Some recent studies suggest that errorless learning principles may be beneficial in memory rehabilitation for people with dementia, while others indicate that effortful processing may be more important. The present study compared the effects of four different learning techniques, varying in level of effort required and number of errors elicited, on free recall, cued recall and recognition of novel and previously known associations among people with early-stage dementia. Ten participants with a diagnosis of early-stage Alzheimer's, vascular or mixed dementia learned novel and previously familiar face–name associations with each of four techniques – vanishing cues, forward cues, target selection, and paired associate learning – in a within-subjects design. All conditions produced significant learning for both novel and familiar associations. There were no significant differences between conditions, although mean scores were slightly higher for errorful conditions. Reducing errors did not produce any benefits. Enhancing level of effort had no significant effects for familiar associations, but high-effort conditions were significantly more effective than low-effort conditions in facilitating cued recall of novel associations. The results confirm that memory rehabilitation techniques can produce significant benefits, but do not support the view that error reduction during learning facilitates greater improvement in early-stage dementia. Effort enhancement may be more important, especially when learning novel associations. We are indebted to a number of people for their assistance. Sue Okell and Julia Brownlie provided valuable support in setting up the research, and Margaret Mooney was instrumental in helping to recruit participants. Professor Robert Logie and Dr Hazel Emslie provided methodological assistance, and Dr Rik Henson generously provided access to stimulus items. Dr Robert Jones offered helpful comments on an earlier draft of this paper. Finally, we would like to thank the staff of the Day Hospital for supporting the research, and most importantly the participants themselves for their enthusiastic and valued contribution. The work reported here formed part of Josephine Dunn's thesis for the doctorate in clinical psychology at University College London, awarded in 2003, which was supervised by Linda Clare.