Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Research Article
When Do Children
Understand “Opposite”?
Catherine I. Phillipsa and Penny M. Pexmana
Purpose: The aims of the present research were to determine opposite. Children demonstrated this understanding only
(a) the age at which children with typical development when asked for the “opposite” one, suggesting that
understand the concept of opposite, (b) whether this is related antonymy was not made salient by stimulus properties
to other cognitive abilities or experiences, and (c) whether alone. Children’s accuracy was not significantly related
there is early implicit understanding of the concept. to their language or working memory skills, to their
Method: Children (N = 204) between 3 and 5 years of age child care experience, or to whether parents reported having
were randomly assigned to 1 of 3 experimental conditions in a books or games about opposites or playing opposite
novel opposite task. Children’s language and working memory word games with children. Eye gaze analyses provided no
skills were assessed, and parents provided information about evidence for early implicit understanding of the concept
children’s access to learning materials about opposites. of opposite.
Results: In the opposite task, 4- and 5-year-olds, but not Conclusion: Children with typical development have a
3-year-olds, demonstrated acquisition of the concept of concept of opposite by 4 years of age.
H
relations, such as synonymy. The challenge of abstracting
a marker of typical lexical and conceptual devel- similarities across exemplars of opposite is illustrated in
opment (e.g., American Academy of Child and recent computational modeling described by Chen, Lu, and
Adolescent Psychiatry, 2008). This concept involves the lex- Holyoak (2010). The antonym relation was particularly
ical semantic relation of antonymy, and acquisition of this hard for Chen et al.’s model to learn, and they attributed
knowledge is a developmental achievement that likely sup- this to the fact that the dimensions of meaning on which
ports, and is itself supported by, the acquisition of other antonyms differ vary across antonym pairs; they stated,
linguistic skills. Antonyms are pairs of words that are both “The shifting relevance of features makes learning a good
minimally and maximally different. They typically differ representation for antonyms challenging” (p. 876).
maximally but only on a single dimension (e.g., length or In the present work we focused on antonymous adjec-
age); therefore, their meanings are simultaneously similar tives. Though antonymous nouns (e.g., life–death) and
and different (Clark, 1970; Jones, 2002; Murphy, 2003; verbs (e.g., increase–decrease) also exist, we focused this
Owens, 2008). There is surprisingly little data on when investigation on adjectives because adjectives lend them-
children actually understand that these words are opposite. selves to manipulation and have been the focus of previous
The purpose of the present study was to address this issue. studies on this topic. Our approach in the current work is
In order to understand the concept of opposite, a child consistent with Murphy’s (2003) view of lexical relations as
must learn that the meanings of some words are related, metalinguistic concepts. Lexical semantic relations such as
that two words that form an antonymic pair have a binary antonymy constitute metalinguistic knowledge—conceptual
relationship with each other (Murphy, 2003), and the spe- knowledge about words and the ways word meanings are
cific relationship that the words have with each other in related. As such, acquisition of the concept of an opposite
order to distinguish antonymy from other lexical semantic can be demonstrated by correct identification of word pairs
that are opposites—that is, identification of word pairs that
share the semantic relation of antonymy. In the following
a sections we review the literature and explain how our research
University of Calgary, Alberta, Canada
builds on what has been done by addressing limitations
Correspondence to Penny M. Pexman: pexman@ucalgary.ca
in the metalinguistic studies of children’s understanding of
Editor: Rhea Paul
opposite that have so far been conducted.
Associate Editor: Jenny Roberts
Received August 11, 2014
Revision received November 14, 2014
Accepted April 25, 2015 Disclosure: The authors have declared that no competing interests existed at the time
DOI: 10.1044/2015_JSLHR-L-14-0222 of publication.
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1233–1244 • August 2015 • Copyright © 2015 American Speech-Language-Hearing Association 1233
Approaches to the Study of Antonym Acquisition who stated that they did not know the concept of opposite
actually did know the relation but did not know the concept
Empirical research assessing the development of the
label, and vice versa. This suggests that studies that rely
concept of opposite generally concludes that appreciation
extensively on children’s verbal responses to reflect under-
begins to develop around the age of 4 years (Clark, 1972;
standing of the word opposite, especially those that do not
Heidenheimer, 1975; Morris, 2003). However, there is also
provide any feedback during the course of the study (as in
some evidence of later (Kreezer & Dallenbach, 1929) and
Clark, 1972, and Morris, 2003), may be unnecessarily diffi-
earlier (Jones, 2007; Jones & Murphy, 2005; Murphy &
cult for children.
Jones, 2008) ages of acquisition and usage. These studies
Other relevant metalinguistic studies have examined
fall into one of two categories: metalinguistic studies and
development of semantic relations with free association
discourse studies.
tasks. These have shown that younger children tend to re-
spond with a word that is likely to follow the stimulus word
Metalinguistic Studies in a sentence (e.g., dark–night), whereas older children tend
Metalinguistic studies evaluate children’s understand- to respond with a word that could replace the stimulus
ing of antonym relations and their ability to work with the word (e.g., dark–light; Entwisle, Forsyth, & Muuss, 1964).
metalinguistic vocabulary of opposition. To date, these stud- This is referred to as the syntactic–paradigmatic shift, is
ies primarily have involved verbal games wherein the child thought to occur between the ages of 5 and 9 years, and is
responds to a question along the lines of “What is the oppo- found mainly for high-frequency adjectives (Nelson, 1977).
site of X?” The present research adopts a related approach. Free association responses involving antonym relations
The first study to investigate children’s appreciation are among the paradigmatic responses produced. The age
for the concept of opposite utilized a verbal response task of the shift may depend on the response task used (Jarman,
and was conducted by Kreezer and Dallenbach (1929). A 1980). Further, although it has been tempting to infer that
child was considered to know the concept of opposite if he the free association task reveals the structure of semantic
or she produced the corresponding antonym (e.g., fast in re- memory, Nelson (1977) noted that the shift may also be
sponse to slow) or if he or she utilized the strategy of nega- attributable to the child’s changing interpretation of the
tion (e.g., not fast). Kreezer and Dallenbach concluded that task, in particular “the child’s changing ability to deal with
50% of children could be taught the concept of opposite be- nonmeaningful verbal tasks” (p. 113).
tween the ages of 5;3 and 5;9 (years;months) and that a child The notion that antonym relations might be salient
could learn the concept without instruction between the for children only after 5 years of age could also be derived
ages of 8 and 9 years. from findings reported by Montgomery, Anderson, and
Further research revealed that the age of acquisition Uhl (2008). Montgomery et al. explored the mechanisms of
of the concept of opposite is probably younger than was interference in the day–night task and found that semantic
initially found (Clark, 1972; Heidenheimer, 1975; Morris, competition for opposites was not strong in their sample
2003). With a similar verbal task, Clark (1972) found of 3- to 5-year-olds. Instead, children’s difficulty with the
that children as young as 4 years could provide a correct day–night task was attributed to response set.
opposite response to some of the stimulus words and that
5-year-olds provided more correct responses than 4-year- Discourse Studies
olds. Morris (2003) also found that children between the ages There is also evidence that children may produce anto-
of 4 and 5 years were able to consistently generate opposites. nyms in their own speech before the age of 4 years. This
However, no effect of age was found; 4- and 5-year-olds suggestion is supported by several studies that identify the
appeared to understand the concept of opposite equally well. ways antonyms are used in discourse by children and adults.
Also in contrast with the learning found by Kreezer Whereas the research reviewed thus far used elicitation tech-
and Dallenbach (1929), Clark (1972) reported a distinct or- niques, these studies used corpus-based approaches and ex-
der of acquisition: When making a semantically appropriate amined the co-occurrence of antonym pairs in turns of speech.
substitution, children always replaced a difficult word with Spontaneous usage of antonyms was found in chil-
an easier word but never the reverse. For example, a child dren as young as 2 years of age (Jones & Murphy, 2005;
might replace tall with big in the tall–short word pair but Tribushinina et al., 2013). Results of discourse studies sug-
never the reverse (i.e., replacing big with tall in the big–small gest that children of all ages use most of the discourse func-
pair). There was a correlation between age and the number tions that have been identified in adult corpora. Overall,
of semantically appropriate responses, which Clark (1972) the discourse functions used by children were found to be
suggested was the result of children in the younger age simultaneously similar to and different from those used by
groups knowing the easier, less complex word pairs and none adults when speaking to children. Antonym use was always
of the more difficult word pairs. These results suggest that proportionally greater in child-produced speech than in
the younger children tested knew the concept of opposite child-directed speech and was found to increase significantly
but did not have a sufficiently developed vocabulary to suc- with age in child-produced speech; this change was not
cessfully perform the task with all word pairs. paralleled in child-directed speech (Jones, 2007; Jones &
Of consequence for all metalinguistic studies of oppo- Murphy, 2005; Murphy & Jones, 2008). On the basis of
site, Kreezer and Dallenbach (1929) found that some children these differences in the patterns of discourse function usage,
1234 Journal of Speech, Language, and Hearing Research • Vol. 58 • 1233–1244 • August 2015
the authors tentatively concluded that children are not merely opposite when it is evaluated with a task that does not re-
reproducing the discourse functions that they hear most quire that children produce a verbal response and that as-
often. sesses children’s understanding of the meaning of the word
Discourse studies also show variability between indi- opposite across several antonym pair exemplars. In addi-
vidual children in their ages and rates of antonym usage tion, we assessed the possibility that children might show
(Jones, Murphy, Paradis, & Willners, 2012; Tribushinina emerging sensitivity to the antonym relation before they
et al., 2013). Jones et al. (2012) noted that this variability show this understanding in their overt behavior and investi-
could be associated with parents’ usage and, possibly, chil- gated whether other cognitive developments or learning
dren’s access to books, organized child care, and other experiences are related to the development of the concept of
learning opportunities. Further, young children who use opposite.
more explicit contrast in their own speech (e.g., “Give me a The present study evaluated the development of the
big piece, not a small piece”) show stronger growth in ad- concept of opposite in 3-, 4-, and 5-year-olds using adjec-
jective production (Tribushinina et al., 2013). This use tives and a novel opposite task that assessed metalinguistic
of contrast is linked to parents’ tendencies to use the same understanding but did not require the child to make a ver-
strategy in their own speech. Contrastive contexts are bal response. By eliminating the need for a verbal response,
thought to be important to adjective learning because they we addressed one limitation of previous metalinguistic stud-
facilitate attention to the specific dimensions on which ad- ies of children’s understanding of the concept of opposite.
jectives differ (Murphy & Jones, 2008; Tribushinina, 2013). On each trial of the opposite task the child was presented
If children’s usage can be taken as a measure of their with three images. The experimenter labeled the images for
understanding, or at least the initial stages of their devel- the child, selected one image, and then, using different
opment of the concept of opposite, it is possible that children wording in each of three conditions, asked the child to
may display an understanding of the concept of opposite in make a selection. In the “opposite-label” condition, the im-
a metalinguistic task before the age of 4 years if the task de- ages were labeled with an adjective (e.g., big dog, small dog,
mands are either changed or decreased (Jones, 2007; Murphy happy dog), and the child was asked to choose “the oppo-
& Jones, 2008). On the other hand, young children’s use site one.” The two remaining conditions were both control
of antonymy in discourse simply may reflect learned asso- conditions in that we included them to assess alternative
ciations between words and be a developmental precursor explanations for children’s behavior in the opposite-label
to children’s metalinguistic understanding of opposite. Last, condition. In the “another-one-label” condition, the images
it is also possible that young children may understand the were similarly labeled, but the word opposite was not men-
lexical semantic relation of antonymy but not in a way that tioned; the child was simply asked to choose “another
is accessible at a conceptual level, which may be required one.” This condition assessed the possibility that provision
for correct responding in metalinguistic studies. of the adjectives alone (without instruction to provide the
Murphy and Jones (2008) suggested that the metalin- opposite) might be sufficient for children to perform above
guistic studies conducted thus far may have failed to find chance in the task. Last, in the “no-label” condition, the
evidence that young children have the concept of opposite images were labeled using basic-level category information
because the response tasks used have been too difficult. only (e.g., dog, dog, dog), and the child was again asked to
Although children may be able to use opposite words at a choose “another one.” This condition assessed the possibil-
young age (as revealed in discourse studies), their usage ity that provision of the images alone (without adjectives or
does not indicate that they have the metalinguistic aware- instructions to provide the opposite) might be sufficient for
ness of opposite and the appreciation that words are ob- children to perform above chance in the task.
jects, separate from their concrete and immediate context, On the basis of prior research (Clark, 1972;
which are required for success in word association and Heidenheimer, 1975; Morris, 2003), and the arguably re-
other metalinguistic tasks (Jones et al., 2012). duced task demands in this task relative to the tasks used
In addition, children’s performance in previous meta- in previous metalinguistic studies, it was hypothesized that
linguistic studies could have been hindered by the require- 5- and 4-year-olds, and perhaps 3-year-olds, would demon-
ment of a verbal response to a prompt (Vilette, 2002). strate a concept of opposite with above-chance perfor-
Children’s verbal recall of an event can lag behind nonver- mance in the opposite-label condition of the opposite task
bal measures of memory for the same event and can lag and that accuracy would improve with age. In contrast,
behind general verbal skill (Simcock & Hayne, 2003). Thus, it was hypothesized that children in all age groups would
there is the possibility that a more engaging and less ver- show chance performance in both control conditions unless
bally demanding task would reveal more accurately the the adjectives and images (in the another-one-label condi-
development of children’s understanding of opposite. This tion) or images alone (in the no-label condition) were suffi-
possibility was tested in the present study. cient to prompt correct responding.
We included eye gaze and latency measures so that
we could examine the process of understanding as well as
Present Research the product. There are numerous examples of research find-
The primary goal of the present study was to deter- ings showing dissociations between implicit or processing
mine the extent to which young children have a concept of measures (e.g., looking duration and response latencies)
1236 Journal of Speech, Language, and Hearing Research • Vol. 58 • 1233–1244 • August 2015
online supplemental materials, Supplemental Appendix A). Figure 1. Schematic of the testing setup for the opposite task.
The images were presented in sets of three (e.g., dirty pig,
clean pig, small pig), and within a set all images were the
same type of animal. Because preschool-aged children may
use visual similarity (e.g., shape, texture, color) to guide cate-
gorization of objects (Baldwin, 1992; Bonthoux & Kalenine,
2007; Morgan & Greene, 1994), care was taken to mini-
mize the visual similarity (e.g., color, posture, size) of the
images within sets so that no one image looked particularly
similar to the other two (e.g., black cat, white cat, orange cat).
The images depicted eight pairs (16 words) of anto-
nyms (see online supplemental materials, Supplemental
Appendix B). Two of the eight word pairs were used for
training, and the remaining six pairs were used for the test
trials; each word pair was presented once. Given our proce-
dure, we needed to select pairs that could be depicted in an-
imal images (rendering pairs such as hard–soft or alive–dead
less suitable). We selected candidate pairs from previous
discourse and metalinguistic studies on antonymy (Bracken,
1988; Clark, 1972; Heidenheimer, 1975; Jones & Murphy,
2005). The only pair we used that had not been used in a
previous study was awake–asleep, but because the words
awake and asleep are both listed in the MacArthur–Bates
Communicative Development Inventory (Words and
Sentences; Fenson et al., 2007) for 30-month-olds, we felt
reasonably confident that children in the present study
would know the meanings of those words. In addition,
we used the pair fat–thin, but in previous studies this was
thick–thin (Bracken, 1988; Clark, 1972; Heidenheimer, 1975).
We replaced thick with fat because many children in Clark’s
(1972) study made this substitution and because fat was a
more appropriate adjective to use with animal names.
In each set of images presented to the child, two of image placed last (or second) performed at chance. The lo-
the images depicted animals in opposite states and thus cation of the other two images was counterbalanced, such
were used to represent the antonymous adjectives (e.g., that in half of the trials the target was placed on the right
dirty–clean). The third image was used to represent an unre- side of the child. This balancing ensured that children who
lated distractor (e.g., small). The distractors were generated always chose the image on the right (or on the left) per-
from the eight pairs of adjectives. One member of each formed at chance. Following the labeling of the images, the
antonym pair was used once as a distractor, with the limi- experimenter put the middle image (e.g., clean pig) into
tation that the words short and tall were not used as the the box and labeled the image again while doing so. The
distractor for the big–small word pair and vice versa. The shallowness of the box ensured that children could see the
distractor was never part of the word pair used in the set selected image throughout the trial.
of images immediately preceding or following it. For some word pairs the target and distractor images
On each of the eight trials in the opposite task, a child were quite similar. For example, the small pig did not have
was presented with a set of three images (e.g., dirty pig, any dirt on it and could, therefore, also be considered a
clean pig, small pig). On the table was a shallow open box clean pig. If the experimenter placed the image of the dirty
positioned directly in front of the child (see Figure 1). All pig in the box, the child would have had to select between
sessions were recorded as digital video with the camera two images that depicted pigs with no dirt on them (i.e., the
angled to record the child’s eye gaze. clean pig and the small pig). To help mitigate this problem,
The experimenter introduced the task to the child the experimenter always placed the image that was con-
through two training trials. The experimenter placed a set fusable with the distractor image (in this example, the clean
of images (e.g., dirty pig, clean pig, small pig) on the table pig) into the box.
in front of the child and labeled each image while pointing Each child saw two training trials (presented in a
to it. The experimenter always placed the first image directly fixed order), which were followed by one of four possible
in front of the child. The placement of the second and third orders of presentation of the six test trials. In the opposite-
images was counterbalanced, such that on half of the trials label condition, the experimenter then asked the child to
the second image was placed to the left of the first image. place “the opposite one in the box” (e.g., dirty pig). This
This balancing ensured that children who always chose the was the signal to the child that it was his or her turn to
1238 Journal of Speech, Language, and Hearing Research • Vol. 58 • 1233–1244 • August 2015
was representative of the U.S. population as reported in t test for each of the conditions. Accuracy in the opposite-
the 2002 Current Population Survey (as cited in Bishop, label condition was significantly above chance, t(67) = 5.02,
2006). The examiner’s manual reports good test reliability p < .001. This effect was driven by both 5- and 4-year-olds
(Cronbach’s coefficient alpha, all rs > .75; Bishop, 2006). performing above chance: t(19) = 5.71, p < .001, and
t(27) = 3.31, p = .003, respectively. Three-year-olds performed
Parent Questionnaire at chance, t(19) = 0.00, p = 1.00. These tests remain sig-
Parents were asked to complete a brief demographic nificant even after applying a Bonferroni correction. In
questionnaire comprising nine questions about the child’s contrast, children’s accuracy was not significantly above
age, siblings, birth order, languages spoken in the household, chance in the another-one-label or no-label conditions (all
child care experience, and access to games and books about ps > .452).
the concept of opposite. Five of the questions were designed To evaluate whether children learned about the con-
to help us describe the sample, and four of the questions cept of opposite in the study (across the six study trials),
were designed to assess characteristics or experiences (i.e., children’s accuracy in the first set of three trials (Block 1;
type of child care attended, age starting child care, avail- M = 0.52, SD = 0.31) was compared with their accuracy in
ability of books or games on opposites, and experience the second set of three trials (Block 2; M = 0.57, SD = 0.30)
playing word games about opposites) that might be related with a 2 (block: 1, 2) × 3 (age: 3, 4, and 5 years old) × 3
to understanding the concept of opposite. (condition: opposite-label, another-one-label, and no-label)
mixed design ANOVA. Accuracy on the two blocks of trials
was not significantly different, F(1, 195) = 2.39, p = .124.
Results Furthermore, accuracy as a function of block did not inter-
Opposite Task Accuracy act with either condition or age (all ps > .903). Together,
this suggests that children’s accuracy did not improve over
Children’s proportions of target selections (i.e., num- the course of the opposite task, suggesting that children
ber of times the target image was selected divided by the did not learn about the concept of opposite from the oppo-
number of valid trials; hereafter referred to as accuracy) in site task itself.
each of the conditions, as a function of age, are presented
in Table 1.
Accuracy was evaluated with a 3 (age: 3, 4, and Correlations and Demographics
5 years) × 3 (condition: opposite-label, another-one-label, and Children’s scores on the measures of working mem-
no-label) between-subjects analysis of variance (ANOVA). ory, receptive vocabulary (TELD-3 raw score), and com-
There was a main effect of age, F(2, 195) = 8.87, p < .001, munication skills (CCC-2 scales) are presented in Table 2,
and a main effect of condition, F(2, 195) = 13.15, p < .001. Of Table 3, and Table 4. We evaluated relationships between
most importance, there was a marginally significant Age × children’s performance in the opposite-label condition of
Condition interaction, F(4, 195) = 2.27, p = .063. Follow-up the opposite task and these measures with partial correlations,
simple main-effect tests for each condition revealed the correcting for children’s age—that is, we tested five corre-
nature of this interaction. In the opposite-label condition, lations, none of which were significant after Bonferroni cor-
there was a significant main effect of age, F(2, 68) = 9.88, rection was applied.
p < .001. Five-year-olds were more accurate than 4-year- We next examined relationships between children’s ac-
olds, t(65) = 2.47, p = .016, and 3-year-olds, t(65) = 4.45, curacy (i.e., proportion of correct responses) in the opposite-
p < .001. Four-year-olds also were more accurate than label condition, and responses to four of the questions
3-year-olds, t(65) = 2.33, p = .023. Fisher’s least significant on the demographic questionnaire were evaluated with
difference correction was applied to these t tests and to all chi-square analyses and Pearson correlations. The one sig-
subsequent follow-up t tests unless otherwise noted. There nificant association that survived Bonferroni correction was
was no significant main effect of age in the another-one- between children performing above chance (four or more
label or no-label conditions (all ps > .141). trials out of six) on the opposite task and parents reporting
Children’s accuracy relative to chance performance having books or games in the home that focus on opposites,
(i.e., 50% target image selection, or three trials out of six) in c2(1) = 6.12, p < .05. That is, children whose parents re-
each condition was evaluated with a two-tailed, single-sample ported that they had books or games focused on opposites
were more likely to perform above chance than those who
Table 1. Children’s mean proportions of correct target item selections did not have access to these materials in the home. It is
in each condition. important to note, however, that there was a trend for chil-
dren with access to books or games on opposites (M = 51.66,
Opposite-label Another-one-label No-label SD = 9.29) to be older than those without access (M = 46.09,
Age
(years) M SD M SD M SD SD = 7.96). To control for this variability, we conducted
a hierarchical logistic regression in which the criterion vari-
3 0.50 0.16 0.44 0.15 0.46 0.13 able was whether children performed above chance on the
4 0.64 0.22 0.54 0.19 0.48 0.21
5 0.79 0.23 0.55 0.22 0.51 0.20
opposite task. As predictors, we entered child’s age on the
first step and access to books or games on the second step.
Age (years)
3 4 5 Overall
Variable M SD M SD M SD M SD
Object memory 6.00 5.71 14.35 5.00 21.79 3.57 15.23 7.33
TELD-3 22.70 4.41 27.78 4.41 30.80 3.62 27.16 5.22
CCC-2 GCC — — 78.70 15.85 80.26 11.70 79.35 14.16
CCC-2 Syntax — — 3.41 3.84 2.21 2.84 2.91 3.48
CCC-2 Context — — 6.59 3.38 5.79 2.37 6.26 3.00
Age started preschool or day care (years) 1.25 0.62 2.65 0.81 2.65 0.86 2.21 1.00
Children in family (N) 2.05 0.52 2.15 0.78 2.45 0.83 2.22 0.74
In child care (%) 75 54 55 60
Have access to books or games about 70 82 95 82
opposites (%)
Have experience playing word games 33 52 70 47
with opposites (%)
Note. The Children’s Communication Checklist–2 (CCC-2) is not intended for use with children under the age of 4 years. Em dashes indicate
data not obtained. TELD-3 = Test of Early Language Development–Third Edition; GCC = general communication composite.
With the child’s age statistically controlled in this way, the image than older children despite being just as likely as older
relationship between access to games or books and chil- children to look first to the target image.
dren’s accuracy on the opposite task was marginal (p = .068)
and thus should not be overinterpreted. Response Latencies
Children’s response latencies were evaluated with a 3
(age: 3, 4, and 5 years old) × 3 (condition: opposite-label,
Opposite Task Response Latency and Eye Gaze another-one-label, and no-label) between-subjects ANOVA.
Eye gaze and latency measures were included to eval- Mean response latencies (in ms) are presented in Table 5.
uate whether children showed evidence of an implicit un- There was only a significant effect of condition, F(2, 130) =
derstanding of opposite before they are able to perform 5.28, p = .006. Children in the opposite-label condition
above chance. In particular, we tested the hypothesis that had longer average response latencies than children in the
young children struggle with the metalinguistic task de- another-one-label condition, t(130) = 2.36, p = .020, and
mands of selecting the opposite one (e.g., Jones et al., 2012) children in the no-label condition, t(130) = 3.33, p = .001.
and that this might be evident in their eye gaze and re- The response latencies of children in the another-one-label
sponse latencies. In short, younger children may be slower condition did not differ from those of the children in the
to respond and may have more total looks at the target no-label condition, t(130) = 1.14, p = .258. This suggests
Table 3. Children’s mean raw scores on additional measures and demographic characteristics in the another-one-label condition.
Age (years)
3 4 5 Overall
Variable M SD M SD M SD M SD
Object memory 14.80 4.02 15.52 5.12 21.16 2.54 17.30 5.00
TELD-3 22.90 4.76 28.96 2.76 28.95 2.87 27.18 4.42
CCC-2 GCC — — 80.22 15.71 79.00 15.44 79.70 15.44
CCC-2 Syntax — — 2.85 2.33 3.40 3.78 3.09 3.00
CCC-2 Context — — 6.89 2.90 5.20 2.95 6.17 3.01
Age started preschool or day care (years) 1.64 0.74 2.68 0.72 2.47 0.87 2.34 0.88
Children in family (N) 1.50 0.63 2.04 0.74 2.45 0.76 2.03 0.80
In child care (%) 55 57 60 57
Have access to books or games about 70 82 95 82
opposites (%)
Have experience playing word games 60 79 75 73
with opposites (%)
Note. The Children’s Communication Checklist–2 (CCC-2) is not intended for use with children under the age of 4 years. Em dashes indicate
data not obtained. TELD-3 = Test of Early Language Development–Third Edition; GCC = general communication composite.
1240 Journal of Speech, Language, and Hearing Research • Vol. 58 • 1233–1244 • August 2015
Table 4. Children’s mean raw scores on additional measures and demographic characteristics in the no-label condition.
Age (years)
3 4 5 Overall
Variable M SD M SD M SD M SD
Object memory 15.22 3.56 18.54 3.40 20.95 4.27 18.87 4.19
TELD-3 23.15 4.58 28.04 3.74 31.25 2.29 27.54 4.80
CCC-2 GCC — — 70.96 13.77 85.53 13.39 76.98 15.29
CCC-2 Syntax — — 4.85 3.44 1.84 2.54 3.61 3.42
CCC-2 Context — — 8.78 4.37 5.16 2.29 7.28 4.04
Age started preschool or day care (years) 1.27 0.70 2.34 0.90 2.76 1.09 2.24 1.06
Children in family (N) 2.11 0.81 1.81 0.74 2.35 0.88 2.06 0.82
In child care (%) 45 54 50 50
Have access to books or games about 95 82 90 88
opposites (%)
Have experience playing word games 56 52 75 60
with opposites (%)
Note. The Children’s Communication Checklist–2 (CCC-2) is not intended for use with children under the age of 4 years. Em dashes indicate
data not obtained. TELD-3 = Test of Early Language Development–Third Edition; GCC = general communication composite.
that children were engaged in a different manner of process- There was a significant effect of age, F(2, 191) = 5.26,
ing in the opposite condition (vs. the other conditions, in p = .006; a significant effect of condition, F(2, 191) = 6.60,
which the word opposite was not used). As such, this result p = .002; and a significant interaction between age and
provides evidence that children carried out additional pro- condition, F(4, 191) = 3.32, p = .012. Follow-up simple
cessing in the opposite condition; presumably, that process- main-effect tests for each condition revealed the nature of
ing is what is required to evaluate the opposite relationship. this interaction. In the opposite-label condition there was
All other effects were not significant (ps > .598). The null a significant main effect of age, F(2, 62) = 9.95, p < .001.
effects of age in this analysis do not support the prediction Five-year-olds made more correct first looks than 4-year-olds,
that younger children would be slower to respond than t(62) = 3.21, p = .002, and 3-year-olds, t(62) = 4.32, p < .001.
older children due to sensitivity to the antonym relation but Four-year-olds did not differ from 3-year-olds, t(62) =
challenges with the metalinguistic task demands. 1.50, p = .139. As such, these results provide little evidence
that younger children are drawn to the correct image in the
Eye Gaze Fixations opposite-label condition. There were no significant effects
Children’s mean proportions of first looks to the cor- of age in the other two conditions (ps > .796).
rect image (target) were evaluated with a 3 (age: 3, 4, and
5 years old) × 3 (condition: opposite-label, another-one-label, Eye Gaze Total Looks
and no-label) between-subjects ANOVA. Mean proportions Eye gaze was also examined in terms of duration of
of first looks to the correct image are presented in Table 5. time spent looking at the target as a proportion of the total
Table 5. Children’s mean response latencies, mean proportions of first looks, and mean proportions of time spent looking at the target image
for each condition.
1242 Journal of Speech, Language, and Hearing Research • Vol. 58 • 1233–1244 • August 2015
of the eye gaze measures: The older children were more 3-year-olds is now supported by empirical test. The results
likely to look first at the correct image, and the youngest allow us to conclude that the majority of children under the
children in our sample did not take longer to respond or age of 4 years do not yet have a metalinguistic concept of
look more often at the correct target image. Thus, it ap- opposite or demonstrate sensitivity to the concept in their
pears that only the older children demonstrated under- processing. The results do not provide clear evidence that
standing of opposite, through overt means (correct target children’s cognitive skills or experiences are related to their
selections) and in processing (more correct first looks to tar- understanding of opposite. As such, there is more to be
gets). In speculating about why we did not find evidence done in future research to identify the key factors that
of implicit appreciation in young children’s processing, at help children learn the concept of opposite and to suggest
least two explanations seem possible. One interpretation is intervention strategies for children who struggle with the
that understanding of opposite, unlike other cognitive skills, concept.
does not develop implicitly first. Of course, it is also possi-
ble that implicit awareness is present in younger children
but that our measures were simply not sensitive enough to Acknowledgments
detect this awareness. We would lean toward the first inter- This research was supported by a Social Sciences and
pretation because the eye gaze measures did show sensi- Humanities Research Council postgraduate doctoral scholarship
tivity to some differences in processing, but we cannot rule to Catherine I. Phillips and a Social Sciences and Humanities
out the second explanation. Research Council Discovery Grant (435-2013-0096) to Penny M.
A limitation of the current design of the opposite task Pexman.
is that, in principle, a child could perform above chance
simply by understanding that there is some kind of associa-
tion between the words that form the antonym pair without References
necessarily understanding that the association is one of American Academy of Child and Adolescent Psychiatry. (2008).
opposite. This is possible because the distractor is unrelated Facts for families. Washington, DC: Author.
to the word pair. Although children could be simply notic- Baldwin, D. A. (1992). Clarifying the role of shape in children’s
ing an association between the word pairs, evidence from taxonomic assumption. Journal of Experimental Child Psychol-
the another-one-label condition in the present study suggests ogy, 54, 392–416. doi:10.1016/0022-0965(92)90027-4
Bishop, D. V. M. (2006). Children’s Communication Checklist–2.
that this is not likely the case. When children were presented
San Antonio, TX: Pearson Education.
with the adjectival labels but not the concept name (oppo- Bonthoux, F., & Kalenine, S. (2007). Preschoolers’ superordinate
site), they did not perform above chance; the adjectival labels taxonomic categorization as a function of individual process-
themselves were not sufficient for above-chance perfor- ing of visual vs. contextual/functional information and object
mance. If the children in the opposite-label condition were domain. Cognition, Brain & Behavior, 11(4), 1–20. doi:10.3758/
simply noticing the relationship between the words in an PBR.15.3.66
antonym pair, the children in the another-one-label con- Bracken, B. A. (1988). Rate and sequence of positive and negative
dition should have also performed above chance. That they poles in basic concept acquisition. Language, Speech, and
Hearing Services in Schools, 19, 410–417.
did not suggests that children in the opposite-label condi-
Case, R., Kurland, D. M., & Goldberg, J. (1982). Operational effi-
tion were paying attention to the specific antonymic relation- ciency and the growth of short-term memory span. Journal of
ship mentioned in the instructions. Experimental Child Psychology, 33, 386–404.
A further limitation of the current study was that Chen, D., Lu, H., & Holyoak, K. J. (2010). Learning and generali-
there could have been variability in the pictures we chose in zation of abstract semantic relations: Preliminary investigation
terms of their ability to convey the meaning of each label. of Bayesian approaches. In S. Ohlsson & R. Catrambone (Eds.),
That is, the picture of the old rabbit may have been a less Proceedings of the 32nd Annual Conference of the Cognitive
fitting illustration of that label than the dirty pig was of its Science Society (pp. 871–876). Austin, TX: Cognitive Science
label. We are reassured somewhat by the fact that we pre- Society.
Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb
sented children with multiple trials and by the fact that,
for evaluating normed and standardized assessment instru-
across children, items were used as both targets and distrac- ments in psychology. Psychological Assessment, 6, 284–290.
ters. Nonetheless, our findings may be influenced to some doi:10.1037/1040-3590.6.4.284
degree by our choice of stimuli. Clark, E. V. (1972). On the child’s acquisition of antonyms in two
semantic fields. Journal of Verbal Learning and Verbal Behavior,
11, 750–758. doi:10.1016/S0022-5371(72)80009-4
Conclusions Clark, H. H. (1970). The primitive nature of children’s relational
The results of the present study provide several novel concepts. In J. R. Hayes (Ed.), Cognition and the development
contributions. To evaluate children’s understanding of op- of language (pp. 269–278). New York, NY: Wiley.
Daneman, M., & Carpenter, P. A. (1980). Individual differences in
posite, we developed a task that allowed collection of both working memory and reading. Journal of Verbal Learning and
implicit and explicit behavioral responses. In addition, to Verbal Behavior, 19, 450–466.
our knowledge this study is the first metalinguistic study Entwisle, D. R., Forsyth, D. F., & Muuss, R. (1964). The syntactic-
that includes 3-year-olds. Therefore, the assumption made paradigmatic shift in children’s word associations. Journal of
by previous researchers about the lack of understanding in Verbal Learning and Verbal Behavior, 3, 19–29.
1244 Journal of Speech, Language, and Hearing Research • Vol. 58 • 1233–1244 • August 2015
Copyright of Journal of Speech, Language & Hearing Research is the property of American
Speech-Language-Hearing Association and its content may not be copied or emailed to
multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.