Semantic dementia is a neurocognitive disorder characterised by a steady and progressive loss of semantic knowledge in the presence of relatively preserved other cognitive skills. Recent treatment studies have proven that language rehabilitation aimed at anomia in semantic dementia can be successful. The objective of this study was to examine the separate and interactive effects of errorless vs. errorful and active vs. passive learning approaches to anomia and their effects on naming and comprehension of treated items, as well as maintenance and generalisation of treatment gains. Seven participants with semantic dementia re-learned two sets of words (one for which participants retained auditory comprehension, and one for which they did not) in each of four different treatment methods based on those approaches. Errorless learning proved more successful than errorful learning in restoring lexical representations in all but one participant while there was no interaction between effects of errorless and active approaches on treatment success. Maintenance of treatment gains showed an advantage for errorless learning at one but not three months post-treatment, although all overall gains were maintained to a significant degree at both time points. Effects of both treatment and maintenance were stronger for items for which participants showed preserved auditory comprehension. The results are discussed in a framework of progressive language disorders and applicability of errorless methods to language rehabilitation in semantic dementia.