Language Competence and Social Preference in Childhood

Language Competence

Oral language competence includes the understanding and production of linguistic utterances and is regarded as a complex system of rules consisting of semantic, syntactic, morphologic, and pragmatic facets of language competence (Saxton, 2010). Developmental psychology often focuses on the difference between receptive and expressive language competence (e.g., Barre, Morgan, Doyle, & Anderson, 2011). This distinction stems from the fact that in communication, there is always a sender and a receiver (McLaughlin, 2006). Whereas the sender encodes, expresses, and produces language, the listener recodes, receives, and comprehends language. In the present meta-analysis we took up this distinction, dividing oral language competence into receptive and expressive language competences.

There are various methods of assessing language competence. In psychological research, language competence is often assessed by language inventories and language tests. Parent and teacher reports, as well as standardized analyses of conversation sequences in natural settings, were used particularly in earlier research on language acquisition and when studying younger children. In the present meta-analysis the whole range of measures (e.g., tests, inventories, mean length of utterances) was included in order to cover as broad a spectrum of oral language competence as possible.

Social Preference

In the present study we focused on the single dimension of social preference, which is part of the multidimensional and broader construct peer relationships. Social preference refers to social status, or the extent to which someone is accepted or not accepted by his or her peers (Coie, Dodge, & Coppotelli, 1982). This differs from competencies through which a child may have acquired this status.

There are various ways to assess social preference. These can include peer nominations, sociometric ratings, number of friends, and evaluations of social acceptance, social rejection, and popularity as rated by peers, teachers, parents, the research assistant, or the respective child, all of which were included in the present analysis. In accordance with the meta-analysis by Dougherty (2006), we were interested in the child’s general popularity within the group and less on classifications of social status (e.g., popular, rejected, controversial, neglected).

Age of Participants

Throughout childhood, peer interactions change dramatically with development. The question thus arises of whether the relation between language competence and social preference is moderated by the child’s age. In preschool, interactions between children are highly physical in nature (Alink et al., 2006). This is evident in situations of conflict, for example, where younger children resort to physical aggression more quickly, whereas older children tend to use verbal means to assert themselves. Accordingly, language competence might become more relevant in social interaction with increasing age.

However, Hay et al. (2004) posit that language competences are highly relevant for peer interactions in early childhood, whereas with increasing age other factors such as shyness, aggressive behavior, or prosocial skills are important for gaining acceptance by peers. Moreover, among older children, oral language competence shows less interindividual variation and should be sufficiently developed in most children to enable effective communication in everyday life (McLaughlin, 2006). There is much more variability in the oral language competence of younger children, which makes it more likely to be predictive for social preference in younger children than in older children. Therefore, we posited that with increasing age, the interrelations between language competence and social preference decrease.

In the present meta-analysis, we concentrated on children between 2 and 11 years. This age range was selected because peer preferences only become visible from toddlerhood upwards (Howes & Phillipsen, 1992) and we wanted to restrict the study to childhood.

Gender of Participants

It is also possible that the relation between language competence and social preference differs depending on gender. For example, Stowe, Arnold, and Ortiz (1999) showed that the relation between language competence and difficulties with peers was more pronounced for boys than for girls. Stowe et al. also demonstrated that compared to girls, boys with low language competences react more frequently with externalizing problem behavior in learning situations. Externalizing problem behavior is visible and thus is more likely to manifest itself in problems in social interactions. However, Von Grünigen and colleagues’ (2010) study yield different results: Although gender was also a moderator of the relation between language competence and peer acceptance, importantly, this relation was stronger for girls than for boys.

Overall, findings on the influence of gender on the relation between language competence and peer interactions are contradictory. Accordingly, we did not posit a specific direction of the impact of gender, but still examined it as a possible moderator of the relation between language competence and social preference.

Language Modality

As detailed above, language competence is a complex construct. It is not entirely understood how receptive and expressive language competences relate to each other (McLaughlin, 2006). For example, in the large cross-national study by Bornstein and Hendricks (2012) correlations between receptive and expressive language competence were positive and significant, similar for different age groups, but varied in size across countries. On average, receptive and expressive language competences were correlated, but with only small to medium effects. Also, receptive and expressive language development involve different processes (McLaughlin, 2006). Therefore, we were interested in whether expressive language competence and receptive language competence are related to gaining acceptance by peers to a similar extent, that is, whether language modality moderates the relation between language competence and social preference.

Methodological Characteristics

Finally, methodological characteristics of a given study, such as the use of age-norms for language and the dimension of social preference used, may also serve as moderators of the relation between language competence and social preference.

First, age-normed values of language competence may influence the strength of the relation between language competence and social preference. The use of age-normed values instead of raw scores is considered to be a quality feature of a study. However, in some study designs (e.g., observations) or in some samples (e.g., bilingual children), age-normed values are nonexistent or the transformation of raw scores creates further difficulties. Nevertheless, it is possible that the strength of the relation between language competence and social preference varies with regard to the use of age-norms, therefore we tested the use of age-norms of language competence scores as a moderator.

Second, it is possible that the assessed social preference dimension moderates the relation between language competence and social preference. As mentioned above, there are several ways to assess social preference, leading to related, yet different aspects or dimensions of peer relationships (e.g., social acceptance, social rejection; Coie et al., 1982).

Literature Search

We used multiple search strategies to identify relevant studies published up to December 31, 2012. First, we scanned the databases PsycINFO, ERIC, PSYNDEX, PsycCRITIQUES, PsycTESTS, Ovid Medline, ISI Web of Science, and LLBA using the following keywords: peer relationship, social preference, social status, sociometric, social interaction, social inclusion, popularity, likability, rejection, rejected, peer acceptance, adjustment, adaption, social problem, and peer problem for the concept social preference. The following keywords were used to search for the topic of language competence: language, communication, verbal ability, verbal test, linguistic, cognition, cognitive test, cognitive ability, intelligence, IQ, mental age, cognitive development, achievement, academic ability, educational status, speech, proficiency, emotion knowledge, false belief, and theory of mind for the concept language competence. The keywords emotion knowledge, false belief, and theory of mind were included because language competence is often used as a covariate in studies investigating these concepts. Second, we sifted through the reference lists of all studies previously identified. Third, we used the above-mentioned databases to identify studies citing the previously identified studies. In cases where relevant information about study design or sample characteristics was missing, authors were contacted and asked for additional information.

Inclusion and Exclusion Criteria

A study had to meet the following standards for inclusion in the meta-analysis: (a) the study had to investigate the relation between language competence and social preference, (b) only studies reported in English were included, (c) the average age of the participants studied had to be between 2 and 11 years, (d) the study had to contain sufficient information to calculate an effect size for the relation between language competence and social preference (see Lipsey & Wilson, 2001), (e) given the small number of longitudinal studies examining the relation between language competence and social preference, we included only studies reporting effect sizes based on the concurrent association between the two variables (maximum time frame of 6 months between the assessments). Studies were included regardless of publication form and year of publication.

Meeting the following criteria led to exclusion from the meta-analysis: Interventional studies were excluded when the effects of the relation between language competence and social preference before intervention were missing. Moreover, we excluded studies with participants who were retrieved from a clinical or a special population such as children with language impairment.

Forty-five studies met inclusion criteria and were not excluded due to the exclusion criteria. In some cases, the results were published in more than one article. To avoid duplication, in cases where articles published data from the same population with the same measures, the more recently published article was included (two studies). In cases where articles were published with the same measures by the same research group, but with different subsamples, the article with the larger sample was included (one study). The remaining 42 studies were included in the meta-analysis. If a same sample was used in a longitudinal study, correlations from wave 1 were included, but not from the other waves. In studies that reported data on different measures of language competence or social preference but from the same sample, effect sizes were averaged across studies to compute a single mean effect size.

Coding

All studies were coded with regard to the following features:

• •
  Form of publication (i.e., journal article: peer-reviewed; journal article: not peer-reviewed; book; book chapter; dissertation; technical report; conference paper; unpublished manuscript);
• •
  Year of publication;
• •
  Country of publication;
• •
  Sample size;
• •
  Mean age of participants in years;
• •
  Proportion of females in the sample;
• •
  Social preference dimension: social acceptance (e.g., most liked, popular, number of friends)/social rejection, (e.g., least liked)/composite of social acceptance and social rejection;
• •
  Source of information for social preference (peers/child/teacher/parents/research assistant);
• •
  Instrument to assess social preference;
• •
  Language modality: receptive language competence (decoding, reception, comprehension)/expressive language competence (encoding, expression, production)/mixed);
• •
  Source of information for language competence (peers/child/teacher/parents/observation);
• •
  Instrument to assess language competence;
• •
  Use of age-normed values to transform raw scores of language competence test results;
• •
  Concurrent correlation coefficient for the relation between language competence and social preference.

All studies that met inclusion criteria were coded for the above study features by the first and the second authors of this meta-analysis. The interrater agreement was κ ≥ .88 for categorical variables (Cohen, 1960) and ICC ≥ .997 for continuous variables (two-way mixed, absolute, average-measures ICC; Hallgren, 2012) indicating a high level of agreement between coders. All diverging assessments were discussed and consensus was reached.

For meta-analyses that examine the strength of the relation between two continuous variables, Pearson product-moment correlation coefficients (r) are usually used as effect size (Field & Gillett, 2010). Thus, in the present meta-analysis, r was used to assess the strength of the relation between the two continuous variables, language competence and social preference. Most studies (84%) reported direct estimates of r. If the correlation coefficient was not reported, we transformed the given effect size using an effect size calculator (see Lipsey & Wilson, 2001). In one case, the study published raw scores (Gertner et al., 1994), which were entered and computed as correlation coefficients.

Meta-Analytic Procedure

We transformed each r into a Fisher’s z score using study weights with ω = n − 3 (see Lipsey & Wilson, 2001). Effect sizes were analyzed using the random effects model, which is appropriate when effect sizes are heterogeneous (Borenstein, Hedges, Higgins, & Rothstein, 2009). SPSS and SPSS macros by Lipsey and Wilson (2001) were used for computation of effect sizes.

Several preliminary analyses were conducted. First, we tested for outliers on the effect size variable. Secondly, we addressed the issue of publication bias in meta-analytic studies. We hypothesized that publication bias was not an issue in the present meta-analysis because the relation between language competence and social preference was not the main focus of most studies, but rather reported interrelations among all study variables. Moreover, we did not apply any restrictions to publication form. Nevertheless, we tested for publication bias assuming that large studies have a higher probability of getting published. On the other hand, studies with low effect sizes should have a lower probability of being published if the sample size is small. The relation between study size and effect size was examined using a funnel graph (Sutton, 2009).

Homogeneity statistic (Q) of effect sizes was examined to assess the variation of the true effect size. The Q statistic is based on a chi-square distribution. Significant heterogeneity indicates that effect sizes vary across studies, which may be explained by study characteristics. As a consequence, moderator analyses are conducted to investigate differences between sample sizes. In the present meta-analysis we focused on age, gender, and language modality as potential moderators. Moreover, we examined whether further methodological characteristics explained variation between studies.

Descriptive Statistics of the Studies Used in the Meta-Analysis

Forty-two studies, with a combined sample of 7,077 children, met the inclusion criteria. Sample sizes varied between N = 19 and 1,090 (M = 144.43, SD = 209.05). The resulting total sample consisted of 49 independent samples (k) yielding 90 effect sizes indicating the magnitude of the relation between language skills and social preference (see Table 1).

Table 1 Articles included in meta-analysis

Study	N	Age	% of females	Language modality	Social preference measure	Effect size (r)
Note. n.r. = Not reported.
Badenes, Estevan, and Bacete (2000)	77	5.58	35.06	Mixed	Composite	.30
Banerjee, Rieffe, Terwogt, Gerlein, and Boutsina (2006), Study 2	60	9.50	41.67	Receptive	Composite	.11
Banerjee, Watling, and Caputi (2011) Older sample	138	11.00	43.48	Receptive	Acceptance, rejection	.05
Banerjee, Watling, and Caputi (2011) Younger sample	72	7.98	56.94	Receptive	Acceptance, rejection	.13
Braza et al. (2009)	98	5.25	56.12	Expressive	Composite	.27
Burleson et al. (1986)	59		46.67	Expressive	Composite	.25
Caputi et al. (2012)	70	7.02	44.29	Receptive	Acceptance, rejection	.15
Cassidy, Werner, Rourke, and Zubernis (2003)	67	4.33	52.24	Mixed	Acceptance, composite	.46
Champion, Lowe, and Cavior (1981)	96	9.04	50.00	Receptive	Acceptance	.16
				Expressive
Coughlin and Vuchinich (1996)	194	9.70	0.00	Receptive	Composite	.22
Curby, Rudasill, Rimm-Kaufmann, and Konold (2008)	347	n.r.	56.48	Receptive	Acceptance	.28
Deutsch (1974)	60	4.00	100.00	Expressive	Acceptance	.22
Doyle, Rappard, and Connolly (1980) Sample 1	31	4.00		Receptive, expressive	Acceptance	.34
Doyle, Rappard, and Connolly (1980) Sample 2	31	4.00	n.r.	Receptive, expressive	Acceptance	.34
Fabes, Eisenberg, Hanish, and Spinrad (2001)	50	4.95	56.00	Expressive	Composite	.41
Flynn and Whiten (2012)	88	3.50	57.95	Receptive	Acceptance, rejection	.21
Gertner et al (1994)	19	4.68	47.37	Receptive, expressive	Acceptance, rejection	.32
Goldman, Corsini, and De Urioste (1980)	38	4.50	42.11	Expressive	Acceptance, rejection, composite	.30
Gulay (2011)	236	5.51	47.88	Mixed	Rejection	.26
Hoglund, Laloonde, and Leadbeater (2008)	114	n.r.	50.00	Expressive	Rejection	.00
Krantz (1982)	47	n.r.	n.r.	Expressive	Acceptance	.01
Ladd (1990)	125	5.35	47.20	Receptive	Acceptance, composite	.25
Ladd, Birch, and Buhs (1999), Study 1	200	5.58	47.50	Receptive	Composite	.26
Ladd, Birch, and Buhs (1999), Study 2	199	5.47	51.76	Receptive	Acceptance, rejection	.18
Mathieson and Banerjee (2011), Boys	31	5.08	0.00	Receptive	Acceptance	.13
Mathieson and Banerjee (2011), Girls	27	5.08	100.00	Receptive	Acceptance	.23
Meece and Mize (2010), Boys	64	4.78	0.00	Receptive	Composite	.25
Meece and Mize (2010), Girls	64	4.78	100.00	Receptive	Composite	.03
Mostow, Izard, Fine, and Trentacosta (2002)	201	7.47	46.77	Expressive	Composite	.09
Murphy and Faulker (2006), Boys	24	6.00	0.00	Expressive	Composite	.12
Murphy and Faulker (2006), Girls	24	6.00	100.00	Expressive	Composite	.36
Nearland (2011)	64	3.00	62.50	Expressive	Composite	.46
Olson and Lifgren (1988)	79	4.67	40.51	Receptive	Acceptance, rejection, composite	.18
Peterson and Siegal (2002)	109	4.67	40.37	Receptive	Acceptance, rejection	.17
Roff and Sells (1965)	862	n.r.	48.61	Mixed	Composite	.51
Rosenthal (1957)	40	7.83	47.50	Expressive	Acceptance	.29
Rubin (1973)	80	8.63	50.00	Receptive, expressive	Acceptance	.08
Rubin and Danielsbeirness (1983)	72	5.42	50.00	Receptive	Composite	.40
Schneider (2008)	52	6.07	46.15	Receptive	Composite	.56
Schultz, Izard, Stapleton, Buckingham-Howes, and Bear (2009)	154	7.31	51.30	Expressive	Acceptance, rejection	.13
Slaughter, Dennis, and Pritchard (2002), Study 2	87	5.29	47.13	Receptive	Composite	.19
Snyder et al. (2008)	267	5.30	49.81	Receptive	Composite	.36
Strayer and Mashal (1983)	20	4.30	55.00	Receptive, expressive	Acceptance	.19
Von Grünigen et al. (2010)	1,090	5.80	47.89	Expressive	Acceptance	.33
Von Grünigen and Kochenderfer-Ladd (2010), Anglo-Sample	160	8.50	45.00	Unclear	Acceptance	.13
Von Grünigen and Kochenderfer-Ladd (2010), Latino-Sample	159	8.50	47.80	Unclear	Acceptance	.24
Watson et al. (1999), Study 2	52	5.06	42.31	Receptive	Composite	.39
Yamamoto, Lembright, and Corrigan (1966)	730	n.r.	52.60	Mixed	Acceptance, rejection	.17
Yeates, Schultz, and Selman (1991), Study 2	49	9.08	40.82	Receptive	Composite	.23

Table 1 Articles included in meta-analysis

View as image HTML

Preliminary Analyses

Publication Bias

As illustrated in Figure 1, large studies were not biased toward small effect sizes. Moreover, the graph showed a symmetrical pattern. Therefore, the data showed evidence against publication bias.

Overall Effect Size

The overall effect size for the relation between language competence and social preference was .25 (CI = .20–.29; SE = .01), which, according to Cohen (1988), is small to moderate in strength, and is significantly greater than zero (z = 23.50, p < .001).

Cochran’s chi-square test indicated that effect sizes significantly differed across studies (Q = 153.14, df = 48; p < .001). Therefore, we investigated whether moderator variables explained variation of effect sizes.

Moderator Analyses

To examine the relation between study characteristics and effect size, we conducted analyses using SPSS macros (Lipsey & Wilson, 2001). For categorical moderators, the SPSS macro METAF with maximum likelihood estimation was used. For continuous moderators, for example, the proportion of females in the sample, the SPSS macro Metareg with maximum likelihood estimation was utilized.

Age of Participants

The exact age was not reported in six independent samples. Therefore, these studies were excluded in the analysis of age as a potential moderator. The Q_between statistic for the moderator age was significant (β = −.52, df = 1; p < .001; k = 43). As illustrated in Figure 2, the older the children, the less important language competence is for social preference.

Directions for Future Research

The present study provides several starting points for further research in this field. First, there are population groups that have received little attention in research to date. For example, research by Von Grünigen et al. (2010) and Von Grünigen, Kochenderfer-Ladd, Perren, and Alsaker (2012) showed that language competence has an important influence on peer acceptance, particularly for children from an immigrant background, functioning as a protective factor against prejudice and rejection by peers. However, based on the few studies on the relation between language skills and social preference, it was not possible to investigate immigrant background and bilingualism as moderators in the present meta-analysis. There is a need for further studies with children from socially disadvantaged backgrounds.

Second, the studies included in our meta-analysis originated almost exclusively from English-speaking countries (North America, UK, Australia) and from three European countries. All of the studies were from culturally Western countries. It may be that the relation between language competence and social preference is different in non-Western cultures. Cross-cultural studies have shown that some behaviors that are highly esteemed in Western cultures are regarded as maladaptive in other cultures. For example, Chen and colleagues showed that shyness is a positive predictor for various adaptive behaviors among Chinese children, whereas in Canada shyness represents a risk factor (Chen et al., 2006). Whether the relation between language competence and social preference differs from culture to culture is a question that the present meta-analysis is unable to answer given the lack of studies, and thus would be an important starting point for future research.

Third, in the present meta-analysis we focused on social preference as a measure of peer relationships. It may be that the relation between language competence and peer relationships differs depending on which aspect of peer relationships is examined. As mentioned previously, there are studies investigating language competence and different aspects of peer relations. However, these different aspects are related to one another only to a moderate extent. For example, in the study by Von Grünigen et al. (2012), the correlation between peer acceptance, victimization, and social withdrawal was significant, but only small to medium in size. Another aspect of peer relationships are close friendships. However, the processes to form and maintain friendships differ from the processes that lead to a high social status within a group (see Rubin et al., 2005, as an overview). Accordingly, whether language competence is of equal relevance to different aspects of social interaction could form the subject of further meta-analyses and lead to a broader understanding of the correlates of peer relationship development.

Fourth, it is conceivable that additional variables such as executive functions or theory of mind mediate the relation between language competence and social preference. In the meta-analysis by Milligan, Astington, and Dack (2007), for example, language competence and theory of mind are significantly interrelated. Caputi, Lecce, Pagnin, and Banerjee (2012) showed that theory of mind is a significant predictor of peer acceptance. Thus, theory of mind might explain the relation between language competence and social preference. Additional variables explaining the relation between language competence and social preference might be further examined in future research.

Strengths and Limitations

The present study had two primary limitations that we would like to address in the following sections. First, it included only studies that examined the cross-sectional relation between language competence and social preference. Accordingly, the results do not provide any conclusive evidence regarding the direction of this relation. Theoretically, we can imagine that a bidirectional relation exists between language competence and social preference. That is, it is possible that children’s limited language skills impede their access to their peer group, and that less well-accepted children interact with other children less frequently and accordingly have fewer opportunities to develop their language skills. Longitudinal or experimental intervention studies are required to test directionality. However, to date, only a few studies have investigated the longitudinal relation between language and social status (e.g., Von Grünigen et al., 2012; Rubin & Danielsbeirness, 1983) and to the authors’ knowledge no experimental intervention study has been conducted on this topic. Thus, the directionality of the relation between language competence and social preference could not be tested within the present meta-analysis.

Furthermore, we were limited in regard to the moderators we examined. As the meta-analysis of Newcomb, Bukowski, and Pattee (1993) demonstrated, correlates of social status differ depending on information source. In the present meta-analysis, most of the information on social preference came from the peers themselves. Thus it was not possible to investigate whether the source of the information captured has any influence on the relation between language competence and social preference. Although capturing peer acceptance through peers themselves is a valid source for the analysis of peer relationships, more varied methods of capturing data might provide further insights into the relation between language competence and social preference. For example, a child might be generally rejected by his or her peer group, yet still have some close friends and thus not feel rejected. In this type of case, the relation between language competence and social preference might correlate differently than when social preference was assessed by peer rating. Moreover, social preference is defined as the social status within a peer group (Coie et al., 1982). However, there are differences between peer groups. For example, in some groups only few children are very liked or disliked whereas in other groups the group cohesion is generally very high and all children are popular to some extent. Thus, it is conceivable that differences between groups might moderate the relation between language competence and social preference.

Despite these limitations, the present meta-analysis casts fresh light on the contradictory findings on the relation between language competence and social preference. The meta-analysis included 49 effect sizes. There were no restrictions as far as the included publication form and year of publication were concerned. Accordingly, this meta-analysis covers a broad range of studies on the relation between language competence and social preference, thus making a valuable contribution to current research on the correlates of peer relations.