The Elephant in the Room: The Psychology of Innuendo & Euphemism

[00:00:00] ANDREW GARRETT:
Good afternoon. I’m Andrew Garrett, the chair of the linguistics department here at UC Berkeley, and I’m happy to welcome you to the second Charles and– Charles M. and Martha Hitchcock Lecture by Steven Pinker. We saw in yesterday’s lecture, and anyone who’s read Steve’s books knows, that he has the rare ability to make the most complex set of ideas understandable, the most challenging problems tractable.

He also has a talent for finding and clearly showing the most vivid examples. This is an especially valuable skill for linguists like me. We spend our days playing in the glorious sandbox of human language.

Our little shovels are digital recording machines, our buckets are scholarly articles and monographs, and a lot of us are scared of climbing out of the sand. Steve comes with a broad sense of the whole human playground and is smart enough to tell us how the lovely grains of sand matter. In each area of linguistic science where he has ventured, therefore, I see a line running straight from the most interesting details of language to very big questions about the mind.

For example, about twenty-five years ago, Steve’s first work that I happened to read explored how individual idiosyncrasies of usage in real sentences like, “That experience grew me up in a hurry,” or General Alexander Haig’s, “Let me caveat that,” cast important light on the cognitive underpinnings of language as used even by those who are less louche linguistically. The actual patterns of idiosyncrasy, Steve argued, support a highly structured model of human linguistic cognition. This has been a theme of Steve’s work, showing that evidence from both natural usage and controlled experiments, and he designs very creative psycholinguistic studies, shows how intricately structured what he has called the human language instinct must be.

People who aren’t language geeks might be startled by the number of papers in our field trying to explain exactly why, even though the past tense of sink is sank, we say things like, “I synced my iPod,” not, “I sank my iPod.” But the implications turn out to be really significant, helping us see whether linguistic cognition is associative or symbolic. In this important debate, Steve has been among the major participants over several decades.

Of course, the nature of the language faculty raises questions of an evolutionary nature. If there is a language instinct, did it evolve? And if it evolved, how could that happen?

Steve is one of the two or three leading participants in that debate as well, stressing again the significance of what he entertainingly calls the raffish idiosyncrasies and seemingly peripheral, but actually omnipresent features of human language. Today we’ll hear about one of the newest strands of Steve’s research on indirect speech acts, a great topic for a Berkeley crowd, as I expect he’ll explain. I’d say more, but a nod’s as good as a wink to a blind bat, if you know what I mean.

For the e- for the elephant in the room, the psychology of innuendo and euphemism, please join me in welcoming Professor Steven Pinker.

(applause)

[00:03:08] STEVEN PINKER:
Thanks very much.

(applause)

Thank you very much, Andrew, for that very kind introduction. Why don’t people just say what they mean? Why do people sidestep, shilly-shally, beat around the bush, veil their intentions in innuendo, euphemism, doublespeak, and other forms of verbal fig leaves, counting on the reader to catch their drift, read between the lines, and connect the dots?

Let me give you some examples of this phenomenon of indirect speech. Any of you who remember one of the opening scenes of the movie Fargo will recall that at one critical juncture, uh, a pair of kidnappers with a hostage tied up in the back seat of the car is pulled over by a police officer because the car is missing its plates. The officer demands that the driver hand over his driver’s license.

He proffers his wallet with the driver’s license showing and a fifty dollar bill extending ever so slightly, and he says to the officer, “I was thinking that maybe the best thing would be to take care of it here in Brainerd.” Which of course everyone recognizes as a veiled bribe. Uh, another example of a veiled bribe can be seen in this New Yorker cartoon in which the driver says to the officer, “What Hershey bar?

I didn’t see any Hershey bar.” Uh, polite requests. If you could pass the guacamole, that would be awesome.

If you think about it, that sentence does not make a whole lot of sense, but somehow, uh, it sounds, uh, it sounds a whole lot nicer than, “Give me the damn guacamole.”

(laughter)

Any of you who have been, uh, at a university fundraising event are familiar with euphemistic schnorring, such as, “We’re counting on you to show leadership in our campaign for the future,” uh, rather than, “You’ve got a lot of money. How about giving some of it to Berkeley?” “Would you like to come up and see my etchings?”

Uh, that has been recognized as a sexual come-on for so long that in the nineteen thirties, James Thurber drew a cartoon in which a confused man says to his date, \”You wait here and I’ll bring the etchings down.\”

(laughter)

And then, uh, fans of The Sopranos might recognize dialogue such as the following. \”Uh, I hear you’re the jury foreman in the Soprano trial. It’s an important civic responsibility.

You’ve got a wife and kids. We know you’ll do the right thing. Which, of course, we all recognize as a veiled threat.

So the puzzle is, why are bribes, requests, seductions, solicitations, and threats so often veiled when presumably both parties know what they mean? Now, there, uh, this is a topic that’s in the general area of a subfield of linguistics called, uh, discourse or pragmatics. And, uh, Andrew made reference to the familiarity of that topic here at Berkeley, in large part because the pioneer of this field was, uh, Paul Grice, who proposed, uh, uh, a gen– overarching principle called the cooperative principle, namely that the speaker and hearer cooperate to advance the conversation.

Uh, Penelope Brown and Stephen Levinson extended Grice’s ideas in their politeness theory, according to which the speaker and hearer cooperate to, uh, not just to advance the message, but to maintain a nebulous quality called face, as in saving face and losing face. Uh, my former colleague at Stanford, Herb Clark, in Using Language, proposes that, uh, this phenomenon falls under a, a, a, um, a phenomenon that he calls joint action. Language use is really a form of joint action.

A joint action is one is, that is carried out by an ensemble of people acting in coordination with each other. As simple examples, think of two people waltzing, paddling a canoe, playing a piano duet, or making love. Well, the– all of these insights, uh, I think, help us understand components of the phenomenon of indirect speech.

But what they leave out is why there should be such a phenomenon in the first place. If conversation was really a matter of pure cooperation, cooperative agents, you would expect, would not bother with, uh, all this indirectness and beating around the bush and, and, uh, treasure hunt to figure out what the speaker really means. When, in general, when you have cooperators, they’re– they can get away with a minimum of communication, like a conspiratorial whisper.

There’s no reason that there should– they should– the speaker should lay out all of these, uh, puzzles of indirectness and euphemism. Um, also, joint actions like waltzing, paddling a canoe, or making love are usually conducted in an atmosphere of relaxation and mutual enjoyment, whereas indirect speech is always suffused with some degree of tension and apprehension.

(cough)

In particular, cooperation theories predict that indirect speech is of a piece with the phenomenon of politeness, a set of tactics by which we bridge gaps in power or social distance. But in fact, uh, a former student of mine, James Lee, now a professor at the University of Minnesota, and I, uh, did, uh, some studies that showed that the phenomenon of politeness and the phenomenon of indirectness are actually qualitatively different. And I’ll– I won’t show you the data, but I think just giving you some of the examples will, uh, uh, show why they are different.

When you have a– We described a number of hypothetical scenarios for our subjects and asked them what form of speech they feel would be most appropriate. What would the character in the scenario be most likely to say?

What would they be most likely to say? And indeed, in conformity with theories of politeness, when there is a big gap in power and distance, such as a junior worker asking a senior worker at a firm for help on a project, Indeed, a, uh, many of the reflexes of politeness were used, but the speech was not indirect. So for the, uh, preferred form in which the request was couched would be something like, “I’m really sorry to bother you, Dave.”

I wouldn’t ask if this wasn’t hugely important to me, but do you think it might be possible to help me with this analysis?” There’s a lot of, uh, cringing and hedging and hesitation, exactly what politeness theory would predict. But notice that the request itself is blurted out.

Do you think you could help me with this analysis? An indirect speech act, such as, “I really admire you, Dave.” You have the perfect background for this.

I wish someone would have told me in school that this stuff would be really invaluable in my work, kind of beating around the bush, was exactly not what people thought was appropriate in that s-setting. Nothing. In contrast, when there is a fraught relationship, such as a sexual come-on, then you get the opposite.

The preferred form of speech is indirect, but it’s not particularly polite. So, for example, what seems plausible in that context would be, uh, Michael saying, uh, “It’s been a really nice evening, Sarah. Would you like to come up to my apartment and see the great view?”

What he would not say would be something like, “I’m really sorry to bother you, Sarah, and I wouldn’t ask if this wasn’t hugely important to me, but do you think that it might be possible to have sex?” So I hope that will convince you that politeness is a real phenomenon, but it is not the same as indirectness. In trying to solve this puzzle, I invoked what I consider to be the fundamental insight of evolutionary psychology, which I attribute mainly to Robert Trivers, which is that all social relationships involve mixtures of cooperation and conflict.

There is no such thing as a purely, uh, a relationship that is purely cooperative in the sense that the ultimate interests of both parties are perfectly aligned. And indeed that the complex coded nature of indirect speech reflects exactly this tension. Now, indirect speech is more than an academic puzzle.

It is also a topic in linguistics with enormous practical applications, such as in the language of diplomacy. Many diplomatic agreements, such as UN Resolution two forty-two, which, uh, came in the wake of the Six-Day Arab-Israeli War in nineteen sixty-seven, are, uh, famously, notoriously, uh, indirect and vague, and the parties have been arguing ever since over what the wording actually means. The language of blackmail is often, uh, framed in indirect speech.

One example is, uh, came when a, uh, man approached David Letterman a few years ago with a, uh, screenplay about a, uh, talk show host who was having an affair with one of his, uh, staffers. This man just offered to sell the rights of the screenplay to David Letterman. Now, it so happened that at the time, David Letterman was having an affair with one of his, uh, uh, staffers.

He interpreted this not as an innocent offer to buy the rights to a screenplay, but as an act of blackmail, namely pay off the guy and he would, uh, keep this, uh, secret. Uh, bribery. Many bribes are, uh, tendered, uh, in indirect speech, not just in Fargo and New Yorker cartoons, but in real life.

Uh, again, I’ll give you one example. Uh, when the Equal Rights Amendment was, uh, just a, a state short of ratification, a lobbyist for the National Organization of Women, Wanda Brandstetter, handed an Illinois congressman a business card that said, uh, “Your vote and, uh, for one thousand dollars and the support of the ERA.” And she was convicted of attempting to bribe a congressman and sentenced to five hundred hours of community service.

Uh, if she had tendered the, uh, same bribe in different language using indirect speech, such as, uh, “As you know, Congressman,” um, organizations such as NOW tend to support candidates whose positions are, uh, in harmony with our goals. And as you know, Congressman, one of our goals is passage of the Equal Rights Amendment, and we are now considering whether to contribute a thousand dollars to your campaign. Then, uh, she would not have been convictable in court.

Uh, and finally, sexual harassment. Those of you who remember the Clarence Thomas confirmation, uh, hearings remember the immortal line, “Who put pubic hair on my Coke?” Which, uh, Clarence Thomas, uh, said to Anita Hill when she was, uh, working for him in the Office of Equal Opportunity, and which she interpreted as a sexual come-on, and hence as an example of sexual harassment.

Now, note by the way, that all of these areas in which indirect speech has practical applications are arenas of conflict, and I think that is not a coincidence. Well, here is a theory that, uh, James Lee and Martin Nowak and I have, uh, proposed to try to make sense of this puzzle. We call it the theory of the strategic speaker, and I will first sum it up in a sentence and then unpack each of the propositions that make up the sentence.

The theory is that indirect speech is a rational strategy to attain plausible deniability of common knowledge of relational models. None of those mean anything to you at this point, right? But that is a preview of the three parts of this talk.

Part one: plausible deniability. Um, the starting point in this analysis is a game-theoretic problem called the identification problem, first pointed out by Thomas Schelling in his classic book, The Strategy of Conflict. Namely, how do you figure out the rational course of action when the outcome depends on another, uh, intelligent agent, but you don’t know the agent’s values?

And as it happens, the example that Schelling used to illustrate this problem was bribing a police officer. So imagine a, a rational actor pulled over by a policeman who is considering two alternatives, either bribe– offer a bribe to the officer in so many words or remaining silent. Which of those two courses of action brings the higher payoff?

Well, the answer is, it depends. It depends on what kind of officer you’re facing. If you’re dealing with a dishonest officer, he would accept the bribe, and you get the very high payoff of going free without a ticket, or in the case of uh, Fargo, without being arrested for kidnapping.

On the other hand, if you have an honest officer and you tender the bribe, uh, then you could get arrested for the crime of attempting to bribe an officer, which carries a much greater cost than the traffic ticket. In contrast, if you remain silent and don’t bother offering a bribe at all, then it doesn’t matter what kind of officer you’re facing. In either case, you get the moderate cost of a traffic ticket.

So, in comparing the two options, one of which has a guaranteed moderate cost, one of which has a high payoff and a high cost, it’s not immediately apparent which of these two is more attractive. But now imagine that there is a third option, namely a veiled bribe, such as, “I was wondering if there were some way of settling this without going to court and doing a lot of paperwork.” Well, now a dishonest officer can sniff out the bribe beneath the innuendo, consummate the transaction, you get the very high, uh, payoff of going free.

An honest officer couldn’t make a bribery charge stick in court by the standard of proof beyond a reasonable doubt, so he, he, would have to let it pass, and the worst that you would end up with is a traffic ticket. So the veiled bribe combines the high payoff of bribing a dishonest officer with the relatively small cost of failing to bribe an honest officer combined in a single option, and so it is the rational choice. Now, Martin, uh, and James and I, uh, tried to formalize this intuition by calculating the average expected cost of direct and indirect bribes, and it turns out a f- fairly simple game-theoretic model can give you this, uh, calculation.

It depends on the proportion of corrupt versus honest cops in the population of policemen, the cost of the ticket, the cost of the bribe, which has to be less than the cost of the ticket, or else it would never pay to bribe. The cost of an arrest for bribery, which has to be worse than the cost of the ticket, otherwise it would always pay to bribe. And crucially, and this is where linguistics comes in, the probability that a cop will treat a statement as an attempted bribe.

Now, it turns out that the– an indirect bribe… I’m sorry, I’m supposed to wait till the-

[00:18:14] CREW MEMBER 1:
We just wanna make sure we can make this sound stop.

[00:18:16] CREW MEMBER 2:
Okay, are we good?

[00:18:17] CREW MEMBER 3:
Okay, sorry for the delay.

[00:18:19] STEVEN PINKER:
Okay. Yes, resume? Thank you.

The result is that an indirect bribe is the optimal strategy if and only if the threshold of explicitness required for an honest cop to arrest a driver is greater than the threshold of explicitness required for a corrupt cop to accept a bribe. Uh, let me, uh, un-unpack that condition. So over here you’ve got the, uh, linguistic variable of how naked the, uh, versus tacit or implicit the bribe is, uh, going from perfectly innocent observation all the way out to a, uh, naked offer of a quid pro quo, and which we assume reflects the probability that a hearer will interpret the, uh, proposition as intended to be a bribe.

Over here we have the expected cost. Now, um, it turns out that if there is a continuous function relating the explicitness of the proposition to the probability that the, uh, officer will interpret it as a bribe, then for an honest cop, here is the payoff. Basically, the more explicit the bribe is, the higher the cost, because with an honest cop, that just means he’s more and more likely to arrest you for bribery.

Uh, for a corrupt cop, the more explicit the proposition, the more likely he is to accept the bribe and the lower the cost, that is the higher the benefit. Uh, the, uh… So which is optimal? Well, it only depends then on the proportion of honest and corrupt cops in the population of, uh, policemen.

If there are more honest cops, then, uh, the optimal strategy, that is the one with the lowest cost, is just to keep your mouth shut and don’t make a uh, don’t say anything. If there are more, uh, corrupt cops, then the rational strategy is to blurt out the bribe in, uh, so many words. So in the case where you have a continuous function relating the probability of interpreting an ambiguous proposition as a bribe, indirect speech, in fact, is not an optimal strategy.

The optimal strategy is either shut up or offer the bribe in so many words, depending only on the proportion of honest and corrupt cops in the population. So under these assumptions, in fact, there is no rational reason ever to use indirect speech. Here’s what makes it rational.

Imagine that there is a nonlinear relationship between the, uh, explicitness of the bribe and the likelihood that it will be acted upon as a bribe. In an extreme case, a step function, so that an honest cop has a threshold. It has to be a certain degree of explicitness before the honest cop will whip out the handcuffs and read you your rights.

Uh, likewise for a corrupt cop, there is a discontinuity and cri-critically, the corrupt throp– cop’s threshold is lower than the honest cop’s threshold. And presumably, this asymmetry is due to the legal requirement that guilt in a criminal proceeding be proven beyond a reasonable doubt. Well, this gap between the corrupt cop’s threshold and the honest cop’s threshold means that the average payoff, uh, shows a local minimum in this region of intermediate explicitness in which the costs are, uh, lowest, meaning that it is optimal for this rational speaker to couch his proposition, uh, i-with a degree of explicitness that falls between those two thresholds.

So that’s the, that’s the model. Another way of putting it is that when you have different payoffs with different hearers and different thresholds for those hearers, then you get the logic of what we informally refer to as plausible deniability. So what I’ve just shown you is a kind of formalization of the common sense notion of plausible deniability.

Now note, by the way, that, uh, this analysis, even though it’s, it’s– I, I hope it’s fairly intuitive, shows why pure cooperation is not enough to explain indirect speech. There is an inherent role of conflict here. Indirect speech is not being used to help an honest officer attain his goal, but rather to confound or thwart him in a-attaining this goal because the interests of the officer and the interests of the driver are not perfectly aligned.

And, uh, the critical prediction, recall, is that there are different thresholds for acting on an ambiguous proposition between the honest and dishonest cops. And in, in fact, uh, James and I, uh, ran a number of experiments in which we gave people a set of propositions varying in explicitness, asked them if they were, uh, a, uh, an officer or if they were expecting an officer to act, at what level in the continuum would they expect them to act. And indeed, uh, everyone, uh, agreed that honest cops would need a greater degree of explicitness before, uh, making an arrest than dishonest cops would need in acting on a bribe.

Okay, that part seems pretty straightforward. Um, but, um, it leaves open the question of why people use indirect speech in non-legal situations when there are no financial or legal payoffs and penalties. It’s not as if we constantly live under the, uh, threat of prosecution for our actions.

Why then do we still use so much indirect speech in everyday life, such as in everyday cases of bribery? Now, you might react, “Everyday cases of bribery? What do you mean?

When would I, an honest, uh, upstanding, law-abiding citizen, ever be tempted to offer a bribe?” Well, how’s this? You want to go to the hottest restaurant in town.

You have no reservation. Why not slip the maître d’ a twenty-dollar bill in exchange for jumping the queue and being seated immediately? This is the assignment that an editor at Gourmet magazine gave the writer Bruce Fit-Feiler.

He dared him to try it at Manhattan’s most exclusive restaurants on a Saturday night and write up his experiences for the readers of the magazine. And I found this, uh, as a psychologist, I found this article utterly fascinating. Uh, first of all, despite the fact that as far as I know, no one has ever been arrested for the crime of attempting to bribe a maître d’.

(laughter)

The assignment filled him with anxiety, and I’m going to read you from the– read to you from the first paragraph of h-the magazine article. “I am nervous, truly nervous. As the taxi bounces southward through the trendier neighborhoods of Manhattan, I keep imagining the possible retorts of some incensed maître d’: ‘What kind of establishment do you think this is?’

‘How dare you insult me!’ “Do you think you can get in with that?” Second interesting psychological, uh, phenomenon displayed in this article is that when Feiler did screw up the courage to, uh, offer the bribe, he instinctively used indirect speech.

He, uh, folded up a twenty dollar bill, held it outside the line of vision, in peripheral vision, and he said to the maître d’, depending on the, uh, restaurant, things like, “I hope you can fit us in.” “Can you speed up my wait?” “Uh, I was wondering if you might have a cancellation.”

So indirect speech really does come naturally to us. And the, uh, third interesting, uh, part of the article was the outcome, which is that it worked every time.

(audience laughter)

He said, “I– We were seated i-in between two and four minutes to the astonishment of my date.” So what’s going on here? What are the intangible costs that drive people to indirect speech?

And here’s a theory, that the cost consists of relationship mismatches. In particular, that human relationships fall into a small number of types, each of which has a rule for distributing resources. Each has a distinct evolutionary and neurobiological basis.

Each applies by default to certain kinds of dyads, contexts, and resources. And this theory of relationship types I have, uh, borrowed, uh, maybe stolen from the anthropologist Alan Fiske, uh, who has worked out this theory in, uh, detail. And of the three natural relationship types, the first is dominance.

Its rule for distributing resources is, “Don’t mess with me.” The dominant, uh, uh, individual, uh, uh, commandeers resources, which, uh, are his prerogative. Presumably, they evolve from the dominance hierarchies that are, uh, ubiquitous among primates.

A very different relationship type is communality or communal sharing, whose rule for distributing resources is, “What’s mine is thine, what’s thine is mine,” which presumably evolved via mechanisms of kin selection and mutualism, and which are found by default among, uh, biological kin, uh, between spouses and among close intimate friends. And, uh, reciprocity, where the rule is, “You scratch my back, “I’ll scratch yours,” and follows the logic of “reciprocal altruism.” Now critically, behavior that is acceptable when one relational model is, uh, in force can be anomalous in another.

For example, at a party, you might go over to your boyfriend or girlfriend or husband or wife and help yourself to a shrimp off their plate. But you wouldn’t go up to your graduate supervisor or your dean and help yourself to a shrimp off, uh, his plate, because what you can get away with in a relationship of communality, namely, um, sharing without account, you can’t get away with in a relationship of dominance. Or if someone, uh, if a good friend had you over for a nice dinner party, uh, at the conclusion of the evening, you wouldn’t pull out your wallet and try to pay him for the cost of the meal because friendships obey the rules of communality, and it would be inappropriate to apply, uh, reciprocity.

Conversely, if you were at a restaurant, you wouldn’t go over to the owner and say, “Oh, that was a great meal. We had a really nice time, and, uh, we’ll have you back, uh, over sometime.” Uh,

(laughter)

Now, those are cases in which the relational model that’s in force is clear to both parties, and so any kind of breach would be the subject of, uh, uh, of comedy or amusement. But there are cases in which the relational model in force, uh, can be ambiguous, and in that case, any divergent understanding can lead to an unpleasant emotion for which we have a name, awkwardness. For example, there can be awkward moments in a workplace in which a, uh, say, a student doesn’t know whether they’re ready to address their advisor on a first-name basis, or they can invite them out after work for a beer.

It’s a well-known bit of, uh, conventional wisdom that close friends should never engage in a business transaction, such as one of them selling his car to the other. The very act of negotiating a price can, as we say, put a strain on the friendship, uh, because the rules of reciprocity are appropriate to neighbors or business partners, uh, or coworkers, but are not appropriate within the dyad of a close friendship. The, um, possible clash between a dominance model and the c– uh, communal sharing of sex, such as when a supervisor solicits sex from, uh, an employee or a student, leads to the infraction of sexual harassment.

And even the clash between two different kinds of communal relationship, friendship or sex, give rise to all of the tensions and complexities of dating, which are often worked out for our amusement in, uh, romantic comedies. Well, this gives rise to a social identification problem in which the social costs of awkwardness from a mismatched relationship type can duplicate the payoff matrix of a legal identification problem, such as bribing a police officer. So let’s work out what–

Why, uh, it feels so nervous to be given the assignment of bribing a maître d’. Now, the key thing in bribing a maître d’ is that the maître d’ ordinarily wields a relationship of authority over his restaurant fiefdom. He’s in charge.

He seats people, uh, where and when he pleases. The act of offering a bribe would change the relationship type into obligating him to seat you four at, at a table in exchange for accepting the bribe. So again, let’s start off with the rational actor model in which your only two options are to tender a naked bribe in so many words, uh, or not to bribe at all.

What is the, um, optimal strategy? Well, it again, it depends on whether you’re facing a corrupt maître d’ who would, uh, accept the b- the bribe, consummate a reciprocity relationship, and show you to a quick table without any emotional costs. On the other hand, you might be facing a scrupulous maître d’ who would maintain his dominance, rebuff the, uh, uh, offer of reciprocity by saying something like, uh, “How dare you insult me?

Do you think you can get in with that? What kind of establishment do you think this is?” Giving rise to the high emotional cost of awkwardness.

In contrast, if you just, uh, wait your turn and show up and, uh, are happy to be seated an hour and a half later, you are ceding the dominance relationship to the maître d’, but you pay the moderate cost of a long wait. So the high payoff of a quick table, high emotional cost of awkwardness have to be balanced against the guaranteed moderate cost of a long wait with no emotional awkwardness. However, if you had the third option, such as, uh, I was wondering if you might have a cancellation, then the, again, the corrupt maitre d’ could sniff out the bribe beneath the innuendo, consummate a reciprocity relationship, show you to a quick table.

A scrupulous maitre d’ could choose to let the comment pass without, uh, um, surrendering his claim to being a rational, sane, uh, user of the English language. Uh, by letting it pass, he maintains his dominance relationship. You still have the long wait, but you don’t have the awkwardness, and so the ambiguous bribe maintains the high payoff of a quick table that you get from bribing a corrupt maitre d’ with the relatively small cost of failing to bribe a scrupulous maitre d’ combined in one option.

And that is why the logic of plausible deniability applies not just to, uh, legal contexts like bribing a cop, but to everyday conte– uh, um, contexts such as bribing a maître d’. Okay, uh, I promised that the theory needed three parts, and here is the third part. There is still a problem with this theory, which is why do people resort to indirect speech even when uncertainty is low or absent?

Uh, when there is no actual identification problem because you know– you can anticipate that all listeners have the same values. To be concrete, you now are in possession of the knowledge that all Manhattan maîtres d’ are bribable. Nonetheless, I suspect that if you ever were to try to duplicate this stunt and try to jump the queue, you probably still would use indirect speech rather than, uh, blurting out your bribe as a, uh, quid pro quo.

And indeed, in our questionnaires where we asked people what kind of language would you, would you use in, uh, these circumstances, even when we tell people one hundred percent of the cops are bribable in this country that you’re driving through, there is– the corruption is so rampant that it is guaranteed that every cop will be, uh, bribable, uh, they still opted for indirect speech as the preferred, uh, method of addressing the officer. Or in cases where the, uh, there is no uncertainty on the part of the hearer as to what the speaker means. It’s hard to believe that any grown woman would really be fooled by the line about the etchings.

Nonetheless, there, I think you would agree that there is something that is much less, uh, awkward, much more, um, uh, comfortable about the line, “Would you like to see my etchings?” than, “Would you like to come up and have sex?” So the deniability is really, when it comes down to it, not all that plausible. As we say, it doesn’t pass the giggle test.

So why should a transparent innuendo still feel less awkward than an overture that is, as we say, on the record? Uh, I mentioned that some of the, uh, these Ambiguities have been, uh, are the subject matter of romantic comedies, and I’ll give you an example from the romantic comedy When Harry Met Sally that illustrates this point. At one point, the couple has just met each other.

Each of them has, uh, has, uh, another boyfriend or girlfriend. They’re confined to a long car trip, and Harry makes a, uh, suggestion that Sally interprets as a come-on, and she accuses him, “You’re coming on to me.” And Harry says, “What do you want me to do about it?

I take it back, okay? I take it back.” She says, “You can’t take it back.”

He said, “Why not?” She said, “Because it’s already out there.” He said, “Oh, geez, what are we supposed to do?

Call the cops?” It’s already out there.” So here’s the puzzle.

What is going on when we have the intuition that an overture is out there or on the record that makes it so much worse than a veiled overture that is implicated indirectly, even in cases where there’s no uncertainty, uh, Uh, real uncertainty as to what the overture means. Uh, and I- And our suggestion is that the key difference is a concept from, uh, economics, semantics, and game theory called common knowledge, which has to be distinguished from shared knowledge.

Common knowledge is also sometimes referred to as mutual knowledge or common ground. Now, in shared knowledge, A knows X and B knows X. In common knowledge, A knows X and B knows X, and A knows that B knows X, and B knows that A knows X, and A knows that B knows that A know, A knows X, uh, ad infinitum. Now, uh, this might seem implausible as a, uh, psychological hypothesis because you can’t fit an infinite, uh, set of nested propositions into a finite brain.

But as in other cases in linguistics in which we attribute infinite knowledge to a finite brain, uh, there is no paradox here because common knowledge can be captured in a recursive formula, a formula that makes reference to itself, Namely, X is common knowledge if people share the proposition Y, where the proposition Y is everyone knows X and everyone knows Y, Y being the proposition everyone knows X and everyone knows Y. That can fit into a finite brain, and it summarizes in a formula the whole set of infinite nested propositions without having to spell them out. Also, common knowledge can be ascertained perceptually from the circumstances in which X is transmitted.

You don’t literally have to have someone explain to you, “Well, you see, he knows that she knows that he knows that she knows,” and so on. There are certain cases, uh, such as, which I will get to, such as a salient public signal in which common knowledge can be generated instantaneously. Now, the reason that common knowledge has been of such interest to

(cough)

economists and game theorists is that it is necessary to find equilibria in coordination games. Well, what does that mean? Well, the original formulation of a coordination game comes from the, uh, philosopher Jean-Jacques Rousseau from his parable of the stag hunt.

Imagine that there are two hunters. Uh, they set out each day either to hunt rabbit, a very small payoff, or to hunt stag, a large payoff. The problem is, while each one individually can, uh, uh, catch a rabbit, it takes the two of them working together to fell a stag.

Uh, however, they have to decide at the beginning of each day which of the two quarry they aim for. Now, uh, and they cannot communicate. So why would a hunter decide to go out and hunt stag?

Well, he has to be assured that the other hunters will also decide that morning to hunt stag. Why would they decide to hunt stag? Well, th-they would only do it if they thought that the other hunters would do it.

But why would they think the other, other hunters would do it? Well, only if they thought that the first one would do it, uh, ad infinitum. In fact, any number of embedded, uh, states of knowledge, I think that he knows that I want to hunt stag, but, uh, I don’t know that he knows that I know that he knows that I know it, is rationally not sufficient to make the decision to hunt stag.

The tragedy would be that each one of them would settle for rabbit, whereas, uh, if only they could attain this common knowledge, they would both, uh, uh, be incentivized to hunt stag to-together. However, if there were some common signal, if you raise a flag, uh, saying, “I’m gonna– I’d like to hunt stag, uh, today,” and you– the other guy can see it, and you know that the other guy can see it, that can solve the problem. There are many examples of coordination games in everyday life that are solved by common knowledge.

Driving on the right or the left side of the road, that is a— uh, there’s no inherent advantage to driving on the right side of the road or the left side of the road, but there is an advantage to everyone driving on the same side of the road. In order— and it’s not enough to think, well, I— If you’re in a country where you don’t know what side it is, I think that the other guy is gonna drive on the right because he may wanna drive on the right, but he’s not gonna do so unless he thinks that everyone else is also gonna drive on the right. Which means that when you have a country that decides to switch sides, such as, uh, as Sweden did in nineteen sixty-nine, there has to be an enormous amount of publicity on when the switchover, uh, occurs.

I believe it was December thirty-first, midnight, nineteen sixty-nine. And that has to be so public that not only do you know it, but you know that all of your fellow citizens know it. Uh, paper currency and other paper assets are valuable to the extent that everyone knows that everyone else knows that everyone else knows that it’s valuable.

Which means that when you have threats to common knowledge, uh, as in hyperinflation, bank runs, and speculative bubbles, uh, you can get the value of a currency very quickly, uh, unraveling. Political action. Uh, why does– is freedom of assembly considered a fundamental right, and why are dictators the world over terrified of the prospect of people assembling peacefully?

Well, if you have a large population that all is disgruntled with a tyrant, then, uh, that is not enough for them to take action because any one of them standing up can be picked off. All of them standing up simultaneously would be too formidable a force for the tyrant to ignore, but people can’t afford to stand up unless they know that everyone else knows that everyone else know, uh, knows that, uh, there’s enough discontent to motivate them all to stand up. Which is why showing up at a public place where not only are you carrying your signs and yelling at the dictator, but you see everyone else around you also carrying the signs, and you know that they can see you and see each other, and that’s what empowers a, uh, crowd to challenge a, uh, the dominance of a dictator.

Nowadays, that can occasionally be done not just by physical premin– uh, presence, everyone seeing each other, but by an electronic equivalent. If everyone knows that a message is posted on, uh, Twitter where everyone else can see it, that can generate the same common knowledge. Now, relational models, that is, whether the rules in force in a given dyad are authority, uh, communality, or reciprocity are a kind of coordination equilibria.

You can kind of, uh, decide how you’re going to distribute the shrimp or the money or the, the sex without having to squabble over it every time. And so where both parties tacitly agree on how to deal with these common resources. So the question is, um, uh, which I alluded to earlier, how do people arrive at the common knowledge necessary to reach equilibria in coordination games?

Where– Well, salient public events are common knowledge generators. And perhaps the– my favorite illustration of this is the story of the emperor’s new clothes.

When the little boy cried out in public, “The emperor has no clothes,” he wasn’t telling anyone anything that they didn’t already know, anything they couldn’t see with their own eyeballs. But nonetheless, he was changing the state of their knowledge because at that moment, everyone knew that everyone else knew that everyone else knew that the emperor was naked. He generated common knowledge, and the common knowledge empowered them to challenge the dominance of the emperor.

The moral of the story being that direct speech, blurting something out, generates common knowledge. And that leads, I submit, to the third piece of the puzzle of innuendo and indirect speech. Namely, indirect speech provides shared knowledge, direct speech provides common knowledge, and relationships, which are a kind of coordination game, are maintained or nullified by common knowledge of the relationship type.

To be more concrete, if Harry says to Sally, “Would you like to come up and see my etchings?” And Sally says, “No.” Well, Sally knows that she has turned down a sexual overture.

She’s a grown woman. She’s no fool. Harry knows that she has turned down a sexual overture.

He’s no fool. But does Sally know that Harry knows? She could still wonder, well, maybe Harry thinks that I’m naive.

And does Harry know that Sally knows that he knows? Harry could be wondering, well, maybe Sally thinks I’m dense. There’s no common knowledge, and that allows them to maintain the fiction of a friendship.

The sexual overture can be rebuffed, and they can resume their life as friends. Whereas if Halle– Harry were to have said, “Would you like to come up and have sex?” Well, now Harry knows that Sally knows that Harry knows that Sally knows.

They cannot maintain the fiction of a friendship. And this is, we suggest, what’s behind that gut feeling that with direct speech, you can’t take it back. It’s out there.

Okay, this is a, uh, what we hope is a nice three-part theory. Is there any evidence that it is, uh, true? So James Lee and I tried to test it by coming up with a number of fictional scenarios.

We, um, uh, in them, a speaker utters a proposition that, uh, varies in level of directness. That’s our main independent variable. We ask the subject to put himself or herself in either the speaker’s shoes or the hearer’s shoes and rate the likelihood of various interpretations of the proposition and various emotional responses, uh, if they– as they imagine themselves to be in that scenario.

We had a bribe scenario where Kyle is pulled over for speeding, and he says to the officer, depending on the condition, “I’m very sorry, officer. I’ve really learned my lesson. From now on, you can be sure that I’ll be more careful,” with a fifty-dollar bill, uh, protruding.

That’s a, uh, proposition that is very, uh, vague and indirect. One notch more direct, “I’m very sorry, officer. I know that I was speeding and that I’ll have to pay for my mistake.”

“I’m very sorry, officer, but I’m actually in the middle of something right now, sort of an emergency, so maybe the best thing would be to take care of this here without going to court or doing a lot of paperwork.” And finally, “I’m very sorry, officer. If I give you a fifty, will you just let me go?”

We had a seduction scenario. Michael and Lisa are coworkers and good friends, have a nice dinner. Michael drives Lisa home.

When passing his apartment, he says, “Wow, I feel like we’ve been talking, uh, so much, but it’s only ten thirty.” Or, “My friend just emailed me those pictures from our trip to Europe that I was telling you about. Do you want to come over and have a look?”

Or, “You know, I have a really terrific view from my balcony. You can see the whole city, the lights, the oceans. “Would you like to come over and have a look?”” Or, “I find you really attractive,” and I enjoyed being with you tonight a lot.

Would you like to come over and have sex?” So that was the, uh, bluntness, uh, continuum in the seduction scenario. Then we also had a threat scenario where a, an unscrupulous professor threatens a student with the loss of a fellowship if she doesn’t work in his lab.

So, uh, I’ve mentioned the theory comes in three parts. Here are some tests of each of the three parts. The first one, and the most basic one, is, uh, speaks to the notion of plausible deniability.

Namely, does the linguistic variable of, uh, directness or indirectness actually correspond to a continuum of probability of the listener, uh, giving various interpretations to the, uh, proposition? That is, is indirect speech just a social ritual like please or thank you, or is it an uncertainty generator about the speaker’s intent? Uh, that was pretty easy to test.

Uh, in these various scenarios, we asked people, um, does the hearer, uh, understand what the speaker really means, namely interpreting it as a bribe, a threat, or a come-on? Uh, the x-axis is how overt the, uh, proposition was. Uh, perhaps not surprisingly, with overt speech, uh, one hundred percent of the subjects deemed, uh, it to be one hundred percent certain that it was a bribe, threat, or come-on, and with varying degrees, whoops, of, uh, indirectness, the likelihood that it was interpreted as a bribe, uh, fell accordingly.

Uh, prediction number two is that indirect speech should be perceived as less likely to get the speaker into legal trouble, such as, uh, being found guilty in a bribery trial or subject to a disciplinary hearing if the student blew the whistle on the professor. And, uh, once again, that is very much the perception of people as they imagine themselves in this scenario. That is, for example, if the driver went on trial for bribery, how likely is it that he would be convicted?

And indeed, uh, people thought it was a near certainty when the bribe was an overt quid pro quo and succeedingly less likely as it was veiled in thicker and thicker innuendo. Uh, prediction three is that the appeal of indirect speech should depend on the payoff matrix. That is, speakers should be a kind of rational actor in calibrating the optimal level of vagueness to the payoff structure of the situation.

Uh, in particular, the, um, as the model predicted, a direct bribe was rated as more attractive relative to an indirect bribe when there were fewer honest cops, that is Ten percent versus ninety percent. The fine for bribery was lower, two hundred and fifty versus ten thousand dollars. The cost of the ticket was higher, two thousand dollars versus two hundred dollars, or the cost of the bribe was lower, uh, twenty dollars as the going rate to bribe a cop versus a hundred dollars.

In fact, here you see the, uh, predicted directness according to the payoff structure that we varied across scenarios, and the y-axis shows the, uh, actual, uh, degree of direct, of directness that subjects judge to be most likely in the situation. Okay, part two, relationship negotiations. Prediction is that indirect speech should be seen as more respectful and better acknowledging the expected relationship with the hearer, namely, uh, affection in the case of communality, deference in the case of authority, collegiality also, uh, communality relationship.

That is that there should be the emotional payoff that the, uh, theory predicts. So this is, uh, in the bribe scenario, for example, how well did Kyle convey deference to the authority? Uh, it was judged that with overt speech, the Kyle was not being deferential, uh, at all.

The vaguer the speech, the more deferential– more deference they thought Kyle was showing to the police officer. Likewise, how, uh, affectionate was Michael toward Lisa? Uh, a, uh, blurting out a sexual, uh, proposition was considered to be not particularly affectionate, whereas being very vague was judged as considerate to her feelings.

Prediction number two, and this is a crucial one, is that just as in the case of the dollars and cents calculation for bribing a police officer, it’s critical if you take the fairly unromantic attitude that indirect speech is a strategy that maximizes expected payoffs, it’s only optimal if there are differing thresholds between a willing hearer and an unwilling hearer. We show that that was true for the police officer, And we had the, um, principle of proving guilt beyond a reasonable doubt as a kind of backup. The question is, in everyday life, in social situations which are unregulated, do we abide by a kind of, uh, proof of guilt beyond a reasonable doubt when interpreting ambiguous propositions?

In particular, um, let me skip that. Uh, imagine two different, uh, hearers for a sexual proposition, one of whom is an unwilling seductee. She has no interest in a sexual relationship.

How explicit does a sexual proposition have to be before she explicitly rebuffs it? Versus if she wants to reciprocate his sexual intentions, how explicit does his proposition have to be before she accepts it? Uh, and the, um, answer is that the, uh, for a willing, uh, uh, hearer, the threshold of, uh, explicitness has to be, um… sorry.

Let me just make sure I’m reading this correctly. This is the first time I’ve actually presented these data. That for a, um, an unwilling, uh, hearer to take offense, uh, the sexual advance has to be much more explicit than for a willing, uh, hearer to accept the proposition.

That is, a– an unwilling hearer gives the speaker a certain amount of benefit of the doubt before calling on it– him, whereas a willing, uh, hearer it will act with a much gentler hint. And it is exactly that asymmetry that is necessary in the model for indirectness to be the optimal strategy. Finally, and this is the hardest part to test, that direct speech, the proposition that direct speech, but not indirect speech, is perceived to generate common knowledge.

Uh, and we tried to test this, um, at least at first, in a very explicit manner by actually asking people to put themselves in the situation of, say, uh, Michael and Lisa, uh, and to assess what they would guess about the other party’s state of knowledge and the other party’s state of knowledge about their state of knowledge. And here’s how we tried to do it. We– since a set of nested propositions, if he knows that she knows that he knows that she knows can make you dizzy very quickly, we tried to lead them through it to these states of knowledge in small baby steps.

So, uh, put yourself in Lisa’s position. Uh, what is Lisa thinking at this point following the proposition from Michael varying in degree of explicitness? We gave them a rating scale from, “I’m absolutely certain that Michael was not asking me to have sex.

I’m virtually certain. I think that he probably wasn’t asking me. Did he just ask me to have sex, or was he just asking me to stay out longer?

I think he was probably asking me, uh, to have sex. I’m virtually certain he was asking me. I am absolutely certain.”

So that was the rating scale. Now, this is for the first-order hearer, namely the hearer judging the intention of the speaker. Second order.

Now imagine that Lisa has politely said she wants to go home. Put yourself in Michael’s position. What is Michael thinking now?

I’m certain that Lisa did not understand that I was, uh, asking her for sex. I’m absolutely certain that she underst-stood, and all of the gradations in between. We also ask then the state, the second-order knowledge on the part of the hearer.

Lisa knows that Michael is asking her to have sex. Now put her– yourself in her position. What is she thinking?

Michael thinks I didn’t understand he was asking me to have sex. I’m absolutely certain of that.” Again, with a set of statements varying in certainty, the other end of the continuum being, “Michael knows that I understood he was asking me to have sex.

I’m absolutely certain of that.” And through comprehension questions after the fact, we had pretty good confidence that people were able to follow us. And we pushed our luck with the third-order speaker knowledge.

Suppose that Michael does realize that Lisa knowingly turned down his invitation to have sex. Put yourself in Michael’s position. What is he thinking?

One end of the continuum, Lisa thinks that I didn’t understand that she turned me down for sex. I’m certain of that. Two, Lisa knows that I understood that she turned me down for sex.

I’m absolutely certain of that. And then the third order hearer state of knowledge. Suppose Lisa is certain that Michael knows she turned down his invitation to have sex.

Put yourself in Lisa’s position. What is she thinking? A set of options.

Michael understands that I turned him down for sex, but he doesn’t realize that I know he understands that. Other end, Michael understands that I turned him down for sex, and he realizes I know he understands that. Okay, here are the, the results as a function of the degree of explicitness of the proposition.

This time, however, I’m plotting it as the parameter, rather, that is the difference between the lines rather than the x-axis. So this line is for a, uh, proposition that is blurted out, this one for, uh, one that is nearly overt in– oops, indirect and very vague. And on the x-axis, we have the degree of embedding of, uh, states of knowledge.

That is, what did he actually mean? What does she think he means? He means, what does he understand, uh, as far as, uh, uh, her understanding of what he means, and so on.

For all of the indirect speech acts, the more deeply embedded the proposition is, the less and less certain the person imagines the actor in the scenario to be. So even if they think it’s pretty likely that it was a sexual come-on, they’re not so sure that he thinks that she thinks that he thinks that it was a sexual come-on. However, when it is blurted out in so many words, the certainty is pinned at a hundred percent.

No matter how many levels of embedding of he says that she says, he thinks that she thinks that he thinks that she thinks you throw at people, they are always confident that everyone knows with complete certainty uh that it was a sexual come-on when it is said uh directly, but not indirectly. So to sum up, um, innuendo, that is polite requests, veiled threats, and bribes, and sexual overtures poses a puzzle for psycholinguistics, and I have offered a solution in three parts. Plausible deniability, namely indirect speech, can minimize expected costs in identification problems when a speaker does not know a hearer’s values.

Relationship negotiation, there are emotional costs to mismatched relationship types, so we seek plausible deniability when exploring relationship changes. And common knowledge, indirect speech is used even when deniability is not so plausible, because direct speech generates common knowledge, which is the basis for coordinating, uh, relationship types. In, uh, another way of putting it is that plausible deniability of common knowledge is possible even when plausible deniability of the, uh, first-order meaning is not.

So some final thoughts on common knowledge. I believe that innuendo is a, uh, window into a much broader world of human social interaction. That is, it’s part of a family of phenomena in which we differentiate, uh, shared knowledge from common knowledge.

Even though we don’t, of course, have the explicit vocabulary to do so, we all tacitly sense that there is an enormous difference between something that everyone knows and something that everyone knows that everyone knows. And that that is, I will throw these out without explaining why, that I think that that is behind phenomena such as hypocrisy, piety, tact, taboo, conventional wisdom, mock outrage, the so-called Washington gaffe, uh, which is, uh, what is the definition of a gaffe in Washington? That’s when a politician says something that is true.

(laughter)

Uh, and political correctness. Uh, now how, uh, how could it be that so many emotionally fraught phenomena are governed by a phenomenon, common knowledge, for which we don’t have an everyday name? You don’t have people explicitly talk about common knowledge.

Well, I suggest that we do talk about it. And again, I will give a shout-out to a, uh, Berkeley linguist. George Lakoff has famously argued that much of our language is suffused with conventional conceptual metaphors.

That is, families of tropes all built around a common image, which capture a common understanding of an abstract, uh, phenomenon. And I think we do have a set of everyday idioms and figures of speech that, uh, indirectly refer to the phenomenon of common knowledge by alluding to the image of a conspicuous object or sound. And, uh, any idiom that makes such an allusion is a little indicator that we have a tacit sense of the difference between common knowledge and shared knowledge.

Such as, of course, the emperor’s new clothes, prototypically. Uh, it’s out there, as in when Harry Met Sally. The cat is out of the bag.

The bell can’t be unrung. Uh, fans of, uh, Seinfeld may remember the episode in which, uh, George says to a woman he’s been dating, uh, “I love you,” and she doesn’t respond, and Jerry says, “That’s a pretty big matzah ball hanging out there.” And of course, uh, the, uh, elephant in the room, the title of this lecture.

And with that, I will, uh, stop and take your questions. Uh, thank you very much.

(applause)

[01:02:17] ANDREW GARRETT:
Thank you, Steve, for a wonderful talk. Um, if you want to ask a question, I’m sure many people do, make them questions and make them brief and line up here at the mic.

[01:02:35] AUDIENCE MEMBER:
Do you see any difference in indirect speech in someone who’s offering a bribe as s-so– as opposed to someone who’s soliciting a bribe? Uh, for a cop who, who is wanting to get the fifty dollars as opposed to the, uh, motorist who, uh, isn’t sure that he should give it or not?

[01:02:56] STEVEN PINKER:
Yes, so is soliciting a bribe also done in indirect speech? So that we, we have not varied that. We haven’t, um, given people the scenario in which a, uh, corrupt officer solicits a bribe.

But, um, I suspect that, uh, much of that is done in indirect speech as well, anecdotally, based on, um, both on, on, um, uh, accounts that I’ve read of corrupt politicians who solicit bribes often in, in, uh, indirect speech, like, you know, we’re looking for contributions to our, uh, campaign fund, uh, uh, on experiences with people encountering bureaucrats and cops in countries that are rife with corruption that, you know, would you like to make a contribution to the, um, uh, po-Policeman’s Benevolent Association, uh, as opposed to, uh, give me some money or, or I will, uh, uh, make your life miserable. Um, this is not to deny that, that, uh, there are cases in which bribes are actually extorted in so many words, but, uh, but I suspect that as you hinted, uh, There’s a lot of indirect speech there as well.

[01:04:03] AUDIENCE MEMBER:
Uh, thank you for a great talk. So I just wanted to ask a, a question about the, the model you presented. Um, you talked about the sort of critical importance of that step function versus a, a linear function.

[01:04:14] STEVEN PINKER:
Uh, it– or it, um, yeah, it doesn’t li– it could be a- Sorry, not- Yes, you know, it could be a sigmoid.

[01:04:18] AUDIENCE MEMBER:
Right. It doesn’t have to be a step.

[01:04:20] STEVEN PINKER:
Any sort of nonlinear function.

[01:04:20] AUDIENCE MEMBER:
Right. Any nonlinear with– Yeah. Just wondering if you’ve thought about sort of ways you could empirically test that if, if you’ve actually looked at, at experiments that you could look for a sort of threshold effect versus just a, a linear or nonlinear.

[01:04:31] STEVEN PINKER:
Yes, you, you certainly could. We, um, and we tried to do so. We, um, the problem is our, our, our, uh, scatter of points, uh, where we plotted perceived likelihood of acting on the proposition as a bribe as a function of linguistic directness measured independently.

it was, um… We– the, the points weren’t orderly enough for us to claim with a high degree of confidence that there was a nonlinearity. Uh, but you’re right that that would be a good test of the model.

And we– if we ran ten times as many subjects then, uh, a-and had many more, uh, levels of indirectness between, uh, naked bribe and, uh, innocent remark, we’d be in a position to test that. But yes, absolutely, that would be a, that would be a good test.

[01:05:22] AUDIENCE MEMBER:
Um, I’ve seen some studies that, uh, that point to, um, creativity as being, uh, related to to– related with dishonesty. Uh, so in other words, uh, people who are primed, uh, to be dishonest, uh, tend to find creative solutions to a problem, uh, more easily. Huh.

And I was wondering, um, if, uh, there might be a similar relationship with tendency to use innu-innuendo versus direct speech. Um, and sort of as a follow-on to that, uh, whether, um, innuendo might, uh… I, I, I haven’t really got my head around the problem yet, but I’m just wondering whether it would lead to, uh, a feeling of greater cognitive ease or greater cognitive strain.

Uh, I’m, I’m thinking cognitive ease, but I don’t know.

[01:06:12] STEVEN PINKER:
Yes. No, I think those are, uh, insightful observations that we– In, in everyday life, we often have a feeling in making, um, judgments about individual differences, that there’s a kind of person who is, uh, perhaps a master at indirect speech. If at, at one extreme, they can be seen as, uh, slick or fast-talking or, uh, you know, oleaginous, and we distrust someone who’s too, too, uh, uh, slick, too unctuous, too, uh, fulsome, um, where you never know what, what they’re thinking.

Uh, we– One step short of that, we value someone who is tactful, which I interpret as keeping fraught propositions out of common knowledge. Um, we– The distinction between being tactful and being, uh, unctuous or slick is often a, a fine one. Then at the other end, uh, we, um, sometimes denigrate people who are too, uh, blunt.

That they can come off as, uh, boorish or uncultured or tactless. But then we can also value someone who’s kind of plain speaking, who can, uh, at least in some circumstances, uh, just get to the point without, uh… And as you note in the second part of your question, there can be a lot of cognitive work when we figure, you know, what, what, what did he really mean by that?

And you go home just wondering, well, you know, how should I interpret that? Uh, which is a kind of imposition, uh, a kind of, uh, lack of consideration. And I, I, uh, you know, I sense, and you can I think you can all confirm this, that a lot of our gossip of what kind of a person someone is, is often how they, uh, fit in those, a-along those two dimensions.

Do they have an optimal degree of tact without being, um, slick? Uh, do they have a refreshing directness without being insulting and boorish? Uh, a-and I bet a content analysis of a lot of our conversation would show us, uh, pinning people in that, uh, um, in that space.

[01:08:13] AUDIENCE MEMBER:
Yeah. No, I’m– I, I just think, uh, sort of a direction for future research might be, um, uh, trying to pinpoint sort of an optimal number of levels of ambiguity- Yes, uh, for, uh…

[01:08:26] STEVEN PINKER:
Yeah, in any way. A-absolutely. And I think, uh, people…

Uh, here’s another con-everyday concept that I think relates to that is when we talk about someone as being socially skilled, uh, versus, you know, graceless, clumsy, I think a lot of it is finding the optimum degree of indirectness so that you– a message gets across, uh, without anyone’s feelings being hurt. Without their fe– without their, as we say, their feathers being ruffled, their noi-nose being put out of joint. And because in social life, you can’t get away…

You, you know, we can’t all, um, cooperate perfectly on everything. There are inherent conflicts of interest in running any organization or for that matter, family, friendship, romantic partnership, and being able to negotiate those conflicts without challenging the relationship. That is, say, treating someone as a peon when they should be your peer, but you still really have to direct them to do something.

Or challenge someone while making it clear you’re still their friend, you still love them. Uh, that’s the very stuff of, uh, relationsh– uh, of dealing with relationships, and I think the socially skilled person is the one who finds that optimal level.

[01:09:34] AUDIENCE MEMBER:
Thanks. Thank you very much for a very in-interesting talk. Well, I have a very short question concerning, well, the seduction scenario you were talking about, uh, in your lecture.

So, uh, in linguistics, we know that euphemisms are used when there is a taboo for something. So there is a concept that is perceived as being unpleasant, and we know that sexuality is obviously a taboo concept. So my question is, do we actually have a choice between directness and indirectness in that seduction scenario?

W-we can, of course, be pretty direct, but this will create the impression that we are not well brought up or not brought up in a proper manner. Uh, yeah. Well, how will you relate this all to, uh, to your theory of

(cough)

the decision between directness and indirectness?

[01:10:18] STEVEN PINKER:
Well, I think that that’s right, but this is an, uh, in what I, in effect, I’ve been trying to do is to come up with a theory of what taboo really is. Uh, why you do have the impression that someone is better brought up if, uh, i-if they find the right level of indirectness instead of just blurting everything out. And it’s not mandatory.

There are people. I suspect there are people in this room who’ve had the experience of someone just blurting out, uh, you know, “I find you really attractive.” “Do you want to have sex?”

Uh, which, um, I, I’ve read interviews with, uh, authors who are famous for just trying it, and if they get rejected, um, nineteen times out of twenty, well, it’s worth it to score on the twentieth. And so, uh, if they have a thick enough skin to withstand the rejection and the, uh, the offense and, uh, because it’s worth it, then they really do opt for more direct speech. And, and then on the other end, uh, again, I’m going to appeal to Seinfeld, not because I’m a Seinfeld fanboy per se, but because romantic comedies r- uh, and situation comedies are often driven by exactly these sorts of conflicts.

But in the other direction from the author who just blurts it out every time with a thick skin for rejection, you’ve got a scene where, uh, George Costanza, uh, is asked by his date at the end of the date, “Would you like to come up for coffee?” And he says, “Oh, no, I have to be at work at nine in the morning, and caffeine would keep me up.” And then walking away, he says, “Oh my God,” coffee doesn’t mean coffee.

Coffee means sex.” Uh, so that’s the danger for being– with being too indirect. Uh, so I think there is, there, there still exists that range, and, uh, I think most people do find some intermediate point.

But, uh, but that is a, a range of options that is available to people.

[01:12:04] AUDIENCE MEMBER:
Hi. Um, woo. Hi.

So I’m wondering if there are– if you’ve identified any linguistic cues for kind of the opposite, as opposed to, um, understanding indirect speech. What if someone is trying to be direct, but then they are misunderstood as having tried to convey an innuendo or a euphemism? Mm-hmm.

And, um, what factors might cause the hearer to, to think that? Factors within the speaker’s speech?

[01:12:40] STEVEN PINKER:
Yes. Well, our, our data suggests that, that, that when you’re direct, at least by our definition of directness, which we confirmed with ratings by, uh, separate groups of subjects, uh, that almost by definition, direct leaves very little room to the imagination. Now, there are, uh, so, um, you know, there are cases, and they’re also s-subject to comedy, where you’ve got the really dense person who doesn’t understand something that’s even very obvious.

There’s the joke about the two psychiatrists, and one of them says to the other, “Have a nice day,” and the other one thinks, “I wonder what he meant by that.” Uh, or, you know, “Do you wanna come up and have sex?” “Oh, what could he possibly mean?”

So I think those are the subject of a-another kind of comedy, but generally directness is harder to misinterpret.

[01:13:22] AUDIENCE MEMBER:
Can I, can I clarify a little bit? With an example?

[01:13:24] STEVEN PINKER:
Yeah.

[01:13:24] AUDIENCE MEMBER:
Like, for example, in a situation where, uh, you… Well, okay, so in a situation where, um, it’s very common for women to say no to sexual overtures, but it’s interpreted as being, uh, playing hard to get, that kind of thing.

[01:13:46] STEVEN PINKER:
Oh, so it’s like- what part of-

[01:13:48] AUDIENCE MEMBER:
this idea of that- Yes, you’re like playing a game or something. Um, or for someone to say, like, “I, I want to be friends with you,” and that being like, “Oh, she wants to get friendly with me.”

[01:14:05] STEVEN PINKER:
Like- That’s right. No, that’s true, and that is actually a significant phenomenon in, um, cases of sexual harassment, of, uh, acquaintance rape, and so on, uh, and in workplace policies. A few years ago, I remember, uh, Safeway Supermarkets told– instructed their cashiers, um, with every customer, look them in the eye and smile, uh, as a way of making Safeway seem to be a nice, friendly place.

And they had to abandon that policy because the male customers interpreted that as a sexual come-on by the, uh, the female ca-cashiers. Uh, and there are lots of data that the threshold there, there’s a whole sexual asymmetry in sexual come-ons that, that, uh, um, the way a male reacts to a, a female proposition and vice versa show enormous, uh, as-asymmetries. And you’re right that there can be…

And because of things like, you know, what part of no don’t you understand, uh, the, the threshold can be… So, uh, so that would be a good example. The threshold can be so f– set so far to the right that that would be a case indeed, good, where, uh, direct speech is interpreted as indirect speech.

And presumably, this in– would be explained again by the, uh, by the payoff matrix. For a man for whom sex has such a high payoff and being turned down has such a low cost, then the, uh, optimum cutoff could be one in which, uh, nothing short of, you know, physical resistance would be interpreted as, uh, as, as a genuine no. So yeah, that’s a, that’s a great example.

[01:15:41] AUDIENCE MEMBER:
Thank you. My question is a little bit related. Uh, it seems like a lot of these scenarios could be interpreted by, in some cases, as being part of a bargain– a broader bargaining scenario.

[01:15:56] STEVEN PINKER:
Mm-hmm.

[01:15:56] AUDIENCE MEMBER:
And I was wondering, have you… Do you… Do you also look into s-sort of at what point people, people perceive something as, as part of, as part of a negotiation and at what point it’s, it’s simply a, a yes or no.

[01:16:13] STEVEN PINKER:
Yeah. No, I, I think that’s, that these are, uh, prototypical cases of negotiation. Uh-huh.

Uh, and in negotiation, there is a nego– negotiation is a coordination problem in the sense that there is a often a range of solutions that would leave both the– both sides better off than if they walked away from the deal, and they, each one wants to find the optimal point within that range of solutions. And often, common knowledge of a salient solution is essential to, um, uh, to consummating, uh, an effective negotiation. Uh, and so that range of, uh, intermediate positions is usually what is being explored in a negotiation, and I think that’s true of, uh, uh, negotiating a relationship as well as negotiating a price.

So yeah. Thank you.

[01:17:05] AUDIENCE MEMBER:
Thanks for a great talk. Uh, so two things. Uh, firstly, I wanted to thank you for explaining my lack of success with American women.

(laughter)

Secondly, uh, I thought it was interesting that, uh, so primarily you focused on the costs and benefits to the speaker-

[01:17:24] STEVEN PINKER:
Yes.

[01:17:25] AUDIENCE MEMBER:
of making shared knowledge common knowledge, And it seems like there’s an interesting angle to looking at the same thing from the perspective of the hearer.

[01:17:35] STEVEN PINKER:
Mm-hmm.

[01:17:36] AUDIENCE MEMBER:
And just to give an example of that, say in the case of the corrupt police officer, there may be costs to the police officer of explicitly appearing corrupt in the sense, for example, the police officer may be subject to disciplinary sanctions if it, if it’s clear in the transaction-

[01:17:54] STEVEN PINKER:
Mm.

[01:17:54] AUDIENCE MEMBER:
that the police officer is corrupt. And then I guess similarly in the case of the, uh, dating scenario, the woman’s morals may be called into question if if it’s, if it’s explicit that she easily accepts the sexual invitation. So I’m just curious if you’ve looked at that cost-benefit equation from the perspective of the hearer as well as the speaker.

[01:18:20] STEVEN PINKER:
Yes. Uh, although not experimentally, but I do explore exactly that question in, uh, my book, The Stuff of Thought, from which this, uh, a lot of this material is taken. And indeed, the, um, a fact that I just threw out, uh, without a lot of justification, namely that there are different thresholds for the speaker and the hearer as to how direct a proposition has to be before they’ll act on it, would itself need an explanation, and indeed the explanation would come from the payoff matrices for– payoff matrix for the hearer.

That is, what are the costs, the potential costs of accepting versus turning down the proposition? And because, uh, those payoffs would be different from the, for the hearer than for the speaker, that would result in the threshold for action showing the discrepancy that is absolutely necessary for, um, indirect speech to be selected. That’s probably why in cases where, um, the payoffs are pretty much identical for speaker and hearer, that’s when it’s just fine to, to kind of blur it out.

You don’t need a lot of, of, um, uh, indirect speech and beating around the bush. in a, you know, married couple when it comes to sex, in a, um, a business transaction where they’re, you know, you’re actually buying something from a, a vendor where everyone benefits to a known degree by the transaction, then, uh, there, you don’t have that, um, slippage between the two sets of interests that define the zone in which indirectness is optimal. So I th-that–

Yes, that’s absolutely right. And for the, for the, uh, there, there’s the built-in asymmetry of male sexuality and female sexuality. In fact, the, the, the extreme case is the, uh, famous experiment by, uh, Elaine Hatfield, in which she, um…

It was actually a wonderful experiment on direct speech, where she hired con– uh, confederates to walk around campus, uh, select a member of the opposite sex and say, uh, “Would you like to have sex with me tonight?” And the, uh, what she found was that seventy-five percent of the men said yes, and, uh, zero percent of the women said yes. So, and they’re, uh, and I think they’re perfectly, uh, well-supported explanations from evolutionary psychology on why that should be true.

Um, so there’s, uh, that– and that’s by the way, one of the reasons why I didn’t make any of these scenarios gender-neutral. When it comes to sexuality, it’s just not gender-neutral. Um, but in addition, there are probably some payoffs that are inherently different between, that are inherently symmetrical on two sides, such as if on either side, if you are rebuffed, that is a kind of, uh, denigration of your, uh, attractiveness, and that works for both, uh, superimposed on a male-female difference.

That’s an aspect that’s the same for, for male and female, that kind of pride insult thing. And, uh, and there are other cer– aspects that could be more s– more or less symmetrical.

[01:21:15] AUDIENCE MEMBER:
Yeah, thanks. Or, or I guess you could say, in other words, the type of the relationship to some degree dictates whether the speaker needs to take into consideration the cost benefits of the hearer as well as their own.

[01:21:29] STEVEN PINKER:
Yes, yeah, very much so. Yeah.

[01:21:33] AUDIENCE MEMBER:
Uh, hi. Um, my question was kind of related. So your example, um, most of your examples in this talk have been about the initial sort of overture, the initial question, like the veiled offer of a bribe or the veiled offer of sex.

But, like, if you, you know, encounter a corrupt cop or a willing sexual partner, it eventually has to like, cross those degrees of indirectness up to the point where you actually hand over the fifty or actually have sex.

[01:22:05] STEVEN PINKER:
Yes.

[01:22:06] AUDIENCE MEMBER:
So I was wondering if you had conducted research or were planning to conduct research on how, um, And under what conditions that step up of ambigu-ambiguity, um, reaches the level of full directness?

[01:22:19] STEVEN PINKER:
Yes, no, that, that, uh, and, um, we, we do discuss that scenario in, um, Martin and James and I in our, uh, PNAS paper of, uh, although we d- we never designed an experiment around it. But yes, one other strat-strategy, which is, in fact, uh, as you suggest, more realistic than you throw out a proposition, uh, gambling on how your speaker will react to it, giving it your best shot, is that you could escalate the degree of directness. Gee, officer, isn’t it a nice day?

Uh, gee, I guess you do have a problem with with speeders here. Uh, I can understand how, uh, uh, you’d be concerned about this. Um, I gue– I’ve heard that policemen have a benevolent association.

Uh, how is it doing lately? Anyway.

(laughter)

Um, you have,

(laughter)

you know, you could up it, up it, up it. And likewise, you know, I don’t even have to play it out for you what it would be like in a seduction scenario, uh, likewise in a threat scenario. And that way, um, you could– the speaker could calibrate the degree or could sort of titrate the degree of directness until they get the first signs of a, uh, rejection or, uh, the deal was consummated.

And I agree with you that in real life, that, um, incremental strategy, moving your way up the line of explicitness may even be more common than, uh, throwing out your best guess as a one-shot offer.

[01:23:39] AUDIENCE MEMBER:
Um, with respect to the, um, interests of the hearer- Mm-hmm. -like, they… A corrupt cop might want to accept a bribe, but, like, might test out and just like make it a little bit more explicit than the offer of the bribe, waiting for them to offer an even more explicit bribe so that at every given point, the, um, level of ambiguity is comfortable for everyone.

Is that kind of-

[01:24:01] STEVEN PINKER:
Ab-absolutely, yes. So it could be be– and because, as in, in response to an earlier question, I mentioned that soliciting a bribe also uses indirectness. Uh, and indeed, that could be done incrementally as well.

Uh, and so you could have, and I suspect again, probably a lot of romantic comedies explore this in the sexual case, probably a set of, uh, potentially escalating, uh, propositions in both directions could at some point meet in the middle. I think Much Ado About Nothing, uh, the, uh, the banter, the, the sexually charged banter has much of that character. Uh, and in, uh, also in Henry V, uh, uh, the denouement at the end.

[01:24:43] ANDREW GARRETT:
So let’s, uh, thank Steve again for a great talk.

(applause and cheering)

[01:24:46] STEVEN PINKER:
Thank you very much. Thank you. Thank you.