The REF (Reorganization of Elementary Functions) model suggests mechanisms of posttraumatic reorganization, and resolves the contradiction between localization and functional recovery. In the process of developing this model, we have reconceptualised the term ‘function’ and introduced a concept of strictly localized, basic processing modules, which we term ‘elementary functions (EFs)’. Uniting the EFs into functional networks gives rise to ‘algorithmic strategies (ASs)’, which constitutes the information flow and processing mediating a particular behavioural surface phenomenon traditionally defined as ‘function’. After brain injury, the EFs of the affected structure are irreversibly lost. However, during neurorehabilitation, the remaining EFs are reorganized into novel ASs. Creation and utilization of these novel ASs constitutes the basis for functional recovery. The degree to which recovery appears possible is a reflection of how successful the new ASs network is in achieving the surface phenomenon under consideration. This model has several implications relative to clinical neurorehabilitation. One of these concerns the degree to which results of rehabilitation training can be expected to generalize across situations and cognitive domains. Since novel ASs are created and selected according to situation dependent feedback mechanisms, dissimilar external conditions may call for different ASs. Consequently, therapeutically established ASs may not be adequate in the settings of daily life. Additionally, upon the loss of a particular brain structure, the structure and networks mediating recovery will not be identical in case of different cognitive and situational demands. Therapeutically, an important implication is that rehabilitative training should include as naturalistic settings as possible.