Content uploaded by Adam Winsler
Author content
All content in this area was uploaded by Adam Winsler
Content may be subject to copyright.
J. Child Lang. (), –. Printed in the United Kingdom
# Cambridge University Press
Age-related changes in preschool children’s
systematic use of private speech in a natural setting*
ADAM WINSLER
George Mason University
MARTHA P. CARLTON
Southern Illinois University at Edwardsville
MARYANN J. BARRY
University of Alabama
(Received March . Revised January )
This study set out to explore the contexts in which preschool children
use private speech, or self-talk, in the naturalistic setting of the
preschool classroom, and age-related changes in the contexts in which
preschoolers talk to themselves. A total of naturalistic observations
of fourteen three-year-old and fourteen four-year-old children were
conducted using a time-sampling procedure in two preschool classrooms
over the course of one semester. Results from logistic regression analyses
revealed that both age groups were (a) more likely to use private speech
during the self-selected activity classroom context as opposed to both
large group and outside free play classroom contexts, and (b) most likely
to talk to themselves when alone, next likely in the presence of peers, and
least likely when in the presence of a teacher. Although the probability
of private speech among three-year-old children did not vary as a
function of the child’s immediate activity, four-year-old children’s
private speech was more likely to occur during sustained and focused
goal-directed activity as opposed to rapidly-changing and non goal-
[*] This research was presented at the annual conference of the American Educational
Research Association, Chicago, IL – March, . This project was supported by faculty
research grants from the College of Education and the University of Alabama. The
authors would like to thank the children and the staff of the Child Development Center,
as well as Susanne Denham, Jose Cortina, and Ann Kruger for their helpful comments on
the manuscript. Address for correspondence : Adam Winsler, Department of Psychology
– F, George Mason University, Fairfax, VA. , USA. e-mail : awinsler!gmu.edu
,
directed activity. The findings suggest that private speech appears
systematically in young children and that, in several ways, four-year-old
children use private speech more selectively than three-year-olds.
Two of the central tenets of Vygotskian sociocultural developmental theory
are that higher-order human psychological functions have social origins and
that children’s cognitive and social development are largely the result of the
internalization or constructive appropriation of sociocultural tools and signs
during joint activities with others (Vygotsky, –\ ; Wertsch,
; Vygotsky & Luria, \). Arguably the most important of the
cultural tools or symbol systems available to the child for internalization is
language (Luria, ; Nelson, ). Vygotsky placed great importance on
the role of language in development and posited that a crucial and formative
period in ontogenesis occurs in early childhood when children begin to use
language not only for communication with others but also as a tool for
thought and the self-regulation of behaviour (Vygotsky, \ ; Berk &
Winsler, ). With the internalization of language, cognitive processes are
restructured and reorganized, the mind becomes mediated by language, and
uniquely human, higher-order psychological functioning becomes possible
(Vygotsky, -\; \; Diaz, Neal & Amaya-Williams,
).
Young children’s private speech, or overt talk to the self, naturally
becomes a focal point for research within this perspective as central
hypotheses concerning the social origins of mental processes and the transfer
from other-regulation to self-regulation can be addressed through ex-
ploration of such speech. Private speech is typically seen by researchers as
reflecting an intermediate phase in the ontogenesis and microgenesis of
speech internalization, wherein language moves from overt communication
with others, to overt communication with the self, and finally to inner speech
or covert verbal thought (Diaz et al., ; Berk, ; Winsler, Diaz, &
Montero, ). Using private speech as a methodological window for
exploring the socially- and linguistically-mediated nature of cognition and
behaviour, research to date has explored such important issues as (a) the role
of private speech as a potential mediating link between adult–child teaching\
learning interactions and children’s task improvement and cognitive de-
velopment (Berk & Spuhl, ; Winsler et al., ), (b) relations between
adult scaffolding and children’s private speech and task performance
(Behrend, Rosengren & Perlmutter, ; Berk & Spuhl, ), (c) the
dynamic relations between private speech, task difficulty, and task per-
formance and improvement over time (Frauenglass & Diaz, ; Winsler et
al., ), and (d) the role of private speech as a tool for behavioural
regulation among children with problems of self-regulation, attention deficit
hyperactivity disorder, and impulsivity (Berk & Potts, ; Winsler, ).
The present study explores two features of preschool children’s private
speech in the naturalistic setting of the preschool classroom: () systematic
contextual variation in the occurrence of private speech, and () age-related
changes in the contexts in which preschool children talk to themselves.
Vygotsky (\) was the first to note that children’s private speech,
rather than being an epiphenomenal event, appeared systematically in
particular contexts. He noted that young children were especially likely to
talk to themselves during moments of task difficulty in problem-solving
activity. Also, he noted early on that children’s private speech use was related
to the presence of others, with young children being more likely to talk to
themselves when others were somewhere nearby than when they were alone.
Since then, other investigators have explored systematically various con-
textual variables associated with children’s private speech, including not only
the presence of others, but also the degree of behaviour regulation provided
by others present, as well as the type of task, task difficulty, and classroom
context. Diaz (), in a review of the research examining contextual
variation in children’s private speech usage, concluded that children’s private
speech is maximized under situations when there is a need for executive
control and there is a relative absence of regulation provided by others. For
example, children have been found to be more likely to engage in private
speech (a) when children are engaged in goal-directed, academic, or problem-
solving activities compared to free play or other activities (Berk & Garvin,
; Winsler & Diaz, ), (b) when children’s problem-solving task is
challenging but attainable as opposed to easy (Berk & Garvin, ;
Vygotsky, \; Behrend et al., ), (c) when they are either alone
or with peers as opposed to in the presence of an adult who is regulating the
child’s behaviour (Rubin, Hultsch, & Peters, ; Berk & Garvin, ;
Diaz, ; Winsler & Diaz, ), and (d) when they are in collaboration
with an adult who is appropriately ‘ scaffolding’ their problem-solving
activity, compared to an adult who provides high degrees of external control
and direction (Goudena, ; Diaz, Neal & Vachio, ; Behrend et al.,
; Berk & Spuhl, ).
In terms of age-related changes in private speech, Vygotsky (–\
; \) hypothesized that children’s private speech changes
ontogenetically in at least three ways. First, that the frequency of children’s
overt private speech follows an inverted-U relation with age, peaking during
the preschool years and decreasing in frequency as children enter the
elementary school years. Research has generally confirmed the existence of
this broad developmental pattern as decreases in children’s overt private
speech toward the end of the preschool years are accompanied by cor-
responding increases in the frequency of partially-internalized manifestations
,
of inner speech, such as whispers and inaudible muttering (Berk & Garvin,
; Bivens & Berk, ). However, three qualifications are in order
concerning this finding : (a) this ontogenetic pattern in frequency is often
observed only among certain subtypes of children’s private speech rather
than children’s overall self-talk, (b) that age-related changes in children’s
private speech use in naturalistic classroom settings have been shown to be
more extended and gradual than those found in laboratory studies, and (c)
that similar curvilinear trends in private speech usage repeat themselves
microgenetically as children of different ages master new challenging tasks
(Frauenglass & Diaz, ; Berk, ; Winsler et al., ).
A second hypothesized developmental change in private speech originating
from Vygotsky’s work is that over the course of the preschool years,
children’s self-talk makes a gradual transition from being less, to more, self-
regulatory in nature. Luria’s research (), for example, demonstrated that
the speech of four-year-old children is more effective in guiding behaviour
than the speech of three-year-olds. Subsequent research exploring the extent
to which the various functional or semantic categories of children’s private
speech (i.e. word play, comments about the self, descriptions of the
environment) change with age, however, has not shown large differences in
the content of private speech produced by children of different ages (Rubin,
). Other studies that have investigated the related issue of whether the
timing of children’s private speech in relation to their ongoing behaviour (i.e.
whether speech first follows, then accompanies, then precedes children’s
actions) as children get older have also provided only mixed support at best
(Rubin, ; Pellegrini, ; Berk, ).
Finally, Vygotsky suggested that over the course of the preschool years,
private speech becomes more and more differentiated from social speech as
it is gradually internalized and integrated with children’s covert mental
activity. Research exploring the relations between social and private speech
and changes in both types of speech over time with young children has
generally supported this claim (Berk & Garvin, ; Furrow, ; Berk,
).
Several limitations and gaps in the above research literature on children’s
private speech are notable. First of all, although researchers have explored
age-related changes in private speech and studied contextual variation in
children’s private speech usage, these two areas of inquiry have occurred in
relative isolation. That is, age-related changes in the contexts in which
children talk to themselves has not been explored. This area is worthy of
study because age-related developments in children’s private speech use
across the preschool years are also clearly accompanied by age-related
changes in the activities in which children engage, and in the immediate
social and behavioural context in which children find themselves. The social
affiliations, sociocognitive play behaviours, and task-related activities of
four- and five-year-old children are likely different from those of three-year-
olds. Older preschoolers are more likely, for example, to sustain their task
behaviour for longer periods of time, demonstrate more complex play, and
interact cooperatively with others than younger preschool children (Johnson
& Ershler, ). It is important empirically to be able to distinguish
between age effects and context effects in understanding children’s private
speech.
A second critical limitation of the research to date in this area is that the
vast majority of studies on contextual variation in children’s private speech
have been conducted in laboratory settings (i.e. Frauenglass & Diaz, ;
Behrend et al., ) or in quasi-laboratory settings such as a separate private
room of a preschool (i.e. Rubin & Dyck, ). The few naturalistic
investigations of children’s private speech that do exist have typically been
conducted with either toddlers age two or under in the home (Nelson, ;
Smolucha, ) or with elementary school age children at school (Roberts
& Tharp, ; Berk & Garvin, ; Bivens & Berk, ; Winsler & Diaz,
). Such lack of information about preschool age children’s use of private
speech in naturalistic settings is most unfortunate given that the preschool
years are those in which much developmental change in private speech is
hypothesized to occur. Given that private speech is known to be quite
sensitive to the setting, and that speech use in the lab may not generalize well
to natural settings (Berk & Landau, ), it appears crucial to collect
naturalistic observational data on children’s self-verbalizations in settings
such as the preschool classroom. Validation of laboratory findings in natural,
real-world settings is critical to progress in the discipline of applied
developmental science (Fisher & Lerner, ).
The present study
If private speech serves important self-regulating functions for children in
early childhood, then the appearance of such speech should be predictable in
terms of the child’s activity and the surrounding context, and this systematic
variation in speech use should be observable in the early childhood classroom
setting. For example, private speech should be more likely to occur during
sustained, focused, goal-directed activities than during unfocused, non goal-
directed activities. Similarly, if children’s private speech becomes more self-
guiding, more internalized, and\or more related to the child’s internal goals
and activities over the course of the preschool years, then four-year-old
children’s self-speech would be expected to occur more systematically across
different natural contexts compared to the speech of three-year-olds. That is,
whereas the private speech of three-year-olds might be expected to be more
evenly distributed across classroom contexts and across children’s activities,
four-year-old private speech may be much more likely to appear sys-
tematically during certain types of activities or contexts.
,
This study contributes to the literature in several ways. First, it provides
much needed naturalistic data on preschool children’s use of private speech
in the preschool classroom. Second, it examines a new dimension of
developmental change in children’s private speech, namely, age-related
changes in the contexts in which children talk to themselves. Third, the
present study introduces two advantageous methodological features to the
private speech literature: a) the use of dichotomous coding of time-sampled
observations on the basis of whether or not private speech occurred (as
opposed to counting number of private speech utterances) which eliminates
the problem, common in the private speech literature (Diaz, ), of large
individual differences in children’s quantity of speech, and b) the use of
multiple logistic regression with partial parameter estimates, which examines
the unique and combined influences of several contextual variables on private
speech simultaneously, including variables which are likely to change with
age, such as the type and duration of children’s activity, and children’s
immediate social context. This study, therefore, has at least four major goals :
() to examine the degree to which preschool children spontaneously use
private speech in the naturalistic context of the early childhood classroom, ()
to explore the extent to which private speech use in preschool classrooms
varies as a function of contextual variables, () to see the degree to which
findings from earlier naturalistic studies with older children (i.e. Winsler &
Diaz, ) replicate and generalize to younger children, and () to investigate
age-related shifts from age three to four in the systematicity of children’s
private speech usage.
A total of naturalistic observations were conducted of three-year-
old children and four-year-old children in two preschool classrooms.
Measures included (a) the classroom context in which the observation took
place (i.e. large group activity period, self-selected activity period, or outside
time on the playground), (b) children’s social context (alone, with peer(s),
teacher present), (c) the goal-directed nature of children’s activity, and (d)
whether or not children’s activity was sustained over time. If private speech
is a tool used systematically by young children for behavioural self-
regulation, and if developmental change takes place between the ages of three
and four in the self-regulatory function of private speech, then the following
predictions can be offered based on the previous literature: () Private speech
will be more likely to occur during self-selected activity periods as opposed
to outside on the playground or large group activities, () Children will be
more likely to use private speech while alone, next likely while with peers,
and least likely in the presence of an adult, () Private speech will more likely
appear when children are engaged in goal-directed activities as opposed to
non goal-directed activities, and () Children will be more likely to use
private speech in the context of sustained activity rather than rapidly-
changing activities, and this effect will be stronger when the child’s activity
is goal-directed rather than not. Finally, it was hypothesized () that each of
the above effects would be stronger among the four-year-old children than
among the three-year-old children.
Participants
Participants included preschool children attending a university-affiliated
laboratory preschool in a medium-sized city in the Southeastern United
States. Half of the sample (N l , % male) consisted of the children
enrolled in the ‘three-year-old room’ at the preschool. The ages in this group
ranged from ; to ; (M l n mos., .. l n) at the beginning of the
spring semester when data collection began. The remainder of the sample (N
l , % male) included children who were enrolled in the ‘four-year-
old room,’ with ages in this group ranging from ; to ; (M l n mos.,
.. l n). The ethnic breakdown of the sample according to parental report
on the preschool application\registration forms was % Caucasian, %
African-American, and % Asian-American. A reasonable range of family
socioeconomic levels was present in the sample (Hollingshead index - Range
l –, M l n, .. l n) as the preschool stratified its enrollment in
the classrooms into three equal thirds : a) children of university faculty\staff,
b) children of university students, and c) children of community members.
Paternal age ranged from to years (M l n, .. l n) and
maternal age ranged from to years (M l n, .. l n). Fathers’
years of education ranged from to (M l n, .. l n) ) and
mothers’ education ranged from to years (M l n, .. l n).
None of these demographic variables varied significantly by children’s age
group.
Setting. The preschool program attended by the children was a NAECP-
accredited, five day a week, morning ( : am– : pm) program, con-
sisting of one three-year-old class and one four-year-old class, with each class
typically including – children. Both classrooms were headed by one lead
teacher and one graduate student assistant (all female). Due to the centre’s
responsibilities as a training and observation site for human development
students, the classrooms also typically included one to five other adult
students who would periodically observe or assist in the classroom. The two
classrooms shared the same teaching philosophy and had similar daily
schedules which, other than the occasional special activity (such as a walk to
a park or a field trip), reliably consisted of certain times each day set aside for
three different types of activities : ) self-selected activities, in which children
would choose which of several activity centers (i.e. block area, house corner,
Lego2 table, dress-up area …) they would go to to play, ) outside play time,
,
and ) large group activities, such as reading or singing together as a group.
Written literature from the center, impressions from the observations, and
interviews with the teachers and the director indicated that the preschool
program holds a ‘child-centred’ or constructivist early childhood edu-
cation philosophy. Within this perspective, children are allowed and en-
couraged to independently explore their environment and design their own
learning and play activities within the boundaries of the set of develop-
mentally-appropriate materials that have been set out by the teacher. The
role of the teacher from this perspective was described by the teachers as
‘facilitator’ – one who chooses the learning materials made available to the
children each day, lets children play and solve problems independently as
much as possible, and one who generally does not get directly involved with
the children’s activities, intervening only when absolutely necessary.
Procedure
Observations. Naturalistic observations were carried out in the two classrooms
over a -week period during the spring semester of the academic year. Two
female research assistants (one per classroom) observed children according to
a predetermined random order using a time-sampling method. Observations
began after a three-week introductory period during which time children
grew accustomed to the presence of the observers in the classroom, observers
were trained, the observational checklist instrument was pilot tested, and the
reliability of the observational checklist instrument was established. Observer
influence effects were minimized in this study by both the presence of the
three-week, rapport-building period and by the fact that children in these
classrooms were generally quite accustomed to the presence of observers in
the classroom given the center’s laboratory responsibilities. By design,
observations took place during the three regularly-scheduled daily activity
periods (discussed above) in the two classrooms: large group (LG), self-
selected activity time (SSA), and outside during recess (OUT). Each child
was observed an average of times yielding a total of observations.
The resulting distribution of observations by classroom context was
observations during SSA (% of the observations), during OUT
(%), and during LG ( %). This distribution of observations
represents well the relative proportion of time children in these classrooms
spend each day in each context.
Observers, who were unaware of the research hypotheses, used a
behavioural observation checklist instrument to record their observations.
Also, to assist in adhering to the time-sampling observation schedule,
observers listened to pre-recorded time signals which projected via head-
phones to one ear from an audiocassette recorder attached to their belt.
Target children were observed according to a predetermined random order
for approximately , -minute periods, with each period consisting of ,
-second direct observation intervals separated by -second recording
intervals. Thus, an observer would observe a target child for seconds, at
which time the audio signal would sound and she would then record her
observations for that second period on the checklist instrument for the
remainder of the minute. Then the audio signal would sound again indicating
that it was time to observe the child again for the second, ten-second
observation. This pattern would continue for ten observations at which time
the observer would go on to conduct a series of ten observations on the next
child on the list. Predominant activity sampling (PAS; Hutt & Hutt, )
was used for the social context and activity variables, meaning that if more
than one type of behavioural class occurred within a -second observation
interval, observers coded the occurrence of only the predominant behaviour
that was present for the larger time period during the observation.
Measures. Four variables were coded on the observational checklist
instrument. First, children’s speech during each -second observation was
coded as either containing social speech, private speech, or no speech. Social
speech was defined as any verbalization intended for communication to
another individual as indicated by the presence of either a) a name or
pronoun reference, b) a gaze at another person during or within one second
of the utterance, c) an intentional physical touch of another person, or d) a
conversation or verbal turn-taking episode. Any verbalization by the child
which did not contain one of these social markers, including inaudible
muttering and silent verbal lip movements, was taken as evidence of private
speech. A dichotomous variable was later constructed which represented
simply whether or not private speech was present during each observation.
Children’s activity was coded as being either explicitly goal-directed or
non goal-directed. Goal-directed activity was defined as behaviour by the
child which appeared focused, organized, and had an identifiable goal or end
point to the activity. The goal being pursued by the child could either be self-
formulated or teacher-provided. Examples of goal-directed activity in the
context of self-selected activity periods (SSA) included, for example, building
a structure out of Legos2 or some other assembly\construction materials,
doing a puzzle, playing a game with rules, or engaging in an organized make-
believe episode of ‘ house.’ Not explicitly goal-directed behaviour in this
context included, for example, aimless wandering around the classroom,
looking on into another group’s activity, repeatedly spinning a puzzle piece
around one’s finger for the apparent ‘ fun of it,’ and making a transition
between one activity and another.
Children’s sustained activity was coded by assessing the relationship
between the child’s activity during the current ten-second observation
interval and the child’s activity during the previous observation period, one
minute ago. For the second through tenth observation in each series of
,
observations on target children, the observer coded whether or not the
activity the target child was engaged in during the current observation was
the same (in terms of goal, materials, and behaviour) as that in which the
child was engaged during the previous observation one minute ago.
Finally, children’s immediate social context was also coded. Observers
noted, for each second interval, whether the child was alone, with one or
more peers, with a combination of one or more peers and a teacher, or one-
on-one with a teacher. Children were coded as being alone if no other person
doing the same general activity was within three feet of the target child and
there were no social interchanges with another person during the observation.
Children were coded as being with a peer if there were one or more other
children present who were either doing the same activity in parallel with the
target child within three feet or who were physically or verbally interacting
with the target child. Children were coded as being with both peer(s) and a
teacher if any adult was included as one of the members of a group, using the
same criteria as those used above for ‘peer.’ Children were classified as being
exclusively with a teacher if they were interacting one-on-one with a teacher
with no other children present within three feet of the target child. Because
no instances of one-on-one direct interaction with a teacher occurred during
the observations which took place in the four-year-old classroom, the
last two categories were combined for the analyses to make one category
indicating teacher presence, broadly defined.
Reliability. Inter-rater reliability for the above coding systems was de-
termined during the last phases of observer training, at which time two
observers independently rated the same children for , second ob-
servations. Reliability was acceptable to good for all category systems.
Percentage agreement across observers was % for children’s activity
(kappa l n), % for sustained activity (kappa l n), % for speech
(kappa l n), and % for social context (kappa l n).
The analysis strategy chosen for addressing the hypotheses was to fit two
appropriate and manageable logistic regression models to the data – each
with private speech (yes\no) as the dependent variable, subject nested within
class (df l ) as a repeated measures effect entered first into the model, and
the relevant contextual variables and interaction terms entered as predictors.
The first model tested hypotheses one and two concerning the influence of
the global contextual variables (classroom context and social context) on the
probability of private speech occurring. The second model tested hypotheses
three and four exploring the influence of specific activity variables (activity
and sustained activity) on the probability of private speech. Significant
effects from these models were interpreted by examining and calculating
odds ratios from the relevant two- or three-way contingency tables (Hosmer
& Lemeshow, ).
One advantage of using logistic regression is it can control for the effects
of other variables within the model and provide partial parameter estimates
for each variable of interest, much like ordinary least squares multiple
regression with continuous predictors of a continuous outcome variable.
Obtaining partial parameter estimates is especially crucial for this type of
naturalistic research as the researcher does not have control over many of the
variables, and indeed many of the predictor variables are not independent
from one another. For example, in this study, age and social context are
related in that the three-year-old children spent more time interacting
directly with their teachers than did four-year-olds. Similarly, classroom
context is not unrelated to social context. Children were understandably
more likely to be in the presence of a teacher during large group activities
than during outside recess time. The logistic regression analyses reported
below are able to tease out the individual and combined contributions of each
relevant variable to the outcome variable of interest, namely private speech.
Classroom context and social context. The first model, which predicted the
log likelihood of private speech on the basis of subject (nested within class),
age (\), classroom context (LG\SSA\OUT), social context (Alone\Peer\
Teacher Present), and the interaction terms for age-by-classroom context
and age-by-social context, was significant [χ
#
(, ) l n, p n].
Likelihood ratio effect tests indicated that each of the variables in the model
contributed significantly to the prediction of private speech [subject within
class χ
#
() l n, p n, class χ
#
() l n, p n, classroom
context χ
#
() l n, p n, social context χ
#
() l n, p n,
age-by-classroom context χ
#
() l n, p n], and age-by-social con-
text χ
#
() l n, p n]. The significant subject effect indicates that
children vary considerably in their base rate frequency of private speech
usage. Other significant effects while the subject variable is present in the
model indicate that the effect is still significant after controlling for the
subject variance.
Table provides the odds ratios for each of the pairwise contrasts within
classroom context and social context, separately by age group and for all
children combined. The odds ratios describe how much more likely private
speech is to occur in one condition compared to another. For example, the
n figure given in Table referring to the comparison of SSA over LG for
four-year-olds indicates that four-year-old children were seven and a half
times more likely to talk to themselves during self-selected activity periods
than during large group activities. Overall, three-year-olds were n times
more likely to talk to themselves than four-year-olds, but as indicated by the
significant age-by-classroom context interaction, this was not true across all
classroom contexts. Four-year-olds were just as likely to talk to themselves
,
. Raw odds ratios (pairwise increase in likelihood that private speech
will occur) for classroom context and social context, combined and by age
Three-year-olds Four-year-olds Combined
Classroom context
SSA over OUT n n n
SSA over LG n . n
OUT over LG n . n
Social context
Alone over n n n
Peers(s)
Alone over n n n
Teacher presence
Peer(s) over n n n
Teacher presence
100
75
50
25
0
3-year olds 4-year olds
Percentage of children’s private speech
Large group (LG)
Self-selected (SSA)
Outside recess (OUT)
Fig. . Distribution of children’s private speech across each of the three contexts (SSA, LG,
OUT), by age group.
as three-year-olds during the self-selected activity context but less likely to
use private speech outside or during large group activities. Figure visually
displays the age-by-classroom context interaction in terms of the overall
proportion of children’s private speech which occurred in each of the three
classroom contexts. Hypotheses and (as they pertain to classroom
context) were, therefore, strongly supported by the data. Children’s private
speech use appears systematically across different classroom contexts and
older preschoolers are more systematic in their spontaneous use of self-talk
than are younger preschoolers.
Regarding social context, children overall were most likely to talk to
themselves if they were alone, next likely in the presence of their peers, and
least likely to use private speech in the presence of an adult. However, this
social context effect also interacted with child age as shown in Figure ,
80
60
40
20
0
3-year olds 4-year olds
Percentage of children’s private speech
Alone
With peer(s)
Teacher present
Fig. . Percentage of observations in which children used private speech in each of the three
social contexts (alone, with peer(s), teacher present), by age group.
which plots the percentage of observations in each of the three social contexts
where children were using private speech, by age group. Interestingly,
however, the direction of the interaction is opposite to that which was
originally hypothesized. Three-year-old children’s private speech use was
much more associated with their immediate social context than was the
private speech of four-year-olds. Thus, although hypothesis was clearly
supported by the data, the social context component of hypothesis , which
predicted that the social context effect would be stronger for four-year-olds,
was not supported.
Children’s activity and sustained activity. The second logistic regression
model exploring the relationship between children’s activity and private
speech was limited to only those observations which took place during the
SSA classroom context. This decision was justified given that (a) the majority
(%) of children’s private speech occurred in the self-selected activity
(SSA) classroom context, (b) the largest number of observations (N l ,
%) took place during SSA, (c) children’s goal-directed behaviour and
private speech likely have different interpretations in the three different
classroom contexts, (d) the SSA classroom context is the setting with the
most theoretical and pedagogical relevance, and (e) the desire to control for
,
other variables to aid in the interpretability of the regression models. This
overall model, which included private speech (yes\no) as the dependent
variable, subject (nested within class), age (\), activity (goal-directed\not),
sustained activity (yes\no), the two-way interaction terms for age-by-
activity, age-by-sustained activity, and activity-by-sustained activity, and
the three-way, age-by-activity-by-sustained activity interaction, was signifi-
cant [χ
#
(, ) l n, p n]. Likelihood ratio effect tests indicated
that the following were significant predictors : subject [χ
#
() l n,
p n], sustained activity [χ
#
() l n, p n], age-by-sustained ac-
tivity [χ
#
() l n, p n] and the three-way interaction [χ
#
() l n,
p n]. Table provides the raw odds ratios for these two variables,
. Raw odds ratios (pairwise increase in likelihood that private speech
will occur) for activity and sustained activity, combined and by age
Three-year-olds Four-year-olds Combined
Activity
On-task over n n n
Off-task
Sustained activity
Sustained task n n n
over New task
separately by age and as a combined group. Figure shows the percentage of
private speech occurrences which took place during goal-directed activity as
opposed to unfocused behaviour. Figure displays the distribution of
children’s private speech as a function of whether or not the child was
engaged in a sustained or a new activity.
As is clear from the graphs, three-year-old private speech is almost equally
likely to appear during goal-directed and non goal-directed behaviour
whereas four-year-old private speech is much more likely to appear when
children are engaged in an explicitly goal-directed activity. Similarly, while
private speech is almost equally likely to appear during sustained or rapidly-
changing activities for the three-year-olds, four-year-old self-talk is much
more likely to occur during sustained activities. The three-way interaction
consisted of the following. Children were more likely to talk to themselves
during sustained activities than during rapidly-changing activities, but this
was only the case when children’s sustained activity was goal-directed in
nature and not true when children’s behaviour was unfocused. The above
pattern was stronger for the four-year-old children than for the three-year-
olds. Thus, hypotheses three, four, and the part of hypothesis five concerning
the relationship between children’s private speech and their sustained and
focused behaviour, were supported by the data.
80
60
40
20
0
3-year olds 4-year olds
Percentage of children’s private speech
Goal-directed
Non Goal-directed
100
Fig. . Distribution of children’s private speech across the two types of children’s activity
(goal-directed, non goal-directed), by age group.
80
60
40
20
0
3-year olds 4-year olds
Percentage of children’s private speech
Same activity
New activity
100
10
30
50
70
90
Fig. . Distribution of children’s private speech across the two temporal dimensions (same
activity, new activity), by age group.
This investigation set out to explore, in the natural setting of the preschool
classroom, contextual variation in children’s use of private speech and age-
related changes in children’s systematic use of self-talk. The results of this
study provide support for neo-Vygotskian developmental theory regarding
,
private speech and the development of self-regulation in the preschool years
by showing that private speech, rather than being epiphenomenal, is used
systematically by young children depending on their activity, immediate
social context, and classroom context. Overall, percent of the classroom
observations contained private speech. This percentage is consistent with, if
not slightly higher than, what has been observed with older, kindergarten-
age children in classrooms (Winsler & Diaz, ), and demonstrates that
three to four-year-old children do spontaneously use self-directed language
in the naturalistic context of the preschool classroom.
Several of the findings of this study suggest that an important transition
may occur between the ages of three and four in the way in which preschool
children spontaneously use private speech, with four-year-old children’s
private speech occurring more systematically (but still spontaneously) as a
function of the classroom context and the type and duration of their activity,
than the self speech of three-year-olds. Most four-year-old children’s private
speech occurred during the tacitly structured, self-selected activity classroom
setting, whereas three-year-old children’s self-talk was more evenly distri-
buted across the three classroom contexts. Most four-year-old private speech
occurred during focused, goal-directed activity whereas three-year-old pri-
vate speech was more equally likely to appear during either goal-directed or
unfocused activities. The majority of four-year-old children’s private speech
occurred in the context of sustained activity, whereas three-year-old private
speech was more equally likely to occur during either sustained or rapidly-
changing activities.
One interpretation of such age-related differences in the systematicity with
which preschool children use private speech is that such speech plays an
increasingly self-regulatory role as children move from three to four years of
age. That is, four-year-old children, more than three-year-olds, may be using
private speech in situations in which it fruitfully serves self-regulatory
functions. Although increased predictability in the occurrence of private
speech clearly does not necessarily imply that such speech is changing
developmentally in its function or regulatory significance, it is certainly one
indication consistent with such an hypothesis. Change in the breadth of
situations for which children employ particular problem-solving strategies
(either in terms of generalization or increased selectivity) is one of several
important change dimensions in developmental inquiry (Siegler, ). The
data reported here suggest only that four-year-olds spontaneously use private
speech more selectively than three-year-olds – not that such children do so in
a conscious, deliberate manner. It is not at all clear (and indeed unlikely) that
four-year-old children are actively choosing to use their self-speech in these
ways. An important area for future study, in fact, would be to determine the
extent to which such variation in children’s verbal self-regulatory strategy
use in the classroom represents conscious choices on the part of the child.
What roles do the child’s conscious awareness and understanding of private
speech play in their private speech usage and in this movement toward
increased systematicity? Is children’s increased systematicity of private
speech dependent upon increased awareness of the potentially self-guiding
nature of private speech, or does children’s unwitting use of private speech
eventually lead to greater metacognitive awareness? Goldin-Meadow ()
has demonstrated that young children’s use of gesture during cognitive
activity appears without children’s awareness, and that such subconscious
gesture use predicts later changes in children’s thinking and strategy use. A
similar pattern might be found to be the case for private speech. Flavell,
Green, Flavell & Grossman () concluded that four-year-olds show
relatively little understanding or recognition of inner speech, or covert verbal
thought, as a possible mental activity existing within themselves or within the
minds of others. However, no empirical information exists at present about
young children’s awareness and understanding of overt private speech.
The theory of mind literature may also provide some insight into
understanding the observed age-related changes in the situations in which
children talk to themselves. Among the most well-documented developments
known to take place during the preschool years is that four-year-olds,
compared to three-year-olds, have notably increased understanding and
awareness of their own thinking and the thinking of others (Astington, ).
By the age of four, children demonstrate a clear understanding of their self
and others as intentional agents and they can be said to possess a represen-
tational theory of the mind (Perner, ; Flavell & Miller, ). Changes
in the contexts in which children talk to themselves regularly may be related
to these developments in social cognition and theory of mind which are also
known to take place between the ages of three and four.
Children’s language development and private speech appear to be
intricately involved with children’s theory of mind development from at least
two perspectives. The first perspective sees children’s theory of mind
development as having social origins, emerging as a result of the child’s social
history interacting with caregivers and from the caregiver and child’s use of
mental state language during joint dialogue (Dunn, Brown, Slomkowski,
Tesla & Youngblade, ; Astington, ; Nelson, ). Parents who
enjoy secure attachment relations with their children, who treat the child as
an independent mental being with their own perspective, who provide rich
language input, and who use much mental state language tend to have
children who are more advanced in their theory of mind or mentalizing
abilities (Jenkins & Astington, ; Meins & Fernyhough, ). From this
perspective, language is seen as the main mediating tool children use to
internalize their understanding of themselves and others as mental agents.
Private speech has recently been implicated as playing a role in assisting
young children with understanding the self and distinguishing the self from
,
others as a personal, active, mental agent (Fernyhough & Russell, ).
Also, increased use of self-regulatory private speech appears to be positively
associated with more advanced mentalizing abilities (Fernyhough, ).
Thus, private speech might be intimately related, if not formative, in the
development of children’s theory of mind.
A complementary perspective is that of Tomasello, Kruger & Ratner
(), who argue that a representational understanding of the self and other
as separate intentional agents is an important social-cognitive prerequisite
which needs to be in place before children are able to fully internalize social
dialogue (i.e. use private speech) and benefit from cultural learning
experiences. Implicit in this formulation is the suggestion that certain
properties of children’s private speech may be expected to change after a
representational theory of mind is established. Qualitative changes in young
children’s understanding of social cognition and mind may lead to changes in
the way in which private speech is used by four-year-old children. For
example, it is possible (although sheer speculation as the research has not yet
been conducted) that increased understanding of the self as an active
intentional agent and increased understanding of mental states such as
beliefs, thoughts, and desires, lead children to be more aware of their own
speech and\or use more intentional, planful, and self-guiding forms of self-
talk. One important question for future research, therefore, concerns the
timing of these developments in ontogenesis – do changes in children’s
private speech precede, accompany, or follow changes in mentalizing
abilities?
Another finding of this study was that the probability of children’s private
speech was found to vary as a function of children’s immediate social context.
As hypothesized, both age groups were most likely to talk to themselves when
alone, next likely to use such speech in the presence of peers, and least likely
when in the presence of a teacher. Although the presence of an adult does not
always imply reduced private speech among children (it depends on the
extent to which the adult is regulating the child’s behaviour [Behrend et al.,
]), it did in the present study. This is likely due to the fact that teacher
presence in these particular classrooms mostly covaried with teacher in-
tervention or regulation of child behaviour. The social context effect was
observed to be significantly stronger for the younger as opposed to the older
age group, which is contrary to the original age-by-social context hypothesis
which predicted that the social context effect would be stronger for the four-
year-olds. The original rationale for this hypothesis was the same as that for
the other interaction with age predictions, namely, that four-year-old speech
across the board will be more systematically related to contextual variables
than the speech of three-year-olds. It appears from this investigation,
however, that the emergence of preschool children’s self-speech does become
more selective with age, but only in relation to the child’s own activities and
goals. Whereas three-year-old children’s private speech is more predictable
on the basis of the child’s social context, four-year-old self-speech is more
predictable on the basis of the child’s own activities, goals, and the
affordances of the classroom context. Such findings, upon further exam-
ination, are not inconsistent with neo-Vygotskian notions of the ontogenetic
pathway for private speech, as the theory sees children’s private speech as
originally branching off from social speech and gradually becoming more and
more integrated with the child’s own cognitive activity and goals (Berk &
Winsler, ). The finding here that social context becomes less important
a predictor of children’s private speech usage with age during the early
childhood years is not only theoretically interesting but it also helps explain
some of the mixed results of previous studies that have explored the influence
of the immediate social context on children’s private speech use (for a review,
see Berk, ). For example, Winsler & Diaz (), who used a similar
methodology and observed five-year-old kindergarten children’s private
speech in classroom situations, did not find the probability of children’s
private speech use to be significantly related to the presence of others.
The numerous limitations of this study are worthy of discussion and they
suggest that replication and extension of these findings would be desirable.
First, this study’s cross-sectional design does not allow for claims to be made
about developmental change in children’s private speech over time. Clearly,
a desirable next step would be a short term longitudinal study addressing
similar questions. Second, the actual content of children’s self-verbalizations
was not measured in the present study. There are certainly multiple forms of
children’s private speech with some types of speech (e.g. noises, fantasy
speech, task-irrelevant statements, crib talk) appearing to be less ‘self-
regulatory’ than others. Such functionally different subcategories of
children’s private speech may follow different ontogenetic pathways (Berk,
), with each worthy of individual naturalistic examination. The as-
sumption which remains untested in the present study is that most of the
private speech used by the preschoolers was self-regulatory in nature. This
assumption may be reasonable on at least two accounts. First, research has
shown that the majority of preschool children’s private speech is task-
relevant and appears to have self-regulatory functions (Berk & Garvin, ;
Frauenglass & Diaz, ; Winsler et al., ). Second, there is some debate
as to whether other types of private speech utterances should be considered
non-regulatory (see Diaz, ). For example, self-stimulating word play
and noises, task-irrelevant affect expressions, and fantasy dialogues can serve
important regulatory functions as well, such as regulating emotions, main-
taining motivation, and processing\consolidating gender roles. Nevertheless,
future investigations along these lines should attempt to use a finer-grained
speech coding system. Obtaining private speech data of sufficient quality to
allow coding of utterances into various functional subcategories while the
,
child is in the natural context of the preschool classroom, however, poses
formidable methodological challenges. Real-time coding of utterances by an
observer in the classroom is likely precluded given that the proximity of the
observer to the child that would be required for hearing and classifying
utterances is likely to be too intrusive for the child. Using audio recordings
of children’s speech in the classroom may be a viable option but obtaining
high quality audio recordings of individual children’s private speech
utterances within the context of a loud preschool classroom is certainly not
without its challenges.
Third, although the time-sampling procedure used in this study avoids at
least one classic methodological problem often found in the private speech
literature, namely, large individual differences among children in the amount
of private speech uttered at any given time, time sampling itself has not
escaped criticism. Mann, Ten Have, Plunkett & Meisels (), for example,
have demonstrated that many time sampling procedures produce relatively
poor estimates of the duration and frequency of specific behaviours, at least
when applied to rich, dyadic social interaction data. Several characteristics of
the time-sampling procedures employed in this study, however, suggest that
many of Mann et al.’s concerns over time sampling either don’t apply or are
significantly reduced in this situation. First, the reliability of time-sampling
increases as the observation interval gets shorter, and the interval used in this
study ( seconds) was relatively short. Second, predominant activity
sampling, a procedure believed to increase the reliability of time sampling
observations (Mann et al., ), was used for the coding of children’s social
context and activity. Third, duration of behaviour, the parameter most
poorly estimated by time-sampling, was not of particular theoretical or
practical interest here. Finally, the events measured and studied here were
not as dynamic and complex as the moment to moment, (state and event),
mother-infant interaction data on which the Mann et al. critique was based.
In terms of implications of this work for practice in the early childhood
classroom, it is clear that increased awareness and understanding of young
children’s private speech use in the classroom is potentially useful for early
childhood educators because such speech may serve as an assessment tool in
at least three senses. First, children’s private speech provides an effective
window through which teachers can observe young children’s thought
processes, emotions, motivational processes, problem-solving strategies, and
their emerging metacognitive skills. Second, given that children are most
likely to talk to themselves when they are in their zone of proximal
development working on challenging but attainable goals, some have
suggested that children’s overt private speech use can serve as a measure of
the extent to which either a child’s goals\activities or the classroom’s
curricular activities are at an appropriately challenging level (Berk & Winsler,
; Winsler & Diaz, ). Third, recent data have demonstrated that
young children’s private speech may serve an important mediating link
between teacher–child interaction\instruction and children’s subsequent
cognitive growth (Berk & Spuhl, ; Winsler et al., ). That is,
children may profit more from scaffolded instruction when they engage in
private speech either during or shortly after the teaching episode, and such
private speech use in the context of a teaching interaction may be a better
predictor of children’s later task improvement than the teaching activities
themselves. Thus, teachers can look toward the self-speech of their children
as an indicator of the extent to which their instructional efforts have been
appropriated or internalized by children.
One of the central goals of developmental psychological science is the
prediction of behaviour. The present study contributed new data on when,
and in which contexts, young children use private speech in the setting of the
early childhood classroom, and how the predictability of such speech changes
with age. The field has now come to a fairly good understanding of when
children talk to themselves. Such knowledge, together with the information
gained from continuing studies exploring the functional and developmental
significance of children’s private speech, will lead us to a better understanding
of this important phenomena of early childhood.
REFERENCES
Astington, J. W. (). The child’s discovery of the mind. Cambridge, MA: Harvard
University Press.
Astington, J. W. (). What is theoretical about the child’s theory of mind ? : a Vygotskian
view of its development. In P. Carruthers & P. K. Smith (eds.). Theories of theories of mind
(pp. –). New York: Cambridge University Press.
Behrend, D. A., Rosengren, K. S. & Perlmutter, M. (). The relation between private
speech and parental interactive style. In R. M. Diaz & L. E. Berk (eds), Private speech: from
social interaction to self-regulation. Hillsdale, NJ: Erlbaum.
Berk, L. E. (). Children’s private speech : an overview of theory and the status of
research. In R. M. Diaz & L. E. Berk (eds), Private speech : from social interaction to self-
regulation. Hillsdale, NJ: Erlbaum.
Berk, L. E. & Garvin, R. A. (). Development of private speech among low-income
Appalachian children. Developmental Psychology , –.
Berk, L. E. & Landau S. (). Private speech of learning disabled and normally achieving
children in classroom academic and laboratory contexts. Child Development , –.
Berk, L. E. & Potts, M. K. (). Development and functional significance of private speech
among attention-deficit hyperactivity disordered and normal boys. Journal of Abnormal
Child Psychology , –.
Berk, L. E. & Spuhl, S. T. (). Maternal interaction, private speech, and task performance
in preschool children. Early Childhood Research Quarterly , –.
Berk, L. E. & Winsler, A. (). Scaffolding children’s learning : Vygotsky and early childhood
education. Washington, DC: National Association for the Education of Young Children.
Bivens, J. A. & Berk, L. E. (). A longitudinal study of the development of elementary
school children’s private speech. Merrill-Palmer Quarterly , –.
Diaz, R. M. (). Methodological concerns in the study of private speech. In R. M. Diaz
& L. E. Berk (eds), Private speech: from social interaction to self regulation. Hillsdale, NJ:
Erlbaum.
,
Diaz, R. M., Neal, C. J. & Amaya-Williams, M. (). The social origins of self-regulation.
In L. C. Moll (ed.), Vygotsky and education : instructional implications and applications of
sociohistorical psychology. New York: Cambridge University Press.
Diaz, R. M., Neal, C. J. & Vachio, A. (). Maternal teaching in the zone of proximal
development: a comparison of low- and high-risk dyads. Merrill-Palmer Quarterly ,
–.
Dunn, J., Brown, J., Slomkowski, C., Tesla, C. & Youngblade, L. M. (). Young
children’s understanding of other people’s feelings and beliefs : individual differences and
their antecedents. Child Development , –.
Fernyhough, C. (, April). Private speech and self-regulation : a link with mentalising
abilities? Paper presented at the biennial meeting of the Society for Research in Child
Development, Washington, DC.
Fernyhough, C. & Russell, J. (). Distinguishing one’s own voice from those of others: a
function for private speech? International Journal of Behavioral Development , –.
Fisher, C. B. & Lerner, R. M. (). Foundations of applied developmental psychology. In
C. B. Fisher & R. M. Lerner (eds), Applied developmental psychology. New York : McGraw-
Hill.
Flavell, J. H., Green, F. L., Flavell, E. R. & Grossman, J. B. (). The development of
children’s knowledge about inner speech. Child Development , –.
Flavell, J. H. & Miller, P. H. (). Social cognition. In W. Damon (ed.), D. Kuhn & R. S.
Siegler (Vol. eds), Handbook of child psychology – th Edition – Volume : Cognition,
perception, and language. New York: Wiley & Sons.
Frauenglass, M. H. & Diaz, R. M. (). Self-regulatory functions of children’s private
speech: a critical analysis of recent challenges to Vygotsky’s theory. Developmental
Psychology , –.
Furrow, D. (). Developmental trends in the differentiation of social and private speech.
In R. M. Diaz & L. E. Berk (eds), Private speech : from social interaction to self regulation.
Hillsdale, NJ: Erlbaum.
Goldin-Meadow, S. (). When gestures and words speak differently. Current Directions in
Psychological Science , –.
Goudena, P. P. (). The social nature of private speech of preschoolers during problem
solving. International Journal of Behavioral Development , –.
Hosmer, D. W. & Lemeshow, S. (). Applied logistic regression. New York: Wiley.
Hutt, S. J. & Hutt, C. (). Direct observation and measurement of behaviour. Springfield,
IL: Charles C. Thomas.
Jenkins, J. M. & Astington, J. W. (). Cognitive factors and family structure associated
with theory of mind development in young children. Developmental Psychology , –.
Johnson, J. E. & Ershler, J. (). Developmental trends in preschool play as a function of
classroom program and child gender. Child Development , –.
Luria, A. R. (). The role of speech in the regulation of normal and abnormal behaviour (J.
Tizard, ed.). New York: Liveright.
Mann, J., Ten Have, T., Plunkett, J. W. & Meisels, S. J. (). Time sampling : a
methodological critique. Child Development , –.
Meins, E. & Fernyhough, C. (). Linguistic acquisitional style and mentalising development:
the role of maternal mind-mindedness. Paper presented at the biennial meeting of the Society
for Research in Child Development, Washington, DC.
Nelson, K. (ed.). (). Narratives from the crib. Cambridge, MA: Harvard University Press.
Nelson, K. (). Language in cognitive development : The emergence of the mediated mind.
New York: Cambridge University Press.
Pellegrini, A. D. (). The development of preschoolers’ private speech. Journal of
Pragmatics , –.
Perner, J. (). Understanding the representational mind. Cambridge, MA: MIT Press.
Roberts, R. N. & Tharp, R. G. (). A naturalistic study of school children’s private speech
in an academic problem-solving task. Cognitive Therapy and Research , –.
Rubin, K. H. (). Impact of the natural setting on private speech. In G. Zivin (ed.), The
development of self-regulation through private speech. New York: Wiley.
Rubin, K. H. & Dyck, L. (). Preschoolers’ private speech in a play setting. Merrill-
Palmer Quarterly , –.
Rubin, K. H., Hultsch, D. F. & Peters, D. L. (). Non-social speech in four-year-old
children as a function of birth order and interpersonal situation. Merrill-Palmer Quarterly
, –.
Siegler, R. S. (). Emerging minds : the process of change in children’s thinking. New York:
Oxford University Press.
Smolucha, F. (). Social origins of private speech in pretend play. In R. M. Diaz & L. E.
Berk (eds.), Private speech : From social interaction to self-regulation. Hillsdale, NJ :
Erlbaum.
Tomasello, M., Kruger, A. C. & Ratner, H. H. (). Cultural learning. Behavioral and
Brain Sciences , –.
Vygotsky, L. S. (). Mind in society: the development of higher mental processes (M. Cole,
V. John-Steiner, S. Scribner, & E. Souberman, eds & trans.). Cambridge, MA : Harvard
University Press. (Original work published , , )
Vygotsky, L. S. (). Thought and language (A. Kozulin, trans.). Cambridge, MA: MIT
Press. (Original work published )
Vygotsky, L. S. & Luria, A. R. (). Studies on the history of behaviour : Ape, primitive, and
child (V. I. Golod & J. E. Knox, eds. and trans.). Hillsdale, NJ: Erlbaum. (Original work
published )
Wertsch, J. V. (). Vygotsky and the social formation of mind. Cambridge, MA: Harvard
University Press.
Winsler, A. (). Parent–child interaction and private speech in boys with ADHD. Applied
Developmental Science , –.
Winsler, A. & Diaz, R. M. (). Private speech in the classroom: the effects of activity type,
presence of others, classroom context, and mixed-age grouping. International Journal of
Behavioral Development , –.
Winsler, A., Diaz, R. M. & Montero I. (). The role of private speech in the transition
from collaborative to independent task performance in young children. Early Childhood
Research Quarterly , –.
