Finding Meaning in Dreams: Chapter 4

The findings to be presented in this chapter are based on studies of American college students over a 45-year period. While such findings may at first glance seem limited because most students come from a narrow age range and a small portion of the socioeconomic ladder, this is in fact the greatest strength of the findings because they are not attenuated by other variables. They provide an anchor point for investigations of the relationship of dream content to such factors as age, ethnicity, race, class, nationality, and individual differences.

Moreover, college students are an ideal group to study if the primary goal is to understand the meaning of dreams because they are introspective, articulate, and often interested in understanding their own dreams. They are at an open and inquiring age, and many are willing to cooperate wholeheartedly in attempts to study their dreams. Although many dream theorists seem to believe that we will learn more about dream meaning from studying the dream reports of children, psychiatric patients, and preliterate people living in small traditional societies, in our experience the new and most provocative findings on dream content come from studying the dream reports of college students or the dream diaries of highly motivated post-college adults.

Hall and Van de Castle (1966: chap. 14) carried out their first study using the new coding system on 500 dream reports from 100 European-American female students and a similar number of reports from 100 European-American male students, all of whom were between 18 and 25. Five reports from each student were randomly selected from dream series containing between 12 and 18 dreams. Dream reports less than 50 words or more than 300 words were eliminated before the selections were made. The findings from this study provide the "norms" for the Hall/Van de Castle system for home dream reports for college students.

Two issues of critical importance have to be dealt with in analyzing findings of the type presented in this chapter. First, dream reports from different groups or individuals may differ in length; since longer dream reports might be expected to have more of everything in them, some kind of correction must be made for this possibility. In Hall and Van de Castle's normative sample, for example, women's dream reports were on the average 8% longer than those from men. It therefore can be expected that the tables presented in this chapter generally will show higher frequencies of dream elements for women than men. In our experience this is a fairly typical gender difference. However, a reversal of the pattern on dream report length in a study by Kramer, Kinney, and Scharf (1983) criticized at the end of this chapter has led to a great amount of confusion in the literature on gender differences in dream content. The second problem in analyzing findings with the Hall/Van de Castle coding system is that the frequencies of social interactions are affected by the frequencies with which various types of characters appear as well as by dream length. For example, since there are more male than female characters in the dream reports of men, there is a greater possibility for social interactions with male characters.

We use several different strategies to deal with these two related problems. Sometimes, as in the case of the normative study to be presented in this chapter, we do not use dream reports under 50 words in length. In one or two instances we have divided the frequency of an element by the number of words or lines in a report. However, we do not think these approaches are the best possible solution, a conclusion based in part on a finding by Trinder et al. (1970) that the relationship between dream report length and frequency of elements is not a linear one for all content categories.

Instead, we apply a variety of solutions to the length and frequency problems that allow us to bypass any correction for the length of the dream report. Basically, we create various ratios, rates, and proportions that are called indexes and percentages in the terminology of this book. For example, when subclasses of such general categories as settings, characters, activities or objects are being compared in different samples, the effect of report length can be eliminated by converting each frequency of a subclass into a percentage of the total frequency of the category. Thus, we determine what percentage of the total human characters in a sample are male or female, what percentage of the combined friendly and aggressive interactions are aggressive ones, what percentage of the total number of aggressions are physical ones, and what percentage of the total number of activities are physical ones (cf. Hall, 1969a:153-154).

Similarly, the comparison of social interactions between the dreamer and other characters can be standardized by creating an index controlling for the different number of characters in different samples. For example, we simply divide the number of aggressive interactions by the number of characters, creating an "aggressive-interactions-per-character" rate, so to speak; we call this the "A/C index." Such indexes can be created for any character subclass to learn if a dreamer has a higher aggressive-interaction rate (A/C index) with men or women, children or adults, relatives or strangers. In short, our "unit of analysis" for aggressive interactions, and all other social interactions, is characters (cf. Hall, 1969a:154).

We do not propose to present all of our specific percentages and indexes at this point. Instead, they will be introduced at the appropriate place in the discussion of various findings, where the context will make them more immediately understandable. The important conclusion for now is one we stated at the beginning of this discussion of percentages and indexes: all of these strategies allow us to bypass correcting for dream length, so we do not have to talk in terms of "mean frequencies" per dream report or per dream-report sample. Whether this approach is the better course to take can only be determined by seeing if the results it provides make psychological sense and inspire new investigators to do further studies using it. However, it is based on long experience.

The issue of how to deal with inevitable differences in report lengths and variations in raw frequencies also brings us to the question of what form our statistical analyses should take. As we explain in more detail in Appendix D, we have chosen statistics that are based in percentages and percentage differences for a variety of reasons. The most important of those reasons, however, is that these statistics allow us to analyze the particular kind of data we had to create in order to correct for dream length and differential frequencies for some kinds of elements.

Rather than discussing issues of statistics in the main text of this book, we ask readers interested in the rationale for our statistical approach to turn to Appendix D before reading the rest of this chapter. Here we will only say the main test of statistical significance we use is called the "significance of differences between two independent proportions" and that it yields the same results as better-known alternatives such as correlation and chi square with our particular kind of data. Beyond that, we need only add here that statistical significance is a minor issue for us, in good part because the large size of our samples insures that even small differences are found to be statistically significant. The main issue for us, therefore, is the magnitude or size of differences, which we measure with Cohen's (1977) h statistic, explained in Appendix D.

We first turn to a presentation of the results of Hall and Van de Castle's normative study of European-American college men and women. Once the findings of this study are discussed, the results of subsequent investigations of American college students by several different investigators will be presented.

The chapter includes three recent innovations with the potential to make dream content studies easier and more interesting. First, we will show that it takes only 100 dream reports from the normative sample of 500 dreams to obtain the same findings. Second, we will argue that one "Most Recent Dream" from 100 students makes it possible to replicate the normative findings. Taken together, these two points mean that comparisons of subgroups with each other and our norms can be done much more efficiently than in the past, and with confidence in any differences discovered. Third, we will present the h-profile as an effective way to display the ways in which a subgroup or individual differs from our norms, making it possible to see possible patterns in the findings with great ease.

To avoid inundating readers with dozens of frequencies and percentages in the main text, only some of Hall and Van de Castle's major findings for each content category are presented in this chapter. The complete findings are displayed in highly detailed tables in Appendix E. Putting the detailed tables together in one place in an appendix makes it easier for readers to use them for research purposes. To make this appendix even more useful, the normative expectations for all our major categories are brought together in Table E.1.

The Hall/Van de Castle Normative Study

Characters

There are several main findings in the realm of characters. We will use these findings to introduce readers to how we use percentages to correct for dream length. We also will show how we use comparative findings to begin to make sense out of our data, but the full implications of most of our findings will become apparent only in the context of the findings presented in later chapters. We begin with the following simple findings, as displayed in Table 4.1.

As the table reveals, there are two to three characters aside from the dreamer in the average dream report. The exact figure is 2.4 for males and 2.8 for females, but we need to recall that women's dream reports are 8% longer than those of men; this finding does not correct for the difference in length. We also need to remember that groups of any size count as only one character. In point of fact, the mean number of characters per dream is not a finding we use very often. It is merely a rough benchmark, as is the fact that less than 5% of dream reports are without any characters aside from the dreamer.

We next see in Table 4.1 that most of the characters in the dream reports of these college students are other human beings. This may not be surprising to most readers, but it is actually a useful finding, experience has taught us, when we use it as part of an equation we call the "animal percent." According to our norms, 6% of the characters in the dream reports of men are animals; the figure is 4% for women dreamers. These findings have no particular relevance in and of themselves, but they start to take on meaning when they are compared with findings for children (where the animal percent is much higher), or dreamers in small hunting and gathering societies (where it is also much higher), or dreamers in Japan (where it is below 1%).

The "finding" on the percentage of animal characters allows us to return to the issue of how we deal with differences in dream length. Here we have used the "total characters" as our unit of analysis. In future studies, whatever the length of the average dream report, and whatever the total number of characters in the dream reports, we can speak of the person or group's "animal percent" in a directly comparative way because it is always a function of "total animals" divided by "total characters."

We next note in Table 4.1 that there is a gender difference in how often men and women include male and female characters in their dreams: men dream twice as often about other men as they do about women (67% vs. 33%), and women dream equally about both sexes (48% men, 52% women). If we put the numbers in the kind of 2 X 2 table discussed in Appendix D, and then perform our statistical operations, we find this difference of 19 percentage points has an h value of .39 (defined as a medium-sized effect size in Appendix D) and is statistically significant well beyond the .01 level of significance (see Table 4.2). However, as Table 4.2 also shows, the confidence interval for this finding ranges from .30 to .48.

We call this finding the "ubiquitous" gender difference in dreams (Hall and Domhoff, 1963a; Hall, 1984), and we label it the "male/female percent." Men have a male/female percent of 67/33 and women have a male/female percent of 48/52. This difference is not found in each and every study, but it is found in most of them. It is a totally unexpected outcome of the coding system. It is a "discovery," and it turns out to have its counterpart in published short stories written by male and female authors (Hall and Domhoff, 1963a) and in stories made up by pre-school children (as reported in the next chapter). There is no obvious or certain "interpretation" of this finding, but it is the type of finding that may lead eventually to a theory of dream meaning, as will be shown in subsequent chapters.

Returning to Table 4.1 for a final time, we see there is also a gender difference in the percentage of "familiar" and "unfamiliar" characters which basically means people we know, such as extended family and friends, or do not know. This information can be used to create a "familiarity percent," determined by dividing the number of familiar characters by the total number of familiar and unfamiliar characters. A comparison of male and female dreamers on familiarity percent is presented in Table 4.3; this same comparison is also used in Appendix D in explaining a 2 X 2 table.

Once again, the magnitude of the difference is in the medium range, the level of significance is well beyond the .01 level, and the confidence interval for the effect size is fairly wide.

If we ask ourselves why there is a gender difference on familiarity percent, we may be given a start toward the answer in the findings on unfamiliar males and females presented in the more detailed table in Appendix E on characters (Table E.2). There we learn that men dream more often of unfamiliar males than do women (28% vs. 15%), but men and women dream equally about "familiar males" (25% vs. 23%). Conversely, women are dreaming more often of familiar females, 29% vs. 16% for the males, but males and females dream about equally of unfamiliar females, 10% for males and 11% for females. In short, these comparisons tell us it is "unfamiliar males" in male dreams and "familiar females" in female dreams that create the difference on the familiarity percent.

We do not want to give the impression from the findings presented so far that our primary interest in the norms is in the gender differences. For our purposes, gender differences are most important as a starting point for the study of individual differences. For example, if we were to study a male dream series that had an animal percent of 18 instead of the usual 6, a male percent of 90 instead of the usual 67, and a familiarity percent of 25 instead of 45, we might start to suspect we were studying a fairly unusual person. In an individual study, the only importance of the gender differences is that we know we must use different norms for uncovering individual differences in males and females.

We now turn to the highlights of our findings on the various types (subclasses) of human characters categorized in the Hall/Van de Castle system. Both genders dream most frequently about known characters. Women are a little more likely to dream about family members and known characters, and men a little more likely to dream about strangers, but we already knew most of this from the findings on familiar and unfamiliar characters. The complete findings on types of human characters can be found in Table E.3.

For the most part, Table E.3 (subclasses of characters) is not used as frequently as Table E.2. However, it is the basis for our normative expectations on one of our possible psychopathology indicators, the "friends percent." Recall from the discussion in chapter 2 that for this indicator we divide the number of known characters by the total number of human characters. The normative percentage is 37 for women and 31 for men.

The findings on characters are useful in and of themselves, but they become of even greater interest in indexes involving aggressive, friendly, and sexual interactions. Thus, we often begin an investigation of a dream series or dream set with an analysis of the characters.

Social Interactions

Social interactions are usually the most revealing aspects of a dream series or dream set. The first consistent finding is that aggressions outnumber friendly interactions for both genders. Sexual interactions are a very distant third, contrary to many popular stereotypes of dream content. Forty-seven percent of men's dreams and 44% of women's contain at least one aggressive interaction. For friendliness the percentages are 38 for men and 42 for women, and for sexuality 12 and 4.

Because aggression is the most frequent social interaction, we begin with a number of findings on aggressions. We first note that men have more aggressive interactions in their dreams than women, a point we will come back to in more detail when we discuss the ways in which we correct for differences in dream length and number of characters in analyzing aggressions and friendliness. In terms of our eight subclasses of aggression, men tend to have more physical aggressions and women far more interactions in the "rejection" category. The net result is that there is a gender difference in the physical aggression percent, calculated by dividing the number of physical aggressions by the total number of aggressions. Males have a physical aggression percent of 50, whereas for females the percentage is 34 (h = .33). Note that the relationship being expressed here, between physical and nonphysical aggression, is entirely independent of dream report length, number of dream characters, or number of aggressions. It is a comparison we often make, and it is one of our hypothesized indicators of possible psychopathology. The detailed findings for the types of aggressions in dream reports can be found in Table E.4.

Turning to the interaction dimension of aggressions, summarized in Table E.5, there are two main findings. First, dreamers are involved in about 80% of the aggressions in their dreams. Percentages that are much higher or lower are of potential interest to us, especially when they are compared with the percentages for involvement in other social interactions. Second, we find that dreamers are more often victims than they are aggressors, as best shown in the victimization percent, which is independent of dream length, number of characters, or number or type of aggressions. If we use only the categories for aggressor and victim in Table E.5, ignoring witnessed, reciprocated, mutual and self aggressions, we can figure the normative Victimization Percent for men and women as follows:

If the victimization percent is above 50, the dreamer more likely to be a victim. If it is below 50, the dreamer is more likely to be an aggressor. This percentage can be determined for every major class of character with whom the dreamer interacts. From this we can learn that the dreamer may be an aggressor in relation to some people and a victim in relation to others. For example, everyone is likely to be a victim of animals and male strangers, a point we will demonstrate after we look at the main findings on friendliness.

Friendliness

Women have more friendliness in their dreams than do men if we look at the raw frequencies alone. The subclass for "helping and protecting" has the highest frequency for both genders, followed by friendly remarks and compliments, which we call "verbal friendliness." Findings on the frequencies and distribution of friendly interactions by seven subclasses are presented in Table E.6.

There are three main findings in terms of the interactions within which acts of friendliness occur. First, dreamers are involved in almost all of the friendly interactions in their dream reports. Second, there is little or no mutual, reciprocal, or self friendliness. Third, we expect the typical college dreamer, whether male or female, to initiate friendly interactions about as often as they are received. This figure, the befriender percent, is 50% for males and 47% for females. The befriender percent does not include "reciprocal" and "mutual" friendly interactions. When looked at in conjunction with the victimization percent, it gives us some idea of the dreamer's assertiveness in her or his dreams. The complete results on friendly interactions are presented in Table E.7.

Aggression and Friendliness

The findings on aggressive and friendly interactions take on even more interest when they are compared with each other. We do this in two different ways that correct for the fact there tend to be more characters in female dream reports. The first way is to divide the total number of dreamer-involved aggressions by the total number of dreamer-involved aggressions plus friendliness. This yields the "aggression/friendliness percent" mentioned in chapter 2 as a possible indicator of psychopathology in dreams. A percentage over 50 indicates a greater amount of aggression; a percentage less than 50 indicates a preponderance of friendliness. The aggression/friendliness percent is over 50 for both men and women; for men the figure is 59, for women it is 51. Men have more aggressive interactions with males, and more friendly interactions with females, whereas women have about equal amounts of aggression and friendliness with both male and female characters. Women are very friendly with familiar males, but have a preponderance of aggressive interactions with unfamiliar males. If we define "enemies" as characters with whom the aggression/friendliness percent is over 60, then male strangers and animals are "enemies" in the dreams of both men and women. These and other findings are displayed in Table E.8 [not available on-line yet].

Our other strategy for comparing aggressions and friendliness while controlling for the number of characters is to divide the number of aggressions or the number of friendly interactions by the total number of characters or the total number of characters in a given subclass. For example, if we divide the total number of aggressions (402) in male dream reports in our normative sample by the total number of characters (1180), we have an A/C index of 34. If we do the same for female dreamers, 337/1423, the A/C index is 24. Thus, male dreamers are found to be even more aggressive in their dream reports than raw frequencies suggest if we control for characters. If we compute the same figures for friendliness, the male F/C index is 21, the female index 22. These indexes are also useful when they are computed for specific classes of characters, especially males and females.

As we noted briefly in chapter 2 when talking about possible indicators of psychopathology, we have found it useful to look at people's patterns of friendliness and aggression with males and females by means of a 2 x 2 table that we call the "A/F square." This table appears as part of Table E.1. It tells a very similar story for males to what we found with the aggression/friendliness percent: men are more aggressive with other men and more friendly with women in their dream reports. However, from the A/F square we learn that women have both more aggression and more friendliness with male characters than they do with female characters. We think that atypical A/F squares might be of great use in studying individuals and groups. There have been a few revealing studies with males that will be presented in chapter 8 but none with females. We are hopeful that the great potential in using this 2 x 2 table in studying women of varying personality and sexual orientations will be tested in the next few years.

Just to reinforce our point about animals as dangerous characters in dreams, we present the A/C and F/C indexes with animals for both men and women in Table 4.4.

Sexual Interactions

After the several interesting and useful findings on aggressions and friendliness, the few findings on sexual interaction will come as a letdown. The two meager findings emerging from Table E.9 are (1) men report more sexual dreams than women and (2) female dreamers have more sexual interactions with familiar characters, whereas male dreamers have more with unfamiliar characters. Not only are the findings meager, but they may reflect the greater likelihood of sexual activity on the part of college males in the past. We think our norms for sexuality in dreams should be used with caution, especially with women, and that not too much should be made out of them. We also note that sexual interactions in dreams are far more likely to be fraught with negative emotions than popular stereotypes suggest (Jerasitis, 1992).

Activities

The main findings on activities can be summarized very briefly. The details can be found in Table E.10. First, activities are frequent in dream reports, averaging 4.7 per dream report for males and 4.9 for females. Second, physical activity, movement, and talking account for 74% of men's activities in dreams and 71% of women's. Finally, we think it may be useful in future studies to divide the number of dreamer-involved physical activities by the total number of dreamer-involved activities to derive a "physical activities percent." The normative expectation for men is 61% and for women 56%. For the most part, we use the activities categories primarily for normative comparisons with other societies, not for individual studies unless we notice some unusual frequencies.

Successes and Failures

Successes and failures are not very frequent in the dream reports of either males or females. Most successes and failures occurred to the dreamer. There was rarely a "consequence" of a success or failure in the normative study. We use these findings to determine if the dreamer has an unusually large percentage of dreams with successes or failures in them. We also use these findings to derive our "success percent," which is the number of dreamer-involved successes divided by the number of dreamer-involved successes and failures; this percentage is 51 for men and 42 for women. The details for the findings on successes and failures can be found in Table E.11.

Misfortunes and Good Fortunes

Misfortunes occur in 36% of men's dream reports and 33% of women's. Good fortunes happen in only 6% of dream reports for both men and women. Misfortunes are followed by good fortune, success, or friendliness about 10% of the time for men and 6% of the time for women. In the normative sample, good fortunes were never followed by any "consequences," positive or negative. Misfortunes are of use to us in two ways. First, we always check the percentage of dream reports in a series or set with at least one misfortune. Second, we use the "bodily misfortunes percent" as a possible psychopathology indicator, as explained in chapter 2. This figure is 29% for men and 35% for women. The specific findings on misfortunes and good fortunes can be found in Tables E.12 and E.13.

Emotions

As mentioned in chapter 2, it is surprising how few emotions are included in dream reports. Women express more emotions than men, but the distribution of types of emotions is about the same. Negative emotions account for 80% of all emotions reported. This figure is our normative expectation when using the negative emotions percent as a possible indicator of psychopathology. The figure is 79% for men and 82% for women if only emotions expressed by the dreamer are included. The main findings on emotions are presented in Table E.14.

Settings

Few dreams lack a setting. There were virtually identical numbers of settings in both samples, 644 in the 500 male dream reports, 654 in the 500 female reports. However, our quantitative findings on the nature of settings are generally prosaic, perhaps because the categories had to be so few and so general to achieve adequate intercoder reliability.

The most useful of the findings on settings in our experience concern the dimensions of indoor/ outdoor, familiar/unfamiliar, and distorted/normal, although there are usually very few distorted settings. To make the findings on indoor/outdoor more clear, we ignore findings on "ambiguous" and "no setting" codings, where there are small numbers and no gender differences. We then can say that when a setting can be coded indoor or outdoor, women are indoors 62% of the time and men 49% of the time. This finding is presented in Table 4.5.

If we look only at settings that can be clearly coded as familiar or unfamiliar, there is once again a gender difference. Although familiar settings predominate for both men and women, women are even more likely to be in familiar settings than men. This finding is presented in Table 4.6. The details on all the findings on settings are presented in Table E.15.

Objects

There are many objects in dreams. The typical dream report has between two and nine objects, with a mean of 4.8 for men and 5.3 for women. Only 20 male dream narratives and 11 female dream narratives in the normative sample had no objects in them. Where we have found these categories of greatest use is in the study of individual dream series in which there is an unusually high percentage of some rare category. Two gender differences in the objects category have been of use in later replication studies: although the frequencies are small, men are three times as likely to dream about implements, such as tools and weapons, and women are more likely to dream about clothing.

Table E.16 presents the large number of specific findings on objects in dream reports. This table provides the basis for determining the "torso/anatomy percent" presented in chapter 2 as a possible indicator of disturbances in bodily image. As noted there, the torso/anatomy percent is determined by dividing the sum of torso, anatomy, and sexual organ references by the frequency of all body parts. The normative torso/anatomy percent is 31 for men and 20 for women.

Remaining Categories

The findings for modifiers are presented in Table E.17; negatives, temporal references, and food and eating are presented in Tables E.18 through E.20. These categories have not been used as frequently as the categories presented earlier. The findings are worthy of only the briefest mention here. The frequencies are not large in most instances, with the exception of "size" (large/small) and "intensity" on the modifiers scale. There are no major gender differences, although women tend to use more negatives in their reports than men. We have created no percentages or indexes using any of these findings. These findings await further exploration and creative development.

The h-Profile

It is often difficult to comprehend a large number of numerical comparisons. A graphic display of results makes the task easier and more interesting. It also allows us to detect patterns more quickly if we are making comparisons among many groups or individuals. We therefore use our effect sizes to create what we call the "h-profile," defined as a display of the differences between one group and another (or our norms) for any combination of our percentages and indexes we may choose to compare. In Figure 4.1 we present the h-profile for male dreamers as it compares with our female norms on several major indexes. As can be seen, it immediately shows the most salient differences, those related to aggression. It gives a sense of pattern.

How Large Should Samples of Dream Reports Be?

As noted at the outset of the chapter, the norms for the Hall/Van de Castle coding system are based on 500 dream reports from 100 females and a similar number of reports from the same number of males. But were that many dream reports necessary to obtain a representative sample of the dream life of these college students? More exactly, how large a subsample is necessary to approximate the overall findings with the full sample? We think the answer to this question is very important because it gives us a good idea of how large samples of dream reports should be in order to be confident that we have an accurate (representative) portrait of a group's dream life.

To answer our question, Adam Schneider compared randomized subsamples of 25, 50, 75, 100, and 250 dream reports coded by Hall and Van de Castle for the original normative study. Schneider used dream reports from the male sample, the only one we could find in Hall's files. The findings for each sample size were compared with each other and the norms.

The important finding that emerges from the many comparisons computed by Schneider is that it takes about 100 dream reports to approximate the norms. At one extreme, sets of 25 have widely fluctuating departures from the norms and are almost certainly not useful. At the other extreme, 250 dream narratives replicate the norms almost exactly; this finding could be of great use to those who might wish to generate new norms for people in other countries. With sets of 50, the findings on most of our measures have settled down considerably compared to sets of 25, but the average departure from the norms for most indicators was usually over ten percentage points. By comparison, the average departures from the norms were generally cut in half with 15 samples of 100 dream reports each, and were small enough to make samples of that size useful on most of our measures.

The main findings from Schneider's analyses are displayed in Table 4.7. The average departure from the norms for many different random subsamples of 50, 75, and 100 dream reports are presented. The table shows that some measures have reasonable stability at 50 or 75 dream reports, such as the A/C and F/C indexes, but that 100 narratives are needed to bring the average departure from the norms to under 10 percentage points for most measures. The information on the stability of specific measures can be used in two ways. First, it can tell researchers what sample size is needed if they are interested in focusing on only a few measures for one theoretical reason or another. Conversely, if researchers find themselves with samples of only 50 or 75 dream reports that cannot be augmented, this table tells them what indicators they can use with some degree of confidence.

Table 4.7. Average Departures from the Male Norms with Numerous Random Samples of Varying Size*

Category		Male Norms	Sets of 250	Sets of 125	Sets of 100	Sets of 75	Sets of 50	Sets of 25


Characters

	Animal Percent	6%	±0.9	±1.3	±1.7	±1.9	±4.4	±7.6

	Male/Female Percent	67/33%	±1.6	±4.3	±4.4	±6.0	±8.4	±13.6

	Familiarity Percent	45%	±0.3	±3.4	±5.3	±4.4	±8.4	±13.4

	Friends Percent	31%	±0.9	±3.1	±5.7	±6.2	±11.7	±16.1

Social interactions

	A/C Index	.34	±.018	±.048	±.044	±.057	±.065	±.136

	F/C Index	.21	±.004	±.021	±.027	±.032	±.050	±.117

	Aggression/Friendliness Percent	59%	±1.3	±4.6	±5.0	±6.2	±8.8	±18.8

	Victimization Percent	60%	±1.3	±7.1	±7.0	±8.2	±14.9	±29.2

	Befriender Percent	50%	±2.0	±8.8	±10.1	±11.8	±18.0	±30.3

	Physical Aggression Percent	50%	±1.2	±4.5	±7.2	±8.1	±15.4	±25.6

Settings

	Indoor Setting Percent	49%	±1.2	±3.2	±5.2	±5.0	±8.7	±13.4

	Familiar Setting Percent	62%	±1.3	±6.5	±5.5	±9.6	±13.1	±24.6

Other categories

	Dreamer-Involved Success Percent	51%	±5.1	±10.9	±10.5	±13.5	±26.3	±48.1

	Bodily Misfortunes Percent	29%	±0.8	±2.3	±4.4	±8.6	±12.1	±25.9

	Torso/Anatomy Percent	31%	±0.8	±7.2	±11.1	±9.2	±17.3	±28.6

	Negative Emotions Percent	80%	±2.6	±6.2	±7.5	±9.1	±13.4	±20.3

Percentage of dream reports with at least one:

	Aggression	47%	±1.9	±5.7	±7.4	±7.6	±11.3	±17.3

	Friendliness	38%	±1.0	±4.0	±4.7	±8.0	±10.7	±18.0

	Sexuality	12%	±2.2	±3.7	±4.0	±5.1	±7.3	±14.7

	Misfortune	36%	±0.4	±2.8	±4.6	±6.2	±8.7	±17.3

	Success	15%	±1.1	±3.2	±5.4	±5.6	±10.0	±15.3

	Failure	15%	±1.9	±4.1	±5.2	±6.2	±8.7	±15.3

* Three sets of random numbers were utilized, making it possible to draw 20 sets of 25, 10 sets of 50, 6 sets of 75, 5 sets of 100, 4 sets of 125, and 2 sets of 250 three different times from the full set of 500 dream reports, thereby creating a total of 60 sets of 25 reports, 30 sets of 50 reports, 18 sets of 75, 15 sets of 100, 12 sets of 125, and 6 sets of 250. The phrase "average" departure from the norms" is used to avoid any confusion that might arise from the use of the statistical terms "deviation" and "range." The numbers in the table are an average of all the departures from the norms for a given set of samples. The use of the plus/minus signs indicates that the departure is, on average, that many points above or below the normative figure in the lefthand column of the table. For example, the ±1.7 for the Animal Percent for sets of 100 dream reports means that the findings for the subsets fell, on average, between 4.3% and 7.7% (the normative figure is 6%). Similarly, the figure ±4.4 for the A/C index for sets of 100 dream reports means that the subsets fell on the average between .296 and .384 (the normative figure is .34).

We think the findings presented in this section also can be used as one dimension in judging the usefulness of past studies of the dream life of a group. We also think these findings could be used to set the minimal sample size for publishable quantitative studies of dream content using the Hall/Van de Castle system. We recognize that statistical tests might be able to detect large differences between two or more groups with sample sizes smaller than 100. However, we think the focus of future research using this coding system should be on what we believe is a more useful kind of comparison, namely, one with the Hall/Van de Castle norms. Sample sizes of at least 100 are needed to approach this goal.

By coincidence or not, findings we will present in chapter 7 lead to a similar figure for what constitutes an adequate sample of the dream life of an individual. That is, it once again takes a subsample of about 100 dream reports for the quantitative findings to resemble those for the entire series of 200 or more dream narratives.

Most Recent Dreams: A New Methodology

Some potential dream content analysts may worry that they will not be able to obtain enough dream reports to do good studies, but we have found there is a simple and excellent way to create very large representative samples. Since the Hall/Van de Castle norms can be approximated with 100 everyday dreams from 100 people, it follows that 100 Most Recent Dreams should provide an adequate sample. We have tested this hypothesis and found support for it with 100 Most Recent Dreams from women students at the University of California, Santa Cruz, in 1992 and 1993. We simply ask students to write down the most recent dream they can remember, whether it was "last night, last week, or last month." To reinforce this emphasis on recency, we first ask people to write down the date and the time of day they recalled the dream. This information also can be used to screen out dreams not recalled within the past few days, thus making even more certain that the dream reports in the sample are recent ones. The instructions we use are reprinted in Appendix C.

The importance of emphasizing Most Recent Dreams in such studies cannot be overestimated because of the tendency for people to report a recurrent dream, nightmare, or an especially unusual dream if they are asked in a general way to report a dream. This is what happened, for example, when Stanley Krippner asked a television audience to send him a dream: 33% of the dream reports he received were described as recurring ones, and many respondents reported that they experienced the dream years earlier, sometimes in childhood (Rubenstein and Krippner, 1991: 41). It is not surprising that the findings differed from the Hall/Van de Castle norms.

Studies based on Most Recent Dreams hold out the possibility of expanding dream content studies to a wide range of subgroups within the college population from skydivers to racial minorities to gays and lesbians. They also hold the potential for cross-national studies in which Most Recent Dreams are collected on the same date in many different universities around the world. Moreover, large samples of Most Recent Dreams can be analyzed very quickly by focusing on the question of what percentage of the reports have at least one aggression, friendliness, sexuality, misfortune, success, or failure. This approach allows investigators to bypass coding the dream narratives for characters and avoids detailed classification of social interactions.

Given the solid rationale for studies of Most Recent Dreams using 100 or more subjects in each group, the day of studies using small samples of two to five dreams from a handful of subjects should be over. Such studies lead nowhere intellectually, as the problems they cause in the dream content literature convincingly show. If it is not possible to obtain lengthy dream series from individuals, then it is far better to work with one recent dream from each of many subjects, as our research demonstrates. It is also easier to obtain one Most Recent Dream from students in a classroom, dorm, or dining hall than it is to obtain dream diaries with even five dreams in them from more than a handful of subjects.

Replications and Follow-ups

We turn now to several follow-up studies that can be seen in part as attempts to replicate the findings presented so far in this chapter. They are more than replications, however, for some of them sought to determine if historical changes, all-female settings, or racial and ethnic differences might influence the findings.

The Richmond Study

Dream reports collected from predominantly European-American students at the University of Richmond in 1979 for a study of emotions in dream reports (Stairs and Blick, 1979) provided an unplanned opportunity to see what if anything had changed in the dreams of college students over a 30-year period (Hall, Domhoff, Blick, and Weisner, 1982). This Richmond sample, as we came to call it, consisted of three, four, or five reports by each of 53 males and 69 females for a total of 263 male reports and 340 female reports. The reports were collected with instructions very similar to those used by Hall in his earlier work. Case Western Reserve University and the University of Richmond, although in different parts of the country, are both private, urban coeducational universities whose students come from the middle and upper-middle socioeconomic strata. The dream reports were collected in psychology classes at both universities.

Due to time constraints on the investigators, only selected categories from the Hall/Van de Castle system were utilized. The categories were selected on the basis of three criteria. First, they had been used in several studies subsequent to the normative study. Second, they are categories with large frequencies. Third, some of them have shown gender differences and others have not.

The findings of this study can be presented in two ways. First, the Richmond males and females can be compared with the Hall/Van de Castle norms for males and females. The results of that comparison show there are few differences and that the differences are relatively small in magnitude even when they are statistically significant. Specifically, Richmond males and females report a higher percentage of familiar characters than the norms. Richmond males have a lower percentage of total friendliness, friendliness with males, and friendliness with unfamiliar characters. Richmond females have a lower percentage of total friendliness. Richmond males have a lower percentage of dreams with at least one sexual encounter and with references to clothing. Richmond females also have a lower percentage of references to clothing. In all, of 40 comparisons, only nine are statistically significant, six for males, three for females.

The second way to compare the findings is to see if the gender similarities and differences for the norms are the same as they are for the Richmond study. Rather remarkably, there was not a single change. Where there were gender differences in the norms, there were gender differences in the Richmond study. Where there were no gender differences in 1950, there were no gender differences in 1980.

Hall et al. were careful to note they did not know why there were so few differences between the two sets of dream reports. They mention three "possible" answers: that "human nature" has not changed, that 30 years is too short a time for the effects of social changes to manifest themselves in dream reports, and that any changes in American society in the past 30 years have not been very significant (Hall et al., 1982:193). They then wrote that "we cannot say which, if any, of these answers is the correct one ..."

Despite this caution, their conclusions have been misinterpreted by some dream researchers. For example, after reporting the findings accurately, Cartwright and Lamberg (1992:82) imply that the authors attribute the lack of differences to the alleged fact that "human nature has not changed." Since few readers of Cartwright and Lamberg's popular book are likely to read the original report, this mischaracterization of the "interpretation" of the findings can be described not only as inaccurate, but as unfortunate. Statements like Cartwright and Lamberg's obscure the fact that extremely important findings were reported. Such statements also steer feminist social scientists away from quantitative content analysis.

The Salem College Study

Salem College is a small, private all-women's college in Winston-Salem, North Carolina. It is very similar to Case Western Reserve and the University of Richmond in that its students are mostly European-Americans who come from middle and upper-middle social strata. In 1987 Linda Dudley and one of her students (Dudley and Fungaroli, 1987) collected dreams from 12 female students enrolled in a special upper-division topics course in psychology to see if the dream reports of students at a women's college differed from those of women in the normative sample. With a sample of 141 dream reports, they found there were both striking similarities and some interesting differences.

The consistencies were found for character and settings. The male/female percent for Salem women was 46/54 as compared with 48/52 in the norms. The familiarity percent was 65 at Salem compared to 58 for the norms. The percentage of indoor settings was 60 at Salem compared to 61 for the norms.

There were some interesting differences in the patterns of social interactions. Whereas women in the Hall/Van de Castle normative study had more aggressive acts with males than with females (22% with males vs. 14% with females in the norms), the women at Salem College had the same percentage of aggressions with male and female characters (14% vs. 15%) due to less aggressive interactions with males. In keeping with this finding for aggressions, it was found the Salem women had a greater percentage of friendly interactions with women than the normative group (21% vs. 15%) even though their level of friendly interactions with male characters remained higher than with female characters (27% for Salem, 24% for the norms). Taken together, these findings lead to the cautious "interpretation" that the all-women's environment led to less aggression with men and more friendliness with women. It is the kind of finding that is clearly worth testing in other social situations.

In spring, 1990, Dudley did a follow-up study with 33 women in an introductory psychology class (Dudley and Swank, 1990). Their average age was 18.5. The findings were very similar to her first study. The women continued to have a higher percentage of friendliness with women than women in the normative group. However, the decline in aggression with male characters was not as great, perhaps because they had not been at Salem College as long. The findings for the two Salem studies are compared with the norms in Table 4.8.

Table 4.8. Comparison of Norm Women with 1987 and 1990 Salem Women Across Content Categories

Content Variable		Hall/VDC Norms	1987 Salem	1990 Salem


Characters

	Percent Male	48	46	49

	Percent Female	52	54	51

	Percent Familiar	58	65	65

Interactions

	Aggression with Males	22	14	19

	Aggression with Females	14	15	14

	Friendliness with Males	24	27	27

	Friendliness with Females	15	21	21

	% dreams with sex	04	Not Studied	08

Settings

	Indoor	61	60	61

	Outdoor	39	40	39

Adapted from Dudley, L., & Swank, M. (1990). A comparison of the dreams of college women in 1950 and 1990. ASD Newsletter 7(5), 3.

Dudley and her students have made an important contribution to the quantitative study of dream content by realizing that they had a unique situation in which to collect dream reports. Their findings have strengthened our confidence in the normative findings and given us new hypotheses to explore about why dream reports may vary in their percentages of friendly or aggressive interactions with male and female dream characters.

The Berkeley Study

The first study of the dream reports of women at a major public university was made by Tonay (1990-91). Moreover, that university was the best-known campus in the state of California, the University of California at Berkeley. Tonay's study included many safeguards to improve the quality of her data. It also included an ethnicity component. It did not include men because not enough of them turned in dream reports.

The subjects were 100 of 220 women attending an upper-division psychology class on the Berkeley campus. They participated voluntarily and received no compensation. Students were invited to submit five consecutive dreams on basically the same forms originally used by Hall and assured of anonymity. After three weeks, 109 dream booklets had been collected. Five subjects said they recalled no dreams during the time period, leaving 104 subjects. Four other subjects later admitted on the post-study questionnaire discussed in the previous chapter that they had made up dreams, leaving 100 subjects and 500 dream reports. The typical subject was 21 years old, single, middle class, and a social science major with a 3.22 average. Most of the subjects were European-American or Asian-American. The volunteer subjects did not differ on any of these variables from the students in the class who did not participate.

The dream reports were put into a common pile and randomized before they were coded by two female undergraduate assistants trained by Tonay. Dreams were randomized again after the coding for each category. A third person compared the coding by the two raters. The rare disagreements were resolved by agreement between the two raters. The findings for 24 comparisons in the categories for characters, settings, friendliness, and sexuality did not reveal a single significant difference from the norms. For emotions and aggressions, there were three differences, all of which may be plausibly interpreted as showing more assertiveness and outgoingness on the part of these women. First, there were more emotions in their dream reports. Second, they were more likely to be victims of women characters than in the past. Third, they were more likely to be aggressors than in the past, meaning in the terms adopted for this book that their victimization percent had declined. However, Tonay counted aggressions reciprocated by the dreamer as instances of being an aggressor, so her findings are not strictly comparable to the norms on this issue. Including reciprocal aggressions in our equation for the victimization percent drops it from 67 to 60 for women and from 60 to 52 for men.

Tonay's study included a comparison of Asian-American and non-Asian-American women. As Table 4.9 shows, there were very few differences between the two samples. The Asian-American women were more likely to be in unfamiliar settings. They were less likely to act as the aggressors in their dreams, and were more often the victims of male characters than non-Asian-Americans. Citing studies of the deferential way in which Asian-American women are often socialized (Hsu, 1971; Fong, 1973; Chow, 1984), Tonay (1990-1991:93-94) hypotheses that the differences on the aggression scores are consonant with subcultural differences.

Table 4.9. Sub-Cultural Comparison: Dream Content of Asian-Americans and Non-Asian-Americans at Berkeley

Category		Percentages

Category		Asian-Americans	Non-Asian-Americans

Characters

	Animals	01	04

	Humans	97	95

	Males	30	35

	Females	40	33

	Familiar	51	56

	Unfamiliar	49	43

Setting

	Indoor	49	41

	Outdoor	32	32

	Familiar	25	28

	Unfamiliar	29*	04

Friendliness

	Dreamer-involved	80	80

	D as befriender	38	39

	D as befriended	54	42

Emotions

	D-experienced	90	81

Aggression

	Dreamer-involved	79	84

	D as Aggressor	27*	49

	D as Victim of males	50*	36

	D as Victim of females	35	35

* p < .05, two-tailed.
Note: Various percentages do not sum to 100 (e.g., indoor/outdoor, familiar/unfamiliar, befriender/befriended) because findings for some categories are not included in the table.

The few interesting differences aside, what stands out in the findings of these first three replication studies is their consistency with previous findings. Dream reports collected at Case Western Reserve (1950), the University of Richmond (1979), the University of California, Berkeley (1986), and Salem College (1987 and 1990) all provide essentially the same results. This is an impressive record of replications. It would seem that the norms for the Hall/Van de Castle system are a rock we can stand on in a sea of uncertainty when it comes to studying home dream reports from European-American and Asian-American college students.

Crane Community College: An African-American Sample

In spring, 1967, an instructor in psychology at Crane Community College in Chicago, Stephen Gornik, offered to collect dream reports for Hall in his psychology class. All of the students were African-Americans. Students were asked, but not required, to turn in their dreams on the standard report form. Gornik received 101 reports from 12 males and 233 reports from 25 females. The students ranged in age from 18 to 29; most were in their early 20s. Unlike our earlier samples, many of these students still lived at home, which may be a factor in some of the differences Hall found. Gornik characterized these students to Hall in the following words:

The student population represents what I would call a rising Negro middle class. While they come from poverty areas, they are strongly motivated to achieve and improve themselves, but in a direction that is not clearly delineated for them (Personal Letter to Calvin S. Hall, August 16, 1967).

Hall analyzed the dream reports for all the Hall/Van de Castle categories except settings, activities, and descriptive elements. The findings for males and females were compared with the Hall/Van de Castle norms and then with each other. There were a great many similarities with the norms, but there also were some interesting differences as well. We begin with a brief summary of the similarities.

There were no differences between the two groups on objects, emotions, misfortunes and good fortunes, success and failure, number of sex dreams, or number of food and eating dreams. There were no differences between African-American males and European-American males on overall aggression or on any of our measures relating to friendliness. Where, then, do any differences lie?

First, African-American males had five differences with European-American males, three in character categories, two on aggression measures:

Before we comment on these differences, let us look at some differences between the African-American and European-American females:

What is most striking between the two sets of women, however, is that the African-American females were both more friendly and more aggressive in their dream reports:

Taken as a whole, what do these differences between European-Americans and African-Americans seem to indicate? We are nowhere near the point in our presentation of the full range of findings with the Hall/Van de Castle system where we want to try to make a case for a relationship between dream content and either personality or culture. Nor do we have enough information on the people in this sample to make a very strong argument even if we were ready to do so. Then too, the higher percentages for family members may be due to living at home instead of in a dorm or separate apartment.

Still, we would hazard two hypotheses, which we will immediately "test" when we compare gender differences within the African-American sample. First, unlike the European-American sample, women and the family seem to be as salient in the dreaming minds of the men as they are in the minds of the women. Second, there is less of a gender difference in social interactions in the African-American sample. Let us now see how these suggestions fit with gender findings within the African-American sample.

For most of our categories, the gender differences within the African-American sample are similar to those in the Hall/Van de Castle norms. This finding is in effect contained within the long list of "no differences" between the two groups of males and two groups of females with which we began a recounting of our findings several paragraphs ago. However, some of the differences we then reported between the two groups of males and females also imply gender differences within the African-American sample. These differences seem potentially quite interesting.

First, the usual gender difference in the male/female percent disappears, showing an equal saliency of male and female characters in both genders. This point is made in Table 4.10, comparing male and female dreamers in both samples.

Second, the gender difference on familiarity percent is reversed. For European-Americans, women have the higher familiarity percent; for the African-Americans in this sample, men have the higher familiarity percent. This reversal is entirely due to the difference between the two samples of males, as can be seen in Table 4.11.

Finally, the gender difference in the A/C index is reversed. For European-Americans, the males have a higher A/C index. For African-Americans in this sample, women have the higher A/C index. This reversal is due to the difference between the two samples of females, as can be seen in Table 4.12.

As we noted a moment ago, we are not in a position to draw conclusions from these findings, but there are interesting and suggestive patterns implied. There seem to be subcultural differences between the two groups, and the findings on gender differences suggest gender relations may be very different in the two groups. It is also noteworthy that African-American women in this sample are both more friendly and more aggressive with each other in their dreams than European-American women are with each other.

The foregoing paragraph is very far from an analysis. We would need behavioral information of the kind we are going to present in chapter 8 before we would begin to take any analysis seriously. We also would want to replicate the findings in other African-American samples. It would make the findings even more powerful if they could be replicated in other tight-knit communities, or in other groups where male/female power relations are thought to be more equal, whatever the race or ethnicity of these groups.

Mexican-Americans at Two Different Universities

There is one study relating to Mexican-American college students using the Hall/Van de Castle system (Kane et al., 1993). It compares Mexican-American and Anglo-American women from four colleges and universities in San Antonio with each other and with Mexican women from four universities in Mexico City. The students were given dream diaries in which they were asked to record the first five dreams they recalled in a three-week period. Many hundreds of dream reports were collected. A random sample of 30 women was drawn from each group of between 40 and 50 women, and one dream was drawn randomly from the series provided by each of the 30 women. The dream reports were compared for number of characters, emotions, aggressions, friendly interactions, sexual interactions, achievement strivings, and misfortunes and good fortunes. There were no statistically significant differences between the Mexican-American and Anglo-American women in any category. There were two significant differences between Anglo-Americans and Mexicans: there were more emotions and more misfortunes and good fortunes in the dream reports of the Mexicans. There is, unfortunately, one serious drawback with this study. There are only 30 dream reports in each sample. The rest of the dream reports need to be analyzed, or else new studies need to be done, before we can be confident of the reported findings.

Although there are no studies using the Hall/Van de Castle system that included Mexican-American male students, Brenneis and Roll (1975) and Brenneis 1976) compared the dream reports of Mexican-American and Anglo-American male and female students at the University of New Mexico using the scale of "34 ego dimensions" developed by Brenneis and discussed in chapter 2. Their results are not strictly comparable with other findings in this chapter, but there is enough overlap in some of the scales to make several of their findings of interest.

For this study there were 42 Chicano males, 65 Chicana females, 61 Anglo males, and 74 Anglo females. All were volunteers. The average age for all four groups was between 20 and 21. The Chicano volunteers were accepted if Spanish was spoken in their homes and if they identified themselves as Chicanos. There was a total of 1123 dream reports, 203 from male Chicanos, 309 from female Chicanos, 283 from male Anglos, and 328 from female Anglos (Brenneis, 1976:281).

If we look first at the findings for the Mexican-Americans, we see there are a number of gender differences that are by now familiar to us. There are more characters in women's dreams, and a greater percentage of them are known to the dreamer. There is a higher percentage of male characters in the men's dream reports. Male dreamers more often have "contentious" or "negative" interactions; there is more aggression, sexuality, and depiction of physical injury in their dream reports. Women are more likely to be in enclosed and familiar settings, men to be in larger and more open settings (Brenneis and Roll, 1975:179; Brenneis, 1976:283-284).

The main cultural difference between the Mexican-Americans and Anglo-Americans was a greater exaggeration of the gender differences in Mexican-Americans (Brenneis, 1976:287). In addition, more familiar settings appear in the dream reports of Mexican-Americans, and there was a tendency -- which did not reach statistical significance -- for there to be more characters (Brenneis, 1976:284). Brenneis (1976:287) also claims "death themes" appear more frequently in the dream reports of Mexican-Americans. He does so on the basis of earlier studies by Roll, Hinton, and Glazer (1974) and Roll and Brenneis (1975). In the first of these two studies, there is no analysis of dream content, nor even any dream reports, but only the results of a questionnaire in which subjects were asked "Have You Ever Dreamed Of ..." for 48 topics.

Perhaps recognizing the weaknesses of questionnaire studies of "dream content," a second study was done in which five dream reports were analyzed from 40 male and 40 female Mexican-American college students and 40 male and 40 female Anglo college students at the University of New Mexico. The study claims there is the same difference in "death themes," but the evidence for this claim is doubtful at best. In this study the problem lies in the coding categories, which include (1) "clear and obvious" death themes and (2) allegedly "symbolic" death themes (Roll and Brenneis, 1975:380). The "obvious" death themes were (1) dead people as though alive; (2) a relative or friend dead; (3) seeing yourself as dead; (4) being hanged by the neck; (5) being buried alive; (6) seeing mortuaries or graveyards; and (7) being in heaven or hell after death. The allegedly "symbolic" death themes are (1) losing a limb; (2) being unable to breathe; (3) seeing lunatics or insane people; (4) having teeth fall out or be pulled out; (5) being tied and unable to move; (6) seeing a creature part animal and part human; and (7) seeing a body change grotesquely (Roll and Brenneis, 1975:380). No evidence is presented that any of these elements relate to death, but this symbolic scale is the key to the whole study, as we shall now show.

The alleged cultural differences achieved statistical significance for women only. However, "if the realistic and symbolic content areas were treated separately, they do not achieve statistical significance" (Roll and Brenneis, 1975:381). In other words, the claim stands or falls, to the degree it can stand when males do not show the difference, on the validity of the allegedly "symbolic" death themes. Roll and Brenneis provide no frequency counts by categories, so it is impossible to know where the two samples of women students may differ. It would be interesting to know out of empirical curiosity just how often these students dream of seeing mortuaries, being hanged by the neck, being in heaven or hell, losing a limb or teeth, or seeing insane people. Our guess, based on our studies, is that it would not be very often.

Further support for the relative lack of differences between Mexican-American and Anglo-American dreamers can be found in a study using the Hall/Van de Castle scales to study aggressions and misfortunes in the dream reports of 140 Mexican-American and Anglo-American teenage delinquents and non-delinquents in Tucson, Arizona (Leman, 1967). Subjects were matched for age, intelligence, and socioeconomic level. There were no ethnic differences on either scale.

All in all, the dream reports of Mexican-Americans seem very similar to those of Anglo-Americans. There may be some cultural differences, but the similarities and the common gender differences seem to be greater.

Gender Changes or Flawed Methods?

Milton Kramer and his colleagues at the University of Cincinnati have applied the Hall/Van de Castle coding system to dream reports collected in the laboratory. Most of their work is innovative and useful. One of their studies concerned the norms for the character categories (Reichers, Kramer, and Trinder, 1970), where they reported a very great similarity between the distribution of characters in their sample of 418 laboratory reports and the Hall/Van de Castle norms.

Another study concerned gender differences in laboratory dream reports (Kramer, Kinney, and Scharf, 1983). It is based on 11 males and 11 females, ages 20-25, who were students at the University of Cincinnati. Subjects were awakened at the end of each of the first four REM periods and asked to report what was going through their minds. In all, 1190 reports were collected from 1659 awakenings, for a recall rate of 72%.

The authors found there were no gender differences in the percentage of reports containing at least one aggression or at least one misfortune suffered by the dreamer. Aggressions appeared in 12% of the dream reports of both men and women. Women had misfortune to themselves in 12% of their reports, men in 9%. The authors conclude that well-known gender differences on aggression have thus disappeared, but their conclusion is mistaken.

According to the Hall/Van de Castle norms, 47% of male dream reports have at least one aggression as compared to 44% of female dreams. Thus, there never was a difference between males and females on aggression if the focus is on the percentage of reports with at least one aggression. What the Kramer, Kinney, and Scharf findings show is a smaller percentage of aggression reports for both sexes, which is in keeping with the differences between laboratory and home reports discussed in the previous chapter. The same analysis obtains for misfortunes, as can be seen in Table 4.13.

Kramer, Kinney, and Scharf also coded the dream reports using Hall/Van de Castle scales for characters (10 subscales), activities (9 subscales), and modifiers (21 subscales). They found male dreams had more male characters and more strangers, which is consistent with the Hall/Van de Castle norms. Males tended to mention large sizes more, also in keeping with the norms. However, Kramer, Kinney, and Scharf also found men higher on five subscales where there had been no gender difference previously, but the reasons are methodological, as we shall soon show. Most of these differences are on minor elements tending to occur infrequently in the Hall/Van de Castle norms. We report the Hall/ Van de Castle findings on these elements in Table 4.14, then show what Kramer, Kinney and Scharf did.

Using raw frequencies for each element, Kramer, Kinney, and Scharf turned the Hall/Van de Castle findings into male/female ratios. They then compared these ratios to their ratios in Table 4.15.

Using the same procedure, Kramer, Kinney, and Scharf found women to be higher on two other minor subscales in the modifier category that previously had shown no gender difference (Table 4.16).

Contrary to Kramer, Kinney, and Scharf (1983:1), their findings do not support their claim that the "sexual revolution of the past two decades has indeed had some psychological impact in altering some of the traditional differences" between the sexes. The different findings are in fact an artifact of their flawed methodology.

The first problem with these findings is that the authors used raw frequencies for male and female dreamers in the Hall/Van de Castle norms to create their M/F ratios. Such a procedure does not correct for report length or different numbers of characters, activities, and modifiers in male and female reports. In the Hall/Van de Castle norms, it will be recalled, the women's dreams were 8% longer and contained more characters (1423 vs. 1180), more activities (2470 vs. 2362), and more modifiers (1458 vs. 1110). Thus, ratios based on raw frequencies will be higher than a comparison based upon indexes that control for differences in length of reports and raw frequencies. Ratios based on raw frequencies may create "differences" where there are none when the proper corrections are not made for differing report lengths and frequencies. For example, the percentage of single characters for female dreamers (72%) is not much higher than for male dreamers (69%) even though there are more single characters in women's dream reports. The Cincinnati study turned this small difference into an M/F ratio of .79 by using raw frequencies.

With this information as a context, we can now see the second reason for the differences from the norms in the Kramer, Kinney, and Scharf study: the men's dream reports turn out to be longer than the women's for a change (116 words vs. 92). Thus, there is likely to be "more" of everything.

There is a third problem. It is not a good statistical practice to use ratios with very low frequencies. Just a few cases can change ratios dramatically. Let us use the mention of straight lines ("straight plus") as an example. The M/F ratio for Cincinnati is 3.55 and only .57 for Hall/Van de Castle, but in the Hall/Van de Castle norms we are talking about a frequency of four out of 1,110 modifiers in male dreams and a frequency of seven out of 1,458 modifiers in female dream reports. That is, the .57 M/F ratio is based on dividing seven into four. Moreover, if we use percentages, as we should, there is no gender difference -- "straightness" is used as a modifier 4% by males and 4% by females in the Hall/Van de Castle norms.

We have discussed the Kramer, Kinney, and Scharf findings and claims at length for two reasons. First, we think it is always a good idea to check for methodological differences when new findings contradict established findings, rather than immediately to assume the differences are real and then begin theoretical speculations. There often are real changes, but all explanations in terms of sampling and statistical procedures should be thoroughly examined before the findings are assumed to be different. In this particular case, the samples are different (men's reports are longer) and percentages are not used. It is essential to follow all the rules of the system to make future findings compatible and cumulative.

The second reason we have discussed these findings at length is that they have been used to question the current accuracy of the Hall/Van de Castle norms (e.g., Lortie-Lussier, Schwab, and DeKoninck, 1985:1012; Rubenstein, 1990:139; Rinfret et al., 1991:180; Lortie-Lussier et al., 1992:81,93). There is thus a danger that those new to dream research will believe that the Hall/ Van de Castle findings are "outdated." But they are not, as the replication studies from the 1980s cited earlier in this chapter have demonstrated, and as a revised analysis of the Cincinnati findings also indicates.

Conclusion

The numerous findings presented in this chapter give us very strong reason to believe that we know what American college students dream about. The findings are robust over 45 years and different regions of the country. They come from both male and female investigators. In short, these findings with the Hall/Van de Castle system are the most reliable, systematic, and potentially useful quantitative findings on dream content ever developed. They deserve to be taken seriously as a starting point for further research investigating the possible relationships of age, class, race, ethnicity, and nationality to dream content. Studies concerning these variables now can be done with confidence with one Most Recent Dream from each of 100 subjects, then displayed in an h-profile.

These findings also give us our first reason to believe that there is meaning in dreams because they show a coherence or regularity in the relatively small but consistent differences in what women and men dream about. In that regard the findings on dream reports are similar to what is found in many other psychological studies comparing women and men. The reasons for these small differences are many and varied, but they are not our concern here. What is important for our purposes is the regularity of the findings in several different samples from various time periods and regions of the country.

Chapter 4: Normative Findings on American College Students

Introduction