Cellulose-binding domains (CBDs) are discrete protein modules found in a large number of carbohydrolases and a few nonhydrolytic proteins. To date, almost 200 sequences can be classified in 13 different families with distinctly different properties. CBDs vary in size from 4 to 20 kDa and occur at different positions within the polypeptides; N-terminal, C-terminal and internal. They have a moderately high and specific affinity for insoluble or soluble cellulosics with dissociation constants in the low micromolar range. Some CBDs bind irreversibly to cellulose and can be used for applications involving immobilization, others bind reversibly and are more useful for separations and purifications. Dependent on the CBD used, desorption from the matrix can be promoted under various different conditions including denaturants (urea, high pH), water, or specific competitive ligands (e.g. cellobiose). Family I and IV CBDs bind reversibly to cellulose in contrast to family II and III CBDs which are in general, irreversibly bound. The binding of family II CBDs (CBD(Cex)) to crystalline cellulose is characterized by a large favourable increase in entropy indicating that dehydration of the sorbent and the protein are the major driving forces for binding. In contrast, binding of family IV CBDs (CBD(N1)) to amorphous or soluble cellulosics is driven by a favourable change in enthalpy which is partially offset by an unfavourable entropy change. Hydrogen bond formation and van der Waals interactions are the main driving forces for binding. CBDs with affinity for crystalline cellulose are useful tags for classical column affinity chromatography. The affinity of CBD(N1) for soluble cellulosics makes it suitable for use in large-scale aqueous two-phase affinity partitioning systems.