The debate about the point of creative writing programmes took a new turn last week. People seem to like this debate – maybe because so many people like taking creative writing classes. Writing in the Atlantic, Richard Jean So and Andrew Piper start by pointing out how much the literary-industrial complex has grown in the last fifty years, and then try to ask a slightly different question about it. Not the usual, 'is the creative writing industry having a pernicious effect on fiction?' but: is it having any effect at all?

They used computational analysis to try answer the question, plugging a couple of hundred books into a computer program to see if it could detect a difference between the novels produced by MFA writers and those written by people who never did an MFA. (‘To make these two groups as comparable as possible’, they ‘only gathered novels by non-MFA writers that were reviewed in the New York Times, which we took as a mark of literary excellence’ – if only.) The program assessed these novels ‘across a range of literary aspects, including diction, style, theme, setting, and even how writers use characters’.

It’s a nice idea. In a course I’m teaching on the American novel, one of the exam questions asks students to measure in some quantifiable way the evolution of style from The Scarlet Letter to The Crying of Lot 49 – it’s always the least frequently attempted essay on the exam. But So and Piper’s results may say more about the limits of computational analysis than they do about creative writing programmes.

For an argument about numbers, their article is short on data. But they start out with one hard fact. By analysing diction, the computer can distinguish MFA novels from non-MFA novels 67 per cent of the time. ‘You don’t need a degree in statistics,’ they write, ‘to know this isn’t very good – you can be right 50 per cent of the time just by guessing.’ For context, So and Piper say that their computer can detect mystery or romance novels 85 and 95 per cent of the time respectively. But I’m not sure how useful this context is. I can spot mystery and romance novels 100 per cent of the time by reading them; but I would never expect to be able tell unfailingly what kind of postgraduate education a writer had.

There are other reasons So and Piper’s algorithms may have struggled. Lots of the non-MFA writers they mention teach on MFA programs or have given enough writing workshops that they might as well have.Akhil Sharma, for example, who didn’t do an MFA, is an assistant professor of creative writing at Rutgers. I would count as one of their non-MFA types: I’ve been reviewed in the Times, don’t have an MFA and never took a workshop in college. But I’ve been teaching creative writing for ten years and have learned from – or been poisoned by – whatever that process does to you. I’ve got my mantras like anybody else, and I’m sure it’s had some influence on my fiction. A lot of editors and agents have been through MFAs, too, so their sense of what works has a significant filtering effect on the fiction that gets published, even if the published writers themselves didn’t get an MFA.

Or perhaps So and Piper weren’t able to identify the MFA novels as a grop because writing programmes don’t all teach the same thing. Then there’s the opposite possibility: writing good fiction may involve mastering certain common techniques, which you discover whether you study them in an MFA or not. As it happens, I believe both of these arguments. Some of what you learn on an MFA is standard issue, and some of it isn’t. It’s like basketball. Every coach will tell you to keep your elbow in when you shoot. Kids who get good on the playground will eventually learn this, too. But coaches also emphasise or teach different things, and because they're different, you can't always generalise about what a formally trained player will do in a game.

If you said to a coach, ‘You can’t teach somebody to play basketball,’ he’d stare at you a little baffled. ‘I can’t promise that they’ll play in the pros,’ he'd say, ‘but I can make them better.’ The debate about the value of creative writing programmes, on both sides of the argument, has tended to make the whole business seem semi-mysterious, like a branch of alchemy – and alchemy may play a part. But much of it is also straightforward, the equivalent of keeping your elbow in and following through on a jump shot.

Here are three things workshops teach students:

1. Pushkin once complained that Byron’s conspirators even say ‘Give me a drink’ conspiratorially. Characters don’t have to be entirely defined by their plot functions.

2. Young writers have a tendency to set down their stories in the order in which the details occur to them, rather than the order in which they make sense: ‘Bob sat at the table. Rain lashed the window. Bob’s shoes were wet. He had just come into the room.’ How about: ‘Bob came in from the rain and sat down, in wet shoes.’ Why is there a table there? Is he in the dining room or kitchen? At a desk? And so on.

3. Dialogue isn’t usually the best way to explain plot. A friend of mine who went to film school calls this the ‘Let’s have a cup of coffee, I’m the king of Spain’ rule.

Most creative writing teachers, and most workshop classes, will give some version of these pieces of advice. You can learn them on your own, but you’ll learn them quicker in company. And they’re worth learning.

There are other standard bits of advice I have mixed feelings about. The most famous is the awful mantra ‘show don’t tell’. It produces a certain kind of fiction and rules out others, which is why I don’t much like it. On the other hand, it works pretty well. Not many writers are good at telling – their explanations are not always that interesting. George Eliot does good explanation. Philip Roth does good explanation. But good explanation is hard to teach: it involves having a sophisticated worldview and finding the moments when that worldview has something specific to say, about psychology, or economics, or the weather. It’s easier to say to a student: let’s cut all that out, stick to the facts, tweak the sequence of events to make it more plausible, prune the dialogue and leave out all the inner thought stuff, which gives the game away, delay the moment of drama, tone it down a little, too, and let’s keep a lid on the hero’s motivations, so we don’t know whether to trust her or not. And at the end of a series of ruthless edits and workshops you have a tight, vivid, suggestive, fine piece of work. You Gordon Lish it.

It’s clear that this kind of advice, no matter how useful, will produce a sameness. One of the seminars I teach involves comparing a Raymond Carver story in its original form (‘Tell the Women We’re Going’) with the version as edited by Lish, to get students to think about what’s being done to them. But that doesn’t stop it from being done. As it happens, I much prefer Carver’s original; my students were split down the middle.

Read more in the London Review of Books

Elif Batuman: Down with Creative Writing · 23 September 2010

Frank Kermode: With Raymond Carver at the Laundromat · 18 October 2000