Chapter 6: Cross-Cultural Studies of Dream Content
In this chapter we will present findings on dream content from a wide range of other nations and cultures. Some of these cultures are very similar to the United States in level of education or degree of urbanization and industrialization. Others are preliterate in nature, including some of the smallest and least structured societies in the world. Whatever the size of the society or its material base, however, the chapter will show there are both similarities and understandable differences in the findings when compared with our norms. The chapter will begin with findings from such modern nation-states as the Netherlands, Argentina, India and Japan, and then turn to findings on dreams collected by cultural anthropologists in past decades from small-scale societies.
Before presenting the findings, it is essential to stress that there are great difficulties in doing cross-cultural research on dream content. These difficulties multiply as the differences between American society and the group under study increase (cf. D'Andrade, 1961; Eggan, 1961; Kracke, 1979, 1987; Dentan, 1983, 1986; 1988; Tedlock, 1981, 1987, 1991). For example, there are issues of access, trust, and establishing rapport. There are problems of common understanding in many cases. Even when these obstacles are overcome, there can be problems obtaining a representative sample of dreamers. Informants, once established, may be eager to report dreams that are part of the cultural lore rather than their own dreams, or be reluctant to report certain dream contents to outsiders.
So, with a clear understanding of the obstacles to cross-cultural studies in mind and the likely uneven quality of the data, we turn to the substance of the findings.
Three Dutch investigators (Waterman, de Jong, and Magdelijns, 1988) have provided us with what is in effect a replication of aspects of the Hall/Van de Castle norms for the Netherlands. The study was designed to see if what they call "sex role orientation," that is, the mixture of "masculine" and "feminine" attributes in a person, was more important than gender itself in predicting dream content. They found gender differences were more important than sex role orientation, and their findings were similar to those of the Hall/Van de Castle norms.
The subjects were students in a college psychology class in Amsterdam, 34 women, 32 men, ages 19 to 31. Each contributed five dreams for a total of 330 dream reports. The reports were coded for aggressions and friendliness. The results were analyzed using some of the percentages and indexes presented in chapter 2 of this book. The formula for the significance of differences between proportions was utilized. The results of this significance test were checked with three separate log linear analyses (2x2x2 design) using the raw data.
As with the Hall/Van de Castle norms, there was more aggression than friendliness in both male and female dream reports, and the aggression/friendliness percents were higher for the male dreamers. A comparison of Dutch and American dreamers on aggression/friendliness percents is presented in Table 6.1.
Table 6.1. Aggression/Friendliness Percents
The American and Dutch dreamers also were similar in their degree of victimization, with women in both cultures having a higher victimization percent than men. Here the one difference for sex role orientation occurred: the most "feminine" women were more often victims than the most "masculine" women.
Finally, there was a far lower physical aggression percent for both males and females in the Netherlands, as can be seen in Table 6.2.
Table 6.2. Physical Aggression Percenta
ah = .37 for males; h= .48 for females.
So, here we have encountered our first cross-cultural difference. If this finding on physical aggression percent were to be repeated in other Dutch samples, then the fact that the United States is far and away the most violent industrialized nation in the world, and the Netherlands one of the least violent, would become of great interest. Furthermore, we would begin to wonder if the national violence rankings that can be derived from the highly detailed homicide and violent crime statistics compiled and analyzed by Dane Archer and Rosemary Gartner (1984) might correlate with the physical aggression percent in dream reports.
Inge Strauch of the University of Zurich, one of the pioneers of the new laboratory dream research in the 1960s, and her collaborator Barbara Meier, studied 500 REM dream reports collected in their sleep laboratory over the years (Meier and Strauch, 1990; Strauch and Meier, 1992, 1995). The 500 dream narratives came from 44 subjects who spent a total of 161 nights in the sleep laboratory, each contributing between two and 19 reports. The 26 women subjects contributed 331 (66%) of the reports and the 18 male subjects contributed 169 (34%). The women's dream reports were longer and more detailed than those of the men (Strauch, personal communication, July 1, 1993). The subjects ranged in age from 19 to 35 years.
The dream reports were coded for settings, characters, and social interactions with the Hall/Van de Castle scales. They also were classified for realism, references to everyday life, and references to leisure activities with Strauch and Meier's own rating system. We will focus here on the findings with the Hall/Van de Castle categories, although it can be said in passing that most dream reports were rated by Strauch and Meier as creative elaborations of realistic situations dealing with leisure, household, and other aspects of everyday life.
As can be seen in Table 6.3, there were very few differences between the Swiss laboratory dream reports and the American norms in the categories of settings and objects. The Swiss women are similar to the American women except for a higher percentage of male characters and strangers, and the Swiss men were similar to the American men except for a higher percentage of indoor settings.
Table 6.3. A Comparison of Swiss and American Dreamers on Selected Categories of Settings and Characters
Indoor Setting Percent
The differences between the Swiss and American samples are greater in the categories of aggression and friendliness, as seen in Table 6.4. This table shows first of all a smaller percentage of the Swiss dream reports contain at least one aggression, with an especially dramatic difference between the males (23% Swiss vs. 47% American). However, this difference has to be interpreted with caution because of the fact, reported in the third chapter, that laboratory dream reports in the United States also show less aggression (Domhoff and Kamiya, 1964a; Hall, 1966a). Thus, the more interesting finding may be the lower physical aggression percent for Swiss men and women, with the largest difference once again between the males (29% Swiss vs. 50% American, h=.43). It is also noteworthy that the figure for Swiss males is close to the 32% for Dutch males reported in the previous section.
Table 6.4. A Comparison of Swiss and American Dreamers on Aggression and Friendliness
Percent of Dreams with Aggressions
Percent of Dreams with Friendliness
Physical Aggression Percent
Inviting and Verbal Friendliness
There are also differences in the types of friendliness expressed by Swiss and American dreamers, a finding made even more interesting by the fact the differences in the percentage of dream reports with at least one friendliness are relatively small for men and nonexistent for women. Swiss dream reports are more likely to contain friendly interactions of a verbal nature, such as invitations, compliments, or greetings. Conversely, the friendliness in American dream reports is more likely to involve gift giving, efforts at help or support, and physical displays of friendliness.
In all, there are both reassuring similarities and intriguing differences in the detailed and careful work of Strauch and Meier. The fact that the Swiss sample is from the laboratory and the American norms from everyday recall injects a note of caution about some of the findings on social interactions, but the Swiss picture on physical aggression percent is also consistent with findings on Dutch dreamers reported in the previous section.
In a series of investigations, Monique Lortie-Lussier and her colleagues and students at the University of Ottawa in Canada have studied French-Canadian and British-Canadian citizens to see if and how social roles influence dream content. One primary focus has been on the dream reports of married, college-educated mothers who are exclusively homemakers and married, college-educated mothers who are employed outside the home. Female college students and college-educated, employed fathers also have been studied (e.g., Lortie-Lussier, Schwab, and De Koninck, 1985; Rinfret, Lortie-Lussier, and De Koninck, 1991; Lortie-Lussier, Simond, Rinfret, and De Koninck, 1992).
There are many similarities between the findings of these investigators and the Hall/Van de Castle norms. For example, findings on various Hall/Van de Castle scales for 15 college-educated French-Canadian housewives (Lortie-Lussier, Schwab, and De Koninck, 1985:1016) and 18 single, French-Canadian undergraduates ages 18 to 22 (Rinfret, Lortie-Lussier, and De Koninck, 1991:187) are compared with the American female norms in Table 6.5.
Table 6.5. A Comparison of French-Canadian and American Women on Selected Content Categories
American Female Norms
Physical Aggression Percent
F/C with Male Characters
F/C with Female Characters
a Recall that the F/C was .04 for women over 30 in Table 5.1.
For the issues of concern to Lortie-Lussier and her co-authors, the variations among their samples are more important. In the first study, however, Lortie-Lussier, Schwab, and De Koninck (1985:1015) found only one statistically significant difference between housewives and wage-earning women out of ten comparisons. This comparison showed more negative emotions in the dream reports of the wage earners. They then turned to a statistical technique called discriminant analysis to see if patterns of differences could be detected in their data. This analysis revealed a four-variable pattern that was created by slightly more residential settings and overt hostility in the housewives' reports, and slightly more indoor settings and negative emotions in the wage earners' reports. In the Rinfret, Lortie-Lussier, and De Koninck (1991:184-186) study, the working mothers ages 27 to 39 dreamed more often of the work environment, their husbands, their children, and unpleasant emotions than did college-age women students. Another paper from this group (Lortie-Lussier, Simond, Rinfret, and De Koninck, 1992) claims "convergence" between the dream content of employed fathers and employed mothers as compared to women homemakers, suggesting that work roles can be as influential as gender roles.
We believe these findings show there are once again considerable cross-national similarities. Nevertheless, the Lortie-Lussier, Schwab, and De Koninck (1985) paper is often joined with the mistaken interpretations by Kramer, Kinney, and Scharf (1983), discussed at the end of chapter 4, to criticize the Hall/Van de Castle norms. For example, Garfield (1988:26) uses the Lortie-Lussier, Schwab, and De Koninck (1985) findings with discriminant analysis to refute her own false claim that Hall and Van de Castle "were the first to say that a woman's place was in the home" in her dreams. Aside from the technical point that no one finding can be isolated from the pattern in a discriminant analysis without losing its significance, and that there were no differences in the mean number of indoor settings in a direct comparison of the two samples, Garfield's comment ignores the fact that it was the working women who had more indoor settings, not the housewives. She also overlooks the fact that the Hall/Van de Castle system only codes for "indoors" and "outdoors," and "familiar" and "unfamiliar." It says nothing about "the home," and Hall and Van de Castle never made the assertion Garfield unfairly attributes to them. Similarly, Moffitt (1990) says that the Lortie-Lussier, Schwab, and De Koninck findings refute the "biological" underpinnings of quantitative content analysis. He thereby ignores the fact that content analysis is a methodology, while at the same time blurring the distinction between "findings" and "interpretations." Rubenstein (1990:139), in addition to accepting uncritically the mistaken Kramer, Kinney, and Scharf (1983) interpretations of their findings, wrongly claims that Lortie-Lussier, Schwab, and De Koninck found "few indoor settings" for the working mothers, which is exactly the opposite of what they actually report: homemakers had more residential settings, working mothers had more indoor settings.
The false claims by Garfield, Moffitt, and Rubenstein, based on a misunderstanding and misuse of very small differences that could be detected only through an analysis of general patterns, have been a source of major confusion in the literature on dream content. Contrary to their conclusion, the similarities between the two French-Canadian samples, and between the American norms and the overall French-Canadian findings, are far greater than the differences.
Three Latin American Countries
The focus of this section is on dream reports from children, teenagers, and young adults in Argentina, Mexico, and Peru. The dream reports were collected in urban schools in the early 1960s and translated for Hall by college students in these countries whom he hired expressly to gather dream reports on forms he provided. The verbal and written instructions to the students were to write down a recent dream. Each student wrote down one dream. In effect, this was the first use of the Most Recent Dream methodology.
The sample for Argentina consisted of 239 males, ages 12 to 17, and 149 females, ages 18 to 21. The Peruvian sample came from 276 males, ages 12 to 19, and 192 females, ages 13 to 20. There were 178 Mexican males, ages 16 to 24, and 54 Mexican females, ages 18 to 21. The findings with these groups were compared with each other and with those for American teenagers. These findings have not been previously published.
By and large the similarities outweighed the differences. For example, the familiarity percent was above 55 for both males and females in all four countries, but there were always more familiar characters in women's dreams, as can be seen in Table 6.6.
Table 6.6. Familiarity Percents
There were similar findings for other content categories. Aggression was more common in male dream reports than female dream reports in all four national groups, but there was always more aggression than friendliness irrespective of gender. Male dreamers in each of the four countries had more aggressive interactions with other males than female dreamers did, but both genders had more aggressions with male than with female characters.
However, there was one striking difference in these comparisons. Peruvians and Mexicans differed from Argentinians and Americans on the male/female percent. Contrary to the usual European-American pattern, male Peruvians and Mexicans dreamed equally about male and female characters, whereas Peruvian and Mexican women tended to dream more about male characters than female characters (Table 6.7).
Table 6.7. Male/Female Percent
ah = .33 between American and Peruvian males; h = .26 between American and Mexican females.
Urbina and Grey (1975) also found a low male/female percent for men in Peru. They used dream reports collected in Lima from 48 male and 48 female college students 18 to 20 years of age. Each student turned in eight dream reports over a period of two weeks to three months. Urbina and Grey argue that the difference in their findings from our earlier American findings is due to the fact they computed the percentage of male characters for each dreamer, and then obtained the mean of these percentages. They believe this method is superior. In fact, they are mistaken. For group comparisons, it does not make any difference whether the percentage for a group of dream reports or the mean percentage is used. Hall (1984:1110) demonstrated this point empirically with findings for 20 dream reports from each of 20 males and 20 females. The percentages for the 400 male and 400 female dream reports were 68 and 48, respectively. The mean percentages for 20 males and 20 females were 67 and 48, respectively. Contrary to Urbina and Grey, then, we have to conclude that the Peruvian college students may be a rather unique group of dreamers.
The possibility that the male/female percent may be higher for Mexican women is supported by an unpublished study Hall did on 152 dream reports from an adult woman in Mexico City; her figure of 55/45 is below the 61/39 finding for his youthful Mexican sample, but higher than the normative figure of 48/52.
Thus, the findings for Peru and Mexico may indicate a genuine cultural difference on male/female percent. The Argentina findings reinforce this conjecture because Argentine culture, with its large number of German and Italian immigrants, and its minimal number of indigenous peoples in Buenos Aires, where the dreams were collected, is very different from that of most other Spanish-speaking countries in Latin America.
Latin cultures are thought to be more "macho" than North European cultures. This may lead to the speculation that the male/female percent should be higher for males, not lower. But even if we assumed Latin cultures are more "macho," are they necessarily more "patriarchal"? Is there a strong "matriarchal" element in Latin cultures? Whatever the answers to those questions might be, we would not want to build a theory on the basis of the small number of findings we have for Latin American countries. Although the findings seem fairly solid, it would be essential to study more dream reports up and down the age scale in Peru or Mexico before serious inferences could be drawn. The Most Recent Dream approach makes such studies feasible.
There have been three studies of dream reports collected from college students in India, one at Allahabad University in Northern India (Grey and Kalsched, 1971), the other two at Andhra University in Southern India (Prasad, 1982; Bose, 1983). All three make use of the Hall/Van de Castle coding system, and all three report findings similar to those for the United States. There are also findings that may be culturally specific.
The dream reports for the Grey and Kalsched study were collected from 45 female and 51 male college students in 1962-63 when Grey was a Fulbright Senior Lecturer at Allahabad. An adaptation of the Hall/Van de Castle instructions was used on the report forms. The students contributed a range of four to 25 reports for a total of 941 dream narratives. They also filled out a "Traditionalism Index" to see how closely they were tied to orthodox Hindu culture.
The reports were later utilized to test Hall and Domhoff's (1963a) claim of an "ubiquitous" gender difference in dream content on the male/female percent. Grey and Kalsched (1971:344) report very similar findings in Table 6.8.
Table 6.8. Male/Female Percent in India and the United States
ah = .15 for males; h = .12 for females.
In making this comparison Grey and Kalsched use the findings from the older article by Hall and Domhoff instead of from the norms, even though they were using the finalized Hall/Van de Castle (1966) system in coding the Indian dream reports. If they had stayed consistent, the findings would have been even more similar, as shown in Table 6.9.
Table 6.9. Male/Female Percent in India and the United States using the Hall/Van de Castle Norms
ah = .09 for males; h = .04 for females.
The authors agree that their finding "seems to confirm" Hall and Domhoff's (1963a) earlier finding with American, Australian aboriginal, and Hopi dreamers (Grey and Kalsched, 1971:344). However, they stress the fact that Indian males dream even more of males and Indian females even less. They think the difference of seven percentage points for males and six for females is significant statistically and theoretically. They "reject" the alleged "assertion of universality" because they predicted the greater gender segregation in India would lead to the larger differences they found.
Grey and Kalsched go on to report some very interesting findings. First, they confirm that there are individual differences in male percent for both male and female dreamers. Second, students who score high on the "Traditionalism Scale," meaning they are devout Hindus and live in highly gender-segregated milieus, have a lower percentage of opposite-sex characters in their dream reports than students making a transition to more secular identities (r = .37 for males, .51 for females). The finding may not be as solid as it sounds because some of the male percents are based on only four or five dream reports, and may therefore be unreliable, but it is suggestive in the context of other findings in this chapter.
The second Indian study is based on 1000 dreams collected in the years 1968-1970, with exactly ten reports from each of 50 males and 50 females (Prasad, 1982). The study was undertaken as a complete normative comparison with the Hall/Van de Castle norms. The Hall/Van de Castle instructions were followed in collecting the dream reports, and every Hall/Van de Castle coding category was used in analyzing the dream reports. The subjects are described as "graduate" and "postgraduate" students at Andhra University in Southern India. They ranged in age from 18 to 25. Volunteers were recruited by the author by going to different classrooms and explaining the nature of the study. He returned to the classrooms each morning to collect dream reports until each subject turned in ten reports. Over 300 students volunteered to be involved, but many subjects dropped out or were not able to remember ten dreams.
The data were analyzed for differences with the American male and female norms using the z score. No analyses are reported for possible gender differences between Indian males and females. The author reports there were a few significant differences between Indian and American males and females in virtually every general category, but the lack of differences on most comparisons is even more striking to us. Unfortunately, it is not easy to gain a full sense of the findings because no percentages are presented. Table 6.10, on settings, is typical:
Table 6.10. Prasad's Table on Settings
This table means that Indian males and females are more likely to be in familiar settings than American males or females. Americans, on the other hand, are more likely to be in indoor settings. But we know nothing of the magnitude of the differences. Nor do we know if Indian males and females differ in the same way American males and females do in their percentages of indoor and familiar settings. Indians (male and female) dream more of familiar characters, and Americans more of unfamiliar characters. Indians dream more of family members and relatives, whereas Americans dream more of strangers. From the tables there appear to be no differences between the two countries in the male/female percent in either men or women's dream reports. However, we cannot be sure because there are no numbers. Here we see again why it would be helpful to have a conventional format for presenting the results of quantitative dream content studies.
Despite the problems of comparing Prasad's findings with the American norms, it seems likely that he has found a cluster of cultural differences where Indian males and females alike differ in the same way from their American counterparts. Indians are more likely to be in familiar settings, interacting with family members and other relatives. They are more likely to witness aggression than to be involved in it (but there is apparently the same frequency of aggression in their dreams). They are more likely to dream of food and nature. They express more happiness in their dream reports. If these are the main differences, however, as his tables imply, then Prasad also has shown there are great similarities between the dream reports of Indian and American college students.
A second study at Andhra University by V. S. Bose (1983) was designed to compare home reports with laboratory reports collected at different times with different awakening strategies. The larger purpose of this sophisticated investigation was to see if the memory processes by which dream reports are revised could be determined. These larger issues need not concern us here. For our purposes, it is sufficient to compare the findings on home dream reports with the American norms. No overall tables are presented because this was not a normative study, so the findings on home dream reports must be taken from comparisons he made between home reports and the various types of laboratory reports. The study is based on 175 dream reports from 44 male college students between the ages of 19 and 26 who kept a dream diary at home for five days and slept in the laboratory for 11 nights. All of the Hall/Van de Castle coding categories were utilized.
The available comparisons are shown in Table 6.11. As can be seen, the difference between the Bose findings for India and the Hall/Van de Castle findings for the United States are minimal.
Table 6.11. Indian and American Comparisons
Mean number of settings
% of all objects that are architectural
% of all objects that are body parts
% of all objects that are Nature
Mean number of modifiers/dream
Percent of dreams that are sexual
In a later analysis using the 175 home-recalled dream reports from the original study, Bose and Pramilia (1993) made a detailed comparison of their findings with the American norms for all Hall/Van de Castle content categories. Their findings are very similar to those of Prasad's (1982) sample from a decade earlier, providing a historical continuity for Indian college students reminiscent of the University of Richmond, Salem College, and University of California, Berkeley, replications of the original Hall/Van de Castle findings from 30 years before. There were more familiar settings and more familiar characters than in the American studies. The percentage of male characters was slightly higher than the American norms by five percentage points, but the results were not statistically significant at the .05 level of confidence. The only differences on social interactions were a greater number of dreamer-involved sexual interactions on the part of the American males; there also was a tendency that did not quite reach statistical significance at the .05 level for Indian college males to witness the aggressions in their dreams more often than the American males. There were no differences in successes and failures, misfortunes and good fortunes, or emotions.
Indian dreamers showed some differences in the categories of objects, activities, and descriptive elements. For example, they were more likely to dream of food and drink, clothes, and communications. They were more likely to engage in verbal activities and experience location changes, and less likely to have movement. They also had more temporal elements and less mentions of intensity.
Bose and Pramilia (1993:6) conclude that there are few differences in the more "dynamic" aspects of the dream reports of the two samples, by which they mean social interactions, activities, and striving, whereas there are some differences in settings, characters, and types of objects. They further argue that the differences are "integrated, consistent, and meaningfully related" in terms of differences between the two cultures (Bose and Pramilia, 1993:7).
If we look at the preponderance of evidence from the three studies in India, we have to conclude that the dream reports of college students at the Indian universities are very similar to those of American college students. This does not deny there are some interesting cross-national differences, or ignore Grey and Kalsched's (1971) finding of a correlation between Hindu traditionalism and a lower percentage of opposite-sex dream characters. However, it does suggest, in conjunction with previous findings in this chapter, that there are some powerful cross-national variables operating as well. For some purposes the similarities may be useful, for others the differences may be of interest.
We will return to these issues after we consider findings from Japan. The pattern of results in the Japanese study helps us to think about national differences, including those on the male/female percent in Peru and Mexico.
There is a major study of dream reports in Japan by Tadashige Yamanaka, Yusuke Morita, and Junji Matsumoto (1982). It is in effect a replication of home and laboratory comparisons in the United States. The main difference is that the "home" reports were collected in an apartment by watching the eye movements of the sleeping subjects and awakening them during what appeared to be REM periods. Given the intensity of this procedure, perhaps it is not surprising that the high rate of home recall (76%) approached that for the laboratory (86%). Nor is it surprising that under these circumstances there were even fewer differences between "home" and laboratory reports.
The subjects in the non-lab portion of the study were six male and 12 female students at the University of Tokushima in the city of Tokushima. They were 19 to 21 years of age. There were 11 males and 10 females of the same age range and university affiliation in the laboratory portion of the study. Subjects were questioned at each awakening about dream contents using a questionnaire based on the Hall/Van de Castle coding system. Subjects in the "home" study were awakened until at least ten dreams were reported; 220 dream reports were collected under this condition. Each subject in the laboratory portion of the study spent one night in the sleep laboratory; 77 dream reports were collected (Yamanaka, et al., 1982:34-35, 38). Once it was determined that the two samples had few differences, they were merged for purposes of comparison with the Hall/Van de Castle findings.
The researchers coded the dream reports for all Hall/Van de Castle categories except settings and friendliness. We begin with the similarities to the Hall/Van de Castle norms and then turn to the interesting differences. Yamanaka, Morita, and Matsumoto did a detailed analysis of objects and activities in the 104 male and 193 female dream narratives. The similarities with the findings for American college students are almost uncanny; comparisons of the Japanese and American results are presented in Tables 6.12 and 6.13. There are more mentions of food and drink in Japanese reports, and more physical movement like running and playing in American reports, but even these differences are not large. The great similarities in objects and activities mean any differences in characters or social interactions will be all the more interesting.
Table 6.12. Frequencies and Percentages of Objects in Dream Reports From Japanese and American College Students (percents in parentheses)
* p < 0.05 (chi-squared test)
Table 6.13. Frequencies and Percentages of Activities in Dream Reports From Japanese and American College Students (percents in parentheses)
ap < 0.05 (chi-squared test). bp < 0.01 (chi-squared test).
There are more characters in Japanese reports, but women have more characters in their reports than men, as in the United States. There is the possibility that the number of characters should be considered slightly inflated because of the intensive method of immediate awakening by which the non-laboratory dream reports were collected. However, if we take it for a real difference, which makes sense in terms of other findings on characters to be presented in a moment, then we can conjecture that the Japanese live in a more socially dense--"peopled"-- environment than do Americans.
Even fewer of the characters in Japanese dream reports are animals than in American reports. The percentage is less than 1 for both men and women, whereas it is 6 and 4 for American men and women. The Japanese dreamers live in an intensely "human" world, which fits with the fact that there are very few pets in large urban areas. Far more of the characters in Japanese reports are familiar to the dreamer, another way of saying there are fewer "strangers" in Japanese dreams. Here it is interesting to note that the Japanese men have the highest proportion of familiar characters and American men the lowest. Both male and female Japanese dreamers live in an intensely familiar world (Table 6.14).
Table 6.14. Familiarity Percent in Japan and the United States
The "ubiquitous" gender difference on the male/female percent disappears for Japanese dream reports because the women have a pattern that is the mirror image of the men's pattern (Table 6.16).
Table 6.15. Male/Female Percent in Japan and the United States
h = .39 between Japanese and American women.
Thus, Japanese dreamers live in an intensely gender-segregated world. We have found no other society, as will be seen in the remainder of this chapter, where women have such a low male/female percent.
The analysis of aggressions was limited to the percentage of dreams in which an aggression occurred. The percentage of the aggression dreams in which the dreamer was the victim or aggressor also was determined. Both males and females have a very low percentage of dreams with aggressions in them--only 14% for males and 26% for females. The men are aggressors in nine of their 15 aggression dreams (60%); the women are victims in 35 of their 51 aggression dreams (68%). On both measures, the males are more atypical than females in terms of our American norms.
The number of sexual dreams is also very low. The men reported three in 104 dream reports (3% vs. 12% in American males) and the women one in 193 reports (.5% vs. 4% in American females). Here we have to be careful, however. We do not know enough about Japanese culture to be certain that subjects would feel comfortable telling sexual dreams in the "home" condition to the experimenter (always of the same sex as the dreamer). As for the laboratory, we know sexual reports there are rare in both the United States (Domhoff and Kamiya, 1964a; Hall, 1966a; Snyder, 1970) and India (Bose, 1983).
It would be fascinating to know if the unreported patterns of aggressive and friendly interactions in these dream reports relate to the atypical findings presented on characters and aggressions. For now, however, we have more than enough data to be certain that there are great similarities in Japanese and American dream reports (objects and activities) as well as intriguing differences (characters, aggressions) worthy of further study to help us understand the relationship between dreams and culture.
Stepping back for a moment, perhaps the numerous similarities in the cross-national findings presented so far in this chapter are not as surprising as they may seem at first. Most of the subjects are college students. Many of them speak English as a first or second language. All are from urban areas in industrialized, class-stratified societies. They probably have more in common than we may realize if we focus only on "culture."
However, the differences may reflect differences in cultural patterns. There may be higher levels of aggression in American dream reports. There seem to be more familiar characters in the reports of Indian and Japanese college students. Most of all, the low male/female percent in Japanese females catches our interest. Does this relate to the extremely strong male dominance in Japan? If so, what do we make of the national cultures at the other extreme on male/female percent, Mexico and Peru? A large sample of Most Recent Dreams collected in those and other nations in the same time period might be a way to answer these questions.
Perhaps further light can be shed on these issues in comparisons with pre-industrial peoples who live in small bands, tribes, and villages, and often hunt and fish for much of their livelihood. Several such comparisons will be made in the remainder of the chapter.
Hall's Yir Yoront and Hopi Study
We begin this section with a consideration of dream reports from the Yir Yoront of Australia and the Hopi of the American Southwest because they provide two of the best dream collections available on small-scale societies. In an unpublished study, Hall (1964) compared the dream reports from these two cultures with each other and with findings on American college students.
The dreams from the Yir Yoront, one of many distinctive aboriginal societies in Australia, were collected in the 1930s as part of field work by Lauriston Sharp (1934, 1939, 1952), who was intimately knowledgeable about this culture. The Yir Yoront put no special emphasis on dreams, and make no use of them in ceremonies. Nor were dreams anything but a peripheral part of Sharp's work (e.g., Sharp, 1969). According to David Schneider (1969:15), who was responsible for giving the dream collection to Hall and later publishing it, Sharp collected them "primarily to bring up new ethnographic material, and to establish and maintain rapport. Dreams were a neutral subject that could be exploited without offending anyone; no one seemed to care enough about them to get upset."
The Hopi dream reports were collected by Dorothy Eggan (1949, 1952, 1961, 1966), who had a strong interest in dreams as part of her focus on the relationship between culture and personality. She had a detailed knowledge of Hopi culture.
The Yir Yoront collection consists of 140 dream reports from 43 men and ten reports from eight women. One man contributed 14, another 11, and the rest from one to five each. The Hopi collection came from four men who contributed 24, 18, 10, and six dream reports for a total of 58, and three women who contributed 60, 18, and 18 dream reports for a total of 124. There is every reason to believe these are good reports due to the familiarity of the anthropologists with the cultures and their close relations with their informants.
Schneider (1969) did his own quantitative analysis of the Yir Yoront dreams. He found sexual content in 19 of the 140 male dreams, mostly with younger women as sexual partners, and only one with the person's wife as the sexual partner. Reports of sexual activity with women in prohibited kinship categories more often include interruptions and unsatisfactory outcomes (Schneider, 1969:22-23). Thirty-nine male dreams contain 73 acts of human aggression. The dreamer witnesses 18 of these aggressions, initiates 14 of them, and is a victim of 41 of them (Schneider, 1969:26-27). In our terms, the victimization percent is very high. Almost all of the aggression (68 of the 73 instances) was directed against adult males.
Eggan (1949, 1952) did not do group or quantitative analyses of the dream reports she collected. Instead, she focused on themes in an individual's dream series and showed how these themes related to the person's personality and cultural beliefs.
Hall's unpublished study comparing Yir Yoront and Hopi dream reports to those from American college students tested the hypothesis that the similarities in content categories for the three different groups would be greater than differences to a statistically significant degree. He then compared American, Hopi, and Yir Yoront males to American and Hopi females. The Yir Yoront females were not included because there were only ten dream reports from them.
Hall used direction of difference between categories rather than magnitude of difference to test his hypothesis. That is, he determined whether the difference between content category A and content category B for each group was systematically positive or negative. To take a specific example, he determined the male percent and compared it to a female percent, as in Table 6.17.
Table 6.16. Male and Female Percents for Yir Yoronts, Hopis, and Americans
Yir Yoront Males
Category one (M/M+F)
Category two (F/M+F)
Direction of difference (one minus two)
In this example, four of the five difference codes are in the same direction. Hall used this method for this cross-cultural study because he believed the magnitude of scores in each content category could be affected by many factors not present in within-culture studies. These include possible sampling problems, the age and gender of the anthropologist, and the type of relationship between the anthropologist and the dream reporters.
This approach allows for a very rigorous statistical test of the hypothesis using a theoretical distribution of difference scores obtained from the expansion of the binomial theorem. Since there are five groups of dream reports being compared, there are five differences in direction for each pair of content categories. If all five comparisons are positive, the difference score is +5. If four of the five are positive, the difference score is +4; if three are positive, +3; if two are positive, +2; if one is positive, +1; and if none is positive, 0. However, for purposes of statistical analysis, the +2 is changed to -3 so it can be grouped with +3, and the +1 is changed to -4 and the zero to -5 so they can be grouped with +4 and +5. There are then three types of scores: +5, -5; +4, -4; +3, -3. In order to test the hypothesis of greater similarity than difference, it was assumed that the empirical distribution of difference scores will differ significantly from the theoretical distribution. Specifically, there should be a preponderance of +5's and -5's if the hypothesis is correct.
Hall developed eight tables containing comparisons involving various proportions using the coding categories for characters, aggressions, friendliness, misfortunes, and good fortunes. There are many more similarities between the three groups (and males and females) than there are differences. These similarities manifest themselves despite the fact that there are such uncontrolled variables as age, number of dream reports per informant, and conditions under which the dream reports were collected. Hall concluded on the basis of his tests of statistical significance that the hypothesis was supported.
Perhaps these findings can be made more concrete by focusing on the few differences among the three societies. Most of the differences occur in the Yir Yoronts. Yir Yoront males dream more about animals, have a higher proportion of aggression with animals, a very high percentage of physical aggressions, and a higher frequency of friendly encounters with humans, particularly with familiar female characters. Much of the friendliness with others consists of the dreamers sharing meat from the animals they have killed.
These findings suggest a correspondence between the dream life and waking life of Yir Yoront males. They do hunt and kill animals far more than Hopi or Americans, who grow their food. They do a great deal of individual squabbling among themselves, and they do share their kills with relatives and friends. This correspondence between dream life and waking life is similar to what we found in the cross-nation comparisons, only on different dream elements. Both types of differences are evidence that there is meaning in dreams.
The only differentiating feature of Hopi male dream reports is a larger number of aggressive encounters with females, for which we have no ready cultural explanation. There are no major deviations for Hopi females or American males and females from the overall findings.
We think the findings of this rigorous study with three good dream collections are impressive evidence the Hall/Van de Castle coding system is applicable to dream reports from anywhere in the world and may pick up cultural differences as well. The biggest drawback to the investigation is the limited number of cultural groups employed. The groups come from different parts of the world and have radically different ways of life, but the fact remains that there are only three of them.
Hall's Larger Cross-Cultural Study
In order to expand the scope of the previous investigation, Hall did a second unpublished study using dream reports from a wider range of societies. Most of the dream collections were provided to him by Schneider, who gathered them from a variety of anthropological sources.
Once again, Schneider (1969) did his own analysis using dream reports from the following societies: Tinguian, Hopi, Yir Yoront, Alor, Ifaluk, Baiga, Navaho, Kwakiutl, Truk, Buku, Hopi-American, and Kiwai. He wrote (1969:55) that his study "seems to support the contention that (1) there are certain regularities in the manifest content of groups of dreams regardless of the society and culture of the dreamer, and (2) there are certain differences between groups of dreams that seem to be a function of the culture of the dreamers." For example, in all groups the dreamer is more often the victim than the aggressor in aggressive interactions, but there are cultural differences within that general context (Schneider, 1969:56). Schneider's method and findings are supportive of the Hall study of Yir Yoront, Hopi and American dreamers.
Hall supplemented the Schneider collection with dream reports obtained from other sources. He then made dozens of comparisons on characters, aggressions, friendliness, and misfortunes. Once again, he found there were more similarities than differences across groups for any given content category, but a few groups show differences.
One of the most consistent similarities concerned the male/female percent. There was no male group where the male percent was below 59 and no female group where it was below 48. There was one small female set of dream reports where the male/female percent was an unusually high 66. All of Hall's findings on this issue in small-scale societies, including the Hopi and Yir Yoront, are brought together in Table 6.17.
Table 6.17. Cross-Cultural Findings on Male/Female Percent
No. of dreams
Male/ female %
# of dreams
Male/ female %
Hall's main findings for all the five small-scale groups where he had both male and female dream reports are brought together in Table 6.18. There is some variability from culture to culture, but for us the similarities are striking. There are always more single than plural characters, more humans than animals, and more familiar than unfamiliar characters. The A/C index is higher than the F/C index with one exception (the Hopi). Dreamers everywhere are more often victims of aggression with two exceptions, one of which is not shown in the table (Yir Yoront), and there is usually more physical than nonphysical aggression.
Table 6.18. Cross-Cultural Findings with Five Small-Scale Societies
Female findings (Males in parentheses)
No. of subjects
No. of dream reports
Percent single characters
Overall A/C index
A/C index with animals
Physical aggression percent
Overall F/C index
Percent sexual reports
Two findings concerning the animal percent are noteworthy. First, it ranges from a high of 34 and 31 for Biaga and Yir Yoront males, not shown in Table 6.18, to 06 for Hopi females. Second, it is always higher for small-scale societies than it is for large-scale ones, although Hopi females (06) come close to our female norms (04). Since these small-scale societies are in greater proximity to animals and unedited nature than the urbanized populations from large-scale societies for whom we have dream reports, we think this difference in animal percent is a cultural difference revealing a relationship between dreams and waking life.
Unfortunately, it is difficult to do much more with these findings because we do not have enough information to explain the deviant cases. The deviations may be due to cultural differences, but it is risky to make such inferences when the sample sizes are small and so little is known about the representativeness of the samples and the methods of collection. Moreover, we do not know enough about the cultures in many cases. In short, these findings take us just about as far as we can go without much more information on the dreamers and the cultures. We need to turn elsewhere for further enlightenment on small-scale groups.
Happily, there are a few investigations of small-scale societies allowing us to see if dream reports and culture are closely linked in them. In particular, Thomas Gregor's (1977, 1986) detailed ethnographic studies of the minuscule Mehinaku Indian village of about 80-85 people in the Amazon rain forests of central Brazil afford us a singularly in-depth look at the dream reports of pre-industrial men and women known personally to the dream collector. Moreover, Mehinaku life in the Xingu National Park preserve where they hunt, fish, and grow a few staples was essentially traditional when Gregor was there: "The presence of an Indian post and a small air force base [several hours from the village] has had a psychological impact that is visible in Mehinaku dream life, but the villagers thus far remain protected from wage labor, contact with missionaries, squatters, and others who would exploit them or alter their lives" (Gregor, 1981a:354).
Gregor found in the course of his general research that the Mehinaku like to remember their dreams and often tell them to their families. They believe, like many preliterate cultures all over the world, that dreams occur when the soul "leaves its home in the iris of the eye to wander about through a nocturnal world peopled by spirits, monsters, and the souls of other sleeping villagers" (Gregor, 1981a:354; see Gregor, 1981b, for a full statement of Mehinaku dream theory). They also believe that dreams can be a clue to the future. Thus, Gregor found it very easy to collect dream reports from them.
Gregor received at least a few narratives from 18 men and 18 women, most of the adult population of the village. The men reported 276 dreams and the women 109, with 25% of the 385 reports coming from two young fathers (ages 22 and 23) who were chosen for intensive study. Their dream content is sufficiently like that of other males in the sample that including them changes overall percentages by less than 5%. Gregor coded the dream reports for physical aggressions and sexuality with categories fairly similar to those in the Hall/Van de Castle system. He also coded the dream reports for the degree of activity or passivity, as determined by whether the dreamer initiated actions (active) or only observed or reacted to the actions of other dream characters (passive).
Gregor found there is more physical aggression in Mehinaku than in American dream reports, especially with animals, but the gender differences are the same in that men report more physical aggressions than women. Mehinaku are more likely than Americans to be victims of aggression. As with our American norms, Mehinaku women are more likely than Mehinaku men to be victims. Men and animals are the most frequent attackers in Mehinaku dream reports, but spirits and monsters sometimes attack as well (Gregor, 1981a:384-385). Sexual elements appeared in 13% of the men's reports and 12% of the women's. The male percentage is virtually identical with the Hall/Van de Castle norms, but the female percentage is three times higher. In men's sexual dreams there is sometimes a fear of assault by jealous male rivals or angry women lovers. Women's dream reports often show a fear of rape and other violent encounters with sexually aggressive men (Gregor, 1981a:388). Sixty-one percent of the men's dream reports were judged as active and 39% as passive. By way of contrast, 42% of women's dream reports were judged as active and 58% as passive.
Gregor believes his overall findings make sense in terms of Mehinaku culture. There is threat from a wide range of animals and insects, and there is conflict and competition among the men. Most men and women have several lovers as well as a spouse, and women do fear sexual pressures and attacks from men (Gregor, 1986). Moreover, "the relative passivity of women's dreams corresponds to women's position in daily life" (Gregor, 1981a:387). The society is highly patriarchal, and men dominate in the home as well as in the religious and public aspects of village life. At the same time, Gregor notes the similarities with the Hall/Van de Castle findings and emphasizes that the gender differences are of the same nature. He suggests that "with additional cross-cultural data it may be possible to show that the dream experience is less variant than other aspects of culture" (Gregor, 1981a:389).
Gregor's study is an outstanding addition to our knowledge of dream content because of his excellent rapport with his subjects and intimate knowledge of their culture. It is impressive independent verification of the generalizability of the Hall cross-cultural findings to other dream collections from small-scale societies.
The Zapotecs of Mexico
O'Nell and O'Nell do not speculate on the reasons for the cultural differences. However, we note that the Zapotec's higher levels of physical aggressions are similar to what Gregor (1981a) found for the Mehinaku Indian village in Brazil.
There were three main similarities between American and Zapotec dreamers:
Males have more aggression in their dreams than females (A/C index);
In both cultures, people are more likely to be victims than aggressors in aggressive interactions;
Both male and female dreamers have more of their aggressive interactions with male characters than female characters.
There were three main differences:
Zapotec dreamers, male and female, had a higher physical aggression percent than did their American counterparts;
Zapotec dreamers had more dreamer-involved aggressions than the American dreamers;
There was little or no tendency for the A/C index to decline with age for the Zapotecs.
O'Nell and O'Nell do not speculate on the reasons for the cultural differences. However, we note that the Zapotec's higher levels of physical aggressions are similar to what Gregor (1981a) found for the Mehinaku Indian village in Brazil.
The Gusii, Kipsigis, and Logoli of East Africa
In 1967, while members of the Child Research Unit at the University of Nairobi in Kenya, the American researchers Robert and Ruth Munroe (1977) did a day of testing and interviewing in English in both a male and female secondary boarding school. Among other measures, they requested a recent dream from the students.
The students came from three male-dominant societies of East Africa: the Gusii, the Kipsigis, and Logoli. Munroe and Munroe (1977:150) emphasize that "the social-structural importance of males, which is based on patrilineal descent and patrilocal residence, is vastly greater than that of females." They therefore predicted there would be a predominance of males in all dream reports they collected.
The Munroes collected 325 dream reports, 182 from males and 143 from females. The breakdown by societal groups is not provided. They coded the dream reports by classifying each report as to whether there was a "preponderance" of male or female characters. Independent judges had a 95% level of agreement on a sample of 40 reports. The frequency of male-character reports among the Gusii was 82% for males and 61% for females. The corresponding figures for the Kipsigis were 82% and 44%, and for the Logoli, 83% and 56% (Munroe and Munroe, 1977:150). The differences were statistically significant using chi square. While this is not an ideal type of analysis from our point of view due to its failure to use total characters as the unit of analysis, it does seem to show the ubiquitous gender difference for three more small societies.
The dream reports were then used in a later investigation to make explicit comparisons with some of the gender differences reported in the Hall/Van de Castle norms. The results were similar to the American findings on four of five comparisons.
The young men dreamed more frequently of outdoor settings, the young women of indoor settings;
male dream reports more frequently mentioned implements;
female dream reports more frequently mentioned clothing; and
female dream reports contained more mentions of emotions.
The results were statistically significant at the level of .05 or lower using a chi square test. In addition, within-society trends were "in the same direction as the overall findings except that Logoli men mentioned clothing in a slightly higher proportion of their dreams than did Logoli women" (Munroe et al., 1985:406).
However, there were differences in aggressions. The East African women students had more dreams with physical aggressions than American college females, and equaled the East African men students in this regard. The East African women also differed from the Hall/Van de Castle norms in having more nonphysical aggression in their dream reports than the young men.
The reasons for the differences from the Hall/Van de Castle norms on aggressions were explored in a later investigation using 187 male dream reports and 149 female reports, for a total of 336 dream narratives. This investigation concluded that "the relatively high frequency of aggression in women's dreams seems due in part to preoccupation with being victimized" (Munroe et al., 1989:728). The detailed investigation of aggression is also noteworthy for a further difference with the American norms that was discovered. Whereas an animal character initiated the aggression suffered by the dreamer in 9% of American dream reports, it did so in 44% of the East African reports. There was no gender difference. This finding is similar to that of Gregor (1981a) on animals as aggressors in Mehinaku dream reports.
Munroe and Munroe (1992) also studied friendliness in the same collection of dream reports. They found female dream reports contained more friendly interactions than those of males, and approximately equal amounts of friendliness with male and female characters. The male dreamers had more friendly interactions with male characters than female ones, the opposite of the American norms. The Munroes suggest that this male pattern "may reflect the strong male bias in social organizations and groupings" (1992:402).
The four analyses of dream reports from East African students do not have the depth of a study like Gregor's, based as they are on one day of dream collecting, and the subjects are clearly acculturating if they speak English and attend schools. Nevertheless, these analyses add range to the findings reported earlier in this chapter.
Gusii Women of Kenya
As part of a study of childhood in Gusii communities in southwestern Kenya in the 1970s by Robert and Sarah LeVine, Sarah LeVine collected 88 dream reports from 22 women. Those reports have been analyzed by both LeVine (1982) and Linda Kilner (1988). Kilner used Hall/Van de Castle coding categories. We will combine findings from the two studies.
The women ranged in age from 18 to 45. All were married and all but one had from one to 11 children; six of the 22 were married to polygamists. None of the women over age 28 ever attended school; all but one of the younger women attended elementary school and two attended high school. All the women had agricultural as well as domestic duties.
Dreams are private matters for Gusii. They are not due to spirits or external forces. They are not told to healers. Although the women were open to reporting their dreams to LeVine, they were not interested in discussing their meaning with her (LeVine, 1982:66-68). The dream reports were very brief, averaging only three sentences in length. The longest is nine lines. Three women provided ten or more dreams apiece.
LeVine looked at settings, characters, aggressions and misfortunes, labor and childbirth, material objects, and the supernatural. We will concentrate on her findings on characters, aggressions, and misfortunes because they have the most overlap with our categories. Kilner used Hall/Van de Castle categories to code for friendliness, sexual interactions, misfortunes, good fortunes, success, and failure. Kilner compared her findings to the Hall/Van de Castle norms and to her own sample of 75 dream reports from an unstated number of adult American professional women who were friends and friends of friends.
LeVine found the women dreamed most frequently about their children (19), their parents (9), and their husbands (8). They dreamed about the women they grew up with, such as mothers, sisters, and aunts, but rarely the female in-laws with whom they have daily contact. Conversely, they dreamed more often of males from their marital home--husband, his brother, and his male cousins--than their own father and brothers (LeVine, 1982:69). Only rarely did they dream of unfamiliar characters.
Sixty-six of the 88 reports (75%) contained job loss, accident, illness, the threat of violence, actual violence, murder, or death; that is, they are full of aggressions and misfortunes. In only two reports is the dreamer the aggressor (LeVine, 1982:70). This finding of high victimization replicates the third of the Munroe and Munroe analyses for this group discussed in the previous section.
The Gusii women had less dream reports with at least one friendliness (33%) than the Hall/Van de Castle norms (42%). They were more likely to be befriended than to initiate friendly interactions when compared to American women. There was sexual activity in two of 88 dream reports (2%) compared to 4% of the normative sample. Gusii women dreamed less often of success or failure, meaning there was very little striving in their dream reports. When they did dream of success, it was more likely to be achieved by some other dream character. Gusii women dreamed twice as often about misfortunes as the Hall/Van de Castle norms (Kilner, 1988:82-83).
Overall, the dream reports of the Gusii women are dominated by a sense of fear, submissiveness, and misfortune. Both LeVine (1982:75) and Kilner (1988:20) argue that the contents of the dream reports reflect the dire life circumstances of women in Gusii culture. As we stated in discussing the Munroe and Munroe studies, the society is highly patriarchal. Women are not respected until they have succeeded as mothers. They leave home to marry a man from an enemy clan:
The man pays bridewealth to her father. She is believed to be devoid of good judgment and incapable of undertaking responsibility of any sort other than child care and manual labor. She wins acceptance through compliance with elders, industry in agricultural tasks for her mother-in-law, and most especially through bearing children. The Gusii look upon their children as economic benefits. As a woman's children grow, especially if the children are male and already circumcised, so does the influence and acceptance of the mother (Kilner, 1988:80).
The interpretations by LeVine and Kilner are very plausible. To make them more convincing, it would be good to have dream reports from comparable adult Gusii males to be sure the differences from American women are not cultural ones shared with males (e.g., low striving, high misfortunes). It also would be good to have a larger sample of women and dream reports. Nonetheless, the findings of these two studies are provocative in terms of the possible continuities between dream content and cultural patterns.
A Missed Opportunity
Julia Levine (1991) did an exceptional job of gathering cross-cultural data that could have led to an excellent comparison of children's dreams in the United States, rural Ireland, an Israeli kibbutz, and a Bedouin tribal group in Israel. Sadly, she has missed this opportunity by using an idiosyncratic and subjective rating system for such dimensions as "reality of dream," "self representation," and "humanness of participants." She did not compare her results with more numerous American findings.
Moreover, she takes a totally uncritical attitude toward poorly done studies of the alleged role of culture in shaping dream content (e.g., Grey and Kalsched, 1971; Roll, Hinton, and Glazer, 1974; Roll and Brenneis, 1975). She goes so far as to claim that questionnaire responses by Japanese and American students on whether they think they have had such common dreams as falling, being chased, or losing teeth (Griffith, Miyago, and Tago, 1958) show the role of "value systems" in shaping dream content; she thereby overlooks the fact that thinking/remembering you once had a dream about something or another is not a study of dream content (Levine, 1991:473).
Levine collected three to five dream reports from each of 26 children from a seminomadic Bedouin settlement in Israel, 24 Israeli children on a large rural Kibbutz, and 27 Irish children from a small fishing village in southwestern Ireland. All the children were between 8.5 and 11 years of age. They were enlisted voluntarily from the third, fourth, and fifth grades of their local elementary school. Levine does not say how many dream reports she collected, but the average was five per Bedouin and Irish child, and 3.5 per Israeli child. The children were asked to write down their dreams in the morning, and the dreams were then told each morning for seven days to Levine and, where necessary, an accompanying translator. The dream reports were tape recorded and later transcribed.
Levine employed three students to code the dream reports on the scales she developed. She trained them to what she describes as a high rate of intercoder reliability on 17 reports, then had each of the three rate one-third of the reports. For the most part, the scales she developed, except for the one on the "reality" of the dream, duplicate the usual content categories found in our system. There are human and nonhuman characters, familiar and unfamiliar characters, activities, and positive and negative emotions. Within her rating system on conflictual dreams, reflecting the usual problems of rating systems, the witnessing of an aggression receives as low a score as being the victim of one. Thus, in one of the sample dream narratives provided by Levine (1991:477), where the dreamer saw his or her father shoot a wolf, the dreamer would receive a low score on "self-representation."
Nonetheless, Levine's findings mostly fit with what we have reported already in this chapter. First, there were twice as many "conflictual" as opposed to "nonconflictual" dream reports in all three cultures (Levine, 1991:479), which seems to parallel the predominance of aggression and misfortune over friendliness and good fortune in most of our samples. Second, dreamers were more "passive and uninvolved" in conflictual dreams, and the "hedonic tone" was more negative in these dreams; here she may well be replicating the well-known findings on victimization in aggressive dreams in most societies.
Levine then turns to cross-cultural differences in conflictual dream reports because she has detected few differences in the one-third of the sample that are nonconflictual. In general, the Bedouin children differed from the Irish and kibbutz children, which fits with our findings in this chapter on differences between citizens of nations and members of small tribal groups. For example, there were more "nonhuman characters" (presumably animals for the most part) and familiar characters in the Bedouin children's reports (Levine, 1991:483). She is describing very typical findings in the dream content literature (e.g., O'Nell and O'Nell, 1977; Gregor, 1981a; Van de Castle, 1983), none of which she cites.
Levine (1991:485) starts her discussion section with the claim that she has "replicated" previous research, meaning Grey and Kalsched (1971), and Roll and his associates, in demonstrating "the relevance of culture in the analysis of manifest dream reports." She then makes several plausible glosses of her findings in terms of cultural differences among the three groups, but in the conclusions section she rightly says that "this study lacks data regarding the waking life of the subjects that could substantiate the relationship of culture to the systematic differences found in dream reports" (Levine, 1991:488). In all, then, this is another lost opportunity in terms of developing cumulative findings in the study of dream content.
We have six general conclusions:
1. There is nothing in the dream reports collected in other societies that cannot be encompassed within the Hall/Van de Castle coding system. Mundane though it may seem, this is probably the most gratifying conclusion from this cross-cultural work. The fact that independent investigators in Europe, India, and Japan found the system useful is part of our evidence for this conclusion.
2. Several of our percentages and indexes provide similar results in most of the cross-cultural samples of dream reports studied to date. There is more aggression than friendliness, for example, and more misfortune than good fortune. Dreamers everywhere are usually more often victims than aggressors in their dreams.
3. There are several cross-cultural gender differences. Men dream more often of other males, women tend to dream equally about males and females. There usually are more familiar characters in women's dreams. Men usually have more physical aggressions in their dreams, and women are more often victims of aggression.
4. Despite the many commonalities of dream reports across societies on many of our coding categories, there is enough variability that the American norms must be used with extreme caution beyond Anglophone Canada and Western Europe. Despite our comments earlier about the fundamental commonalities in all large industrialized societies, there are too many differences with French Canada, Latin America, India, and Japan to use the norms in studying individuals from those countries. Instead, new norms should be created for those countries. In terms of aggression scores, there may be important differences with Europe and Anglophone Canada as well.
5. The variations in dream content from culture to culture seem to relate to unique cultural patterns in the few cases where we have enough information to hazard an opinion. The variations in animal percent certainly support that conclusion, as do the findings on such distinctive and carefully studied groups as the Yir Yoront and Mehinaku. In other words, the differences found among cultures are a further reason to believe that there is meaning in dreams because these differences "make sense" in terms of what is known about the cultures.
6. Finally, we reluctantly conclude that the problems of doing good studies of dream reports from small-scale societies are so great that the understanding of dream meaning is unlikely to be advanced much further by such studies. The studies presented in the last section of this chapter have taken us about as far as we can go in this area. The assembled studies have shown us cross-cultural similarities and differences, giving us confidence in our coding system and reinforcing the claim that there is meaning in dreams. However, the hopes of finding highly revealing and distinctive dream content once held by those who studied culture and personality have not been realized. Moreover, there are too many difficulties standing in the way of convincing studies with small-scale studies, except in rare cases like Gregor's study of the Mehinaku. Those difficulties include mutual suspicions, power differentials, establishing rapport, cutting through stereotyped responses based on cultural beliefs, and understanding linguistic and cultural subtleties. Dream research in small-scale societies is probably best left to those who want to study people's folk beliefs and practices relating to dreams.
If the goal is to understand the meaning of dreams, as it is in our case, then it is probably better to study adults who share the same language and culture as the researcher, although the Most Recent Dream methodology may be a promising approach to cross-national studies. Either way, within nations or cross-national, the future of dream content research lies primarily in studies of citizens in large-scale industrialized societies. We turn to such citizens in the remainder of this book, reassured by cross-cultural studies that we have a powerful and universalistic coding system.
Tedlock (1991:162), in an article suggesting content studies are outmoded in anthropology and should be replaced by studies of how dreams are understood and used in different cultures, claims cross-cultural content studies by people such as Hall (1951, 1953c) are based on dream reports collected in an inadequate fashion. However, Hall used only American data in the publications she cites, and never made more than a few passing comments in print on cross-cultural findings. His cross-cultural findings never have been published until now. It is not clear from Tedlock's paper just why she made these inaccurate gratuitous comments about Hall's work.
Although its findings are in the expected direction, the brief report by Robbins and Kilbride (1971) on one or two dreams from each of 50 male and 44 female Bantu primary and secondary school students in Uganda is not substantial enough to be discussed here. The authors asked the students to "describe the one dream you remember the most," which is likely to generate atypical reports. Furthermore, the results are not statistically significant. Hall later obtained these reports and scored them for male/female percent. The percentage of male characters in 53 male reports was 82; the percentage in 43 female reports was 43 (Hall, 1984:1111). Both figures, of course, are typical of what we usually have found elsewhere.