``Mind over matter'' is a frequently used phrase, but is there any evidence suggesting that mind can exert some influence over the behaviour of physical, material systems? Two meta-analyses dealing with such effects will be reviewed, both of which suggest that mind can directly interact with matter. Both these databases involve participants attempting to make a random system behave in a non-random manner.
The first of these databases involves studies in which people tried to influence the outcome of falling dice. This work was initially suggested by claims of gamblers that they were able to influence the outcome in dice throwing situations in gaming casinos. Radin and Ferrari  conducted a meta-analysis of 148 dice studies conducted between 1935 and 1987. This database also included 31 control studies in which no conscious influence of outcome was attempted. The results showed a significant overall effect for the experimental influence studies ( es = .012, Stouffer , and chance results for the control studies (Stouffer z = 0.18). To obtain a homogeneous distribution of effect sizes, 53 studies (35 per cent) of the database had to be deleted. Of these deleted studies, 33 had positive and 19 had negative effect sizes. Eleven study quality measures were considered. While the relationship was not significant, the authors did find that effect size decreased as study quality increased.
Another methodological problem affecting this database is that the probability of obtaining a specific outcome is not necessarily equally distributed across all the die faces (e.g., if using pipped dice, the six typically has the least mass and is thus most likely to come up). To examine the possible influence of this ``non-random'' aspect of dice throwing, the results for a subset of 69 studies, in which targets were balanced equally across the six die faces, were examined. A significant overall effect was still obtained (Stouffer z = 7.617, p ). For these 69 studies, the effect size was relatively constant across the different measures of study quality, and a file drawer analysis revealed that a 20:1 ratio of unreported, nonsignificant studies for each reported study would be required to reduce the database to chance expectations.
The second ``mind over matter'' meta-analysis involves studies in which a person attempts to influence a microelectronic random number generator (RNG) to behave in a non-random manner. This meta-analysis, conducted by Radin and Nelson , involves the largest parapsychological database to date, with 832 series, of which 597 were experimental series and 235 control series. The general protocol of these studies involves having a RNG drive a visual display, which an observer tries to influence, by means of mental intention, in accordance with prespecified instructions. The randomness of the RNG is usually provided by radioactive decay, electronic noise or pseudorandom number sequence seeded with true random sources; the RNG's are frequently monitored to ensure true random output in these studies. The observer initiates a ``trial'' by means of a button push, which starts the collection of a fixed length sequence of data. For each data sequence, a z score may then be computed. The mean effect size per trial for the experimental series was very small, but very robust ( es = .0003, combined z = 15.58, p = and significantly higher ( z = 4.1, p = 0.00004) than the effect size for the control series ( es = --.00004). Sixteen study quality measures were investigated; effect size did not significantly co-vary with study quality. The file drawer estimate for this data base is enormous, requiring 54,000 null, unreported studies to reduce the observed effect to chance levels. Given these findings, Radin and Nelson concluded that ``it is difficult to avoid the conclusion that under certain circumstances, consciousness interacts with random physical systems'' (p. 1512, ).