Chapter 1
"You can know the name of that bird in all the languages of the world, but when you're finished, you'll know absolutely nothing whatever about the bird. So let's look at the bird and see what it's doing — that's what counts." — Richard Feynman, What Do You Care What Other People Think? (1988)
Interstitial 1 — October 1989
The costume is two cardboard boxes and a length of dryer hose. The big box has dials drawn on in marker, the small one has a slot cut out for her eyes, and the hose runs from the top of her head to the small of her back, doing nothing, which is the part she likes best. At every house the porch light comes on, a grown-up leans into the doorway, and the question arrives, and at every house she has the answer ready before the candy hits the pillowcase.
"And what are you supposed to be?"
"A robot."
That's the whole transaction. She says what she is; they pay her.
By nine o'clock she is home, the robot's head on the carpet beside her, the pillowcase upended on the living-room floor. She sorts. Chocolate in one pile — the real pile, the one the other piles exist to protect. Hard candy in another. A pile for trading at school: things she doesn't like that other kids, inexplicably, do. A short stack of full-size bars that need to be looked at more than eaten. And a last small pile, pushed back against the foot of the couch where the shadow is: the ones to hide from dad.
Her mother, stepping over the piles on her way through, says the sour ones should go with the hard ones. They should not. They go with the trading pile, obviously, because that's what they're for.
By Thanksgiving the piles will be gone — traded, eaten, forgotten. Tonight they are exactly right.
The psychologist's question sounds like a parlor game. Your house is on fire. You have one minute. What do you take?
Nobody hesitates. The kids. The dog. The photo albums. The passport, the laptop, the shoebox of letters. People produce the list as fluently as if it had been sitting in their heads all along, waiting.
It hadn't. In the early 1980s, a psychologist named Lawrence Barsalou started putting questions like this to people, and he found that "things to take out of a burning house" behaves like a real category in every way that matters.1 It has better and worse members, and everyone agrees on the ranking: the children outrank the photographs, the photographs outrank the toaster, and nobody saves the toaster. It has a center and fuzzy edges — the wedding album is obviously in; the guitar is a judgment call. People can say what belongs in it as quickly as they can say what counts as a bird.
And yet, unlike birds, the category did not exist until the moment somebody needed it. No one walks around with a fire list. The mind assembled the category on demand, complete with rankings and borderline cases, used it, and stood ready to let it dissolve the moment the question went away.
Barsalou called these ad hoc categories, and once he'd named them (he made them a category), they turned up everywhere: things to pack for a beach trip, ways to make a small apartment feel bigger, things to sell at a garage sale, things to eat on a diet.2 None of them is written down anywhere. None of them needs to be. Each one is built for a purpose, fitted to that purpose, and thrown away afterward like scaffolding.
A child's Halloween piles are ad hoc categories in their purest form. To eat first. To trade. To hide. No adult taught the scheme; no adult could improve it. The piles are perfectly tuned to the politics of one living room in one week of October — and they're gone by Thanksgiving, because by Thanksgiving there is nothing left for them to do. The piles dissolve, the instinct that made them doesn't.
This book is about that instinct: where it comes from, how it actually works — which is not how anyone assumed it worked — and what happened when it was handed filing cabinets, census forms, and, eventually, machines.
Barsalou, L. W. (1983). Ad hoc categories. Memory & Cognition, 11(3), 211–227. The "things to take out of a burning house" example is Barsalou's own; the finding is that such categories show typicality gradients (better/worse members, agreed-on rankings) "as salient as those structuring common categories." The result rests on typicality ratings, not reaction-time measurements — the text above has been kept to that claim. ↩
Barsalou's own examples of ad hoc/goal-derived categories (1983); see also Barsalou (1985) on graded structure organized around an ideal rather than around features that cluster in the world — the trait that distinguishes ad hoc categories from world-tracking ones like "bird" or "fruit" (see Chapter 3–4). ↩