In some ways, the term “science fiction” is as problematic as “religion.” “Religion” would be an example of polysemy: the same word can mean very different things. For example, if you study religion in psychology, anthropology, sociology, and philosophy classes, you’ll encounter very different behaviors, different concepts, and assumptions about what is important.
“Science fiction” has a similar problem. There’s a reason Clute & Nicholls’ mammoth Encyclopedia of Science Fiction devotes three pages to the question. The modern term dates from about 1930, as a replacement for (magazine editor Hugo) Gernsback’s “scientifiction”–his label for the writings he bought from new authors and/or collected from old authors and published in magazines like Amazing. The implication seems to be that his choices weren’t random: something distinguished “scientifiction” from the rest of fiction. But what?
If one tries to make a list of the plot features characteristic of science fiction, the first complication is that the characteristics change over time. In 1947, J.O. Bailey described the basis of a science fiction story as a technological invention which results in adventures (quoted in Clute & Nicholls p. 312). In 1987, Kim Stanley Robinson described it as historical literature pushed forward (ibid. p. 314). Both definitions describe a fair number (but never all) of their contemporary science fiction, but neither work well in the other’s generation.
Using a list of characteristics proves to be too inclusive. For example, when you say “science fiction” most of us assume stuff written from the late 19th century on. After all, Verne and Wells are commonly described as the ‘fathers’ of science fiction. (Poe and Shelley sometimes get nods.) But some of the plot elements are a lot older: Lucian was writing about traveling to the moon in the 2nd century. There are some books that are accepted as science fiction, but don’t seem to have a significant number of items from that hypothetical list.
Ironically, the same strategy can be too exclusive. Greg Bear’s Darwin’s Radio, for example, won the Nebula for best science fiction novel, so I’m pretty confident it really is science fiction–but it doesn’t have any spaceships, no aliens, and not even much in the way of technology except for some rather obscure medical epidemiology research, so would probably be left out. Michael Crichton’s Andromeda Strain, in contrast, has some speculative medical epidemiology research–has satellites bringing back ‘space viruses’ that makes people crazy and suicidal. But Crichton is rarely counted as a science fiction author: bookstores shelve him in general fiction (or alongside “techno-thrillers”), not science fiction.
On the other hand, like a certain Supreme Court judge on pornography, we know it when we see it. Authors like Crichton, Margaret Atwood and Kurt Vonnegut just don’t feel right, and are rather unsatisfying as science fiction. “We”–which is to say, those who read a lot of science fiction–seem to make some rather sharp (and pretty consistent) judgments on what ‘counts,’ and these judgments are different than outsiders’. (“You read science fiction? Oh, I just love Margaret Atwood . . . “) I’d argue that there seem to be a large number of implicit criteria in place, which only become ‘visible’ when activated. The analogy would seem to be the difference between American and Irish fiction: it’s not just the setting, it’s cultural references, attitudes, terms, habits . . . impossible to list, but individually clear enough when made conscious.
So, my strategy is to turn it around. Mainstream science fiction (mSF) has plot elements the casual reader associates with science fiction, and includes the Atwoods and Crichtons of the world. Genre science fiction (gSF) is defined socially: a population of individuals who, having read overlapping bodies of literature, have acquired overlapping schema. That is, “science fiction” is what(ever) science fiction readers read. (Compare Norman Spinrad’s [frustrated but serviceable] definition: “science fiction is anything published as science fiction” (ibid. p. 314).
Thus, gSF is a subset of mSF; science fiction is a subset of mainstream culture, with distinct literary expectations, attitudes about the value of technology, and (I’ll extensively argue) characteristic and normative beliefs about how religion should be treated.