Introduction
Why I read this book
Range: Why Generalists Triumph in a Specialized World is a book about generalists. And it’s pretty much the only book about generalists that’s gotten substantial traction in popular culture. Since I started this whole blog around being a generalist, Range was kind of required reading for me.
Based on the branding, I assume most people who read it suspect themselves to be generalists and are hoping to receive some validation and encouragement toward that identity. I myself went in already knowing that I’m a generalist, not really caring whether the book’s contents validate that, because it’s a core part of myself that I’m not willing (or maybe even able) to change.
I was interested to see what positive aspects of generalism the author would highlight because, aside from what I know of my own experience, I don’t really see those highlights in popular culture.
TL;DR
Most of the content in this book is vignettes of generalists throughout recent history—what they did and how they achieved greater things than their specialist peers. It also contains a lot of informal reports on studies and surveys that show the traits of generalism succeeding in different contexts.
The book’s main conclusions are that 1) people with range, especially early in their careers, do better in the long run, and 2) organizations with range in their constituents also do better.
Yeah, the conclusions are pretty predictable. You know what all the studies and anecdotes are going to show, and unfortunately it can get pretty dry as there are a lot of studies and anecdotes. Ironically, I’d say David Epstein approached this subject from a single angle, and it would’ve benefited from a more varied approach. I would’ve liked to see some analysis on the psychology of generalists, how generalism was viewed throughout history (why it was idealized in some periods and not others), and at least some lip service toward the pitfalls of generalism.
Anyway, don’t worry about the dullness. There are a handful of very compelling ideas in this book, which I’ll present below in their original contexts, with a bit of extra analysis and some connections to other useful ideas I’ve found elsewhere.
Context of the work
This book, published by pop science journalist David Epstein in 2019, inherited the world of what I’d call the Neat Tricks Backed By Science books. Or the Stephen-Pinker-Adjacent books. In different places, Epstein mentions and gives direct responses to the works Thinking Fast and Slow by Daniel Kahneman, Grit by Angela Duckworth, and Battle Hymn of the Tiger Mother by Amy Chua. In terms of style and structure and the way it justifies its claims, Range very much belongs in that genre.
It’s interesting, though, to see a work of this kind explicitly go against the other works. The Neat Trick genre is supposed to overturn old ways of thinking, not other new ways. For Angela Duckworth saying, “IQ doesn’t matter as much as you think; grit matters more,” Epstein then says, “Grit doesn’t matter as much as you think, match-quality decisions matter more.” In my view, his going against contemporary thinkers, when he really didn’t need to mention them at all, demonstrates a bit of extra conviction.
Is it actually true?
With books like this, I’m more skeptical than I think the target reader is. I can imagine generalist-curious people eagerly interpreting it as confirmation that their constant task-switching is a good trait and not just a low attention span or the fear of failure.
I wouldn’t say I’m overwhelmed with the evidence, because even though there’s an overwhelming volume of it, this kind of content is not the most reliable form of evidence. All the anecdotes about great generalists, well, they’re anecdotes. But they’re inspiring, and they do move the needle a bit, mainly by showing that a lot of generalists are “hiding in plain sight.”
The studies and surveys, well, I’ve seen enough non-replicating studies to take all of this with a lot of salt. There are no meta-analyses here, so you just have to hope that the studies were done well and without any bias toward the “generalists are better” conclusion. The same goes for Epstein’s curation of the studies. If you want a quick benchmark for “How reliable are these claims,” I’d look at this paper finding R-index values for Thinking Fast and Slow. Like I said, Range is in the same genre as TFaS, written for the same audience, so I’d expect the same level of epistemic rigor (that is, not stellar).
The major updates I personally made are something like:
- A lot more top achievers are generalists than you might think, because their generalism often doesn’t make it into the biographical content that you’d see at a glance.
- Specialists today are more specialist than at any other time in history, and the pitfalls of specialism (to summarize: narrow thinking) are more pronounced. This makes generalists more valuable as a way to counter that.
- The kinds of problems that lend themselves to specialization are often the less interesting problems. (Incidentally, they are also most easily solved by AI, which is something that problem-solvers might want to consider in the modern age).
Now let’s look at the fun parts.
Kind and wicked
Epstein notes that some activities, like chess and golf, lend themselves to specialization, and so the people at the top of these fields are usually people who started at an early age and focused on little else. Epstein surfaces a conclusion from Kahneman’s psychology research: the difference between kind learning environments (those with a limited, defined set of parameters, where patterns repeat and feedback is immediate) and wicked learning environments (where the rules are unclear or incomplete, feedback is delayed or inaccurate, and patterns aren’t obvious).
The kind:wicked dichotomy sort of matches finite:infinite games. It’s not a perfect match—a wicked learning environment can still be a finite game, like “Double your company’s monthly active users, somehow.” But infinite games have a lot of the characteristics that wicked environments usually have: a wide array of choices, unpredictable outcomes, and no well defined end state to work backwards from.
Kind environments lend themselves to what Epstein calls “chunking”—this is really just “data compression” or “relying on categories” or “using broad heuristics.” The game of chess is a kind learning environment, where the mind can learn to identify patterns that repeat often. Chess grandmasters can look at a given game-state for a split second and then replicate it on the board in front of them, but when the pieces are arranged into an impossible game-state, they can’t do it. This demonstrates that it’s not raw memory that sets chess masters apart, but the efficient compression of data that is familiar to them.
Now, Epstein mentions that AI systems are very good at performing in kind learning environments. I really like how he drew a timeline of AI performance progress—chess, Jeopardy, Starcraft, driving, statistical analysis—and then pointed out that this is also a spectrum from kind to wicked learning environments. Chess was solved by AI decades ago, Starcraft only recently, and driving is still in progress. The larger and more varied the game, the more sophisticated the systems need to be to navigate it.
What I also like is that Epstein avoids saying, “AI can only win in kind learning environments,” which would’ve been more tempting to say in 2019 than it is now. He makes no claim that AI progress will halt somewhere, but only points out that wicked learning environments are taking longer, as a way to show that they’re qualitatively different.
Shots fired
I promised some beef with those other pop science books, so I’ll relay that here. These weren’t the most compelling chapters for me, but I’ll outline Epstein’s takes on two of the big ideas that have been floating around in this idea ecosystem.
Against tiger mothering
Apparently, early specialization in musical practice and performance was a major part of Amy Chua’s parenting. If it wasn’t, well, this is the only angle from which Epstein opposes the whole Tiger Mom method. He brings up a list of prolific musicians who all spent significant time in a state of “musical generalism,” as opposed to the very early, highly structured training that Chua put her daughters through.
He points to the figlie del coro, a group of female musicians in 18th Century Venice. They were possibly the greatest musicians the Western world had ever seen at the time, and they each played three to five different instruments. They didn’t have extensive practice regimens—they lived simple lives and spent a lot of time doing chores in the orphanage they belonged to. What set them apart? Epstein says it was their wide range of musical experience, and also the opportunity to test and figure out which instrument was truly the best fit.
He goes on to highlight the importance of a “sampling period” in the lives of other top musicians. And he points to a plethora of jazz musicians who never took formal lessons and instead learned just by being around music all the time, picking it up through osmosis and developing it in unstructured play.
Amy Chua’s daughter did become a violin prodigy, Epstein concedes, but as an adult she has since quit playing. He also suggests, perhaps controversially, that drilling and performing classical music to precision is missing something essential about music as a human experience. “Jazz is creative, classical is re-creative.” I’m inclined to agree, but it’s hard to build much of an objective argument on a foundation of, “This genre of music is more important than that other one.”
Against grit
I shouldn’t say Epstein is “against grit,” but he’s against elevating grit as an unqualified virtue and using it as the primary predictor of success. This view seems to be backed by some good data.
Angela Duckworth developed Grit while observing the US Military’s Academy’s challenging “Beast Barracks” training program. This new attribute, grit, a combination of “work ethic and resilience” and “consistency of interests,” ended up being the best predictor of which individuals would complete the training program—better than the academy’s Whole Candidate Score, the metric that had previously been used to select cadets.
But, fast-forward a few years, and that military academy was suffering a shortage of officers—cadets who had previously passed the training kept leaving the military for other careers. Did they lose their grit? No, Epstein says, it’s just that they were finally empowered to make “match-quality decisions” by switching to other career paths. The military had no talent-matching mechanism, while corporate America was investing heavily in talent-matching. You can’t expect people to never quit anything, he reasons.
The academy developed the Officer Career Satisfaction Program as a way to give officers more work-style choices within the military. It was effective at retaining officers—more effective than offering them more money (though it’s unclear how much money was offered).
To sum up Epstein’s finding: a singular focus on grit will select candidates who are resilient and whose interests are aligned right now. But it can’t promise ongoing future success, because you can’t predict how their interests will change in the future. That ties in with a theme later in the book: the end-of-history illusion, how people systematically underestimate how much they’ll change in the future. You need to build flexibility into people’s careers if you want to ensure they keep showing up over the long haul. Perfect grit is an unrealistic solution.
Analogical thinking
In the chapter called “Thinking Outside Experience,” the subject is astronomer Johannes Kepler, struggling to understand the complex motion of the planets in the night sky. His approach was interesting to me: he’d first search for an analogy that could qualitatively match his observations, and once that fit, he’d pin down the math later. He tried many analogies: Are the planets sliding on transparent spheres? Are they like floating objects in a whirlpool? Does the sun’s light “pull” on everything it illuminates? They were all incorrect, but in subtler and subtler ways. This was how Kepler eventually conceptualized the idea of cosmic gravity.
What follows in this chapter are some lessons about how thinking analogically helps people solve hard problems. There’s a story (riddle?) about a radiation beam that can kill cancer cells at a certain intensity, but it also kills healthy cells at that intensity. So how can we use it? Epstein then prompts you with a story: several groups of soldiers, each small enough to travel without raising alarm, take many roads to converge on a city from all directions, where they are then numerous enough to conquer it. Then I curse to myself for not seeing the radiation-beam solution beforehand, and concede that yes analogical thinking is an effective tool.
This surprised me. I think analogies are… out of style, nowadays? In popular culture and popular science, analogies are mostly used to dumb down a known, precise answer, so novices can grasp it. How many times have you heard, “Spacetime is like a trampoline with a bowling ball on it,” with no attempt to address the obvious, “Yeah, but the trampoline only behaves that way because of gravity and isn’t gravity the thing we’re trying to analogize here?” So it was interesting to see that paradigm inverted: you can start with analogies as a way to conceptually grasp a situation, which will then help you arrive at the accurate mathematical model more quickly (or at all).
Trapped priors
“Trapped priors” is a concept that’s basically a more under-the-hood description of “confirmation bias.” It’s the unofficial theme of the chapter “Fooled by Expertise,” which describes how specialists can think too narrowly and make really bad predictions as a result.
Paul R. Ehrlich, a butterfly ecology specialist, said in 1968, “The battle to feed all of humanity is over. In the 1970s hundreds of millions of people will starve to death in spite of any crash programs embarked upon now. At this late date nothing can prevent a substantial increase in the world death rate.” That, of course, didn’t happen. And so we learned that human population growth is more complicated to model than butterflies fighting over flowers, or whatever.
A surprising detail that again shows a bit of extra rigor from Epstein: economist Julian Simon, the foil of Ehrlich who said something like, “More humans will mean more ingenuity, and we’ll figure out a way to sustain ourselves,” is also pinned as a specialist with trapped priors. Why? To operationalize his bet against Ehrlich, Simon proposed using the price of metals as a proxy for mass human flourishing. I don’t fully understand his reasoning, but the thrust of this story is that he measured human flourishing in a way that naively made sense from a narrow economics point of view. Later on, other experts basically said, “That wasn’t a good proxy, you actually won the bet by pure luck.” Simon never amended his claims.
Then we get an introduction to political scientist Philip Tetlock, his search for “superforecasters,” and the Good Judgment Project. His work started with him asking political scientists during the Cold War to make probabilistic predictions of events in the coming years (and finding that they were bad at it). Tetlock called them “hedgehogs” for their narrow, on-the-ground viewpoints. Hedgehogs are bad at making predictions, but good at spinning narratives that protect their worldviews in the face of new evidence.
The Good Judgment Project came about when Tetlock started asking non-experts to make the same predictions about the Cold War. Some were better and some were worse, but of course the real value came in identifying and analyzing those better—the “superforecasters.” Superforecasters were found to have the following traits: numeracy (the ability to think and express oneself in terms of numbers), a range of experiences, genuine curiosity about many things, polite disagreement, a desire to falsify their own beliefs, and a willingness to change positions often.
I didn’t immediately see why generalists would be more likely to be superforecasters, but Epstein raises the idea of “science curiosity, as opposed to science knowledge.” People who have an abundance of genuine curiosity are perhaps more likely to be generalists, and also more likely to hold their beliefs loosely in the way that superforecasters do.
Trapped epistemologies
The chapter that immediately follows, called “Learning to Drop Your Familiar Tools,” describes a different kind of “trapped priors.” Not just individual beliefs, but whole ways of knowing, can become trapped by too much narrow, self-reinforcing experience.
Epstein describes the US Challenger disaster. Key decision makers gravely misunderstood what a “factor of safety” meant, they decided to launch the shuttle despite some warning signs on their equipment, the shuttle was destroyed, and then famous physicist Richard Feynman yelled at everyone. But Epstein reveals an angle that I’d missed the first time I’d heard this story.
Part of the reason NASA went ahead with the launch despite “warning signs” was that they had a systematically data-driven culture, and the warning signs were not quantified in any data. “The O-rings did something weird in the test, and although the equipment didn’t fail the test, it was unexpected. What if they do something unexpected again, but worse?” That kind of objection ran up against a culture of, “Nice idea, come back to us when you’ve got data.”
You can imagine how well a data-only culture had worked for NASA, and how much waste it had successfully dismissed. But it blinded them to a black-swan-type threat. A quote from Feynman: “When you don’t have any data, you have to use reason.” That pesky “reason,” refusing to be mapped to any well-defined process! Before these eleven virtues is a virtue which is nameless.
Expanding on the theme of “specialists being too hesitant to ditch their preferred tools,” Epstein describes how, in deadly forest fires, the bodies of firefighters have often been found still clutching their unwieldy chainsaws. Navy personnel have drowned, still wearing their steel-toed boots. Fighter pilots have refused orders to eject. The specialist’s tool becomes a part of their body, such that, in the heat of the moment, they don’t consider tearing it off. Likewise their mental tool becomes a part of their mind.
The freshest, most surprising example of a trapped tool is organizational congruence. It’s completely taken for granted in modern workplace culture that the company should have some core values that are adhered to down the whole chain of command, at every level. Amazon has its leadership principles, Microsoft has its company values, and so on. Epstein claims this has gone too far and that the optimal amount of “workplace ambiguity” is more than zero. He warns of professionals getting too attached to the process, asking, “Does this match the official company values?” but never, “Do I predict this will actually work?” Some amount of incongruence forces professionals to develop agency, Epstein argues, and individual agency must be balanced with group compliance, not replaced by it.
Now, it’s not exactly a bold claim to say, “The bureaucracy of large organizations slows things down and stifles creativity,” but I never considered that even the act of writing out company-wide values starts that process. It gives people a template to follow in all their decisions—a template other than “will this work.”
The overspecialization of science
In Current Year, you have probably already heard of this problem. Every modern scientist is doing something like “isolate this one particular protein in this one particular strain of yeast,” and virtually no one is researching “How to extend human life” or “How to control the weather.” I exaggerate.
But it’s evident, if you ask around, that most scientists today are working on extremely narrow questions. Are they just really passionate about that one yeast protein? Epstein doesn’t think so.
He recalls sitting in on a grant hearing by a Senate subcommittee on science and space research, observing that only the proposals with an immediate and tangible application were considered. He reflects on the fact that so many of the most applicable breakthroughs in science came about as the result of non-directed experimentation, such as Oliver Smithies discovering gel electrophoresis by playing around with starch paper, and Tu Youyou discovering an effective malaria treatment chemical from a traditional herbal remedy.
He quotes biologist Yoshinori Ohsumi, “Scientists are increasingly required to provide evidence of immediate and tangible applications for their work. They’re expected to say what they’ll find before they look for it.”
Obviously, Epstein’s solution is to grant more money to open-ended research and experimentation. It’s hard for me to ignore the other side of the equation, though: how does a government agency justify, to its taxpayers, the funding of research that has no real indication of paying off? It’s a conundrum, because plenty of useful research will start off that way, but so will most useless research. I’m reminded that so many great scientists in history have completely avoided this question by being more like “self-funded hobbyists playing around in their basement” than “straight-A students who are great at winning grants.”
Maybe in the recent past this problem was easier, because we could all coordinate around broad national goals like “send man to the moon,” so any free experimentation vaguely related to that effort got funded. Nowadays, nobody agrees on the broad goals, much less which sub-goals to fund, so the only socially permissible things to fund are projects that have a very quick, profitable turnaround.
Or maybe it’s a bunch of other reasons.
Widely cited microbiologist and “generalist scientist” Arturo Casadevall has a slightly different view of things (which Epstein seems to want to merge with his own). Casadevall’s emphasis is not on funding dynamics forcing specialization, but rather on specialization in schools producing bad scientists. He raises a claim that kids these days (paraphrasing) are taught a list of scientific facts but not how to think scientifically, and that this is extra silly because we now have smartphones that can look up all the facts anyway. Casadevall founded the R3 initiative to try to reintroduce scientific thinking and a philosophy of science to the research industry.
I’m mostly happy to pile on and say, “Scientists today don’t understand scientific thinking,” but I’m also suspicious of how good it feels to pile onto that particular pile, and how hard it is to verify the actual claim.
Hidden generalists?
Are most high-achievers actually natural generalists who specialized late in their careers? Epstein hints so. I’m not convinced it’s most, but definitely more than I thought.
Why? For one thing, a bunch of the generalist figures that he highlights in this book are people I would’ve guessed were lifelong specialists. So there’s at least a slight update I need to make to my expectations.
Secondly, there are all those studies he cites which show some variation of, “The majority of top Xers started in fields other than X.” As I said, I’m suspicious of both poor study quality and selection effects, but nevertheless, it’s such a high volume of studies that I’m willing to believe a handful of them are true (similar to the studies in Thinking Fast and Slow).
But thirdly, there’s a qualitative argument that I really like. He notes that when someone’s story is told at the end of their successful career, it’s just A-to-B, because we know what A and B are, and it makes for a more coherent story that way. The story isn’t being told when it’s in the middle, when the person is just hopping around between interests, having yet to make that winning connection or excel at that final project.
I’m reminded of a tricky dynamic in financial investing. When you look at the price chart of an asset, you can draw a line between, say, the price 30 years ago and the current price (let’s assume it’s higher now). And then you can come up with a valuation metric in the form of CurrentPrice/some-other-value, and it will look like a predictive metric for future returns, regardless of what the other value is. The metric, in retrospect, will appear to be mean-reverting, correctly predicting higher returns when it dips and lower returns when it spikes. But that’s because the “future returns” are, by definition, FuturePrice/CurrentPrice. So CurrentPrice is present in both sides of the equation, and that is the reason for the apparent correlation. I’m paraphrasing the content in Valuation and Stock Market Returns: Adventures in Curve Fitting; check that out if you want an explanation that’s not so rushed.
The general point is that anything can look mean-reverting if it’s noticed in hindsight and scaled to past data. And I think the same goes for popular narratives:
Roger Federer is the greatest tennis player? Ah, let’s look at his past. Here’s the point where he first started playing tennis—that’s important. Look at this friend of his who got him into tennis, he was an important influence on him. And Roger played soccer in high school? Ah, but then he mean-reverted back to tennis, didn’t he? Soccer just didn’t suit him. And boom, you’ve got a biography of a tennis specialist.
But in reality, Roger Federer played just about every sport growing up. His mother was a tennis coach but she didn’t train him, and he was bad at tennis, worse than the other sports, when he first tried to learn. But there’s no reason to acknowledge that information when you’re talking about Roger Federer, because you’re probably talking about his tennis.
So I guess the point here is to be aware of the memetic effects—what makes an idea attractive. Specialist narratives are memetically more attractive than generalist narratives, all else being equal, so we should adjust our thinking accordingly.
Conclusion
Although I didn’t love the style of Range and would’ve liked to see fewer, smaller claims argued more strongly, I appreciate David Epstein taking on this project and bringing lots of compelling content into a topic that’s lacking it.
I am personally trying to overcome that memetic disadvantage myself, by writing content about generalists here on TrueGeneralist.com.