BACKGROUND: Personalised cueing is a training method designed to facilitate naming of unknown, realistic visual stimuli (dog breed names). Creation of a personalised cue is similar to the use of mnemonic devices by normal individuals to remember important bits of information. Theoretical support for the method comes from Craik and Lockhart's depth-of-processing model of memory (1972). Several studies have shown that training with personalised cueing methods results in significantly higher levels of long-term naming accuracy than when subjects are trained with phonological cueing techniques. However, it has also been observed that all individuals are not equally proficient in creating personalised cues and that the nature of the information in personalised cues varies markedly from individual to individual. AIMS: The objective of this study was to determine if the type of information contained in a personalised cue (cue form) affects the degree to which these cues facilitate learning of subordinate category names (dogs). METHODS AND PROCEDURES: 600 personalised cues developed by 15 non-brain-damaged (NBD) and 15 aphasic individuals to learn the names of unknown dog breeds (e.g., Maltese) were examined. The cues were classified as one of five cue forms by three judges on two separate occasions approximately 1 month apart. Examination of intra-judge agreements for the cue forms yielded a total of 251 cues for analysis of cue form effects (127 aphasic; 124 non-brain-damaged). OUTCOMES AND RESULTS: To examine the effects of cue form on facilitation of naming, weighted recall scores were calculated for each cue based on accurate naming on probes one week, one month, and 6 months after training. Kruskal-Wallis analysis of variance by ranks (KWANOVA) was used to determine the effects of cue form on learning of the subordinate category names. Significant cue forms effects were found for aphasic but not non-brain-damaged participant cues. Mann-Whitney post-hoc comparisons of aphasic cues revealed that cues containing semantic information had significantly higher mean rankings than those containing phonological information and those containing a combination of phonological and semantic information. CONCLUSIONS: Two implications arise from the study with respect to the use of personalised cueing as a clinical procedure. One is that it may be necessary to exert some limited control over the creative process of developing a personalised cue to ensure the inclusion of semantic information in the cue itself. A second is that individuals who have problems accessing semantic information may require some training before attempting to create personalised cues. In such instances, it would be important to obtain information about the source of the individual's anomic deficits through careful testing before embarking on a training programme featuring personalised cueing.