You are on page 1of 14

Child Development, July/August 2009, Volume 80, Number 4, Pages 13011313

A Good Story: Children With Imaginary Companions Create


Richer Narratives
Gabriel Trionfi

Elaine Reese

Clark University

University of Otago

In line with theories that childrens pretend play reflects and extends their narrative skills, children with
imaginary companions were predicted to have better narrative skills than children without imaginary companions. Forty-eight 5-year-old children and their mothers participated in interviews about childrens imaginary companions. Children also completed language and narrative assessments. Twenty-three of the children
(48%) were deemed to have engaged in imaginary companion play. Children with and without imaginary
companions were similar in their vocabulary skills, but children with imaginary companions told richer narratives about a storybook and a personal experience compared to children without imaginary companions.
This finding supports theories of a connection between pretend play and storytelling by the end of early
childhood.

Imaginary companion play is captivating for children who engage in it, for parents who are surprised by it, and for developmental researchers
who want to understand it. Early research often
attributed imaginary companion play to psychopathology, personality defects, or deficiencies in social
skills (Ames & Learned, 1946; Svendsen, 1934).
Taylor (1999) argued that these early studies, taken
largely from clinical populations, were flawed
because they did not include comparison samples
from nonclinical populations. Indeed, contemporary wisdom is that imaginary companion play
may even offer some developmental benefits for
children (e.g., Bouldin & Pratt, 2002; Gleason, Sebanc, & Hartup, 2000; Taylor & Carlson, 1997). In
this study, we explore the potential developmental
benefits of imaginary companion play for childrens
narrative skills.
Traditionally, the definition of an imaginary
companion was restricted to repeated play with
an invisible other (Svendsen, 1934), but later the
definition of an imaginary companion expanded
to include certain instances of play with personiGabriel Trionfi is currently with IDEO.
The Marsden Fund of the Royal Society of New Zealand
funded this research. We thank Kathleen McMenamin, Smita Srivastava, and the members of the Language and Memory team
for data collection, transcription, and coding. Special thanks go
to Marjorie Taylor and Eric Amsel for their insights throughout
the project. We are also grateful to the families for sharing with
us their imaginary friends and their stories, both imagined and
real.
Correspondence concerning this article should be addressed to
Elaine Reese, Psychology Department, University of Otago, P.O.
Box 56, Dunedin, New Zealand. Electronic mail may be sent to
ereese@psy.otago.ac.nz.

fied objects (Singer & Singer, 1990; Taylor, 1999).


Critical to both forms of pretend play is that children view their imaginary companion as a separate other. Harris (2000) characterized both types
of imaginary companion play as sustained forms
of role play. In both, childrens interactions with
their imaginary companions are a form of simulated social exchange. Indeed, the similarity
between imaginary companions and real friendships is a defining feature of this play: As with
real friends, children play with their imaginary
companions, pretend with them, involve them in
their daily routines and occasionally argue with
them (Gleason, 2004a, p. 205; cf. Gleason et al.,
2000; Taylor, 1999).
Ones definition of imaginary companions influences ones estimate of the number of children who
have an imaginary companion. Some researchers
have reported imaginary companion play to be
engaged in by half of all children (e.g., Singer &
Singer, 1990), whereas others using stricter definitions have reported that only one fifth of all children have imaginary companions (Newson &
Newson, 1968). A longitudinal study revealed new
information about the occurrence of imaginary
companion play over time (Taylor, Carlson,
Maring, Gerow, & Charley, 2004). Three- and
four-year-olds were interviewed about their
engagement in imaginary companion play, and the
same children were interviewed again 3 years later.

 2009, Copyright the Author(s)


Journal Compilation  2009, Society for Research in Child Development, Inc.
All rights reserved. 0009-3920/2009/8004-0026

1302

Trionfi and Reese

Using Taylors (1999) definition, a surprising 31%


of 6- and 7-year-olds reported engaging in imaginary companion play, and 49% of children were
identified as engaging in imaginary companion
play at some point across the duration of the study.
Even under the most stringent double-interview
process, Taylor and Carlson (1997) still found a
28% rate of occurrence of imaginary companion
play in their sample of one-hundred and fifty-two
3- to 4-year-olds. Despite some variations in estimating how many children engage in imaginary
companion play, it is clear that researchers have
been able to identify consistently two groups of
children: those who engage in imaginary companion play at some point in development and those
who do not engage in imaginary companion play.
The question then becomes: How are children with
imaginary companions different from those children without imaginary companions?
Imaginary companion play is engaged in most
often by firstborn children (Bouldin & Pratt, 1999;
Gleason et al., 2000; Manosevitz, Prentice, &
Wilson, 1973) and is somewhat more common in
the play of young girls than young boys (Carlson
& Taylor, 2005; Gleason, 2004a; Pearson et al.,
2001; Taylor & Carlson, 1997; but see Taylor
et al., 2004). Boys, in contrast, are more likely
than girls to impersonate a character, another
form of role play (Carlson & Taylor, 2005; cf.
Harris, 2000). Imaginary companion play is also
positively related to childrens sociability, such
that children with imaginary companions are
reported to have just as many or more real
friendships (Bouldin & Pratt, 1999; Gleason et al.,
2000; Taylor, Cartwright, & Carlson, 1993) and
experience the same level of peer acceptance as
children without imaginary companions (Gleason,
2004a). Adults who provide retrospective reports
of imaginary companions have a stronger orientation toward others than adults who do not report
having had an imaginary companion in childhood
(Gleason, Jarudi, & Cheek, 2003). Moreover, children with imaginary companions are no more or
less likely to be shy than children without imaginary companions (Bouldin & Pratt, 2002), and
adults who report having had an imaginary companion do not rate themselves as shyer than
adults who report not having had an imaginary
companion (Gleason et al., 2003). In one study,
parents reported that children with imaginary
companions were more anxious than children
without imaginary companions, but the reported
levels of anxiety were not of clinical concern
(Bouldin & Pratt, 2002). Children who engage in

imaginary companion play do show an inclination


toward other forms of fantasy (Bouldin, 2006;
Bouldin & Pratt, 1999; Taylor & Carlson, 1997),
and 4-year-old children with imaginary companions show an advanced understanding of minds,
even after controlling for verbal ability (Taylor &
Carlson, 1997).
Language skills are another established correlate
of imaginary companion play. Preschool boys with
imaginary companions speak more and use longer
utterances than boys without imaginary companions (Singer & Singer, 1990). Preschool children
with imaginary companions also have more
advanced receptive vocabularies than children
without imaginary companions (Taylor & Carlson,
1997). Finally, children with imaginary companions
demonstrate more complex syntax in their speech,
with higher rates of adverbial and relative clauses,
compared to children without imaginary companions (Bouldin, Bavin, & Pratt, 2002). These linguistic
correlates of imaginary companion play may reflect
a broader and well-founded developmental connection between childrens play and language. From
an early age, childrens symbolic play is linked to
their language use (e.g., Shore, OConnell, & Bates,
1984), although this link can at times be quite specific. For instance, Tamis-LeMonda and Bornstein
(1994) found a correlation between toddlers early
symbolic play and their later language, but only
with a measure of semantic diversity, not with total
productive vocabulary.
The main aim of this study is to extend prior
research on the language skills of children with and
without imaginary companions by moving into the
realm of narrative. Play researchers theorize that
children need to use explicit language to negotiate
meaning in the pretend scene (Bruner, 1996; Fein,
1989; Pellegrini, 1985). These linguistic negotiations
are found in the co-construction of roles, settings,
goals, and conflicts (Goncu, 1993; Goncu &
Katsarou, 2000), as well as in discussion of other
elements, such as the meaning of props used
(Garvey, 1977). Such linguistic negotiations highlight the similarities that theorists have noted
between play and narrative in form and composition, particularly in the use of characters, a scene, a
goal, and a conflict (Bruner, 1990; Kavanaugh &
Engel, 1998; Pellegrini & Galda, 1993; Sutton-Smith,
2001). Symbolic play and narrative have been characterized as two distinct expressions of symbolic
thought (Kavanaugh & Engel, 1998, p. 81). Alternatively, pretend play and storytelling have been
identified as elements in a continuum of narrative
activities (Nicolopoulou, 2002). Nicolopoulou (2006)

Play and Narrative

proposed that pretend play and storytelling have


different origins in social interaction, and only late
in the preschool period do the two coalesce into
complex sociodramatic play. Critically, pretend
play interventions can improve childrens narrative
skills (e.g., Baumer, Ferholt, & Lecusay, 2005).
Our specific goal in this study was to explore the
relations between imaginary companion play and
childrens narrative skills. Imaginary companion
play is theorized to be a sustained and richly
detailed form of role play (Harris, 2000), which
over time can become more layered or abstract. We
view imaginary companion play as one of the most
complex expressions of play in early childhood.
Similarly, narrative, as an instance of decontextualized language, is one of the most complex linguistic
expressions of early childhood (Dickinson & Snow,
1987; Snow, 1983). Somers and Yawkey (1984) theorized that imaginary companion play facilitates
cognitive development through a number of
abstract thought processes. One such process is
decontextualization, which they identify as the
use of real situations out of their contexts during
play (p. 86). Following this thinking, we further
offer that imaginary companion play is highly decontextualized because it primarily involves real
social interactions with an imagined other during
play. Play with an invisible companion may even
promote social interactions with real others when
children attempt to share details about their invisible friend with interested adults (see Gleason,
2004b). Conceptually, then, both imaginary companion play and narrative rely heavily on mental
and linguistic constructions to create context. Based
on these similar requirements for decontextualization in the two activities, we predicted that children
with imaginary companions would demonstrate
more complex narrative skills than children without imaginary companions.
This extension to narrative is especially important given increasing evidence that narrative skill in
early childhood is linked to reading success and to
school achievement (e.g., Griffin, Hemphill, Camp,
& Wolf, 2004; ONeill, Pearce, & Pick, 2004; Reese,
Suggate, Long, & Schaughency, in press), even after
childrens vocabulary and syntax skills are taken
into account. Mastery of the elements of a canonical
narrative is essential for story comprehension, both
when reading independently and when listening to
stories. Skill at telling stories in a culturally appropriate way is also appreciated by teachers
(Michaels, 1981) and, we suspect, by peers,
although research has not yet addressed the latter
hypothesis. Thus, if we are correct that children

1303

with imaginary companions show stronger narrative skills, it is possible that children with imaginary companions will develop into stronger readers
in the elementary school years.
To explore our initial hypothesis that imaginary
companion play is linked to childrens narrative
skill, we drew upon measures from a larger longitudinal study of childrens language and socialcognitive development from ages 1 to 5 (see
Reese, 2002, for an overview of this study). At the
5-year data point, children and their parents participated in interviews about the childrens imaginary companions, past or present. Children also
completed two standardized language measures, a
measure of story understanding, and two narrative
production tasks at this same age. We hypothesized that children who engaged in imaginary
companion play would produce narratives that
were qualitatively richer and more detailed than
children who did not engage in imaginary companion play. Peterson and McCabe (1983), drawing upon Labov and Waletzkys (1967 1997) highpoint analysis, noted that a child narrators use of
orientations and evaluations could be used to classify the complexity and quality of a narrative. In
contrast to referential statements that simply tell
what happened in the story, orientations (to characters, time, and place) contextualize the story for
the listener, whereas evaluations (adjectives, emotions, and dialogue) emphasize meaning or high
points within the narrative. Both orientations and
evaluations in childrens narratives are positively
correlated with their later reading skill (Griffin
et al., 2004). We predicted that these individual
differences in narrative quality would be robust
and evident across the two contexts, whether or
not they involved an element of pretense. Finally,
we predicted that these differences in narrative
skill would still be evident even when we controlled for potential differences in the vocabulary
levels of children with and without imaginary
companions.

Method
Participants
All participants were part of a larger longitudinal study conducted in New Zealand (see Reese,
2002, for more details). Participants included 48
motherchild dyads (24 males and 24 females)
recruited initially by posting fliers on community
notice boards and contacting parents through
public birth records. Data for this study were taken

1304

Trionfi and Reese

from the 65-month time point (M = 65.2 months,


SD = 13.7 days). Firstborn children represented
37% of the sample (18 of 48). All participants identified themselves as living in households where English was the primary language. Forty-four mothers
identified their childrens ethnicity as New Zealand
European and four mothers identified their childrens ethnicity as New Zealand M
aori. Demographic information collected at the beginning of
the study when children were 19 months old
revealed that the sample was on average of middleclass status (Elley & Irving, 1976) and that mothers
had an average of 13.3 years of education
(SD = 2.26 years). Mothers ranged in their education levels from not having completed high school
to having completed a postgraduate degree. Fewer
than half of the mothers (45.8%) had continued
their education past high school.
Materials
A video and audio recorder was used during
data collection. An unfamiliar story, A Perfect
Fathers Day (Bunting, 1991, 32 pages), and a puppet
were used during the story retelling and comprehension tasks. At the end of the second visit, a
small gift was presented to the child.
Procedure
Two female researchers met with participants in
their homes over two sessions within 2 weeks of
children reaching 65 months of age. Each session
lasted an average of 1 hr.
Imaginary Companion Interview
The imaginary companion interview was based
on the procedure utilized by Taylor et al. (1993).
After interacting with the child for a few minutes,
the researcher said Now Im going to ask you
some questions about friends. Some friends are
real, like the kids who go to your school, the ones
you play with. And some friends are pretend
friends, ones that are make-believe, or imaginary.
Do you have a pretend or imaginary friend, or have
you ever had one? If the child answered yes,
then the researcher asked the child 15 questions
about the pretend friend, including its gender, age,
physical appearance, likes, and dislikes. Meanwhile, another researcher conducted the imaginary
companion interview separately with the childs
mother with a slightly different preamble, in which
the researcher stated an interest in childrens imagi-

nation and imaginary companion play. The


researcher explained that children of this age were
often secretive about their imaginary companions
and that it was common for parents to be unaware
of them. Mothers were then asked if they were
aware of their child engaging in imaginary companion play currently or at a previous time. If they
responded affirmatively, the remainder of the imaginary companion interview was conducted. If mothers or children identified multiple imaginary
companions, they were asked to describe only their
three favorites.
Children were considered to have an imaginary
companion if either the children or their mothers
currently acknowledged an imaginary companion
or indicated the presence of an imaginary companion in the past. Mothers reviewed childrens imaginary companion interviews and clarified if children
had mistakenly identified a real friend. These childrens reports were not included as evidence of an
imaginary companion. We were unable to use a
more stringent double-interview process (Taylor &
Carlson, 1997) due to time constraints associated
with the larger longitudinal study.
Coding of imaginary companion interviews was
completed by two independent coders for 22% of
the interviews. Interrater reliability on the presence
or absence of an imaginary companion was 100%.
The remaining interviews were coded by one of the
coders.
Vocabulary Skills
All children completed the Peabody Picture
Vocabulary TestIIIB (PPVTIIIB; Dunn & Dunn,
1997) during the first session and the Expressive
Vocabulary Test (EVT; Williams, 1997) during the
second session. The PPVT measures childrens
receptive vocabulary and the EVT measures childrens expressive vocabulary. Standard scores were
used in analyses.
Story Comprehension
During the second session, researchers read an
unfamiliar story (A Perfect Fathers Day; Bunting,
1991) to children. Children were asked 12 comprehension questions as a researcher read the story to
them. Comprehension questions focused on plot
information (what happened), inferences (causal
connections between plot events and inferring
meaning from pictures), and real world knowledge
that impacted story understanding (see Reese, 1995,
for a full description of this task and scoring).

Play and Narrative

Children could receive a total of 15 points for their


answers, with one plot question worth 2 points and
a formal definition question (What is a surprise?)
worth 3 points. Coding of the story comprehension
task was completed by two independent coders for
25% of the sample, achieving an inter-rater agreement of Cohens j = .97. Each of the coders proceeded to code half of the remaining transcripts
independently.
Story Retelling
At the conclusion of the story comprehension
task, children were introduced to a puppet. Children were told that the puppet did not hear the
story and were asked to retell the story to the puppet from beginning to end (see ONeill et al.,
2004, for a similar procedure). Researchers assisted
the children in their retelling by first reminding the
children of the storys title. Researchers provided a
maximum of two supportive, yet nonspecific,
prompts per page. If children did not respond to
the second prompt, researchers turned the page.
Researchers prompts included generic questions
(e.g., Now whats happening?), and generic support (e.g., Thats neat), but avoided any elaborations of childrens retelling. Retellings were
recorded and transcribed for coding.
Childrens retellings were then coded for story
memory and narrative quality (adapted from Peterson & McCabe, 1983; Reese, 1995). Initially, the text
of the storybook was divided into 63 propositions.
Childrens retellings, which ranged from 1 to 40
propositions, were first compared to the text propositions. One point was issued for each text proposition children mentioned in their retelling, termed a
memory unit. Each memory unit was then coded for
narrative quality in terms of evaluative and orienting devices that were additional to those provided
in the text. Five separate categories of narrative
quality included: descriptors (adjectives and
adverbs), dialogue, character names, temporal
locativecausal, and verbatim recall from text.
Children received a maximum of 1 point for each
category of narrative quality per memory unit. This
practice prevented unfairly inflating the scores of
children who repeatedly used the same narrative
device in one given memory unit. For example, a
big red balloon found in one memory unit would
only receive 1 point for the descriptor category.
Coding of memory units and narrative quality was
completed by two independent coders for 25% of
the story retelling transcripts, achieving an interrater reliability of Cohens j = .83. Any disagree-

1305

ments that arose during the reliability process were


then resolved, and these resolved scores were used
in analyses. Each of the two coders proceeded to
code half of the remaining transcripts independently. Following this coding, narrative quality
points were divided by the number of memory
units the child provided to create a story retelling
narrative score for each participant. This score represents the average narrative quality score per
memory unit produced by the participant during
the retelling narrative. These average narrative
quality scores reduced the variability associated
with sheer talkativeness.
Past Event Narratives
Mothers selected three past events that children
had experienced in the past year (typically within
the past month) for researchers to discuss with children (see Cleveland & Reese, 2005, for a full
description of the procedure). The events were
selected to be unique and to have occurred over the
time span of 1 day. Thus, birthdays and holidays
were excluded because they are either routine or
extended events. Researchers discouraged selection
of movies and plays because these events contain
an inherent storyline. Past event narratives were
provided at the end of the same session as the story
retelling for half the children, and at the end of the
other session for the remaining children. Researchers started off the discussion with a general prompt
Your mum told me you went to Tunnel Beach but
she didnt tell me what happened. Tell me everything that happened when you went to Tunnel
Beach. Researchers followed up this initiation with
prompts such as Tell me some more about that
and What else? and confirmed childrens
responses (Wow) but did not ask specific questions about the event. Childrens responses were
recorded and transcribed for coding.
Childrens performance across the three narratives in this task was variable, probably because
we were relying on mother-nominated events, not
all of which were of similar interest to children.
Therefore, we selected the highest scoring narrative from each child for analysis to capture childrens best performance on this task. A memory
unit consisted of a childs utterance containing
unique information about the past event, as
opposed to repeated or off-topic information (see
Cleveland & Reese, 2005, for coding details). Similar to the story retelling coding, these memory
units were then coded for narrative quality. The
narrative quality coding scheme for the past event

1306

Trionfi and Reese

narratives parallels the coding scheme used in the


story retelling task but did not include verbatim
recall because there was no documentation of the
original events. Thus, the four categories of narrative quality for past event narratives were descriptors (adverbs and adjectives), dialogue, character
names, and temporallocativecausal terms. As in
the story retelling task, children only received 1
point for each subcategory of narrative quality
within a given memory unit to prevent the skewing of scores in favor of using only one type of
narrative device.
Coding of memory units was completed by two
independent coders as part of a more extensive
coding scheme for a previous study (Cleveland &
Reese, 2005). On 25% of the past event narrative
transcripts, the two coders achieved an interrater
reliability of Cohens j = .77. For narrative quality,
a different team of two independent coders
achieved an interrater reliability of Cohens j = .80
on 25% of the transcripts. Any disagreements that
arose during the reliability process were then
resolved, and these resolved scores were used in
analyses. Each of the two coders proceeded to code
half of the remaining transcripts independently. As
with the story retelling narratives, the past event
narrative quality points were divided by the number of memory units to create average past event
narrative scores for each participant. These scores
represent the average narrative quality scores per
memory unit produced by participants during their
past event narratives and adjusts for talkativeness.
Coding for imaginary companions and narrative
quality was completed independently. Specifically,
imaginary companion interviews were coded from
video by participant name, whereas narratives were
scored from transcripts using participant ID as the
sole identifier. The first author was the primary
coder on all tasks except for story comprehension
and past event narrative memory units; five
research assistants served as reliability coders.

Results
Analysis of the imaginary companion interviews
identified 23 of the 48 children in this study as having an imaginary companion (48% of the sample).
These childrens imaginary companions were identified in the following manner: Eight were reported
by children only (35% of the imaginary companion
sample), nine by mothers only (39% of the imaginary companion sample), and six by both mother
and child (26% of the imaginary companion sam-

ple). Sixty-five percent of children with imaginary


companions identified having only one, 13% identified two, and 22% acknowledged having three or
more. Of the 36 imaginary companions described
by children and their mothers, half were identified
as people, with the remainder identified as animals
(14%), fantasy beings such as a cucumber boy
(25%) or unclear (11%). Only three imaginary companions were based on personified objects (toys).
Almost all described imaginary companions were
current companions with only three identified as
former companions.
Children with and without imaginary companions did not differ as a function of their mothers
level of education, t(46) = 0.36, ns (Ms = 13.44 and
13.20 years, respectively). Among children who
engaged in imaginary companion play, 10 were
male (42% of all males in the sample) and 13 were
female (54% of all females in the sample). This gender difference was not statistically significant, v2(1,
N = 48) = 0.75, ns. Childrens imaginary companion
play, however, did differ as a function of birth
order. Thirteen (56%) of the children with imaginary companions were firstborns, but only 5 (20%)
of the children without imaginary companions
were firstborns. This birth order difference was
statistically significant, v2(1, N = 48) = 6.81, p < .01.
Participants descriptions of imaginary companions were often detailed and colorful. In almost all
cases, the gender of the imaginary companion
matched that of the child. Cross-gender friendships
occurred only among female participants, particularly when more than one imaginary companion
was identified. Every male (n = 10) who described
an imaginary companion identified the companion
as male even when multiple imaginary companions
were described. One female participant classified
her imaginary companion named Batman as both
a girl and a boy. Participants description of
imaginary companions ages ranged from a baby
puppy (named Brown Puppy) to a 1-million-yearold wrestler (named Giant Strongman, who had no
toenails), with one participant describing Creka
(who looks like a shadow) as both old and
young. Imaginary companions names ranged
from nondescript (e.g., Her) and mundane (e.g.,
Emily) to the truly unique (e.g., Holibola). When
asked who would win in a race, most participants
predicted imaginary companions to be the winners,
although predictions were often justified based on
factors that would determine the outcome (i.e., the
size or age of the imaginary companion, or accessories such as a flying car or high-heeled shoes).
When asked who was more likely to be naughty,

Play and Narrative


Table 1
Language and Narrative Scores for Participants Without Imaginary
Companions (No IC) and With Imaginary Companions (With IC)
No IC
(n = 25)
M

SD

With IC
(n = 23)
M

SD

Vocabulary

lary and story comprehension skills of participants


who engaged in imaginary companion play and
those who did not.
Measures of Narrative Skill

t(46)

PPVTIII
106.32 10.62 105.22 14.45 )0.30 0.09
EVT
106.71 12.41 106.13 12.46 )0.16 0.05
Story comprehension
9.15 2.85
9.02 2.73 )0.16 0.05
Story retelling
Memory units
10.60 4.30 15.57 12.54 1.87 0.50
Narrative quality
2.76 2.49
9.43 14.61 2.25* 0.64
Narrative score
0.22 0.18
0.51 0.42 3.10* 0.90
Past event narratives
Memory units
10.36 7.36 11.22 9.57 0.35 0.10
Narrative quality
8.6
6.6
10.65 8.70 0.92 0.27
Narrative score
0.79 0.23
0.97 0.36 2.02* 0.60
Note. PPVTIII = Peabody Picture
EVT = Expressive Vocabulary Test.
*p < .05.

1307

Test

III;

almost all participants (81%) identified imaginary


companions as the likely offender.
Main Analyses
Table 1 contains descriptive statistics for the
vocabulary, story comprehension, and narrative
tasks for children with and without imaginary companions. One child did not complete the EVT and
the story comprehension task. Group means were
substituted for these two missing scores in analyses. Unpaired t tests were used to compare the two
groups of children on all measures. For several of
the narrative subcategories (dialogue in both narrative tasks and verbatim recall for story retelling),
significant heterogeneity of variance was found
using Levenes test. We reanalyzed these variables
using t tests that did not assume equal variances,
but in all cases the pattern of significant findings
was the same. Therefore, we present only the
results with equal variances assumed.

Because the literature on narrative performance


during story retelling tasks is distinct from the literature on childrens personal narratives (see ONeill
et al., 2004; Peterson & McCabe, 1983), we first analyzed these two narrative tasks separately as a function of childrens imaginary companion status.
Supporting the distinct nature of the two narrative
tasks, the childrens overall narrative scores in
the two tasks were not significantly correlated,
r(46) = .10, ns.
Story retelling. Some children in each group had
narrative scores of zero in this task, but children in
the two groups differed dramatically in the maximum narrative scores obtained (0.70 for children
without imaginary companions and 2.0 for children
with imaginary companions). Children with imaginary companions were significantly higher in their
overall story retelling narrative score than children
without imaginary companions, t(46) = 3.10, p < .01
(see Table 1), with an effect size (d = 0.9) that qualifies as large (Cohen, 1988). Children with and without imaginary companions did not differ
significantly in the number of memory units produced, t(46) = 1.87, p < .10, d = 0.5, but the trend
was for children with imaginary companions to
produce more memory units than children without
imaginary companions.
Based on the difference in story retelling narrative scores between the two groups and the large
effect size associated with that difference, we

Table 2
Story Retelling Narrative Subcategory Scores for Participants Without
Imaginary Companions (No IC) and With Imaginary Companions
(With IC)
No IC
(n = 25)

Measures of Vocabulary Development and Story


Comprehension
Participants without imaginary companions and
those with imaginary companions did not differ in
their performance on the PPVTIII, t(46) = )0.30,
ns, or on the EVT, t(46) = )0.16, ns. Neither was
story comprehension a function of engagement in
imaginary companion play, t(46) = )0.16, ns. Thus,
there were no significant differences in the vocabu-

Descriptors
Dialogue
Characters
Temporallocative
causal
Verbatim recall
*p < .05.

With IC
(n = 23)

SD

SD

t(46)

0.17
0.02
0.14
0.01

0.38
0.05
0.10
0.03

0.36
0.09
0.23
0.03

0.58
0.14
0.30
0.70

1.35
2.23*
1.35
0.95

0.39
0.63
0.38
0.29

0.02

0.07

0.11

0.27

1.57

0.44

1308

Trionfi and Reese

conducted an analysis of the subcategories that


comprise this composite narrative score (see
Table 2). Although these analyses identified only
the subcategory of dialogue to be significantly different between the two groups, mean scores for
children with imaginary companions were higher
than children without imaginary companions in all
narrative quality subcategories. The probability for
children with imaginary companions to exceed the
narrative skills of children without imaginary companions in all five subcategories was only .03 (binomial test). Effect sizes for the story retelling
subcategories were all in the small to medium
range.
Past event narratives.. Similar to the story retelling
task, although the minimum best past event narrative scores were similar for both groups (0.42 for
children without imaginary companions and 0.43
for children with imaginary companions), the same
was not true of maximum best past event narrative
scores (1.17 for children without imaginary companions and 2.0 for children with imaginary companions). Indeed, there was a significant difference
between the best past event narrative scores of participants who engaged in imaginary companion
play and those who did not, t(46) = 2.02, p < .05
(see Table 1). This significant difference between
the two groups past event narrative scores was
supported by a medium effect size (d = 0.60). There
was no significant difference between the two
groups in the number of memory units they provided for their best past event narrative,
t(46) = 0.35, ns.
Based on the difference in past event narrative
scores between the groups and the medium effect
size associated with that difference, an analysis of
the subcategories that comprise this composite narrative score was completed (see Table 3). These
Table 3
Past Event Narrative Subcategory Scores for Participants Without
Imaginary Companions (No IC) and With Imaginary Companions
(With IC)
No IC
(n = 25)

Descriptors
Dialogue
Characters
Temporallocative
causal
*p < .05.

With IC
(n = 23)

SD

SD

t(46)

0.58
0.01
0.11
0.09

0.20
0.03
0.15
0.12

0.63
0.03
0.12
0.18

0.29
0.08
0.20
0.17

0.73
1.07
0.32
2.02*

0.21
0.31
0.09
0.58

analyses revealed a significant difference between


the groups in their use of temporallocativecausal
terms and again showed scores that were higher
for children with imaginary companions than those
without imaginary companions in all narrative subcategories. The probability for children with imaginary companions to exceed the narrative skills of
children without imaginary companions in all four
subcategories was .06 (binomial test). Most effect
sizes were in the small to medium range.
Although childrens overall narrative scores in
the two narrative tasks were not correlated, an
identical pattern of a narrative advantage for children with imaginary companions was demonstrated across all narrative subcategories for both
narrative tasks. The probability for children with
imaginary companions to exceed the narrative skills
of children without imaginary companions in all
nine subcategories across the two tasks was only
.002 (binomial test). To capture this general pattern
in a single score, we created an overall narrative
score composed of an average of the childrens narrative scores across the story retelling and past
event narrative contexts. Children with imaginary
companions had a significantly higher overall
narrative score (M = 0.74, SD = 0.27), than children without imaginary companions (M = 0.51,
SD = 0.16), t(46) = 3.69, p < .001. The effect size for
this comparison was large (d = 1.06). Children with
imaginary companions did not provide significantly more memory units across the two narratives
compared to children without imaginary companions, t(46) = 1.5, ns. All subsequent analyses are
performed on childrens overall narrative score.
Imaginary Companion Play as a Unique Predictor of
Childrens Narrative
Because children who engaged in imaginary
companion play also tended to be firstborns, however, we needed to explore the unique role of
imaginary companion play in childrens narrative
skills. Could these differences instead be
accounted for by childrens birth order? We conducted a hierarchical regression analysis to predict
childrens overall narrative skill as a function of
birth order and imaginary companion status, after
controlling for childrens vocabulary skill. Because
childrens PPVT and EVT scores were correlated,
r(46) = .47, p < .01, we computed an overall vocabulary score composed of the average of the two
vocabulary scores. We entered this vocabulary
score in the first step of a regression analysis predicting childrens narrative score. In a second step,

Play and Narrative


Table 4
Predicting Childrens Narrative Skill From Vocabulary, Birth Order,
and Imaginary Companion Status
B
Step 1
Vocabulary
Step 2
Birth order
Step 3
Imaginary companion status

.004

SE

0.003

nR2

.18

.02

).16

0.07

).32

.21**

.18

0.06

.37

.12**

Note. Final beta weights are reported.


**p < .01.

we added childrens birth order, and in a third


step, childrens imaginary companion status. Childrens vocabulary was not correlated with either
their imaginary companion status, r(46) = ).05 or
with their birth order, r(46) = .08, nor was vocabulary a significant predictor of childrens narrative
skill (see Table 4). In contrast, both birth order
and imaginary companion status uniquely predicted childrens narrative skill (see Table 4). Children who were firstborns and children with an
imaginary companion had higher narrative skills
across the two contexts than children who were
laterborns or who did not have an imaginary companion.
We also conducted a second regression analysis
with an interaction term (birth order by imaginary
companion status) entered in the final step to determine if firstborns with imaginary companions were
primarily responsible for the main effect of imaginary companion status on narrative. The results of
this analysis indicated no significant interaction
between birth order and imaginary companion status (b = ).05, ns). Because only 5 firstborns did not
have imaginary companions, however, we consider
this analysis exploratory.
Social Interaction as a Potential Mechanism of the
Narrative Advantage for Children With Imaginary
Companions
Nearly all (92%) of the imaginary companions in
our sample comprised invisible friends, not personified objects. If children with invisible imaginary
companions want to share their imaginary companions with the real people in their lives, they must
describe their imaginary companion to others
(Gleason, 2004b). We hypothesized that these
decontextualized conversations may serve to
further develop childrens narrative skills. We conducted an exploratory analysis of childrens overall

1309

narrative score along a continuum of social interaction possibilities, from those children without imaginary companions (n = 25), to those children with
imaginary companions of which only the child had
knowledge (child knowledge; n = 8), to those children whose mothers knew about their imaginary
companions (mother knowledge; n = 15). Only children with imaginary companions could benefit
from social interactions with those companions, but
children whose mothers know about their imaginary companions may have an added source of
social interactions about the companion.
We used nonparametric analyses to test this prediction due to the small number of children with
child-knowledge-only imaginary companions. A
KruskalWallis test indicated a significant difference in childrens narrative scores as a function of
social-interaction possibilities, v2(2, N = 48) = 12.85,
p < .002. On their overall narrative scores, children
without imaginary companions had a mean rank of
18.38, children with child-knowledge-only imaginary companions had a mean rank of 24.38, and
children with mother-knowledge imaginary companions had a mean rank of 34.77. A follow-up
MannWhitney U test indicated a significant difference in the narrative scores of children with
mother-knowledge imaginary companions and children without imaginary companions (Z = )3.74,
p < .01). Children with child-knowledge-only companions did not differ significantly from either
group in their narrative scores.

Discussion
The main finding of this study was that 5-yearold children who currently or previously engaged
in imaginary companion play had more advanced
narrative skills than children who did not engage
in this type of play. Although children in the two
groups did not differ significantly in their vocabulary skills or in their story understanding, the children with imaginary companions told richer stories
in two different contexts compared to children
without imaginary companions. Firstborns also told
richer narratives than laterborns, but imaginary
companion play uniquely predicted childrens
narrative skill, even after accounting for birth
order and vocabulary skill. These findings add to
the growing body of evidence that imaginary
companion play is associated positively with childrens linguistic and social-cognitive development
(Bouldin et al., 2002; Singer & Singer, 1990; Taylor
& Carlson, 1997). These findings also fit well with

1310

Trionfi and Reese

Nicolopoulous (2006) proposal that pretend play


and storytelling have become integrated by the end
of early childhood. Because the PPVT in particular
can be viewed as a measure of childrens general
cognitive ability (Cunningham & Stanovich, 1997),
these findings point to the unique role of imaginary
companion play, over and above cognitive ability,
in childrens narrative skills.
Specifically, participants who engaged in imaginary companion play used more dialogue in their
story retelling narratives and more temporal
locativecausal terms in their past event narratives
when compared to participants who did not engage
in imaginary companion play. Although we did not
have specific predictions about the aspects of narrative on which children with imaginary companions
would excel, both dialogue (a type of evaluation)
and temporallocativecausal terms (types of orientation) are advanced narrative elements. The storybook for the retelling task, A Perfect Fathers Day,
included a generous amount of dialogue in the text.
In their use of created dialogue in their retellings,
children with imaginary companions appeared to
be better at emulating the style of the text compared to children without imaginary companions.
In the past event narrative context, temporal, locative, and causal terms are essential for making the
event narrative understandable to the listener, and
it is not until middle childhood that most children
can use these elements coherently (Reese et al.,
2008). Thus, we believe that the children with imaginary companions in our study were especially
good at using the most advanced narrative elements in each context. Moreover, they excelled at
different elements in each contextthose elements
that were most essential for conveying that particular story effectively to the listener.
Furthermore, mean scores of children who
engaged in imaginary companion play were higher
than those of children who did not engage in imaginary companion play for all of the narrative subcategories across both tasks, and effect sizes for
these differences were almost all in the small to
medium range. The effect size for the overall difference in childrens narrative skill as a function of
imaginary companion status was large. All of these
findings confirm our hypothesis and identify a clear
link between imaginary companion play and childrens storytelling abilities. This finding has practical importance given a growing body of evidence
that childrens narrative skill upon school entry
strongly predicts their later reading ability, especially their reading comprehension in mid-elementary school (e.g., Dickinson & Tabors, 2001; Griffin

et al., 2004). A word of caution is in order, though.


Imaginary companion play has not yet been implicated directly in childrens reading achievement,
nor is there causal evidence for the benefits of
imaginary companion play in any developmental
domain. We are not advocating that parents or
teachers encourage children without imaginary
friends to create such friends. Rather, if a child has
already created an imaginary companion, parents
and teachers could allow this play to flourish.
Although the link between imaginary companion
play and narrative quality is clear, the reason for
this link is not. We speculate that one reason that
children with imaginary companions have richer
narrative skills is that they gain practice in decontextualized conversations both during interactions
with their imaginary companions, and when they
tell others about their imaginary companions.
Advanced narratives rely on decontextualized
language skill. All of the anecdotal and case study
evidence indicates that at some point during the
duration of childrens imaginary companion play,
most will tell their parents and family members a
story about their imaginary companions. It is possible that this storytelling leads to conversations in
which adults promote childrens engagement in
and understanding of complex decontextualized
language practices. Because children have sole
knowledge of their imaginary companions, adults
may be required to ask a greater number of questions about imaginary companions than they ask
when conversing about topics for which both adult
and child share knowledge. Overcoming this lack
of shared knowledge is practically impossible
unless children describe these absent entities, a task
that involves decontextualized language. Children
who participate in these storytelling exchanges may
begin to produce richer narratives in general as a
result of these unique exchanges. The results of our
exploratory analyses were in line with this interpretation, in that only children whose mothers knew
about their imaginary companions showed a narrative advantage over children without imaginary
companions. Only a small number of children had
mothers who did not know about their companions, however, so additional research needs to be
conducted with children who do not share their
imaginary companions with parents to garner firm
support for this hypothesis.
Synergistic support for these hypothetical conversations can be found in Gleasons (2004b) work
on the correspondence in the reports about imaginary companions of parents and children who
engage in imaginary companion play. Gleason

Play and Narrative

found that parents descriptions of imaginary companions reached a higher level of agreement if a
childs imaginary companion was a classically
invisible rather than a personified object. Gleason
theorized that intangibility of invisible friends
may promote better efforts by parents to learn
about or to describe their childrens invisible
friends. Parents can see personified objects and
might even hear their children interacting with
them, meaning that neither parent nor child may
think to discuss a personified objects age or activities because both witness its existence (p. 211).
Recording actual conversations of this sort would
confirm their occurrence beyond the anecdotal
report of parents. It would also provide conversational evidence that could be used to support our
theory regarding childrens increased use of
decontextualized language when speaking about
their imaginary companions.
This stance echoes other theorists positions that
the developmental benefits of play are not solely a
result of engagement in play activities, but play can
also promote secondary interactions with added
developmental benefits (Pellegrini & Galda, 1993).
Although it would be difficult to execute such a
study, collecting childrens spontaneous narratives
about their imaginary companions would allow for
the elaboration of the proposed theory and potentially reveal how parents influence the development
of these childrens higher order language practices.
In previous analyses with this sample (e.g.,
Cleveland & Reese, 2005; Farrant & Reese, 2000), as
well as in other samples (e.g., Haden, Haine, &
Fivush, 1997), parents who adopt an elaborative,
open-ended questioning style during reminiscing
have children with stronger personal narrative
skills. A next step in our research program will be
to explore a range of parentchild storytelling practices in early childhood that support childrens
imaginary companion play and their narrative
skills. Given the enriched language environment
for firstborns (e.g., Jones & Adamson, 1987), parentchild storytelling practices may also help us
understand the higher incidence of imaginary companions and better narrative skills for firstborns in
our study. Once again, however, we found that the
link between imaginary companion status and narrative was independent of childrens birth order. A
richer language environment for firstborns cannot
completely account for our findings.
Because our main finding is limited to a concurrent correlation between childrens imaginary companion play and their narrative skills, however,
there is always the possibility that children with

1311

better language or narrative skills are driven to create an imaginary companion as an outlet for their
verbal expression, or that a third variable can
account for this relation. Although we have ruled
out childrens vocabulary skills in accounting for
the relation we found between imaginary companions and narrative, other potential candidates are
childrens personality characteristics or their perspective-taking skills. In earlier time points from
this longitudinal study, we found no unique relations between childrens early temperament at ages
13 and their imaginary companion play at age
5 once birth order had been taken into account
(McMenamin, 2000). Nor did we find any significant links between childrens earlier language or
understanding of mind at ages 3 and 4 years and
their imaginary companion play at age 5. These
null findings mirror those of Taylor et al. (2004),
the only other study to explore older childrens
imaginary companions in depth. In that study, childrens imaginary companion play at ages 67 was
not predicted by their earlier characteristics or theory of mind, nor was imaginary companion play
correlated with concurrent personality and emotion
understanding. It is possible that the imaginary
companions of older children arise through different processes and for different reasons from the
imaginary companions of early childhood. Taken
together with Taylor et al.s results, we propose
that imaginary companion play arises from a more
general predilection for fantasy. The reason that
different correlates of imaginary companion play
are found at different ages could result from the
necessary skills to support fantasy at that age (e.g.,
perspective-taking at younger ages and complex
storytelling at older ages). The developmental outcomes of engaging in imaginary companion play
at different ages could also be expected to differ.
Longitudinal research on imaginary companion
play and narrative in the early elementary school
years is needed to tease apart cause and effect in
this relation.
We acknowledge several other limitations to this
study. Our sample was relatively small, and we
were unable to conduct the more stringent doubleinterview process for identifying children with
imaginary companions (Taylor & Carlson, 1997;
although see Taylor et al., 2004). Despite these limitations, we were able to replicate several findings
from previous research on imaginary companion
play. For instance, our identified rate of 48% of 5year-old children with past or present imaginary
companions is in line with Taylor et al.s
(2004) identification of 49% in a sample of 6- to

1312

Trionfi and Reese

7-year-olds, and with Pearson et al.s (2001) identification of 46.2% in a sample of 5- to 12-year-olds.
Also similar to Taylor et al.s research with older
children, we found no significant gender difference
in the incidence of imaginary companions in our
school-age sample (but see Pearson et al., 2001). As
in other studies, there was no difference in the
mothers education levels as a function of their childrens imaginary companion play (e.g., Bouldin &
Pratt, 1999). Finally, similar to other studies, the
majority of the children in our study with imaginary companions were firstborns (e.g., Bouldin &
Pratt, 1999; Gleason et al., 2000).
In conclusion, this work reveals a connection
between childrens engagement in imaginary companion play and the qualitatively richer stories that
children tell about fictional and personally experienced events. The value of this finding is threefold.
First, this finding extends our understanding of
imaginary companion play and its developmental
correlates. Second, this finding deepens the literature on the theoretical relation between play and
language. Third, this finding highlights childrens
engagement in play as an important factor to consider in understanding their developing narrative
skill, which in turn is critical for childrens reading
skill.

References
Ames, L. B., & Learned, J. (1946). Imaginary companions
and related phenomena. Journal of Genetic Psychology,
69, 147167.
Baumer, S., Ferholt, B., & Lecusay, R. (2005). Promoting
narrative competence through adult-child joint pretense: Lessons from the Scandinavian educational practice of playworld. Cognitive Development, 20, 576590.
Bouldin, P. (2006). An investigation of the fantasy
predisposition and fantasy style of children with
imaginary companions. Journal of Genetic Psychology,
167, 1729.
Bouldin, P., Bavin, E. L., & Pratt, C. (2002). An investigation of the verbal abilities of children with imaginary
companions. First Language, 22, 249264.
Bouldin, P., & Pratt, C. (1999). Characteristics of preschool and school-age children with imaginary companions. The Journal of Genetic Psychology, 160, 397410.
Bouldin, P., & Pratt, C. (2002). A systematic assessment of
the specific fears, anxiety level, and temperament of
children with imaginary companions. Australian Journal
of Psychology, 54, 7985.
Bruner, J. (1990). Acts of meaning. Cambridge, MA: Harvard University Press.
Bruner, J. S. (1996). The culture of education. Cambridge,
MA: Harvard University Press.

Bunting, E. (1991). A perfect fathers day. Boston: Clarion.


Carlson, S. M., & Taylor, M. (2005). Imaginary companions and impersonated characters: Sex differences in
childrens fantasy play. Merrill-Palmer Quarterly, 51, 93
118.
Cleveland, E. S., & Reese, E. (2005). Maternal structure
and autonomy support in conversations about the past:
Contributions to childrens autobiographical memory.
Developmental Psychology, 41, 376388.
Cohen, J. (1988). Statistical power analysis for the behavioral
sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Cunningham, A., & Stanovich, K. (1997). Early reading
acquisition and its relation to reading experience and
ability 10 years later. Developmental Psychology, 33, 934
945.
Dickinson, D. K., & Snow, C. E. (1987). Interrelationships
among prereading and oral language skills in kindergartners from two social classes. Early Childhood
Research Quarterly, 1, 125.
Dickinson, D. K., & Tabors, P. O. (2001). Beginning literacy
with language. Baltimore: Brookes.
Dunn, L. M., & Dunn, L. M. (1997). Peabody Picture Vocabulary TestIII. Circle Pines, MN: American Guidance
Service.
Elley, W. B., & Irving, J. C. (1976). Revised socio-economic index for New Zealand. New Zealand Journal of
Educational Studies, 11, 2536.
Farrant, K., & Reese, E. (2000). Maternal style and childrens participation in reminiscing: Stepping stones
in autobiographical memory development. Journal of
Cognition and Development, 1, 193225.
Fein, G. G. (1989). Mind, meaning and affect: Proposals
for a theory of pretense. Developmental Review, 9, 345
363.
Garvey, C. (1977). Play. Cambridge, MA: Harvard
University Press.
Gleason, T. R. (2004a). Imaginary companions and peer
acceptance. International Journal of Behavioral Development, 28, 204209.
Gleason, T. R. (2004b). Imaginary companions: An evaluation of parents as reporters. Infant and Child Development, 13, 199215.
Gleason, T. R., Jarudi, R. N., & Cheek, J. M. (2003). Imagination, personality, and imaginary companions. Social
Behavior and Personality, 31, 721738.
Gleason, T., Sebanc, A., & Hartup, W. (2000). Imaginary
companions of preschool children. Developmental Psychology, 36, 419428.
Goncu, A. (1993). Development of intersubjectivity in the
dyadic play of preschoolers. Early Childhood Research
Quarterly, 8, 99116.
Goncu, A., & Katsarou, E. (2000). Constructing sociocultural approaches to literacy education. In K. A. Roskos
& J. F. Christie (Eds.), Play and literacy in early childhood
(pp. 221230). Mahwah, NJ: Erlbaum.
Griffin, T., Hemphill, L., Camp, L., & Wolf, D. P. (2004).
Oral discourse in the preschool years and later literacy
skills. First Language, 24, 123147.

Play and Narrative


Haden, C., Haine, R., & Fivush, R. (1997). Developing
narrative structure in parent-child conversations about
the past. Developmental Psychology, 33, 295307.
Harris, P. L. (2000). The work of the imagination. Malden,
MA: Blackwell.
Jones, C. P., & Adamson, L. B. (1987). Language use in
mother-child and mother-child-sibling interactions.
Child Development, 58, 356366.
Kavanaugh, R., & Engel, S. (1998). The development of
pretense and narrative in early childhood. In O.
Saracho & B. Spodek (Eds.), Multiple perspectives on play
in early childhood education (pp. 8099). Albany: State
University of New York Press.
Labov, W., & Waletzky, J. (1997). Narrative analysis: Oral
versions of personal experience. Journal of Narrative &
Life History, 7, 338. (Original work published 1967)
Manosevitz, M., Prentice, N. M., & Wilson, F. (1973). Individual and family correlates of imaginary companions
in preschool children. Developmental Psychology, 8, 72
79.
McMenamin, K. (2000). Predicting imaginary companions
in five-year-old children: The multivariate use of language, temperament, attachment, and birth order. Unpublished dissertation, University of Otago, Dunedin,
New Zealand.
Michaels, S. (1981). Sharing time: Childrens narrative
styles and differential access to literacy. Language and
Society, 10, 423442.
Newson, J., & Newson, E. (1968). Four years old in an
urban community. London: Allen & Unwin.
Nicolopoulou, A. (2002). Peer-group culture and narrative development. In S. Blum-Kulka & C. E. Snow
(Eds.), Talking to adults: The contribution of multiparty
discourse to language acquisition (pp. 117152). Mahwah,
NJ: Erlbaum.
Nicolopoulou, A. (2006). The interplay of play and narrative in childrens development: Theoretical reflections
and concrete examples. In A. Goncu & S. Gaskins
(Eds.), Play and development: Evolutionary, sociocultural,
and functional perspectives (pp. 247274). Mahwah, NJ:
Erlbaum.
ONeill, D. K., Pearce, M. J., & Pick, J. L. (2004). Preschool
childrens narratives and performance of the Peabody
Individualized Achievement TestRevised: Evidence of
a relation between early narrative and later mathematical ability. First Language, 24, 149183.
Pearson, D., Rouse, H., Doswell, S., Ainsworth, C.,
Dawson, O., Simms, K., et al. (2001). Prevalence of
imaginary companions in a normal child population.
Child: Care, Health and Development, 27, 1322.
Pellegrini, A. D. (1985). The relations between symbolic
play and literate behavior: A review and critique of the
empirical literature. Review of Educational Research, 55,
107121.

1313

Pellegrini, A. D., & Galda, L. (1993). Ten years after: A


reexamination of symbolic play and literacy research.
Reading Research Quarterly, 28, 162175.
Peterson, C., & McCabe, A. (1983). Developmental psycholinguistics: Three ways of looking at a childs narrative.
New York: Plenum.
Reese, E. (1995). Predicting childrens literacy from
mother-child conversations. Cognitive Development, 10,
381405.
Reese, E. (2002). Social factors in the development of
autobiographical memory: The state of the art. Social
Development, 11, 124142.
Reese, E., Haden, C. A., Baker-Ward, L., Bauer, P. J.,
Fivush, R., & Ornstein, P. (2008). Coherence of personal
narratives across the lifespan: A multidimensional model.
Manuscript under review.
Reese, E., Suggate, S., Long, J., & Schaughency, E.
(in press). Childrens oral narrative and reading skills in the
first three years of reading instruction. Reading and Writing.
Shore, C., OConnell, B., & Bates, E. (1984). First sentences
in language and symbolic play. Developmental Psychology, 20, 872880.
Singer, D. G., & Singer, J. L. (1990). The house of makebelieve: Childrens play and developing imagination. Cambridge, MA: Harvard University Press.
Snow, C. E. (1983). Literacy and language: Relationships
during the preschool years. Harvard Educational Review,
53, 165189.
Somers, J. U., & Yawkey, T. D. (1984). Imaginary play
companions: Contributions of creative and intellectual
abilities of young children. Journal of Creative Behavior,
18, 7789.
Sutton-Smith, B. (2001). The ambiguity of play. Cambridge,
MA: Harvard University Press.
Svendsen, M. (1934). Childrens imaginary companions.
Archives of Neurology and Psychiatry, 32, 985999.
Tamis-LeMonda, C. S., & Bornstein, M. H. (1994). Specificity in mother-toddler language-play relations across
the second year. Developmental Psychology, 30, 283292.
Taylor, M. (1999). Imaginary companions and the children
who create them. New York: Oxford University Press.
Taylor, M., & Carlson, S. M. (1997). The relation between
individual differences in fantasy and theory of mind.
Child Development, 68, 435455.
Taylor, M., Carlson, S. M., Maring, B., Gerow, L., &
Charley, C. M. (2004). The characteristics and correlates
of fantasy in school-age children: Imaginary companions, impersonation, and social understanding. Developmental Psychology, 40, 11731187.
Taylor, M., Cartwright, B. S., & Carlson, S. M. (1993). A
developmental investigation of childrens imaginary
companions. Developmental Psychology, 29, 276285.
Williams, K. (1997). Expressive Vocabulary Test. Circle
Pines, MN: American Guidance Service.