Much ado about very little: The small effect sizes when home and laboratory collected dreams are compared.

Abstract

We report a reanalysis of the original codings of dream reports collected at home and in the sleep laboratory from the same participants studied by Hall and Van de Castle in 1964. We used Cohen'sh statistic for effect sizes to argue that, even when statistically significant, most of the differences between the two samples are small to medium in magnitude. This finding suggests that past arguments over the relative usefulness of the two types of samples might not have occurred if the magnitude of effect sizes had been taken into account. The one exception concerns aggressions of various kinds, which also show the greatest variability with age, gender, and culture. It is concluded that useful dream samples for studies using the Hall and Van de Castle coding system can be collected in the laboratory or from normal recall at home and that effect sizes should be calculated in all dream content studies.