The first part of the paper reviews and criticises drift samplers. The samplers are divided into three broad categories:

1. Samplers without flow-meters (Fig. 1).

2. Samplers with flow-meters (Fig. 2).

3. Tube samplers (Fig. 3) and other more complex samplers.

The second part of the paper considers some problems associated with drift sampling.

As the volume of water flowing through a drift sampler increases, the size of the catch increases. The average relationship between the two variables can be expressed by a regression line (e.g. Fig. 5). A distinction is made between drift rate (the number of invertebrates passing a sampling point in unit time) and drift density (the number of invertebrates per unit volume of water). The relationship between drift rate and drift density is discussed. A summary is given of all the factors which affect drift density.

The sampling efficiency of drift samplers is discussed, and the results of efficiency tests are described. As the performance and sampling efficiency of a drift sampler vary with local stream conditions, they must be thoroughly tested before the start of regular drift sampling.

The study of the diel periodicity in invertebrate drift is discussed. The diel drift pattern greatly varies when different sampling periods are used (Fig. 6).

The following statistical problems are discussed:

1. Temporal and spatial variations in drift rate and drift density, using Poisson series as a statistical model.

2. The estimation of total numbers drifting downstream. Total drift is estimated from drift rate when agreement with a Poisson series is accepted for samples across the stream. Total drift is estimated from drift density when agreement with a Poisson series is rejected, and the catches in a series of nets are proportional to the volumes of water sampled by the nets.