This excerpt from

Mind Readings.
Paul Thagard, editor.
1998 The MIT Press.
is provided in screen-viewable form for personal use only by members
of MIT CogNet.
Unauthorized use or dissemination of this information is expressly
forbidden.
If you have any questions about this material, please contact
cognetadmin@cognet.mit.edu.

2
The Cognitive

Science of Deduction

Philip N . Johnson-Laird and Ruth M . J. Byrne

The late Lord Adrian, the distinguishedphysiologist, once remarked

that if you want to understandhow the mind works then you had
better first askwhatit is doing. This distinction hasbecomefamiliar in
cognitive scienceasone that Marr (1982) drew between a theory at
the "computationallevel" and a theory at the "algorithmic level." A
theory at the computationallevel characterizeswhat is being computed, why it is being computed, and what constraintsmay assistthe
process
. Sucha theory, to borrow from Chomsky(1965), is an account
of human competence. And, ashe emphasizes
, it should alsoexplain
how that competenceis acquired. A theory at the algorithmic level
specifieshow the computation is carriedout, and ideally it should be
preciseenoughfor a computer program to simulatethe process
. The
algorithmic theory, to borrow againfrom Chomsky, should explain
the characteristics
of humanperformance- where it breaksdown and
leadsto error, where it runs smoothly, and how it is integratedwith
other mental abilities.
We have two goalsin this chapter. Our first goal is to characterize deductionat the computationallevel. Marr criticizedresearchers
for trying to erect theories about mental processeswithout having
stopped to think about what the processeswere supposedto compute. The samecriticism can be levelled againstmany accountsof
deduction, and so we shall take pains to think about its function:
what the mind computes, what purpose is served, and what constraintsthereareon the process
. Our secondgoalis to examineexisting
algorithmic theories. Here, experts in severaldomains of enquiry
have something to say. Linguists have consideredthe logical form
of sentencesin natural language. Computer scientistshave devised

30

Chapter2

programs

that

confronted
logic

make

discrepancies

. Psychologists

experimental
in

order

what
the

it

mind

: A

What

happens

that

they

duce

a novel

from

some

times

how

of

it

have

and

from

account
of

they

formal

based

work

Computational

make

on

their

these

dis -

deduction

might

be

carried

to
out

one

is

of

states

from

to

them

a single
to

action

. Their

to

true

be

of

affairs

is to

given

In

many

of

affairs

,
pro

, they

arrive

that

is

and

some

practical
and

their

at a valid

their

argue

, though

state

aim

. Typically

another

is a perceived

answer

observations

conclusion

point

bound

short

perceptual

imagined

follows

? The

proposition

starting

, which

a deduction

, or

that

from

Level

information

, beliefs

is a course

starting

con

point

is

.
long

are

depends

trary

of

( see

Spinoza

defended

because

extraneous

people

all

, there

anism

. The

leads

naturally

logic

( Leibniz

the

is some

the

, 1765

view

).

1800

attributed

that

the

In
in

an

: deduction

to

recent

years

mind

is furnished
authors

an

other
,

Henle

, or

import

asserts

( 1981

"

with

( Henle

) has

con

mech

. This

, impressed

, " which

- processing

) . These

claims

them

fault

, she

reasoning

at

con
advo
be

from

reasoning

information
be

so

been

seems

faulty

, and

seems
has

, " she

which

content

inference

L . J . Cohen

cannot

, 1854

What

errors
to

of

).

idea

, re - interpret

philosopher

to

any

, it

a valid

found

competence

; Boole

. This
suspect

premises

malfunction

to

might

. Mistakes

never

cases , the

underlying

Kant

extent

is impossible

infallibly

than

view

forget

be

such

to

1677

. " I have

unambiguously
) . In

more

a similar

material

error

psychologists

nothing

the

applying

, as you

( and

is

logical

principles

that

philosophers

) has

that

concerns

principles

sense

inference

premises

say

universal
these

common

by

invalid

controversy

. Some

a set

exercises

to

cated

- standing

logical

on

everyone

curred

the

propositions

, their

people

1978

inference

review

theories

some

conclusion

One

could

people

with

merely

clusion

occur

when

initial

conclusion

( 1962

at

, statements

inferences

will

theories

a preliminary

outline

Theory

start

memories

philosophers

algorithmic

. We

to

, like
everyday

proposed

establish

is , and

and

Deduction

true

to

between

have

investigations

ciplines
show
by

deductions

doctrine
an

inborn
by

the

31

The CognitiveScience
of Deduction

human

invention

think

of

rationally

logic

The

Psychologism

Mill

that

is

1843

judge

to

is

not

Freud

mind

seems

Yet

view

is

its

idea

discovery

relations

properly

no

so

logic

but

it

between

principle

but

and

stances

possible

mistakes

. g

they

are

intuitions

self

of

logic

are

they

are

is

not

1975

is

of

the

however

computational

ally

thinking

examination

level

them

able

with

Dosto

can

intellectual

to

it

Laird

in

People

not

concede

Evans

they

for

checking

discipline

the

point

The

prepare

requires

no

logical

have

the

validity

result

problem

make

1982

where

they

valid

circum

as

them

certain

only

that

1972

deduc

made

errors

to

in

valid

have

these

pushed

rational

make

forgetting

be

because

methods

are

they

away

to

important

an

Revlis

theories

deductions

or

it

as

that

their

people

that

explain

prepared

conscious

full

that

know

refute

Johnson

are

of

not

reasoning

1974

but

the

to

They

premises

that

even

and

that

these

at

invalid

could

Wason

of

is

can

is

attempt

make

observation

do

theorists

manoeuvre

Erickson

makes

sometimes

the

this

deduction

practice

also

misunderstanding

with

. g

competence

in

authors

Freud

they

course

Human

behaviour

of

conclusion

any

and

They

Of

human

that

follow

their

theories

never

The

of

logical

fallible

Nietzsche

ways

not

that

of

the

competence

in

does

valid

it

moreover

deduction

account

of

logical

It

logical

foibles

of

behave

merely

draw

repudiate

view

and

Dostoyevsky

proposed

because

Nietzsche

about

premises

might

Our

but

view

follies

irrationality

or

irrational

separate

they

yevsky

interests

may

rational

provide

have

of

tions

it

They

consequences

with

the

with

best

inherently

1977

the

disagree

thought

by

darker

stressed

psychologists

people

Evans

much

reconcilable

their

guided

render

development

Stuart

inferences

this

for

objective

impossible

rational

Some

to

the

serve

of

invariably

invention

John

those

attacked

contemplates

have

desire

optimally

logical

mind

one

who

this

may

when

hard

those

incapable

see

1884

concerns

take

Indeed

it

and

beings

must

of

human

it

commentators

humanity

of

logic

view

generalization

Frege

the

matter

Other

and

people

of

century

is

valid

that

laws

nineteenth

on

argue

the

competence

not

be

subjective

proposItIons

are

logic

depend

thought

related

that

ultimately

mathematics

of

believed

people

may

and

laws

done

so

These

meta

way

for

Thus

logicians

the

the

to

32

Chapter 2

be capable of sound pre - theoretical intuitions . Yet , logic would hardly

have been invented if there were never occasionswhere people were

uncertain

about

the

status

of

an inference

. Individuals

do

sometimes

formulate their own principles of reasoning, and they also refer to

deductions in a meta-logical way . They say, for example: " It seemsto
follow that Arthur is in Edinburgh , but he isn't , and so I must have
argued wrongly ." These phenomena merit study like other forms of
meta - cognition

(see e.g. Flavell , 1979 ; Brown , 1987 ) . Once the meta -

cognitive step is made, it becomes possible to reason at the meta- metalevel, and so on to an arbitrary degree. Thus , cognitive psychologists
and devotees of logical puzzles (e.g. Smullyan, 1978; Dewdney , 1989)
can in turn make inferences about meta - cognition . A psychological
theory of deduction therefore needs to accommodate deductive

competence, errors in performance , and meta- logical intuitions (cf.

Simon , 1982 ; Johnson - Laird , 1983 ; Rips , 1989 ) .
Several

ways

computational

exist

to

characterize

level . Many

deductive

competence

theorists - from Boole

at the

(1847 ) to Mac -

namara (1986)- have supposed that logic itself is the best medium .
Others , however , have argued that logic and thought differ . Logic is
monotonic, i .e. if a conclusion

follows

subsequent premise can invalidate

from some premises , then no

it . Further

premises lead mono -

tonically to further conclusions, and nothing ever subtracts from

them . Thought in daily life appearsnot to have this property . Given
the premises :
Alicia

has a bacterial

infection

If a patient has a bacterial infection , then the preferred treatment for

the patient is penicillin .
it follows validly :

Therefore , the preferred treatment for Alicia is penicillin .

But , if it is the

case that :

Alicia is allergic to penicillin .

then

common

- sense dictates

that

the conclusion

should

be withdrawn

But it still follows validly in logic . This problem suggeststhat some

inferences in daily life are " non - monotonic " rather than logically valid ,

e.g. their conclusions can be withdrawn in the light of subsequent

33

The CognitiveScience
of Deduction

information . There have even been attempts to develop formal systems of reasoning that are non -monotonic (see e.g. McDermott and
Doyle , 1980). We will show later in the book that they are unnecessary. Nevertheless, logic cannot tell the whole story about deductive
competence.
A theory at the computational level must specify what is computed , and so it must account for what deductions people actually
make. Any set of premises yields an infinite number of valid conclusions. Most of them are banal. Given the premIses
,.
~.

Ann is clever.
Snow is white .
the following conclusions are all valid :
Ann

is

clever

Snow

is

They

must

vidual

when

and

white

be

apart

more

true

given

constraints
The

The

of

that

1964

; johnson

Joe

is

at

an

conveys

more

affairs

Joe

is

which

at

only

home

in

Joe

is

valid

decrease
deductions

1983

Mary

) .

Thus

at

her

is

of

its

must

at

be

least

indi

Hence

guided

three

extra

information

which

false

can

be

of

( see

Bar

conjunction

office

away

traced

possible

- Hillel

such

and

back

to

states

of

Carnap

as :

information

one

they

proportion

sane

the
as

no

them

semantic

out

that

throw

on

Yet

(i .e .

constituents

rules

out

more

states

of

turn

disjunction

semantic

than

to

drawing

life

conclusions

not

rules

and

daily

true

of

suggests

depends

assertion

home

in

their

white

are

dream

information

is

premises

evidence

semantic

- Laird

snow

would

deduction

The

is

philosophy

affairs

the

constraint

concept

medieval

and

that

govern

first

clever

logician
a

logic

white

is

make

than

logical

is

Ann

from

reasoners

by

snow

and

conveys

more

semantic

information

than

the

inclusive

at

home

or

Mary

deduction

it

is

at

cannot

One
that

datum
do

decrease

her

increase

in

support
semantic

office

or

both

semantic

of

information

the

information

constraint

but

is
,

such

as

that
:

it

can

valid

34

Chapter2

Joe is at home .
Therefore , Joe is at home or Mary is at her office , or both .
seem odd or even improper

(see Rips , 1983 ) .

A second constraint is that conclusions should be more parsimo nious than premises . The following

argument violates this constraint :

Ann is clever .
Snow is white .
Therefore , Ann is clever and snow is white .
In fact , logically
conclusion

untutored

individuals

declare that there is no valid

from these premises . A special case of parsimony is not to

draw a conclusion

that asserts something

that has just been asserted.

Hence , given the premises :

If J ames is at school then Agnes is at work .
James is at school .
the conclusion :
J ames is at school and Agnes is at work .
is valid , but violates this principle , because it repeats the categorical
premise . This information

can be taken for granted and , as Grice

(1975 ) argued , there is no need to state the obvious . The develop ment of procedures for drawing parsimonious

conclusions is a chal -

lenging technical problem in logic .

A third constraint is that a conclusion
something

new , i .e., something

should , if possible , assert

that was not explicitly

stated in the

premises . Given the premise :

Mark is over six feet tall and Karl is taller than him .
the conclusion :
Karl is taller than Mark , who is over six feet tall .
is valid but it violates this constraint because it assert nothing

new . In

fact , ordinary reasoners spontaneously draw conclusions that establish

relations that are not explicit in the premises .
When there is no valid conclusion that meets the three con straints , then logically

naive individuals

say, " nothing

follows " (see

35

The Cognitive Scienceof Deduction

e . g . Johnson
is wrong

- Laird

ises . The
three

are

point

is

constraints

straints

or

play

no

that

reasons
, Ch

posits

in

, an

Formal

Rules

Three

main

been

proposed

1 . Foffilal

of

the

that

Fonnal
assumed

rules

are

used

the

role

to

argued
.

children

and

Piaget

operations

No
to

further
the

soning

There

are

demonstrated
pure

logic

in

there

is

as the

rule

Jean

to

deduction

mental

have

) of

originally
containing

ponens
to

e . g . Piaget
actions

set
by

models

logic

modus

own
a

develop

mental

psychologist

of

emphasize

and

early

about

the

, which

, 1953

) . He

reflect

" formal

their

unequivocal

be

introduced

to

the

algebra

than

grounds

order

conclusion

. Theorists

( see

their

to

( or

for

first

yields

) are

Piaget

need

more

that

is

on

operations
teens

,"

. lnhelder

nature

of

formal

inherent

is nothing

constraints
deduce

dominant

. The

late

supposed

operations

calculus

that

internalize

write

of
To

a new

of

conclusions

been

, such

, p . 305

. They

process

interpretations
to

ultimately

are

( 1958

for

long

the

process

( Johnson

search

conclusions
was

children

This

which

inference

logic

that

them

inference

have

derive

of

in -

.
of

question

of

quite

Level

the

counterexamples

theories

may

competence

a set

to reach

con

. They

them

the

the

for

to

Algorithmic

about

way

themselves

, and

of

deductive

, and

for

scientists

that

are

without

formal

theory

conform

rationality

draw

the

any

prem

meets

aware

, which

of

response
any

that

are
in

theory

, to simplify

at

rules

procedures

premises

of

people

inference

3 . Semantic

that
, our

from

conclusions
people

deduction

deductions

cognitive

- specific

of

, the

follow

represented

summary

of

by

valid
that

process

Theory

speaking

that

no

claim

awareness

classes

rules

2 . Content

the

information

: A

) . Logically

mentally

that

semantic

is

not

yields

conclusions

maintain

there

. 3 ) . In

rationality
the

, 1984

conclusions

do
are

part

dependent
, 1983

that

they

direct

Bara

always

. We

Laird

on

and

. There

the

for

deductive
to

since
of

propositional

rejecting

out

this

banal

operations

though

. In

short

, rea -

itself

account
must

correspond

logic

calculus

competence
rule

these

propositional

: we

depend
valid

have
on
,

already

more
conclusions

than
.

36

Chapter2

Moreover

, Piaget

1975
in

; Braine

and

sufficient

detail

for

had

genius

ments

to

is

it

so

to

asking
,
even

great

the

from

that

he

in

to

a
and

vagueness

of

have

; Ennis
his

. He
experi

theory

The

masked

effort

energy

theory

inventing

his
.

no

1960

program

for

himself

often

describe

computer

questions

Piaget

readers

Parsons

failed

modelled

right

but

( see

) , and

be

the

them

Form

more

tics

in

to

left

to

its

under

detect

its

and

Keenan

1969

, 1971

had

in

is

of

quantifier

of

films

equivalent

Some

to

of

his

scope

Everybody

is

two

different

two

quantifiers

Everybody

is

we

the

predicate

For

any

and

can

not

natural

by

loved

by

is
by

some
y .

. As

( see

these

logic

accounts

conform
in

1970

to

the

, a universal

outside

has

, such

no

in

scope

of

negation

clear

- cut

devices

the

relative

for

as :

depending
mean

the

on

scopes

of

somebody

grammar

language

paraphrase
calculus

loved

admired

can

Such

somebody

. It

of

; Lakoff

) . What

of

rules
them

1970

quantifiers

indirectly

linguis
form

fonnal

from

, 1972

quantifier

sentence

of

- Laird

negation

interpretations

x , there
is

admired

are

loved

the

of

in

logical

transformational

English

only

scope

existential

. A

has

which

an

logic

indicating

that

found

the

derived

; Jackendoff

notion

are

films

unlike

by

be
of

existence
be

; Johnson

quantifiers
the

his

the
to

1969

, 1972
the

logical

within

all

can

analyses

inspired

Seuren

; Harman

common

behaviour

Not

analysis

deductions

originally

logical
proposed

presupposed

enable

were
Leech

to

have
often

that

analyses

But

guide

linguists

sentences

e .g .

Linguistics

orthodox

Many

inference

is

for

idiosyncratic
1983

Logical
A

perhaps
it

flaws

was

Rumain

answer

inadequacy
stand

' s logic

or

other

" Loglish

"

( the

is

language

that

resembles

as :

y , such

that

if

a person

then

is

a person

37

The CognitiveScience
of Deduction
It can also mean :
There is somebody whom everybody is loved by .

(There is some y , for any x , such that y is a person and if x is a person ,

then x is loved by y .)
Often , the order of the quantifiers in a sentence corresponds to their
relative scopes, but sometimes it does not , e.g.:
No - one likes some politicians .
(For some y , such that y is a politician , no x is a person and x likes y .)
where the first quantifier
second .
Theories
many different

in the sentence is within

of logical form
linguistic

the scope of the

have more recently

frameworks , including

emerged within

Chomsky 's (1981 )

" government and binding " theory , Montague grammar (Cooper ,

1983), and Kamp 's (1981) theory of discourse representations. The
Chomskyan theory postulates a separate mental representation of
logical form (LF), which makes explicit such matters as the scope of
the quantifiers, and which is transformationally derived from a representation of the superficial structure of the sentence (S- structure).
The sentence, "Everybody is loved by somebody," has two distinct
logical forms analogous to those above. The first corresponds closely
to the superficial order of the quantifiers, and the second is derived by
a transformation that moves the existential quantifier , " somebody," to
the front - akin to the sentence:
Somebody, everybody is loved by .
This conception of logical form is motivated by linguistic considerations (seeChomsky , 1981; Hornstein , 1984; May , 1985). Its existence
asa level of syntactic representation, however , is not incontrovertible .
The phenomena that it accounts for might be explicable, as Chomsky
has suggested (personal communication , 1989), by enriching the
representation of the superficial structure of sentences.
Logical form is, of course, a necessity for any theory of deduction that depends on formal rules of inference. Kempson (1988) argues
that the mind 's inferential machinery is formal , and that logical form
is therefore the interface between grammar and cognition . Its structures correspond to those of the deductive system, but , contrary to

38

Chapter2

Chomskyan

theory , she claims that it is not part of grammar , because

general knowledge

can playa

role in determining

represents . For example , the natural interpretation

the relations it

of the sentence :

Everyone got into a taxi and chatted to the driver .

is that each individual
interpretation

chatted to the driver of his or her taxi . This

, however , depends on general knowledge , and so log -

ical form is not purely a matter of grammar . Kempson links it to the

psychological

theory of deduction

advocated by Sperber and Wilson

(1986 ) . This theory depends on formal rules of inference , and its

authors have sketched some of them within the framework of a
" natural deduction " system .
One linguist , Cooper (1983 ) , treats scope as a semantic matter ,
i .e. within the semantic component
grammar , which is an application

of an analysis based on Montague

of model - theoretic semantics to

language in general . A different model - theoretic approach , " situation

semantics ," is even hostile to the whole
formal manipulation offormal
and Etchemendy , 1989a ,b) .
Formal Logic in Artificial

notion

of reasoning as the

representations (Barwise , 1989 ; Barwise

Intelligence

Many researchers in artificial intelligence

have argued that the predi -

cate calculus is an ideal language for representing

knowledge

(e.g.

Hayes , 1977 ) . A major discovery of this century , however , is that

there cannot be a full decision procedure

for the predicate calculus .

In theory , a proof for any valid argument

can always be found , but

no procedure

can be guaranteed to demonstrate

is invalid . The procedure

that an argument

may , in effect , become lost in the space

of possible derivations . Hence , as it grinds away , there is no way of

knowing

if , and when , it will stop . One palliative

is to try to mini -

mize the search problem for valid deductions by reducing the number
of formal rules of inference . In fact , one needs only a single rule to
make any deduction , the so- called
1965 ) :
A or B , or both
C or not - B , or both
... A or C , or both .

" resolution

rule " (Robinson ,

39

The CognitiveScience
oj Deduction

The rule is not intuitively

example:

obvious , but consider the following

Mary is a linguist or Mary is a psychologist.

Mary is an experimenter or Mary is not a psychologist.
Therefore , Mary is a linguist or Mary is an experimenter .
Suppose that Mary is not a psychologist, then it follows from the first
premise that she is a linguist ; now , suppose that Mary is a psychologist , then it follows

from the second premise that she is an experi -

menter . Mary must be either a psychologist or not a psychologist, and

so she must be either a linguist or an experimenter .
Table

2 .1 summarizes

proving , which
showing

the

main

steps of

relies on the method

that the negation

resolution

theorem

of reductio ad absurdum, i .e.

of the desired conclusion

leads to a

contradiction . Unfortunately , despite the use of various heuristics to

speed up the search, the method still remains intractable : the search
space tends to grow exponentially with the number of clausesin the
premises (Moore , 1982 ) . The resolution

method , however , has be -

come part of " logic programming " - the formulation of high level
programming

languages in which

programs consist of assertions in a

formalism closely resembling the predicate calculus (Kowalski , 1979).

Thus , the language PROLOG
and Mellish , 1981 ) .

is based on resolution (see e.g. Clocksin

No psychologist would suppose that human reasoners are

equipped with the resolution rule (see also our studies of " double
disjunctions " in the next chapter). But , a psychologically more plausible form of deduction hasbeen implemented in computer programs.
It relies on the method of " natural deduction ," which provides separate

rules

of inference

for

each

connective

. The

programs

maintain

clear distinction between what has been proved and what their goals
are, and so they are able to construct

chains of inference

forwards from the premises and working

working

backwards from the con -

clusion to be proved (see e.g. Reiter , 1973; Bledsoe, 1977; Pollock ,

1989). The use of forward and backward chains was pioneered in
modern times by Polya (1957) and by Newell , Shaw, and Simon
(1963); as we will see, it is part of the programming language,
PLANNER

40

Chapter2

Table

simple

example

The

of

deduction

Mary

All

to

is

has

read

( Psychologist

For

any

Not

the

read

some

books

books

into

aim

of

deduction

the

Mary

Step

For

showing

reductio

ad

that

the

absurdum

resultant

set

. e

of

negate

the

propositions

is

"

all

"

Psychologist

all

"

Any

the

"

Some

be

is

by

its

read

its

( Read

Mary

into

have

as

( Read

deleted

replaced

psychologist

book

&

connectives

can

"

&

( Book

psychologists

given

books

some

some

variables

( for

Translate

quantifiers

disjunctions

work

is

function

done

books

returns

( f

so

"

and

by

( the

some

argument

of

Skolem

value

the

presence

called

requires

eliminate

the

function

function

consisting

of

which

some

( Psychologist

Mary

( Not

( Psychologist

( Not

( Read

Step

clauses

is

not

cancels

( Psychologist

they

to

in

simulation

1983

the

logical

common

enable

desired

be

assertion

inconsistent

disjunctions

to

leave

Assertion

of

to

the

the

empty

resolution

set

follows

at

once

in

this

way

because

its

Theories

advocated

as

psychologists

1975

it

these

fonn

for

of

the

both

. g

forward

an

up

their

most

plausible

Braine

1986

posit

premises

make

differences

to

the

Macnamara

theories

their

readers

application

absurdum

been

many

of

further

reduced

conclusion

ad

uses

All

outweighs

to

by

is

reductio

Laird

program

Rips

containing

to

Psychological

by

Johnson

premises

in

assertions

has

logic

The

deduction

mental

any

assertions

out

of

Rules

1975

ing

set

inconsistent

Natural

to

cancel

led

disjunct

Mary

had

Read

both

second

rule

assertions

Fortnal

of

the

Whenever

negation

or

for

Mary

are

resolution

( Psychologist

two

necessary

out

Not

These

rule

( fMary

the

it

Mary

Apply

thus

them

proving

some

with

evaluated

theorem

Translate

InconsIstent

. g

"

have

Mary

conclusion
.

be

resolution

psychologist

psychologists

. .

Step

"

and

we

what

they

outline

least

of

will

minds

Osherson

at

process

Indeed

own

1978

and

backward

initial

but

account

one

chaining

recover

have

three

in

of

The

Cognitive

Science

Johnson

soning

Deduction

Laird

partly

in

of

( 1975

based

Table

. 2

on

along

introducing

41

proposed

natural

with

those

disjunctive

theory

of

deduction

of

the

conclusions

Its

two

propositional

rules

rea

are

other

summarized

theories

The

rule

000

or

leads

( or

to

both

deductions

infonnation

this

If

it

is

I t

is

frosty

frosty

Johnson

or

rules

such

fail

Rumain

ways

their

P2

Second

played

P1

next

make

' t

throw

semantic

many

the

won

"

for

people

inference

be

Yet

played

the

"

the

according

up

of

They

their

differ

and

"

or

the

can

are

from

is

Table

1978

the

to

make

of

forI

. 2

rules

of

has

) 1al

and

formal

number

is

Braine

Johnson

any

in

pri

exploit

series

Braine

format

rule

. g

connect

first

theory

it

for

consequences

approach

in

"

his

its

like

way

for

described

( see

others

the

procedures

to

follow

( and

prepare

Where

have

heart

. 2

and

and

P2

and

avoids

then

to

rule

to

the

Laird

' s

in

proposi

following

the

above

includes

step

example

rule

He

At

the

only

deduction

Braine

used

colleagues

theory

disjunctive

or

game

that

be

natural

so

Therefore

to

to

the

ponens

his

be

and

Table

First

and

in

' t

can

the

on

in

form

then

modus

then

1983

presented

remarked

difficult

proposed

that

and

based

If

won

as

Braine

theories

rules

foggy

assumption

tions

have

unacceptable

be

therefore

one

hypothetical

is

we

seem

would

game

Laird

rule

two

as

thus

it

it

the

auxiliary

ing

mary

and

rule

Therefore

P1

that

away

without

an

for

n .

need

by

for

some

building

their

example

the

auxiliary

effects

rule

rules

directly

into

such

as

the

the

main

Therefore

again

antecedent

allowing

for

This

any

idea

number

is

also

of

adopted

propositions

by

Sperber

in

and

the

Wilson

disjunctive

( 1986

42

Chapter 2

Table 2.2
The principal fonnal rules of inferenceproposedby three psychological
theoriesof deduction

Johnson
-Laird Braine

Rips

Conj unctions
A , B
A

... A

& B

&

... A

Disj unctions
A

or

.' . A

B , not - A
or

... B

Conditionals
If A then
If A
A

or

~ B

B , A
B then

.' . B
C , A

. ' . If A then

,' . C
B

Negated conjunctions
not (A & B ) , A ... B
not (A & B ) 0.. not - A or not - B

A & not - B ... not (A & B )

Double
not

negations

not - A

. 0. A

De Morgan 's laws

A & (B or C) ... (A & B) or (A & C)

Reductio
A

ad absurdum

~ B &

not - B

. . . not - A

Dilemmas
A

or

B , A

~ C , B ~ C

... C

or

B , A

~ C , B ~ D

. . . C or

Introduction of tautologies
. .. A

or

not - A

Notes
" + " indicates that a rule is postulated by the relevant theory .
" A ~ B" means that a deduction from A to B is possible. Braine 's rules
interconnect any number of propositions , as we explain in the text . He
postulates four separate rules that together enable a reductioad absurdumto
be made. johnson - Laird relies on procedures that follow up the separate
consequences of constituents in order to carry out dilemmas.

43

The Cognitive Scienceof Deduction

Braine, Reiser, and Rumain (1984) tested the theory by asking

subjects to evaluate given deductions. The problems concerned the
presence or absence of letters on an imaginary blackboard , e.g.:
If there

is either

There

a C or an H , then

there

is a P .

is a C .

Therefore

, there

is a P .

The subjects' task was to judge the truth of the conclusion given the
premises. The study examined two potential indices of difficulty the number of steps in a deduction according to the theory , and the
" difficulty weights " of these steps as estimated from the data. Both
measurespredicted certain results: the rated difficulty of a problem ,
the latency of response (adjusted for the time it took to read the
problem ), and the percentage of errors. Likewise , the number of
words in a problem correlated with its rated difficulty
of response

and the latency

Rips (1983) has proposed a theory of propositional reasoning,

which

he has simulated

in a program

called ANDS

(A Natural

Deduction System). The rules used by ,the program- in the form of

procedures -

are summarized

given conclusions

in Table 2 .2 . The program

evaluates

and it builds both forward - chains and backward -

chains of deduction , and therefore maintains a set of goals separate

from

the

assertions

that

it

has

derived

. Certain

rules

are

treated

as

auxiliaries that can be used only when they are triggered by a goal , e.g.:
A , B
Therefore

, A and

which otherwise could be used ad infinitum at any point in the proof .

lfthe program can find no rule to apply during a proof , then it declares
that the argument is invalid . Rips assumes that rules of inference are

available to human reasonerson a probabilistic basis. His main method

of testing the theory has been to fit it to data obtained from subj ects
who assessedthe validity of arguments . The resulting estimates of the

availability of rules yielded a reasonable fit for the data as a whole .

One surprise , however , was that the rule :
If A or B then
A
Therefore

, C

44

Chapter 2

had

higher

worth

nothing

called

for

these
the

that

problems

major

are

rules

that

ought

to

apply

from

assign

It

leads

is

content

Content

It

one

out

is

of

Conversely

14

information

content

assertions

use

logic

of

. Yet

is

what

once

it

formal

the

rules

been

rules

argue

interpretation

of
,

radical

of

inference

at

the

they

has

information

premise
(, f

That
,

during

formal

problem

additional

to

on

deductive

proponents

import

make

of

only

form

based

of

of

The

to

reasoners

experiment

Only

chance

theories

form

his

semantic

content

influence

logical
that

than

regardless

reasoners

different

however

specific

its

ponens

logical

content

exerts

away

maintained

the

the

modus

in

thrown

performance

apply

to

of

deductions

be

which

by

their

content

premises

valid

chance

for

affected

rule

better

,
than

are

that

abstracted

the

evaluated

difficulty

rules
:

simple

to

problems
better

people

formal

of

was
16

evaluated

that

half

the

information

other

were

than

semantic

16

of

is

availability

or

to

alternative
that

,
have

- Specific

Rules

Second

Theory

Algorithmic

Level

Content

- specific

artificial

intelligence

gramming

that

each

rely

series

of

leads

to

made

Hence

of

Mary

Paul

on

the

is

is

Mark

which

is

( Programmer

of

can

derived

PLANNER

simple

1971

is

gone

in

much
- like

assertions

such

series

of

Mary

in

Mark

the

following

the

plans

way

proof

actions

goal

proof

- base

before

to

data

de

gone

as
a

pro

many

leading

has

notation

in

hypothetical

and

represented

its

has

same

language
as

and

what

before

the

and

be

workers

in

It

proofs

from

plan

has

) .

programmer

Paul

by

implemented

following

what

be
in

linguist

originally

between

each

by

psychologist

( Psychologist

( Linguist

pioneered

( Hewitt

resemblance

can

set

were

were

conclusion

plan

They

assertions

written

sisting

inference

PLANNER

possible

program

of

language

scendants

is

rules

.
con

A
-

45

The Cognitive Scienceof Deduction

The

assertion

to

this

"Mary is a psychologist," is obviously true with respect

data

All

psychologists

are

expressed

base

are

formulate

not

such

the

if

has

Mary

is

first

( Goal

a
The

goal

tence

rules

of

inference

One

way

to

to

infer

the

goal

truth

base

the

consequent

that

of

for
in

is

is

that

is

psychologist

an

If

the

matches
and

an

to

above

with
sets

assertion

the

up

in

experimenter

"

of

with

conclusions

assertion
base

that

and

effect

so

it

sentence

the

to

following

It

looks

be

goal

fails
for

eval

) )

- chains

supplemented

specific
data

that

by

an

the

matches

Mary

satisfied

Another

is

data

satisfied

inference

specific

the

base

too

using

such

heuristic

advice

formulate

and

which

The

rules

so

the

sen

program

about

con
can

how

to

way

(x )

in

which

to

its

Mary

is

derive

content

- specific

rule

is

is

This

an

to

experimenter

has

program
.

satisfied

by

an

input

assertion

such

as

springs

The
rules

is

psychologist

response

x )

x ) ) )

antecedent

procedure

Mary

( Psychologist

( Experimenter

Wherever

(x )

( Assert

even

( Antecedent

such

data

above

backward

follows

base

the

consequent

is

certain

satisfy

assertion

rule

" Mary

structs

the

can

( Psychologist

This

be

its
an

with

uated

as

experimenter

such

rule

x )

evaluate

searches

find

but

procedure

program

it

to

an

( Experimenter

enables

program

as

by

such

x ) ) )

experimenter

to

is

( Psychologist

which

assertions

rule

(x )

( Goal

assertions

experimenters

as

( Consequent

it

General

life

and

asserts

that

is

an

experimenter

the

effect

can

of

construct

adding

the

forward

further

- chains

assertion

of

to

inference

the

data

using

as

46

Chapter2

Content -specific rules are the basis of most expert systems,

which are computer programs that give advice on such matters as
medical diagnosis, the structure of molecules, and where to drill for
minerals. They contain a large number of conditional rules that have
been culled from human experts. From a logical standpoint , these rules
are postulates that capture a body of knowledge . The expert systems,
however , use them asrules of inference (seee.g. Michie , 1979; Duda ,
Gaschnig, and Hart , 1979; Feigenbaum and McCorduck , 1984). The
rules are highly specific. For example, DENDRAL , which analyzes
mass spectrograms (Lindsay, Buchanan, Feigenbaum, and Lederberg,
1980), includes this conditional rule :
If there is a high peak at 71 atomic massunits
and there is a high peak at 43 atomic massunits
and there is a high peak at 86 atomic massunits
and there is any peak at 58 atomic massunits
then there must be an N - PROPYL - KETONE3 substructure.
(seeWinston , 1984, p. 196). Most current systemshave an inferential
" engine" which , by interrogating a user about a particular problem ,
navigates its way through the rules to yield a conclusion . The condi tional rules may be definitive or else have probabilities associatedwith
then , and the system may even use Bayes theorem from the probability calculus. It may build forward chains (Feigenbaum, Buchanan,
and Lederberg, 1979), backward chains (Shortliffe , 1976), or a mix ture of both (Waterman and Hayes- Roth , 1978).
Psychologists have also proposed that the mind uses content specific conditional rules to represent general knowledge (e.g. Ander son, 1983). They are a plausible way of drawing inferences that depend
on background assumptions. The proposal is even part of a seminal
theory of cognitive architecture in which the rules (or "productions "
as they are known ) are triggered by the current contents of working
memory (see Newell and Simon , 1972, and Newell , 1990). When a
production is triggered it may, in turn , add new infoffilation to
working memory , and in this way a chain of inferences can ensue.
A variant on content - specific rules has been proposed by Cheng
and Holyoak (1985), who argue that people are guided by "pragmatic reasoning schemas." These are general principles that apply to

The Cognitive Scienceof Deduction

47

a particular domain . For example, there is supposedly a permission

schema that includes rules of the following sort:
If action A is to be taken then precondition B must be satisfied.
The

schema

is intended

to govern

actions

that

occur

within

a frame

work of moral conventions , and Cheng and Holyoak argue that it

and other

similar

schemas

account

for

certain

aspects of deductive

performance.
Content plays its most specific role in the hypothesis that reasoning is based on memories of particular experiences (Stanfill and Waltz ,
1986 ) . Indeed , according to Riesbeck and Schank 's (1989 ) theory of

" case-based" reasoning, human thinking has nothing to do with logic .

What happens is that a problem reminds you of a previous case, and
you

decide

what

to do on the

basis

however , that when an activity

of this

case . These

theorists

allow

has been repeated often enough , it

begins to function like a content - specific rule . The only difficulty

with this theory is that it fails to explain how people are able to make
valid deductions that do not depend on their specific experiences.
General knowledge certainly enters into everyday deductions,
but whether it is represented by schemas or productions or specific
casesis an open question. It might , after all, be represented by asser
tions in a mental language . It might

even have a distributed

repre -

sentation that has no explicit symbolic structure (Rumelhart , 1989).

Structured representations , however , do appear to be needed in order

to account for reasoning about reasoning (seeJohnson-Laird , 1988,

Ch . 19) .

Mental Models : A Third Theory at the Algorithmic

Level

Neither formal rules nor content - specific rules appear to give complete explanations of the mechanism underlying deduction . On the
one hand, the content of premises can exert a profound effect on the
conclusions that people draw , and so a uniform procedure for extracting logical form and applying formal rules to it may not account
for all aspects of performance . On the other hand , ordinary individ uals are able to make valid deductions that depend solely on connec -

tives and quantifiers, and so rules with a specific content would have
to rely on some (yet to be fonnulated ) account of purely logical

48

Chapter2

competence . One way out of this dilemma is provided by a third sort

of algorithmic
theory- , which depends
on semantic -procedures .
Consider this inference :
The black ball is directly behind the cue ball . The green ball is on
the right of the cue ball , and there is a red ball between them .
Therefore , if I move so that the red ball is between me and the
black ball , the cue ball is to the left of my line of sight .
It is possible to frame rules that capture this inference (from Johnson Laird , 1975 ) , but it seems likely that people will make it by imagining
the layout

of the balls . This idea lies at the heart of the theory

of

mental models . According to this theory , the process of deduction

depends on three stages of thought , which are summarized in Figure
2 .1. In the first stage, comprehension , reasoners use their knowl edge of the language and their general knowledge
premises : they construct an internal

to understand the

model of the state of affairs that

Premises and generel knowledge

COMPREHENSION
Models
DESCRIPTION
Putatlve concluslon
URLIDRTION
:search fo
altematiue models
falsifying conclusion
Valid conclusion
Figure 2.1
The three stagesof deduction according to the model theory .

The CognitiveScience
of Deduction

49

the premises describe. A deduction may also depend on perception ,

and thus on a perceptually -based model of the world (see Marr ,
1982). In the second stage, reasonerstry to fonnulate a parsimonious
description of the models they have constructed. This description
should assert something that is not explicitly stated in the premises.
Where there is no such conclusion , then they respond that nothing
follows from the premises. In the third stage, reasoners search for
alternative models of the premises in which their putative conclusion
is false. If there is no such model , then the conclusion is valid . If there
is such a model , then prudent reasoners will return to the second
stage to try to discover whether there is any conclusion true in all the
models that they have so far constructed. If so, then it is necessaryto
search for counterexamples to it , and so on, until the set of possible
models has been exhausted. Because the number of possible mental
models is finite for deductions that depend on quantifiers and connectives, the search can in principle be exhaustive. If it is uncertain
whether there is an alternative model of the premises, then the conclusion can be drawn in a tentative or probabilistic way . Only in the
third stage is any essential deductive work c;arried out : the first two
stagesare merely normal processesof comprehension and description .
The theory is compatible with the way in which logicians for mulate a semantics for a calculus. But , logical accounts depend on
assigning an infinite number of models to each proposition , and an
infinite set is far too big to fit inside anyone's head (Partee, 1979).
The psychological theory therefore assumesthat people construct a
minimum of models: they try to work with just a single representative sample from the set of possible models, until they are forced to
consider alternatives.
Models form the basisof various theories of reasoning. An early
program for proving geometric theorems used diagrams of figures in
order to rule out subgoalsthat were false (Gelernter , 1963). Although
this idea could be used in other domains (see Bundy , 1983), there
have been few such applications in artificial intelligence . Charniak
and McDennott (1985, p. 363) speculate that the reason might be
becausefew domains have counterexamples in the fonn of diagrams.
Yet , as we will see, analogous structures are available for all sorts of
deduction .

50

Chapter2

Figure2.2
The Euler circle representation of a syllogism .

Deductions from singly- quantified premises, such as "All psychologists are experimenters," can be modelled using Euler circles
(see Figure 2.2). Psychological theories have postulated such representations (Erickson , 1974) or equivalent strings of symbols (Guyote
and Sternberg, 1981). These deductions can also be modelled using
Venn diagrams (see Figure 2.3) or equivalent strings of symbols, and
they too have been proposed asmental representations (Newell , 1981).
A uniform and more powerful principle , however , is that mental

51

The CognitiveScience
of Deduction

Figure2.3
The Venn diagram representation

of a syllogism .

models have the same structure as human conceptions of the situations they
represent (Johnson - Laird , 1983 ) . Hence , a finite set of individuals

is represented, not by a circle inscribed in Euclidean space, but by a

finite

set of mental

tokens

. A

similar

notion

of a " vivid

" representa

tion has been proposed by Levesque (1986) from the standpoint of

developing efficient computer programs for reasoning. But , there are
distinctions

between the two sorts of representation , e.g. vivid rep -

resentations cannot represent directly either negatives or disjunctions

(see also Etherington et al., 1989). The tokens of mental models may
occur in a visual image, or they may not be directly accessible to
consciousness. What matters is, not the phenomenal experience, but
the

structure

of the

models

. This

structure

, which

we

will

examine

in

52

Chapter
2

detail in the following

chapters , often transcends the perceptible . It

can represent negation and disjunction .

The general theory
accounting

of mental

models has been successful in

for patterns of performance

in various sorts of reasoning

(Johnson - Laird , 1983 ) . Errors occur , according to the theory , because

people fail to consider
therefore

all possible models of the premises . They

fail to find counterexamples

to the conclusions

that they

derive from their initial models , perhaps because of the limited

cessing capacity of working
The

model

theory

adherents of formal

memory

pro -

(Baddeley , 1986 ) .

has attracted

considerable

criticism

rules . It has been accused of being

from

unclear ,

unworkable , and unnecessary . We will defer our main reply to critics

until the final chapter , but we will make a preliminary

response here

to the three main charges that the theory is empirically

inadequate :

1. Mental models do not explain propositional reasoning: " No clear mental

model theory of propositional reasoning has yet been proposed" (Braine ,
Reiser , and Rumain , 1984; see also Evans, 1984, 1987; and Rips , 1986).
2. Mental models cannot account for performance in Wason's selection
task. The theory implies that people search for counterexamples, yet they
conspicuously fail to do so in the selection task (Evans, 1987). The criticism
is based on a false assumption. The theory does not postulate that the search
for counterexamples is invariably complete - far from it , as such an impeccable performance would be incompatible with observed errors. The theory
explains performance in the selection task.
3. Contrary to the previous criticism , Rips (1986) asserts: "Deduction - assimulation explains content effects, but unfortunately it does so at the cost of
being unable to explain the generality of inference ." He argues that a modus
ponens deduction is not affected by the complexity of its content , and is
readily carried out in domains for which the reasoner has had no previous
exposure and thus no model to employ . However , the notion that reasoners
cannot construct models for unfamiliar domains is false: all they need is a
knowledge of the meaning of the connectives and other logical terms that
occur in the premises. Conversely , modus ponens can be affected by its
content .

Conclusion
We have completed

our survey of where things stood at the start of

our research . There were - and remain -

three algorithmic

theories

of deduction . Despite many empirical findings , it had proved impos sible to make a definitive

choice among the theories .

The Cognitive Scienceof Deduction

53

References
Anderson , J. R . (1983 ) . The architecture of cognition. Hinsdale , NJ : Lawrence
Erlbaum

Associates

Baddeley , A . D . (1986 ) . Working memory. Oxford : Clarendon Press.

Bar - Hillel , Y ., and Carnap , R . (1964) . An outline of a theory of semantic
information . In Y . Bar - Hillel (Ed .) LAnguage and Information . Reading , MA :
Addison - Wesley .
Barwise , J. (1989). The situation in logic. Stanford : Center for the Study of
Language and Information .
Barwise , J., and Etchemendy , J. (1989a) . A situation -theoretic account of reasoning
with Hyperproof Unpublished paper . Center for the Study of Language and
Information , Stanford University .
Barwise . J., and Etchemendy , J. (1989b ) . Model - theoretic semantics . In M .
Posner (Ed .), Foundations of cognitive science
. Cambridge , MA : Bradford Books ,
MIT

Press .

Bledsoe , W . W . (1977) . Non - resolution theorem proving . Artijicial Intelligence, 9,

1 - 35 .

Boole , G . (1847/ 1948 ) . The mathematical analysis of logic, being an essaytowards a

calculusof deductive reasoning. Oxford : Basil Blackwell .
Boole , G . (1854) . An investigation of the laws of thought on which arefounded the
mathematical theories of logic and probabilities. London : Macmillan .
Braine , M . D . S. (1978 ) . On the relation between the natural logic of reasoning
and standard logic . PsychologicalReview, 85, 1- 21 .
Braine , M . D . S., Reiser , B . J., and Rumain , B . (1984) . Some empirical justifi cation for a theory of natural propositional logic . The psychology of learning and
motivation (Vol . 18) . New York : Academic Press.
Braine , M . D . S., and Rumain , B . (1983) . Logical reasoning . In J. H . Flavell and
E . M . Markman (Eds.), Carmichael's handbook of child psychology. Vol. III : Cognitive
development(4th Edition ) . New York : Wiley .
Brown , A . (1987 ) . Metacognition , executive control , self- regulation , and other
more mysterious mechanisms . In F. E . Weinert and R . Kluwe (Eds.) , Metacognilion , motivation, and understanding. Hinsdale , NJ : Lawrence Erlbaum Associates.
Bundy , A . (1983) . The computer modelling of mathematical reasoning. London :
Academic

Press .

Charniak , E ., and McDermott , D . (1985) . Introduction to artijicial intelligence.

Reading , MA : Addison - Wesley .
Cheng , P. W ., and Holyoak , K . J. (1985). Pragmatic reasoning schemas. Cognitive Psychology, 17, 391 - 416 .
Chomsky , N . (1965) . Aspects of the theory of syntax. Cambridge , MA : MIT Press.

54

Chapter
2

Chomsky, N . (1981). Lectures

ongovernment
andbinding
. Dordrecht: Foris.
Clocksin, W . and Mellish, C. (1981). Programming
in Prolog
. New York:
Springer-Verlag.
Cohen

Behavioral

Cooper

( 1981

260

Duda

. ,

of

reasoning

NJ

and
&

Evans

journal

Evans

ledge

&

Evans

St

St

of

St

Theme

Dordrecht

and

( 1979

variations

Model

Reidel

design

Michi

Scientijic

41

in

( Ed

.) ,

Press

Ameri

handle

the

Prospector

Expert

systems

in

the

Piaget

' s

propositional

logic

theory

.) ,

) .

In

to

1 -

analysis

of

Loyola

behavior

symposium

in

on

formal

cognition

syllogistic

Vol

2 ) .

Hills

. ,

Brachman

. J

. ,

Preliminary

and

Kautz

report

Mimeo

( 1989

Murray

) .

Vivid

Hill

) .

Toward

a
,

) .

( 1984

The

29A

statistical

theory

297

psychology

306

of

6 ,

. ,

Heuristic

451

( 1987
,

) .

75

) .
74

systems

in

and

468

of

reasoning

Quarterly

deductive

reasoning

London

Rout

processes

in

reasoning

British

In

Beloff

and

Colman

( Eds

.) ,

case

study

the

analytic

Reasoning
93

Buchanan

solving

Expert

Associates

( 1982

University

( Ed

Borgida

demonstrated

ability

45

set

Solso

( 1977

reasoning

theory

Edinburgh

Psychology

problem

micro

. ,

and

using

Lederberg
the

- electronic

( 1979

DENDRAL

age

) .

On

-program -

Edinburgh

generality
In

Edinburgh

Michi

University

Feigenbaum

intelligence

Flavell

Frege

' 5

( 1979

( 1884

. ,

) .

and

- Hill

1959

( 1963

) .

) .

and

Realization

of

Feldman

( 1984

the

) .

world

The

arithmetic

geometry

( Eds

.) ,

generation

Pan

34

Oxford

906

theorem

Computers

Books

monitoring

Psychologist

of

fifth

London

cognitive

American

Thefoundations

to

Metacognition

and
.

challenge

inquiry

Feigenbaum

McCorduck

computer

developmental

Gelernter

andJapan

cognitive

McGraw

Psychology

In

. ,

Paul

Feigenbaum

Press

Survey

.) ,

syntactic

' s

Laboratories

Hart

tractable

Psychological

and

experimentally

exploration

Children

Experimental

Kegan

Journal

Evans

Bell

St

of

and

) .

be

370

Erlbaum

knowledge

puzzles

91

. ,

) .

In

317

People
-

Edinburgh

Lawrence

Etherington

AT

( 1974

and

Research

irrationality

mineral

( 1975

tasks
,

( Ed

Educational

Erickson

88

for

) .

) ,

age

Review

( 1989

Gaschnig

- electronic

Ennis

NJ

system

micro

human

Quantification

( January

consultant

dale

) .

Can

Sciences

( 1983

) .

Brain

Dewdney
can

and

Basil

new

911

thought

area

of

Blackwell

- proving

and

Artificial

machine

New

Yark

The CognitiveScience
of Deduction

55

Grice, H . P. (1975). Logic and conversation. In P. Cole and] . L. Morgan (Eds.),

Syntax and semantics
. Vol. 3: Speechacts. New Yark : Seminar Press.
Guyote, M . J., and Sternberg, R . J. (1981). A transitive- chain theory of syllogistic reasoning. CognitivePsychology
, 13, 461- 525.
Harman, G. (1972). Deep structure as logical form . In D . Davidson and
G. Harman (Eds.), Semantics
of naturallanguage
. Dordrecht : Reidel .
Hayes, P. (1977). In defense of logic . Proceedings
oj the Fifth InternationalJoint
Conference
on Artijicial Intelligence
, 559- 565.
Henle, M . (1962). The relation between logic and thinking . Psychological
Review,
69, 366- 378.
Henle, M . (1978). Foreward to Revlin , R . and Mayer, R . E. (Eds.), Human
reasoning
. Washington, D .C .: Winston .
Hewitt , C. (1971). PLANNER : A language for proving theorems in robots.
Proceedings
of the SecondJointConference
on Artificial Intelligence
.
Hornstein , N . (1984). Logic as grammar: An approachto meaningin language
.
Cambridge, MA : Bradford Books, MIT Press.
lnhelder, B., and Piaget, J. (1958). The growth of logicalthinkingfrom childhoodto
adolescence
. London : Routledge & Kegan Paul.
jackendoff , R . (1972). Semanticinterpretationin generativegrammar
. Cambridge,
MA : MIT Press.
Johnson- Laird, P. N . (1970). The interpretation of quantified sentences. In G. B.
Flores d' Arcais and W . J. M . Levelt (Eds.), Advancein psycholinguistics
. Amsterdam:
North - Holland .
Johnson- Laird, P. N . (1975). Models of deduction. In R . J. Falmagne (Ed.),
Reasoning
: Representation
andprocess
in childrenand adults. Hillsdale, NJ: Lawrence
Erlbaum Associates
.
Johnson-Laird, P. N . (1983). Mental models
: Towardsa cognitivescience
of language
,
inference
and consciousness
. Cambridge: Cambridge University Press.
Johnson- Laird, P. N . (1988). The computerand the mind: An introductionto cognitive
science
. London : Fontana.
Johnson- Laird, P. N . and Bara, B. (1984). Syllogistic inference. Cognition, 16,
1- 61.
Kamp , J. A . W . (1981 ) . A theory of truth and semantic representation . In
J. Groenendijk , T . Janssen, and M . Stokhof (Eds.) , Formal methods in the study
of language. Amsterdam : Mathematical Centre Tracts .
Kant , I . (1800/ 1974 ) . Logic. Trans . R . S. Hartman and W . Schwartz . New York :
Babbs

- Merrill

Keenan , E . L . (1971 ) . Quantifier

255 - 284 .

structures in English . Foundations of Language, 7,

56

Chapter2

Kempson , R . M . (1988) . Logical fonn : The grammar cognition interface . journal

ojLinguistics , 24 , 393 - 431 .
Kowalski , R . A . (1979) . A logicfor problem solving. Amsterdam : Elsevier .
Lakoff , G . (1970 ). Linguistics and naturallogic . Synthese, 22 , 151- 271 .
Leech , G . N . (1969) . Towards a semanticdescription of English. London : Longmans .

Leibniz, G. (1765/ 1949). New essays

concerning
humanunderstanding
. Trans. A . G.
Langley . LaSalle , IL : Open Court .

Levesque
, H. J. (1986). M aking
30, 81- 108.

believers out of computers . Artificial Intelligence,

Lindsay , R ., Buchanan , B . G ., Feigenbaum , E . A ., and Lederberg , J. (1980) .

Applications of artificial intelligencefor chemical inference: The DENDRAL
project.
New

Yark

: McGraw

- Hill

Macnamara , J. (1986 ) . A border dispute: The place of logic in psychology. Cambridge ,

MA

: Bradford

Books

, MIT

Press .

Marr , D . (1982) . Vision: A computational investigation into the human representation

and processingof visual information. San Francisco : W . H . Freeman .
May , R . (1985) . Logical form : Its structure and derivation. Cambridge , MA : MIT
Press .

McDermott , D ., and Doyle , J. (1980) . Non - monotonic

logic . I . Artificial Intelli -

gence , 13 , 41 - 72 .

Michie , D . (Ed .) (1979) . Expert systems in the micro-electronic age. Edinburgh :

Edinburgh University Press.
Mill , J. S. (1843 ). A system of logic. London : Longman .
Moore , R . C . (1982 ) . Automatic deduction : A . Overview . In P. R . Cohen and
E . A . Feigenbaum (Eds.) , The handbook of artijicial intelligence ( Vol . 3) . Los Altos ,
CA : William

Kaufmann

Newell , A . (1981) . Reasoning , problem solving and decision processes: The

problem space as a fundamental category . In R . Nickerson (Ed .), Attention and
performance( Vol . 8) . Hinsdale , NJ : Lawrence Erlbaum Associates.
Newell , A . (1990) . Unified theories in cognition. Cambridge , MA : Harvard Uni versity Press.
Newell , A . and Simon , H . A . (1972) . Human problem solving. Englewood
NJ : Prentice - Hall .

Cliffs ,

Newell , A ., Shaw , J. C ., and Simon , H . A . (1963) . Empirical explorations with

the Logic Theory Machine . In E . A . Feigenbaum and J. Feldman (Eds.), Computers and thought. New York : McGraw - Hill .
Osherson , D . (1975 ) . Logic and models of logical thinking . In R . J. Falmagne
(Ed .), Reasoning: Representation and process in children and adults. Hillsdale , NJ :
Lawrence

Erlbaum

Associates

57

The CognitiveScience
of Deduction
Parsons

, C . ( 1960

logician

' s viewpoint

Partee
Egli

and

) . Logic

, J . ( 1989

Press

Bradford

Polya

, G . ( 1957

Reisbeck

Intelligence

, R .

( 1973

Rips

, L . J .

it

( 2d

: A

? In

different

points

"

II : A

R . Bauerle
of

: Manchester

prolegomenon

).

ed . ) . New

, R . C . ( 1989

Associates

) . Two
of

in

view

, U .

. Berlin

University

. Cambridge

Press

, MA

: MIT

York

: Doubleday

) . Inside

case - based

reasoning

. Hillsdale

semantically
of

guided
the

models
Verbal

Third

of

deductive

system

InternationalJoint

syllogistic

Learning

) . Cognitive

( 1986
the

and

for

automatic

Conference

reasoning

Verbal

processes

, L . J . ( 1989

Robinson
ciple

in

Rumelhart

) . A

P . A . M

: Feature

Behavior

on

Artificial

selection

, 14 , 180

propositional

Shartliffe

, E .

A .

- 195

reasoning

and

and

and

. Psychological

a microstructural

Operators

and

Computer

of

bounded

the

.) ,

Arizona

, 31 , 8 ~ - 116
resolution

prin

.
-

12 , 23 - 41 .

account
Similarity

of

and

human

analogical

reason

reasoning

.
nucleus

University
:

.) ,

on

( Eds
of

. Cognition

based
,

. Harnish

: University

knaves

logic

( Eds

R . M

. Tucson

Machinery

Press

) . MYCIN

. Brand
belief

knights

Ortony

University
).

M
and

- oriented

: Cambridge

( 1976

In

Computing

) . Towards

( 1969

. Cambridge

Elsevier

machine

and

: Cambridge

grammar

of

Associationfor

S . Vosniadou

Cambridge
Seuren

of the

psychology

, D . E . ( 1989

In

muddles
of knowledge

) . The

. J . A . ( 1965

. Journal

) . Mental

representation

Rips

: A

Press
- based

contribution

to

the

theory

of

.
medical

consultations

. New

Yark

Simon

, H . A . ( 1982
: MIT

Press

Smullyan

Sperber

Schank

( 1975

other

psychology

from

. Manchester

a person

to solve

. journal

Press

MA

thinking

, 90 , 38 - 71 .

Problems

ing

build

. Proceedings

, L . J . ( 1983

Review

logical

. Pp . 41 - 46 .

conversion
Rips

or

Semantics

of

, 51 , 75 - 84 .

mathematics
.) ,

psychology

to

Erlbaum

R .

growth

) . How

- proving

Revlis

and

, C . K . , and

theorem

( Eds

) . How
Books

: Lawrence

Reiter

Stechow

' s " The

, J . ( 1953

Pollock

Piaget

of Psychology

) . Semantics

A . von
- Verlag

Piaget

and

. BritishJournal

, B . H . ( 1979
,

Springer

NJ

) . Inhelder

, R . M

logical
,

puzzles
D .,

) . Models

rationality

( Vols

and

2 ) . Cambridge

and

. ( 1978

).

What

. Englewood
Wilson

D .

is

the

name

Cliffs

, NJ

( 1986

) .

of

this

book

: Prentice
Relevance

- Hall
:

? The

riddle

of

Dracula

and

Communication

and

cognition

Oxford: BasilBlackwell.
Spinoza
, B. (1677/ 1949). Ethics
. Trans. W. H. White and Ed. by J. Gutmann.
New York: Hafner.

58

Chapter2

Stanfill, C., and Waltz , D . (1986). Toward memory-basedreasoning. Communicationof theAssociationforComputingMachinery, 29, 1213- 1228.
Wason, P. C., and Johnson- Laird, P. N . (1972). Psychology
of reasoning
: Structure
and content
. London : Batsford. Cambridge, MA : Harvard University Press.
Waternlan, D . A ., and Hayes- Roth , F. (Eds.) (1978). Pattern directedinference
systems
. New York : Academic Press.
Winston , P. H . (1984). Artificial intelligence(2d ed.). Reading, Massachusetts
:
Addison- Wesley.

This excerpt from

Mind Readings.
Paul Thagard, editor.
1998 The MIT Press.
is provided in screen-viewable form for personal use only by members
of MIT CogNet.
Unauthorized use or dissemination of this information is expressly
forbidden.
If you have any questions about this material, please contact
cognetadmin@cognet.mit.edu.