Mental State Talk: A Developmental Perspective

A thesir submittled in conformity with the requirements for the degree of Master of Arts

Department d Human Development and Applied Psychology Ontario Institute for Studies in Education of the

University of Toronto

O Copyright by Shed L. fmll(2000)

53 uisitions and Acquisitions et 6' bgnphie Sewices ~nriœs b~Wiogmphques

The author has granted a non- exclusive licence dowing the National Library of Canada to reproduce, loan, distniute or seii copies of this thesis in microform, paper or electronic formats.

The authof retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts fiom it may be printed or othawise reproduced without the author's permission.

L'auteur a accordé une licence non exclusive permettant à la Bibliothèsue nationale du Canada de reproduire, prêter, distn'buer ou vendre des copies de cette thèse sous la forme de microfiche/film, de reproduction sut papier ou sur fonnat électronique.

L'auteur conserve la proprittd du droit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.

Mmhl Statib Tdk: A DmlopmenW Pmpecüve

Master of Arts, 2000

Shed Lynn Twell

Deparamit of Humn OIvrlopnwnt and Applied Psychology

Ontrrlo Instîtuto for Studles in Education of the

University d Toronto

ABSTRACT

We investigated the developmental traiedories of mental state talk (cognitive, desire

and feeling) in children, and influences on mental Mate talk Thirty-seven Caucasian

families, each with two parents and two children, pertidpated in this study over a two-

year period. Oeta wre collecteci during six 90niinute sessions in the participants'

homes. Each speaker tum was coded for the presenœ of mental state ternis and the

referent (e.g., seif, other).

Results indicated that cognitive state talk inaeases in children and parents as the

children get older. Mothers and older girls talked more than fathers and older boys

about the mental states of othrs. Finally, children with an older sibling, when

compared to those withwt older siblings are advantaged in their cognitive state talk at

age 4 after antrolling br MLU. Children with an older sibling are rot hawever

advantaged in t h i r tdk about ôesires and feelings.

First. I would like to thank Dr. knnifer Jenkins for teaching me so vey much about

the research process. This was a challenging project, requiring considerable teamwrk,

support and encouragement until the vefy end. Dr. Jenkins' insight and assistance will

never be forgotten. I also want to thank my second reaâer, Dr. Janet Astington, whose

research acumen was greetly appreciated, as were her words of encouragement,

particulariy as the project neared completion.

Secondly, I must thank Dr. Susan Lollis and Dr. Hildy Ross for allowing me access

to this substanüal database. In addition, a sinœre thank you to Yuiko Takei at Guelph

University, for the energy and devotion she demonstrated in helping my devise the

coding scheme and in coding a phenamenal amount of data.

Third, I must thank my farnily for suppofting and humoring me as I completed draft

after draft Mer draî?. This was a daunting process, which could not have bemn

completed without the endless love, patience and understanding from my husband,

Jerome, my father Charles, Ravel, and my mother-in-law Angela.

1 have reœived support from friends too numerous to mention them ail, but special

thanks to Tania, Narslk, Maggie and Carol, and al1 of my Riends a t b Hospital for Si&

Childmn. Much gratitude is extended toward Isabel and Fiona for ywr humour and

editonal eflorts.

Last, but œrteinly not Ieast, e simm d&t of gratitude to my dear Wend Rosemary

Condillac and hdr husband Dave. Wrthout their boundleu fiiendship. emrgy and

cornmitment, this thesis w l d never have barn tompleteâ.

iii

List of Tables

List d Figums

Chiptrr 1 : Intmduction

Develapment of childrm's mental state talk

Development of parents' mental state tak

Gender diferences in mental state talk

Diffemces between boys 8 girls

Oiffwmces be-n mothen 8 fathem

Referent of mental state talk

Effect of Siblings on mental state talk

A i m of Prisent Study

Chapter 2: Method

Participants

Procedure

* Behwiwral observstions

Codim for mental strtb U k

C d i ~ for refmnt

ii

iii

vi

vii

8

8

15

19

19

22

23

26

31

34

34

35

35

37

39

42

42

42

Chaptei 3: Resultr

Conversational turns

Mean length of utterance

Relationship between MLU & mental state talk

Question #1

Youngei children

Older children

Parents

Cognitive, desire & feeling talk as a

percentage of mental state talk

Question #2

Parents

Younger children

Older children

Question #3

Chaptei 4: Dlrcu8sion

Relationship between MLU 8 mental state talk

Development of children's mental state talk

Development of parent's mental state talk

Gender differences in mental state talk

Referent of mental state talk

Effect of siblings on mental state talk

Reféwnces

List of Tables

Table 1. Means & standard deviations for changes in

conversation tums over time.

Table 2. Means & standard deviations for changes in

mean length of utterance over time.

Table 3. Correlations between MLU & younger children's speech. 48

Table 4. Correlations between MLU & older children's speech 48

Table 5. Means & standard error for changes

in children's mental state talk over time.

Table 6. Means & standard error for changes in parents'

mental state talk over tirne.

Table 7. Means & standard mors for gender

differences in talk about the 'other'.

Table 8. Means & standard error for the effect

of an older sibling on mental state talk.

List of Figures

Figure 1. Changes in younger siblings' mental state talk

over tirne.

Figure 2. Changes in older siblings' mental state talk

over time.

Figure 3. Changes in parents' mental state talk

over time.

Figure 4. Gender differences in parents' talk about 'other'.

Figure 5. Gender differences in oMer siblings' talk about 'other'. 60

Figure 6. Effect of an older sibling on mental state talk. 62

vii

1: INTROOUmolq

Developmental changes in children's talk about the mind have becorne a topic of

interest for investigetom over ment pan. To understand why people do what they do

requires that w& understand the events of the mind which pmvob bhaviow. As

adults, o u kliefr, desires, motivations and feelings &ive ouf actions. An examination

of children's production of mental state ternir aliows us to explore h w children's

unâerstanding and talk of mental events devdops over time. Existing studies point to

important trends in children's changing abilities to discuss various types of mental

states, however, there io still much to discover. For example, little is known about

developmental changes after four yeen of age. Furthemiore, the literatun provides

only a glimpse into the changes in parents' mental state tdk as thdr children grow and

develop. The pesent shrdy s d t s to uncover new information, Mich will extend whet is

already krown. We examine not only children's devdopmental changes in mental state

talk, but also that of their parents, as wdl as factors that may be associated with the

production of mental state tenns.

The Devdopment of Children's Mental State Talk

Previous research has provided us with trends in the devdopmental changes in

belief, desire and M i n g stata talk Thme studies are delcri*bd below in some deteil.

As Astington and Pelletier (1996) pointad out. talk about mental $tatas cen k divided

into many categories, such as perceptiont cognition, volition, morality, sbf@ct and

physiology. Howevert talk about Wids (e.g., think, know), deaires (8.g.. want, hop),

and fedings (e.g., Sad, happy) are mpwted to be cornmon components of al1 linguistic

taxonomies and are therefom, the foais of the present study. It is important to note

that there is little consistency in the way that the fbquency of mantal state tems is

presented in previous researh. For example, some studies calculate the fiequency os

a proportion of al1 utteranœs, while othen calculate mental state terms as a proportion

of only those utterances containing mental state tenns. The method employed by each

remarcher wiil be describeci.

Talk about Beliefs

Talk about beliefs or cognitions (these tems will bo uwd interchangeably)

generally does not appear in children's vocakrlary until the middle of the third year

(Furrow, Moore, Oavidge & Patriquin, 1992; Hughes 8 Dunn, 1998; Shatz, Wdlman &

Silber, 1983). For exampk, Shalr et al. fowsed on the development of cognitive

speech in one child over a W-year period. Findings revealed that cognitive state talk

first appeared betMHM 2 yean, 4 months and 2 yean, 11 months. This nsult was

corroborateci when the same authors studied 30 2-year old children over a sixinonth

period, and found that cognitive tems did not appar until the rniddle of the third year.

Once talk about cognitive States tuas been estaMished, it continues to inuease in

Wequency (Furraw et d., 1992; Hughes & Dunn, 1998). Fumw et al.. for example.

obscwved ninetmm mottiw4ild d y d s whm the childm wem between 2 and 3 years

old. Results indicated a signifiant imieam in the production d cognitive t e m

betwwn 2 and 3 years of age, as a perc8ntqp of il1 utterances producd at each time

pefiod.

Hughes and Dunn (1998) found that the fiequency of cognitive ternis inweased in a

sample of 50 dder children who were observd from the time they unre 3 years, 1 1

months to 5 years d age. Results indicateâ mat children used cognitive state ternis et

the initial observation petiod, and aiet the frequency inwased until the final

observation period. The authors pierented the quantity of mental state talk as the

number of ternis pmduced pw hour of observation. These studies provided important

information about the developmental changes in cognitive talk from its onset until age

five, but what do we kmw about children's talk about other mental states and their

respective developmental changer?

Talk about Beliefs and Desires

Desin tenns have been found in children's speech prior to cognitive tenns and are

more frequent until the third year. In an investigation of cognitive and desite state talk

in fourteen mothechild dyads, Moore, F U ~ W , Chiasson and Patriquin (1 994) reported

that desire temis mwo the most cornmon type of mental state talk until aie third year.

Between the third and fouai year, desin t m s leveld off while cognitive tems

increased and were more muen t thur desire tems. Moore et al. pfesented the

ftequency of mental state talk as a propoflion of participants' total number of

utterances.

Other studies that have examineci concomitant developmental changes in belief and

dasire &te talk have found that amnd the age of thm, childisn inmase their

production of bdief terms, hawsvw, &sire t m s rmain the most frequent type of

menW state talk (Bartsch & Wellman, 1995). Barbch and Wellman studieâ cognitive

I l

and desire talk in ten children from the time they were 1 W to 6 years old. When the

frequency of belief and desin tems were compared to each other, it was repted that

the onset of desire talk preceded belief talk. and was Vie most cornmon type of talk

between 18 and 72 months of age. This pattern emerged whether belief and desire

tenns were expressed as a percentage of al1 utterances wntaining refermces to

beiiefs and desiros, or as a percentage of the child's total utterances. While the

developmental changes in this study appear to contradict those of Moore et al. (1 994),

it is important to bear in mind that Bartsch and Wellman reported only on genuine

references to psychological states, whereas Moore and his colleagues did not

distinguish between various fundionr of mental state tems. The differences in the

inclusion criteria may have affeded the ftequency of mental state terrns for example, by

decreadng the freguency of cognitive tems in Bartsdr and Welhan's study. Oesires

and cognitions are two distinct types of mental state and, taken together, these results

indicate that diildren use dwire talk More they talk of beliefs. Moreover, the

production of cognitive tems increases from the age of three, both as a percentage of

al1 utterances and os a percentage of all mental state talk

falk about Feelings

Feeling talk ha8 been found to appear as earîy as 18 monais of age and to inuease

in frequency from that time (Dunn, Brethcwton & Munn, i987). This was examined in

two longitudinal studies. The fint study induded 43 motherchild dyads, when the

second-born childfen wem 18 and 24 months of aga The second study consisted of

16 mother4ild dyaâs whm the fimt-boln childm wmre 25 and 32 months of age.

00th studies yielded similar nrultr, with childm speaking more about feelings as thay

got older. The results Mered similar developmental trends whaher the frequency of

f-ling talk war presented as the number of uttefances or the numbr of conversation

turns that wntained a feeling bm. However, Dunn et al. only examineâ feeling taik in

children until the age of three and w could not provide details of developmental

changes after that time. Additionally, because only feeling terms were attended to, no

information was provided about changes in feeling talk in combination with changes in

cognitive and desire state talk.

Talk about Beliefs, Desires and Fmlings

In studies thet have examined concomitant changes in cognitive, desire and feeling

talk, results have indiatecl a prepandeance of desire terms, with both cognitive and

feeling ternis king less frequent, but inueasing with time, when each category of

speech was repoed relative to the remaining categories (Bretherton 8 Beeghly, 1982).

Bretherton and Beeghly obtained their results by training 30 mothen to record their

children's use of six cetegories of intemal state language in the home. The study was

longitudinal, beginning Men children wre 10 months of age and ending at 28 months

of age. Bretherton and Be8ghly assessed children's language production dwing four

situations (e.g., Srninute snack and S-minute play sessions) and found that these

scores wwe highly comlated udth rnotfmrs' reports of their children's mental state talk

Brown and Oum (1991) extended the work of Bmtnerton and Beeghly by studying

slightly older childm aged 24-36 monthr. The auVwm nporled that cognitive tc~mr

fint appmred at 30 monais of age, while feeling twmr w m pmmnt from 24 monüis.

60th types of mental state tak i m a d in hquency from the time they appeamd.

The results were expressed as mental state ternis per 100 conversation tums and

supported the findings of studies that have exarnined changes in either cognitive or

feeling talk (Bartsch & Wellman, 1995; Dunn et al., 1 987; Fumw et al., 1992; Hughes

8 Dunn, 1998; Moore et al., 1994). This study provided importpnt new information

about the interplay between various types of mental state talk.

Summary

To summarize, references to cognitive states have b e n rare until the third year end

have increased from that time (Bartsch 8 Wellman, 1995; Fumw et al., 1992; Hughes

& Dunn, 1998; Moore et al., 1994). Talk about desires appeared earlier than that of

cognitions and continued to be the most m m o n type of mental state talk until36

months of age (Brown d Dunn. 1991 ; Moore et al., 1994) and beyond (Battsch 8

Wellman, 1995). Finally, talk about feelings began as young as 18 months and

continued to increase in frequency; howcrver, it remained less cornmon than talk about

desires (Brown & Dunn, 1991). The aforemsntioned developmental changes have

been reliably reported despite mtioddogical dinarences.

Remaining Questions

While the previously r e v i d studies addressed important contbmporaneous

changes in cognitive, desin and fmling talk, them are several developmcmtd

questions that have becm kft unaôâfesseâ. First, we kiow little about the devetlopmcmt

of mental -ta talk kyond 36 months d Many 8tudies began masuring mental

state talk whn childm wmm a8 y a n g a8 two yeam of aga Wth only hno exceptions

ho~ever, (Bartsch & Wellman, 1995; Hughes & Dunn, 1 W8), developmental changes

have not been examined in children beyond four yean, of age.

Some researchers have observecl mental state talk as it occurs in a school context

(Hughes 8 Dunn, 1998), while othen have cfeated contrived situations in drildren's

homes (Furraw et al., 1992; Moore et al., 1994). Mentil state discourse in contrived

environments may be falsely elevated or reduced, as children are askd to participate

in conversations that rnay or may not be relevant to them, and with people they may not

know. Some studies have, howver, obsewed children in their own homes (Bartcich &

Wellman, 1 995; Shatz a al., 1 983). The home environment is where children can

interect with familiar people, in a familier and cornfortable place. Children rnay be more

likely to talk about things that interest them and engage in play situations that bettw

demonstrate their understanding of the mind when in their own homes.

Fathers have been consistently ignored both as a -pient of talk and as a generator

of talk in the study of childien's talk about the rnind. Children's mental state ta#< has

been examimd, almort exclusively, as it occurs with mothem (Furrow et al., 1992) or

peen (Hughes & Dunn, 1998). These studies have provided convecging evidence of

developmntal changes in mental state talk T h y Iaava unanswered, hmmver, the

infiuence of fathen.

The prasent study was designeâ to addms the pmvious methodological gaps in

several ways: 1) by studying the concomitant changes in cognitive, desire and fmling

talk, 2) by utiliting data cdl8ded in the hotne 3) by oûm ing conversations ocairring

between the target child, their sibling, mother and father, 4) by including children who

are 2 and 4 years old a time 1, and following them up Mer 2 years.

fha ûevelopnnnt of Purnts' Mental Shte Trlk

To more fully comprehend aie changes in children's understanding of the mind,

attention must be paid to the mental state terms that chiidren are 8xp0sed to in their

daily environments. Existing research has found that maternai talk about mental States

benefits children in a variety of ways. Specifically, positive correlations have been

found bebmen matemal mental state talk, children's use of mental state tems and

diildren's performance on measures of language production and comprehension

(Beeghly, Bretherton & Mervis, 1968). Additionally, Moore et al. (1 994) showed that

mothen' production of cognitive tems when children were two years old conelated with

children's language comprehension scores at age four. Given the apparent influence

that matemal mental state taik has on children's understanding of the mind, it is

important to map such developmental changes.

Talk about Beliefs

The resub of eadiw studies indicate that mothers tend to produce mors cognitive

tennr, as a proportion of al1 utteranœs, as theit childfen appmch 3 years of age.

Funow et al. (1992) reporteci thi8 trend mer studying patterns of mt8mBl cognitive

state talk in nineteen rnother&ild dyads when the childm ware two and t h m years of

age.

Talk about Oesires

When desire and cognitive state talk were exarnined simultaneously, desire talk was

reported to decrease and cognitive to increase, as children approached three years of

age. The changes in the frequency of mental state ternis were presented as a

percentage of total utterances. Moore et al. (1 994) explored the cognitive and desire

ternis produced by 14 mothen when their children were 2, 3, and 4 years old. It rnay

be that mothers increased the frequency and therefore, the saliency, of cognitive ternis,

just prior to children's' increased production of such terms.

The issue of directionality is both important and cornplex, and in its entirety, is

beyond the scope of this paper. However, one must consider that children may be

better able to understand cognitive terms as they gel older. As a result, children may

use more cognitive terms, which mothers may be mirroring in their own speech, or the

opposite may be true. Regardless of the position one favors, it does look as though

mothen exect an irnporlant influence on their children's developing knowledge about

the mind. The influence of mothen' speech on their children's speech must be

considered in light of the fact that mothers' production of cognitive terms, when children

were two yean old conelated with children's production of cognitive ternis at age three

(Moore et al., 1 Q94),

Talk about Feelings

Changes in matemal production of feeling terms were examined in one study, which

found that mothers increased their use of feeling words over t h e (Dunn et al., 1987).

Dunn et al. obsewed the naturally occumng speech of 43 mothen and their second-

born children Wen their childien wen 18 and 24 months of age. Obsewations were

also conduded with 16 mothen and their first-bom children, when the children were 25

and 32 months of age. As children approached 32 months of age, mothem increased

their production of feeling state talk, representeâ by the number of utterances and the

number of conversation tums containing a rderence to feeling states.

Talk about Beliefs, Desires and Feelings

Investigators have found that maternel references ta cognitive states increased,

while the patterns for desire and feeling talk were leu clear, when cognitive, desire and

féeling deto talk w i e observeci simultaneously. Bemghly et al. (1986) researched the

perception, physiologie, affect* desire. cognition and moral judgement terms used in

conversations behusen 26 motherchild dyads. Observations were mduded Men the

children wem 13,20 and 28 months of age. Results suggesteâ that cognitive state talk

increased, as a proportion of all mental dete talk, as children neared 28 months of age.

The patterns for desire and fwling $tate talk m e less obviour end did not change

significantly over time.

Brown and Dumi (1 991) obsenred six mother-child dyads in order to identify

changes in matemal ref-8 to cognitive, desire and feeling states. Obsewations

were conducted whm childm wre 24-26,2&50 snd 33-36 months of age. As others

have reported (Beeghly et al., 1986; Fuimw et al., 1992), moünin incrc~sed their

production of talk about cognitions u childm appmached 36 months of age. In

addition, the fnquency of talk about desires decreasd Mer children raich84 3û

months of age, 8nâ this wpported the findings of Moore et (il. (1994). who found desire

terms to be mort Vequent when children wenr two years old. It is important to note that

these results were consistent with pnviously established patterns, despite the fact that

Brown and Dunn presented mental state ërms p r 100 conversation turns, a method

dissimilar to other investigators. Finally, there was a non-significant increase in feeling

state talk as children approached 36 months of age. These tesults extended the

findings of Beeghly et al. (1 986) by providing a glirnpse into the develapmental

changes that occuned in maternal mental state talk as children went beyond 28 months

of age.

Summary

ln surnmary, metemal references to cognitive States have been fowid to increase as

children approach three years of age, while references to desires decreased (Furrow et

al., 1992; Moore et al., 1994). The pattern of change for feeling state talk is less well

establisheâ (Dunn et al., 1987). These patterns have emerged despite the fact that

some researchets have presented changes as a proportion of al1 uttefances (Brown &

Dunn, 1991 ; Funrnn et al., 1 892; Moore et al., 1 994) h i l e othen have expressed

changes relative to al1 mental state talk (Beeghly et al., 1986).

Remaining Questions

fhere are a small number of studies that have examined developmental changes in

maternal mental state talk. Many questions remain to k annnrered and we shall

addreu two of them in the pmmt study. Fint, axisting shrdiw have not addmsseâ

maternal changes in mental da8 talk aftw childmn am fow yeam of age. The pmsent

study indudes childmn from 8ge two to six in the hopes of extending what we know

about the developmental changes in mothen' mental state trlk. Second. we have

included fathers, as none of the previously desdbd studies have done so, leaving

unexplored fathers' influence on childm's developing understanding of the mind.

Gmder Dïffemnces in Talk about the Mind

In a review of emotional development, Brody (1 985) describecl emotionat

development according to socialcognitive theory, in which it is suggesteâ that one's

social environment can exert a significant infiuence over emotional development. In

adhmrence with this theory, Brody suggested that girls and boys would speak differently

about feelings and that this dflerence wwld stem from mothem and fathm gender-

specific socialization of their children's emotional expression. The study of the

relationship between gender and the development of talk about mental states has

produœd tonflicting resuk.

Gender Differences betW88n Girls and Boys

Some studbs t h 1 have exwnind gmdef d i f f e m s in how boys and girls talk

about the mind have found no evidence of gender 8ft8ds. Kuebli and Fivush (1 992)

involved both mothen and fathers, as they obsemeâ c o n v ~ i o n r about past events

with each parent and their chiid. The study involveâ 24 children and their parents when

the children wem 40 months of aga With regard to childrsn's talk about emotions, no

gender differenas wers fouid; howevw, the authors nportrd th& th childm

producd fw M i n g wwds overall, which may have pmvented wch a finding.

Cewantes and Callanan (1998) examinai feeling ternis in 84 children at age 2,3

and 4 in conversatiom with thrir mothen. Girls were found to speak more about

emotions than boys did. but only at age 2. The authors suggested that the la& of

genâer difference may have resulted from the fact that girls started off using higher

rates of ernotion talk than boys and did not show an increase in the frequency of feeling

words over tirne. Furthemore, Cervantes and Callanan postulated thet the gender

differences in 'starting points' of emotion words might have been due to differences in

language ability, and no sudi measures were included in the study.

Other studies of mental state talk have found that girls used mon emotion words

thon boys did. Kuebli, Butler and Fivush (1 995) investigated gender differences in 18

children with their mothers, speaking of past events. Children wen, 3 years, 4 months,

4 years, 10 rnmths and 5 yeam, 10 months at the time of the observations. Results

indicated that girls used mon motion words and a greater variety of emotion words

than boys did, and that this differenœ was most pronounced at the third observation

time.

Dunn et al. (1 987) eiamined feeling state talk, naturalistically in children a g a 18,

24,25 and 32 months. Gender difrences wem not examineâ when children w e 32

months of a- as only 16 diildren in that age group participated. At 24 months of age,

however, girls rebmd more Men to fWing states than did boys.

Thwe is rame evidenco that giils umd more cognitive temis t(lM boys did. Hughes

and Dunn (1998) invmtigated g e W diifimœs in cognitive state talk in 50 children

as they interacted with a peer. Obse~ations were conducted when the target chiid was

3 Yeats, i l rnonths, 4 yeafs, 6 months, and 5 yeam of aga Results indicated that girls

produced more re&wences to cognitive states than did boys and u s a a greater variety

of cognitive state tems. Gender difterences in cognitive state talk penisted even aftw

vocabulary scores (British Pidure Vocabulary Scale) wen covaried out of mental state

talk. This pattern of iesults is sirnilar to those reported for gender differences in feeling

state talk (Dunn et al., 1987; Kuebli et al., 1995).

Studies have revealed that mothers talk to daughters more than to sons about

motions and this may contribute to gender differences in how childien talk about

mental states. For exemple, Kuebli et al. (1 994) observed the way in which mothen

spoke to their children, who wem between 3 yean, 4 months and 5 yeen, 10 months.

Results indicated that mothen used more feeling tems with daughters than with sons

and that rnothers who used more emotion words in the first phase had children who

used more ernotion words during the final phase. Cetvantes and Callanan (1 998)

confirmed the findings of Kuebli et al., as they reported mothers spoke more to

daughters Vian to sons about feelings when children were 2, 3, and 4 years old.

Surnmary

Analyses of gender differenws in children's hlk about the mind have produœd

unclear resuîts, with some mporting gender differencer in motion talk (Cervantcis &

Callanan, 1998; Dunn et al., 1987; Kuebli et al., 1995) and cognitive talk (Hughes &

Dunn, 1998), while othem have not (KUebli & F i W , 1992).

Remaining Questions

The mixed pattm of gender differenœs in diildren's emotion talk may be a result of

the low frequency of femling ternis used by childmn in some studks (Kuebli CL Fivush,

1992), or the context in which children'i speech w s emmineâ. For example, Kuebli

and Fivush found gender dmrences Men asking Children to speak of past events.

With the exception of Kuebli and Fivush, studks of gender diflerences in mental state

talk have not included fathers. The present study hopes to address these

methodological differences by: 1 ) examining gender differences in natwally m m n g

speech, 2) including siblings, rnothers and fathers 3) obsming children over an

extemled p e d of time to offset the conœms exprersed by pnwious investigaton, that

the lack of gender differences in motion talk may be a rewlt of Vie low frequency of

such terms.

Gender Differences behween Mothers and Fathen

It has been repoRed that mothen and fathen do not talk difrerently about emotions.

Kuebli and Fivush (1992) reported on gander differmces between mothem and fathers

involved in canvemtiort about past events with their children, a g a 40 months.

Results indimted no difference in the numkr of motion mxds u s d by mothem and

fathem. These msub a n confuring, givem aie prcwrknt viw that wmen are more

emotionally expressive than mm am (Brody, 4985). Howmw, the s-c context

used in this study miy have contributed to the la& ofgendar diffemnœ in emotion talk

Summary

There is very little research that has investigated gender differences in parents' talk

about mental states. Kuebli and F ivush (1 992) conducted one such study and reported

that mothen and fathen did not diffw in tkir talk about emotions when interacting with

their children.

Remaining Questions

The present study will add new information to what is known about gender

differences in mental state talk First, we will contrast mothen and fathen on their

production of cognitive, desire and feeling talk.

Second, we will explore the possibility of gender differences in parents' speech as it

occurs naturalistically. This context will allow us to determine if parents are diffeiently

socializing an understanding of the mind during routine, daily interactions with their

children.

Finally, because wm have a $ample of childen age two to six, we can begîn to

addnss the issue of gender diffwnces ovw a broed developmental period.

ni. ReWent d Mental Stato Tdk

Gendw differemcms hava been examid not only in t m s of the fieqwncy of

mental state tdk, but also in how childm talk about the mental states of other people.

Giiligan (1982) propored thrt m m and w o m di* in the way in which they M n e

themselves and othen. She suggested that women define who they are in tems of

their relationships and connecteâness with others, and that women are more likely to

act out of cuncem for others. In contrast, men are more likely to act in a way that is

consistent with laws and rules.

It has been report& that the conversations betwwn girls are markedly different

than those between boys. Tannen (1990) examined videotapes of girls and boys in

grade 2,6 and 1 O, conversing with a best fn'end. In summarizing hei observations,

Tannen no td thet girls of al1 ages engaged in conversations that were relationship

oriented, Mile boys tendeâ to tease each other, and discuss reœnt events and sports.

Furthemore, girls tendeâ to share issues relevant to one girl at a time, whilo boys

tended to talk about their own issues. Given the importance of relationships for girls

and women, it is thought that girls and women may refer more to the intemal states of

others when compamd to boys and men.

Referent of Belief Talk

Gender differences have not bmn reporled in shrdies that have examined children's

talk about the mental sûates of 'othen'. Hughes and Dunn (1998) recently exomined

the cognitive state talk of 25 tafget childnm, in conversation with a friend. Three

cetegories d referent were u d : ' s e W (the speaker / child), 'othet (usually the ChiM's

friend) and 'child plus friemf. The authon collapseâ the latter two group together to

obtain a meawre of cognitive tems tM iduded a referenœ to "other' (othei than the

drild). The results did not support Gilligan's ümry, as no significrwit ~Wtd8f

diFerences uvefe found in fehmnœs to ''dhd. lt would be interesting to detmine if

the rewlts persisted if aie catqories of 'othd and 'child plw friend had not b e n

collapsed. providing an indication of how offen boys and girls refend to sotneone

other than themselves, without being wntaminateâ by shared references.

Referent of Feeling Talk

Gender diffemnces have not been found in either aiildren's talk or parents' talk

about the feeling states d 'others' (other than child). Kuebli and Fivush (1992) studied

the refennt of feeling state talk in 24 mothen, fathers, and their children, during

conversations about past events. The children, half girls and half boys, wece 3 years, 4

months of age. Referent was codd as 'child' (child or parent refem'ng to child), 'other'

(child or parent rderring to oaier then child) or 'group' (shared by child and other).

Results did not indicate that one gender spoke more about the feeling states of 'osiers'.

Summary

The rewlts of the previoudy desaibed studies, Mile consistent, hould be

considerd inconcIusive because of their low nurnbet and methodological differences.

Previous research has obsewd children whL talking about past events (Kuebli &

Fivush, 1992). We know vwy littk about gendet dwerenws btween girls and boys

and be- moaCm and fWws in how they Wk about the mental states of others

during natunlly ocarmng conversations in mir home environment. In addition, the

conversation partmm hava differed betWB(M dudies. Kueôli and Fivush (1 992)

obsewd children wRh their parents, *il8 Hughes and Dunn (1998) obsewed children

interacting with a fnend. Futthennom, 'oam' wu categorized d i i t l y in =di study.

Hughes and Dunn collapsed the cetegories of 'child plus ftiend' and 'othet, while

Kuebli and Fivush did not include shared references in their 'other' cetegory.

Remaining Questions

The present study hopes to wmplement previous work and provide new information

about gender-specific patterns of taik about others' mental states by: 1) examining aie

proportion of cognitive, desire and feeling state talk that refers to 'other than the child',

2) observing families during naturalistic observation periods, 3) by involving both

siblings and parents.

The Effect of Siblings on Mental Sfit. Trlk

In previous woik, investigators have concludd that being the oldest sibling is

advantageous in ternis of intelledual abilities. Belmont and Marolla (1973) conducteci

a retrospective examination of family size, birth order and intelligence, using the Raven

Progressive Matrices as a measure of intellectual ability. Results indicated that as

family size and birth order position increasd scores on the Ravm decreased,

suggesting a disadvantage of larger fmily size and of being the youngest child.

Zajonc and Markus (1975) described aie optimal birth position in ternis of

intelledual ability, as king the firstborn of two, with the younger sibling appearing

closely in time to the older sibling. In particulat, it was suggested that the younger

sibling should be bom, optimally, Men the older sibling ha8 reBCh8d su(licicwit

cognitive maturity to play the role of a teacher, which will, in tum, k sdvantsg.0~~ for

the older child's intelleetual abilities. H~owever, a specific gap in time between sibling

births has not been identifid.

Siblings and Theory of Mind

It has recently b e n reportecl that children from large families do better on theory of

mind tasks. Pemer, Ruffman and Leekam (1994) investigated the effects of family size

and age on pHormance on false belief tasks. Results indicated that older children and

those from larger families demonstrateâ superior performance on false belief tasks.

The nlationship was hrrttier investigated by disceming the role of younger verses older

siblings in false belief understanding. Analyses failed to reveal an advantage of

younger verses older siblings.

Additional studies have revealeû that children with an older sibling are advantaged

in their falss belid understanding. The data in Pemer et al.% study (1 994) was re-

examined by Ruffian, Pemer, Naito, Parkin and Clements (1998) for two parts of e

fourgart study. Farnily size was divided into hnro variables, the number d youngw and

number of older siblings. The numbw of older siblings predicted performance on false

belief tasks. Moreover, Rulhan et al. tested a $ample of children who had comparable

numben of younger and older siblings. Analyses of this data indicated a signifiant

Mect for oldw siblings, but mt for younger. The authors mnduded that older, but rot

younger, siblings exerted a hclitative d e in falw k l W understanding.

0t)m res8arCtHHs have rwealed that it w m n d simply having an oldw sibling that

wa8 advantaamus, but nthrr the number d oldef childm me intmcted with. Lewis,

Frwman, Kyriakidou, Maridaki-Kassotaki and Berridge (1 996) exemind how the

number of kin and the number of daily interactions with those kin contributecl to false

belief understanding. Performance on false belief tasks favored thow children who

interaded with a greater number of older siblings the day prior to testing. When birai

order was considered, then, provd to be an advantage to being the thirdbom as

opposed to the secondborn, and of being a younger sibling verses being a first bom.

So, contrary to studies that addresseci the relationship between family sire, birHi order

and intelligence, then appeared to be an advantage of being the younger sibling in

terms of false belief understanding.

It has been reported by some that the advantage of having a sibling only applies to

children with lower language ability. Jenkins and Astington (1 996) sampled 32 girls

and 36 boys between the ages of 2 yean 11 months and 5 years 5 months, to

detemine the relationship between false belief understanding and a variety of cognitive

and family structure measures. Resulb were consistent with thpl of Pemer et al.

(1994), in that farnily size was a signifiant predictor of false belief understanding. The

authon suggested mat the presenœ of siblings in the home might advance false belief

understanding by providing a greater number of interactions duting which false bdief

understanding may k facilitated. In addition, language ability was also associated with

felse bdief understanding, in that false klief understanding was more strongly

associated with the pmsence dsiblingr for childm with laver language ability, thrn it

was for children with grnater languq8 ability. Furthemore, the nwl ts feild to

demonstrate a superiof W8d for olckr verses youngef riblin~s when falw bdief

understanding was wmpared in children with one older sibling verses one younger

sibling.

Sumrnary

Earlier studies have found that k ing the younger sibling is not advantagews in

tens of intellectwl ability (Belmont & Marolla, 1973; ajonc & Markus, 1975), whereas

discrepant trends have been identified for the development of theory of mind. Some

investigators have reporteâ an advantage in being the younger sibling (Lewis et al.,

1996; Ruffman el al., 1998), while others have not (Jenkins 8 Astington, 1996).

Remaining Questions

There were several methodological differences in the studies of sibling status and

theory of mind task performance, which impede a definitive conclusion. First, Ruthan

et al. (1 998) and Lewis et al. (1 996) did not include a measure of language ability, so

we do not know whether task performance w u influenced by an interaction behnreen

language ability and sibling status. Secondly, Rulhnan et al. found that the advantage

of an older dbling wcis @vident only for childrm over thme yean of age. Jenkins and

Astington did not provide s-c age8 of the children with either one older or one

younger sibling. Finally, cultural influencer on falm bdM understanding munt be

consided. Lewis et al. u s d a samplo of children living in the Ore& communities of

Cyprur and Crete, while Jenkins and Astingtm used a sample from Toronto, Canida.

It b as yet unclerir as to how diffisrent cultures may dbct family cohesion, intendion

and kin netirmrks, and how these f&tm may in tum nrSIii.ncs false ûelief

understanding.

The Ralationrhip k(nmn Oldw Siblings urd Mental State Trlk

It has been found that children with an older sibling tend to outperfom those without

on measures of false belief (Lewis et al., 1 996; Ruffman el al., 1 998). Furtherrnore,

passing false belief tasks has been associated with the frequency of mental state talk

(Brown et al., 1 QQ6; Hughes & Dunn, 1998). A natural extension of these associations

would be to study the relationship between having an older sibiing and the frequency of

different types of mental state talk.

Rernaining Questions

To date, there are no studies of which we are aware that have examined the

relationship between having an older sibling and mental Mate talk. The present study

aims to addresr the methodological issues that have h n identifid in past studies of

sibling order and theory of mind tasks. First, we have induded a correlate of langwge

ability, the Mean Length d Utterance (MLU) in order to determine whether the

advantage of an older sibling is related to language ability. Second, all of the children

we compare are over four yeam of qe , thus taking into account Ruthnan et al.'s

(1 998) finding that older siblings only bendit younger siblings over the ege of three.

The design of the present study complements that of other studies and may provide

answers to questions that have not previously h n asked. First, this study is

longitudinal, spanning a two-year period and including children from 2 to 6 years of

age. We focused on the naturally ocarning we of concomitant cognitive, desire and

feeling state talk. The conte* of the observations was consistent among families, in

that al1 observations were made in the farnily's home, duting naturally accuming periods

of both play and conflict. Unlike previous research, the present study examina

interactions between mothers, fathen, and sibling pairs, providing a comprehensive

examination of intemal state talk in families.

This study had Viree goals. The first was to map the relative changes in children's

cognitive. desire and feeling state talk be-n the agas of two and six. The data set

was such that the onset of talk about mental States could rot be examineci. Instead,

developmental trajectories wem examined using meastires of frequency. This

particular rnethodology does not dlow us to map the ôeginning of children's

understanding of mind. However, it does allow us to identify developmental changer at

an age *en children's pmdudion of mental state tems might be more ftequent.

Given the findings fmm pmvious mmarch, wa expected to find an incn,am in cognitive

talk over time for both ywnget anâ older siblings. In light of the contradidory findings

of Bartsch and Wellman (1995) ami Moore et al. (1994), w expected only that desire

tems would be the mort common type of mental state tilkwbn childm were two

yean old. We preâided that feeling talk would increase over time, but remsin Iess

common than talk about cognitions or desires.

Changes in parental cognitive, desire and feeling talk are also examined in this

study. Ptevious research has examined only matemal mental state talk, and found

that cognitive talk increased as children's age increased, while talk about desires

decreased ffom the time children were two years old. Wo predicted a similar trend for

cognitive and desire talk Studies of feeling talk have been less conclusive and so we

made no particular prediction. Furthemore, previous studies have not involved

fathers. However, we predided that fathers would demonstrate changes in mental

state talk similar to that of mothers.

Our second goal was to examine gender differences in talk about others' mental

states. Given the differenœs in how males and fernales define thernselves and others,

we predicted that girls would talk more about the mental States of other people than

would boys. We also predided that mothen, more so than fathen, when talking to a

child, would talk mon about the mental states of wmeone othw than the diild.

Furthemore, given the Wnd for talk about desires to begin earlier than that of

cognitions, we pfedicted that talk about the desims of others would k greater thon talk

about aie cognitions of dhers, et Imst when childm wbn îwo yeam of age.

The third goal war to examine the dF8d of having an older sibling on meMd state

talk. At the 2. the mean age of the ywnger childm who had an older sibling was 4.4

years. At time 1, the man age of the 01- siblings, who thmseIve8, did not have an

older sibling, was 4.4 years. This similarity in age a l l d us to examine aie

relationship between the presence of an older sibling and mental state talk. Given the

trend for children with an older sibling to be advantaged in fatse bdief understanding

(Lewis et al., 1996; Rdman et al., 1998), we predicted that those with an older sibling

would produce more cognitive terms than those without. We predided a specific

advantage for cognitive state talk over other types of mental state talk based on the

proposal that cognitive talk requires more advanced representational skills than talk

about desires (Barlsch & Wellman, 1995).

Pu(lcipwit8

Forty Caucasian families, eech consisting of two p a r 0 and hwo children living

together, participatecl in air study. F amilies were rmi ted based on birth

announments in the local newspaper. me data were originally collecteci to examine

the concept of moral orientation (Lollis, Ross 8 Lefou, 1996). ln the initial intenriew,

parents wem told thet investigatws wsn, interested in the rdationship between their Iwo

children, as well as in how children leam funily rules and expectations for interpersonal

behavior. The childm wen told that the obsewews were coming into their homes and

would watch how they played together. They were asked not to interact with the

observer.

At time 1, the older children were ôebmen 3.8 and 4.9 years of age (M = 4.4 years)

and the younger childrm were betwwn 1.9 and 2.6 w= 2.4 y-). At time 2, the

older children were between 5.4 and 7.0 yem (Y = 6.3 y@-) and the younger children

were between 3.8 and 4.8 yean = 4.4 yeam). The gender of older and younger

chilben was balanced for an eqwl numbr of riIl possible brother/sistm combinations.

Families lived in a medium-sizd industrial city in SOUfhWBstern Ontario. Caneda. All

fathen ware mployed Usîâe the home on a full-tirne basb, and 29 mothem were

employed outside the homa on a hl1 or pmt-tirne baais. Fathers or other family

membem genwally cand for the childm, in üie mothors' absence. Parents' educational

backgrounds varid wiâely in the mple: 29% hrd completed a univemity degm, 15%

haâ completed r collage progmm, 41% had completed high s c h d and

15% had not gmduated from high schod. At time 1, parents' ages rangd between 23

and 48 yeam of age (M = 30.8 for mothers; M = 32.6 for fathers). At time 2, thrw

families wem unavailable to participate, as one family had moved away and Iwo others

were in the process of divorce or sepration. Therefom, al1 analyses wen, conducteci

with the remaining 37 families.

Procedure

The data used in this study was ofiginally collecteâ for an investigation of moral

orientation carried out by Susan Lollis and Hildy Ross. The majority of the procedure

that follows is taken directly (with permission) frorn a publication pertaining to the

originel data collection (Lollis, Ross & Leroux, 1 996). Tho section titled coding mental

$ta& talk marks aie beginning of the procedurer carriecl out for the present study.

Behavioral observations.

Data were collected during six Qûmin observational ses8ions in the homes of the

participants, at each of the hno tirne pieds. Duing haif the sessions both parents

were present (mother-father sessions) and during the remaining sessions the mother

and children were obsewed without the father (moaHKonly sessions). These two

situations wwe used as they wrwe thought to mpresmt the most common

con8tellations in the famil y.

During the sessionr, an observer follannd the childm nd dictated ont0 one tradc

of a stem audiotape a descriptive rccount d all interactions botwem the childm and

36

of al1 parental behaviots that related to the childrm's interaction. On the second track

of the tape, a recording was made of the s m that occurred in the home. Obrervers

did not participate in family interaction and nsponded as little as possible to camments

of family mernbers. For obsmations to proceeâ the children had to k in the same

room and parents had to be either in the rame or an adjacent room, although in both

cases allowances were made for brief absences of up to 2 min. Televisions, video

games, or other major distractions wn not allowed. Whenever these requirements

were not met, observeni stopped recording and either waited until the participants

cornplied with these provisions or arranged to obsenie again et a tirne that was more

wnvenient for the families. To maintain stability and rapport. two obsewers were

assigned to each family. To limit the intrusiveness of the observations, only one

observer was present during each observation session. These procedures were

repeated two yeats later. The content of the speech that family membws directed to

one another and a descn*ption of each person's adions were accounted for in coding

the transcripts.

There is some research suggesting that girls use mon, feeling words than boys

(Kuebli et al., 1995) and thrt mothm use mon, fmling wordr than fathers (ffiebli &

Fivush, 1992). In addition, Condry and Condry (1976) demonstmted that when

observen know a child's geridet, they diffmntly Mbute emotions. To prevent biased

coding of tranraipts, codefs wcne blind to th8 g d e r of parents and the gender of

childm. For example, if r child indiCBfd gmdw whm refwing to thair siMing (e.g.,

het she), the mtarence was rspl8cad by a code for ribling, with no indication d gendw.

Coding Mental State talk

Mental state talk was divideâ into three categories; cognitive, desire and fwling

talk M a t follows is a description of tems included in each category.

Cognitive talk includeâ tems wed to denote the thougMs, mernories or knowledge

of the speaker, listener or a third person. nie tems included in this category were the

tems think, know, believe, wonder, remember, bfget, guess, pretend, understand and

expect, and aall variations. Shatz et al. (1983) found these to be the most cornmon

cognitive state tems uttered by young children, and these ternis have been included as

examples of cognitive talk by leter researchen, (Bartsch & Wellman, 1995; Hughes 8

Dmn, 199û; Moore et al., 1994).

There is rom8 variation in the coding schernes of past research, with respect to

whidr ternis to include and which to exdude. Battsch and Wellman (1995) were

concerned with the onset of references to genuine mental states, and so used a

conservative toding scheme that included only mentalistic uses of ternis and excluded

convenational uses (ag., 'Do you know ...'). Conversely, other researchem have

included convemational uses (Brown, Donelan-McCall8 Dunn, 1996; Fuww et al.,

1 992; Hughes & Dunn, 1998; Pemer, 1 991 ). Much of the support for the inclusion of

conversational uses atise8 fmm work by Moore and hir wlleagues (Moore, Bryant &

Fumm, 1989), who establishd that childm as young as four wem able to distinguish

betwcnm varying Ievels of œrteinty in cognitive terms, and used this kndedge to

modulate asseftio~. Fu-, a8 B m et al. pointed out, it would be enoi#ous

to assume that statmmnts rudr as '1 dont k n W am nothing m m than idiomatic

axprsuionr, whm umd by adub or childnn. We induded conv«rrtional usas of

cognitive tennr basa on these arguments and the recent finding of Hughes and Dunn

that both 'genuine' and 'conversational' uses of cognitive terms were associated with

performance on theory of mind tasks (Hughes & Dunn, ? 998).

Consistent with Shatz et al. (1983)' we included "know Mat" if it was used to direct

an interaction by introducing information (ag., "know what, 1 have a... ') and references

to '1 know" and "1 dont knoW if a descriptive statement is made irnplicitly, not only

explicitly, such as 'the big snake is dangerous," " I know." Consistent with Bartsch and

Wellman (1995), "1 knog and "1 don't know' were included if they were linked with a

description of ignorance. Consistent with Pemer (1991), we included know as it r e f h

to an ability (e.g., '1 know how to tie my show'), or to facts (0.0.' '1 know rny socks are

in the drawef) or to access information.

Instances of cognitive state tems which were excluded were undeai meanings of a

tenn or sentence fragment, repetitions of own or others' utterance and ternis useâ in

direct response to a question (e.g., 'Where do you think the sock is?" "1 think the sock

is in the drawef) and '1 knaw Boô" if it c m be paraphrased to mean '1 met Bob"

(Bartsch EL Wellman, 1995; Shatz et al., 1 983).

Oesire t m s induâed went, hop?, wish and c8m, and al1 variations of these termr

used to capture diildmn's desires or goab (Bartsch & Wellman, 1995; Shatz et al.,

1983). Specifically, wanf was induded as a refmnce to a goal directed behwior (e.g.,

'I wsnt to rit down'), h o p r it refend to a wish or wint (ag., '1 h o p Santa canes

mona), and a m as it referred to a prof- or Iack of pmbmme (e.g., 'I dont cus

which crayon I usen) (Bartsch 8 Wellman, 1995). As in Bartsch and Wellrnan and

Shatz et al., unclear rneanings of a term or sentence fragment and repetitions of own or

othen' utterances were excluded, as were idiomatic phrases such as "taking caren or

"wish upon a star."

Feeling tems included those that referred to an emotional state. Consistent with

previous coding criteria, we included al1 variations of sad, hurt, angry, happy, excited,

love, dislike, afraid, enjoy, fun, glad, mad, scared, upset, suprise and fear (Dunn et al.,

1987). In addition, disgust was included, consistent with Dunn. Brown and Beardsall

(1 991). The terrn like was included when it referred to a state of enjoyrnent or dislike,

and good was included only if it denoted a feeling state (Dunn et al., 1987). Phrases

that connote a feeling state were also included, such as "make a fuss" (Dunn et al.,

1987). Excluded feeling tems consisted of non-verbal expressions (e.g., crying or

laughing), specific and non-specific expletives (e.g., ÿuck"), like when it indicated

volition and good if used in a moral sense (Dunn et al., 1991). As in Bartsch and

Wellman (1 995), unclear rneanings of a tenn or sentence fragment and repetitions of

own or others' utterances were excluded.

Coding Referent

The category of referent of the mental state talk described whose mental state the

speaker was refemng to. Speech was divided into three categories, consistent with

Kuebli and Fivush (1 992); (1) other (other than the chiid), when the parent speaking

referred to him I her self; or to a third person; (2) child, when the child who was the

conversation partner was being referred to; (3) group, indicating that the mental state of

the speaker and anothw wwe k ing refeweâ to. The catqory of 'groupr wat also used

W n the mental state of the Iistener and the speaker wen, cornbined in an embedded

construction. For example, '1 know that you want to go " is a linguistic structure known

as an embedded sentence. Such uttefances w l d k coded as a cognitive terni (with

an embedded desin, proposition) ref-ng to a group (deVilliers, 1995).

As we were interested only in references to people other than the target child,

references to 'child' and 'group' were not included in analyses. To examine parental

references to 'other' (other than the child) for cognitive talk, we divided r8fwenc8s to

'othet by the total number of references to cognitive states to create a proportional

score. The same produre war cameâ out for desire and feeling talk and for fathers.

All variables wbre transformeci using a square root transformation to account for

positive skem.

Based on gender differencm in the vmy that males and fernales define themselves,

we predicted that mothen, more than fathen, would mode1 for childm how to talk

about the mental states of amthar person. AS such, we are interest8d in wmparing

mothem and fbthws in mir piodudion of mental state temr aâdresaed to 'other,'

meaning 0 t h ~ th8n the Child. It may bu that by talking to childm about the mental

states of m e o m 0 t h ~ than the child, parents will promot8 mir child's interest in the

mental states of 0 t h ~ people.

Consistent with the coding $chme af W b l i and Fivu* (1992), thme categories of

refetent w m usud for childmfs speech: (1) wlf, W m the child mhmd to Mm or her

self (2) other, when the child referteâ to their conversation parfner (2d person), or

made a 3d peson reference (e.g., other parent, sib) (3) gmup, when child refened to

self and another.

To examine diffemnœs btween boys and girls in their references to somebody

other than themselves, cognitive talk direcîed at 'osier' (second I third person) was

divided by the total amount of cognitive talk to mate a proportional score. This

procedure was repated for desire and feeling speech and was performd seperately

for younger and older siblings. All variables were positively skwed and so a square

root transformation was perfomeû.

Based on the literature, we predided that girls would talk more than boys would

about the mental states of peopk other than themselves. Therefore, we are interested

in comparing gender dWemnces in the frequency of mental state talk for the category of

'other.

Inter-rater reliability was calculateâ for the t y p of mental state talk and the refefent,

using 1 S pet cent of the families, each a time 1 and time 2. For time 1, the per cent

agreement for cognitive state talk was 97%, for desin alk, 98% and for feeling talk

88%. For time 2, the par cent agreement was 98% for cognitive talk and desim talk,

and 97% for fmling state talk. Kappa values for the -nt at tirne 1 and time 2 w m

0.91 and 0.93, respedively.

Conversat ional Tums

The original tmnscripts w m divided into convenationrl tums, which wem defined

as al1 of one speakedr uttefance8 bounded by the utterances of another speaker

(Shatz 8 Gelman, 1973). Sentence fragments and mpetitions were not included. A

conversetion tum may have contained mors than 1 utterance.

Utterances

As in Shah and Gelman (1973) an utterance was defined as a complete clause.

For the purposes of mis study, an utterance was oonridered to be a group of words

bounded by a period or question mark on the trenscriptr, end includeâ both forneci

sentenas and conversationil phrases. Utterances mat wfe sentence fragments or

exact repetitions of pmviour utterances were not included. tf non-worddbehaviour was

al1 that occurreâ, it war not counted as an utturanco or as a conversation tum. Singing

was not counted a i an uttefance, a convemtion tum, or a word.

Mean Lemgth of Utteranœ (MLU)

The numW of words in the 100 utterances follming the fint ten conversation tums

was avefageâ and constituted the MLU (Shatz & Gelmen, 1973). This was calculateâ

for each child at time 1 and time 2.

When counting the n u m k of wofd8 in an uttemnœ, the aiteria set by Shatz and

Gelman (1973) wem fdluwed: (1) cmtmüons (O& an't, wanna) and pioprr name8

(e.g., Julia Webtw) nnn courited as hno wwds; (2) hyplmated wwdr (ebg., beep-

bop) wwe counted r me w#d. ExpIatives and rwndr PiggW, 'Ah hrh' wem mt

counted as words. The interiater nliability for MLU calculetions was .97 at time 1 and

.74 at tirne 2, for the younger sibling. For the older sibling, the mliability was .97 et

time 1 and .85 at time 2.

As al1 of ouf main questions involved diifences between mothers' and fathen'

mental state talk. oniy date h m rnother-faaier sessions wen used in analyses. In

order to account for the significant diflerences in conversation tums between family

mernbers, proportional scores were used for al1 variables. In accordance with Brown et

al. (1 996), we created proportional scores by dividing the number of cognitive state

ternis by the number of conversation tums. This process was repeated for desire and

feeling ternis and for each family member. As al1 variables mre positively skewed,

they were transformed using a square mot transformation.

In order to examine differences betWBBn mothers and fathers in the mental state talk

that they direded to their younger and older children, summed scores of mothers'

cognitive, desire anû féeling talk to either the younger or older diild was created and

divided by her total number of conversation tums. The same procedure war carrieû out

for fathen' speech. As al1 variables wre positively skwad, they wre transfomied

using a square mot transftmnation. In one family, the fathgr, at time 2, used mom

cognitive and fWing tems thon he had speaker tums (multiple uses per tum).

Although scores for al1 pamnts w e le81 than 1 (ranged fiom O airough .82), scores for

this fatherwere ovw 1. In oder to reduce the dthis outlier, it was riscoded to '1'.

P amts alm directed r#nirkri to both of their childm sirnuttane~usIy. At üme 1. the

proportion of directed to both childm w# -4% d th. mental $tate Uk. At time

2, it had flsen to 16.5%. Because the proportion of talk was so lw et time 1, it was not

possible to maintain thb as a separate category of talk in subrequent analyses.

Cons8quently, wo droppd speech dir-eû 1 both childrsn from the analyses that am

reported below. In ofder to ensun that mis did not Iead to erroneous conclusions, we

always fan subsequent analyses for time 2 wparately, in which speech directed to both

children was included as a third Ievel of ouf age of child variable.

Convmaüon Tumr

Over time, the number of mothw conversation tums inueased significantly,

1 (36) = 4.24, p < ,001, as did fathers, (36) = -3.99, Q c -001. The means and standard

deviations far the- variables can k found in Table 1. Mothm engaged in mon,

conversation tums than fathers st time 1, f (36) = -2.76, g c -01. There was a trend for

fathen to engage in mon conversation tums than rnothen at time 2, f (36) = 2.01, p <

.M. It is important to note that on, father at time 2 was an outlier in the very large

number of conversation tums that he pmduced. Rether than changing the value of this

outlier, t h t-test was conducted bo(h with and without this f a W . As the results of the

t-test did not change substantively, he mmains in the analyses. Moreover, dl

subsequent analyses wbm conduded both with and without this father. As none of the

results reporteci below change substantively, ho ma ins in al1 analyses.

Over timo youngor and older childm increased the number of c o n v ~ t i o n tums in

which they engageô, f (36) = -7.56, p < .W1 and f (38) = -7.48, p .ml, respecüvely.

The means and standard daviationr for these variables cari al- be fowid in Table 1.

Fu--, olâer childm producd m conversation hrms thn younger childm et

time 1, t (36) = 4.25, Q .001, and time 2, f (58) = -3.07, Q < -005.

Younger Child 7.56"

Mother &! 40.46

sp 28.83

Mean Length of Uttomce (MLU)

The MW foi youngw diildm was significwrtly greatw at üme 2 than a time 1, f

(36) = 6.31, < .ml. For oldw childm, MLU did not change significantly over time.

The rneans and standard deviationr for these variables can be found in Table 2. Older

children had a higher MLU than their ywnger ribling at time 1, (36) = -7.55, p c .al,

and again at time 2, t (36) = -2.3, p c .03. When ma MLU of younger siblingt at time 2

(men they wen 4 y w r old) and 01- siblings at tim 1 (when they wwe 4 ymn old)

wem compudd, üwe was no rignificant d i i n œ .

Younger !Y! 2.72

sQ 0.85

The Relrtion8hip k(nmn MLU and Montal State Talk

ln order to examine the reletimship behmen children's use of cognitive, desire and

feeling tems and their mon, general use of Ianguage, we conduded bivariate

correlations between MLU and each type of speech, for younger and older children

separately. The only significant C O K ~ ~ ~ O ~ S were between the younger chilci's desire

state talk and MLU a time 1, and cognitive stata talk and MLU at time 2, as cen be

seen in Tables 3 and 4. So, as the younger children are b m i n g mon, verbally

fluent, the strength of the mlationship betwaen MLU and cognitive state talk increaser,

while the relritionrhip btween MLU and desires and feelings decreases. For the older

children, as they becom more verbally fiucmt, the stmngth of the mlationrhip betwvm

MCU and cognitive $tate talk dmeam~, whib the nlationthip between MLU and

desire and fwling state trlk imams.

Conalrtionr behum MLU and toeoch of vounau riblipoIiPOI

Time 1

Cognitive

Desin

Feef in9

Time 2

Cognitive 0.35.

Derin 4.23

Feeling -0.13

*Q < .O5

TaMa 4. Conelrtions between MCU and sgoach of oldu Jblinaa.

Time 1

Cognitive

Desire

Fding

Time 2

Cognitivs

DmCn

F d i n g

According to Cozby (1 997), the strength of a correlation will increase if a broad

range of scores is involved. So, to better understand the conelations between mental

state talk and MLU, the range of the different types of mental state talk was examined.

Upon closer inspection, the conelations that decreased in strength do in fact have a

reduction in the range of mental state talk at time 2. For example, the range of desire

and feeling talk for younger siblings nartowed ove? time, as a result of individual

differences in development. For those correlations that are stronger at time 2, such as

the younger siblings' cognitive talk, there was an increese in the range of cognitive state

talk.

Question #1: How does the balance between different types of mental state talk in

children and in parent's change with children's age? Outcome variables were the

proportion of conversation tums that involved cognitive, desire and feeling state tens.

Younaer Children

A 2(time) x 3(type of speech) x 2(gender) repeated measures analysis of variance

(ANOVA) for the younger children revealed a significant main effect of time, (1,35) =

52.96, p < .001, with children using more mental state talk at t h e 2 than at time 1.

Type of speech was also significant as a main effect, E (2,34) = 52.81, p < .001, with

children talking most about desires at both time periods. Gender was significant as a

main effect, (1,35) = 6.54, p < .02. There were no significant interactions between

gender and time or type of speech. so girls spoke more than boys dld about cognitions,

desires and feelings at time 1 and time 2.

There was a significant interaction between time and type of speech, E (2,34) =

27.50, p c .W1. A doubly muhivariate N O V A was used as a posthoc test, to

detemine which type(s) of speech changed over time. The doubly muMivariate ANOVA

revealed that the proportion of cognitive speech (as a proportion of the nurnber of

conversation tums) increased significantly, E (1 , 35) = 141.51 , Q c .O01 , as did feeling

speech, E (1, 35) = 8.94, Q c .0W. The proportion of speech about desires did not

change significantly over time, E (1,35) = 0.73, n.s. By time 2, cognitive talk is

ocairring almost as frequently as desire talk. The data are presented graphically in

Figure 1 to enable the patterns to be seen more clearly. See Table 5 for means and

standard errors.

To further explore gender differences in mental state talk, the MLU of boys and girls b

was compared et time 1 and time 2. Results of t-tests indicated that at time 1, the

verbal fluency of girls was significantly greater than that of boys, 1 (35) = 2.55, e < .02.

At tirne 2, although girls spoke more than boys did, the difference was no longer

significant. Overall, the magnitude of the gender differenœs was greater at time 1.

when children were 2 yeers of age, than et time 2, when they were 4 yean of age.

These results suggest that the gender dMlerence in mental state talk may be best

explained by girls' genenl cornpetence in language when compared to boys rathef than

a cornpetence that is specific to mental state talk

Older Children

A 2(time) x 3(type of speech) x 2(gmder) repeated meesures N O V A for older

childmn revealed a signifiant main effect of type d speech, E (2,34) = 87.79, Q c 001,

with children speaking most about desires, followed by cognitions and then feelings.

Time was not significant, E (1, 35) = 1.63, ns., nor was gender, E (1, 35) = .6, n.s.

The interaction between time and type of speech was signifiant, E (2, 34) = 28.37,

e e .W1 (Mer to Table 4 for means and standard deviations). The post-hoc doubly

muitivariate ANOVA revealed a signifiant increase in talk abwt cognitions, 5 (1, 35) =

15.69, g c ,001, and a significant decrease in talk about desires, E (1, 35) = 20.54, Q <

.O01 and feelings, E (1, 35) = 6.26, < -02 (as a proportion of the total number of

conversation tums). As cen be seen in Figure 2, older children spoke proportionately

more about cognitions over time, whereas they spoke proportionately less about

feelings and desires. At time 2, although older children spoke marginally more about

desires than cognitions, the difrence belween the Iwo was slight. The means and

standard mors for these variables can be found in Table 5.

Although there are no gender differences in mental state talk, a t-test was

conducteci to compare the MLU of older boys and girls at each time period. There was

no signifiant difference in MLU between boys and girls at either time. So, the older

girls did not demonstrate greater ampetence in verbal fluency or frequency of mental

state talk than the older boys.

Table S.

Mean IM) and Standard Error (SEI for chanaes in children'~

mental state talk at time 1 and time 2,

Younaer

cognitive M - SE

feeling !Y!

SE

cognitive hd s€

feeling M s€

Figura 1.

Mean ~ro~ortion of vounaer children's coaruve. desire and feeling s~eech at tirne 1 and time 2,

t i m e 2

O

CO g n itive desire feeling

Type of Spcrch

Figure 2.

M e a n p r a D a r t i o n o f oldetctiildren'seech at time 1 and tirne 2.

O .1 t i m e 2

O

CO g n itive desire fe e lin g

Type of Speech

Parents

A 2(time) x 3(type of speech) x 2(age of child) x 2(parent) repeated measures

ANOVA revealed a significant main effect of type, (2,35) = 19.43, g c .001, with

parents speaking most about cognitions and desires, and least about feelings. Parent

was significant as a main effect, with mothers speaking more about mental States than

fathers, E (1, 36) = 7.1 3, c .02. There were no significant interactions between parent

and time, type or age of child, so mothers spoke more than fathers did about

cognitions, desires and feelings under al1 conditions. lime was not signifiant as a

main Med, E (1 , 36) = .26, n.s., nor was age of child, E (1, 36) = .92, n. S.

The type by time interaction revealed a trend for talk about cognitions, desires and

feelings to change from time 1 to tirne 2, E (2, 35) = 2.87, g < .08. The doubly

mulüvariate ANOVA revealed that the proportion af cognitive state telk imeased

significantly over time, E (1, 33) = 5.6, g < .03, while speech about desires, E (1, 33) =

.37, n.s. and feelings, (1, 33) = .53, n.s. did not change significantly. At time 1,

parents spoke proportionately more about desires, followed by cognitions and feelings.

At time 2, parents spoke proportionately more of cognitions, followed by desires and

feelings. Refer to Table 6 for means and standard enw and ta Figure 3 for a graphical

representat ion.

Table 6.

Mean LM) and Standard Emr ISE) of ~arents' mental state talk to children at tirne 1 and 2.

State Talk Ttme 1 Time 2 F (1,331

Cognitive M E

Feeling M -LI

SE

Mean ~rogortion of parents' coanitive. desire and feeling s~eech at time 1 and time 2.

cognitive desire feeling

Type of Speech

Cognitive, desire and feeling tilk as a percantage of mental state talk

The frequency of cognitive, desire and feeling talk was also calculated as a

percentage of the frequency of mental state talk, separately for each family member at

each time.

For younger children, cognitive talk represented cl % of al1 mental state talk, desires

81 % and feelings 18% at time 1. At tirne 2, cognitive talk represented 39%, desires

48% and feelings 13%.

These results indicate that desire and feeling talk, as a proportion of mental state

talk decreased over time, as cognitive talk increased with age.

For older children, the pattern was similar to that seen for younger children. The

percentage of cognitive, desire and feeling talk at time 1 was 18%, 66% and 16Oh.

respectively. At time 2, the percentages were 40%,49% and 11 %, respectively.

As children get older their mental state talk is increasingly about cognitions. These

patterns are consistent with those produced when mental state ternis were expressed

as a proportion of the total number of conversation tums.

The cognitive, desire and feeling state talk of mothen and fathers was summed

together and then calculated as a percentage of mental state terms. At time 1, the

percentage of mental state talk that was cognitive talk was 38%, desire talk was 42%

and feeling talk was 20%. At time 2, the percentage of cognitive talk increased to 49%.

with a decrease in desire talk to 33Oh and feeling talk to 18%.

57

Question 2: We are interested in answering two questions. First, do mothers and

fathen differently socialize their children to speak of the mental states of people other

than the child? Second, are thwe differences in how boys and girls refer to the mental

states of people other than themselves?

Parents

A repeated measures P(time) x 3(type of speech) x 2(parent) x 2(age of child)

ANOVA revealed a significant main effect of age, E (1, 36) = 7.16, < .02, with parents

talking more to older children about the mental states of someone other than the child.

Type of speech was also significant, (2, 35) = 7.88, e < .002, with parents talking

most about othen' cognitions, fo l lmd by desires and feelings. Parent was significant,

F (1, 36) = 13.16, p e .002. As can be seen in Figure 4, mothers talked more about the 9

others' mental state than fathers did. T i m was not significant, E (1, 36) = -15, n.s., and

there were no significant interactions, so rnothers spoke more about the mental states

of people other than the target child, at both time 1 and time 2, irrespective of the type

of mental state that was k i n g referred to or age of child. The means and standard

error c m be found in Table 7.

Children

Younger Children

A 2(time) x 3(îype of speech) x 2(gender) repeeted rneasures ANOVA on younger

children's talk about the other revealed a significant main M d of time, E (1 , 35) =

7.36, p < -02, with children showing highet levels of talk about sorneone other than

themselves at time 2 than at time 1. Type of speech was not significant, E (2'34) = .6Q,

58

n.s., nor was gender, E (1, 35) = -41, nos. Them was a significant interaction between

type of speech and gender, E (1, 35) = 4.18, < .03, with younger girls speaking more

about others' cognitions and desires. Means and standard enar can be found in Table

Older Children

A 2(time) x 3(type of speech) x P(gender) repeated rneasures ANOVA on older

children's talk about the mental states of someone other Vian themselves revealed no

effect of type of speech, (2, 34) = .47, n.s. Gender was significant as a main effect,

(1, 35) = 4.03, p = .05, with girls speaking more about the mental states of others' than

boys (se8 Figure 5). There was no signifiant main effect of time, (1,35) = 1.75, n.s.

There were no significant interactions, so girls spoke more of others' mental states

regardless of time and type of speech. Means and standard erroi can be found in

Table 7.

Mean lM1 and Standard Err~r (SB of talk about the 'othet fother than 1118 child) for ~arents and children at tirne 1

and time 2. as a function of aender,

Time 1 Time 2 F

Parents

Mother M -

SE

Father M -

SE - Older Children

Boys

Girls M -

s!E Younaer Children

Girls M s€

Mean differences in the morti ion of ~arents' talk about 'other' (other than tamet chil&

mother father Gendrr of Parent

Figure 5.

Mean differences in the ~ro~or t ion of older siblinas' talk about 'other' (other than themselves).

girls b o p

Gendrr o f ChilJ

Question 3: Are 4-year old children with an older sibling advantaged in their cognitive

state talk, relative to other types of mental state talk?

Children's talk when they were 4 years old was examined. At time 2, the mean age

of the younger children who had an older sibling was 4.4 years. At time 1, the mean

age of the older siblings, who themselves did not have an older sibling, was 4.4 years.

This meant that talk at time 1 for childten without an older sibling and talk et time 2 for

61

children with an older sibling war the dependent vaieble. A mked within and belween

3 (type of speech) x 2(psition) ANOVA revealed a signifiant main eflect of type of

speech, (2.71) = 73.62, p < ,001, with children talking most about desires, then

cognitions and then feelings. Position was not significant as a main effect, 5 (1,72) =

1.77, n.s.

There was a significant interaction b e W n type of speech and position, E (2, 71) =

12.52, e < .MM. As a n be seen in Figure 6, children with an older sibling spoke more

about cognitions and less about desires and feelings. Means and standard error can

be found in Table 8.

In order to ensure that the results of position were not explained by differenoes

between groups in MLU or age, a 3(type of speech) x P(position) x 3(MLU) x3(age)

ANOVA was conduded, with MLU and a g each recoded to trichotomized variables.

The position by type of speech interaction temained signifiant, E (2, 55) = 9.32, p c

.ml, demonstrating that the position effect wae not explained by differences between

group on MLU or on age.

Table 8.

Mean IM) and Standard Error (SB of M 8 m State Talk. With and Without an Older Sibling.

With an Older Sibling Without Older Sibling

State Talk - M SE - n - M SE -. n

cognitive

desire

feeling

Figure 6.

Mean ~roportion of coanitive. desire and feetina talk in children with and without an older sibltna.

cognitiw desire feel hg Type of Speech

CHAPTER 4: Di8curdon

Many important developrnental changes in children and mothers' talk about mental

states have been identified in the literature. The present study sought to elucidate

previous findings and provide new information about mental state talk by addressing

several significant gaps in Our understanding. First, little is known about changes in

children's or parents' talk about mental states after children reach four years of age, and

so we included children until age six. Second, we examined concomitant changes in

belief, desire and feeling talk to augment what is known about the balance between

different types of talk as children get older. Throughout this discussion, when changes

in children's talk is discussed, we refer to the type of mental state talk as a proportion of

total conversation tums. Third, many studies have addressed mental state talk in

contrived, time limited contexts. By studying families naturalistically and over an

extended period, we hoped to uncover patterns of mental state talk as it occurs in

everyday conversations. Finally, fathers have been largely ignored in studies of mental

state talk. For this study, we have included only dual-parent families so that we may

begin to identify changes in fathen' mental state talk as a function of children's

developmental stage.

The Relationship Between MLU and Mental State Talk

There are two factors to consider when interpreting the correlations between MLU

and mental state talk. Fint, the relationship between MLU and mental state talk is

affected by the range, which decreases in cases where conelation strength decreases

('e.g., desire talk for younger siblings) and increases in cases where conelation strength

increases (e.g., cognitive talk for younger siblings).

Second, as children are growing older, the proportion of the different types of mental

state talk is changing. In some cases, individual children use proportionately less

desire talk while using propoitionately more cognitive talk.

The Development of Children's Mental State Talk

The first aim of this study was to examine, sirnultaneously, the developrnental

changes in children's cognitive, desire and feeling state talk. The present results

indicate that cognitive talk, expressed as a proportion of conversation tums, burgeons

at age four and continues to increase until age six. These resuks provide converging

evidence with previous studies, with cognitive talk presented either as a proportion of al1

utterances (Bartsch 8 Wellman, 1995; Moore et al., 1994) or as the number of terrns

per hour (Hughes & Dunn, 1998).

Despite the methodological diffennces, the results of the present study do agree

with the pattern reported by Bartsch and Wellman, that talk about desires is the most

frequent type of mental state talk between ages 2 and 6. With few exceptions, the

majority of past research has studied children only until the age of four. This study not

only supports previously established trends, but also off ers new information about

changes in cognitive talk beyond four years of age.

These latest results also indicate that feeling talk, as a proportion of conversation

tums, increases until the age of four, and then begins to decline. Our results are

consistent with those of Dunn et al. (1987), who found that children's production of

feeling temis increased between 18 and 32 months of age, when represented as the

number of conversation tums containing a reference to feelings. That we found

references to feeling talk decreasing after the age of four suggests that feeling state talk

reaches a peak by 4 years of age, and decreases as cognitive state talk increases.

By the age of four, the children in our sample were able to discuss the events of the

mind in terrns of beliefs, desires and feelings. It is important to map the developmental

changes in a normative sample of children so that in the future, developmental

deviations might be identified and remediated.

The Development of Parents' Mental State Talk

The mental state talk of parents was examined for changes over time. The resuîts

of the current study find that parents increase their use of cognitive state talk over time.

as a proportion of conversation tums. This finding is consistent with those of other

researchen who expressed cognitive ternis differently, as a percentage of utterances

(Furrow et al., 1992). While the current study does not address the issue of

directionality between children and parents' speech, it is worthwhile noting that such

changes are synchronous and may be indicative of modeling.

The present resuk indicate that mothers and fathers do not change their frequency

of talk about desires over time, as a propoition of their convenation tums. Earlier

studies have reported that mothers decrease their proportion of desire talk, albeit with

younger children (Brown & Dunn, 1991 ; Moore et al., 1994), and with speech presented

in nurnber of terms pet 100 conversation tums (Brown & Dunn, 1991) or as a proportion

of total utterances (Moore et al., 1994).

Our data indicate that parents do not change the frequency with which they engage

in feeling talk as a proportion of their conversation tums. Feeling talk is the hast

frequent type of talk produced by parents at each time period. A previous longitudinal

study of feeling state talk indicated that the number of conversation tums in which

mothers referred to feeling States increased between the time children were 18 and 32

months of age (Dunn et al., 1987). The present results, with older participants than

Dunn et al. extend what is known about developmental changes in feeling talk up to the

time when children are six yean of age.

The present study not only supports past results, but also provides distinct

information about the changes in fathers' mental state talk over time. Fathers'

developmental changes in cognitive, desire or feeling state talk were not significantly

different from mothers'. For the cuvent study, al1 analyses were based on observation

sessions during which both parents were present. A natural extension would be to

obseive and compare the mental state talk of parents separately, with their children, to

examine their patterns of change.

Gender Oifferences in Mental State Taik

Although it was not a specifc aim of this study, we did examine the data for gender

differences between boys and girls and mothers and fathers, in their production of

mental state talk. The present resuîts provide converging evidence that gender

differences are apparent in the speech of young childm and are consistent with

previously reported findings (Cervantes & Callanan, 1998; Dunn. Bretherton & Munn,

1987; Hughes & Dunn, 1998; Keubli, Butter 8 Fivush, 1995). These previous studies

have reported that girls spoke more than boys did about feelings and cognitions. For

the present study, gender differences were found only for the younger sibling and may

represent gender differences in verbal fluency, suggesting that girls and boys may have

different baselines for mental state talk. Interestingly, at time 2, there was no significant

diffeience in the verbal fluency of boys and girls (MLU), arguing against this being the

only explmation of what was found.

That gender differences are not found in the older siblings suggests that there may

be an interaction between birth order and parental socialkation of mental state talk. It

would be important to replicate the present findings with a larger sample of children and

using a more varied measure of verbal ability.

It appears that mothers and fathers differ in their production of mental state ternis in

everyday conversations with their children. The present results are in keeping with the

suggestion that mothers use more motion labels than fathers during interactions with

children (Parke, 1994). In addition, mothers also use more cognitive and desire labels

than fathers in this study. Why this is the case needs to be explored further. All of the

fathers in this study were employed full time, hi le 29 of the mothers were involved in

full or part-time work. Perhaps by virtue of being home more often, mothen take on a

different role than fathers, resulting in more frequent conversations between mothers

and children about mental states. Fathers, on the other hand, may develop

relationships with childien that do not involve the labeling mental states. Our

knowledge of gender differences in mental state talk would benefit from a more detailed

investigation of the contexts and styles that parents use to talk about mental states.

The present study examined only feeling words or labels, and it may be that fathers

express feelings and other mental states in ways other than labeling. In addition, the

present findings should be replicated with a larger sample sire.

The Referent of Mental State Talk

A second airn of this study was to investigate possible gender differences in parents

and children's talk about 'others' (other than the child). In the present study, mothers

speak more about the mental states of people other than the target child, than do

fathen. These results are consistent with Gilligan's (1 982) views, and inconsistent with

those of other researchers who found no differences (Kuebli & Fivush, 1992).

However, several methodological differences rnay account for Our findings. First, this

study examined references to 'other' in a naturalistic context, during everyday

conversations, as opposed to Kuebli and Fivush's (1 992) more structured context.

Second, we observed the production of cognitive, desire and feeling ternis about 'other'

among families with two children and two parents simultaneously. Kuebli and Fivush

conducted their obsewations of feeling ternis with each parent separately.

The current study highlights new information about differences between mothers

and fathers talking about a variety of mental states. Female children may selectively

attend to and model the speech of their mothers, while boys may attend to their fathen.

Given differences in parental modeling of talk about people other than the child to

whom they are speaking, girls would then be more likely than boys to attend to the

mental states of people other than themselves and this is supported by gender

differences in the older siblings' talk about people other than themselves.

Our analyses did not reveal an overall gender difference in the younger siblings' talk

about 'othen', at either time point, however, younger girls did speak more than younger

boys about the cognitions and desires of othen. Results also indicate that older girls

speak more about othen' mental states than older boys. at both time points. That

gender differences are not apparent for al1 types of speech in the younger children

suggests that talk about others' mental states may develop along a different time line

for each type of talk. One must interpret these results with caution, however, as the

distribution for younger children's talk about 'othet remained positively skewed. Most

children spoke very little of others' mental states and a few spoke much more. Gender

differences may in fact be a function of the higher scores. A larger sample sire might

help to clarify this issue. The gender differences in older siblings support our prediction

and are consistent with Gilligan's (1982) perspective that girls and women show more

of a focus on the intemal states of other people. Given that talk about others shows a

clear developmental pattern, as does talk about cognitions, it is likely to be how girls are

more advanced in language than boys. This advantage is picked up in subtle ways,

such as talking about cognitions, talking about others, and the combination.

While other researchers have not reported gender differences in talk about 'other',

(Hughes & Dunn, 1998), methodological differences make a direct cornparison difficult.

Specifically, the definition of 'other' is inconsistent in the literature. Hughes and Ounn

for example, collapsed refennces to 'other' (other than child) and to 'child plus friend'

when they examined references to 'ottiei': whereas for the cunent study, 'other' did not

include shared refennces between the speaker and another person. Fuither, the

context of talk about the other in previous studies diffen from the present study.

Whereas Hughes and Dunn obsewed children with a friend. the present study involved

siblings, mothers and fathers.

The results of the this study illuminate gender differences in talk about the mental

states of others, as occun naturally in the home, an environment which has the

potential to influence perspective taking and social develo pment. It would be

worthwhile to examine the developmental trajectories of children, as they refer to the

mental states of 'others', over a longer period of time, in order to detemine the

existence of gender differences in adolescence.

The Effect of Siblings on Mental State Talk

The final aim of this study was to examine the effects of the presence of an older

sibling on mental state talk. Children with an okler sibling were found to be advantaged

in cognitive talk, but not desire or feeling talk, Previous studies (Ruffman et al., 1998)

have found an increase in false belief understanding amongst children with an older

sibling in the home. It may be that cognitive talk with an older sibling is the mechanism

involved in the association between the presence of an older sibling and false belief

understanding. When older children talk to younger children about cognitions they may

foster an understanding of the nature of cognitions and more specifically the nature of

belief,

While the present study contributes new and important information about

developmental changes in mental state talk, aiem is much to be leamed about mental

state talk beyond age six. Furthemore, we have no measures, at any time point, of

false belief understanding, perspective taking or other measures of social skills to

correlate with the frequency of mental state talk. Because our design was correlational,

we cannot make causal statements about the rote of gender or parents in talk about

mental states. Additional longitudinal studies are necessary to more fully understand

developmental changes in the mental state talk of family members.

This is the first study that we are aware of that examines, longitudinally, the naturally

occumng speech of children and both parents when children are between two and six

years old. These results add important information to what is known about changes in

mental state talk, as a function of age, gender and birth order. The present study also

represents an important first step in identifying the developmental changes in fathen'

mental state talk.

Refemnces

Astington, J. W. (1 993). The child's discoverv of the mind. Cambridge, MA: Harvard

University Press.

Astington, J. W., & Pelletier, J. (1996). The language of mind: Its role in

teaching and leaming. In D. Re Olson & Ne Tonance (Eds.), The handbook of

education and human develo~rnent (pp.593-619). Cambridge, MA: Blackwell

Publishers.

Bartsch, K., 8 Wellman, H. M. (1 995). Children talk about the mind. New York:

Oxford University Press.

Beeghly, M., Bretheflon, I., & Mervis, C. B. (1986). Mothers' intemal state language

to toddlers. British Joumal of Develoornental Psvcholoay, 4, 247-261.

Belmont, L., & Marolla, F. A. (1973). Birth order, family size, and intelligence.

Science. 182, 1096-1 101.

Bretherton, l., 8 Beeghly, M. (1 982). Talking about intemal states: The acquisition

of an explicit theory of mind. Developmental Psvcholoay, u, 906-921.

Brody, L. R. (1 985). Gender differences in emotional developrnent: A review of

theories and research. Journal of Personal@, 102-1 49.

Brown, J. RD, Oonelan-McCaIl, N., 8 Dunn, J. (1 996). Why talk about mental

states? The significance of children's conversations with friends, siblings and rnothers.

Child Develo~ment, a, 836-849.

Brown, J. R., & Dunn, J. (1 991). You can cty, m m : The social and developmental

implications of talk about intemal states. British Journal of üevelo~mental Psvcholoay,

73

Ceivantes, C. A., & Callanan, M. A. (1998). Labels and explanations in mother-

child emotion talk: Age and gender differentiation. s, M, 88-

98.

Condry, J., & Condry, S. (1976). Sex differences: A study of the eye of the

beholder. Child Develo~ment, qZ, 81 2-81 9.

Cozby, P. C. (1 997). Methods in Behavioral Research. Mountain View, CA:

Mayf ield Publishing Company.

deVilliers, J. (1 995). Questioning minds and answering machines. In D.

MacLaughlin & S. McEwen (Eds.), Proceedinas of the 19' Annual Boston Univenitv

Conference on Lanauaae Develo~ment (pp. 20-36). MA: Cascadilla Press.

Dunn, Je, Bretherton, I., & Munn, P. (1 987). Conversations about feeling states

between mothers and their young children. Oevelo~mental Psvcholoqy, a, 132-1 39.

Dunn, J., Brown, J.. 8 Beardsall, L. (1 991). Family talk about feeling states and

children's later understanding of others' emotions. Develo~mental Psvcholoay, 27,448-

455.

Furrow, D., Moore, C., Davidge, Je, & Chiasson, L. (1992). Mental ternis in mothen'

and children's speech: Similarities and relationships. Journal of Child Lanauaae, I&

61 7-631.

Gilligan, C. (1 982).

develo~ment. Cambridge, MA: Harvard University Press.

Hams, P. (1 996). Desires, beliefs and language. In P. Carruthen & P. K. Smith

(Eds.), Theories of theories of mind (pp.200-220). Cambridge, MA: Cambridge

University Press.

Hughes, CD, 8 Dunn, J. (1998). Understanding mind and motion: Longitudinal

associations with mental-state talk between young friends. Develo~mental Psvcholoay,

34, 1 026-1 037. - Jenkins, J. M., & Astington, J. W. (1 996). Cognitive factors and family structure

associated with theory of mind development in young children.

Psvcholoay, a, 70-78.

Kuebli, J., Butler, S., 8 Fivush, R. (1995). Mother-child talk about past emotions:

Relations of matemal language and child gender over time. Coanition and Emotion, 2,

265.283.

Kuebli, J., & Fivush, R. (1 992). Gender differences in parent-child conversations

about past emotions. Sex Roles, a 683698.

Lewis, C., Freeman, N. He, Kyriakidou, C., Maridaki-Kassotaki, K., & Berridge, O. M.

(1 996). Social influences on false belief access: Speclic sibling influences or general

apprenticeship? Child Develoment, 6Z, 2930-2947.

Lollis, S., Ross, H., & Leroux, Lm (1 996). An observational study of parents'

socialization of moral orientation during sibling conflicts. Merrill-Palmer Quarte* a, 475494.

Moore, C., Bryant, O*, Funow, O. (1989). Mental tens and the development of

certainty . Child Develoomen~ 1 67-1 7 1 . Moore, Cm, Çunow, D., Chiasson, L., & Patriquin, M. (1 994). Developtnental

relationships between production and comprehension of mental ternis. First Lanauaae,

14,l-017. - Parke, R. D. (1 994). Pmgress, paradigrns, and unresolved problems: A

commentaiy on ment advances in Our understanding of childm's ernotions. Memll-

Palmer Quarterly, 4ût1 S M 69.

Perner, J. (1 991).

Press.

75

Understandina the re~resentational mind. Cambridge, MA: MIT

Pemer, J., Ruffman, T., & Leekam, S. R. (1994). Theory of mind is contagious: You

catch it from your sibs. Child Develo~ment, ê5, 1228-1 238.

Ruffman, T., Pemer, J., Naito, M., Parkin, L., & Clernents, W. A. (1 998). Older (but

not younger) siblings facilitate false belief understanding. Develo~mental Psvcholoay,

34, 161 -174. - Shatz, M.. & Gelman, R. (1973). The development of communication skills:

Modifications in the speech of young children as a function of listener. Monoara~hs of

the Society for Research in Child Developrnent, a ( 5 , Serial No. 152).

Shatz, M., Wellman, H. M., & Silber, S. (1983). The acquisition of mental verbs: A

systematic investigation of the first reference to mental state. Coanition, u, 301-321.

Tannen, D. (1 990). Gender differences in topical coherence: Creating involvement

in best f riends' talk. Discourse Processes, s, 73-90.

Zajonc, R. B., & Markus, G. B. (1975). Birth order and intellectual development.

Psvcholoaical Review, &, 74-88.