Situational features influencing mentalistic explanations of action


Material Information

Situational features influencing mentalistic explanations of action
Physical Description:
ix, 74 leaves : ; 29 cm.
Montgomery, Derek E., 1965-
Publication Date:


Subjects / Keywords:
Human behavior   ( lcsh )
Causation   ( lcsh )
Perception in children   ( lcsh )
Attribution (Social psychology)   ( lcsh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1993.
Includes bibliographical references (leaves 71-73).
Statement of Responsibility:
by Derek E. Montgomery.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001938138
notis - AKB4273
oclc - 30935139
System ID:

Full Text









I wish to thank my wife, DeMaris, for her unfailing

encouragement and support throughout the span of this

project. Jennifer Bedore, Michelle DeLuca, Marc

Essinger, Kevin Miller, and Steve Wu each shared their

time and talents in preparing and/or administering the

studies. Thanks also go to my friend James Forbes for

his assistance with Study 2. I would also like to

gratefully acknowledge my committee members, Dr. John

Biro, Dr. M.J. Farrar, Dr. Ira Fischler, and Dr. Patricia

Miller, for giving of their valuable time to assist with

this study. More generally, Drs. Farrar and Miller have

shared their expertise in developmental psychology again

and again in both formal and informal settings, on

matters both general and specific.

Finally, special thanks go to Dr. Scott A. Miller

for his encouragement and his invaluable methodological

and conceptual suggestions in all phases of this project.

His expert guidance throughout my graduate career is

greatly appreciated. He has exemplified the wise and

virtuous mentor one generally encounters only in

fictionalized terms. His example is one I will try to

emulate throughout my career.


This research was supported by a NICHD predoctoral

traineeship to the author in the Department of

Psychology, University of Florida (National Research

Service Award No. 1T32HD07318).



ACKNOWLEDGMENTS ....................................... iii

LIST OF TABLES ......................................... vi

ABSTRACT .............................................. vii

INTRODUCTION ............................................ 1

Literature Review.. ....................................2
Present Study ....................................... 11

EXPERIMENT 1 ........................................... 13

Method .............................................. 13
Subjects ......................................... 13
Materials.............. ........................... 13
Procedure ........................................ 13
Scoring .......................................... 22
Results ............................................. 22
Discussion .......................................... 33

EXPERIMENT 2 ......................... .................. 40

Method .............................................. 43
Subjects ......................................... 43
Materials ........................................ 43
Procedure ........................................ 44
Scoring .......................................... 52
Results ............................................. 53
Discussion .......................................... 59

GENERAL DISCUSSION............. .......................... 64

REFERENCES ............................................. 71

BIOGRAPHICAL SKETCH ..................................... 74


Table Page

1. Description of Scenarios Used in Experiment 1...15

2. Control Events Used in Experiment 1............. 19

3. Mean Percentage of Mental Explanations Across
Actors in Experiment............................ 23

4. Age Comparisons Within Actor in Experiment 1....26

5. Actor Comparisons Within Age in Experiment 1....27

6. Stories Used in Experiment 2 ....................45

7. Examples of Target-Directed Mental Explanations
in Experiment 2 ................................. 49

8. Explanations Used by Preschoolers Other than
Target-Directed Mental Explanations in Experiment
2 ............................................... 51

9. Mean Scores in the Presence and Absence of the
Situational Features in Experiment 2............ 57

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of
the Requirements for the Degree of Doctor of Philosophy




Derek E. Montgomery

August 1993

Chairperson: Scott A. Miller
Major Department: Psychology

Actions are often assumed by humans to be compelled

and directed by mental desires and beliefs. Although

young children often attribute desires and beliefs to

actors, little is known about how they calibrate those

attributions, knowing when an occurrence should be

explained on the basis of an actor's mental state and

when an occurrence has other causal mechanisms (e.g.,

biological or physical). Accordingly, the goal of this

research was to assess how children know when an action

is mentalistically caused. Specifically, the research

assessed whether variable self-initiated movement and

perceptual access to goals act as cues guiding children's

distinctions between actions in the world that should and

should not be construed as mentalistically caused.


In Study 1, preschoolers, first graders, and adults

chose whether an occurrence was caused by mental or

physiological states of the actor. The occurrences

differed in whether they were accompanied by the

situational features listed above and in whether the

actor was an artifact, insect, mammal, or human. At all

ages mental explanations were preferred in the presence

(but not the absence) of the three situational cues (ps <

.001). Also, when the three cues were present, first

graders and adults differentiated between artifacts and

living organisms, providing mental explanations

significantly more often when the actor was alive (ps <

.001). Only adults, however, differentiated between

"higher" and "lower" organisms, using mental explanations

significantly more often (ps < .001) for mammals and

humans than for insects.

In Study 2, preschoolers and first graders witnessed

scenarios in which perceptual access or variable self-

initiated movement was either present or absent. They

were then asked to verbalize why the action occurred. In

both age groups, mental explanations that appealed to the

volition of the actor were more common when the feature

of interest was present than when it was absent, ps <

.001. Taken together, these studies suggest that young

children are capable of abstracting features of action

when construing its cause. Doing so apparently plays a


vital role in learning how (and when) to generalize

mentalistic explanations to new situations.


Throughout each day the perceptual world of humans

is filled with the occurrence of physical actions taking

place in varying forms and substance. A central task for

adults and children is to understand and interpret the

nature of these various actions by inferring their

causes. One way to set about solving this task is by

postulating that an actor's beliefs and desires are

causally connected to that individual's actions. When

mental states such as these are viewed as causal

constructs, they are generally embedded within the

various postulates and principles comprising a "theory of

mind" (or "folk psychology") (see Gopnik, 1993; Wellman,

1990, for recent discussions about the nature of theories

of mind).

In general, the central principles of a theory of

mind state that unobservable mental states are causally

connected to (a) sensory perception, (b) other mental

states, and (c) behavioral output. For example, seeing a

toy placed in a red drawer causes Billy's belief about

its location. If one is aware that Billy desires to play

with the toy, then it can be inferred that his belief

about the toy's location will direct the action,

motivated by his desire, of approaching and opening the

red drawer. In this example, Billy's perception of the

toy, his belief and desire about the toy, and his action

toward the toy are all interconnected, forming a

framework within which one can predict and explain his

action. Construing action within this framework is, in

essence, folk psychology.

Literature Review

Although much research has demonstrated that young

children can explain actions by appealing to actors'

mental beliefs and desires (see Astington & Gopnik, 1991;

Flavell, Miller, & Miller, 1993, for recent reviews),

little is known about how children calibrate those

attributions, knowing when an occurrence should be

explained or expected on the basis of the mental states

of an actor and when an occurrence has other causal

mechanisms such as internal biological mechanisms or

external physical forces. Accordingly, the aim of the

following research is to explore potential situational

cues which potentially guide children in parsing

occurrences and phenomena into those which are and are

not explicable via particular mental desires and beliefs.

The question being explored is whether young children

employ mental explanations of action indiscriminately or

in a manner that is principled and constrained according

to the presence of specific situational features.

What systems of explanation might compete with folk

psychology in children's determinations of an action's

cause? Following Dennett (1987), two systems of

explanation which are alternatives to folk psychology are

the Design Stance and the Physical Stance. Explanations

from the Design Stance are based upon the assumption that

certain outcomes and actions occur because the actor,

whether an artifact or natural kind, is specifically

designed to do the particular action being explained.

That is, design explanations are based upon the

assumption that the actor is designed, or hardwired, with

parts (biological or mechanical) that have a task to

perform. For instance, one might explain a moth's

evasive behavior in the presence of a bat, a chief

predator, in terms of the moth's functional design

("moths are made to survive and that is why this moth is

avoiding the bat" or "something inside moths compels them

to survive and flee from bats"). The ringing of an alarm

clock would be explained as occurring because the clock

is designed to perform that very function. As considered

in this paper, a Design Stance is not so much a

teleological argument for an activity, but instead a

rudimentary assumption that features intrinsic to an

organism or machine are causally responsible for an

observed activity or occurrence.

There is evidence that young children will appeal in

a general way to the nature of an actor's design as an

explanation of action and occurrences. When asked to

explain the action of natural kinds (e.g., "why do

rabbits hop?"), preschoolers (72%) and young grade school

children (73%) spontaneously offered a cause intrinsic to

the actor (Gelman & Kremer, 1991). Examples of such

explanations include "rabbits are made to hop" and "the

leaf just makes itself change colors" (Gelman & Kremer,

1991, p. 411). When asked directly, children generally

indicated that something "inside" of both artifacts and

natural kinds caused events such as a bird flying or a

car going up a hill (Gelman & Kremer, 1991). In

addition, Springer and Keil (1991) have found that

preschoolers believe that external characteristics of

organisms are caused by internal, natural forces (e.g., a

flower is red due to inheritance) rather than external

causes such as a human painting the flower.

Similarly, Gelman and Wellman (1991) found that

preschoolers' explanations of the behavioral functions of

objects and organisms were more influenced by the

internal constitution of the actor than by its external

characteristics. For example, if one removed the insides

of a dog, preschoolers realized that the dog could no

longer physically function (e.g., bark or eat food).

Further, they understood that removing external features

of the dog, such as its fur, did not have a similar

effect. These examples illustrate young children's

assumption that functional design parts of organisms and

artifacts can cause actions and outcomes.

R. Gelman (1990) suggests that preschoolers have an

"innards principle" which assumes that natural objects

have something inside them which governs their movement

and change. Interviews with preschoolers (Gelman, 1990)

reveal that they have realistic ideas and beliefs about

the internal constitution of animate and inanimate

objects. Animates were often said to have blood and

bones inside them whereas the insides of inanimates were

said to consist of human-made materials like cotton and

plastic. Thus, even by preschool age, children appear to

have-expressible beliefs about the mechanical and

biological design features characteristic of a variety of


In addition to a Design Stance, one may adopt a

Physical Stance, explaining actions, for example, by

appealing in a rudimentary way to physical impingements

responsible for movement. To illustrate, a common

assumption is that movement can be caused when one object

physically contacts another with a sufficient speed and

force to propel it (Michotte, 1963). In the absence of

such contact, movement is regarded as self-initiated.

The contrast between these differences in the initiation

of movement is salient even to infants, who dishabituate

in response to such changes in movement (Leslie, 1984).

More generally, children are often capable of

distinguishing intentional from overt physical causes of

behavior (see Wellman & Gelman, 1992, for a review of

this literature). To illustrate, Bullock and Gelman

(1979) found that 4-year-olds explain the popping of a

jack-in-the-box as resulting from the impact of a ball

rather than the intentions of the jack-in-the-box.

Given the various systems of explanation possible

for any action, how does one decide which explanation is

most appropriate for any given action? Is there a clear

preference across situations for one 'type of explanation

over another, do preferences vary across situations, or

do children simply regard all types of explanations as

equally preferable, adequate, and acceptable in every

instance? As noted earlier, the following research is

designed to begin exploring these various possibilities

by attempting to assess whether children are sensitive to

the presence or absence of specific situational features

which influence the manner in which an action's cause is


Scenarios were developed in which situational

features believed to implicate various mental

explanations were either present or absent. If these

features influence the manner in which an action is

interpreted, then their presence should predictably

result in mental explanations, whereas their absence

might suggest the plausibility of searching for an

alternative explanation. Studies 1 and 2 postulate three

such features and explore their effects on children's

explanations of action.

What situational features might imply that an action

can be explained mentalistically? One feature is the

self-initiation of action (Premack, 1990). Self-

initiated action, considered in its broadest sense, is

action that begins without any external forces acting on

it (Mandler, in press). It has long been pointed out

that adults commonly infer that self-initiated movement

is psychologically or voluntarily caused (Heider &

Simmel, 1944; Michotte, 1963, ch. 13). Generally, if an

action is not self-initiated, then the only alternative

explanation must involve appealing to the external

forces) impinging upon the actor (e.g., a Physical

Stance) rather than an internal mental state or design


Secondly, in a related fashion, the action must be

variable; in other words, if an action is clearly

intentional, then for every stimulus more than one

response should be possible. If an action is instinctual

or a reflex, then it can easily be explained by appealing

to the physical stimulus eliciting the action or the

biological instinct compelling the action (see Dretske,

1988, ch.2). For example, if moths are known to be

instinctually drawn to flames, one could easily explain

the moth's flight toward flames by appealing to its

biological design ("it has to," "that's its nature," "all

moths are drawn to light").

A third feature which may implicate mental

explanations is the presence of perception toward a

critical fact or event underlying an action. In folk

psychology, beliefs and desires typically originate via

perceptions; therefore, it is implausible for a

particular belief or desire to be invoked as an

explanation of an action if there has never been a

perceptual experience by which that belief and/or desire

could have been formed. For example, if Mary is walking

in the direction of an ice cream store, one cannot

explain Mary's action in terms of her belief about the

location of the store or her desire for ice cream if she

has never seen nor heard about the store and therefore

does not know it exists. Of course, some beliefs are

mentally constructed but, as a general rule, even the

preceding beliefs from which the new belief is derived

were formed at some point by a perceptual stimulus.

Thus, to ascertain the specific beliefs and desires

responsible for an action one must note the actor's

perceptions responsible for forming the relevant beliefs

and desires. The types of desires that can be used to

explain a given action are limited by the knowledge one

brings to a situation.

To address the extent to which these three features

may influence the way action is construed, Study 1

explores their influence across a wide range of actors,

varying from artifacts to humans. It is possible, for

example, that there may be a positive relation between

the neurological sophistication of the actor and a

willingness among subjects to impute mental states as

explanations of that actor's behavior. In this case,

subjects may use the biological differences among species

of actors to infer how the action is caused instead of

using the situational features surrounding the action.

How far-reaching the features actually are can be

illustrated by exploring whether or not they override in

importance the biological nature of the actor.

The relation between the biological nature of the

actor and inferring an action's cause can be framed by

reviewing two ways in which the essence of folk

psychological terms is defined by some philosophers of

mind. Dennett (1987) suggests that beliefs and desires

can be imputed to any system, ranging from thermostats to

humans, so long as the actions of those systems can be

reliably predicted and explained by appealing to mental

constructs. The Instrumentalist position (Dennett, 1987)

states that since the action of any number of biological

or artifactual systems can be accurately predicted and

explained by the constructs of beliefs and desires, the

utility, and hence importance, of these constructs is

independent of whatever ontological assumptions one might

hold about them. The real properties of mental states,

whatever they may be, are secondary considerations when

compared to their utility in explaining action. In

contrast, others (e.g., Fodor, 1987) emphasize that

beliefs are significantly more than instrumental

constructs: Mental states are said to have real physical

properties found only in highly sophisticated

representational systems such as the human brain. In

other words, a defining attribute of a mental state is

its real physical properties. Thus, physical

considerations might be regarded as preconditions for

deciding whether an action has a mentalistic explanation.

These two philosophical positions can be taken to

suggest different ways in which subjects might invoke

mental explanations of action when the situational

features are present. On the one hand, they might impute

mental states to explain action regardless of the

biological features of the actor, so long as certain

contextual preconditions are met (i.e., the three

situational features above). They may do so even if they

believe that only complex organisms truly have a mind.

However, it is also possible that one's assumption about

the literal physical properties of beliefs and desires

may affect one's decision to use mental states as

explanations of action. In other words, there might be a

direct relation between judgments of whether a particular

organism is sophisticated enough to be sentient and a

willingness to use a mental state as a causal explanation

for the action of that organism.

If subjects in Study 1 are Realists, then even

preschoolers, who have correctly denied mental states to

artifacts in past research (e.g., Berzonsky, Miller,

Woody-Ramsey, & Harris, 1988; Gelman, Spelke, & Meck,

1983), should mentalistically explain the actions of

artifacts infrequently (especially when a plausible

alternative explanation is provided for them). In

contrast, most children (as well as adults) accept that

"higher" mammals and humans have sentient experiences

(e.g., Carey, 1985; Inagaki & Sugiyama, 1988), meaning

that, if they are Realists, one should expect a fairly

high degree of mental explanations for the actions of

nonhuman mammals. The number of mental explanations for

insects should be intermediate if subjects are Realists

since the percentage of children and adults who indicate

that insects can have thoughts and feelings rarely

approaches 100% and varies greatly with differences in

wording and target (e.g., Carey, 1985; Inagaki &

Sugiyama, 1988).

Present Study

The purpose of Study 1 is to examine whether the

presence of the three situational features detailed above

is a sufficient condition for compelling mentalistic

explanations of action. Equally important is the

question of whether the four categories of actors used in

this study (human, mammal, insect, artifact) interact

with the presence of the three features in subjects'

determinations of whether the action is to be explained

mentalistically. The presence of such an interaction

will be tested by determining whether there is a

significant difference in the number of children's

mentalistic explanations among the four categories of


Subjects were provided with both design and mental

explanations following each action. 'Design explanations

appeal to functional, physical properties that are

intrinsic to the actor. If subjects consistently use

them primarily for specific categories of actors, then it

can be inferred that the effect of the three principles

in producing mentalistic explanations is mediated by the

category of the actor. Explanations from a Design rather

than Physical Stance are chosen as alternatives to mental

explanations because the movement in each scenario is

self-initiated, meaning that only internal or intrinsic

features of the actor are plausible explanations.




Three age groups, each consisting of 24 subjects,

were tested. There were 13 male and 11 female

preschoolers (range 4-8 to 5-8; mean = 5-2), 14 female

and 10 male first graders (range = 6-4 to 7-8; mean = 7-

0), and 12 male and 12 female adults (introductory

psychology students). Children were drawn from

predominantly middle-class populations.


For each of 16 scenarios, an illustration depicting

the action to be explained was drawn on 20 x 28 cm paper.

Every actor was depicted realistically, with no

suggestion of anthropomorphic features on the nonhuman

actors. Across drawings each actor was sized roughly

proportional to one another. Actors did not directly

face the viewer, preventing the presence of emotional

cues which might be derived from facial expressions

(Lillard & Flavell, 1990).


Actions in each of the 16 scenarios were self-

initiated, accompanied by sensory perceptual access to

facts surrounding the action, and were one of two

occurrences that potentially could have taken place (see

Table 1). Upon describing each action to the subject,

the experimenter concluded by asking, "Why is (the actor)

doing that?" Design and mental explanations of the

action were then provided and subjects were asked to make

a forced choice between the two.

Across all stories, each of the three situational

features was presented as similarly as possible. To

indicate an action's variability, each story contained

the phrase, "(The actor) can either (action A) or (action

B)." For the "sensory perception" feature, every story

mentioned a specific sensory ability in the present

tense: "(The actor) sees/feels/hears etc. (an object)."

In each story perception formed a belief about the

nature, location, and/or existence of the object toward

which the action was directed. The "self-initiating"

feature was established by explicitly stating the verb

describing the action (e.g., "The actor is

walking/eating/flying/running" etc.). In no case was a

potential external causal agent mentioned, implied, or


Altogether there were four scenarios for each of the

four actor categories. Of the four scenarios within each

actor category, two were "neutral" stories and two were

"anomalous" stories (Bartsch & Wellman, 1989). Neutral

stories merely presented the occurrence of an action.

For example, in one scenario a cat ran inside a house.

Table 1

Description of Scenarios

Actor Perception

Billy sees bed

Judy sees clouds

Harvey sees snake

Susie sees lollipops

Horse sees apples

Squirrel sees a nut

Cat sees open door

Dog sees car keys

hopper hears loud

Spider sees juice

Ant smells sugar

Bee sees colors

Orange sees orange
Machine rocks

Train wheels feel

Computer hears inquiry

Car feels water

Used in Experiment 1


search outside/
search under bed

carry umbrella/
leave it behind

pick it up/
run away

purchase lollipop/
purchase candy bar

run toward barn/
run toward apples

stay in tree/
retrieve nut

fall asleep/
go through door

stay home/
follow driver to car

remain stationary/
climb tree

walk to web/
drink juice

walk home/
walk to sugar

fly to colors/
fly home

put rocks in bag/
leave rocks on ground

go short way/
go long, safe way

print answer A/
print answer B

keep top down/
put top up


searching for toy
under bed

removes umbrella
from closet

runs away

purchasing a lollipo

running toward

retreiving nut
to tree

running through

following driver
to car

climbing a tree

walks to juice

walks to sugar

flying to colors

placing rocks in

going long, safe way

printing answer A

putting car top up

No details were provided to suggest that the action was

somehow inconsistent with whatever desires the cat may

have had. In neutral stories the action could

sufficiently be explained as motivated by the actor's

desire (e.g., "the cat wants to go into the house"). In

contrast, anomalous stories described an action that,

placed in the context of the actor's apparent desires,

would actually contradict those desires. For example, in

one scenario the protagonist was looking for his airplane

under a couch although it was really outside. In another

story, a bee flew toward colorful rocks that, from a

distance, had the misleading appearance of flowers. In

anomalous stories a desire-based explanation is

insufficient because the action does not bring about the

outcome apparently desired by the actor. However,

reference to the actor's belief (e.g., "he thinks the

plane is under the couch") explains the seemingly

erroneous direction of the action. The two different

types of stories were presented to increase the

generality of the study by providing settings in which

both desires (neutral stories) and beliefs (anomalous

stories) are appropriate and commonly used explanations.

Accordingly, "wants to" was the mentalistic

explanation provided for subjects in the neutral

situation and "thinks that" was the mental alternative

for the anomalous stories. The design explanations

paired with the mental explanations were "made so that it

will do X" and "something inside makes it do X."

Preschoolers have used and understood these terms in past

research (Gelman & Kremer, 1991; Gottfried & Gelman,

1992). The four explanations were consistently worded in

the following fashion: "Because the actor is made so that

(it will do the behavior)," "Because something inside the

actor makes it (behavior)," "Because the actor wants to

(behavior)," and "Because the actor thinks (fact

explaining behavior)." For example, in a scenario in

which subjects were told about a horse moving toward an

apple that looked real but was actually plastic, they

were asked, "The horse is running toward the plastic

apple. Why is the horse doing that? Because the horse

is made so that it will go to things that look like

apples or because the horse thinks the plastic apples are

real?" In another scenario, in which a convertible car

automatically put its top up after water reached the

seats, children were asked, "Why is the car doing that?

Because the car thinks it is raining or because something

inside the car makes it put the top up when it gets wet?"

Overall, the two mental explanations were paired

equally often with the two different design explanations.

The four resulting pairs were equally distributed within

each of the four categories of actors. Thus, in the

anomalous stories (there were two per each actor

category), "thinks" was paired once with each of the two

design responses within every actor category. Similarly,

in the neutral stories "wants" was paired with the two

design explanations for each actor category. For each

age group, the 16 scenarios were presented in four random

orders, with the only constraint being that no more than

two scenarios with the same actor category were presented

in a row. The order of presentation of the design and

mental choices was successively alternated across

scenarios. For half of the subjects, the design

explanation was presented first in the initial scenario

while the reverse was true for the remaining subjects.

Following the 16 scenarios, five control events were

presented (see Table 2). The control events were natural

or mechanical events occurring without the presence of

the three critical situational features. The events were

not illustrated and were constructed primarily to examine

whether children would reject a mental explanation and

choose a plausible alternative (design explanation) when

none of the three situational features were present.

This is a necessary control because choosing the

mentalistic explanation when the three features are

present is interpretable only if the alternative choice

is also shown to be a meaningful and comprehensible term.

The five events were administered in an invariant

order. For each of the control events one of the two


Table 2

Control Events Used in Experiment 1

1. Why does a light bulb shine?

2. Why does a toaster get hot?

3. Why does a telephone ring?

4. Why does a puppy grow?

5. Why does a baby grow?

design explanations was paired with "wants." Each design

explanation was used at least once for each of the two

types of events (animate and artifact). The order in

which design and mental explanations were presented was

alternated across trials. The control events always

followed the 16 scenarios since pilot testing revealed

that even the youngest children had a clear preference to

use design explanations for these occurrences. It was

feared that placing the events early in the procedure

might bias children's subsequent responses toward design


Finally, the relation between subjects' beliefs

about the ability of artifacts to have sentient

experiences and a willingness to mentalistically explain

artifacts' actions was assessed. Three short questions

concluded the testing session, probing children's beliefs

about whether artifacts ("machines like cars and trains")

(a) have brains, (b) can think, and (c) can feel happy

and sad. To ensure that children did not respond to the

items with a "yes" or "no" bias, two filler questions

were also asked, designed to evoke a "no" and "yes"

response respectively.

Each testing session began by telling subjects, "I'm

going to show you some pictures and tell you some

stories. In each story something happens." Subjects

were informed that they would hear two answers explaining

why the action occurred and were to choose the better of

the two. Adult subjects then proceeded to read the

scenarios themselves, examine the accompanying picture

depicting the action, and then mark their responses on

individual answer sheets. For children, two warm-up

trials were administered to familiarize them with

choosing between two causal explanations. First, they

were shown a musical top and told, "Here is a toy. When

it spins, music plays." As the musical top was spinning,

children were asked, "Why is music playing? I'm going to

give you two answers and you tell me which one is right.

Is music playing because the toy is red or because the

toy has something inside that makes it play music?" To

avoid biasing children's subsequent responses in the test

trials, the experimenter did not indicate whether the

subject's response was correct or incorrect. Moreover,

the design and mental explanations were not pitted

together so that a response bias toward one explanation

over the other could not be established in the warm-up


For the second warm-up trial, children were told to

"pretend you see some ice cream in the refrigerator and

you take it out and eat some. Now why would you do that,

because you wanted to eat ice cream or because the

refrigerator is white?" Again, the experimenter did not

reinforce or correct the child's answer to this question.

Following the second warm-up trial, children were told,

"Sometimes 'wants' and 'thinks' might be the better

answer and sometimes 'made that way' and 'something

inside' might be the better answer."

The warm-up trials were presented in the same order

as above for all children since pilot testing revealed

that young children found the spinning top particularly

engaging and thus an effective way to begin the testing

procedure. Every first grader answered both warm-up

trials correctly. For preschoolers, 83% were correct

with regard to the spinning top, but only 50% correctly

chose that one would retrieve ice cream from the

refrigerator because one wanted to eat it.


Scores were assigned to subjects by calculating the

percentage of mental explanations and design explanations

for each of the four categories of actors, as well as for

the five control events.


There were no main effects for gender or order, nor

did they interact with any of the variables of interest

(age and actor), so they will not be considered further.

Table 3 illustrates the percentage of mental

explanations of action when the three features were

present for each category of actor. Compared against a

50% baseline (chance-level), the percentage of children's

mentalistic explanations across all four actor categories

is significantly greater than chance. The percentage of

preschoolers' mental explanations when the situational

Table 3

Mean Percentage of Mental Explanations Across Actors in
Experiment 1


Pre-K 1st Adult

Type of Actor

Artifact .71** .65* .11**

Insect .74** .91** .49

Mammal .88** .98** .82**

Human .85** .96** .85**

* Means are significantly different from chance (50%) at
p < .01, two-tailed significance level

** Means are significantly different from chance (50%) at
p < .001, two-tailed significance level

features were present ranged from 71%, t(23) = 4.46, p <

.001, for artifacts to 88%, t(23) = 9.42, p < .001, for

mammals. Similarly, mental explanations for first

graders ranged from 65%, t(23) = 2.60, p < .025, for

artifacts to 98%, t(23) = 33.05, p < .001, for mammals.

Clearly, children preferred to explain action surrounded

by the three situational features with mental rather than

design explanations and, in general, ,this preference held

regardless of the biological category of the actor.

In contrast to children, adults' mental explanations

of action were reserved primarily for mammals (82%) and

humans (85%). Mental explanations in these two

categories were significantly greater than chance level

(50%), t(23) = 7.84, and t(23) = 7.89 respectively, ps <

.001. Unlike children, mental explanations for insects

(49%) were at chance-level, t(23) = .16, n.s. Adults

also differed from children by using design explanations

significantly more often than chance (89%) to explain the

actions of artifacts, t(23) = 9.08, p < .001.

A 3 (age) x 4 (actor) mixed analysis of variance

revealed a main effect for actor, F(3,71) = 70.67, p <

.001, as well as for age, F(2,69) = 33.35, p < .001.

There was also an age x actor interaction, F(6,207) =

13.06, p < .001.

Overall, preschoolers (M = .79), t(23) = 5.83, and

first graders (M = .87), t(23) = 7.87, were much more

likely to use mental explanations than were adults (M =

.57), ps < .001. The difference in the number of mental

explanations between children and adults is primarily the

result of children using mental explanations for artifact

and insect actions significantly more often than adults

(Table 4). The difference between first graders and

preschoolers was marginally significant, t(23) = 2.03, p

= .046.

Regarding the main effect of actor, mental

explanations (collapsed across age) were used equally

often for mammals (M = .89) and humans (M = .89). These

percentages of mental explanations are significantly

higher than the percentages for insects (M = .71) and

artifacts (M = .49), ps < .001. The difference between

insects and artifacts is also significant, p < .001.

Differences in the number of mental explanations

between different actors were also assessed within each

age group. Post-hoc comparisons between actors were

conducted using Bonferroni-corrected t-tests (dividing

the .05 alpha level by the number of comparisons within

each age group, 6). Adults discriminated among all of

the actors in their explanations except between mammal

and human actors (see Table 5). First graders did not

significantly differentiate in their use of mental

explanations among living organisms, but they did

differentiate between artifacts and the remaining three

animate actors (see Table 5). In contrast, although the

percentage of mental explanations by preschoolers

Table 4

Age Comparisons Within Actor in Experiment 1



A < P**

A < 1**

P & 1 n.s


A < P*

A < 1**

P < 1*


A & P n.s

A < 1**

P & 1 n.s


A & P n.s

A & P n.s

P & 1 n.s

Note. The differences in mental explanations between
actors were calculated using Bonferroni-corrected t-
tests. A = Adults, 1 = First graders, and P =

* p < .01, two-tailed significance level
** p < .001, two-tailed siginficance level
n.s. p > .016

Table 5

Actor Comparisons Within Age in Experiment 1


Comparison Pre-K 1st Adult

Between Actors

Artifact < Insect n.s *

Artifact < Mammal n.s *

Artifact < Human n.s *

Insect < Mammal n.s n.s *

Insect < Human n.s n.s *

Mammal < Human n.s n.s n.s

Note. The differences in mental explanations between actors
were'calculated using Bonferroni-corrected t-tests.

* p < .001, two-tailed significance level

increased from artifacts to mammals and humans, their

differentiations among these different actors were not

statistically significant.

Since children tended to explain artifacts' actions

mentalistically when the three features were present, it

is possible to explore whether there is a relation

between such explanations and their judgments about

whether artifacts can think, be happy and sad, and have

brains. For the most part, subjects ,answered these

questions sensibly, although three preschoolers and one

first grader said "no" to every question (including the

filler questions) and one preschooler answered "yes" to

every question. The question of whether artifacts have

brains proved to be easiest as every subject, with the

exception of one preschooler, answered it correctly.

Similarly, Johnson and Wellman (1982) found that a

majority of 4- and 5-year-olds believed that dolls do not

have brains. Taken together, such results indicate that

preschoolers are clearly aware that a fundamental

biological difference exists between artifacts and

animate objects. However, as this study suggests, when

children explain action they apparently do not regard the

presence of a brain as a necessary prerequisite to

imputing mental explanations.

None of the adults indicated that artifacts could

have mental experiences. Similarly, only 6/24 (25%) of

first graders indicated that artifacts could experience

emotions and 1/24 (04%) said that artifacts could think.

Compared to first graders, a significantly higher

percentage of preschoolers (58%) indicated that artifacts

could think, X2 (1, N = 48) = 16.38, p < .001. Of these

children, all but one of them also believed that

artifacts could be happy (54%), X2 (1, N = 48) = 4.27, p

< .05. Thus, there was consistency in preschoolers'

judgments in that they either assumed that both mental

states were present or that both were absent, p > .10

(McNemar's test). Overall, the correlation among

preschoolers between affirming that artifacts can

experience emotions and thoughts (range 0-2) and choosing

mentalistic explanations of artifact behavior when the

three cues were present (range 0-4) was not significant,

r (24) = .23, p > .10 (this correlation is also not

significant when removing the data of the four

preschoolers who demonstrated a "yes" or "no" bias).

This suggests that beliefs about whether artifacts can

have sentient experiences were not driving preschoolers'

decisions to mentalistically explain artifacts' action.

This also appears to be true for first graders, who used

mental explanations for artifacts a majority of the time

even-though they generally denied that artifacts can have

sentient experiences.

Turning to the control events, the overall

percentage of mental explanations when the three cues

were present (M = .75) was significantly greater than

when they were absent (in the control events) (M = .11),

F(1,69) = 665.11, p < .001. Mental explanations were

much more frequent when the three cues were present than

in the control events for preschoolers (M = .79 and M =

.28 respectively), t(23) = 10.07. p < .001, first graders

(M = .87 and M = .05), t(23) = 18.27, p < .001, and

adults (M = .57 and M = .00), t(23) = 20.02, p < .001.

For each age group, the percentage of design explanations

for the control events was significantly greater than

would be expected by chance (50%), ps < .001.

Further analysis of performance on the control

events revealed a main effect of age, F(2,69) = 30.43, p

< .001 and an interaction between presence of the three

features and age, F(2,69) = 14.38, p < .001.

Preschoolers were more likely to use mental explanations

for the control events than were first graders, t(23) =

3.90, p < .01, and adults, t(23) = 5.41, p < .001. The

difference between adults and first graders was not

significant, p > .10. The interaction reflects that

preschoolers' used mental explanations more often than

first graders for the control events, while the opposite

was true for the scenarios featuring the three cues.

Altogether then, preschoolers tended to use mental

explanations for the control events most often among the

three age groups. However, each age group clearly used

different explanations for the control events compared to

the scenarios with the three situational features.

Within the control events, preschoolers were

significantly more likely to attribute a mental

explanation for the animate occurrences ("why does a

puppy/baby grow?") (M = .42) than for the three

artifactual occurrences (M = .17), t(23) = 2.62, p <

.025. In contrast, first graders and adults were equally

likely to give design explanations for both animate and

inanimate events. Nevertheless, mental explanations

among preschoolers were more likely in the presence of

the three features than for animate occurrences in the

control events, t(23) = 3.96, p < .01. In fact, artifact

actions occurring when the three cues were present were

more likely to be explained mentally by preschoolers than

were animate occurrences in the control events, t(23) =

2.81, p < .01.

The preference among the age groups for design

explanations in the control events is notable for two

reasons. First, it indicates that, even for children,

the design explanations were not inherently awkward or

meaningless. Subjects showed a clear preference for them

in situations where perceptual access, self-initiated

movement, and behavioral variability were not factors

related to the occurrences being explained. Secondly, it

suggests that children's mental explanations are not

indiscriminate, but are sensitive to-the presence or

absence of situational constraints. That is, one can

infer that the three situational features implicate

mentalistic explanations since such explanations are

common in their presence and substantially decrease in

their absence. This issue is pursued more directly in

Study 2.

One final issue, regarding the validity of the

scenarios, is whether the four scenarios within each

actor category were explained similarly to one another.

Such similarities would suggest that variables among the

stories which were peripheral to (a) -the biological

category of the actor and (b) the presence of the three

features were not influencing how actions were explained.

The four scenarios within each category were paired and a

McNemar's test for significance was used to compare them.

Alpha-level was adjusted by using a Bonferroni correction

(dividing the .05 Alpha-level by the number of

comparisons within each actor category for each age

group, 6). For preschoolers., there were no significant

differences in the percentage of mental or design

explanations among the story pairs within each of the

four categories. In other words, within each of the

actor categories the four stories were generally

explained in similar ways by children. For the first

graders, only one pairing of items (in the Artifact

category) was answered differently. Adults were somewhat

more discriminating than children, treating three pairs

of items differently, one each in the insect, mammal, and

human categories. Because so few (4 of 72) comparisons

were statistically significant, no reliable patterns

emerged to adequately explain why the four pairs were

explained differently by subjects. In sum, subjects

apparently were not overly influenced by extraneous

variables in the scenarios. Instead, they were equally

likely to explain pairs of scenarios within a category in

the same fashion, using either the design terms or the

mental terms for each item within a pair.


Study 1 explored the extent of the influence of

situational conditions hypothesized to implicate mental

states as causes of action. Results suggested that when

actions were self-initiated, variable, and accompanied by

perceptual access informing beliefs and desires, children

consistently chose mental explanations of the action over

design explanations. They did so across a wide range of

actions and actors. Adults were more selective than

children when the situational features were present,

reserving mental explanations primarily for humans and

mammals. In the absence of the three situational

features, the proportion of mental and design

explanations was reversed for all age groups. A number

of significant points follow from these results.

To begin, this is the first study to directly

investigate the generality of children's theory of mind

by equating on important dimensions the actions of a wide

range of actors and examining the extent to which mental

explanations are used for such actions. Both

preschoolers and first graders used mental explanations

more often than would be expected by chance for all

categories of actors, doing so significantly more often

than adults for insects and artifacts. Thus, children's

theory of mind generalizes across biological categories

of actors. The generality of the theory of mind can be

regarded as robust since children were given a choice

between mental explanations and a plausible alternative

explanation. The design explanations provided an

important control for a possible linguistic bias toward

using mental explanations since the content of children's

earliest causal explanations are generally mentalistic

(Hood & Bloom, 1979).

With development, the biological nature of the actor

becomes an increasingly influential element in children's

causal inferences, overriding in importance the presence

of the three features. There is an emerging delineation

among first graders between artifacts and living actors.

Although they mentalistically explain the actions of

artifacts more often than would be expected by chance

(Table 3), and significantly more often than adults

(Table 4), they also use mental explanations

significantly less often for artifacts than for animate

actors (Table 5). Thus, although first graders are

primarily mentalistic (somewhat more so, in fact, than

preschoolers), they are nevertheless beginning to refine

and limit such explanations, with the initial demarcation

being drawn between living and nonliving actors. By

adulthood, this distinction is nearly absolute and a

further distinction between "lower" and "higher" animate

organisms is also made. Clearly, adults inferred causes

of action on the basis of category membership, inferring

mental causes significantly less for insects and

artifacts than for mammals and humans.

Children's failure to infer causes on the basis of

the actor category is likely due to factors other than an

inability to make the categorical distinctions being

tested. Preschoolers and first graders can conceptually

differentiate among artifacts, insects, mammals, and

humans (see Keil, 1992; Wellman & Gelman, 1992 for recent

discussions), selectively attributing characteristics to

each and, further, believing that the internal

constitution of a species can make it distinct and unique

from other species (see Keil, 1989). Also, one might

protest that children mentalistically explained

artifacts' action because the presence of perceptual

abilities suggested to them that the artifacts were

alive. However, this argument misses the point in that

the goal of the study was to assess what situational

features would influence children to use mentalistic

explanations. When perceptual abilities were a feature

of the action of the artifact, mentalistic explanations

were used. When they were not (in the control events)

design explanations were used.

More to the point, it is unlikely that children's

mental explanations for artifacts' actions are the result

of a pervasive animism. For example, in this study, even

the youngest children used design explanations instead of

mental explanations when explaining the activities of

artifacts in the control events. Another potential

explanation for children's preference for mentalistic

explanations across actor categories is that the

alternative design explanations were meaningless.

However, the consistent use of design explanations for

the control events demonstrates that mental explanations

were not being chosen by default when the three features

were present, but rather were being chosen over terms

equally meaningful and comprehensible to children.

Instead, it may be that children base inferences

about the causes of action more on the situational

features surrounding the action than,on the biological

category of the actor. A similar finding illustrating

the importance of situational features relative to

category membership, when drawing causal inferences,

comes from Gelman and Kremer (1991). They found that the

type of occurrence being explained (e.g., a rabbit

hopping or milk freezing) was as important as the

biological category of the actor (animate or inanimate)

in influencing such inferences. When asked whether an

occurrence had an internal cause, children were more

likely to answer affirmatively for the actions of

artifacts than for the actions of natural kinds.

However, children also used a second dimension by which

to infer causes of an occurrence: They were also more

likely to attribute internal causes to occurrences that

appeared to be self-generated than to outcomes which

appeared to be externally caused, regardless of the

biological category of the actor. Thus, internal causes

were considered more likely for artifacts than for

natural kinds and more likely for self-generated activity

than for externally generated activity.

Along the same lines as Gelman and Kremer (1991),

the present study suggests that, for children, the manner

in which an action occurs may initially be an important

basis from which to infer its cause, even more so,

perhaps, than the biological category of the actor.

Although many children in this study did not believe that

artifacts could think or experience emotion (see also

Carey, 1985), they were more flexible than adults in

their willingness to use these mental states to explain

the action of artifacts. The generality and power of the

theory of mind for children seems to come from

abstracting features of action, not from associating

categories of actors with mental states and categorically

imputing mental states as causes only to particular

species of actors. In this respect, children have the

appearance of being Instrumentalists, using mental states

as explanations of action independent of whether the

actor is sentient. Perhaps they do so because their

theory of mind has proven to generally be a successful

way to make sense of their world and, therefore, as a

matter of parsimony is extended to a variety of actors.

In contrast, adults make use of biological

categories to infer the cause of action, perhaps being

influenced by the neurological differences distinguishing

each category. Thus, with development, category

membership and features of action are integrated, meaning

that the features surrounding an action and the

biological nature of the actor are used to infer causes

of action.

The findings from Study 1 are also significant in

that they are the first attempt to systematically compare

children's use of design explanations with mental

explanations. If it is true, as Gelman (1990) argues,

that children have an "innards principle" which governs

their explanations of action, then it is important to

specify the conceptual distinctions derived from such a

principle. That is, (a) do children distinguish design

features which are internal or intrinsic to the actor

from internal mental states and (b) if so, how do they

determine when each system of explanation is causally

responsible for an action? The results of Study 1

indicate a clear distinction is made, even among

preschoolers, between the different types of

explanations. That is, preschoolers, like adults, used

the design explanations significantly more often when the

three situational features were absent than when they

were present. Thus, young children not only appeal to

internal, nonobvious, causes of action (Wellman & Gelman,

1992), but they also discriminate between different types

of nonobvious causes by consistently associating mental

and design explanations with different types of

phenomena. Future research should further address

children's comparisons between design and mental

explanations, more closely examining situational features

which implicate to children that design states are

responsible for the action being explained.

In sum, Study 1 illustrates that the three

situational features are conducive to mental explanations

of action. With development, biological considerations

increasingly become important relative to the three

features. The clear reversal in the percentage of mental

and design explanations when the three features were

absent suggests that the features are important

preconditions for explaining an action mentalistically.

Study 2 examines this issue directly. In that study,

situations are equated on every dimension except for the

presence or absence of a relevant situational feature.

The question posed is whether causal explanations change

depending on the presence or absence of each feature.


.Experiment 2 assessed whether varying the presence of

the situational features affected the nature of children's

spontaneous causal explanations of action. The features

were systematically varied so that 1/4 of the actions

children explained were externally caused (SI-), 1/4 were

self-initiated (SI+), 1/4 occurred when perceptual access to

a key event was lacking (PA-), and 1/4 occurred when

perceptual access was present (PA+).

Since the presence of behavioral variability is

generally linked with self-initiated movement, the two were

considered as simultaneously present or absent in this

study. Only in unusual circumstances, such as being

commanded to move in a particular fashion, might action be

self-initiated (physiologically, at least) without being

considered variable or optional. Similarly, externally

caused movement is not usually considered optional. An

exception occurs when one is given a choice of how to be

transported and chooses an external force as a causal agent.

Nothing in the scenarios in Study 2 suggested exceptional

instances in which self-initiated movement and behavioral

variability did not co-occur. There was no implication that

the self-initiated movement was the only action possible nor

was there any indication that the external causal agent was

deliberately chosen by the actor. Thus, self-initiated

movement and perceptual access were the primary variables of

interest in this study.

If children's theory of mind is adult-like, then they

should be sensitive to the removal of either self-initiated

movement or perceptual access, adjusting their explanations

accordingly. As discussed earlier, removing these

situational features often makes mentalistic explanations of

action implausible within an adult-like folk psychology.

For example, when an action is not self-initiated, such as

when someone is pushed into a swimming pool, the action is

readily explained as being caused by the physical event

initiating the movement (e.g., "he's going into the pool

because someone pushed him"). In a similar fashion, it is

implausible to indicate that an actor's desire for an

outcome is causing an action if perceptual access necessary

for awareness of the outcome is lacking. In such instances,

a number of other types of explanations are possible,

including appeals to either physical events, false beliefs,

or ignorance. However, an action is generally not

considered intentional if the actor is unaware that he is

accomplishing the action.

An important control in the following study is that

children's explanations of action cannot be based solely on

the intrinsic desirability of the action. This is so

because, even though each pair of actions differs in either

the presence of perceptual access or the manner in which

they are initiated, the actions themselves are identical

within each pair. In this way, any differences which emerge

in how a pair of actions is explained can only be attributed

to the differences in how each action was initiated or

perceived by the actor. For example, in one pair of stories

a bracelet is being thrown away. The manipulation is

whether the actor saw the bracelet amidst the trash before

throwing it away. Thus, the action of throwing away the

bracelet is identical for each story in the pair and only

the presence of perceptual access varies. By equating the

actions on every dimension except the feature being

manipulated, this study differs from a previous

investigation of similar issues (Smith, 1978).

For example, in that investigation (Smith, 1978) the

presence or absence of self-initiated movement was

confounded with the type of action that was occurring. To

illustrate, one self-initiated movement that children

witnessed was chewing, an activity generally presumed to be

pleasurable. There was no corresponding action that was

externally caused. In contrast to the familiar act of

chewing, an externally caused movement was an actor's hand

being hooked and pulled up by the handle of an umbrella.

Since the goal of this action is unclear and the action

itself is unfamiliar, it is not surprising that children

would be less likely to regard it as something the actor

"wants" to do when compared to a more familiar activity such

as chewing. In other words, it is important to control for

the possibility that children's desire-based explanations

are influenced more by the intrinsic-desirability of the

action than by how the action was initiated.



Subjects were drawn from predominantly middle-class

elementary schools and preschools. There were 18 children

in each of two age groups: 9 female and 9 male preschoolers

(range 4:7 to 5:6; mean age 5:1) and 10 male and 8 female

first graders (range 6:9 to 7:6; mean age 7:1). None of the

subjects from Study 1 participated in this study.


Pencil drawings were used to depict each sequence of

action. The same guidelines governing depictions of action

in Study 1 were used in Study 2. The actors in the drawings

were humans since, along with mammals, mental states were

attributed to them most often in Study 1. Refusal to

explain their actions mentalistically in Study 2, when the

situational features are absent, would be a more impressive

demonstration of the influence the features have on

mentalistic explanations than if "lower" organisms were the

actors. Also, keeping the category of actors constant

(human) permits a further test of the extent to which

children infer types of causes solely on the basis of the

biological nature of the actor. If they do so, then there

should be relatively little variation in explanations across

different situations in this study.


Altogether, there were eight pairs of scenarios (see

Table 6). Each scenario was illustrated by three separate

drawings. The first drawing depicted the protagonists and

the setting in which the action eventually occurred. The

second drawing depicted the initiation of the action and the

third picture depicted its occurrence. The pictures were

presented to children in a top-to-bottom column proceeding

from the first to last picture (Kun, 1978). After the

action in the third picture was stated, children were asked

to explain why the actor was doing the action and their

responses were transcribed by the tester. Each picture

remained on the table until the child's explanation was


To introduce the experimental procedure, children were

given two warm-up trials, one in which a physical answer was

implied (why does laughter occur when one is being tickled?)

and one in which a desire-based explanation was implied (why

does one remove ice cream from the freezer?). The

experimenter did not reinforce specifics of subjects'

answers. Following the warm-up trials, the sequential

nature of the three pictures comprising each scenario was

illustrated by telling children that presenting each new

picture in the story is "just like turning the pages of a


In 4 of the 16 scenarios the action was caused by a

physical force (SI-) (see Table 6). The external causal

Table 6

Stories used in Experiment 2













entering a pool headfirst

moving into a sandbox

going into a playground

reclining on a couch

throwing away a bracelet




carrying a cat in a laundry basket

smashing a worm with a wagon

entering a kitchen upon one's


Note. SI+ = scenarios in which self-initiated movement is
present. SI- = scenarios in which the action is externally
caused. PA+ = scenarios in which perceptual access to
either the bracelet, cat, worm, or request is present. PA-
= scenarios in which perceptual access to either the
bracelet, cat, worm, or request is absent.





event, depicted in Picture 2, was physical contact with

either a pedestrian human (couch, SI- and pool, SI-) or a

human riding a bike (sandbox, SI-, playground, SI-). In

four matching scenarios (SI+) the action was self-initiated

in Picture 2 rather than externally caused. However, the

causal agent from the SI- scenarios also appeared in Picture

2, but without contacting the protagonist. Thus, in each

pairing either a causal event (e.g., "Jill pushes Allison

with the bike") or a non-causal, equally visible physical

event (e.g., "Mary rides by on her bike") occurred in

Picture 2 and was mentioned by the narrator. If children

generally assume that salient physical events are causal,

then they could use the depicted physical event to explain

the action for either story within a pair. In contrast to

the variations in Picture 2, Pictures 1 (orienting children

to the scenario) and 3 (depicting the occurrence of the

action being explained) were identical in each matching pair

of stories.

For the remaining four pairs of scenarios, perceptual

access to a physical event shaping the action was

manipulated (see Table 6). These events were a cat crawling

into a laundry basket, a worm being dropped in front of a

moving wagon, a bracelet being dropped into a trash bag, and

a child verbally calling for another child to enter a room.

Pictures 1 and 3 were once again identical within each pair,

while the presence of perceptual access to the physical

event was varied in Picture 2. For example, in the cat

story the second picture depicted a cat crawling under

clothes in a laundry basket. For half of the subjects, the

protagonist saw the cat's action (PA+) while for the other

half the protagonist's back was turned away from the

critical physical event (PA-). In each case, the narration

accompanying the second picture noted the presence or

absence of perceptual access. Children were reminded of

this variable in the third picture (e.g., "She doesn't see

Kitty in the basket") before the action was described.

One story from each of the eight pairs in Table 6 was

administered to each subject. Of these eight stories, two

were drawn from each of the four categories (SI+, SI-, PA+,

and PA-) and were administered in one of six random orders.

Each story was present in three of the six orders, ensuring

even distribution among the subjects. The only constraints

on the ordering were (a) no more than two stories in which

the same critical feature was manipulated (perceptual access

or self-initiated movement) were administered in a row and

(b) no more than three stories were given in a row in which

a critical feature was successively either present or



There were two key elements coded from children's

spontaneous explanations of action: (a) whether the

explanation was mentalistic and (b) whether the mental

explanation was directed toward the target action being

explained. Regarding the first element, mental explanations

were defined as appeals to an internal drive, desire, or

belief of the agent. Such explanations included appeals to

beliefs (e.g., think, know), preferences/emotions (e.g.,

like, hate), desires/wishes (e.g., want to, gonna), traits

(e.g., dumb, nice), perceptions (e.g., see, hear), and

psychophysiological states (e.g., sleepy, hungry) (see

Bartsch & Wellman, 1989; D'Andrade, 1987). Also, at times

children used implicit expressions of desires, mentioning

acquisition of the goal without explicitly specifying the

actor's desire for the goal. For example, when asked why an

actor was going into a sandbox, some children merely stated

a goal ("to play," or "so she can build a sandcastle").

Table 7 provides various examples of the different types of

mental explanations given by children.

Regarding the second element of coding, it was

important to determine whether or not the explanations were

"directed" toward the target action. Specifically, target

actions, listed in Table 6, were the depicted actions which

children were explicitly asked to explain. A mental

explanation was target-directed if it (a) implied awareness

of the target action and/or (b) a desire to accomplish the

target action. Such explanations were considered to be

plausible only in situations where the target action was

either self-initiated or accompanied by perceptual access

(see Table 7). In contrast, mental explanations that were

not target-directed were explanations which did not suggest

Table 7

Examples of Mental Explanations Directed Toward the Target


Belief/Knowledge: "Because he thinks it's garbage"

(bracelet, PA+)

Feelings/Emotions: "Because she hates worms" (worm, PA+)

"Because she likes to play" (playground (SI-)*

Desires/Wishes: "She wants to play outside" (playground,


"Because she sees it and she wants to play" (sandbox,


"Because she wants to kill it" (worm, PA-)*

Desires (implicit): "So she can play" (playground, SI+)

Traits: Trait-like explanations were not used

Psychophysiological States: "Because he's getting sleepy"

(couch, SI+)

"Because he's tired" (couch, SI-)*

Perceptions: "She saw him crawl under the clothes into the

basket" (cat, PA+)

* = Mental explanations directed toward the target were
considered as inappropriate in these instances

Note. Explanations were in response to the question of why
an action, listed in Table 6, was occurring.

that the actor was motivated by a desire for the target

action and/or that the actor was unaware that the target

action was occurring. Examples of such explanations are

given in Table 8.

In contrast to mentalistic explanations of action, a

second general category of explanations used by some

children were appeals to physical events. These occurred

most often in SI- stories. Such explanations generally

included mentioning the external cause of an action or

describing an event that resulted in the ongoing action

being explained. Examples of physical explanations are also

shown in Table 8.

Sometimes, children's explanations were ambiguous in

stories where perceptual access was manipulated. That is,

it was unclear from the explanation if the child felt that

the action was intentional. To use the cat story as an

example, some explanations for why the actor was carrying

the cat toward the car included, "to go on a trip," "because

she's going somewhere," or "because she's going to do the

laundry." From these explanations, it is not readily

apparent whether or not the child believes that the actor

knows that the cat is being carried in the basket.

Therefore, when an explanation was applicable for situations

in which perceptual access was either present or absent, the

response was probed by asking the child whether the actor

knew about the key event (e.g., "Does she know the cat is in

Table 8
Explanations Used by Preschoolers Other than Target-Directed
Mental Explanations in Experiment 2

A. Other Mental Explanations

Couch (SI-) "Because he was hurt (upon being pushed)"

Cat (PA-) "Because the Kitty's in the basket and she

doesn't know it"

Cat (PA+) "Because she doesn't know he's in there"*

Worm (PA-) "Because she (the actor) doesn't watch out

where she's going"

Worm (PA+) "Because she didn't see it"*

Bracelet (PA+) "Because he doesn't know it's in


B. Physical Explanations

Couch (SI-): "Because he pushed him on the couch and

he bumped his head"

Pool (SI-): "He fell in because he pushed him"

Cat (PA-): "Because Kitty got into the basket"

These explanations were considered as inappropriate, since
they were not target-directed in the presence of self-
initiated movement or perceptual access

Note. Explanations were in response to the question of why
an action, listed in Table 6, was occurring.

the basket?"). Children's answer to the "know" question was

taken as clarifying evidence regarding their belief about

the intentionality of the act when the explanation was

ambiguous. Thus, children's initially ambiguous responses

were counted as target-directed mental explanations only if

the subject said "yes" to the "know" question.

Finally, in rare instances two different types of

answers were given by the same child. For example, one

child explained that a person who was propelled toward a

sandbox (SI-) was going into the sandbox "because Jill

pushed her in and she fell. Then she wanted to play." In

such cases, if the two answers differed in plausibility, the

more plausible of the answers (given the context of the

situation) was used in all analyses. Thus, in the above

example the child's explanation was regarded as an appeal to

a physical event.

Twenty-five percent of children's explanations were

randomly chosen and coded by'a second rater (blind to

whether perceptual access or self-initiated movement was

present). The responses were coded with regard to whether

or not the explanation was a mental explanation directed

toward the target. Interrater agreement was 92%.


One point was given for each target-directed mental

explanation. Therefore, scores could range from 0-2 for

each of the four types of scenarios, PA+, PA-, SI+, SI-.

Total scores could range from 0-4 when a key feature was

present (combining PA+ and SI+ stories) and also when a key

feature was absent (combining SI- and PA- stories). If

children correctly use target-directed mental explanations

more often when self-initiated movement and perceptual

access are present than when they are absent, their scores

should be significantly higher for the PA+ and SI+ stories

than for the PA- and SI- stories.


A 2 (age) x 2 (presence or absence of a situational

feature) mixed analysis of variance was conducted. Age was

the between subjects variable and presence was the within

subjects variable. The dependent variable was the number of

mental explanations directed-toward the target action. The

variables of gender and order were also tested for their

effect on the number of target-directed mental explanations.

However, there was not a main effect of these variables, nor

did they interact with any of the variables of interest (age

and presence), all ps > .09; therefore, they will not be

considered further.

The effect of age was not significant, F (1,34) = .89,

n.s., meaning that target-directed mental explanations were

as likely to be given by preschoolers as by first graders.

It is noteworthy that a substantial number of target-

directed mental explanations explicitly referred to the

actor's desire, using the mental verb "want" (see Table 7

for examples). When the situational features were present

(SI+ and PA+), the percentage of target-directed mental

explanations containing the mental verb "wants" was 49% for

preschoolers and 36% for first graders. Preschoolers and

first graders were just as likely to explicitly refer to the

actor's desire this way in the SI+ stories as in the PA+

stories (ts(17) = .57 and .94 respectively, ps > .10). The

percentages are comparable to Bartsch and Wellman (1989) who

found that 30% of the time preschoolers and adults

spontaneously invoked a desire term (generally "wants") to

explain an action.

In contrast, target-directed mental explanations

featuring expressions of belief (i.e., "think" and "know")

were used infrequently by children (10% for preschoolers and

1% for first graders) in the PA+ and SI+ stories. These

percentages are lower than the 18% spontaneously given by

subjects in Bartsch and Wellman (1989). The difference

between studies is likely due to the absence of anomalous

belief stories in this study. Recall from Study 1 that

anomalous belief stories require children to explain an

action that is contradictory to an actor's apparent desire.

A majority of belief terms in Bartsch and Wellman (1989)

came in response to such stories.

A main effect of presence, F(1,34) = 164.57, p < .001,

indicates that children used target-directed mental

explanations selectively, being influenced by whether a

situational feature was present. Overall, subjects were

much more likely to use target-directed mental explanations

in the four stories marked by the presence of a situational

feature (PA+ and SI+) (M = 3.64) than in the four

contrasting stories (PA- and.SI-) (M = .97). The effect of

presence was significant regardless of whether self-

initiated movement or perceptual access was the variable

being manipulated: Target-directed mental explanations were

much more common for SI+ stories (M = 1.92) than for SI-

stories (M = .53), F(1,35) = 130.99, p < .001, and,

similarly, for PA+ stories (M = 1.72) compared to PA-

stories (M = .44), F(1,35) = 101.04, p < .001.

In the SI- stories, children often appeared to adopt

what might be characterized as a Physical Stance, explaining

the action in terms of the physical contact initiated in

Picture 2 (see Table 8 for examples of such explanations).

The percentage of times in the SI- stories that an action

was explained in terms of the actor being "pushed" or

propelled in some way was 61% for preschoolers and 61% for

first graders. Such explanations were common even though

subjects could have explained action in the SI- stories by

appealing to the mental states of the actor. For example, a

few children explained the action as an attempt to avoid

further contact with the external causal agent (e.g., the

act of moving toward a playground is occurring so that the

causal agent on the bike "won't do it again").

In addition to the main effect of presence, there was a

significant interaction between age and presence, F (1,34) =

7.14, p < .025. The interaction reveals that preschoolers

were significantly more likely to use target-directed mental

explanations when the features were absent (PA- and SI-

stories) (M = 1.33) than first graders (M = .61), F(1,34) =

4.17, p < .05. In contrast, first graders were

significantly more likely to use them (M = 3.83) than

preschoolers (M = 3.44) when the key feature was present

(PA+ and SI+), F(1,34) = 5.17, p < .05.

Examining the interaction of age and presence more

closely shows that the age differences in the frequency of

target-directed mental explanations primarily occurred in

the PA stories. An age by presence interaction was not

significant in the SI stories, F(1,34) = 1.94, p > .10 but

was significant in the PA stories, F(1,34) = 9.36, p < .01.

The age difference in the number of target-directed mental

explanations in the PA- stories (M = .67 for preschoolers

and M = .22 for first graders) was significant, F(1,34) =

5.44, p < .05 (see Table 9). The reverse occurred when the

crucial feature was present (PA+) as first graders were more

likely to use target-directed mental explanations (M = 1.89)

than preschoolers (M = 1.56), F(1,34) = 4.14, p = .05 (see

Table 9). These results suggest that even though children

at both ages correctly discriminated between PA+ and PA-

stories, first graders did so more often than preschoolers.

They were more likely to use target-directed mental

explanations when perceptual access was present and less

likely to use them when perceptual access was absent.

Unlike the differences in performance in the PA stories,

children were equally successful in the SI stories. That is,

Table 9
Mean Number of Target-Directed Mental Explanations in
Experiment 2



















Note. For both age groups, the differences between the SI+
and SI- stories, and the differences between the PA+ and PA-
stories, are significant, p < .01. Scores ranged from 0-2.

children at both ages were equally likely to use target-

directed mental explanations in the SI+ stories and equally

likely to use other explanations in the SI- stories.

Preschoolers' difficulties in the PA stories, relative

to first graders, are further illustrated by examining the

varying number of times they explicitly referred to the

actor's lack of perceptual access in the PA- stories. In

these stories, 44% of first graders' explanations explicitly

mentioned the actor's lack of perceptual access (e.g.,

"Because he doesn't see it [the bracelet] and he's taking

the garbage out"; "Because she didn't see him [the cat]").

In contrast, preschoolers explicitly noted the lack of

access only 22% of the time.

Overall, 36% of preschoolers' explanations in the PA

stories were probed and 42% of first graders' explanations

were probed. Accurate judgments of the actors' knowledge,

on the basis of perceptual access, occurred 69% of the time

for preschoolers and 93% of the time for first graders. For

preschoolers 44% of the probes were in the PA+ stories, as

were 41% of the probes for first graders. The remainder

(56% and 59% respectively) were in the PA- stories.

A final level of comparison involves assessing

differences in explanations between each pair of stories

(eight pairs altogether) (see Table 6). First graders

discriminated between the stories in nearly every pair (the

exception being the playground story), offering more target-

directed mental explanations when the situational feature

was present than for the matching story in which the feature

was absent, all ps < .01 (Bonferroni-corrected Fisher test).

For preschoolers, significant differences in explanations-

emerged in three stories (the Pool, Couch, and Bracelet

stories), p < .01 (Bonferroni-corrected Fisher test).


The primary goal of Study 2 was to investigate the

circumstances influencing when and how children

mentalistically explain action. To this end, children's

spontaneous explanations of action were examined to see if

they differed as a function of the presence or absence of

(a) perceptual access to a key event and (b) self-initiated

movement. Results clearly demonstrated a sensitivity among

preschoolers and first graders to these situational

features, since target-directed mental explanations were

given significantly more often in their presence than in

their absence. A number of points follow from, and amplify,

this finding.

First, Study 2 extends Bartsch and Wellman's (1989)

examination of children's spontaneous explanations of action

in two ways. It replicates, under a set of stories and

circumstances different from Bartsch and Wellman (1989), the

assertion that young children's explanations of action often

spontaneously refer to the mental states of the actor. This

extension is important because, to a large degree, studies

of children's theory of mind have been limited to procedures

in which children predict (rather than explain) action on

the basis of another's belief or desire. Few studies

address young children's spontaneous explanations of action

even though the richness and importance of a theory of mind

often comes from its usefulness as a tool to construe and

explain particular actions performed by oneself or another.

Although other rudimentary systems of explanation exist for

children, such as the Physical and Design Stances, the

results of this study illustrate that the explanatory system

of choice, spontaneously invoked to explain action in the

SI+ and PA+ stories, was mentalistic -in nature.

In a complementary sense, Study 2 also illustrates that

young children predictably refrain from imputing target-

directed mental explanations under specific sets of

circumstances (i.e., when either perceptual access to a key

event or self-initiated movement is lacking). Such evidence

makes clear that the theory of mind of preschoolers and

first graders is not overextended and uniformly applied to

all action. Instead, young children were sensitive to which

actions were explicable via specific concepts within their

theory of mind and which actions were inexplicable given the

purview of their theory of mind. They did so by accounting

for how the action was occurring and inferring its cause on

that basis. Consistent with Study 1, such considerations

were more important than who was doing the action, as

explanations varied even though the biological category of

the actors remained constant.

Preschoolers properly adjusted their explanations of

action just as often as first graders when the feature of

self-initiated movement was manipulated, but did so less

often when perceptual access was varied. It is possible

that preschoolers were simply less sensitive than first

graders to whether perceptual access was present and,

therefore, less accurate in their judgments of knowledge.

For example, in this study preschoolers accurately

attributed knowledge, when probed, 69% of the time compared

to 93% for first graders. If they were not fully attuned to

the perceptual activity of the actor, its occurrence or

absence may have been overlooked and quickly forgotten,

resulting in target-directed.mental explanations when the

actor was unaware of the target (and vice versa).

More generally, the present methodology had a greater

likelihood of producing false negatives in children's

performance rather than false positives. For example,

reliance on open-ended explanations instead of providing a

forced-choice may have resulted in underestimating the

abilities of some children. In addition, the key features,

perceptual access and self-initiated movement, were embedded

within a story-like context and perhaps were not as salient

as they might have been otherwise. It is easy to imagine

ways to make them more salient; for example, by telling

subjects to pay particular attention to perceptual access or

self-initiated movement and then reminding them of their

presence or absence before asking for an explanation. Doing

so, however, increases the risk of overestimating the

importance children spontaneously attribute to the

situational features when inferring the cause of an action.

Future research could employ these methodological

changes and address the nascent abilities of even younger

children to parse the world according to the situational

features examined in Study 2. Methodological changes, such

as employing a forced-choice between explanations, would

serve to assess whether 3-year-olds, and perhaps even 2-

year-olds, construe the causes of action differently

depending upon whether or not self-initiated movement or

perceptual access is present. Given the results of Study 2,

in which preschoolers tended to do better in the SI stories

than in the PA stories, it might be expected that

sensitivity to self-initiated movement would emerge prior to

any sensitivity to the presence of perceptual access.

Evidence of the early ability to map mental concepts onto

the distinction between self-initiated and externally caused

movement would support the contention of Premack (1990) that

the foundational concept of agency emerges in infancy. He

argues that infants' ability to perceptually distinguish

between self-initiated and externally caused movement is the

basis upon which they interpret the intentionality of an

action. If these rudimentary interpretive capabilities

emerge during infancy, then when children's first

verbalizations of words like "wants" and "gonna" occur

(around age 2 and 3) (e.g., Hood & Bloom, 1979), they should


use these terms only when explaining self-initiated



The focus of this research was to explore a basic

question regarding children's thinking about the mind: When

do children impute mental states to others? This is an

important question because it speaks to the issues of how

theory-like children's conception of the mind is and, also,

the mechanisms by which such'a theory might develop. In the

following discussion, these topics are discussed in turn.

If, as often argued, children have some form of a

theory of mind (e.g., Gopnik, 1993), -it is reasonable to

suspect that principles exist which signify if (and how) an

action is explicable by that theory. Theories must specify

the situations in which their constructs can and cannot

meaningfully explain an occurrence. For example,

psychologists do not apply their cognitive theories to

astronomical phenomena because broad principles (e.g.,

cognitive theories are directed toward explaining human

behavior) delineate the spheres of inquiry for which the

theories are relevant. In the same way, a theory of mind-

must meaningfully delineate when and how phenomena are

explicable by its explanatory constructs (e.g., belief and


Consider, for example, a situation in which two movers

are carrying a table from one room to the next. What is it

about that action which would normally compel one to explain

it in terms of the beliefs and desires of the movers rather

than (or in addition to) the beliefs and desires of the

table, the biological makeup up the movers, the time of day,

wind velocity, etc.? If children's thinking about the mind

is theory-like, one would suspect that situational features

exist which consistently enable them to extract from such

novel, perceptually complex situations the information

needed to correctly infer that the action is occurring

because of the beliefs and desires of the movers.

To answer whether this is the case for young children,

Studies 1 and 2 examined whether self-initiated movement,

when variable, and perceptual access to a key event were

characteristics of actions that made them particularly

conducive to target-directed mentalistic explanations.

Altogether, children witnessed a number of actions in which

(a) the actions were committed by a wide range of actors,.

(b) plausible design explanations were pitted as

alternatives to mental explanations, and/or (c) the presence

of the features was varied. Given these manipulations, four

possible patterns of explanation could have emerged:

Mentalistic explanations could have (a) never occurred, (b)

always occurred, (c) occurred randomly, or (d) occurred

systematically. The results of Study 1 and Study 2 clearly

illustrate that the last pattern is actually the one that

characterizes the explanations of young children and adults.

Target-directed mental explanations were common in the

presence of the features (even when a plausible alternative

was present) and were also generalized, by children, across

different actor categories. In comparison, the removal of

self-initiated movement or perceptual access resulted in a

marked decrease in target-directed mental explanations.

These results suggest that children abstract features

of action when construing its cause and that doing so plays

a vital role in learning to generalize mentalistic

explanations to new situations. Indeed, without the ability

to rely on situational features in inferring the cause of an

action, children would have to learn, on a case by case

basis, why an action was occurring. Since mental states are

unobservable, such learning would likely require an overt

confession on the part of the actor signifying her beliefs

or desires. However, young children capably generalize

across situations, identifying defining features that

immediately categorize an occurrence as potentially

explicable or inexplicable by their theory of mind.

Not only do children accurately generalize across

situations (Study 2), but they also tend to inaccurately

generalize across biological categories of actors (Study 1).

This overgeneralization is consistent with Premack (1990),

who has suggested that self-initiated movement is not a

domain-specific cue of intentionality for infants. He

argues that the self-initiated motion of various objects,

varying from boxes to humans, is regarded by them as

intentional. In this study, children's overattribution of

beliefs and desires to artifacts and insects, compared with

adults, may be the result of a rigid adherence to the

importance of the situational features surrounding an

action. In fact, one feature of theories is that they can

lead to the misintepretation of information when a person is

"biased" by their precepts (Astington & Gopnik, 1991). It

may be that children's theory of mind biases them to note

particular features of action that, when present, override

perceptual/biological features of the actors. Thus,

although in their appearance and biological make-up humans,

ants, and computers have little in common, when their

actions were construed within the theoretical framework, a

specific commonality between them may have surfaced because

features of their actions satisfied children's theoretically

important preconditions. As a result, even though young

children are not always animistic (as noted in the

Introduction), they nevertheless tended to impute

mentalistic explanations to artifacts.

Although children imputed target-directed mental

explanations systematically across Studies 1 and 2, one

might argue that consistency alone ddes not mean that

children's thinking about the mind is "theory-like." In

fact, consistency may simply be the result of learning over

time to associate specific mental states with specific types

of action. For example, children may learn from their own

experiences that action which is variable, self-initiated,

and accompanied by perceptual access is often driven by

their beliefs and desires directed toward the action. They

may extrapolate mentalistic causes of action to others on

the basis of this association. Therefore, consistency in

their explanations may simply be a byproduct of

consistencies among their own phenomenological experiences.

Even so, it is important to dissociate the mechanisms

by which consistency emerges from the "theory-like" function

the situational features perform. To return to the example

of the "moving table," one might argue that children explain

the action by simulating their own experiential history of

carrying objects and extrapolating such experiences to the

movers (e.g., Harris, 1991). Even if this is true, at a

more basic level one must ask how children know that this

action is one that can be explained by simulation whereas

another action, perhaps the movement of leaves by the wind,

requires no simulation. Clearly, features such as self-

initiated movement and perceptual access narrow the field of

occurrences that can be explained by simulation, thereby

acting as theoretical principles in carving the world into

phenomena explicable and inexplicable via mental concepts.

Nevertheless, if these situational features act as

theoretical principles delineating phenomena into those

which are and are not explicable via beliefs and desires,

the question of how they come to play the role they do is

still open. Answering this question addresses the nature of

the mechanisms by which a theory of mind develops. As noted

earlier, it may be that through experience, children learn

that most times when their action is self-initiated,

variable, and accompanied by perceptual access, then it is

volitional. They may assume that the same experiences are

true for others under those same circumstances, including

the actions of insects and artifacts (children's inferences

about the phenomenological experiences of species other than

humans are interesting and, for the most part, also an open

questions). Another possible mechanism, however, is that

sensitivity to these situational features is hardwired, and

perhaps innate (see Premack, 1990), suggesting that

sensitivity to how an action occurs may be the basis for the

later emerging concepts of belief and desire. For example,

children may first come to divide the world into objects

which are either self-propelled or externally caused.

Concepts of volition and desire may then emerge as ways to

interpret this distinction. That is, opposite the

simulation view, the mental concepts may come after

sensitivity to the situational features. The question then

is whether (a) the concepts are derived strictly from

phenomenological experience and then, over time, gradually

become associated with variable self-initiated movement and

perceptual access or (b) distinguishing the situational

features comes first and gradually, children come to learn

that these distinctions among different actions can be

explained by the beliefs and desires of the actors. This

realization may emerge from learning that target-directed

mental states only covary with action characterized by the

situational features.

To disentangle the interrelation between

conceptualizing the mind and parsing the world according to

the situational features investigated in this study, one

must address how very young children distinguish and explain

action. Future research should examine the various features

of action that are salient to toddlers and used by them to

discriminate among occurrences. The consistency of the

discrimination across situations, the timing of the

emergence of sensitivity to different features, and the

salience of the various situational features are all issues

worthy of investigating. Following such investigations, it

would then be possible to assess whether distinctions among

actions are made independently of children's target-directed

mental explanations and whether such explanations emerge

prior to the perceptual distinctions of interest.


Astington, J.W., & Gopnik, A. (1991). Theoretical
explanations of children's understanding of the mind.
British Journal of Developmental Psychology, 9, 7-32.

Bartsch, K., & Wellman, H. (1989). Young children's
attribution of action to beliefs and desires. Child
Development, 60, 946-964.

Berzonsky, M.D., Miller, P.H., Woody-Ramsey, J., Harris,
Y. (1988). The relationship between children's
judgments of animacy and sentiency: Another look.
Journal of Genetic Psychology, 149(2), 223-238.

Bullock, M., & Gelman, R. (1979). Preschool children's
assumptions about cause and effect: Temporal
ordering. Child Development, 50, 89-96.

Carey, S. (1985). Conceptual change in childhood.
Cambridge, MA: MIT Press.

D'Andrade, R. (1987). A folk model of the mind. In D.
Holland & N. Quinn (Eds.), Cultural models in language
and thought (pp.112-150). New York: Cambridge
University Press.

Dennett, D.C. (1987). The intentional stance.
Cambridge, MA: MIT Press.

Dretske, F.I. (1988). Explaining behavior: reasons in a
world of causes. Cambridge, MA: MIT Press.

Flavell, J.H., Miller, P.H., & Miller, S.A. (1993)
Cognitive development (3rd ed.), Englewood Cliffs,
NJ.: Prentice-Hall.

Fodor, J. (1987). Psychosemantics: The problem of
meaning in the philosophy of the mind. Cambridge, MA:
MIT Press.

Gelman, R. (1990). First principles organize attention
to and learning about relevant data: Number and the
animate-inanimate distinction. Cognitive Science, 14,

Gelman, R., Spelke, E.S., & Meck, E. (1983). What
preschoolers know about animate and inanimate objects.
In D. Rogers & J.A. Sloboda (Eds.), The acquisition of
symbolic skills (pp.297-326). New York: Plenum Press.

Gelman, S.A., & Kremer, K.E. (1991). Understanding
natural cause: Children's explanations of how objects

and their properties originate. Child Development,
62, 396-414.

Gelman, S.A., & Wellman, H.M. (1991). Insides and
essences: Early understandings of the non-obvious.
Cognition, 38, 213-244.

Gopnik, A. (1993). How we know our minds: The illusion
of first-person knowledge of intentionality.
Behavioral and Brain Sciences, 16, 1-14.

Gottfried, G.M., & Gelman, S.A. (1992, April). Domain
differences in children's understanding of internal
cause. Presented at the Conference on Human
Development, Atlanta, GA.

Harris, P.L. (1991). The work of the imagination. In
In A. Whiten (Ed.), Natural theories of mind (pp.
283-304). Oxford: Basil Blackwell.

Heider, F., & Simmel, M. (1944). An experimental study
of apparent behavior. American Journal of Psychology,
57, 243-259.

Hood, L., & Bloom, L. (1979). What, when, and how about
why: A longitudinal study of early expressions of
causality. Monographs of the Society for Research in
Child Development, 44 (6, Serial No. 181).

Inagaki, K., & Sugiyama, K. (1988). Attributing human
characteristics: Developmental changes in over- and
underattribution. Cognitive Development, 3, 55-70.

Johnson, C.N. & Wellman, H.M. (1982). Children's
developing conceptions of the mind and brain. Child
Development, 53, 222-234.

Keil, F.C. (1989). Concepts, kinds, and cognitive
development. Cambridge, MA: MIT Press.

Keil, F.C. (1992). The origins of an autonomous biology.
In M.R. Gunnar & M. Maratsos (Eds.), Minnesota
symposia on child psychology: Vol.25. Modularity
and constraints in language and cognition (pp. 103-
138). Hillsdale, NJ: Erlbaum.

Kun, A. (1978). Evidence for preschooler's understanding
of causal direction in extended causal sequences.
Child Development, 49, 218-222.

Leslie, A.M. (1984). Spatiotemporal continuity and the
perception of causality in infants. Perception, 13,

Lillard, A.S., & Flavell, J.H. (1990). Young children's
preference for mental state versus behavioral
descriptions of human action. Child Development, 61,

Mandler, J.M. (in press). How to build a baby: II.
Conceptual primitives. Psychological Review.

Michotte, A. (1963). The perception of causality.
London: Metheun.

Premack, D. (1990). The infant's theory of self-
propelled objects. Cognition, 36, 1-16.

Smith, M.C. (1978). Cognizing the behavior stream: the
recognition of intentional action. Child Development,
49, 736-743.

Springer, K., & Keil, F.C. (1991). Early differentiation
of causal mechanisms appropriate to biological and
nonbiological kinds. Child Development, 62, 767-781.

Wellman, H.M. (1990). The child's theory of mind.
Cambridge, MA: MIT Press.

Wellman, H.M., & Gelman, S.A. (1992). Cognitive
development: Foundational theories of core domains.
In M.R. Rosenzweig & L.W. Porter (Eds.), Annual Review
of Psychology (Vol. 43) (pp. 337-375). Palo Alto, CA:
Annual Reviews Inc.


Derek Montgomery was born July 2, 1965, in Toronto,

Ohio. He graduated summa cum laude with a B.A. in

psychology from Ohio State University in 1987. In 1991

he received his M.S. in developmental psychology from the

University of Florida. His research interests center

around children's metacognitive development and their

understanding of invisible mechanisms of causation.

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.

Scott A. Miller, Chair
Professor of Psychology

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.

ohn Biro
Professor of Philosophy

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.

M. r F rrar
Ass ciate Professor of

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.

Pra Fischler
Professor of Psychology

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.

Patricia Miller
Professor of Psychology

This dissertation was submitted to the Graduate Faculty
of the Department of Psychology in the College of Liberal
Arts and Sciences and to the Graduate School and was
accepted as partial fulfillment of the requirements for the
degree Doctor of Philosophy

August 1993

Dean. Graduate School

3 1262 08553 8I 883I
3 1262 08553 8683