In this paper it is argued that an expert system requires more than factual knowledge before it can display expertise in a given domain. The additional knowledge consists of the heuristics or 'rules of thumb' used by an expert to manipulate and interpret the factual knowledge. The knowledge acquisition phase of an expert system project involves determining the factual knowledge (which may be obtained from published sources) and the heuristics used by an expert to manipulate that knowledge--these heuristics can only be obtained from an expert. In reviewing existing biological expert systems it is apparent that many contain only the factual knowledge relating to the domain, and lack the heuristics that enable such systems to show expertise. This paper reviews a number of knowledge acquisition techniques which could be used for acquiring heuristic knowledge and discusses when their use is appropriate. The knowledge acquisition techniques discussed are those suitable for the development of small-scale expert systems as these are most likely to be of interest to biologists. The techniques include the use of questionnaires, interview techniques and protocol analysis; particular emphasis is placed on a modification to the 'twenty questions' interview technique which was developed specifically to elicit taxonomic knowledge relating to water mite identification.

The role of expert systems in species identification, with particular reference to the problems posed by damaged specimens and inexperienced taxonomists, is discussed. Of the three main types of expert systems available, the frame-based system is shown to provide the most appropriate model for a taxonomic expert system rather than a logic- or rule-based system. The advantages of an expert system over other computer-aided methods of identification are considered. A rule-based system requires the original knowledge (species descriptions) to be structured into rules, whereas a frame-based system can store the generic and specific descriptions in a series of frames. The frames fall into a hierarchy which closely resembles the taxonomic hierarchy, and down which information can be inherited. Two aspects of frame-based systems considered are the use of probabilities in identification, and the optimum structure of the knowledge base. The conventional use of probabilities is to provide an indication of the correctness of the result. However, in some studies involving the identification of many specimens, the speed of identification may be increased (with a reduction in accuracy) if identifications are made to a predetermined probability level. Although frames allow accurate representation of the taxonomic hierarchy, a semantic net, incorporating structures of the organism and/or details of the habitat may result in a more efficient expert system.