, the same tripartite structure can be described by a lesser number of intersecting arcs. Even if the sublamina samples are considered to be a combination of lognormal distributions, the truncation points within and between laminae fluctuate in grain size enough so as to reclassify as much as twenty percent of the sample from one subpopulation to another. Size frequency distributions therefore must be considered highly composite. The large amount of information expressed at microlevels suggest this may prove to he the link between experimental studies and depositional process.