The Gricean Screw

Speranza

In her online study, "The Griceian Turn", C. J. Lee writes as per below. Some running comments.

Cheers,

----


The Gricean Turn in Psychology
Carole J. Lee
University of Michigan, Ann Arbor
Psychologists’ work on conversational pragmatics and judgment suggests a
refreshing approach to charitable interpretation and theorizing. This charitable
approach—what I call Gricean charity—recognizes the role of conversational
assumptions and norms in subject-experimenter communication. In this
paper, I outline the methodological lessons Gricean charity gleans from psychologists’
work in conversational pragmatics. In particular, Gricean charity
imposes specific evidential standards requiring that researchers collect
empirical information about (1) the conditions of successful and unsuccessful
communication for specific experimental contexts, and (2) the conversational
norms governing communication in experimental contexts. More
generally, the Gricean turn in psychological research shifts focus from attributional
to reflexive, situational explanations. Gricean charity does not primarily
seek to rationalize subject responses. Rather, it imposes evidential
requirements on psychological studies for the purpose of gaining a more
accurate picture of the surprising and muddled ways in which we weigh
evidence and draw.
Keywords: Gricean charity; methodological rationalism; interpretation; principle
of charity; cognitive psychology; conversational pragmatics; heuristics
and biases; reflexive analysis
Traditional accounts of charitable interpretation that rely on norms of
rationality to guide interpretation have typically invoked rules of logic
and probability, as well as principles of evidence or justification—while
overlooking norms governing the social and communicative relationships
between the interpreter and interpreted.
Psychologists working on conversational pragmatics and judgment have
observed the same oversight in their field: researchers, especially those
Philosophy of
the Social Sciences
Volume 36 Number 2
June 2006 193-218
© 2006 Sage Publications
10.1177/0048393106287235
http://pos.sagepub.com
hosted at
http://online.sagepub.com
193
Author’s Note: Thanks to Elizabeth Anderson, James Joyce, Peter Railton, Paul Thagard,
Norbert Schwarz, Paul Roth, Alison Wylie, James Bohman, and the audience members at the
Seventh Annual Philosophy of Social Science Roundtable for comments that have sharpened
up this paper.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
from the heuristics and biases tradition, who tend to argue that subjects are
systematically irrational, have neglected to consider how social or conversational
norms may influence subjects’ interpretations of and communications
with experimenters.1 In response, psychologists such as Norbert
Schwarz, Denis Hilton, and Gerd Gigerenzer have invoked Paul Grice’s
principles of cooperative communication to attribute alternative interpretations
of experimental tasks to subjects—interpretations for which subject
responses may be said to be “conversationally rational.”2
My account of charitable interpretation broadens traditional charitable
accounts by recognizing conversational norms as rational principles of conversational
inference. I will call the general method of using conversational
principles to guide interpretation Gricean charity. Gricean charity provides a
naturalized account of charity which looks to facts about natural language and
communication in the interpretation of subject responses. I take psychologists’
work in conversational pragmatics as primary exemplars of Gricean charity at
work. This work demonstrates that a broader perspective on rationality and the
nature of subject-experimenter communication imports specific evidential
requirements on psychological studies: namely, that subject responses be
interpreted in light of empirical information about (1) successful and unsuccessful
communication in specific experimental contexts, and (2) the conversational
norms governing communication in experimental conditions.
The genealogy of this charitable approach may be traced back to “the
presentation problem” faced by Ward Edwards.3 In order to test human performance
on rational choice tasks, psychologists have to put the task and
options into words. However, turning the decision task into a word problem
adds an additional level of complexity for both the subject and researcher.
For the subject, natural language expressions are often ambiguous, and may
support any number of meanings. So, the subject must interpret the intended
meaning of the stated task, and provide a response under that interpretation.
As a result, the subject’s choice behavior is influenced by her interpretation
of the experimental task. Edwards’s observation may be captured by a more
general Davidsonian lesson: any psychological theory on human judgment
194 Philosophy of the Social Sciences
1. Denis J. Hilton, “The Social Context of Reasoning: Conversational Inference and
Rational Judgment,” Psychological Bulletin 118, no. 2 (1995): 249.
2. Norbert Schwarz, “Judgment in a Social Context: Biases, Shortcomings, and the Logic
of Conversation,” Advances in Experimental Social Psychology 26 (1994): 123-62.
3. Ward Edwards discusses this kind of problem, though not under the rubric “presentation
problem.”Ward Edwards, Harold Lindman, and Lawrence D. Phillips, “Emerging Technologies
for Making Decisions,” in New Directions in Psychology 2 (New York: Holt, Rinehart &
Winston, 1965).
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
“must include a theory of interpretation” about subjects’ beliefs about the
experimental task.4
In the first part of this paper, I will lay out my account of Gricean charity.
First, I will argue that conversational norms are indeed relevant to the questionnaires
and surveys used in psychological testing. To illustrate, I will
reinterpret some portions of Kahneman and Tversky’s Linda questionnaire
in light of Gricean conversational maxims, and use this reinterpretation to
rationalize subject responses. This analysis serves to highlight the methodological
lessons of the Gricean turn in psychological research. In the second
part of this paper, I will consider and respond to methodologically motivated
objections to Gricean charity. In the course of responding to these objections,
I will argue that Gricean charity generates new psychologically interesting
questions, phenomena, and methods, without harboring scientifically
illegitimate forms of bias.
1. Gricean Charity
A. Questionnaires as Forms of Cooperative
Communication
Conversational pragmatics invokes normative principles of communication
to account for how subjects arrive at their interpretations of the experimental
task. For an account of these norms, researchers have turned to Paul
Grice’s account of cooperative communication. According to Grice, cooperative
communication aims to use language efficiently and effectively to
further a common goal or set of goals. Communication is said to be rational
insofar as it conforms to conversational principles that are themselves
instrumental in furthering these cooperative ends.5 The most general principle,
the Cooperative Principle (CP), directs conversants to “[m]ake your
conversational contribution such as is required, at the stage at which it
occurs, by the accepted purpose or direction of the talk exchange in which
you are engaged.”6 Grice analyzes CP into the following general principles:
Quality: (1) do not say what you believe to be false, and (2) do not say that for
which you lack adequate evidence.
Lee / The Gricean Turn in Psychology 195
4. Donald Davidson, “Belief and the Basis of Meaning (1974),” in Inquiries into Truth and
Interpretation (New York: Oxford University Press, 2001), 147. The italics are mine.
5. Paul Grice, “Logic and Conversation,” in Studies in the Way of Words (Cambridge,
Mass.: Harvard University Press, 1989), 28.
6. Ibid., 26.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
Quantity: (1) make your contribution as informative as is required for the current
purposes of the exchange, and (2) do not make your contribution more informative
than is required.
Relevance: make your contribution relevant.
Manner: (1) avoid obscurity of expression, (2) avoid ambiguity, (3) be brief (avoid
unnecessary prolixity), and (4) be orderly (provide information in a sequentially
accessible way).
The conversational maxims provide interpretations of utterances of natural
language expressions. Natural language expressions often imply multiple
meanings, or imply meanings that are not captured by the literal
statement as expressed. The conversational maxims provide a way of inferring
a speaker’s intentionally implied meaning—that is, the conversational
implicature—even when this meaning goes beyond the literal meaning of
what she has said.7 Conversational implicatures are calculable in the sense
that, for every putative implicature, it is possible to construct an inductive
argument showing how the implicature follows given the literal meaning of
the utterance, mutually recognized facts about the particular context of
communication, and the conversational maxims.8 If the hearer discovers
that any of these premises are false, then the implicature is cancelled.
196 Philosophy of the Social Sciences
7. The special case Grice considers is one in which a conversant flouts a conversational
maxim. Here, a conversant provides a contribution that, when taken literally, violates one of
the maxims. For example, let’s say Dick and Jane have plans to meet with a third person, Tom,
who is very late (and usually so). Dick asks, “What happened to Tom?” Jane responds, “Tom’s
watch must operate counterclockwise.” Jane’s contribution, when taken literally, is false—she
knows Tom’s watch does not operate counterclockwise. Jane flouts the maxim of quality
which enjoins her to provide true or well-founded contributions. In order to construe Jane’s
contribution as conforming to the maxim of quality, we infer that she intends to imply something
beyond the literal meaning of her statement—something that is actually true—perhaps
that “Tom has profound trouble keeping track of time.” Jane expects Dick to be able to infer
this implied meaning, given what he knows about Jane, the conversational context, and the
conversational maxims. The interpretation of a wide range of linguistic phenomena (such as
figures of speech, hyperbole, metonymy, irony, and metaphor) may be subsumed under the
more general problem of interpreting implicatures.
8. Stephen Levinson analyzes the inductive argument constructed by the hearer in the following
way:
(i) S has said that p.
(ii) There’s no reason to think S is not observing the maxims or at least the co-operative
principle.
(iii) In order for S to say that p and be indeed observing the maxims or the co-operative
principle, S must think that q.
(iv) S must know that it is mutual knowledge that q must be supposed if S is to be
taken to be co-operating.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
Psychologists working in conversational pragmatics have argued that
questionnaires can be understood as forms of cooperative communication.
As Hilton points out, psychological experiments and surveys used to test
cognitive competence “are forms of social interaction between the experimenter
and participant, which invariably involve communication through
ordinary language.”9 Experimenters provide subjects information via instructions,
questionnaires, and surveys; and subjects provide responses in the
form required. As in other conversational settings, these forms of information
exchange are carried out in ordinary language.
Researchers in conversational pragmatics have argued that experimenters
and subjects share a mutual recognition that their efficient, effective
acts of communication serve mutually understood goals. Subjects
know that experimenters aim to collect data—namely, subjects’ responses
to predesigned questions and experimental tasks. Since subjects cannot ask
for clarification or paraphrases of the question, the experimental conditions
encourage them to assume that the meanings of the questions and tasks are
self-evident.10With this knowledge, subjects can expect that experimenters
expedite this process by stating their questions and tasks clearly, concisely,
and sufficiently. That is, subjects may reasonably expect questionnaires and
surveys to conform to the maxims of quantity, relevance, and manner. For
subjects, experiments are conducted by their academic and epistemic
authority figures: much current research is conducted on undergraduate
psychology students by graduate students and professors. Because of the
asymmetry in authority, expertise, and knowledge between experimenters
and subjects, subjects may reasonably expect experimenter-provided information
to be especially truthful and well supported, in accordance with the
maxim of quality.11
Experimenters know that subjects characteristically participate in the
experiments to fulfill requirements for their introductory psychology courses,
or for small monetary rewards. Coming into a psychological experiment,
subjects come with prior expectations about the experimenter’s goals. In
Lee / The Gricean Turn in Psychology 197
(v) S has done nothing to stop me, the addressee, from thinking that q.
(vi) Therefore, S intends me to think that q, and in saying that p has implicated q.
Stephen Levinson, “Conversational Implicature,” in Pragmatics (Cambridge: Cambridge
University Press, 1983), 113-14.
9. Hilton, “The Social Context of Reasoning,” 249.
10. Herbert H. Clark and Michael F. Schober, “Asking Questions and Influencing
Answers,” in Questions about Questions: Inquiries into the Cognitive Bases of Surveys, ed.
Judith M. Tanur (New York: Russell Sage Foundation, 1992), 26.
11. Hilton, “The Social Context of Reasoning,” 254.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
particular, they know that the experimenters have designed special questions,
with the goal of evaluating and explaining their answers. In order for their
answers to be constructive or relevant toward the experimenters’ goals, the
subjects are expected to answer sincerely, in conformance with the maxim
of quality; and, in order to help the experimenters carry out this goal efficiently
(given the number of subjects involved), they are expected to answer
in conformance with the maxims of relevance, quantity, and manner.
Even when subject responses take the form of checking boxes, circling
multiple-choice options, or ranking outcomes, these maxims still apply.
Even with such regimented forms of response, the maxim of quality dictates
that subjects provide honest rather than dishonest answers. The maxim
of relevance directs subjects to answer the experimenter’s intended question,
and not a different question that may be more amusing to entertain.
The maxim of quality enjoins subjects to provide sufficient answers by
answering all questions. And, in accordance with the maxim of manner,
subjects are expected to provide unambiguous, clearly marked, and quickly
deciphered responses: for example, it would be inappropriate for a subject
to provide a long-winded essay in answering a multiple-choice question.
In the experimental context, subjects can rely on special kinds of clues
in interpreting the experimenters’ intended meanings—clues such as the
wording of the task, the questionnaire’s previous questions, the formal
structure of the questionnaire, and interactions with and assumptions about
the experimenter.12 Such evidence grounds key implicatures about the
meaning of the experimental task or question.
B. The Linda Problem
To see how conversational norms and assumptions can inform our interpretation
of subjects’ responses, consider Daniel Kahneman and Amos
Tversky’s Linda problem:
Linda is 31 years old, single, outspoken, and very bright. She majored in philosophy. As
a student, she was deeply concerned with issues of discrimination and social justice,
and also participated in antinuclear demonstrations.
Please rank the following statements by their probability, using 1 for the most probable
and 8 for the least probable.
198 Philosophy of the Social Sciences
12. For a nice review of this research, see Norbert Schwarz, Cognition and Communication:
Judgmental Biases, Research Methods, and the Logic of Conversation (Mahwah, N.J.: Lawrence
Erlbaum, 1996).
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
Linda is a teacher in an elementary school.
Linda works in a bookstore and takes Yoga classes.
Linda is active in the feminist movement.
Linda is a psychiatric social worker.
Linda is a member of the League of Women Voters.
Linda is a bank teller.
Linda is an insurance salesperson.
Linda is a bank teller and is active in the feminist movement.13
According to probability theory, the probability of two events A and B is
equal to or less than the probability of each of its conjuncts: p(A + B) !
p(A) and p(A + B) ! p(B). If we identify Linda’s being active in the feminist
movement as event A, and Linda’s being a bank teller as event B,
subjects should rank the probability that Linda’s both a bank teller and
active in the feminist movement (A and B) as equal or lower than the ranking
for A or B considered alone. They found that the vast majority of statistically
naïve and statistically sophisticated subjects rated the conjunction
of events as more probable than the conjunct, in violation of the conjunction
rule.
However, it isn’t clear that subjects can be said to have violated the conjunction
rule, if we reinterpret the questionnaire in light of Grice’s conversational
maxims.14 The only information required to rank the probabilities
in accordance with the conjunction principle is two particular outcomes: the
outcomes “Linda is a bank teller” and “Linda is a bank teller and is active
in the feminist movement.” The rest of the cover story, such as the personality
description and the other outcomes, are irrelevant, unnecessary, and
superfluous.15 The experimenters’ question, as stated, violates the maxims
of relevance, quantity, and manner.16 These violations have implications for
Lee / The Gricean Turn in Psychology 199
13. Amos Tversky and Daniel Kahneman, “Judgments of and by Representativeness,” in
Judgment under Uncertainty: Heuristics and Biases, ed. Daniel Kahneman, Paul Slovic, and
Amos Tversky (New York: Cambridge University Press, 1982).
14. The line of argument here is due to Ralph Hertwig and Gerd Gigerenzer, “The
‘Conjunction Fallacy’ Revisited: How Intelligent Inferences Look like Reasoning Errors,”
Journal of Behavioral Decision Making 12 (1999): 275-305.
15. Jonathan E. Adler, “Abstraction Is Uncooperative,” Journal for the Theory of Social
Behavior 14, no. 2 (1984): 165-81.
16. Notice that the personality description and the rest of the outcomes are relevant to the
question of how the outcomes “Linda is a bank teller” and “Linda is a bank teller and is active
in the feminist movement” rank relative to the other outcomes. However, the personality information
and other outcomes are not relevant to the question Kahneman and Tversky are primarily
interested in: namely, how the outcomes “Linda is a bank teller” and “Linda is a bank
teller and is active in the feminist movement” rank relative to each other.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
the questionnaire’s validity. If subjects enter the experiment under the
assumption that experimenters are cooperative communicators, they would
rule out this interpretation, since it construes experimenters as violating
conversational norms.
Subjects are thus put in the position of interpreting the meaning of the
experimenters’ question so that it conforms to the conversational maxims.
Such an interpretation must render all of the information provided by
experimenters as useful and relevant to solving the intended question. One
way to render the extra personality and outcome information useful and relevant
is to take the experimenters as asking something other than a mathematical
probability question. The term probable is polysemous: it can also
be interpreted as meaning plausible, having an appearance of truth, or reasonable
in light of the evidence.17 Subjects, faced with the problem of inferring
which of these meanings experimenters implicate, might interpret the
question as a plausibility problem, where an outcome is said to be more
“plausible” insofar as it has something to speak in favor of it. Under this
interpretation of the question, it is not incorrect to judge the conjunction of
events as more plausible than its conjunct: given the personality description,
the outcome “Linda is a bank teller” has nothing to speak in favor of
it; however, the outcome “Linda is a bank teller and is active in the feminist
movement” does have something to speak in favor of it, since we expect
Linda’s commitment to liberal values would be expressed in her choice of
occupation and/or hobbies.18
The notion of plausibility might also be understood as having something
to do with how well the personality information might explain the different
outcomes: the personality description provided has more explanatory
strength when it comes to explaining why “Linda is a bank teller and active
in the feminist movement,” as opposed to explaining why “Linda is a bank
teller.” If subjects interpret probability to mean something like degree to
which they can be explained, then subjects’ responses cannot be said to be
incorrect. Not only does the personality description better explain the conjunction
than the conjunct, but—as Kahneman and Tversky themselves
200 Philosophy of the Social Sciences
17. Hertwig and Gigerenzer, “The ‘Conjunction Fallacy’Revisited.” See also Gerd Gigerenzer,
“On Narrow Norms and Vague Heuristics: A Reply to Kahneman and Tversky (1996),”
Psychological Review 103, no. 3 (1996): 284-308.
18. An alternative interpretation of “plausibility” might be provided by the notion of “conceptual
coherence.” For a model of conceptual coherence that bears on how stereotypes and
individuating information can inform each other, see Ziva Kunda and Paul Thagard, “Forming
Impressions from Stereotypes, Traits, and Behaviors:A Parallel-Constraint-Satisfaction Theory,”
Psychological Review 103, no. 2 (1996): 284-308.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
point out—best explanations and most probable outcomes often have an
“inverse relationship”:19 the value of an explanation can be improved by
increasing its content and scope, even though the probability of its truth is
reduced thereby.20
C. Methodological Implications of the Gricean Turn
Gricean charity provides an interpretation of subjects that rationalizes
their apparently irrational responses. Grice’s conversational maxims suggest
that Kahneman and Tversky’s statement of the Linda problem misleads
at least some subjects to solve a different problem—a problem for which
subjects’ responses are rational. Accordingly, Kahneman and Tversky are
not entitled to describe subjects as providing irrational responses.
This is the first methodological lesson I draw from the Gricean turn in
psychological research: the conclusions of any psychological study can
only be as valid as its questionnaires and surveys. If experimenter communications
violate the conversational maxims, leaving greater room for the interpretation
of unintended implicatures, then we cannot take subjects’ responses
at face value—much less generalize or explain them. The Gricean turn
suggests researchers studying human reasoning create semantically clear
questionnaires and surveys, in conformance with conversational maxims.
Even Kahneman and Tversky seem to accept this point. In the postscript of
their canonical book Judgment under Uncertainty: Heuristics and Biases,
Lee / The Gricean Turn in Psychology 201
19. Amos Tversky and Daniel Kahneman, “Extensional versus Intuitive Reasoning: The
Conjunction Fallacy in Probability Judgment,” Psychological Review 90, no. 4 (1983): 312.
20. Conversational implicature may also play a role in subjects’ interpretation of the outcome
“Linda is a bank teller.” Subjects might interpret this outcome, when stated alone, as the
most information the experimenters can assert with confidence according to the maxim of
quantity. However, when presented with the accompanying outcome “Linda is a bank teller
who is active in the feminist movement,” the meaning of “Linda is a bank teller” is unclear.
Subjects may have taken the additional information “who is active in the feminist movement”
as indicating a kind of contrast, where “Linda is a bank teller” is supposed to mean that “Linda
is a bank teller who is not active in the feminist movement.” This interpretation would distinguish
the outcomes “Linda is a bank teller” and “Linda is a bank teller who is active in the
feminist movement” as non-overlapping events. To prevent subjects from adopting this interpretation,
Tversky and Kahneman rephrased “Linda is a bank teller” as “Linda is a bank teller,
whether or not she is active in the feminist movement.” However, Don Dulany and Denis
Hilton found that only 28% of subjects interpreted “whether or not she is active in the feminist
movement” in the way Tversky and Kahneman hoped. See Tversky and Kahneman,
“Judgments of and by Representativeness”; and Don E. Dulany and Denis J. Hilton,
“Conversational Implicature, Conscious Representation, and the Conjunction Fallacy,” Social
Cognition 9, no. 1 (1991): 85-110.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
Kahneman and Tversky admit that apparently irrational responses elicited
from subjects could have arisen from the subjects’ misunderstandings of the
question.21 And, they concede that the “conversational aspect of judgment
studies deserves more careful consideration than it has received in past
research, our own included.”22
Another methodological lesson of the Gricean turn is that experimenters,
in arguing for the validity of their questionnaires and surveys, must collect
data about subjects’ interpretations of the experimental tasks. Kahneman and
Tversky’s research up to that point had been conducted using a questionanswering
paradigm, where “the subject is exposed to information and is
asked to answer questions or to estimate values, orally or in writing.”23
Subjects were provided uniform information from experimenters, and experimenters
assumed subjects’ interpretations would be uniformly identical
with their own. However, because communicating clearly in experimental
conditions is itself a challenge, this is not a safe assumption.24
The Gricean turn also reminds us that effective communication requires
cooperation by both experimenter and subject. Indeed, experimenters shoulder
a greater responsibility for clear communication because of the asymmetrical
nature of experimenter-subject communication. Unlike ordinary
conversation, the course and content of communication in experimental
conditions are predetermined by only one conversant, namely, the experimenter.
25 Experimenters can direct the subjects and ask any number of clarificatory
questions, though the converse is not true. As a result, experimenters
are especially responsible for making “correct assumptions about the codes
and contextual information that the audience will have accessible and be
likely to use in the comprehension process.”26
202 Philosophy of the Social Sciences
21. Daniel Kahneman and Amos Tversky, “On the Study of Statistical Intuitions,” in
Judgment under Uncertainty: Heuristics and Biases, ed. Daniel Kahneman, Paul Slovic, and
Amos Tversky (New York: Cambridge University Press, 1982), 493.
22. Ibid., 504; emphasis added.
23. Ibid., 501.
24. Indeed, subsequent testing has found that subjects usually misinterpret the meaning of
key words and phrases in the Linda problem and other problems from the heuristics and biases
tradition. Hertwig and Gigerenzer used a multiple-choice method in checking subjects’ interpreted
meaning of “probability” by asking them to check which of a list of terms best reflected
their understanding of “probability” in the Linda problem. They found that only 12% of
checked choices were mathematical (e.g., “expectancy,” “frequency,” “percentage,” “logicality,”
or “certainty”). See Hertwig and Gigerenzer, “The ‘Conjunction Fallacy’ Revisited,” 280-82.
25. Clark and Schober, “Asking Questions and Influencing Answers,” 26.
26. Dan Sperber and Deirdre Wilson claim this asymmetry in communication exists in all
communicative contexts. I do not agree with this stronger claim. Dan Sperber and Deirdre
Wilson, Relevance: Communication and Cognition (Oxford: Basil Blackwell, 1986), 43.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
The Gricean turn changes the nature of the investigation—away from
sweeping charges of rationality or irrationality lodged against subjects—
toward exploration of the communicative conditions that either tend to mislead
subjects or tend to facilitate their successful performance.27 Such
investigation, undertaken by psychologists studying conversational pragmatics,
is situational since it focuses on experimental conditions to explain
the apparent rationality or irrationality of subject responses. The situational
explanation here is of a special kind, in which the experimenter and the
experimenter’s relationship with the subject partly constitute the explanatory
situation/context. The situational explanation suggested by Gricean charity is
reflexive, insofar as it considers how the experimenter and the experimenter’s
relationship with the subject play a role in subjects’ observed behavior.
Such situational explanations shift investigation away from attributional
explanations, which focus on characteristics of the subjects themselves—
such as flawed or limited memory, attention, search, or reasoning strategies.
The great irony here is that explanations that unfairly blame subject
responses on internalized judgment heuristics risk committing the fundamental
attribution error.28
D. Naturalized Conversational Norms
The great methodological insight of conversational pragmatics is that,
rather than make unsubstantiated assumptions about subjects’ interpretations
of experimental tasks, researchers ought to gather evidence to identify those
interpretations and identify conditions of successful versus unsuccessful
communication. However, in hypothesizing about subjects’ conversational
inferences, critics might argue that researchers in conversational pragmatics
have adopted a few working assumptions of their own about subjects’ conversational
assumptions and norms. In particular, researchers in conversational
pragmatics have uncritically adopted Grice’s maxims of conversation
as the relevant norms of conversation in subject-experimenter communication.
One possible explanation for this is that researchers have regarded
Grice’s conversational maxims as universal norms of conversation.
Regarding Grice’s maxims as universal norms of conversation would fail to
respect the context and cultural relativity of such norms. Everyday experience
demonstrates that the content of conversational norms varies depending on the
context and goals of communication. For example, we recognize that in contexts
where conversation is made for the sake of mutual entertainment, the
Lee / The Gricean Turn in Psychology 203
27. Thanks to Elizabeth Anderson for this insight.
28. Hilton, “The Social Context of Reasoning,” 249.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
maxims of manner, quantity, and quality do not apply: many of the anecdotes,
jokes, and tangents we trade are valued precisely for the creative ways in
which the descriptions of true or merely hypothetical events are drawn out,
elaborated, and exaggerated. Additionally, the conversational norms may
depend on broader cultural norms or goals: the research of linguistic anthropologists
demonstrates that conversational norms we hold dear are not universal,
but reflect the goals and cultural norms of specific social milieus.29
Regarding Grice’s maxims as universal norms of conversation would
also fail to capture the dynamic relationship between conversational norms
and subjects’ goals, questionnaire content, and communicator identity. To
illustrate, consider the following case of subject-experimenter communication.
Subjects familiar with psychological testing may know that the ostensible
purpose of an experiment is often different from its real purpose.
Subjects who know this may think it likely that experimenters are trying to
deceive them in the sense that the experimenter aims to gather information
about the subject that is not directly asked for. Since this assumption is true
more often than not, especially in questionnaires from the heuristics and
biases tradition, it would be reasonable for subjects not to assume that
experimenters’ conversational contributions conform to the maxims of
quality, quantity, or relevance. Subjects who look skeptically upon the
questions asked of them may adopt the goal of figuring out what the experimenters
are really after, and provide answers that seek to uncover or frustrate
the experimenter’s goals.30 In this scenario, subjects would best be
described as rejecting Grice’s maxims of quality, quantity, and relevance.
Although researchers in conversational pragmatics have not explored
conversational norms that differ in content from Grice’s maxims, it would
be unfair to construe them as adopting Gricean norms as universal norms of
conversation. These researchers have gone to great lengths to cite a broad
range of studies suggesting that subjects respect something like the maxims
of quality, relevance, manner, and quantity.31 They have also sought to
204 Philosophy of the Social Sciences
29. For example, Michelle Rosaldo’s fieldwork suggests that the Ilongots do not share
Grice’s commitment to the maxim of quality. She argues that the primary conversational norm
in Ilongot culture directs conversants to provide conversational contributions that maintain
social roles and relationships, regardless of the truth of those assertions. Michelle Z. Rosaldo,
Knowledge and Passion: Ilongot Notions of Self and Social Life (Cambridge: Cambridge
University Press, 1980). For similar kinds of critiques in the context of Malagasy society, see
Elinor Ochs Keenan, “The Universality of Conversational Postulates,” Language in Society 5
(1976): 67-80.
30. Thanks to Peter Railton for raising this possibility.
31. For extensive reviews, see Hilton, “The Social Context of Reasoning.” See also Schwarz,
Cognition and Communication.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
empirically support their account of subjects’ conversational assumptions.
For example, they cite experiments studying how subjects’ judgments of
trustworthiness (on the part of the experimenter) vary with changes in the
content of communication and the identity of the communicator.32 The
implication of citing this research is supposed to be that subjects’ judgments
of trustworthiness may have important effects on what conversational
norms govern subject-experimenter communication.
The best interpretation of researchers’ commitment to Grice’s maxims is to
interpret them as adopting Grice’s maxims as empirically compelling formulations
of the conversational norms that seem to govern subject-experimenter
communication in certain types of experimental contexts. Grice’s maxims
are not presented as uncontestable social facts, but as working hypotheses.
However, in embracing a naturalized account of conversational norms,
conversational pragmatics would be well served by checking its assumptions
about subjects’ conversational assumptions and norms with respect
to particular cases of subject-experimenter communication. Such research
would build on the lessons of an older psychological literature on source
effects that concerned itself with how subjects’ goals influence research
results.33
E. Gricean Charity and Naturalized Interpretation
Like conversational pragmatics, Gricean charity adopts conversational
norms as norms of rationality, and uses these norms to guide the interpretation
Lee / The Gricean Turn in Psychology 205
32. See, for example, Zvi Ginossar and Yaacov Trope, “Problem Solving in Judgment
under Uncertainty,” Journal of Personality and Social Psychology 52 (1987): experiment 5.
Also, Eleanor Singer, Hans-Jürgen Hippler, and Norbert Schwarz discovered that increasingly
emphatic confidentiality assurances decrease the rate at which subjects are willing to respond
to survey questionnaires. Their explanation is that such confidentiality assurances—when construed
as relevant to the ensuing survey questions—suggest that the survey will ask questions
that are personal, embarrassing, and/or incriminating. Subjects’ decreased willingness to participate,
and decreased willingness to provide identifying information in order to participate in
future surveys, seems to indicate a decreased level of trust in the surveyors’ confidentiality
assurances. Eleanor Singer, Hans-Jürgen Hippler, and Norbert Schwarz, “Confidentiality
Assurances in Surveys: Reassurance or Threat?” International Journal of Public Opinion
Research 4, no. 3 (1992): 256-68.
33. This literature attributes to subjects a broader range of possible goals, including the
attainment of private rewards, the discovery of the experiment’s true rationale, the presentation
of the self in the best light, and the desire to contribute to the experiment’s success by
assisting the experimenter in proving her point. For a review, see Arie W. Kruglanski, “The
Human Subject in the Psychology Experiment: Fact and Artifact,” Advances in Experimental
Social Psychology 8 (1975): 101-47.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
of subjects’ beliefs. Gricean charity invokes these norms of conversation to
justify alternate interpretations of the stated question or task—an interpretation
for which subjects’ expressed beliefs can sometimes be construed as
rational. Implicit in this approach to intentional explanation is the empirical
assumption that subjects are very unlikely to violate naturalized conversational
norms. Gricean charity’s difference with conversational pragmatics is
merely one of emphasis: Gricean charity explicitly embraces naturalized
norms of conversation, and recommends the continued collection of evidence
about the conversational assumptions and norms guiding subjectexperimenter
communication.
This naturalized approach to interpretation builds on important themes
in naturalized accounts of interpretation by contemporary philosophers
such as David Henderson and Mark Risjord. Like these accounts, Gricean
charity is naturalized in the sense that it allows and encourages the use of
empirical knowledge taken from the human sciences to guide interpretive
theory. Henderson argues that empirical knowledge, especially psychological
theory, should be the primary guide in interpretation. Risjord’s interest
in interpreting group-level events in terms of cultural norms expands the list
of human sciences relevant to interpretation to include anthropological,
sociological, and historical theories. Gricean charity’s interest in naturalized
conversational norms connects psychological research with findings
and theories in linguistic anthropology and sociolinguistics.
Henderson’s and Risjord’s naturalized accounts of interpretation also
embrace the Davidsonian lesson that any psychological theory on human
judgment “must include a theory of interpretation” about subjects’beliefs about
the experimental task.34 Henderson paints a picture of “interpretation-cumexplanations,”
where we “construct interpretive schemes so as to be yoked
with our psychological and sociological theories to the end of modeling and
accounting for the behavior and behavioral dispositions of our subjects.”35
Under Risjord’s account, appeals to meanings (common to a group of
speakers) are crucial to the interpretation of group-level phenomena.36
Gricean charity’s contribution to this common ground is in providing a positive
account of the kind of “integral role” that interpretation should play in
the co-development of interpretive and psychological theory. In particular,
Gricean charity draws important methodological lessons from current research
206 Philosophy of the Social Sciences
34. Davidson, “Belief and the Basis of Meaning (1974),” 147; emphasis added.
35. David K. Henderson, Interpretation and Explanation in the Human Sciences (Albany:
State University of New York Press, 1993), 73-74.
36. Mark Risjord, Woodcutters and Witchcraft (Albany: State University of New York
Press, 2000), 137-38.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
on conversational pragmatics and recommends a naturalized account of
conversational norms that respects the cultural relativity of conversational
norms, and their dynamic relationship with questionnaire content and communicator
identity.
2. Objections
A. Universal Rationality
An important test of any account of charitable interpretation is whether
it can allow for and even prefer interpretations that describe others as being
systematically irrational. Some might object that Gricean charity fails this
test, since it can be enlisted to rationalize just about any case of apparent
irrationality. This possibility stems from a particular step in the inductive
argument for any conversational implicature. In drawing an implicature, a
hearer must arrive at the belief that, in order for some speaker S to say that
p, and still be observing the maxims or the cooperative principle, S must
think that q. This premise about q is itself arrived at by means of an inductive
argument about what claim q is implicated, given the literal meaning of
p, the conversational context, and beliefs mutually held by speaker and
addressee.37 Without any principled way of deciding what gets to count as
a reasonable or best candidate implicature q, acceptable conversational
implicatures are restricted only by the limitations of the human imagination.
Since we can always find some reinterpretation for which subject
responses can be said to be rational, Gricean charity always provides a way
to rationalize subject responses.
This worry overlooks an important lesson of the Gricean turn, namely,
the role of evidence. Gricean charity puts the onus on the researcher to identify
subjects’ interpretations of experimental tasks and to identify conditions of
successful communication. This evidence can sometimes undermine experimenters’
claims that subjects do in fact draw particular conversational inferences.
Additionally, such evidence can speak in favor of describing subjects
as engaging in irrational lines of reasoning.
For example, open-ended protocols ask subjects to describe their interpretations
of the experimental task and/or explain their answers, in their
own words. This method is helpful because it allows experimenters to capture
the cognitive processes associated with specific semantic inferences,
Lee / The Gricean Turn in Psychology 207
37. Levinson, “Conversational Implicature,” 113-14.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
problem reasoning, and judgment.38 Using this method, Don Dulany and
Denis Hilton asked subjects, “What did ‘Linda is a bank teller’ mean or
imply to you? Be as clear as you can.” They found that many subjects indicated
that Linda’s being a bank teller was irrelevant to answering the question.
Their reasoning seemed to go as follows: because Linda’s being a
bank teller was a property shared between the two outcomes, it was “a constant
that cancelled,” and was not supposed to affect the outcomes, thus
reducing the problem to judging whether Linda was a feminist or not.39 This
line of reasoning is very problematic. Events common to different outcomes
do not cancel out in the way subjects imagine. This additional self-reported
information about subjects’ beliefs and inferences suggests describing their
judgments as profoundly mistaken.40
Before my response to this objection, it seemed that Gricean charity’s
modus operandi was to rationalize what seemed to be irrational responses
in famous studies on human judgment. However, some studies do seem to
demonstrate some kind of systematic irrationality in human reasoning. This
is a great strength of Gricean charity—that it recommends the collection of
evidence that may speak in favor of the rationality or irrationality of human
reasoning under various conditions. It is Gricean charity’s deference to evidence
about communication in experimental contexts that prevents the
hyperrationalization of subject responses.
The symmetrical treatment of rational and irrational beliefs can also be
found in Henderson’s and Risjord’s accounts. Under Henderson’s account,
both rational and irrational belief are held to the same standard of explicability,
which seeks to explain beliefs and actions in terms of the subject’s
causally relevant intentional states and psychological dispositions. In cases
where we happen to adopt a “rationalizing explanation,” the explanatory
force of this explanation derives not from the rationality of what the subject
208 Philosophy of the Social Sciences
38. How experimenters ought to use information from open-ended protocols is not clearcut,
since subjects do not always have direct access or insight into their conversational inferences.
Psychologists are generally skeptical about whether subjects have insight about their
inferences or about the factors that do and do not influence their judgments/choices. For the
classic paper on this, see Richard E. Nisbett and Timothy DeCamp Wilson, “Telling More
Than We Can Know: Verbal Reports on Mental Models,” Psychological Review 84, no. 3
(1977): 231-59.
39. Dulany and Hilton, “Conversational Implicature, Conscious Representation, and the
Conjunction Fallacy,” 102.
40. This self-reported information also suggests that subject responses cannot be rationalized,
even if they interpret “probable” to mean “plausible”: even if subjects are concerned with
the relative plausibility of outcomes, the outcomes still do not cancel in the way they believe
they do. Thanks to James Joyce for this point.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
believes, but from citing causal antecedents and dispositions in terms that
feature in psychological generalizations. Gricean charity provides an example
of what Henderson would identify as a kind of “modest” rationalizing explanation:
Gricean charity hypothesizes that subjects are likely to arrive at a
particular interpretation of an experimental task, in virtue of their disposition
to conform to conversational principles (where claims about their dispositions
and about the conversational principles are empirical claims).41
The force of Gricean charity’s explanations draws strength from empirical
evidence about subjects’ interpretations and their conversational assumptions,
goals, and norms.
For Risjord, theories that attribute rational rather than irrational beliefs are
held to the same standard of explanatory coherence—a standard that is not
committed to interpreting all beliefs and actions as rational.42 In contrast to
Henderson, however, Risjord recognizes that norms of rationality have a legitimate
place in intentional explanation. Interpreters bring with them “interests
constitutive to the interpretive enterprise” such as interests in the agents’ point
of view and in the structure of the society in which they live.43 Risjord
observes that our interest in these perspectives requires explanations invoking
norms.44 This is because intentional explanations and group-level explanations
invoke reasons, where reasons may count as reasons only insofar as they conform
to norms recognized by the agent or the society. It is in this indirect way
that norms figure in the content of explanations.45 Gricean charity draws
strength from Risjord’s analysis. As an account of interpretation, Gricean charity
has an interest in understanding communication and psychological experimentation
from the subject’s point of view. As such, Gricean charity is
interested in the reasons subjects have for their interpretations of the task,
where these reasons count as reasons insofar as they relate to conversational
norms governing subject-experimenter communication. These conversational
norms figure in the content of Gricean charity’s intentional explanations.
B. Biased Applications of Gricean Charity
Keith Stanovich and Richard West observe that some charitable strategies
seeking to rationalize subject responses function by reacting to findings of the
heuristics and biases research approach: they aim to restore the rationality of
Lee / The Gricean Turn in Psychology 209
41. Henderson, Interpretation and Explanation in the Human Sciences, 135-36.
42. Risjord, Woodcutters and Witchcraft, 182.
43. Ibid., 177.
44. Ibid., 187-88.
45. Ibid., 155.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
subject responses in the face of research purporting to demonstrate otherwise.
What is suspicious about this pattern of theorizing is that it criticizes studies
purporting to demonstrate irrationality, but rarely—if ever—critiques those
where modal subject response coincides with the normative response.46 The
worry is that charitable researchers in psychology are biased insofar as they
hold higher standards of experimental design and evidence for psychological
theories claiming to demonstrate human irrationality.
However, Gricean charity does not require a higher standard of evidence
for theories claiming that human reasoning is irrational in some way. Rather,
what speaks for the strength of Gricean charity is that it recommends an evidential
standard that applies generally to studies on human judgment, irrespective
of their conclusions about the rationality or irrationality of subject
responses. The interviewing methods used for attaining key interpretive evidence
are also shared across the rationality divide. I will discuss such interviewing
techniques in greater detail in subsequent sections of this paper.
C. Testing for the Effects of Irrelevant Information
Some object that conversational norms overly constrain what it is that
cognitive psychologists can test since these norms may circumvent asking
questions in ways that prove psychologically interesting. In particular,
Kahneman and Tversky expressed the concern that Grice’s maxim of relevance
poses an “exceptionally difficult” problem for experimenters interested
in studying the effects of irrelevant information on cognition.47 Citing
Richard Nisbett et al.’s work, they observe that subjects can mistakenly
construe nearly any piece of information as relevant.48 From a methodological
point of view, the worry is that the maxim of relevance’s constraint on
questionnaire and survey design precludes the very possibility of testing the
influence of irrelevant information on human cognition.
However, there are ways in which to test the effect of irrelevant information
on cognition without having to violate conversational norms. Psychologists
210 Philosophy of the Social Sciences
46. Stanovich and West make this observation in the context of the “reject-the-norm” strategy,
which rejects the experimenter’s normative theory for a different one to which modal
subject response conforms. Keith E. Stanovich and Richard F. West, “Individual Differences
in Reasoning: Implications for the Rationality Debate?” Behavioral and Brain Sciences 23
(2000): 650.
47. Kahneman and Tversky, “On the Study of Statistical Intuitions,” 501-2.
48. Richard E. Nisbett, Henry Zukier, and Ronald E. Lemley, “The Dilution Effect:
Nondiagnostic Information Weakens the Implications of Diagnostic Information,” Cognitive
Psychology 13 (1981): 248-77.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
working on conversational pragmatics have forged ingenious experimental
methods to do this. One way is to undermine subjects’ assumption that the
source of information in the experimental context is intentional and cooperative.
Recall that the conversational maxims apply in the special case where
information exchange occurs between intentional, cooperative communicators
in ordinary language. If experimenters can undermine this key assumption,
then all conversational implicatures should be cancelled, which would
allow experimenters to study the effects of irrelevant information on human
cognition.
Norbert Schwarz and his colleagues discovered an ingenious way of
using a computer interface to do just this. They focused on Kahneman and
Tversky’s famous lawyer-engineer question:
A panel of psychologists have interviewed and administered personality tests
to 30 engineers and 70 lawyers, all successful in their respective fields. On
the basis of this information, thumbnail descriptions of the 30 engineers and
70 lawyers have been written. You will find on your forms five descriptions,
chosen at random from the 100 available descriptions. For each description,
please indicate your probability that the person described is an engineer, on
a scale from 0 to 100.
The same task has been performed by a panel of experts, who were highly
accurate in assigning probabilities to the various descriptions. You will be
paid a bonus to the extent that your estimates come close to those of the
expert panel.49
In Kahneman and Tversky’s original study, the low-engineer group was told
that there were 30 engineers and 70 lawyers. The high-engineer group was
told that there were 70 engineers and 30 lawyers. Both groups were provided
the same five personality descriptions, most of which were stereotypical of an
engineer or lawyer.50 Kahneman and Tversky found that subjects’ predictions
about how probable it was that a given person was an engineer or lawyer were
Lee / The Gricean Turn in Psychology 211
49. Daniel Kahneman and Amos Tversky, “On the Psychology of Prediction (1973),” in
Judgment under Uncertainty: Heuristics and Biases, ed. Daniel Kahneman, Paul Slovic, and
Amos Tversky (New York: Cambridge University Press, 1982), 53.
50. Kahneman and Tversky (ibid.) offered the following as an example of one of the personality
descriptions:
Jack is a 45-year-old man. He is married and has four children. He is generally conservative,
careful, and ambitious. He shows no interest in political and social issues and
spends most of his free time on his many hobbies which include home carpentry, sailing,
and mathematical puzzles.
The probability that Jack is one of the 30 engineers in the sample of 100 is ———%.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
independent of the base rates of engineers/lawyers described in the questionnaire:
subjects ignored the base rate information in violation of Bayes’s rule.
Norbert Schwarz et al. predicted that subjects’ violation of Bayes’ rule
resulted from conversational implicatures reasonably inferred from the
original wording of the questionnaire. They point out that subjects who
identify the experimenter as cooperative are in the position of trying to render
the communicated information about Jack’s personality relevant to their
interpretation of the experimental task; and, by the maxim of quantity,
subjects are left to infer that all the detailed information provided about
Jack’s personality is meant to play into the proper solution of the task.51
To undermine the assumption that the individuating information was
relevant and informative, Schwarz et al. ingeniously created a computer communicator
condition, where subjects were told that a computer—an uncooperative
and unintentional communicator—had created the personality description
by randomly drawing sentences from psychologists’ or researchers’ files pertaining
to the target person.52 By undermining the assumption of cooperative
212 Philosophy of the Social Sciences
51. Special clues in the question underscore the relevance and importance of the personality
description in solving the task. The first paragraph of the instructions “informs subjects that
the individuating information was compiled by psychologists on the basis of respected procedures
of their profession, namely interviews and tests.” Schwarz et al. observe that, since psychologists
are stereotypically perceived as experts on issues of personality rather than
probability and base rates, identifying the authors of the personality descriptions as psychologists
emphasizes the relevance and informativeness of the individuating information rather
than the base rate information in solving the experimental task. Kahneman and Tversky reinforce
the importance and relevance of the personality descriptions by going on to state that
“[t]he same task has been performed by a panel of experts, who were highly accurate in
assigning probabilities to the various descriptions.” This sentence underscores the relevance of
the individuating information by pointing out that the stereotypical descriptions are sufficiently
diagnostic for experts to succeed in solving the experimental task. Although the professional
identity of the experts is left unspecified, subjects might reasonably infer the experts
are psychologists, based on the following facts: the experts are highly accurate in personalitybased
predictions, and the experts are so-called by experimental psychologists in the context
where they seem to use the personality tests to predict outcomes. The further claim that “[y]ou
will be paid a bonus to the extent that your estimates come close to those of the expert panel”
suggests that the subjects are encouraged to make judgments in a similar manner as the expert
panel. If subjects have already identified the experts as psychologists, this statement would
encourage subjects to study the personality traits described, and use them to diagnose professional
identity. Norbert Schwarz, Fritz Strack, Denis Hilton, and Gabi Naderer, “Base Rates,
Representativeness, and the Logic of Conversation: The Contextual Relevance of ‘Irrelevant’
Information,” Social Cognition 9, no. 1 (1991): 67-84.
52. To undermine the assumption that the individuating information was relevant and
informative, Schwarz and his colleagues told subjects that a computer had created the provided
personality description by randomly drawing sentences from psychologists’ or researchers’
files pertaining to the target person. Ibid., 74.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
communication, subjects were freed from having to construe the personality
information as relevant to the experimental task. They found that in the
computer communicator condition, subjects weighed the individuating information
less, and considered the base rate more: the mean probability estimate
for subjects judging the probability that the target was an engineer was only
40%, compared to the control group’s mean probability estimate of 76%.
D. Conversational Clarity and Conceptually
Difficult Tasks
Kahneman and Tversky have objected that efforts to make the experimental
task as semantically unambiguous as possible reveal key clues about
solving the task. Such clues, the objectors maintain, compromise researchers’
abilities to test whether subjects can solve the task without undue help. The
best way to understand this objection perhaps is by way of example. In the
Linda case, one way to clear up the ambiguity about what probable means
is to paraphrase with the more precise mathematical word frequency in the
following way:53
In an opinion poll, the 200 women selected to participate have the following
features in common: They are, on average, 30 years old, single, and very
bright. They majored in philosophy. As students, they were deeply concerned
with issues of discrimination and social justice and also participated in antinuclear
demonstrations.
Please estimate the frequency of the following events.
How many of the 200 women are bank tellers? ___ of 200
How many of the 200 women are active feminists? ___ of 200
How many of the 200 women are bank tellers and active feminists? ___
of 20054
As the researchers Ralph Hertwig and Gerd Gigerenzer expected, none of
the subjects provided answers in violation of the conjunction rule under this
formulation of the Linda problem.
However, Kahneman and Tversky suggest this formulation of the question
makes judgments of probability a piece of cake. They ask subjects for
a numerical estimate of the ratio of women who are bank tellers and/or
active feminists in the total population of Linda-like people. That is, they
Lee / The Gricean Turn in Psychology 213
53. This idea is due to Hertwig and Gigerenzer, “The ‘Conjunction Fallacy’ Revisited.”
54. Ibid., 291.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
ask subjects to estimate the number of people belonging to the set of bank
tellers, the set of active feminists, and the intersection of both these sets,
within a fixed population. Being able to judge the probability of a conjunction
of events requires conceptualizing how classes of events are related in
this set theoretical way. At the very least, understanding the conjunction
rule requires seeing that the number of members in the intersection of two
sets must be less than or equal to the number of members in each of those
individual sets. Hertwig and Gigerenzer’s way of posing the question not
only clarifies the meaning of the question, it also takes subjects through
the hardest step of understanding how to think about and solve the task.55
So the question becomes, are there ways of clarifying the meaning of the
Linda question without giving away what Kahneman and Tversky take to
be key clues?56
Hertwig and Gigerenzer discovered a way of clarifying the intended
meaning of the probability question without giving away key clues to solving
the problem. They observed that, in the original question, the meaning
of probability was obscured by the maxim of relevance: in particular, the
maxim directs subjects to make irrelevant personality information pertinent
to the experimental task. They hypothesized that one way to indicate that
the personality information was not necessarily relevant to the probability
task is to ask a different question for which Linda’s personality description
is relevant before asking the probability question. This leaves subjects free
to interpret the meaning of the subsequent “probability” question without
reference to the personality description. To test this, Hertwig and Gigerenzer
214 Philosophy of the Social Sciences
55. This finding replicates Kahneman and Tversky’s discovery that subjects are much less
likely to violate the conjunction rule when the conjunctions were represented by the intersection
of concrete, finite classes rather than by an abstract combination of properties. They found
that only 25% of subjects violated the conjunction fallacy when they were asked to estimate
the frequencies rather than single-event probabilities. But, by representing the question in this
frequency format, Tversky and Kahneman took themselves as encouraging “subjects to set up
a representation of the problems in which class inclusion is readily perceived and appreciated.”
Tversky and Kahneman take their findings to demonstrate that “[t]he formal equivalence of
properties to classes is apparently not programmed into the lay mind”—that subjects do not
“evaluate compound probabilities by aggregating elementary ones” unless provided a representation
“in which different relations and rules are transparent.” Tversky and Kahneman,
“Extensional versus Intuitive Reasoning,” 294-309.
56. Notice that Kahneman and Tversky’s question and objection presuppose no obligation
to satisfy Gricean norms. From a Gricean perspective, the lesson to be drawn is not that
“people are irrational,” but that “if you do not want to mislead people, then speak more
clearly.” An experiment without clearly specified questions does not demonstrate that subjects
are innately irrational, but that certain representations are less effective at communicating a
task than others.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
provided subjects with Linda’s personality description, and asked subjects
to judge how good an example Linda’s personality was of “an active feminist,”
“a bank teller,” and “a bank teller and an active feminist.” Subjects
were then asked the probability question, “Which of the following statements
is most probable?” Possible answers were “Linda is an active feminist,”
“Linda is a bank teller,” and “Linda is an active feminist and bank
teller.” Under this version of the Linda problem, they found that subjects
were less likely to violate the conjunction rule: only 59% of subjects violated
the conjunction rule, compared to 88%. Notice that the effect of successful
communication here is not to manipulate subjects into providing the
normatively correct answer, since the majority of subjects continue to violate
the conjunction rule. Rather, the effect is to get a more accurate picture
of subjects’ probability judgments.
3. Conclusions
Gricean charity builds on recent accounts of naturalized interpretation
by issuing positive methodological recommendations for proper interpretation
and explanation in psychological experimentation. In particular, the
Gricean turn reminds us that researchers are responsible for successful
communication in the experimental context. In order to promote successful
communication, Gricean charity recommends creating questionnaires in
conformance to naturalized conversational norms, and recommends seeking
evidence about (1) the conditions of successful communication and
(2) the conversational norms governing subject-experimenter communication.
The Gricean turn also shifts psychological explanations toward situational
explanations that focus on the experimental conditions responsible for subject
responses. And, this situational perspective engages in reflexive analysis
about the influence of the experimenter (and experimenter’s relationship
with the subject) on observed behavior.
Gricean charity’s evidential recommendations are unbiased in the sense
that they apply symmetrically across the rationality divide: they apply to
studies on human judgment, irrespective of the studies’ conclusions about
the rationality or irrationality of subjects’ responses. Evidence gained by these
methods does not necessarily speak in favor of the rationality or irrationality
of subject responses. Thus, Gricean charity passes an important test for
any account of charitable interpretation: namely, it allows for the possibility
of and even prefers interpretations that construe others as being systematically
irrational. As a result, Gricean charity does not illegitimately favor
Lee / The Gricean Turn in Psychology 215
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
psychological theories that construe human reasoning as being generally
rational.
Gricean charity is biased in its special focus on conversational pragmatics.
However, this is not a problematic bias, but a focus of research. Such a
focus is found in normal scientific research programs that use background
theory and hard-core assumptions as tools to generate new questions, predictions,
evidence, and hypotheses.57 Gricean charity suggests new questions
about the nature of conversational inference: under what conditions
can we undermine subjects’ assumptions of the experimenter’s cooperativeness?
How can we undermine the assumption of intentionality in questionnaires,
surveys, or experimental communications in general? How can
we emphasize the right information so as to elicit valid ways of reasoning
and correct judgment? How is task interpretation tied into reasoning about
how to solve the task? And how do conversational norms vary under different
experimental conditions?
These questions motivate the creation of new methods for collecting
evidence about subjects’ conversational inferences. For example, Schwarz
et al.’s computer communication condition allowed researchers to undermine
subjects’ assumption of cooperativeness and intentionality on the part
of experimenters. Hertwig and Gigerenzer’s subtle double-question technique
provided a way of manipulating the assumption of relevance.58
Finally, the questions, hypotheses, and interests motivated by Gricean
charity have led to the discovery of new kinds of psychologically interesting
phenomena. For example, Schwarz et al.’s computer communicator
condition revealed new conditions that sensitize subject probability judgments
to base rate information. His computer communicator condition also
revealed other fascinating evidence. When the wording of the questionnaire
216 Philosophy of the Social Sciences
57. Philip E. Tetlock, “The Impact of Accountability on Judgment and Choice: Toward a
Social Contingency Model,” Advances in Experimental Social Psychology 25 (1992): 331-76;
and Gerd Gigerenzer, “From Tools to Theories:A Heuristic Discovery in Cognitive Psychology,”
Psychological Review 98, no. 2 (1991): 254-67.
58. In addition to these methods of checking subject task construal, experimenters may
employ a multiple-choice check method, which asks subjects to check which of a provided set
of interpretations best matches their interpretation of the task. One weakness of this approach
is that subjects may draw different implicatures from the more precisely stated multiple-choice
options. For examples relevant to the conjunction fallacy, see Hilton, “The Social Context of
Reasoning,” 260. Experimenters may also check subject interpretations by asking subjects to
paraphrase the experimenter’s original question. One problem with this kind of method of
checking subject interpretations is that the paraphrases provided may themselves be vague,
polysemous, and/or indeterminate. For examples relevant to the conjunction fallacy, see
Hertwig and Gigerenzer, “The ‘Conjunction Fallacy’ Revisited.”
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
was framed as a statistics problem, the computer communicator condition
impaired subjects’ probability judgments: subjects again ignored or underweighted
the base rate information.59 Schwarz et al. suggest that subjects
relied more on the individuating information “presumably because in a
statistical framework random sampling suggests that the resulting selection
is representative of the population of descriptive information from which it
is drawn.”60 By communicating more successfully with subjects, researchers
in conversational pragmatics have opened the door to getting a clearer
picture of the surprising and muddled ways in which we weigh evidence,
draw inferences, and make choices.
References
Adler, Jonathan E. 1984. Abstraction is uncooperative. Journal for the Theory of Social
Behavior 14 (2): 165-81.
Clark, Herbert H., and Michael F. Schober. 1992. Asking questions and influencing answers. In
Questions about questions: Inquiries into the cognitive bases of surveys, edited by Judith M.
Tanur, 15-48. New York: Russell Sage Foundation.
Davidson, Donald. 2001. Belief and the basis of meaning (1974). In Inquiries into truth and
interpretation, 141-54. New York: Oxford University Press.
Dulany, Don E., and Denis J. Hilton. 1991. Conversational implicature, conscious representation,
and the conjunction fallacy. Social Cognition 9 (1): 85-110.
Edwards,Ward, Harold Lindman, and Lawrence D. Phillips. 1965. Emerging technologies for
making decisions. In New directions in psychology 2, 259-326. New York: Holt, Rinehart
& Winston.
Gigerenzer, Gerd. 1991. From tools to theories: A heuristic discovery in cognitive psychology.
Psychological Review 98 (2): 254-67.
———. 1996. On narrow norms and vague heuristics: A Reply to Kahneman and Tversky
(1996). Psychological Review 103 (3): 592-96.
Ginossar, Zvi, and Yaacov Trope. 1987. Problem solving in judgment under uncertainty.
Journal of Personality and Social Psychology 52:464-74.
Grice, Paul. 1989. Logic and conversation. In Studies in the way of words. Cambridge, MA:
Harvard University Press.
Henderson, David K. 1993. Interpretation and explanation in the human sciences. Albany:
State University of New York Press.
Hertwig, Ralph, and Gerd Gigerenzer. 1999. The “conjunction fallacy” revisited: How intelligent
inferences look like reasoning errors. Journal of Behavioral Decision Making 12:275-305.
Hilton, Denis J. 1995. The social context of reasoning: Conversational inference and rational
judgment. Psychological Bulletin 118 (2): 248-71.
Kahneman, Daniel, and Amos Tversky. 1982. On the psychology of prediction (1973). In
Judgment under uncertainty: Heuristics and biases, edited by Daniel Kahneman, Paul
Slovic, and Amos Tversky, 48-68. New York: Cambridge University Press.
Lee / The Gricean Turn in Psychology 217
59. Schwarz et al., “Base Rates, Representativeness, and the Logic of Conversation,” 72-73.
60. Ibid., 76.
distribution.
© 2006 SAGE Publications. All rights reserved. Not for commercial use or unauthorized
Downloaded from http://pos.sagepub.com at UNIV WASHINGTON LIBRARIES on July 23, 2008
———. 1982. On the study of statistical intuitions. In Judgment under uncertainty: Heuristics
and biases, edited by Daniel Kahneman, Paul Slovic, and Amos Tversky, 493-508. New York:
Cambridge University Press.
Keenan, Elinor Ochs. 1976. The universality of conversational postulates. Language in Society
5:67-80.
Kruglanski, Arie W. 1975. The human subject in the psychology experiment: Fact and artifact.
Advances in Experimental Social Psychology 8:101-47.
Kunda, Ziva, and Paul Thagard. 1996. Forming impressions from stereotypes, traits, and behaviors:
A parallel-constraint-satisfaction theory. Psychological Review 103 (2): 284-308.
Levinson, Stephen. 1983. Conversational implicature. In Pragmatics, 97-166. Cambridge:
Cambridge University Press.
Nisbett, Richard E., and Timothy DeCamp Wilson. 1977. Telling more than we can know:
Verbal reports on mental models. Psychological Review 84 (3): 231-59.
Nisbett, Richard E., Henry Zukier, and Ronald E. Lemley. 1981. The dilution effect:
Nondiagnostic information weakens the implications of diagnostic information. Cognitive
Psychology 13:248-77.
Risjord, Mark. 2000. Woodcutters and witchcraft. Albany: State University of New York Press.
Rosaldo, Michelle Z. 1980. Knowledge and passion: Ilongot notions of self and social life.
Cambridge: Cambridge University Press.
Schwarz, Norbert. 1994. Judgment in a social context: Biases, shortcomings, and the logic of
conversation. Advances in Experimental Social Psychology 26:123-62.
———. 1996. Cognition and communication: Judgmental biases, research methods, and the
logic of conversation. Mahwah, NJ: Lawrence Erlbaum.
Schwarz, Norbert, Fritz Strack, Denis Hilton, and Gabi Naderer. 1991. Base rates, representativeness,
and the logic of conversation: The contextual relevance of “irrelevant” information.
Social Cognition 9 (1): 67-84.
Singer, Eleanor, Hans-Jürgen Hippler, and Norbert Schwarz. 1992. Confidentiality assurances
in surveys: Reassurance or threat? International Journal of Public Opinion Research 4 (3):
256-68.
Sperber, Dan, and Deirdre Wilson. 1986. Relevance: Communication and cognition. Oxford:
Basil Blackwell.
Stanovich, Keith E., and Richard F. West. 2000. Individual differences in reasoning:
Implications for the rationality debate? Behavioral and Brain Sciences 23:645-726.
Tetlock, Philip E. 1992. The impact of accountability on judgment and choice: Toward a social
contingency model. Advances in Experimental Social Psychology 25:331-76.
Tversky, Amos, and Daniel Kahneman. 1982. Judgments of and by representativeness. In
Judgment under uncertainty: Heuristics and biases, edited by Daniel Kahneman, Paul
Slovic, and Amos Tversky, 84-98. New York: Cambridge University Press.
———. 1983. Extensional versus intuitive reasoning: The conjunction fallacy in probability
judgment. Psychological Review 90 (4): 293-315.


Carole J. Lee is currently a philosophy PhD candidate at the University of Michigan, Ann
Arbor. She will be joining the faculty at Mount Holyoke College in the fall of 2006. Her current
research provides an account of methodological rationalism for interpretation in psychological
research