The study, therefore, requires replication of sites of each type of sediment to
estimate natural variation that is not associated with type of sediment. It is some-
times argued that such a study as described is replicated because there were several
cores in each patch. This is an example of what Hurlbert (1984) called ‘pseu-
doreplication’ – the replicate units in each sample are at the wrong scale and are
estimating variability at tens of metres, not at the hundreds of metres at which
there are differences among patches of the same type. A better-constructed design
would solve the problem by sampling several patches of coarse sediment spaced,
say, 2 km apart and several patches of finer sediment at similar spacing, about 2 km
from the nearest area of coarse sediment sampled. The two types of patch should
be interspersed, i.e. chosen to be higgledy-piggledy on a map, so that no systematic
trend or gradient makes them different for reasons other than the type of sediment
(see Underwood (2000) for examples and illustrations of this issue).