New Maps of Science Fiction|
by William Sims Bainbridge
and Murray M. Dalziel
Analog Yearbook, 1977, pages 277-299.
Many science fiction stories are written about social conflict, social-scientific speculations, and specific ideas taken from the social sciences, yet very few of the factual articles that appear in SF magazines are about the social sciences. This article will be an exception. We will describe some of the most basic and most modern techniques of sociological survey analysis, demonstrating them with data provided by science fiction fans, arriving at a scientific sketch of science fiction literature.
It is strange that science fiction literary criticism is entirely unscientific. Anybody can read a book and give a personal opinion on it. A few readers write intelligent book reviews for magazines like Analog. The best of them bring great writing skill and considerable experience in SF to their job, but none use science. We do.
The nearest thing to true scientific literary research on SF was probably the monthly Analog feature "The Analytical Laboratory," which ended with the October 1976 issue. For more than twenty years Analog tabulated reader preferences for the stories in each issue and awarded bonuses to the authors that scored highest. The first question that naturally comes to mind about stories and authors is "How much do you like them?" Literary critics try to go far beyond this simple query, but it is the one that people ordinarily care most about, and for us it is the most important sociological question. Using modern techniques of analysis we can recover a tremendous amount of hidden information from statistics of people's likes and dislikes. Perhaps the basic trick in survey research is how to ask questions without asking them. A corollary of that is the ability to ask many questions while only asking a few. Those sound like impossible stunts, but they are really quite easy, given the appropriate methods. We want to construct maps of SF literature, identifying types, describing them, and outlining the ways they fit together. If we simply asked people to draw such maps, we would wind up with an unintelligible mess. We couldn't fit their different ideas together! But if we ask people a few questions about themselves and their general opinions, and have them rate authors in terms of how much they like them, then we have all the data we need to construct clear maps.
Two years ago, when we began this project at Harvard one of us was already an avid science fiction reader. But we realized we needed broad knowledge of science fiction culture before we would know what questions to ask. We attended six major science fiction conventions, joined clubs, read books, conducted over a hundred interviews, and sent a preliminary questionnaire to members of The New England Science Fiction Association. This gave us the necessary background. Then we produced our main questionnaire, including 112 questions asking respondents their opinions of a variety of authors and types of fiction, seeking information about their involvement in SF fan activities, and probing for a general picture of life-style. We mailed copies to editors of science fiction amateur magazines (fanzines) and their associates. One hundred thirty of them were kind enough to return completed questionnaires. With the assistance of 130 expert SF readers we were ready to do collective literary criticism in a scientific manner.
That may sound less than exciting. Questionnaires sent through the U.S. Mail cannot compete for drama with Apollo rockets fired toward the Moon. Science is often mundane. We hope, however, you will find our results interesting! Some of the material was used in the science fiction chapter of Dr. Bainbridge's new book, The Spaceflight Revolution, a study of the social movement that produced modern space rocketry, but the data reported here have never been published before.
Half of the questions focused on a selected list of twenty seven well-known authors. Twenty were simply the authors who won Hugo awards for fiction in the ten years 1963-1972. We added classic writers Jules Verne, H. G. Wells, and Edgar Rice Burroughs to get a historical dimension. J. R. R. Tolkien and H. P. Lovecraft were put in to extend the list in the direction of fantasy John W Campbell, Jr. was included because he was such a towering influence for so long. Because of our interest in the history of spaceflight, and because of his popularity, we included Arthur C. Clarke; as it happened, he received a Hugo award after our 1972 cutoff date.
One list of twenty-seven questions asked the 130 fan respondents how much they liked each author. After each name in the list, the fan was supposed to circle one of seven numbers, from -3 through 0 to + 3 indicating how he felt. Plus 3 meant "like strongly;" -1 was "dislike slightly;" -2 meant "dislike somewhat;" and -3 was a dismal "dislike strongly."
Before we could analyze the 27 x 130 = 3,510 bytes of data generated by these questions, we had to punch the questionnaire responses onto IBM cards and feed them to a computer. We use the machine to do tedious calculations for us. The obvious first step was to take a look at the average scores. Figure 1 shows the list of twenty-seven authors, arranged from the best-liked to least-liked, with the average score for each and the percent of the 130 fans who liked each author "strongly." Surprise, surprise! Isaac Asimov wins.
|1. Isaac Asimov||2.208||50.4%|
|2. Arthur C. Clarke||2.128||47.2%|
|3. Robert A. Heinlein||2.024||46.3%|
|4. Larry Niven||1.992||37.7%|
|5. Poul Anderson||1.967||41.3%|
|6. Theodore Sturgeon||1.960||40.3%|
|7. Roger Zelazny||1.924||42.0%|
|8. Fritz Leiber||1.893||34.7%|
|9. J. R. R. Tolkien||1.861||53.5%|
|10. Ursula K. LeGuin||1.809||39.1%|
|11. Robert Silverberg||1.680||31.2%|
|12. Clifford D. Simak||1.675||25.8%|
|13. Jack Vance||1.552||25.0%|
|14. Harlan Ellison||1.443||30.3%|
|15. Frank Herbert||1.395||13.4%|
|16. John Brunner||1.383||26.7%|
|17. Philip Jose Farmer||1.373||26.3%|
|18. Anne McCaffrey||1.345||18.5%|
|19. Samuel R. Delany||1.308||26.5%|
|20. Gordon R. Dickson||1.248||15.4%|
|21. Philip K. Dick||1.169||23.7%|
|22. H.G. Wells||1.122||17.9%|
|23. R. A. Lafferty||1.119||23.7%|
|24. John W. Campbell||1.042||11.9%|
|25. Jules Verne||0.659||10.6%|
|26. Edgar R. Burroughs||0.587||16.5%|
|27. H. P. Lovecraft||0.574||19.1%|
All the authors can take heart that their averages came out on the plus side. No author was disliked by the majority. The percentages of the 130 that gave a top rating to each author allow us to spot variations in how concentrated authors' fans are. Tolkien got the highest percentage of "like strongly" responses, 53.5%, even higher than Asimov's 50.4%. But Tolkien still fell near the middle of the list on average score. This means that Tolkien is liked very much by half the fans, but liked little or even disliked by the other half. There is great disagreement about him. Frank Herbert shows the opposite pattern. Hardly anybody is fanatically enthusiastic about his work, but most like it pretty much. There is little disagreement about him.
Agreement and disagreement are the meat of what survey analysis is about. Pollsters would have a very boring job if everybody always agreed, and sociologists would face an impossible task if disagreement were entirely random, showing no meaningful social patterns. We want to find patterns of disagreement. One of the tricks of survey analysis is finding agreement in disagreement: groups of individuals who seem to agree with each other while disagreeing with others. Social scientists working in this tradition continually search for the opportunity to make statements like, "People who agree about x also agree about y. " Or, "People who disagree with w tend to be like z. " Linking such observations into general theories of human behavior is one of the highest aims of social science.
The scores for the authors may be of interest to some fans and to Isaac Asimov, but we haven't gotten very scientific or controversial yet. Let's put one more question in the computer to thicken the plot. We asked the following simple question to uncover the political orientations of the 130 respondents:
- In general, how would you classify yourself with respect to current political issues? (please check just one)
- [ ] Radical
- [ ] Liberal
- [ ] Moderate
- [ ] Conservative
This looks like a dumb question. People often scribble nasty comments in the margins of our questionnaire next to it, complaining they can't squeeze themselves into just one box, or informing us that we are idiots for including such obvious junk. Sorry folks, but this is a great question! The problem is that people imagine we care about them as individuals, are trying to make friends with them or to capture their immortal souls in our questionnaires. Not so. We are looking for abstract social scientific relationships between variables. This is the same thing physicists do. They don't care about the personalities of the electrons they count or the souls of the experimental equipment they use. The point is, our political question works, and has been used successfully many times in other research. Even if it does not capture the complex range of political opinion, it does divide a large group like our 130 into significantly different camps.
About 15% of our sample were Radicals, 45% Liberals, 27% Moderates, and 13% Conservatives. Interesting things start to happen when we run this question back against the twenty-seven author questions to see if people with different political orientations like different authors. They do. Take Poul Anderson and Harlan Ellison, for example. In the table below we show the percentages in each political category who said they liked each author "strongly:"
A glance at the table shows that reviews are mixed. In both cases, the Radicals agree with the Liberals; around 30% like Anderson strongly, while about 40% give this rating to Ellison. But Moderates and Conservatives like Anderson a lot better than Ellison. If we call Moderates and Conservatives political Righties, and call Radicals and Liberals political Lefties, then we can summarize the political differences as follows. Righties like Anderson strongly about twice as often as do Lefties. Lefties like Ellison strongly about twice as often as do Righties. This is a big political difference.
If we make one little assumption, we can say something about the political orientations of the authors themselves. Let's assume, "People like authors that have the same opinions they do." This sounds reasonable. It follows that Anderson is a Righty and Ellison is a Lefty. Many readers familiar with both would agree. Perhaps the authors would also agree. In any case, we have discovered something about the politics of the authors by merely adding one question to the twenty-seven on how fans felt about them. This is much more efficient than asking twenty-seven extra questions about the politics of each author. The preceding table is a little awkward to read, however, and we could really use some standard measure to indicate correlations like the ones it shows. We could combine percentages, and present a simpler table, like this:
That makes the picture clear, but we have lost data by combining Moderates and Conservatives. We need a good, simple, mathematical index of the association between politics and fondness for an author. Happily, the statisticians have given us just what we need, the correlation coefficient. Actually, there are several different correlation coefficients, with names like tau, gamma, Pearson's r, Spearman's rho -- each designed for use with different kinds of data. But they all look the same and can be interpreted in about the same way. For the record, all our coefficients are gammas and Pearson r's.
A correlation coefficient is a number usually between -1 and +1 describing the relationship between two variables, x and y. A plus sign in front of the coefficient means that if variable x goes up, y is likely to go up, too. A minus sign means that if variable x goes up, y goes down. Zero, or any number close to it, means there is no relationship between x and y.
Suppose you want to know the correlation between weight and height among our twenty-seven SF authors. For each one, you record the weight in pounds (variable x) and height in inches (variable y). You put all fifty-four numbers, properly labeled, into the computer and push the "correlate" button. Out comes a correlation coefficient, let's say 0.60. Since the figure is plus, it means that taller authors tend to be heavier than shorter ones. But the relationship is not exact. A short fat author may be heavier than some tall skinny one. If taller authors were always heavier than shorter ones, the figure would have approached 1.00. We never find that kind of perfect relationship in sociology, because every human characteristic is the result of several causes, no one of which is a complete explanation in itself. Social scientists usually have to content themselves with figures between +/- 0.20 and +/- 0.50. Very small figures, below about +/- 0.15 for this study, are not "significant." That is, they may be the result of pure luck in the run of figures we got, and not represent a real sociological finding.
We produced correlation coefficients showing the relationship between degree of liking for the twenty-seven authors and the political question. Fourteen of the twenty-seven figures were significant, and we have listed them in Figure 2. As you can see, Anderson was the most right-wing author, and Ellison the farthest to the left. The figures are our best estimates, and are not really precise, but it is okay to compare figures if they are not too close together. We can say Anderson is more to the right than Niven, for example. Look through the list yourself and decide if the "+" ones are really Righties and the "-" ones Lefties.
Six other items in the questionnaire asked fans to give their preferences on types of literature, following the same format as the twenty-seven author questions. Listed in order from the most-liked to least-liked, the types were: "hard-science" science fiction, science-fantasy, sword-and-sorcery, classic science fiction from the early days of SF, "New Wave" science fiction, and horror and weird. Our stalwart computer quickly produced 6 x 27 = 162 correlations, describing associations between authors and types. The sixty-six largest of them are shown in Figure 3. Some of the figures should come as no surprise. The positive 0.57 correlation between liking Ellison and liking New Wave is a reflection of the fact that Ellison is the leading propagandist for this type of literature.
|Sword & Sorcery|
|Horror & Weird|
The good doctor Asimov should be pleased at the 0.54 correlation between him and hard-science SF, because he writes it and also produces more factual science books than any man has a right to. The negative correlation between Ellison and hard-science (-0.34) means that if you like it, you probably won't like him. In effect these six lists provide a kind of conceptual map of the authors. The heading of each list describes the authors with positive (+) figures under it. The authors with negative (-) coefficients are in some way opposite to the descriptive heading.
We can turn these tables into real maps simply by graphing the figures for any two types of literature. In Figure 4 we have graphed the political variable against the hard-science variable. Notice that the points for the twenty-seven authors are not scattered at random all over the map. They form a long cluster, slanting from the lower-left to upper-right corners. This is a graphic representation of the overall fact that there is a positive correlation (0.26) between politics and hard-science. That is, right-wing fans and authors tend to prefer hard-science writing. No relationship like this ever turns out exact. Frank Herbert is the main exception in this case.
A Political-Scientific Map of the Authors
Politics is the horizontal dimension, with the right-wingers at the right and the left-wingers at the left. "Hard-Science" science fiction is the vertical dimension, with hard-science authors at the top and anti-hard-science authors at the bottom.
We also found a strong correlation (0.50) between science-fantasy and sword-and-sorcery. Both also showed some affinity to horror and weird. None of these three show any association with politics. While hard-science tended to be somewhat Righty, New Wave was strongly Lefty, having a correlation of -0.51 with politics. Not surprisingly, there is a negative figure (-0.25) between hard-science and New Wave. Politics has not only reared its ugly head, but pushed much of its slimy body into the world of science fiction.
In addition to the six types of SF and fantasy fiction, we also asked about several other kinds of literature. Fans who liked "Utopian political novels and essays" tended to prefer Silverberg, Simak, Sturgeon, and Wells, and were apt to dislike McCaffrey. Those who liked "avant-garde fiction which experiments with new styles " also liked Silverberg, Dick, LeGuin, Farmer, Delany, and Ellison. Those who liked "fiction which deeply probes personal relationships and feelings" tended to go for Herbert, Silverberg, Dick, LeGuin, Delany, Sturgeon, Zelazny, and Ellison. "Fiction concerned with harmful effects of scientific progress" was given a better than average score by fans of Silverberg, Dick, and Ellison. These four types of literature are all Lefty and they all correlate strongly with New Wave. They form an intercorrelated cluster and have ties with "Mainstream literature." Mainstream itself is not Lefty. New Wave looks like the center of the cluster, and the other types really describe several of the aspects of New Wave.
There is some question of what New Wave science fiction really is. When Harlan Ellison announced New Wave to the world in his 1967 anthology Dangerous Visions, he hoped to produce a revolution "intended to shake things up. It was conceived out of a need for new horizons, new forms, new styles, new challenges in the literature of our times." All of our New Wave authors except LeGuin were included in Dangerous Visions, but Asimov, Niven, Leiber, Anderson, and Sturgeon were also there. We were able to learn the years in which thirty-two of the anthology's authors first published in science fiction magazines. The average is 1954 which meant most of the authors weren't new even back in 1967! Ellison wanted old and new authors to experiment with new styles, which meant in many cases the importation of contemporary styles from other branches of literature. So New Wave doesn't mean new authors, and it may not mean new styles. Perhaps it just means an emphasis on certain styles new to science fiction, but familiar elsewhere. Perhaps it means literature containing opinions previously uncommon in science fiction. In any case, despite Ellison's insistence that New Wave is a good thing, our questionnaire respondents liked it less well than hard-science SF or other varieties.
James Blish, an author close to the hard-science tradition, called Ellison "not only the most audible but possibly the most gifted of the American members of the New Wave. " Blish has Said New Wave consists of the following elements:
(1) Heavy emphasis upon the problems of the present, such as overpopulation, racism, pollution, and the Vietnam war, sometimes only slightly disguised by SF trappings; (2) Heavy emphasis upon the manner in which a story is told, sometimes almost to the exclusion of its matter, and with an accompanying borrowing of devices old in the mainstream but new to science fiction, such as stream of consciousness, dadaism, typographical tricks, on-stage sex, Yellow Book horror, and naughty words; (3) Loud claims that this is the direction in which science fiction must go, and all other forms of practice in the field are fossilized; (4) Some genuinely new and worthy experiments embedded in the mud.
Our findings seem entirely in line with what Blish said. New Wave expresses many of the social concerns of the political Left. It can be described as avant-garde fiction which experiments with new styles. New Wave often speculates about utopian dreams or the possible harmful effects of scientific and technological progress. In some ways New Wave is tied closely to the styles of writing taught in universities or often called "Mainstream."
Another way to chart the field is to compare the characteristics and readerships of different science fiction magazines. They differ significantly in editorial policies, styles, and reputations.
In one question we asked respondents to rank the current science fiction magazines, telling us which is their favorite, which they liked second best, and so on. A third liked Analog best, while another third chose Fantasy & Science Fiction. Of course Analog outsells F&SF by a wide margin. But in award votes at science fiction conventions, as in our questionnaire, the two always come close to a tie. This indicates a significant difference between the opinions of the average SF reader and the more single-mindedly dedicated fan. Readers go for Analog, while fans like both magazines equally. There are significant differences between Analog fans and F&SF fans.
People whose favorite magazine is Analog are much more likely to prefer hard-science fiction, and are less likely to go for other types. Politically, the Analog crowd included more Conservatives while the F&SF bunch included more Liberals, but both magazines had support across the spectrum. Those fans who said they worked at technical occupations included 56% of the Analog fans but only 46% of the F&SF fans. Analog supporters were more likely to have seen an Unidentified Flying Object at least once in their lives, and more likely to report they were satisfied with the future facing themselves and their families. Don't ask us what those two facts mean!
The F&SF contingent were more likely to prefer New Wave and all the kinds of literature associated with it. They do not have a monopoly on this attitude, since Amazing Stories fans and others may feel the same way. F&SF fans also gave high scores to science-fantasy and to horror and weird. In other words, Analog fans liked hard-science, the generally best-liked style, while F&SF fans went for everything else.
Our study was not limited to questions dealing with opinions and preferences. We learned all kinds of odd facts about fans and the things that inspired them to like different authors and styles. For example, students like both Ellison and Heinlein more than the average, and like Vance and McCaffrey less. Female fans are more likely than men to prefer McCaffrey and sword-and-sorcery fiction. But rather than hunt around for random findings, let us map SF using the purest and most straightforward data we have -- the author preferences.
We correlated each of the twenty-seven authors with the twenty-six others, producing a 27x26 table with 702 correlation coefficients. (Really, there are only 351 because they come in symmetrical pairs.) Of course, a huge table like this cannot be seen as a whole by the human mind. Pieces of it are easy enough to read, but the table contains too much information to be intelligible as a whole. Looking through it we discover, for example, that fans who like Harlan Ellison are also in favor of Silverberg, Dick, LeGuin and Delany. Those who like Anderson tend to prefer Niven and Heinlein. We should point out that when we talk about "liking" and "preferring, " we don't use these terms in precisely the everyday sense. When we say a fan likes Ellison, we mean he likes Ellison better than the average fan does, not necessarily better than he likes a lot of other authors. We have avoided extremely technical language in this article, so we stand the risk of sometimes saying something we don't quite mean. Like most sociology, this article is a little awkward and a little imprecise to avoid being completely one or the other.
Our correlation table can be used as a reader's guide. Say a friend of yours has just started reading science fiction. He has finished a book by Anderson and liked it very much. You can give him two kinds of objective advice. First, show him Figure 1 and advise him to try a couple of the best-liked authors at the top of the list. Second, look through our correlation table (which is too big to publish here) and tell him to try Niven and Heinlein because they are somewhat similar to Anderson.
Figure 5 is a greatly simplified version of our huge table, given in diagram form. It represents all the correlations of 0.30 or better as straight lines connecting the authors that happen to be correlated. The lengths and directions of the lines have no significance. Pay attention only to which authors they connect. Lovecraft, Tolkien, and Vance are not shown, because they are not closely related to any other authors. Brunner and Herbert are only connected to each other. If you like Brunner, you are apt to like Herbert, and vice versa. If you like Clarke, you are apt to like Asimov and Heinlein. If you like Silverberg, you are apt to like Lafferty, LeGuin, Farmer, Ellison, and Dick.
|Figure 5: Significant Correlations between the Authors. Each straight line indicates a correlation of 0.30 or better between a pair of authors. If you like one of the authors, you will tend to like every other author to which his or her name is directly connected.|
Our computer has generated many more maps of science fiction, enough for a small atlas, several using rather sophisticated mathematical techniques: factor analysis, similarity matrix analysis, and multidimensional scaling. But we think we have already made our point: standard sociological survey techniques can be used to make reliable maps of science fiction or any other kind of literature. When social scientists explain their findings to the public, they walk an unpleasant tightrope. If the findings make too much sense, people will say, "Oh, we knew that all along. Big Deal. " If the findings are strange and surprising, people will refuse to believe them. Damned if we do and damned if we don't.
Many of the results we have presented here will make sense to science fiction readers. We hope that fact will give you some confidence in our methods. Perhaps you also have a sense that questionnaires can produce a tremendous amount of unexpected information when subjected to scientific analysis.
Social science surveys are not just polls undertaken to find out which candidates the average American will vote for. They can be used to uncover hidden relationships between psychological and cultural variables, to chart and even explain important but obscure social relationships. Perhaps everybody knows that politics plays a part in science fiction writing, but now we see what a great part. People are not entirely free to decide what authors they will like. Their backgrounds and general opinions push them in one direction rather than another.
In another questionnaire we asked science fiction fans whether they agreed or disagreed with the following statement: "Questionnaires of this type are valid tools of social-scientific research." The response was cool. The average fan was just a third of the way from "indifferent " toward "slightly agree." Radical fans distrusted the questionnaire most, while Moderates had more faith in it than the average. See! Even attitudes toward sociology can be analyzed sociologically. We hesitate to predict which readers will like this article. Our findings do make sense. We have produced new maps of science fiction.
Bainbridge, William Sims. The Spaceflight Revolution. New York: Wiley-Interscience, 1976.
Blish, James (William Atheling, Jr.). More Issues at Hand. Chicago: Advent, 1970, pages 123,131.
Ellison, Harlan (ed.). Dangerous Visions. Garden City, New York: Doubleday, 196 7.
Rogers, Alva. A Requiem for Astounding. Chicago: Advent, 1964.