Representative Samples and the Quality of the Data
How good is the data?
Dream research is often accused of working from biased or otherwise inadequate samples: Maybe the people who remember dreams are different from other people? Perhaps the dreams that people remember are not an accurate sample of their actual dream life? Maybe people don't report their dreams honestly? Aren't the sample sizes usually too small?
In the face of all this, what makes us think we have representative people giving us representative samples of dream reports in numbers that are large enough to give us confidence that we have solid findings?
As far as the representativeness of the people, the many studies of high and low dream recallers can provide us with an answer. There are only small differences -- if any -- between high and low dream recallers on a variety of personality and cognitive tests. It therefore follows that those who contribute dream reports are a representative sample of people for every personality and cognitive test devised so far by psychologists. Furthermore, there is evidence that high recallers differ primarily in their interest in dreams and their motivation to recall them, which are factors restricted to the issue of dreams, and therefore not a challenge to the general representativeness of the people who report dreams. Put another way, "interest" and "motivation" concerning dreams do not correlate with personality and cognitive variables.
Turning now to the representativeness of the dream reports that these representative people give us, there are several steps to the argument, most of which are based on dream reports collected from awakenings throughout the night in sleep laboratories.
We know from studies using dreams collected in the laboratory that they do not differ very much, if at all, from the reports of dreams at home from the same subjects.
We know from studies in the sleep lab that dream reports from early in the night do not differ much, if at all, from dreams collected later in the sleep period, which is essential to the next point.
We have reason to believe that the dreams we recall upon awakening in the morning are most likely to be the ones we just dreamed, but this poses no problem because of the previous finding, that the last dream reports of a sleep period to not differ from those earlier in the night.
We know from studies of reports collected from awakenings throughout the night in sleep labs that "recency" and "length" are the main factors in determining which of the earlier reported dreams are recalled once again in the morning. The "dramatic intensity" of the report is a third factor, but a study in which subjects immediately telephoned a report when they recalled a dream during the day showed that minor cues in our environment can trigger the recall of very mundane (i.e., those of low "dramatic intensity") dreams, thereby balancing the influence of dramatic intensity on morning recall. We also know that dramatic intensity is not a big influence on dream recall because our normative studies show that a majority of dream reports do not contain the aggression, sexuality, or bizarreness that supposedly make dreams "dramatic." When all is said and done, then, the laboratory studies, the telephone study, and our norms give us confidence that recency, length, and everyday cues are leading us to a representative sample of dream life.
Why do we believe that subjects are giving us full and honest reports? There are three answers. First, almost all of our subjects provide dream reports anonymously, reducing any tendency to misreport. Second, people express a lack of personal responsibility for their dreams, which leaves them quite willing to report whatever they experience. For most people, dreams are something that happen to them, an experience, so they do not see them as a reflection on their self-image. Third, the consistency of our findings from sample to sample suggests that people are reporting honestly; otherwise, they are all misreporting in the same ways, which would be quite surprising.
As far as sample sizes, they have been too small in the past, but by drawing many subsamples of varying sizes from large samples where the overall results are known, we now know that it takes 75 to 100 reports to sample anindividual's dream life adequately and 100 to 125 for a sound comparison of a group's dream reports to our norms.
For individuals, we used dream series where we had findings with anywhere from 150 to nearly 1000 dream reports, then compared the overall findings for each series with the results for subsamples of 25, 50, 75, and 100 dream reports, and found we could match the overall findings within a few percentage points when the subsample reached 75 and 100.
For groups, we first compared the overall findings in Hall and Van De Castle's original normative sample of 500 male dream reports from 100 individuals with the results for subsamples of 25, 50, 75, 100, and 125, and 250, and found a similarity to the overall norms within a few percentage points with subsamples of 100 and 125. Then we compared 100 Most Recent Dreams from 100 women college students at the University of California, Santa Cruz, in the early 1990s with Hall and Van De Castle's normative findings on 500 dream reports from 100 women, and found that the overall results in our percentages and indexes were almost exactly the same for both samples.
Click here to see a table with the actual data from our subset analysis of the Male Norms.
Move on to learn about our statistical approach.
Go back to the Content Analysis page.