Abstract This paper provides a reference to estimate the representation of large-sized items (seeds and fruits, mainly) in samples of larger and smaller volume in wetland sites with the aim of proposing a minimum sample size to recover these remains in a representative way. For this, almost 100 samples from a late Neolithic settlement phase found at the lakeshore site of Parkhaus Opera (Zurich, Switzerland) were subsampled into one larger subsample (A-sample, of ca. 3 l of volume) and one smaller subsample (B-sample, of ca. 0.3 l of volume). We compared how large and small-sized items were represented in the different fractions of large and small subsamples on the basis of ubiquity, concentration and proportions between the taxa. Large-sized remains (like Prunus spinosa or charred fragments of Corylus avellana) and some medium-sized remains (Najas, Aethusa cynapium) were more often represented in larger subsamples and therefore are considered to be underrepresented in smaller samples. Average concentration values were similar in both groups of samples (and therefore comparable) but large differences were observed on a one-to-one sample basis, finding no positive monotonic correlation between them. Our observations also prove that in order to obtain data that are comparable to dryland sites concerning charred remains (including cereals and large-seeded wild fruits), large volume samples of at least ca. 3 l are needed. Counting units per taxon in each fraction were re-defined on the basis of the results obtained. Finally, some clues to interpret results concerning large-sized items in sites with samples of small volume are also proposed following our observations.