The interplay of referential function and character primacy on referring expressions in children’s narratives

Abstract

This study considered the linguistic forms used by 63 English-speaking Canadian children from kindergarten to second grade (ages 5;6–8;8) to introduce, maintain reference to, and reintroduce primary and secondary characters throughout their narratives The expected referring forms were used more frequently for the best-matching referential function: indefinites for introduction, pronouns and null forms for maintenance, and identifiables (i.e., definite and possessive NPs, and proper names) for reintroduction. Developmental changes in form–function mappings were present for both introduction and reintroduction. Many children were also influenced by the relative prominence of story characters in their use of pronominals. Nonetheless, function constraints exerted a much stronger influence on referential choice than did character primacy in all grades. By systematically exploring the interplay of referential function and character primacy on referring expressions, this study adds to existing findings on many levels. It also invites future research that manipulates various features of both primary and secondary characters.

Keywords

Character primacy function constraints narrative production referring expressions school-aged children

Creating and sustaining clear reference to story characters require the narrator to monitor the listener’s familiarity with the various characters so as to present the information in a way that is easily interpretable. Character reference can occur at different points in a narrative and be accomplished using different linguistic forms. Bamberg (1987) proposed that references be grouped by referential function, namely introduction, maintenance, and reintroduction. Referential function influences which referring expressions can typically be interpreted unambiguously by the listener. Nonetheless, young children in particular may rely instead on an organizational strategy that gives special status to primary characters in their narratives. The current study looked at the impacts of both referential function and character primacy on the use of referring expressions for character reference in the narratives of young school-aged English-speaking children.

The development of form–function mappings for character reference

When referring to a character, the speaker ought to choose a linguistic form that coincides with the level of cognitive accessibility of the intended referent in the listener’s knowledge of the story (Ariel, 1996; Cornish, 1999). This level of accessibility will vary depending on whether the character is being referred to for the first time or again, and how recently it was previously mentioned. To construct and interpret reference, both parties presumably track the various characters throughout the story by creating and modifying a discourse model – a specific mental representation of an individual’s knowledge of the discourse experience (Brown & Yule, 1983; Johnson-Laird & Garnham, 1980).

Referring expressions may be considered on a continuum of presupposition or givenness (Gundel, Hedberg, & Zacharski, 1993). Introduction involves referring to new information, and so it invites forms that mark newness and are lowest in presupposition level, such as indefinite noun phrases (NPs). Subsequent mentions involve continued reference to given information. Maintenance of reference to a recently mentioned character is usually accomplished unambiguously by referring expressions with the highest level of presupposition such as pronouns and null forms, whereas reintroduction of a character that has been temporarily out of attentional focus usually requires forms with more referential strength, such as definite NPs.

The patterns of form–function mappings mentioned above are generally expected in older school-aged children and adults (Hickmann & Hendriks, 1999; Hickmann, Hendriks, Roland, & Liang, 1996) and correspond to what Bamberg (1987, p. 57) labeled an anaphoric strategy, whereby the narrator chooses nominal expressions for introduction and reintroduction, while maintaining reference predominantly with pronouns. Younger children, however, tend to rely excessively on the nonlinguistic context to disambiguate (e.g., pointing; Beliavsky, 2003; Karmiloff-Smith, 1985). They also favor forms with lower levels of referential strength (i.e., pronouns) regardless of the function (Hickmann et al., 1996; Orsolini, Rossi, & Pontecorvo, 1996; Warden, 1976).

A few studies have considered the linguistic forms that children use to mark new vs. given information in their narratives. Looking across studies, it seems that not until the ages of 9 or 10 years do children who speak English, French, or German use indefinite forms to introduce characters with adult-like proficiency, although they begin to differentiate between forms for new and given information much earlier (Hickmann et al., 1996; Kail & Hickmann, 1992; Warden, 1976). For instance, in their cross-linguistic study, Hickmann et al. (1996) had adults and children aged 4–10 years tell two stories from short sequences of pictures. By 7 years of age, the children in the three Indo-European language groups produced a majority of indefinite nominals to introduce characters (⩾ 54%), whereas they used almost exclusively (⩾ 98%) referring expressions lower in referential strength (i.e., definite nominals, pronouns, and null forms) for subsequent mentions (maintenance and reintroduction combined). For English-speakers, the proportion of indefinite nominals used to introduce characters increased significantly between the ages of 4, 7, and 10 years (from 27% to 88%); only 10-year-olds did not differ from adults.

The other level of contrastive use of referring expressions involves differentiating the forms that best match maintenance rather than reintroduction of referents. In a companion study to the one mentioned above, Hickmann and Hendriks (1999) found that the 4- to 10-year-old English-speakers used a clear majority of pronouns or null forms (73%) to maintain reference to characters, whereas these forms accounted for a minority (17%) of reintroductions. No significant age differences were reported for any language group.

Two additional studies have examined the referring expressions used for maintenance compared to reintroduction of story characters, both of which did find developmental differences. Orsolini et al. (1996) considered character reintroduction by Italian children aged 4–10 years who generated multi-episode stories from a wordless picture book involving multiple characters, Frog, where are you? (FWAY; Mayer, 1969). The authors found evidence of form–function mappings in all age groups: null forms were the most frequent referring expression across age groups for maintenance (as expected, given that person and number are marked on the verb in Italian), whereas full NPs were most frequent for reintroduction (⩾ 60%). For reintroduction, the two younger groups (aged 4 and 5 years) used null forms significantly more often (25%–30%) than did the three older groups (aged 6, 8, and 10 years; 15%–20%); this suggests that older children were more sensitive to differences between referring expressions best suited for maintenance or reintroduction.

Vion and Colas (1999) specifically manipulated whether a character was maintained or reintroduced in the last picture of eight-frame wordless picture sequences. French-speaking children aged 7–11 years demonstrated that they were sensitive to this difference: participants used forms that did not explicitly identify the referent (mostly pronouns) for maintenance and, in contrast, forms with greater referential strength (definite or possessive NPs, etc.) for reintroduction. Pronouns never represented more than 6% of reintroductions for any age group but decreased with age to fewer than 2%; null forms were all but absent. The use of definite nominals to reintroduce increased with age throughout the developmental period considered, as did the magnitude of the difference in the use of definite nominals between maintenance and introduction. Hence, the few existing studies agree that children as young as 4 years of age show emerging form–function mappings for given referents depending on whether or not they have been consistently in the focus of attention of the narrator and the listener. There is some uncertainty, however, regarding whether this sensitivity continues to be refined during the early school years.

Mature form–function mappings are presumed to involve selection of referring expressions that are best suited for each referential function: introduction, maintenance, and reintroduction. Yet, studies that have considered contrastive use of referring expressions across the three referential functions for children of different ages are rare. From the results of their companion studies (Hickmann & Hendriks, 1999; Hickmann et al., 1996), Hickmann and Hendriks (1999, pp. 445–446) argued that children are sensitive to referential ‘continuity vs. discontinuity’ earlier than they distinguish new from given information. They made this claim based on the lack of developmental trends for forms used for maintenance versus reintroduction, whereas contrastive use of referring expressions for introduction and subsequent mention was not present in the youngest age group (4- to 5-year-olds) and became adult-like by 10 years. These results require confirmation, however, particularly given that the stories used to elicit the narratives were short and thus may not have been demanding with regard to managing character reintroduction.

The possibility that longer, more complex stories may produce different results is supported by the work of Serratrice (2007), who elicited FWAY from English-speaking 7- to 8-year olds. Participants exhibited sensitivity to function constraints in their referential choice across all three functions, as they clearly opted for a majority of indefinite nominals for introduction (⩾ 63%), pronouns or null forms for maintenance (⩾ 66%), and NPs for reintroduction (⩾ 80%). Nonetheless, approximately one-third of introductions were accomplished with referring expressions that are usually reserved for given information, and one-fifth of reintroductions in the subject role involved pronouns. These data reflect contrastive use of referring expressions across the three referential functions, but also suggest that children’s referential systems may still be consolidating in this age range, particularly for marking introduction and reintroduction. This hypothesis is supported by results from Chen and Lei (2013), who found that somewhat older children (aged 8–10 years) produced a strong majority of indefinite NPs (87%) for introduction and of definite NPs (84%) for reintroduction.¹

Character primacy and referential choice

Beyond function constraints, the relative standing of characters has also been hypothesized to influence children’s referring expressions. Karmiloff-Smith (1980, 1985) proposed that between the ages of 4 and 10 years, French- and English-speaking children who produced simple two-character stories from a short picture sequence progressively moved from a strategy that gave special status to the primary character – the thematic subject – to an anaphoric strategy that gave precedence to referential function. She coined the term thematic subject strategy to characterize reserving pronouns for reference to primary characters.

Bamberg (1987) made the point that to highlight any distinctions in the forms used to refer to characters of differing prominence, one should contrast maintenance and switching (i.e., introduction and reintroduction combined), as only in cases of ‘switching reference from one character to a new or displaced character’ (p. 63) would an overuse of referring expressions too high on the continuum of presupposition become evident. Two studies have found developmental differences in the forms used to either switch reference to or reintroduce primary characters. Wigglesworth (1997) looked at how English-speaking children switched reference to the thematic subject in FWAY (the boy), and categorized each participant by dominant strategy. Most adults used an anaphoric strategy (i.e., few pronouns for switching to the boy), whereas fewer than half of 10-year-olds did so. Moreover, nearly half of the 8-year-olds and three-quarters of the 6-year-olds used a high proportion of pronouns to switch reference to the boy. Orsolini et al. (1996) looked at reintroductions of primary and secondary characters by 4- to 10-year-old Italian-speakers who produced FWAY and found that the two younger groups (ages 4 and 5 years) were more likely to reintroduce the boy using pronouns or null forms than were older groups (aged 6, 8, and 10 years). In fact, the older children tended to do so only when it was possible to disambiguate the identity of the referent via other means (e.g., singular vs. plural marking on the verb; semantic relatedness such as with bee and stinging). Nonetheless, all age groups produced mostly full NPs to reintroduce any primary or secondary character on its own, including the boy.

Bamberg (1987) proposed that the degree of animacy of characters could also influence children’s referential choices. Whereas the adults who told FWAY consistently used an anaphoric strategy for the boy and the dog, the children treated these characters differently when switching reference. Children in all three age groups showed a strong tendency to maintain reference to either the boy or the dog using pronouns. In contrast, only 9- to 10-year-olds showed a preference for nominal forms when switching reference to the boy or the dog, whereas the two younger groups (aged 3 and 5 years) were more likely to use pronouns for the human compared to the animal. Hence, although even the youngest group of children showed some sensitivity to form–function mappings for both characters, a difference appeared between the two characters for switching. Hickmann, Kail, and Roland (1995) obtained similar results for French-speaking 6- to 11-year-olds. Bamberg (1987) speculated that young children were more likely to give special status to characters that were ‘more active and dynamic’ and that accordingly played a primordial role in ‘carrying the plot forward’ (p. 96). Hence, although both the boy and the dog could be deemed primary characters in FWAY based on factors such as frequency and consistency of appearance, and point of first appearance (McGann & Schwartz, 1988), the boy may have ranked higher in primacy given his role in the plot. These results from narrative elicitation find support in the experimental research of Fukumura and van Gompel (2011), who considered factors affecting referential choice in a sentence completion task. The adult participants frequently used pronouns to refer to both inanimate and animate (human) referents but did so significantly more often for the humans.

Results from the studies of Wigglesworth (1997), Orsolini et al. (1996), and Bamberg (1987) provide some support for the hypothesis that, at least at some points in development, function constraints and character features may be in opposition, although there was no evidence that pronouns were strictly reserved for primary characters even in the youngest groups. Primacy seems to affect which referring expressions are used for character reference by young school-aged children, and the relative degree of animacy or agentivity of characters may contribute to this effect. With age will come narrative experience, adult modeling, increased cognitive resources, and greater awareness of the needs of the listener (Kang, Kim, & Pan, 2009; Karmiloff-Smith, 1985; Kelly & Bailey, 2012; Matthews, Lieven, Theakston, & Tomasello, 2006; Orsolini et al., 1996; Whitely & Colozzo, 2013), which together should progressively lead children to overcome the tendency to favor reduced forms for primary characters and become more sensitive to function constraints. The relative influences of function constraints and character standing have not, however, been explored systematically within a single developmental study that considered all three referential functions and that contrasted primary and secondary characters.

The present study considered how function constraints and character primacy influenced referential choice in the narratives of English-speaking Canadian children from kindergarten to grade 2. Each child produced two stories from wordless picture books, both of which presented multiple opportunities for introduction, maintenance, and reintroduction of primary and secondary characters. The first goal of this study was to document to what extent participants were sensitive to function constraints in their use of referring expressions. Based on prior research, we anticipated that the children in the three grades would show emerging form–function mappings across all three referential functions. The limited existing data regarding reintroduction made it difficult to foresee whether developmental differences would be present for this function. Nonetheless, given that the stories required that the narrator keep track of numerous characters over multiple episodes, we predicted a greater degree of form–function matches with increasing grade for both introduction and reintroduction. This study also investigated whether the influence of referential function on referential choice was present across character groups, and specifically to what extent participants were influenced by character primacy in their use of referential pronominals. In the face of limited prior evidence, our predictions were tentative regarding the effect of character primacy, but we expected that function constraints would exert a stronger effect on referential choice than primacy across the grades.

Method

Participants

Sixty-three monolingual English-speaking Canadian children (27 boys, 36 girls) from kindergarten through second grade (range 5.50–8.67 years of age) participated in the study: 21 in kindergarten (M age = 5.93, SD = 0.28), 23 in grade 1 (M age = 6.97, SD = 0.32), and 19 in grade 2 (M age = 8.09, SD = 0.31). They all spoke primarily English both in the home and at school and were typically developing based on teacher and parent reports. The grades did not differ regarding level of maternal education, F(2, 62) = 1.37, p = .26, η_p² = .04, or gender distribution, χ² (2, N = 63) = 1.60, p = .45, Cramer’s V = .16.

Procedures

Narratives were elicited using two wordless picture books that are similar in many ways, including visual presentation (black and white pictures), length, and complexity (multiple episodes). Critically, both stories invite numerous mentions of various story characters in each referential function – introduction, maintenance, and reintroduction. These stories were also chosen because they differ in relevant ways regarding character features. In April fools (AF; Krahn, 1974), two boys play a prank on the townspeople by scaring them with a monster that they have constructed. All animate characters in AF are human. There are two equipotent primary characters (i.e., the boys), who are often performing shared activities. The secondary characters consist of a woman, a man, and crowds of townspeople.

In Frog on his own (FOHO; Mayer, 1973), a pet frog leaves his owner in a park and embarks on a journey of wreaking havoc for a number of other park users. FOHO has an animal as the only main protagonist. There are multiple (seven human and five animal) secondary characters who are mostly acting independently. Hence, in this story, degree of animacy (Comrie, 1989) and primacy – two factors that are presumed to increase the likelihood of using pronouns for reference – are in competition. See Appendix for a detailed description of each story.

Each child was seen individually. The experimenter instructed the participant to look through the pages of the entire book before telling the story. She explained that the story would be video-recorded for a friend who would listen to it later without the book. She then positioned herself so that she could not see the pictures. The physical setup and the instructions were designed to reduce the shared contextual information between the participant and the listener, and thus increase the likelihood that the children would be more precise when referring to story characters (Hickmann et al., 1995; Kail & Hickmann, 1992; Matthews et al., 2006). AF was always presented before FOHO.

Narrative transcription

The stories were orthographically transcribed using the Systematic Analysis of Language Transcripts software (SALT; Miller & Iglesias, 2012). Unintelligible and abandoned utterances, story closings (e.g., the end, that’s it), tangential comments, and mazes (i.e., false starts, retraces, and repetitions) were excluded from the texts. A second listener reviewed transcripts for accuracy; any disagreements were resolved by discussion.

Referential coding

All mentions to an animate story character (including those in dialogue) were tagged according to the character’s identity and coded on two dimensions: referential function and referring expression. Following Bamberg (1987), each mention was classified into one of three categories of REFERENTIAL FUNCTION. The unit of analysis was the clause, defined as a predicate and its arguments, and following Berman and Slobin (1994, pp. 660–664).

Introduction: The speaker refers to a story character for the first time.

Maintenance: The speaker continues to refer to the same character within a clause, in successive clauses, or across a clause (i.e., a descriptive statement) that does not advance the story (e.g., it was a really nice forest). Clauses were not systematically considered to be intervening, however. The guiding principle for differentiating between maintenance and reintroduction was judging which character(s) could be presumed to be active within the discourse model at a given point within the story. Cases of continued accessibility were coded as maintenance. Specifically, relative clauses (e.g., He [frog] found a little boy that was rowing a boat. The frog [Maintenance] had a nasty idea) and nonfinite clauses (e.g., They [the boys] painted it yellow and green to look like a monster’s head. They [the boys; Maintenance] carried it out) were not generally considered to be intervening.

Reintroduction: The speaker refers back to a character who was previously mentioned but was temporarily out of the focus of attention. Thus, a character reference was coded as reintroduction if it was judged to correspond to a ‘displaced non-new topic’ (Bamberg, 1987, p. 44). This most often arose when the character was not mentioned in the preceding clause (see above for exceptions). Reintroduction also occurred when the speaker changed from referring to two or more characters together (as a single referent) to referring to a subset of these characters, and vice versa (e.g., There was a dog chasing them [frog and cat]. It was the dog chasing the cat [Reintroduction]).

The type of REFERRING EXPRESSION corresponds to the lexical form used to refer to a story character. We used a classification system modified from Halliday and Hasan (1976) and Gundel et al. (1993) for this coding.

Indefinite nominals: Nouns preceded by indefinite articles (e.g., a frog), numerals (e.g., one boy), or the demonstrative determiner this used colloquially (e.g., One day this boy was walking in a park).

Indefinite pronominals: Indefinite (e.g., everybody, someone, all, other) and interrogative (e.g., who, what, which) pronouns.

Identifiable nominals: Nouns preceded by definite or demonstrative determiners (e.g., the dog, that frog), by possessive pronouns (e.g., their dad), or by nouns (with the clitic ’s) marking a possessor (e.g., the boy’s frog); proper names (e.g., Bob, Froggie).

Referential pronominals: Personal (e.g., he, she, they, it), possessive (e.g., her, hers, his, their, theirs), demonstrative (e.g., this, that), and relative (e.g., which, who, that) pronouns, as well as ellipses of the NP (e.g., he turned very pale and ø fainted).

A second coder double-checked all transcripts for coding of character identity, referential function, and referring expression. All disagreements were discussed between the two coders, who came to a final decision.

Primary and secondary characters

For some analyses, references to animate characters were grouped into primary or secondary characters based on factors that have been proposed to be associated with main protagonists (see McGann & Schwartz, 1988). In AF, the boys are both clearly main characters, as they are driving the plot forward, and they also appear earlier, more frequently, and more consistently compared to the other human characters. In FOHO, the frog, whose actions advance the plot, also meets the primacy criteria of frequency and consistency of appearance, and point of first appearance. The frog is also anthropomorphized in subtle ways (i.e., facial expressions clearly reflecting emotions; waving to a human character). The secondary characters differ in terms of degree of animacy, as some are humans and others are animals. See Appendix.

Measures

Using the SALT program, we calculated the total number of references for each child according to multiple dimensions: referential function, category of referring expression, and character. We used proportional data to control for differences in numbers of references across children and between functions or characters, and performed an arcsine transformation prior to all statistical analyses. We nonetheless report marginal means based on untransformed values for ease of interpretation.

Reliability

Reliability was evaluated using the stories of 7 (11%) of the 63 children. Based on the story texts that were independently transcribed by a second coder, agreement was 95% for the word-level transcription. An estimate of inter-rater reliability for referential coding was obtained based on discrepancies prior to changes stemming from consensus coding. Agreement was 98% for the presence of a character reference (i.e., accounting for omissions by one of the coders). For references identified by both coders, agreement was 99% for character identity, 99% for referential function, and 99% for referring expression. The range of reliability across children for all referential coding variables was 94%–100%.

Results

Function constraints on referring expressions

All participants told stories from two wordless picture books. Each participant produced on average 15.3 introductions, 58.8 maintenances, and 30.4 reintroductions. The following sections report how referring expressions varied depending on referential function and whether the children in the three grades showed the same patterns of referential choice by function. Referring expressions were grouped into three categories: (1) indefinites (85% nominals and 15% pronominals), (2) referential pronominals (86% pronouns and 14% ellipses), and (3) identifiable nominals.

Distributions of referring expressions by function

Each function has an expected category of referring expression that is assumed to best fit its requirements: indefinites for introduction, referential pronominals for maintenance, and identifiable nominals for reintroduction. The descriptive data presented in Table 1 illustrate by what means the children referred to story characters. We calculated the proportion for each category of referring expression (number of referring expressions of a given category [indefinites, referential pronominals, or identifiables] divided by the total number of character references) for each function individually for each child. This neutralized differences regarding the number of character references across children and functions.

Table 1.

Mean (SD) proportions of referring expressions, by function, and by grade.

Referential function	Referring expression	Grade						All grades (N = 63)
Referential function	Referring expression	K (n = 21)		1 (n = 23)		2 (n = 19)		All grades (N = 63)
Introduction	Indefinites	.53	(.20)	.59	(.22)	.69	(.14)	.60	(.20)
	Pronominals	.12	(.12)	.08	(.10)	.06	(.09)	.09	(.11)
	Identifiables	.36	(.18)	.33	(.18)	.25	(.11)	.31	(.17)
Maintenance	Indefinites	.04	(.02)	.03	(.02)	.02	(.02)	.03	(.02)
	Pronominals	.78	(.11)	.77	(.09)	.74	(.10)	.77	(.10)
	Identifiables	.18	(.10)	.20	(.09)	.23	(.10)	.20	(.10)
Reintroduction	Indefinites	.14	(.08)	.09	(.06)	.12	(.06)	.12	(.07)
	Pronominals	.36	(.12)	.29	(.12)	.28	(.13)	.31	(.12)
	Identifiables	.50	(.11)	.62	(.11)	.60	(.14)	.57	(.13)

Indefinites represented the majority (M = .60) of introductions. The participants tended to select the expected category of referring expression for new information most often, although it was not a strong majority due mostly to the high level of identifiable nominals (M = .31). For maintenance, the children used almost exclusively forms that are deemed appropriate for given information: referential pronominals (M = .77) and identifiable nominals (M = .20). Finally, for reintroduction, the most frequent category was identifiable nominals (M = .57). When reintroducing characters, the children were most likely to choose a referring expression that signals given information while also acknowledging that the referent had been temporarily out of attentional focus. This was not a strong majority, however, mostly on account of a high level of referential pronominals (M = .31). The distributions of referring expressions were similar across grades in each function (see Table 1).

We used inferential statistics to test whether referential function systematically influenced referential choice, by considering the expected dominant category of referring expression for each function. We also considered whether any such influences varied by grade.

Expected dominant referring expression by referential function

For each category of referring expression, the referential function where that category would be expected to dominate was compared to the two other functions where it should not be used as frequently. This resulted in three pairs of comparisons: (a) for indefinites: introduction vs. maintenance, and introduction vs. reintroduction; (b) for referential pronominals: maintenance vs. introduction, and maintenance vs. reintroduction; (c) for identifiables: reintroduction vs. introduction, and reintroduction vs. maintenance (see Figure 1). We completed three two-way mixed ANOVAs of function (3) by grade (3), one for each category of referring expression (indefinites, referential pronominals, and identifiables), using the proportion of the relevant referring expression as the dependent variable. The focus for each ANOVA was on two pairwise planned contrasts (comparing the referential function where the referring expression would be expected to dominate with each of the other functions, separately) and the corresponding interaction contrasts between pairs of functions and grade (see below for details regarding each ANOVA). Significant contrasts would indicate sensitivity to function constraints in referential choice (i.e., differences in the proportion of the referring expression between pairs of functions), and significant interaction contrasts with grade would signal differences between grades regarding this sensitivity. There was a main effect of function for each category of referring expression (ps < .001, η_p² ⩾ .88). We thus considered the planned contrasts. Only significant results are reported in detail, based on a critical p-value of .05 – the criterion for statistical significance used throughout this study.

Figure 1.

Mean proportions of referring expressions by referential function. Error bars correspond to ± 2 SE. (a) Indefinites. (b) Referential pronominals. (c) Identifiable nominals.

Indefinites: Introduction vs. maintenance and introduction vs. reintroduction

Indefinites are expected to dominate in introduction but not in either of the other functions. When comparing the use of indefinites between introduction and maintenance, both the pairwise planned contrast between functions, F(1, 60) = 650.23, p < .001, η_p² = .92, and the corresponding interaction contrast with grade, F(2, 60) = 5.35, p = .007, η_p² = .15, were significant. To examine the interaction, we completed simple main effects tests, one per grade. Each grade produced a significantly higher mean proportion of indefinites for introduction (M_K = .53, M_Gr1 = .59, M_Gr2 = .69) compared to maintenance (M_K = .04, M_Gr1 = .03, M_Gr2 = .02); although the differences between functions were large in all cases, the effect size was larger for the eldest grade (ps < .001; values of η_p² = .89, .89, and .96, for kindergarten to grade 2).

The same pattern emerged when comparing the use of indefinites between introduction and reintroduction. Both the pairwise planned contrast, F(1, 60) = 357.89, p < .001, η_p² = .86, and the corresponding interaction contrast with grade, F(2, 60) = 3.96, p = .024, η_p² = .12, were significant. To examine the interaction, we completed simple main effects tests, one per grade. Each grade produced a significantly higher mean proportion of indefinites for introduction (M_K = .53, M_Gr1 = .59, M_Gr2 = .69) compared to reintroduction (M_K = .14, M_Gr1 = .09, M_Gr2 = .12); although the differences between functions were large in all cases, the effect size increased with grade, and particularly between kindergarten and the two older grades (ps < .001; values of η_p² = .76, .87, and .91, for kindergarten to grade 2).

Referential pronominals: Maintenance vs. introduction and maintenance vs. reintroduction

Referential pronominals are expected to dominate in maintenance but not in either of the other functions. When comparing the use of referential pronominals between maintenance and introduction, the pairwise planned contrast between functions was significant, F(1, 60) = 750.06, p < .001, η_p² = .93, but the corresponding interaction contrast with grade was not (p = .57). The proportion of referential pronominals used for maintenance (M = .77) was consistently higher than for introduction (M = .09).

The same pattern emerged when comparing the use of referential pronominals between maintenance and reintroduction. The pairwise planned contrast was significant, F(1, 60) = 571.26, p < .001, η_p² = .90, but the corresponding interaction contrast with grade was not (p = .49). The proportion of referential pronominals used for maintenance (M = .77) was consistently higher than for reintroduction (M = .31).

Although neither of the interaction contrasts was significant, there was a marginally significant main effect of grade, F(2, 60) = 3.10, p = .052, η_p² = .09, suggesting that the use of referential pronominals differed with grade. Post-hoc tests using a Bonferroni correction revealed that the kindergarten group had higher proportions of referential pronominals than did grade 2 (p = .049) across functions; the other pairwise comparisons were not significant (kindergarten and grade 1, p = .42; grade 1 and grade 2, p = .87). See Table 1 for details.

Identifiables: Reintroduction vs. introduction and reintroduction vs. maintenance

Identifiable nominals are expected to dominate in reintroduction but not in either of the other functions (although they are also appropriate for maintenance). When comparing the use of identifiables between reintroduction and introduction, both the pairwise planned contrast between functions, F(1, 60) = 81.74, p < .001, η_p² = .58, and the corresponding interaction contrast with grade, F(2, 60) = 4.02, p < .023, η_p² = .12, were significant. To examine the interaction, we completed simple main effects tests, one per grade. Each grade produced a significantly higher mean proportion of identifiables for reintroduction (M_K = .50, M_Gr1 = .62, M_Gr2 = .60) compared to introduction (M_K = .36, M_Gr1 = .33, M_Gr2 = .25), and the effect size increased considerably with grade (ps ⩽ .010; values of η_p² = .29, .59, and .77, for kindergarten to grade 2).

When comparing the use of identifiables between reintroduction and maintenance, the planned pairwise contrasts was significant, F(1, 60) = 463.69, p < .001, η_p² = .89, but the corresponding interaction contrast was not (p = .13). The proportion of identifiables was consistently higher for reintroduction (M = .57) than for maintenance (M = .20).

To summarize, participants consistently preferred indefinites for character introduction, referential pronominals for character maintenance, and identifiables for character reintroduction. There were nonetheless noteworthy differences between grades. The older grades showed greater degrees of contrastive use of referring expressions depending on the referential function for both indefinites and identifiables; in fact, only the two older grades achieved a clear majority for both indefinites for introduction and identifiables for reintroduction. Furthermore, although there was no indication of grade differences in sensitivity to function constraints for referential pronominals, the kindergartners used these referring expressions more frequently compared to children in grade 2 across functions.

The effect of character primacy on referring expressions

The previous section examined sensitivity to function constraints in referential choice for all characters combined. The current section considers whether the influence of referential function was consistent across character groups. The focus here is on the use of referential pronominals given that it has been hypothesized that young children might have a tendency to reserve pronouns for reference to primary characters across functions, and thus to be influenced relatively more by character prominence than by function constraints. Thus, the following analyses served to examine to what extent character primacy had an effect on the use of referential pronominals for referring to animate story protagonists, and to compare any such effect with the influence of referential function. Referential pronominals are not generally assumed to have a good fit with either introduction or reintroduction (due to their high level of presupposition), whereas they are expected to be used frequently for maintenance.

For introduction, primary characters (the boys and the frog) could be introduced only once per participant (or occasionally twice if the boys were introduced separately in AF²), whereas this was not the case for secondary characters (i.e., these categories included more than one character). For maintenance and reintroduction, all character groups could be referred to multiple times by each participant. We thus report proportions based on total counts (by grade and for all participants combined) for each character group (see Table 2 and Figure 2).

Table 2.

Proportions of referential pronominals (and total counts) for primary and secondary characters, by story and by function.

Story	Referential function	Character group	Grade			All grades
Story	Referential function	Character group	K	1	2	All grades
April fools	Introduction	Boys	.43	.24	.33	.33	(73)
	Introduction	Humans	.09	.09	.08	.09	(230)
	Maintenance	Boys	.96	.94	.96	.95	(896)
	Maintenance	Humans	.79	.84	.86	.84	(455)
	Reintroduction	Boys	.56	.55	.48	.53	(449)
	Reintroduction	Humans	.30	.25	.37	.31	(219)
Frog on his own	Introduction	Frog	.00	.00	.00	.00	(63)
		Humans	.13	.05	.01	.06	(374)
		Animals	.03	.05	.00	.03	(221)
	Maintenance	Frog	.69	.65	.61	.65	(1025)
		Humans	.84	.82	.81	.82	(880)
		Animals	.55	.54	.45	.51	(432)
	Reintroduction	Frog	.17	.15	.16	.16	(480)
		Humans	.35	.23	.28	.28	(502)
		Animals	.18	.07	.11	.11	(230)

Notes: Total numbers of children contributing data varied depending on the function and the character. Kindergarten, n = 19–21; grade 1, n = 22–23; grade 2, n = 18–19; all grades, N = 60–63. Proportions are based on total counts per character group (number of referential pronominals/total references), by grade or for the grades combined.

Figure 2.

Proportions of referential pronominals (based on counts) for primary and secondary character groups by referential function and by story. (a) April fools. (b) Frog on his own.

April fools

For AF, we grouped the characters as follows: the boys (primary) and secondary human characters. The few instances when the boys were mentioned with other humans were included with the secondary characters (30 of 2322 of total references ≈ 1%). For references to the boys, although the absolute levels of referential pronominals were unusually high for introduction and especially for reintroduction of these primary characters, the proportion of referential pronominals was much higher for maintenance (.95) than for either introduction (.33) or reintroduction (.53). The same pattern emerged for the secondary human characters, as the proportion of referential pronominals was again much higher for maintenance (.84) than for either introduction (.09) or reintroduction (.31).

For each function, the proportion of referential pronominals was higher for the boys than for secondary humans, with differences ranging from .11 to .24. The within-function differences between the character groups were smaller than the differences between maintenance and either introduction or reintroduction, which ranged from .42 to .75. These patterns were consistent across grades (see Table 2 for details).

Despite the strong and consistent effect of function across grades, further consideration of the data for individual participants revealed that approximately one-third of the children (23 of 63; 48%, 30%, and 32% in kindergarten to grade 2) appeared to be influenced by primacy (i.e., the thematic subject) when introducing animate characters in AF, as they used a referential pronominal to introduce the boys. In contrast, only one child used a majority of referential pronominals to introduce the secondary human characters. For reintroduction, half the children (30 of 60; between 48% and 53% per grade) appeared to be influenced by primacy in AF, as they used a majority of referential pronominals to reintroduce the boys and a minority to reintroduce the secondary humans, with large differences (⩾ .25) between these character groups. Overall, based on their use of referential pronominals to introduce and/or reintroduce the boys, 60% of the participants (68%, 57%, 56% in kindergarten to grade 2) seemed to treat the thematic subject differently when switching reference in AF.

Frog on his own

For FOHO, the three character groups consisted of the frog (primary), secondary human characters, and secondary animal characters. The rare cases of mixed references that crossed animacy (human/animal) boundaries (56 of 4263 total references, ≈ 1%) were excluded from these analyses. In all three cases, the proportion of referential pronominals was much higher for maintenance than for either introduction or reintroduction: frog, maintenance = .65, introduction = .00, reintroduction = .16; secondary humans, maintenance = .82, introduction = .06, reintroduction = .28; secondary animals, maintenance = .51, introduction = .03, and reintroduction = .11.

For each function, secondary humans obtained the highest proportions of referential pronominals, with differences with the frog or the secondary animals ranging from .03 to .31. In comparison, the differences between maintenance and either introduction or reintroduction ranged from .40 to .76, and were thus much larger than within-function differences between the character groups. These patterns were consistent across grades (see Table 2). Finally, there was little evidence for an effect of primacy for this story. No child introduced the frog with a referential pronominal and only 5 children used a majority of referential pronominals to reintroduce the frog. The effect of animacy was only somewhat stronger, as 4 children introduced and 14 children (23% overall; 43%, 13%, and 11% in kindergarten to second grade) reintroduced secondary humans using a majority of referential pronominals, whereas fewer than 3 children did so to either introduce or reintroduce secondary animals.

To summarize, for both stories, the influence of referential function on referential choice was consistent across character groups, as the proportions of referential pronominals were considerably lower for introduction or reintroduction than for maintenance for all character groups; the differences were large in all cases, and larger than between character groups for any function. Results regarding any effect of primacy on the use of referential pronominals depended on the characteristics of the primary characters. In AF, where the primary characters were two boys, an effect of primacy was present for both introduction and reintroduction: the use of referential pronominals was higher for the boys compared to secondary humans, and unusually so for both introduction and reintroduction on account of one-third and one-half of children favoring referential pronominals for introducing and reintroducing the boys, respectively. In FOHO, where the primary character was a frog, there was little evidence for an effect of primacy, as the proportions of referential pronominals were consistently low to moderate across character groups for introduction and reintroduction; furthermore, secondary humans had the highest proportion of referential pronominals for every function. Finally, in contrast with these differences for primary characters between stories, the proportions of referential pronominals for each function for secondary humans were comparable across stories (see Figure 2).

Discussion

This study considered the influence of both function constraints and character primacy on the referring expressions used for character reference by children in kindergarten to second grade. All three referential functions – introduction, maintenance, and reintroduction – were considered in the analyses of the narrative texts that each child produced from two long and complex wordless picture books. Participants in all three grades demonstrated clear sensitivity to referential function in their referential choice. Form–function mappings nonetheless more closely followed the expected patterns for the older grades. Although character primacy also influenced the relative frequency of referential pronominals, this effect was small in comparison to that of referential function. The results from this study also add to the literature regarding the diversity of referring expressions used for introduction and reintroduction. The discussion will address each of these topics in turn.

Function constraints on distributions of referring expressions

The first goal of this study was to consider to what extent young school-aged children were sensitive to function constraints in their referential choices across all three referential functions. Participants were clearly influenced by referential function as the expected categories of referring expressions represented a majority of character references for the function with the matching level of presupposition: indefinites (i.e., indefinite NPs; indefinite and interrogative pronouns) for introduction; referential pronominals (i.e., personal, possessive, demonstrative, and relative pronouns; ellipses) for maintenance; and identifiable nominals (i.e., definite and possessive NPs; proper names) for reintroduction. Furthermore, the expected categories of referring expressions were used more frequently for the best-matching referential function compared to the two other functions.

These results were mostly consistent across grades, yet there were indications of developmental changes in contrastive use of referring expressions by referential function. First, the tendency to reserve indefinites for introduction increased with grade. Although all three grades produced a majority of indefinites for introduction, only second graders showed a strong preference for this form–function mapping, with indefinites making up more than two-thirds of introductions. Second, the participants in the two older grades more strongly favored identifiables for reintroduction compared to introduction. Furthermore, identifiables represented a majority of reintroductions only for first and second graders. Third, kindergartners generally used more referential pronominals across referential functions compared to the older grades. This suggests that participants in the youngest grade were less discerning in their use of referring expressions with low referential strength.

In accord with existing studies, children between 5 and 8 years of age were sensitive to differences between forms best suited to refer to new information (introduction) compared to given information (maintenance and reintroduction), and this sensitivity was still developing in this developmental range (Hickmann et al., 1996; Kail & Hickmann, 1992; Serratrice, 2007; Warden, 1976). Regarding form–function matches for marking given information depending on whether or not it had been continuously in the focus of attention (maintenance vs. reintroduction), the results are generally in line with previous studies where children in this age range used fewer pronominals for reintroduction compared to maintenance and a minority of pronominals for reintroduction (Chen & Lei, 2013; Hickmann & Hendriks, 1999; Orsolini et al., 1996; Serratrice, 2007; Vion & Colas, 1999). Additionally, the data revealed grade differences in the selectivity of identifiables for reintroduction. This finding is in accord with the studies of Orsolini et al. (1996) and Vion and Colas (1999), who found a developmental change in sensitivity to function constraints for given information in this age range. It is, however, at odds with the results of Hickmann and Hendriks (1999), who found no age differences in the distributions of nominals vs. pronominals to reintroduce. It may be that such developmental differences will only come to light if the narrative task is sufficiently demanding regarding reintroduction (i.e., by requiring the narrator to keep track of various characters over multiple episodes) and if it provides sufficient data regarding this function. Both these conditions were met in the current study as well as that of Orsolini et al. Vion and Colas used simple picture sequences but specifically manipulated whether a character was maintained or reintroduced in the last picture and used multiple stories per condition.

Hence, although the children in the three grades showed sensitivity to function constraints in their referential choices, there were clear developmental differences in the tendencies to introduce with indefinites and to reintroduce with identifiables. These results are generally in accord with the findings of prior studies. Yet, the examination of contrastive use of referring expressions for all three referential functions within a single study and the use of multi-character, multi-episode stories provide new insights regarding developmental trends for form–function mappings for introduction and reintroduction. As the next sections will show, they also point to smaller yet noteworthy effects of character features, and provide new data regarding the variety of forms used for each function.

Influence of character primacy on distributions of referring expressions

This study also investigated to what extent participants were influenced by character prominence in their use of referential pronominals. Such an effect would be most obvious when considering reference switching to a new or displaced character (i.e., introduction or reintroduction), where an overuse of referring expressions too high on the continuum of presupposition would stand out. We compared primary and secondary characters and also considered the relative degree of animacy within this dichotomy. In AF, where the primary characters were a pair of boys and all secondary characters were humans, the proportion of referential pronominals was higher for the primary characters for every function. Most notably, referential pronominals were used very frequently to introduce (.33) and reintroduce (.53) the boys, which would not be expected given the low referential strength of these forms. Furthermore, the group analyses were supported by individual-level data. A developmental difference emerged only for introduction, where nearly half of kindergartners referred to the boys using referential pronominals compared to approximately one-third of first and second graders; for reintroduction, approximately half the children in each grade used a majority of referential pronominals when referring to the boys.

FOHO included three groups of characters. Secondary humans were consistently referred to most often with referential pronominals, secondary animals least often. The frog – unmistakably the primary character – held an intermediate position. The proportions of referential pronominals were low for introduction (⩽ .06) and moderate for reintroduction (⩽ .28) for all character groups. Other than for maintenance, only in the case of reintroductions of secondary humans did a sizeable group (one-fifth) of children – mostly kindergartners – use a majority of referential pronominals. Thus, for this story, there was no obvious effect of primacy, but a small effect of animacy, on referential choice.

The data from AF are in accord with the findings from other studies regarding the tendency of at least some children in this age range to preferentially (but not exclusively) use referential pronominals for the thematic subject across referential functions (Orsolini et al., 1996; Wigglesworth, 1997). There was also some indication that this tendency may decrease in the developmental range considered, at least for introduction. On the other hand, for FOHO, where primacy and degree of animacy came in competition, the data are in accord with prior research suggesting that the second factor can influence referential choice (Bamberg, 1987; McGann & Schwartz, 1988). Bamberg (1987) proposed that the likeliness of a specific character being referred to with a pronoun depends on how active it is and the extent to which it moves the plot forward. In the current stories, all animate characters were active at some point and, accordingly, participants regularly maintained reference to them using referential pronominals. Yet, they also frequently introduced or reintroduced the boys in AF using referential pronominals – and did so much more than any other character group.

Given the marked difference regarding the use of referential pronominals for the primary characters in AF and FOHO, it is all the more striking that the data for secondary humans were very similar for every function across stories. Hence, what stands out when considering the various groups of characters is an inordinately high level of referential pronominals used to refer to the boys in AF across referential functions. In all likelihood the facts that the boys were primary characters and humans (rather than animals) likely contributed to this effect. The boys were also clearly driving the plot and present throughout the story, but these two factors would apply equally to the frog. Although AF always came before FOHO, there is no obvious reason why the children would tend to use so many referential pronominals only when referring to the primary characters in the first story simply because they were telling it beforehand. The listener provided neither feedback nor any indication suggesting that the child should be more specific when referring to characters when telling the second story.

One additional plausible reason for the high use of referential pronominals for the boys can be entertained – their participation in shared activities. Previous studies have suggested that co-agentivity influences the selection of referring expressions. Orsolini et al. (1996) found that children who narrated FWAY in Italian used majoritarily person/number inflection on the verbs only when they reintroduced the two primary characters together (the boy and the dog), but mostly full NPs when any story character was reintroduced on its own. Wigglesworth (1997) similarly noted that many 6- and 8-year-old English-speaking children tended to simplify their FWAY narratives by grouping the major protagonists and referring to them with pronouns when their activities could be collapsed. In AF, the boys appeared together at the outset and shared activities most of the time. Together these facts likely pushed children to talk about them together (which they did in 81% of cases) and to use pronouns when doing so.

Assuming that co-agentivity played a role in the high level of referential pronominals used to refer to the primary human characters in AF, it is worth noting, however, that it did not have the same effect for secondary human characters. In AF, the secondary humans often consisted of a group of townspeople acting together, and 65% of references to this group designated two or more people compared to only 11% of references to humans in FOHO. Yet, the data were remarkably similar regarding the use of referential pronominals for the secondary human characters across stories.

The data obtained for each story separately would provide distinct impressions regarding the effect of primacy on character referencing for the current study participants. What to expect in terms of form–function mappings for different character groups is clearly highly task dependent, at least for children who are continuing to develop their referencing abilities. The present study underscores this point by having systematically reported within-child data for more than one story that contrasted the features of primary characters, something that has rarely been done. One implication of these results is that the observed referential strategies cannot simply be interpreted as corresponding to a specific developmental level, given that the same children told both stories within a few minutes. Future research could further investigate these effects by systematically manipulating not only the degree of animacy, but also the co-agentivity of both primary and secondary characters.

Interplay of referential function and character primacy on referring expressions

The participants in the current study clearly showed that they could preferentially match linguistic forms with the appropriate levels of presupposition to their referential functions, although these abilities were continuing to develop. Yet, there is no denying that character features also influenced referential choice. Still, for all grades, the effect of function far outweighed the influence of either primacy or degree of animacy on referring expressions, as the proportion of referential pronominals was considerably higher for maintenance than for either introduction or reintroduction for all character groups. Crucially, there was little evidence that any of the children were using an organizational strategy that gave special status to primary characters to the point that it overrode function constraints. Even in AF, where a primacy effect was evident, participants used referential pronominals more frequently when maintaining reference to the boys than when introducing or reintroducing them (∆ ⩾ .42), and referential pronominals were used frequently to maintain reference to primary and secondary characters alike. Thus, for complex stories involving several active characters, the 5- to 8-year-olds in the current study were in most cases showing evidence of applying an imperfect anaphoric strategy to their character references. This was most obvious for the primary characters in AF, where three-fifths of the children seemed to employ a mixed strategy where sensitivity to function constraints dominated, but the thematic subject also exerted an influence on referring expressions. Thus, by kindergarten, most of the participants in the current study were showing sensitivity to function constraints suggesting that they had the cognitive resources to consider the perspective of the listener when referring to characters in their stories while constructing the plot of their narratives, although they still had to consolidate these abilities.

Diversity of forms within functions

Prior studies have highlighted that English-speaking children could mark newness in diverse ways (Chen & Pan, 2009; Hickmann, 1980; Hickmann et al., 1996; Kail & Hickmann, 1992), including varied indefinite NPs (with indefinite articles, numerals, and the colloquial demonstrative this), definite NPs using possessives or relative clauses, and proper names. The data from the current study concur with these findings and add to them by revealing that indefinite pronouns (e.g., somebody, no one) can also figure among the referential forms used by children for introduction. This study also enriches existing evidence by revealing a similar diversity for reintroduction, which is in accord with the findings of Vion and Colas (1999) for French-speakers. In addition to using the expected definite or possessive NPs and proper names, participants in the current study often reintroduced characters using indefinite NPs with numerals, NPs with relative clauses, indefinite pronominals, and even referential pronominals that were further specified cataphorically (e.g., They saw it was Bob and Joe).

Indefinite pronominals were used unexpectedly frequently in AF, where 55% of all indefinites were pronominals, compared to only 7% in FOHO. The participants sometimes referred to the boys using indefinites such as one, other, and both. They also occasionally used interrogative pronouns to reflect the perspective of the townspeople, who did not know the identity of the pranksters (e.g., Everyone was wondering who was scaring everybody). As the previous example also illustrates, many participants referred to the townspeople using indefinite pronouns (everybody, nobody) across functions – perhaps because the crowd did not have a prominent identity compared to individual characters or simply because the children did not have an easily accessible label for the townspeople.

Even in the eldest grade, the data do not reflect conventional form–function mappings for either introduction or reintroduction. This suggests that referential abilities were continuing to develop across the developmental range considered in the current study. Still, a more complete picture of children’s abilities requires considering to what extent they were able to introduce, maintain reference to, or reintroduce characters in an interpretable manner using diverse linguistic forms. We undertook this task in a companion study that considered referential adequacy (Colozzo & Whitely, 2014) and found that in many instances, unexpected referring expressions used for introduction and reintroduction were easily interpretable.

Conclusion

This study considered the linguistic forms that children in kindergarten to second grade used to introduce, maintain, and reintroduce reference to primary and secondary characters in two multi-episode narratives. By systematically exploring the interplay of referential function and character primacy, this study adds to existing findings on many levels. Participants used a variety of linguistic forms not only to indicate new information but also to draw attention to given information that had temporarily been out of attentional focus. They also showed sensitivity to function constraints in their referential choices across all three referential functions, as well as developmental changes in this sensitivity for both introduction and reintroduction. Function constraints exerted a stronger influence on referring expressions than did character primacy for all three grades. Generally, the vast majority of children were applying an emerging anaphoric strategy in their character references – choosing mostly nominal expressions for introduction and reintroduction, while maintaining reference predominantly with pronouns. They were also showing a growing ability to distinguish between nominal forms of differing referential strength for introduction or reintroduction. When considering the various groups of characters, the pair of primary human characters in AF stood out due to the inordinately high level of referential pronominals used to refer to the boys across referential functions. No such effect was present for the primary animal character in FOHO. This finding suggests caution when drawing any strong conclusions regarding developmental levels or stages based on a single task given the potentially large effects of specific features of characters on referential cohesion. It also invites future research that systematically manipulates multiple variables related to characters, including degree of animacy and co-agentivity for both primary and secondary characters.

Footnotes

Appendix: Story descriptions

Acknowledgements

We are most grateful to the children, families, and school personnel, who with their participation and assistance, made this project possible. We would like to thank Judith R. Johnston and Carolyn E. Johnson for their mentorship and support. We also wish to acknowledge Sarah Blois, Rebecca Kowalenko, Heather Morris, and Heidi Logan for their valuable work on this project.

Funding

This work was supported in part by a scholarship to the second author from the Natural Sciences and Engineering Council of Canada and by grant support from the Johnston family.

Notes

References

Ariel

(1996). Referring expressions and the +/- coreference distinction. In Fretheim

Gundel

J. K.

(Eds.), Reference and referent accessibility (pp. 13–35). Amsterdam, The Netherlands: John Benjamins.

Bamberg

M. G. W.

(1987). The acquisition of narratives: Learning to use language. Berlin, Germany: Mouton de Gruyter.

Beliavsky

(2003). The sequential acquisition of pronominal reference in narrative discourse. Word, 54, 167–189.

Berman

R. A.

Slobin

D. I.

(1994). Relating events in narrative: A crosslinguistic developmental study. Hillsdale, NJ: Lawrence Erlbaum.

Brown

Yule

(1983). Discourse analysis. Cambridge, UK: Cambridge University Press.

Chen

Lei

(2013). The production of referring expressions in oral narratives of Chinese–English bilingual speakers and monolingual peers. Child Language Teaching and Therapy, 29, 41–55. doi:10.1177/0265659012459527

Chen

Pan

(2009). Development of English referring expressions in the narratives of Chinese–English bilinguals. Bilingualism: Language and Cognition, 12, 429–445. doi:10.1017/S1366728909990216

Colozzo

Whitely

(2014). Keeping track of characters: Factors affecting referential adequacy in children’s narratives. First Language, 34, 155–177. doi: 10.1177/0142723714522164

Comrie

(1989). Language universals and linguistic typology (2nd ed.). Chicago, IL: The University of Chicago Press.

10.

Cornish

(1999). Anaphora, discourse, and understanding: Evidence from English and French. Oxford, UK: Oxford University Press.

11.

Fukumura

van Gompel

R. P. G.

(2011). The effect of animacy on the choice of referring expression. Language and Cognitive Processes, 26, 1472–1504. doi:10.1080/01690965.2010.506444

12.

Gundel

J. K.

Hedberg

Zacharski

(1993). Cognitive status and the form of referring expressions in discourse. Language, 69, 274–307.

13.

Halliday

M. A. K.

Hasan

(1976). Cohesion in English. London, UK: Longman.

14.

Hickmann

(1980). Creating referents in discourse: A developmental analysis of linguistic cohesion. In Kreiman

Ojeda

A. E.

(Eds.), Papers from the parasession on pronouns and anaphora (pp. 192–203). Chicago, IL: Chicago Linguistics Society.

15.

Hickmann

Hendriks

(1999). Cohesion and anaphora in children’s narratives: A comparison of English, French, German, and Mandarin Chinese. Journal of Child Language, 26, 419–452.

16.

Hickmann

Hendriks

Roland

Liang

(1996). The marking of new information in children’s narratives: A comparison of English, French, German and Mandarin Chinese. Journal of Child Language, 23, 591–619. doi:10.1017/S0305000900008965

17.

Hickmann

Kail

Roland

(1995). Cohesive anaphoric relations in French children’s narratives as a function of mutual knowledge. First Language, 15, 277–300. doi:10.1177/014272379501504502

18.

Johnson-Laird

P. N.

Garnham

(1980). Descriptions and discourse models. Linguistics and Philosophy, 3, 371–393. doi:10.1007/bf00401691

19.

Kail

Hickmann

(1992). French children’s ability to introduce referents in narratives as a function of mutual knowledge. First Language, 12, 73–94. doi:10.1177/014272379201203405

20.

Kang

J. Y.

Kim

Y. S.

Pan

B. A.

(2009). Five-year-olds’ book talk and story retelling: Contributions of mother-child joint bookreading. First Language, 29, 243–265. doi:10.1177/0142723708101680

21.

Karmiloff-Smith

(1980). Psychological processes underlying pronominalization and non-pronominalization in children’s connected disourse. In Kreiman

Ojeda

A. E.

(Eds.), Papers from the parasession on pronouns and anaphora (pp. 231–250). Chicago, IL: Chicago Linguistics Society.

22.

Karmiloff-Smith

(1985). Language and cognitive processes from a developmental perspective. Language and Cognitive Processes, 1, 61–85. doi:10.1080/01690968508402071

23.

Kelly

K. R.

Bailey

A. L.

(2012). Becoming independent storytellers: Modeling children’s development of narrative macrostructure. First Language, 33, 68–88. doi:10.1177/0142723711433582

24.

Krahn

(1974). April fools. New York, NY: Dutton.

25.

Matthews

Lieven

Theakston

Tomasello

(2006). The effect of perceptual availability and prior discourse on young children’s use of referring expressions. Applied Psycholinguistics, 27, 403–422. doi:10.1017/S0142716406060334

26.

Mayer

(1969). Frog, where are you? New York, NY: Dial Press.

27.

Mayer

(1973). Frog on his own. New York, NY: Dial Press.

28.

McGann

Schwartz

(1988). Main character in children’s narratives. Linguistics, 26, 215–234. doi:10.1515/ling.1988.26.2.215

29.

Miller

J. F.

Iglesias

(2012). Systematic Analysis of Language Transcripts (SALT), Research Version 2012 [Computer Software]. Middleton, WI: SALT Software LLC.

30.

Orsolini

Rossi

Pontecorvo

(1996). Re-introduction of referents in Italian children’s narratives. Journal of Child Language, 23, 465–486. doi:10.1017/S0305000900008886

31.

Serratrice

(2007). Referential cohesion in the narratives of bilingual English–Italian children and monolingual peers. Journal of Pragmatics, 39, 1058–1087. doi:10.1016/j.pragma.2006.10.001

32.

Vion

Colas

(1999). Maintaining and reintroducing referents in French: Cognitive constraints and development of narrative skills. Journal of Experimental Child Psychology, 72, 32–50. doi:10.1006/jecp.1998.2475

33.

Warden

D. A.

(1976). The influence of context on children’s use of identifying expressions and references. British Journal of Psychology, 67, 101–112. doi:10.1111/j.2044-8295.1976.tb01501.x

34.

Whitely

Colozzo

(2013). Who’s who? Memory updating and character reference in children’s narratives. Journal of Speech, Language, and Hearing Research, 56, 1625–1636. doi: 10.1044/1092-4388(2013/12-0176)

35.

Wigglesworth

(1997). Children’s individual approaches to the organization of narrative. Journal of Child Language, 24, 279–309.