Objective: There is limited evidence about the efficacy of cognitive interventions for Alzheimer’s disease (AD). However, aside from the methodological quality of the studies analyzed, the methodology used in previous meta-analyses is itself a risk of bias as different types of effect sizes (ESs) were calculated and combined. This study aimed at examining the results of nonpharmacological interventions for AD with an adequate control of statistical methods and to demonstrate a different approach to meta-analysis. Method: ESs were calculated with the independent groups pre/post design. Average ESs for separate outcomes were calculated and moderator analyses were performed so as to offer an overview of the effects of bias. Results: Eighty-seven outcomes from 19 studies (n  812) were meta-analyzed. ESs were small on average for cognitive and functional outcomes after intervention. Moderator analyses showed no effect of control of bias, although ESs were different from zero only in some circumstances (e.g., memory outcomes in randomized studies). Cognitive interventions showed no more efficacy than placebo interventions, and functional ESs were consistently low across conditions. Conclusions: cognitive interventions delivered may not be effective in AD probably due to the fact that the assumptions behind the cognitive interventions might be inadequate. Future directions include a change in the type of intervention as well as the use of outcomes other than standardized tests. Additional studies with larger sample sizes and different designs are needed to increase the power of both primary studies and meta-analyses.