UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The development of coherence and complexity in the narratives of preschool-age children Gibney, Kimberley Marise 1995

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1995-0602.pdf [ 5.27MB ]
Metadata
JSON: 831-1.0099021.json
JSON-LD: 831-1.0099021-ld.json
RDF/XML (Pretty): 831-1.0099021-rdf.xml
RDF/JSON: 831-1.0099021-rdf.json
Turtle: 831-1.0099021-turtle.txt
N-Triples: 831-1.0099021-rdf-ntriples.txt
Original Record: 831-1.0099021-source.json
Full Text
831-1.0099021-fulltext.txt
Citation
831-1.0099021.ris

Full Text

T H E D E V E L O P M E N T O F C O H E R E N C E A N D C O M P L E X I T Y IN T H E N A R R A T I V E S O F P R E S C H O O L - A G E C H I L D R E N By K I M B E R L E Y M A R I S E G I B N E Y B.A., The University of British Co lumb ia , 1992 A T H E S I S S U B M I T T E D IN P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F M A S T E R O F S C I E N C E in T H E F A C U L T Y O F G R A D U A T E S T U D I E S T H E F A C U L T Y O F M E D I C I N E T H E S C H O O L O F A U D I O L O G Y A N D S P E E C H S C I E N C E S W e accept this thesis as conforming to the required standard T H E U N I V E R S I T Y O F BRIT ISH C O L U M B I A October, 1995 © Kimberley Mar ise Gibney, 1995 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of A^c^ \ £AOCJ v| a. y\ pi *S p-eec-U S ^ - e ^ c i e S The University of British Columbia Vancouver, Canada Date Oaf IV/IC DE-6 (2/88) II A B S T R A C T The general purpose of this study was to investigate the development of storytelling abilities in 3-and 4-year-old children. The goal plan analysis developed by Trabasso and his colleagues was supplemented with portions of Berman and Slobin's plot-based model of narrative structure (1994). This composite model was used to address the specific purposes of the study. The first purpose was to determine if the children had knowledge of goal plans of action, as reflected in their narration of a complex story. The second goal was to determine if this underlying structural complexity facilitated the production of a coherent narrative and if developmental differences could be seen between the two age groups on these measures. A final purpose of the study was to compare the results with those obtained in similar studies of plot and goal plan development by Berman and Slobin (1994), Trabasso and Nickels (1992), and Trabasso, Stein, Rodkin, Park Munger and Baughn (1992). The procedures for narrative elicitation follow those laid out by Berman et al. (1986) and used in the comparison studies mentioned. Ten English-speaking 3-year-olds (ages 3;2.10 to 3;9.19) and eleven English-speaking 4-year-olds (ages 4;4.3 to 5;0.11) participated in this study. Five of the children who participated as 3-year-olds also participated as 4-year-olds, providing longitudinal data. Each child told the story two times, first when they were unfamiliar with the book (Part 1), and again after reading the book with their caregiver(s) over the course of a week (Part 3). Results from the age-based comparisons showed some small group differences in the 3- and 4-year-olds' goal plan knowledge and their ability to narrate coherent and complex narratives. Since many of these differences did not capture the qualitative differences seen between the two groups of children, composite z-scores were used to calculate complexity and coherence rankings for the subjects based on the mean of the combined group. The results of these rankings indicated that the 3-year-olds were in transition from Ill producing locally coherent narratives with a single global goal plan to producing globally coherent and complex narratives. Ten of the eleven 4-year-old children told relatively coherent and complex narratives. These results were substantially different from those reported in the Trabasso and Berman and Slobin studies, which proposed that 3- and 4-year-olds have poor knowledge of goal plans of action and produce very simple narratives of the event or action-sequence type. It is proposed here that the differences are due to differences in experimental methodology, particularly for story and experimenter familiarity. iv TABLE OF C O N T E N T S A B S T R A C T ii LIST OF F IGURES x LIST OF T A B L E S xiii A C K N O W L E D G E M E N T S 1 C H A P T E R O N E INTRODUCTION AND REVIEW OF THE LITERATURE 1 Introduction 2 Review of the Literature 2 Narrative Production 7 Narrative Structure 7 Story Grammars 10 Plot-based Analyses 13 Goal Plan Analysis 15 The Development of Narrative Complexity 17 Narrative Coherence 18 The Development of Narrative Coherence 20 Summary 21 Statement of Research Questions 21 Goal Plan Knowledge 22 The Development of Plot Coherence 22 Primary Plot Coherence 22 Secondary Plot Coherence 23 The Development of Narrative Complexity 23 Primary Goal Plan Complexity 23 Secondary Goal Plan Complexity 24 Nongoal Plan Units 25 C H A P T E R TWO METHOD 25 Overview 26 Subjects 27 Data Collection Procedures 28 Part 1: Initial Story Elicitation 29 Part 2: Adult Story Modelling 30 Part 3: Story Retelling 30 V Transcription Procedures 30 Coding 31 Goals 34 Primary Plot Line 34 Goal Plan Unit Categories 35 Plot Opening Constituents 36 Plot Unfolding Constituents 39 Plot Concluding Constituents 41 Secondary Plot Line/Nonfrog Goal Plan Units 43 Nongoal Plan Units 44 Analysis 44 Story Selection 45 Age-based Comparisons 45 Goal Plan Knowledge 47 Primary Plot Coherence 52 Secondary Plot Coherence 53 Primary Goal Plan Complexity 55 Secondary Goal Plan Complexity 56 Nongoal Plan Units . 57 Content-based Comparisons 57 Coherence Analysis 58 Complexity Analysis 59 C H A P T E R T H R E E R E S U L T S 59 Overview 59 Age-based Comparisons 59 Goal Plan Knowledge 62 The Development of Narrative Coherence 62 Primary Plot Line 66 Plot Opening Coherence 69 Plot Unfolding Coherence . 72 Plot Concluding Coherence Measures 72 Secondary Plot Coherence 74 The Development of Narrative Complexity . . 74 Primary Goal Plan Complexity 74 Global Measures of Narrative Complexity 76 Complexity of the Plot Opening 80 Measures of Plot Unfolding Complexity 82 Measures of Plot Concluding Complexity 84 Nonfrog Goal Plans 87 Nongoal Plan Units 88 Content-based Comparisons 89 Coherence Analysis 93 Complexity Analysis 95 Comparison of Complexity and Coherence Rankings 97 vi C H A P T E R F O U R DISCUSSION 97 Overview 97 Age-based Comparisons 98 Goal Plan Knowledge 99 The Development of Narrative Coherence 100 Global Narrative Coherence 102 Plot Opening Coherence 104 Plot Unfolding Coherence 106 Search Coherence 108 Plot Concluding Coherence 108 Secondary Plot Coherence 110 The Development of Narrative Complexity 110 Global Goal Plan Complexity 112 Plot Opening Complexity 113 Pattern of Initiating Events 113 Plot Unfolding Complexity 115 Plot Concluding Complexity 116 Secondary Goal Plan Complexity 116 Goal Plan Complexity 117 Pattern of Secondary Goal Plans 117 Nongoal plan Units 119 Content-based Comparisons 119 Coherence Analysis 119 Complexity Analysis 120 Comparison of Complexity and Coherence Rankings 121 Summary 122 Data Comparison Considerations 124 Directions for Future Research 127 BIBLIOGRAPHY 130 A P P E N D I X A 130 Frog, Where Are You? (Mayer, 1969): Picture by Picture Description 130 vii LIST OF F IGURES Figure Page 1.1: Basic rewrite rules for generating story structures. 8 1.2: Schematic diagram of a hierarchical story structure. 9 3.1: Proportion of narrators encoding attempts at each location. 71 3.2: Distribution of narrators by number of Initiating Events. 77 viii LIST OF T A B L E S T A B L E P A G E 3.1a: 3-year-old subjects' performance across goal plan knowledge measures 61 3.1b: 4-year-old subjects' performance across goal plan knowledge measures 61 3.2: Number of narrators encoding take, find and search goals. 62 3.3a: Individual scores for 3-year-old subjects on essential story elements. 63 3.3b: Individual scores for 4-year-old subjects on essential story elements. 63 3.4a: Proportion of goal plan, nonfrog goal plan and nongoal plan units for 3-year-old subjects. 65 3.4b: Proportion of goal plan, nonfrog goal plan and nongoal plan units for 4-year-old subjects. 65 3.5a: Plot opening coherence measures for 3-year-old narrators. 68 3.5b: Plot opening coherence measures for 4-year-old narrators. 68 3.6a: Plot unfolding coherence measures for 3-year-old narrators. 70 3.6b: Plot unfolding coherence measures for 4-year-old narrators. 70 3.7: Number of narrators encoding find or both find and take outcomes by age group. 72 3.8: Individual and group results for secondary plot line coherence. 73 3.9a: Global measures of narrative complexity for 3-year-old narrators. 75 3.9b: Global measures of narrative complexity for 4-year-old narrators. 76 3.10a: Plot opening complexity of 3-year-olds' narratives. 77 3.10b: Plot opening complexity of 4-year-olds'narratives. 78 3.11: Rank order of initiating events for 3- and 4-year-old subjects. 80 3.12: Summary of mean scores for plot unfolding complexity measures. 81 3.13: Three and 4-year-old narrators grouped according to number of encoded attempts. 82 3.14a: Individual and group plot concluding scores for 3-year-old narrators. 83 ix T A B L E P A G E 3.14b: Individual and group plot concluding scores for 4-year-old narrators. 83 3.15: Three and 4-year-old narrators grouped according to number of nonfrog goal plans. 85 3.16a: Complexity of nonfrog goal plans for 3-year-oid narrators. 85 3.16b: Complexity of nonfrog goal plans for 4-year-old narrators. 86 3.17: Rank order of 3- and 4-year-old narrators encoding nonfrog goal plans. 87 3.18: Mean number and proportion of nongoal plan units by type. 88 3.19: Coherence ranking of 3- and 4-year-old narrators combined. 90 3.20: Coherence scores for 3- and 4-year-old narrators as a whole. 92 3.21: Complexity ranking of 3- and 4-year-old narrators combined. 93 3.22: Complexity scores and ranking for 3- and 4-year-old narrators. 94 3.23: Categorization of narratives by coherence and complexity. 96 X A C K N O W L E D G E M E N T S I would like to express my sincere thanks to the many people who have helped to make this project possible. Without their support and understanding I would never have made it this far. In particular, I would like to thank: Dr. Carolyn Johnson for her constant support, encouragement, and time over the past many months, Dr. Judith Johnston - who got more than she bargained for, John Nicol who solved many a mystery, my family for their concern, support, and frequent favours, my friends who rarely mentioned the "T" word, and Matthew Findlay who put up with me and believed in me, even when I didn't. A special mention is due to the children who participated in this study and their families, and to Norma Jean Gomme who collected the 3-year-old data and generously allowed me to use it. 1 C H A P T E R O N E INTRODUCTION AND REVIEW OF THE L ITERATURE Introduction Narratives play an important role in society and are used to teach, entertain and record past events. In order to tell a good story, a narrator must employ sophisticated linguistic, cognitive and social skills. A rich literature has developed around how children acquire these skills and subsequently apply them to narratives. The focus of this study was to investigate children's knowledge of goal plans of action and how that developing knowledge affects narrative coherence and complexity. Several linguistic and cognitive models have been proposed to account for narrative structure, production and comprehension. Each model has advantages and disadvantages, the source of much debate in the literature. The resolution of these debates has been hampered by disparities in experimental and task design that have made the results difficult or impossible to compare. In an attempt to resolve this problem, at least for picture-based narratives, some investigators (e.g. Bamberg, 1987; Bamberg & Marchman, 1990; Berman, 1988; Berman & Slobin, 1994; Trabasso, van den Broek & Suh, 1989; and Trabasso & Nickels, 1992) have chosen a standard task design using the wordless picture book, Frog, Where Are You? by Mercer Mayer (1969), to elicit narratives. A common task design enhances our ability to compare stories produced by different subjects, at different ages, and even in different languages. Most of the published analyses of young children's English-language frog stories have been carried out on stories that were collected by Virginia Marchman and Tanya Renner from twelve 3-year-olds 2 and twelve 4-year-olds (e.g. Berman & Slobin, 1994; Trabasso et al., 1992; Trabasso & Nickels, 1992). This study expands the database by a further ten 3-year-olds and eleven 4-year-olds, five of whom were longitudinal subjects. Modified versions of the plot and goal plan analyses presented in Trabasso et al. (1992), Trabasso and Nickels (1992), Trabasso and Rodkin (1994) and Berman and Slobin (1994) were used to measure and describe the children's narrative performance. The results of these analyses were compared to the original studies and others reported in the literature. Chapter One provides a review of the research literature concerning narrative production, structure, coherence, and complexity, and outlines the specific research questions addressed by the study. The experimental method, coding scheme, analysis procedures and subject information are detailed in Chapter Two. The results of the study, as they relate to the research questions posed in Chapter One, are provided in Chapter Three. The final chapter discusses the results with respect to the research questions and previous research in the areas of goal plan knowledge and narrative coherence and complexity. Review of the Literature Narrative Production The formulation and expression of a good narrative requires the integration of linguistic, cognitive and social skills. Each skill represents a possible performance variable on an experimental task that may have a significant impact on the length, complexity, or coherence of the narrative. Therefore, researchers investigating narrative competence must consider the effect of the task design on these variables, both in designing experiments and in interpreting and comparing results. 3 A wide variety of narrative tasks has been used over the years. Each one places different demands upon the individual's short and long term memory, knowledge of narrative structure and events, and cognitive and linguistic processing abilities. Some of the task variables to consider include: (a) Story genre: Is the narrative about real, fantasy or pictured events, and is the event sequence constrained by a script or must the narrator generate the sequence? (b) Story familiarity: Are the events forming the narrative familiar to the subject and/or the listener? (c) Story elicitation: Was the narrative told spontaneously, elicited using props (e.g. pictures, dolls, or films) or retold from a story previously heard or seen? Story genre affects both the subject's production of the narrative and the researcher's system of analysis. In their study of 4- to 8-year-old children's use of fantasy scripts, Seidman, Nelson and Greundel (1986) found that different genres of narrative required different event structures. This alteration at the event or local level of the narrative's structure in turn affects the global organization of the story, because the ties needed between events change as well. For example, the global structure of a script-based narrative about making a sandwich is linear because it is based on a sequence of steps that take place over a period of time, whereas a fantasy narrative about finding a talking frog is more likely to have a more complex, hierarchical structure. In addition, the temporal sequence of the script-based narrative is familiar and, therefore, predictable to both the narrator and the listener. However, the plot of the nonscript-based narrative must be generated by the narrator and pieced together on-line by the listener. The result is that the scripted narrative is easier to process for both the narrator and the listener, which in turn affects the complexity and coherence of the narrative. 4 From an analysis viewpoint, it is more difficult to draw conclusions from stories about imagined events because each narrator has generated the story in a very different way. The story may have a very simple or a very complex underlying structure, depending on the events chosen and how the narrator interprets the task. In this case, the narration of a simple story must be interpreted carefully and is not an absolute measure of narrative competence since it may underestimate the narrator's abilities. As mentioned above, the genre of the narrative is closely tied to the variable of story familiarity. Narratives about real, past, and scripted events are the most familiar subject matter for both narrators and listeners. The basic semantic content, temporal order, and relations between events are already known, which greatly simplifies the task of incorporating them into a good story. In contrast, fantasy narratives require a great deal of semantic and linguistic preplanning to create an event sequence that can then be adapted to the appropriate story structure (Orsolini, 1990). Picture-based narratives may be considered intermediate, in that they provide basic event information such as the characters involved and the main events of the story. Depending upon the picture sequence used, narrators may need to make inferences about the story context and the temporal and causal relations between the events. Narrators may interpret this type of task as a picture description instead of a narrative, and one should be alert to signs of descriptive voice in the text (e.g. frequent use of deictic pronouns or progressive aspect). Narrators have limited capacities for linguistic and cognitive processing, and generating story content and structure uses varying amounts of these capacities. Increasing event and/or story familiarity should reduce the linguistic processing required to generate a story because the narrator has the propositions and the causal and temporal relations between events stored in memory. In 5 turn, this type of task is thought to be more memory intensive than spontaneous narration. In a comparison of twenty 9- to 11-year-old normal children and twenty age-matched language-delayed children's spontaneous narratives and story retellings,1 Merritt and Liles (1989) found that familiarity with the content of the story produced a positive effect on plot and episode complexity for both groups. The familiar stories, which were based on videotaped readings of a short story the children heard and saw, contained a higher number of story grammar components (see pages 7-10 in this chapter for explication) and more complete episodes than the narratives generated from story stems, such as "Once upon a time, two friends were in a deep and dark cave." They concluded that the retellings were more complex because more of the child's processing resources could be allocated to structuring the content into coherent episodes. Therefore, the story retellings more accurately represented the children's competence with underlying story structure. Listeners can also be 'familiar' with narratives - either with the semantic content of a particular narrative and/or with the underlying structure of narratives in general. In a study of the effects of listener familiarity on twenty 7- to 10-year-old normal and twenty age-matched language-delayed children's narratives, Liles (1987) found that it had an opposite effect to narrator familiarity on episodic complexity. When the normal children were aware that their adult listener was familiar with the story, their story retellings contained fewer complete episodes and they used more pronouns to refer to people and events across sentences. Liles concluded that these narrators were sensitive to the background knowledge of their listeners and adjusted the content of their stories accordingly. This adjustment implies that narrators are aware of their listeners' needs and 1 Story retelling tasks can be structured in different ways, depending on how children are familiarized with the story they are to retell. The initial story can either be generated by the child and then retold, or the child may retell a standard story presented to her by the experimenter either auditorily, visually, or both. 6 employ the Gricean principle of avoiding redundancy in their stories (Grice, 1975). The result is that certain events (e.g. complex reactions and beginning events) can be reduced or deleted from the narrative because the information can be inferred by the listener from her own knowledge (Johnson & Mandler, 1980). Interestingly, the language-impaired subjects in Liles's study did not employ this strategy, indicating that they had not acquired some of the more subtle discourse skills required to be good narrators. It appears that familiarity has different effects on narrative complexity and coherence depending on whether it is the narrator or the listener who is familiar with the story. In the first case, the processing demands are reduced and, therefore, the individual has resources available for telling a more complex and/or coherent narrative. In the second case, the narrator does not want to bore the listener with too much detail or repetition and is aware that there are story components that can be eliminated. The resulting reduction in the surface complexity of the story is compensated for by an increase in the cohesiveness of the narrative, particularly through the use of pronouns. The effects of the third task variable are similar to those discussed for story genre and familiarity. As mentioned above, the elicitation method plays a large role in determining the underlying semantic structure of the narrative. In the case of spontaneous narratives, no props are provided, and, therefore, the story must be generated from either previous or fantasized experiences. In her analysis of 3-,4-, and 5-year-olds' spontaneous conversational narratives, Umiker-Sebeok (1979) found them to be more variable than those based on props because of their interactive nature. The listener actively participates during the narrative, rather than listening to it passively. Depending on how the listener responds to a particular event, the narrator has the flexibility to either pursue or abandon that episode because there is no predetermined 'plot' to the story. 7 In contrast, when a narrative is based on props such as puppets or a picture sequence, the narrator is more constrained in terms of story .content. Therefore, more of the processing load can be devoted to discourse planning (Orsolini, 1990). As a result, narratives elicited in this manner may be more coherent and contain more local cohesive devices such as pronouns. In story retelling tasks, the content of the narrative is fixed and the majority of the processing load is placed on the individual's memory. Assuming the narrator's memory of the story is good, then the attentional focus can shift to developing the linguistic and textual complexity of the narrative. Narrative Structure In order to understand how children acquire narrative abilities, we must have a theory of how stories are structured. Story structures have been represented in different ways depending on the investigators' goals for the analysis and the narrative genre studied. The three main types of story structure models are: story grammars, based on underlying mental story schemas, (e.g. Johnson & Mandler, 1980; Stein & Glenn, 1979; van Dijk, 1980), plot analyses of story content (e.g. Botvin & Sutton-Smith, 1977; Berman & Slobin, 1994; Gillam, 1989) and goal plans of action (e.g. Trabasso & Nickels, 1992; Trabasso, van den Broek & Suh, 1989; Trabasso, Stein, Rodkin, Park Munger & Baughn, 1992; Trabasso & Rodkin, 1994). Story Grammars Story grammars were developed to explain the structural regularities seen in stories. Story grammars are an integration of previous research in the areas of story structure, particularly on folk tales, discourse processing, and generative grammar. While several different grammars have been published (e.g. Johnson & Mandler, 1980; Stein & Glenn, 1979; van Dijk, 1980), they all share a number of components. The grammar consists of a set of story constituents generated by a set 8 of recursive formation or rewrite rules, which create a hierarchical story structure. The resultant story structure can be altered by a second set of rules called transformational rules. This level of the narrative structure has also been called the macrostructure (van Dijk, 1980). Each proposition in the narrative is part of the microstructure and is the linguistic expression of a story constituent. The knowledge basis of the grammar is a story schema, which is "a mental structure consisting of sets of expectations about the way in which stories proceed" (p. 17 Mandler, 1984). The schema forms the highest level of narrative organization and corresponds to van Dijk's (1980) notion of narrative superstructure. The model developed by Johnson and Mandler (1980) is one of the most comprehensive story grammars. In addition to the standard set of rewrite and transformational rules, they developed rules for embedding and coordinating episodes and movement and deletion rules for constituents. The rewrite rules shown in Figure 1.1 are a subset of those included in the model. S T O R Y ^ Setting And EPISODE EPISODE -+ fBEGINNING Cause D E V E L O P M E N T Cause ENDING IEPISODE And/Then EPISODE BEGINNING -»• Beginning Event D E V E L O P M E N T -> fCOMPLEX REACTION Cause G O A L PATH ISimple Reaction Cause Action C O M P L E X REACTION -> Simple Reaction Cause Goal G O A L PATH Attempt Cause O U T C O M E O U T C O M E -> Outcome Event ENDING -• fEnding Event [EPISODE Figure 1.1: Basic rewrite rules for generating story structures (adapted from Johnson & Mandler, 1980). 9 S T O R Y 1 Setting (A) Episode Episode 1 (?) Episode 1 Begin (Q) Development © 1 End Begin @ Development (c) End Complex (b) Goal Path Complex (C) Goal Path Reaction 1 Reaction 1 Simple © G o a l Attempt (S) Outcome Simple (c) Goal Attempt © O u t c o m e Reaction Reaction (A> AND Relation © = THEN Relation © = C A U S E Relation Figure 1.2: Schematic diagram of a hierarchical story structure (adapted from Johnson & Mandler, 1980). Figure 1.2, above, provides an example of how the grammar in Figure 1.1 could be used to describe a story in terms of episodes, story constituents and the coordinate, temporal and causal relations between them. In this case, the story has a hierarchical structure, with the main episode consisting of two temporally related subepisodes. Each of the subordinate episodes starts with a beginning event that causes the protagonist to react and establish a goal and a means to achieve that goal. An attempt is made to achieve each of the goals. Each of these attempts is causally related to its outcome. In turn, these outcomes are causally related to the endings of their episodes. The advantages of this type of analysis are that it captures the hierarchical structure of stories, links the underlying structure of narratives to models proposed for syntax, phonology and semantics, and captures many of our intuitions about what a "good" story is. However, these grammars also have some limitations. Story grammars evolved from analyses of folk tales and do not adequately 10 account for the structure of either more modern or non-European narratives. Therefore, unlike other generative linguistic models, story grammars have reduced potential for use in crosslinguistic studies or investigations of language universals (Black & Bower, 1980). In addition, since the grammar models are structurally biased toward literacy-based narratives, they may also fail to adequately describe the earliest narrative structures and narrative development in preliterate children. Plot-based Analyses Unlike story grammars, which focus on the structural interactions among story constituents, plot-based analyses focus more on the semantic content of the narrative (e.g. Applebee, 1978; Berman & Slobin, 1994; Botvin & Sutton-Smith 1977; Gillam 1989; Propp, 1968). These analyses developed from roots similar to those of story grammars and share many of the same concepts, such as episodes and the initiation of action. However, most plot analysis models are more focussed towards cognitive, rather than linguistic, aspects of storytelling (e.g. levels of conceptual development, processing demands of complex language). Three concepts shared by these analyses are centering (the central theme of the narrative), chaining (associative relations between events) and causality (temporal and causal links between events), proposed by Applebee (1978). In his book, The Child's Concept of Story (1978), Applebee outlines a hierarchy of plot development based on analyses of 120 2- to 5-year-old children's fantasy narratives. He argues that narrative complexity develops in conjunction with conceptual development, as described by Vygotsky (1934). He terms children's earliest stories heaps, which are collections of unrelated events with no motivating purpose or theme. Over time, children learn to center the sequence events on a theme and to bring the narrative to a logical conclusion through the use of associative relations, and 11 temporal and causal links. These processes correspond to the goal plan, cohesion, and causal chain portions of the model proposed by Trabasso and his colleagues. In another early plot-based model, proposed by Botvin and Sutton-Smith (1977), each clause in the narrative is analysed for its role in the story and it relationship to other story elements. The aim is to obtain a description of the narrative's plot structure. Unlike a story grammar, there are no constraints on what the function of the next action element should be. In their study of 220 children aged three to twelve, they broke down the children's spontaneous fantasy narratives into primary and secondary plot units. They define primary plot units as elements that "serve to delimit the action of a narrative or episode within a narrative and always occur in pairs, forming dyads" (e.g. lack + lack-liquidated) (p. 378). Secondary plot units serve to "mediate the action established in the initial primary plot unit and lead the action of the narrative to the final primary plot unit" (p. 378). As in a story grammar, more complex narratives are formed by sequencing or embedding subordinate dyads to form subplots within the main plot. Botvin and Sutton-Smith found that the narratives fit into seven levels of narrative complexity, with the children generally producing longer and more complex narratives as they got older. In addition, they make the important observation that structural complexity appears to facilitate the production of a longer narrative. They argue that increased structural complexity reflects a higher level of cognitive organization. This organization eases the child's planning load, thereby allowing her to allocate more processing resources to encoding more information in the story. The complexity hierarchy proposed by Botvin and Sutton-Smith (1977) is adapted from the work of Propp (1968) and expands the Applebee hierarchy to seven levels of story structure complexity. Level one narratives lack both coherence and structural relationships, as in Applebee's heap 12 category. The second level of story structure has one nuclear dyad, which centers the story. The only secondary plot units occur before the members of the nuclear dyad are stated (i.e. setting information). In the third level, the narrative is elaborated by secondary plot units that are interspersed between the primary units. At the fourth level, fewer secondary plot units are included, but two or more nuclear dyads are concatenated using 'and'. Two or more well-developed episodes characterize the fifth level of story complexity. Both the primary and secondary plot units are elaborated at this level. The first hierarchical story structure occurs at level six with the use of embedded subplots. The highest level of story complexity is characterized by multiple embeddings of dyads with numerous subplots. This level of complexity describes the plot of Frog, Where are You?, which was used in the present study. There are several advantages to a model of story structure like the Botvin and Sutton-Smith model. In particular, it is less prescriptive about story structure and content than the story grammars, and the levels of complexity they describe correlate well with Vygotsky's levels of cognitive development. This link with cognitive development provides a possible explanation for the developmental changes seen. However, they do not address the structural similarities accounted for by generative models and schema theory, and the analysis becomes awkward when applied to narratives with multiple episodes. Berman (1988) devised a plot-based model of narrative analysis for use in her cross-linguistic research into narrative development from age three to adulthood. She attempts to address the issue of universal narrative structures, raised above, by proposing a basic structural framework that is common to all narratives. In her model, the plot of the narrative consists of three primary categories: onset of the plot, unfolding of the plot, and resolution of the plot. The plot elements 13 needed for a good narrative are assigned to the appropriate primary category. The onset should include all of the setting information and the events establishing the story's central conflict. The unfolding includes the protagonist's attempts to resolve the conflict, while the final outcome of the attempt is assigned to the resolution. Because of this improved structure, this model is better suited to the analysis of long texts such as Frog, Where are You?. Berman supplements the structural analysis with linguistic analyses of tense, connectivity, given/new information, and evaluative comments, to form a more comprehensive description of narrative development. In general, several goals are achieved with plot-based analyses. They describe the semantic content children choose to include in their stories and account for both the local and global cohesion strategies used in narratives (e.g. temporal and causal conjunction and central plot dyads). They may also be used to analyse a wider variety of narrative types, because they do not rely on a set of generative rules to account for the story's structure. This flexibility has also been shown to be useful for analyzing the narratives of children younger than five (Berman, 1988; Berman & Slobin, 1994). In addition, they investigate the effects of linguistic and cognitive development on the development of narrative competence. A small subset of the analyses reported by Berman and Slobin (1994) were used in conjunction with a goal plan analysis in this thesis research. Goal Plan Analysis The goal plan analysis developed by Trabasso and his colleagues (Trabasso, van den Broek & Suh 1989; Trabasso & Nickels, 1992; and Trabasso, Stein, Rodkin, Park-Munger & Baughn, 1992) combines many of the concepts of story grammars, plot-based analyses, and Schank & Abelson's causal chain theory (1977). Trabasso et al. (1992) adopt the story constituent categories defined 14 by Stein and Glenn (1979) and Johnson and Mandler (1980), such as setting, initiating event, attempt, purpose, goal, reaction, and outcome, and use them to characterize the content of the narrative. They also analyse the causal, temporal, and enabling links between narrated events and states, story constituents, and episodes. These links are inferred and may be psychological, motivational, or physical causes. Trabasso and his colleagues call this system of links a causal network. They propose that the causal network model can be used to represent and generate hierarchical goal plans. Hierarchical goal plans are generated when more than one goal is established for the protagonist. Essentially, a goal plan represents the narrator's understanding of how to interpret and integrate events in a coherent way. The result of this understanding is the plot of the story, as outlined by Berman (1988). Trabasso and his colleagues agree with Berman and Slobin, in that they view narration as a "joint process of event comprehension and language production" (Trabasso & Rodkin, 1994; p.87). In order to produce a narrative, the storyteller must understand what has happened, why it happened, and what might happen next and.then formulate coherent, grammatical sentences to convey this information to the listener. They believe that we can infer what children know about planning and hierarchical goal plans through the content they choose to encode in their narratives (Trabasso & Rodkin, 1994). However, unlike Berman, they include in their analysis plot elements that are not explicitly stated in the narrative but can be inferred from the elements that the narrator has included. For example, a failed outcome for an attempt can be inferred if a second attempt is narrated. Given Liles's (1987) finding that 7- to 10-year-old children adjust the content of their narratives to avoid redundancy in a condition similar to that used in this study, it seemed prudent to assume that 15 the children in the present study might behave in a similar fashion. Therefore, goal plan analysis was chosen as a measure of narrative structure - and, hence, complexity - because it credited narrators using this strategy, rather than penalizing them, as the Berman plot analysis does. This analysis presumes that local and global goal plans constitute the underlying representation of the plot of the story. By analysing the events and states of the narrative to determine their narrative functions and relationships with other constituents, we ought to be able to determine the extent of the narrator's knowledge of goal plans. This reasoning forms the foundation for the measures used to answer the research questions on complexity outlined at the end of this chapter. The Development of Narrative Complexity Applebee (1978) found that most 2-year-olds could tell narratives with a central theme, and 3- and 4- year-olds' narratives generally had a central theme and chaining relations between events. In contrast, Botvin and Sutton-Smith (1977) found that almost all of their 3-year-olds told narratives of the lowest level of complexity. A substantial increase in the complexity of the narratives was seen between ages four and five, with the majority of the 5-year-olds producing narratives with a single elaborated episode. Given that the tasks used in these studies were quite similar (spontaneous fantasy narratives), a reasonable explanation for the disparity in results is sampling error, since the Botvin and Sutton-Smith study had only four 3-year-olds, six 4-year-olds and ten 5- year-olds compared to Applebee's groups of 30 children in each of the 2-,3-,4-, and 5-year-old age categories. If this explanation holds, it underlines the importance of basing age expectations on substantial numbers of children. Stein and Glenn (1979) described the development of story structures using a grammar similar to Johnson and Mandler (1980). Unlike the previous hierarchy, each new level entails the structures 16 found in the previous level plus an additional degree of complexity. However, the stories still progress from unrelated sequences to coherent narratives with logical conclusions and embedded episodes. At the earliest ages, children produce primarily descriptive sequences, which describe the event and its context but do not include any logical or cohesive relationships between elements. The second level of structure is the action sequence, in which the elements of the story are related temporally. In the third level, the reactive sequence, causal relations are added to the story structure. For an abbreviated episode story structure, an implicit goal structure is added to the narrative, using an initiating event/consequence pair. If the goal sequence is explicit, the story structure is called a complete episode. Complex episodes contain an embedded episode and/or plan within the central episode of the story. The final level of this hierarchy is the interactive episode, in which more than one plot line is maintained. As mentioned above, task has been shown to have a significant impact on the complexity of narratives. The hierarchies mentioned above were based primarily on children's productions of fantasy narratives, which require much more linguistic and cognitive processing than picture-based narratives. As a result, they may underestimate children's competence with other narrative tasks. However, they remain useful as illustrations of the process of change, particularly in the early stages of narrative development. In fact, the developmental results reported for studies using the story picture-elicitation method fit the same general pattern seen for fantasy narratives. The consensus is that 3- and 4-year-olds produce mainly narratives of the type defined above as descriptive and action sequences. The narratives are either made up of descriptions of the characters and their surroundings or are a series of unrelated events. More of the 4-year-olds produce narratives with event sequences that 17 are temporally related and script-like, and a few show evidence of increasing knowledge of goal plans. It is not until age five that children provide consistent evidence that they have knowledge of goals and goal plans in their stories. This result was obtained using both Trabasso's causal chain/goal plan model and Berman's plot line analysis. However, these analyses were done on a small number of English-speaking subjects (twelve children in each age group), so the possibility of a sample effect exists. This possibility is particularly important to consider for the 3-, 4-, and 5-year-olds, who appear to be at ages of significant narrative development (Applebee, 1978; Berman & Slobin, 1994; Botvin & Sutton-Smith, 1977; Trabasso & Rodkin, 1994). Narrative Coherence Coherence occurs at both the local and global levels in narrative discourse. The definition of coherence used here is in keeping with that used by Berman and Slobin (1994) and Trabasso et al. (1992). Likewise, Bennett-Kastor defines coherence as "that quality which makes the discourse stand as a whole text; that is, which makes the utterances of the text related to one another in some salient way, and not just random strings" (1983, p. 136). Local coherence is the result of linguistic elements that signal inter-relationships between clauses and sentences (e.g. cohesive devices such as and and then). Global coherence is created through devices such as sustained local coherence, selection of an anchor tense, world knowledge and plot organization. Narrators use both bottom-up processes, such as event description, and top-down processes, such as knowledge of goals and plans of action, to produce coherent narratives. For example, Trabasso and Nickels argue that global narrative coherence is achieved "when children use naive theories about intentional action to infer and encode information about goals and goal plans of action" (1992, p. 250). The coherence of the narrative is closely intertwined with its complexity, in that the child's knowledge of story structure has a major effect on both the complexity and the coherence 18 of the narrative told. Including more goals and goal plans of action makes the narrative - by-definition - more complex. These same story elements provide relational information that makes the narrative more coherent to the listener. The Development of Narrative Coherence Following Trabasso and Nickels's argument that coherence is the result of both top-down and bottom-up processes, what might the pattern of development look like? Intuitively, we would assume that narrative coherence ought to be highly correlated with age, and several studies have verified this basic finding across different narrative genres (e.g. Applebee, 1978; Botvin & Sutton-Smith, 1977; Berman & Slobin, 1994). Because of the ages of the children involved in my study, I will limit my discussion of the development of narrative coherence to children aged two to nine. Since the focus of this research is the effect of goal plan knowledge on narrative coherence and complexity, I will not discuss the many linguistic elements that contribute to coherence (e.g. tense, conjunction, reference). Berman and Slobin (1994) propose that 3-year-olds have little knowledge of goals or goal plans of action, and therefore their narratives are poorly structured and only minimally coherent at the local level, with no global coherence. At four years of age, Trabasso and Rodkin state that the children are in a state of transition from "describing isolated states, actions unrelated to the plan, and neutral outcomes" to "taking the perspective of the main protagonist, encoding initiating events, encoding attempts without purposes, and encoding the ending" (1994, p. 102), indicating they have some knowledge of goal plans. By age five, they are able to access and use underlying goal plans to selectively encode elements of the story. 19 Several changes have been reported between the ages of two and nine with respect to local coherence devices, which in turn contribute to global coherence. Berman and Slobin (1994) found that the preschool age children in their study progressed from "spatially-motivated linking of utterances as picture-by-picture description (3-year-olds)" to "temporal organization at a local level of interclausal sequential chaining of events (most 5-year-olds)". It was not until age nine that the narratives contained "sequential and/or causal chaining of partially elaborated events" (p. 58). However, in her study of 3-year-olds' use of referential pronouns, Gomme found that they were developing the ability to "create textual coherence through pronominal reference" by distinguishing between how they referred to the main protagonist versus any other character (1994, p.112). This "thematic advancement strategy" was proposed by Bamberg (1987) in his study of German children's narrative development. For fantasy narratives, many children acquire the ability to center their narrative around a central conflict or theme by about age two (Applebee, 1978). That is to say that they understand that a narrative must have an underlying purpose or "point" that motivates its telling. However, their narratives contain few, if any, causal or temporal links between utterances and, therefore, require a great deal of inference on the part of the listener. By four years of age, the narrators assume more of the responsibility for conveying the main components of the story and use a higher proportion of causal links, creating, first local and then textual, coherence. By age six, almost all of the key components of the plot are included and narrators are able to provide motivational links for the events in their stories. (Kemper & Edwards, 1986). 20 Summary People are able to tell stories because they have an underlying concept of how narratives are structured and what a 'good' story is. These underlying concepts are reflected in the content and organization of the narratives they produce. A good narrative must have a certain level of complexity in order to properly motivate the initiation of a goal for the story, convey all of the major events that occur, and provide some resolution for the goal. In addition, that underlying structure should be coherent, with a central theme, and local temporal and causal links between components. Three common models of narrative structure are story grammars, plot-based analyses, and goal plan analyses. The model of underlying narrative structure motivating this thesis research is the goal plan of action developed by Trabasso and his colleagues (1989, 1992, 1994). It captures the structural generalizations seen in narratives, while avoiding many of the problems associated both with rigid, generative models, such as story grammars, and plot-based analyses, which focus on the cognitive factors of narration at the expense of the social aspects (e.g. the ability to avoid redundancy in order to accommodate listener knowledge). The output of the goal plan is the plot of the narrative, as defined by Berman (1988). Her concept of a plot complements the goal plan analysis by describing how the underlying narrative structure and the event knowledge are combined to produce a coherent narrative. For the analysis done here, portions of the two models were combined in order to best capture the interaction of structural complexity and story content. The goal plan was divided into three plot components -- the onset, unfolding and resolution of the plot. Within each of these components are the plot elements of a narrative (e.g. setting information, initiating events, attempts to achieve a goal, outcomes of the 21 attempts). The number of plot units that are included, and their functions within the narrative dictates the complexity of the goal plan. Similarly, the integration of these plot units into the narrative as a whole (i.e. the formation of temporal, causal, and associative relations between story components) dictates the relative local, and global, coherence of the plot. Certain expectations for how the children will perform on the task can be formed, based on the Berman and Slobin (1994) and Trabasso et al. (1992) studies, as well as other studies of narrative development. In these studies, 3-year-olds produced narratives that had a central theme but few, if any, relations among utterances. The narratives were made up largely of event and state descriptions, and naming of objects and characters in the various scenes. The 4-year-olds performed slightly better with a few more narrators including locally coherent portions of the plot. However, global coherence and complexity were still poor. Statement of Research Questions The following questions were asked to determine the extent of 3- and 4-year-olds' knowledge of goal plans, and how this knowledge affected the coherence and complexity of their narratives. The coherence and complexity issues were broken down according to the portion of the narrative in question and the plot line or goal plan involved. Goal Plan Knowledge Question 1: Do preschool aged children have knowledge of underlying goal plans of action, as measured by: (a) the presence of goal-relevant attempts, purposes and outcomes, (b) encoding multiple attempts at a goal, (c) goal-attempt-outcome sequences, and (d) encoding sufficient information for the listener to infer superordinate (Take) and subordinate (Search and Find) goals? 22 The Development of Plot Coherence Primary Plot Coherence Question 2a: How does the pattern of global narrative coherence change from ages 3 to 4, as measured by a) the number of essential story elements, and b) the distribution of global goal plan, nonfrog goal plan, and nongoal plan units? Question 2b: Did the 4-year-old subjects encode more information relevant to the opening of the plot than the 3-year-old subjects? Question 2c: Did the 4-year-olds' narratives unfold in a more coherent manner than the 3-year-olds', as demonstrated by: a) more frequently sustaining the search for the frog until the end of the story, b) including more attempt-outcome pairs, and c) dedicating a larger proportion of the story to achieving the three goals? Question 2d: Do the two groups show different patterns of goal plan continuation over the length of the story, as indicated by the locations of the attempts? Question 2e: Did more 4-year-olds than 3-year-olds encode a successful outcome for the superordinate goal plan? Secondary Plot Coherence Question 3: Did the 4-year-olds have more coherent secondary plot lines than the 3-year-olds, as measured by a) the number of nonfrog Goal-Attempt-Outcome sequences, b) the proportion of G A O ' s to total nonfrog episodes, and c) the number of shared elements between nonfrog episodes? 23 The Development of Narrative Complexity Primary Goal Plan Complexity Question 4a: Did 4-year-olds produce either a) longer narratives, b) narratives containing more goal plan units, or c) more G A O sequences than younger subjects? Question 4b: Was the opening of the plot more complex in the 4-year-olds' stories than in the 3-year-olds' stories, as measured by a) the characters they introduce, b) the number of opening events they include, or c) the number of initiating events motivating the goal plan? Question 4c: Did the 3-year-olds and 4-year-olds differ in the particular initiating events they choose to include? Question 46: Does the pattern of narrated attempts change between three and four years of age, as measured by: a) the number of attempts, b) the number of attempts made to achieve superordinate and subordinate goals, c) the number of purposeful attempts, and d) the number of attempts made by the boy versus those attempts made by the dog? Question 4e: Is there a discernable difference in how the 3-year-old and 4-year-old groups encoded outcomes for either local or global goal plans? Secondary Goal Plan Complexity Question 5a: Is there a difference between the two groups in the complexity of their nonfrog goal plans, as measured by a) the number of nonfrog goal plan units, b) the number of nonfrog goal plans, and c) the mean number of units per goal plan? Question 5b: Is there a difference between the 3-year-old and 4-year-old groups in the pattern of nonfrog episodes that the children included in their stories? 24 Nonqoal Plan Units Question 6: Did the 3-year-old and 4-year-old groups differ in the kinds of nongoal plan units they encoded? C H A P T E R TWO 25 METHOD Overview The main purposes of the study were to investigate preschoolers' knowledge of goal plans, as reflected in their narration of a complex story, and to determine if there were developmental differences in the coherence and/or complexity of the 3- and 4-year-old subjects' stories. A secondary purpose of the study was to compare the results with those obtained in similar studies of plot and goal plan development by Berman and Slobin (1994), Trabasso and Nickels (1992), and Trabasso, Stein, Rodkin, Park Munger and Baughn (1992). The procedures for narrative elicitation used in this study follow those laid out in Berman et al. (1986), and used by several other researchers (e.g. Bamberg, 1987; Berman & Slobin, 1994; Trabasso & Nickels, 1992; Trabasso, Stein, Rodkin, Park Munger & Baughn, 1992). Data was collected from two subject groups, 3-year-olds and 4-year-olds, on a longitudinal (five subjects) and cross-sectional (eleven subjects - five 3-year-olds and six 4-year-olds) basis. A wordless picture book entitled Frog, Where are You? by Mercer Mayer (1969), was used to elicit narratives by the children. The study consisted of three parts conducted over a one-week time period: (1) Initial Story Elicitation (2) Adult Story Modelling (3) Story Retelling. For the first part, children accompanied the experimenter to a research room just outside their regular preschool classroom. The children were allowed to look through the book until they were ready to tell the story. The stories elicited were video and audio recorded. The second part of the study involved caregivers reading the book at home with the child a minimum of four times. The third portion of 26 the study involved re-elicitation and recording of the story. The data was then transcribed and coded according to the categories outlined below. Subjects Longitudinal and cross-sectional data were gathered from sixteen subjects by the author and a previous graduate student researcher, Norma Jean Gomme. An initial subject group often 3-year-olds (age range 3;2.10 - 3;9.19, mean 3;5.26) was selected and one year later, a group of eleven 4-year-old subjects (age range 4;4.3 - 4 ; 12.11, mean 4;8.13) was chosen. Half of the 3-year-old subjects (two girls and three boys) were available to participate again the following year as 4-year-olds. All subjects were selected from the preschool program at the U.B.C. Child Study Centre with the assistance of the Program Coordinator of the Centre. Only monolingual English-speaking children with no known developmental or language delays were nominated for consideration as subjects. A letter of introduction outlining the procedures and purposes of the study and a consent form were sent to the parents or guardians of each of the children who were nominated to be subjects. The 3-year-old subjects (six girls and four boys) and 4-year-old subjects (six girls and five boys) were chosen depending on their availability, willingness, and prior participation as 3-year-olds. Provided in Table 2.1 is a description of the two subject groups, with "X" indicating a cross-sectional subject and "L" indicating a longitudinal subject. The 4-year-old subjects are listed in the order in which their narratives were elicited, from earliest to latest. The longitudinal 3-year-olds were assigned the same subject number given to them as 4-year-olds and, therefore, they occur on the same line in both age columns. The cross-sectional subjects were assigned the remaining subject numbers. 27 Table 2.1: Age and gender of longitudinal and cross-sectional subjects 3-year-old Subject Group 4-year-old Subject Group Subject Gender Age* (y;mm.dd) Subject Gender Age* (y;mm.dd) 1 BT ? 3;08.21 1 BT (L) ¥ 4;11.21 2 KS ? 3;08.12 2 JM (X) ¥ 4;09.08 3 S K ? 3;06.09 3 S K (L) ¥ 4;09.02 4 DM o" 3;03.27 4 DM (L) 4;06.03 5 JT ¥ 3;05.24 5 R P (X) ? 4;10.26 6 DA 3;09.19 6 DA (L) 5;00.11 7 JB ¥ 3;05.09 7 R W (X) ¥ 4;04.03 8 EH ¥ 3;02.10 8 RC (X) d* 4;04.18 9 R S c? 3;06.25 9 G H (X) ¥ 4;10.21 10 RL 3;04.30 1 0 R L ( L ) 11 LH (X) cT d 1 4;07.06 4;07.24 * Ages given are subjects' ages at the time the narrative selected for analysis was elicited. Data Collection Procedures In order to facilitate the comparison and interpretation of results, a common experimental design has been used by a number of researchers (e.g. Berman, 1988; Trabasso et al., 1992). Mercer Mayer's Frog, Where are You? was selected because it provides a standard plot and visual support for generating the story. In addition, it contains both global and local episodes, involving two different characters, which allows for a very complex narrative. A synopsis of the story is provided in Appendix A. 28 The story data was collected in a research room at the University of British Columbia Child Study Centre, with the exception of one story by subject EH, which was elicited at home. The data from the 3-year-old subjects was collected by Norma Jean Gomme as part of her Master of Science thesis research and were used for the present study with her permission. The study involved a sequence of three parts, two parts involving data collection from individual children at the preschool site and one part involving parental participation at home over a time span of one week. Part 1: Initial Story Elicitation Before asking the child to leave his or her classroom to tell the story, the experimenter spent at least three hours in the child's classroom, participating in regular classroom activities. Only when the child seemed comfortable with the experimenter's presence was he or she invited to play a story game. If the child chose not to participate despite gentle encouragement, he or she was invited again another day. No child refused to participate in the storytelling task. Each child was allowed to look through the picture book Frog, Where are You? by Mercer Mayer (1969) until he or she knew the book well enough to tell the story. The child and experimenter looked at the book together while the child told the story. When the child finished telling the story, the experimenter asked if he or she would like to tell the story again. The majority of the children did not choose to tell the story a second time, presumably because they felt that they had already completed the experimenter's task. Because research has shown that listeners' comments can affect the content and structure of the narrative (e.g. Berman et al., 1986; Berman, 1988), the experimenter restricted her participation in the task to a preselected set of phrases outlined in Berman et al. (1986). These phrases were 29 used to encourage the child to continue participating (e.g. "What's happening?" or "Then what happened?"), to respond to narrator attempts to elicit listener comment/reaction (e.g. "Wow!", "Uh-oh!", "Yeah?" or "Uh-huh.") and to clarify a child's utterance for transcription purposes (e.g. repeating a portion of child's utterance or requesting repetition from the child). The stories were audio and video recorded except story 1 for subject 9 (RS) due to battery failure and stories 1 and 2 for subject 8 (EH) due to her distractibility. The audio recording equipment consisted of an audio recorder (Marantz, model PMD420), lapel microphone (Samson remote), audio transmitter (Samson VHF FM ST-2) and an audio receiver (Samson VHF FM SR-2). Fuji FR-II Super C r 0 2 audio tapes were used. The audiotapes were played on a T A S C A M 112 audio cassette recorder over Sennheiser (HD 520II) headphones. The video recording equipment consisted of a camera (JVC Super VHS GF-S 550) and a tripod (Davis and Sanford, Model F-10). Professional quality video tapes were used (3M Scotch Colour Plus High Grade). The videotaped stories were viewed on a Sony (Trinitron KV1367) television using a J V C (HR D670U) video cassette recorder. Part 2: Adult Story Modelling Following the initial story elicitation, a copy of the picture book was sent home with the child. The parents were instructed to read the book with the child at least four times over the next week. Parents were not provided with a standard script to use when telling their child the story. Parents were asked to encourage the active participation of their child during the story. A note reminding the parents of the procedures for this portion of the study was sent home with the book along with the researcher's telephone number in case of questions. Parents were encouraged to tape record one of these story-telling sessions for possible future analysis. 30 Part 3: Story Retelling One week after the original story elicitation, the experimenter invited the individual children to tell the same story. As in Part I, the children told the story while looking through the picture book with the experimenter. Each child was encouraged to tell the story again, although only two of the 3-year-old subjects opted to do this. The experimenter's participation in the task was restricted to the utterances mentioned above (Part 1). The second story elicitation was also recorded using audio and visual recording equipment. Transcription Procedures Each story was transcribed according to the transcription format and conventions laid out for the CHILDES database. Each line in the transcript consists of a main clause and any dependent clauses. This transcription criterion is similar to the "unified predicate" (p.37) described by Berman et al. (1986) and used in many of the other "Frog Story" narrative studies. Coding Unless otherwise mentioned, the unit of analysis used was the plot unit. The defining characteristics of a plot unit are that it is either part of the narrative discourse or can be inferred from it (e.g. a failed outcome for an attempt can be inferred if another attempt is made to resolve the same goal) and it has an identifiable role within the narrative. Clauses functioning outside the discourse, such as negotiation of task completion, were omitted from analysis. A plot unit could consist of a portion of a clause or several predicates over two or more utterances. Each line of the transcript was coded according to the plot units it contained. To avoid confusion, only the first predicate was coded if more than one made up the plot unit, for example: 31 Portion of an Utterance: "... for the frog." (functions as Purpose) Simple Utterance: "He looked in the tree." (functions as Attempt) Complex Utterance: "He couldn't find the frog anywhere, not even under the log." (functions as one Failed Outcome) Each plot unit was called either a goal plan, nonfrog goal plan or nongoal plan unit, depending on its function within the discourse. Any plot unit that functioned as an introduction, opening event, initiating event, attempt, nonattempt event, purpose, reaction or outcome for one of the frog goals was counted as a goal plan unit. Any plot unit that was a constituent of a nonfrog goal plan was counted as a nonfrog goal plan unit. Nongoal plan units were plot units that were part of the narrative discourse but were not members of a goal plan (e.g. description of part of a picture). The following coding categories are based on those used in a study of goal plan development by Trabasso, Stein, Rodkin, Park-Munger and Baughn (1992) and in Berman and Slobin's (1994) study of plot development. Additional categories were developed to account for certain observations of the story data and to address more specifically the areas of interest outlined in the research questions. Utterances that were incomplete or contained some unintelligible words were only coded for the information explicitly stated or which could be reasonably inferred. Goals The narration of the frog's escape in each story was coded for the explicit statement or reasonable inference of each of the three frog goals: Searching for the frog, Finding the frog and Taking the frog back home. The criteria for determining if a superordinate goal could be inferred were drawn from the work of Trabasso and Nickels (1992) and Berman and Slobin (1994). The purpose of these codes was to determine the presence and frequency of occurrence of each of the three frog 32 goals across the two age groups. According to Trabasso's work, the inclusion of goals in the narrative is strong evidence that the narrative has a global goal plan, as well as local structural coherence. In order for a goal to be inferred, an antecedent (i.e. initiating) event and a consequence (i.e. outcome) of an attempt must be included elsewhere in the text.2 Some of the possible information used to infer each frog goal is listed below that goal's description. If a Find goal was inferred, that entailed a Search goal as well. Likewise, if a Take goal was inferred, both a Find and a Search goal were entailed. Diverging from Trabasso and Nickels (1992), the three primary goals were not coded each time they were reinstated. Nonfrog goals could be coded at any time during the narrative. 1. Take the Frog Back Home This goal was coded when there was either an explicit mention of getting the frog back 3 or sufficient information provided in the text to infer the existence of the goal. The information was considered to be sufficient if three or more of the following items were included: - a stated relationship between the boy and the frog 2 Trabasso and Nickels (1992) initially state that the criteria for inferring a goal are: "either an event that causes a goal and an attempt to achieve it, or by failed outcomes and instances of goal plans to search in particular locales" (p. 261). However, later they indicate that "without knowing that the frog is of value, inferences about the goals of wanting it back and finding it are unwarranted" (p.271). Berman and Slobin's (1994) criterion is also more stringent, in that the goals must be explicitly stated in the narrative. The first of these criteria will be used in the present study as it is more consistent with the criteria used in later studies by Trabasso et al. (1992) and Trabasso and Rodkin (1994). 3 Trabasso et al. (1992) are somewhat unclear as to whether the frog must be the same frog that was lost. Berman and Slobin (1994, p.46) indicate that the frog they take home may either be the same frog or a replacement frog. Since Trabasso developed his study based partly on Berman and Slobin's work, this definition of "the frog" will be the one used in the current study. 33 - noticing that the frog was missing from the jar or any other initiating events that could act as an antecedent event for the boy's realization that the frog was gone - an attempt to take the frog home at the end of the story - the existence of other goal plans in the story (Find, Search goals) Examples: "I'm gonna go get my froggie." (explicit) "The frog was nowhere to be seen." (antecedent event) "He looked and looked for him." (attempt) "But nowhere he could find him." (consequence, inferred goal) 2. Find the Frog The Find goal was coded when it was either explicitly stated in the text or could be inferred from the rest of the text. For example: - a Take goal was explicitly stated, which entails Finding the frog - attempts were made to find the frog - the search was continued after a failure (which also implies a Search goal) - a frog was found at the end of the story Examples: "He looked to see if he could find his frog." (explicit) "His frog was gone!" (antecedent event) "He looked in his boot. No frog." (attempt, failed outcome) "Next he looked outside. No one." (resumed attempt, failed outcome, inferred goal) 3. Search for the Frog A narrative was coded for the Search goal if the search was maintained after at least one failed outcome. A search began when the boy or dog searched, looked, called, or yelled for the frog. 34 Attempts to find the frog were assumed to create local goal plans, which together provided evidence for inferring higher-order goals. A Search goal is entailed by both the Take and Find goals. Examples: "He called outside for the frog." (explicit) "He called and called in the tree." (renewed attempt) 4. Nonfrog Goal If an event occurred that initiated a goal state other than one of the three primary goal plans, that goal state was coded as a nonfrog goal. Because nonfrog goals related to local goal plans, they were coded wherever they occurred in the narrative (see also Secondary Plot Line). Examples: "The dog fell out the window." (initiating event, sets up "rescue" goal) "The boy climbed out and saved him." (attempt, successful outcome) Primary Plot Line Goal Plan Unit Categories The goal plan coding categories were adapted from those outlined in Trabasso et al. (1992). Clauses that were part of a goal plan were assigned to three categories, depending on the portion of the narrative being coded. These categories were Plot Opening Constituents, Plot Unfolding Constituents, and Plot Concluding Constituents, as defined by Berman (1988). Within each of these three categories were additional categories for coding the specific function of the utterance in the goal plan. 35 Plot Opening Constituents At the beginning of good narratives, the narrator establishes the setting for the story and provides the motivation for the inference of a goal. The setting information includes the circumstances under which the story occurs, the characters involved, and the relationships between them. The goals can be inferred by the listener from the initiating events if they are not stated explicitly in the narrative. 1. Character Introductions Narrators were given credit for introducing each of the three main characters - the frog, the boy, and the dog -- if they named them within the first two pictures. 2. Opening Events An opening event was any event that contributed to the listener's understanding of the story's context but did not function to motivate a goal state. These utterances always accompanied pictures 1 and 2, as seen in the following examples. "They sat down on the stool." "He looked in the jar at the frog." 3. Protagonist-Frog Relationship In Frog, Where are You, there is a possession relationship between the frog and the boy. This relationship serves to motivate the formation of the three goal plans for the story: taking the frog home, finding the frog, and searching for the frog. Each story was coded once for this relationship following the narration of the first page. 36 Relationship stated, e.g. "Froggie I love you." or "The boy was looking at his frog." No relationship stated, e.g. "They are looking at a frog." 4. Initiating Events This coding category included the six events listed by Trabasso et al. (1992) plus an additional category for events 'other' than these six. These events are relevant to the establishment of the Search, Find, and Take goals. These events are stated explicitly within the narrative text corresponding to the second and third pictures. Sleep: The boy and/or the dog go to bed or fall asleep, allowing the frog to escape. Frog Leaves: The frog leaves/escapes from his jar and/or the room. Wake-up: The boy and/or dog wake-up after the frog has escaped. Frog Gone: The boy and/or dog realize that the frog has escaped. Frog's Jar is Empty: The boy and/or dog notice that the jar is empty. 4 Reaction: The boy and/or dog's emotional reaction to the frog's disappearance. Other: Some initiating event other than the above is given by the child, e.g. distinguishing between the frog getting out of the jar and escaping from the room. Plot Unfolding Constituents The continued search for the frog constitutes the main body of the story. A s the plot unfolds for the listener, the boy and the dog have the opportunity to make several attempts at different locations to find the frog. There are also several incidents that are not related to their search, such 4 As mentioned by Trabasso & Rodkin (1994, p.87) the Frog Gone event entails the Frog's Jar is Empty event but not the reverse. 37 as knocking over the beehive and getting scared by the owl. Only those events and locations related to the three frog goals were coded as plot unfolding constituents. The nonfrog events were coded separately as constituents of nonfrog goal plans (see Secondary Plot Line). 1. Attempt An attempt to achieve a goal was made. Attempts to achieve the "Search" goal included clauses containing verbs such as look, search, call, and yell, from which a search could reasonably be inferred. An attempt to achieve the Take goal required either a verbal request for the frog to return or explicit mention that the frog was being taken home. Search/Find attempt, e.g. "The boy looked in the hole." Take attempt, e.g. "The boy tried to take him home." 2. Purpose A purpose for the attempt was coded if the narrator included a reason for the action. Often these reasons were encoded in for or infinitival phrases (e.g. "for the frog"). Purposes were coded with the relevant attempt. Examples: "He called for his frog." "He climbed up to see if he could find a frog." 3. Nonattempt Event Any event that was relevant to the frog story but was not an actual attempt to resolve a goal plan was coded as a nonattempt event. Examples: "He went to the window." (nonattempt event) "He called out the window." (attempt event) 38 4. Search Locations These codes refer to the locations where the boy and/or dog could search for the frog. Narrators who encode attempts at various locations can be viewed as having a sustained goal-plan underlying their narrative. The children in the present study narrated a number of Attempts at locations other than those included in the Trabasso et al. study (1992). Also, several subjects distinguished between the boy's and the dog's attempts to find the frog. Therefore, the Room and Ground locations were divided into two categories depending on the character involved, and three new locations were added: 'Landing', in which the boy searches for the frog after first landing in the pond; 'Pond', in which the boy makes an attempt to recover the frog after he has discovered the family of frogs; and 'Other', in which an Attempt is made to find or recover the frog at some idiosyncratic location (e.g. under the bed). Repetitions of the same location were coded separately unless they were elicited repetitions. Room: The boy/dog look for the frog in the bedroom. This coding category was divided into two subcategories, depending on the character involved: A. Boot: The boy looked in the boot. B. Jar: The dog looked in the jar. Window: The boy/dog searched out the window for the frog. Field/Outside: The boy/dog called for the frog in the field. Gopher Hole: The boy searched in the gopher hole. Bee Hive: The dog called in the beehive. Tree: The boy searched in the tree. Rock: The boy climbed the rock to see if he could see the frog or he called from the top of the rock. Landing: The boy or dog searched for the frog after landing in the pond. 39 Log: The boy looked behind the log because he thought he heard frogs. Pond: The boy or dog made an attempt at the pond to take the frog home. Other: The boy or dog searched someplace other than at a locations listed above. Plot Concluding Constituents Utterances that served to conclude either local or global goal plans were included as plot concluding constituents. The main coding categories that served these purposes were Outcomes of attempts and Reactions to events, both of which were adapted from Trabasso's work on goal plans (Trabasso et al., 1992; Trabasso & Nickels, 1992). 1. Outcomes While Trabasso et al. (1992) included successful goal attainment (outcomes) in their analysis, they made no mention of how the narrators communicated failed outcomes. Since the inclusion of this information could be used as evidence for a hierarchical goal plan, the outcome coding was expanded to include both explicit and inferred failures. A failed outcome could occur after any attempt, so that each narrative could be coded for several failures. Only one successful outcome was given per superordinate goal, unless a separate, successful attempt was narrated. Since the successful outcome for the Search and Find goals occurred simultaneously, it was only coded once, as Frog Found. Frog Found: The narrator explicitly stated that the boy/dog found a frog. This statement was coded as a positive outcome or resolution to the 'Find the Frog' and the 'Search for the Frog' goals. The frog did not have to be the same frog that the boy lost earlier in the story. Example: "He looked over the log and there was his frog." 40 Frog Taken Home: The narrator explicitly stated that the boy took the frog with him when he left the pond. No mention of where he was taking it was considered necessary for a successful outcome. Example: "He took the frog and left the pond." Explicit Failure: The narrator explicitly stated that an attempt failed to achieve the goal which motivated it. Example: "The boy looked in the hole." (Attempt) "Nobody in there." (Outcome) Implied Failure: Following an attempt to achieve a goal, a second attempt was made. Although no outcome was given for the first attempt, a failure was implied by the fact that a second attempt was made. Example: "The boy called for his frog." (Attempt #1) "Then he looked in the hole." (Implied Failure #1, Attempt#2) 2. Reactions A narrator's inclusion of character reactions is considered by many researchers to be an indicator of a more mature level of cognitive development because the child is able to tell a story from a character's perspective as well as a personal one (**ref here). Reactions were coded as either cognitive or emotional reactions to an event. They generally functioned as conclusions to local or global goal plans. Examples: "He was so happy to find his frog." (emotional reaction) "He wondered where his frog went." (cognitive reaction) 41 Secondary Plot Line/Nonfrdg Goal Plan Units An episode that was not related to the one of the three frog goals was coded for a nonfrog goal, and its elements were coded as constituents of nonfrog goal plans. These local goal plans (e.g. saving the dog, escaping from angry bees) are differentiated from the three frog goal plans solely on the basis of their scope. They do not form the underlying structure for the entire narrative, but rather single episodes that occur during the search for the frog. These local goal plans were important to include in the analysis because they also reflect the children's knowledge of goal plans, but on a local level. Nonfrog goal plans could also be used as evidence that the narrator was capable of using higher order goal plans in the story if they were complex enough. For a local goal plan to be inferred there had to be an initiating event that established a goal and an attempt to achieve that goal. An explicit outcome was optional. 1. Nonfrog Goal As mentioned previously in the Goals section, a goal is established by an initiating event unrelated to the boy's search for the frog was coded as a nonfrog goal. Examples: "The dog fell out the window." (initiating event, sets-up "rescue" goal) "The boy climbed out and saved him." (attempt, successful outcome) 2. Initiating Event One or more events that created a goal for a character and caused that character to act. Example: "The dog fell out of the window." 42 3. Internal Reaction A character's emotional or cognitive reaction to an attempt, initiating event, or outcome of a nonfrog episode. Example: "He was scared that his puppy was hurt." (emotional reaction) 4. Attempt An action that attempted to achieve the goal of a nonfrog episode. Example: "The boy jumped out of the window." (attempt to retrieve the dog) 5. Purpose A statement about why an attempt was made that clarified the goal of the episode. Example: "The boy jumped out to get his dog." 6. Nonattempt Event An event that was relevant to the nonfrog episode but was not an attempt to achieve the goal. Example: "The dog ran beside the deer." 7. Outcome The implied or explicit result (positive or negative) of an attempt. A failed outcome was implied if another attempt was made. Examples: "The boy saved his dog." (explicit successful outcome) "The boy tried to pick him up." (implicit outcome, renewed attempt) 43 Nongoal Plan Units According to Trabasso et al. (1992), there are several types of clause that are not members of a goal plan. This category includes events and states that have no motivating goal or contextual information. Due to their frequency, rhetorical strategies used by the narrator to maintain or gain listener interest were added to this category. Previous studies have found that 3-year-olds' narratives consist primarily of nongoal plan clauses. Descriptions may be differentiated from narrations of event or states by the use of the present tense and/or progressive aspect, the use of an indefinite pronoun to refer to a previously mentioned character or location, lack of previous goal plan structures, and level final intonation contour. 1. Description of Event An event that is not a member of any goal plan, often a description of a pictured event. Example: "He climbs the tree and then he climbs the rock." (two described events) 2. Description of State A description of a pictured or imagined state which had no antecedent goal plan. Example: "He's standing in a field." 3. Name of Person/Object A clause that served solely to identify a person or object. Example: "There's bees and that's a beehive." 44 4. Narrator Strategy Rhetorical devices used by the narrator to engage or maintain the listener's attention or participation, for example, asking the listener questions about the story or explaining a pictured event outside the narrative mode. Examples: "Who is that?" "What do you think that is?" Analysis Story Selection Children's stories can vary considerably from telling to telling. If the research goal is to determine what children know, it is appropriate to base the analysis on their best efforts. To ensure that the subjects' maximal performance was used for the goal plan analysis, the researcher selected the best story each child had told. Six story components served as the basis for deciding which of the stories was most complete and, therefore, most coherent. These story elements are adapted from those outlined in Berman (1988) and Gomme (1994). Where two stories contained an equal number of story elements, the story that was judged to be the most complete by a naive reader was chosen. The story elements are: Initial Event Chain [Plot Opening]: 1. The frog escapes from the jar. 2. The boy and/or dog discover that he is missing. Search Occurs [Plot Unfolding]: 3. The boy and dog search for the frog. Story Ending [Plot Conclusion]: 4. The boy finds the frog. 5. The boy takes the frog. 6. The new frog is the same or a substitute for the frog who escaped. 45 Age-based Comparisons The following comparisons were done based on the age of the narrators. In some cases, further calculations and analyses were necessary to determine individual and group scores. Where that was true, a description of the analysis or calculation is included below. The number of the relevant research question is provided above each description. Goal Plan Knowledge Trabasso and Rodkin (1994) and Berman and Slobin (1994) found that their 3-year-old subjects did not base their narratives on goal plans of action, and that the 4-year-olds' used only local goal plans. This lack of global and local goal plan knowledge would have a substantial effect on the level of structural complexity of the narrative, in terms of the number of story components and episodes included. It would also limit the local and global coherence of the plot, in terms of encoding the initiating, elaborating, and concluding events of the story and the relationships between these events. Therefore, before doing comprehensive goal plan and plot analyses, it was necessary to determine if the 3- and 4-year-olds in this study had underlying knowledge of goal plans of action. Question 1: Goal Plan Knowledge The measures used assume that the information encoded by narrators reflects their underlying knowledge of goal plans of action. Several different measures have been profiled in the literature, a subset of which were applied to the 3- and 4-year-olds' narratives to determine the presence or absence of underlying goal plan knowledge for both local and global goal plans. 46 Trabasso et al. (1992) used two measures of global goal plan knowledge. The first was the sum of the explicit goal plan units functioning as attempts, purposes and outcomes in the primary plot line. Since these units form the central core of the goal plan, I termed them principal goal plan units. The number of principal goal plan units provided a gross measure of how well the narrator had captured the protagonist's plan of action for resolving the goal of the story. Research by Bamberg (1987) and Bamberg and Marchman (1990) has shown that renewed attempts were "judged by adults to be highly important in moving along the plot" (1987, p. 139). Based on this finding, the second measure of goal plan knowledge chosen was the number of times the Search/Find goal plan was renewed by an attempt to find the frog. The attempts to find the frog were chosen because they form the majority of the possible attempts in the narrative and, therefore, provide a large enough sample so that developmental change may be seen. A narrator who had poor or no knowledge of goal plans would have few, if any, attempts to find the frog. Trabasso and Nickels (1992) used goal-attempt-outcome (GAO) sequences to infer local goal plan knowledge. The explicit mention of a goal (i.e. a purpose or an adjacent explicit fail), an attempt to resolve that goal, and an outcome for that attempt (implied or explicit) forms a local episode. Following the assumption that goal plans are the underlying structure of episodes, then each episode represents a local goal plan. G A O sequences were also used to measure global goal plan knowledge when several sequences involved the same goal. Only G A O sequences related to the three primary goals were included in this measure. Nonfrog G A O sequences were analysed with respect to secondary plot coherence. 47 The final measure of goal plan knowledge is based on the criteria for goal inference presented in the Goals section of the coding scheme. Narratives that satisfy the criteria for the inference of both superordinate and subordinate goals indicate knowledge of hierarchical goal plans of action. These measures indicate the approximate complexity of the narrative — is the goal plan hierarchical or not (measure c)? How many episodes are there (measure b) and how complex are they likely to be (measure a)? They also indicate its coherence - are the episodes well-motivated and locally coherent (measures b and c)? How coherently does the plot unfold (measure a)? For this question and for the remaining analyses, the two issues of coherence and complexity frequently overlap, with changes in one affecting the other. In theory, good goal plan knowledge results in a more detailed 'scaffold' for story components, thereby reducing the individual's processing load and allowing her to encode more of the story content and to encode it in a more unified, meaningful way. Primary Plot Coherence Question 2a: Global Primary Plot Coherence At the most fundamental level, globally coherent narratives need a setting, a change of state, a goal, an attempt to achieve the goal and an outcome. Without any one of these components, the narrative would be difficult to understand. The narratives were scored for each of five criteria, selected by Trabasso and Rodkin (1994), for a globally coherent frog-story narrative. 1. The boy has a relationship to the frog and that possession relation is stated. 2. The boy loses the frog, which initiates goals to search for, find and retrieve the frog, and a plan of action (e.g. the frog leaves, his jar is empty, boy realizes the frog is gone). 48 3. The boy attempts to find the frog (i.e. a search for the frog is begun). 4. The boy continues his search despite failure (i.e. the search is repeated in different locations). 5. The boy eventually finds his frog and takes him home (i.e. Find and Take outcomes are stated). Another measure of global coherence is the proportion of the narrative that is relevant to the primary plot line of the story. Stories that contain more information relevant to the goals of the narrative are more coherent than stories without much relevant information because there are semantic, causal, and temporal ties between goal plan units. Nonfrog goal plan units contribute to the local, but not the global, coherence of the narrative. The total number of goal plan, nonfrog goal plan, and nongoal plan units was calculated and then divided by the total number of plot units to obtain the proportions reported. Question 2b: Plot Opening Coherence According to Berman (1988) and Berman and Slobin (1994), for a goal plan to be initiated, the narrator must introduce the characters and encode the protagonist's relationship to the goal object — in this case the boy's possession of the frog - and the initiating events leading up to the goal state. They refer to these components of the narrative as the "opening of the plot" and predict that the more listeners know about the motivating circumstances of a story, the less difficulty they will have inferring the correct goal plan. I chose to include goal plan units functioning as opening events in the plot opening units measure, because they contribute to the event context of the story and, therefore, promote coherence. The proportion of plot opening units is primarily a measure of local narrative coherence, since it is based on the first portion of the narrative. 49 The proportion of plot opening units was calculated for each narrative as follows: Total of All Plot Opening Units = Proportion of Plot Opening Units to Goal Plan Units Total Number of Goal Plan Units In determining the coherence of the opening of the plot, it is important to look at both the number and proportion of plot opening units. Because the length of the narrative can vary substantially, the proportion alone can be deceiving. A low proportion of plot opening units may not indicate poor coherence if it is due to the overall length of the narrative, rather than a lack of proper setting and goal initiating information. Similarly, a high proportion of units may reflect a short narrative rather than inclusion of all the crucial plot opening information. Within the larger category of plot opening units are two important subcategories - the setting and the initiating episode. The setting orients the listener to the characters involved, their actions prior to the establishment of the goal, and the temporal location of the narrative. The number of plot units functioning as character introductions, opening events, and temporal markers was the second measure of plot opening coherence. The primary plot is initiated in the episode in which the frog escapes. How the narrator encodes this initiating episode dictates to a large extent how well the listener will understand what the theme of the narrative is. A locally coherent episode at this early point in the narrative helps to establish later, global coherence. The final measure of plot opening coherence was the coherence of the initiating episode of the child's narrative. The coherence criteria were modified from those established by Berman and Slobin (1994, p.52) to allow the boy's realization to be inferred. This constituent could only be inferred if a purposeful attempt was made to find the frog or a reaction was explicitly stated, since these entail that the boy knows that the frog is gone. 50 Each initiating episode was scored for the following constituents: 1. The boy wakes up (e.g. "He waked up."). 2. The temporal location of the episode is given (e.g. "Then in the morning ..."). 3. The boy sees/realizes something (e.g. "He saw that..."). 4. The state of affairs is given (e.g. "The frog was gone," or "The jar was empty"). 5. The boy takes action or reacts (e.g. "He looked in his boot," or "He started to cry"). Question 2c: Plot Unfolding Coherence In the Trabasso model, goal plans, and therefore plots, are sustained by protagonists' attempts to reach the underlying goals of finding the frog and bringing him back home. Narratives in which the plot is maintained until the end would be judged as more coherent than narratives with an initial search that is later abandoned. In Frog, Where are you?, there are at least 11 locations where an attempt to find or regain the frog could occur. A story was scored as having a sustained search if a character missed no more than two attempts in a row to achieve a goal, where the sequence is determined by the sequence of locations pictured in the book. A suspension of the search after two missed attempts was chosen as the criterion because it required the narrator to turn the page and miss the two attempts shown on facing pages. The only exception to the two-attempt location criterion was for the Landing and Log locations. The picture of the boy landing in the pond does not actually depict an attempt to find the frog. However, enough narrators encoded an attempt there to warrant separating it from the Other location category. The attempt actually takes place at the log, so an attempt at either location was accepted as a continuation of the search if the previous attempt had been missed. 51 In adult narratives, storytellers create narrative coherence by not only sustaining the search, but by also informing listeners of the outcome of each attempt. They use both implied and explicit outcomes to provide the motivation for the next attempt. Each child's story was scored for the number of attempt-outcome pairs, to determine if these young narrators were able to use this same strategy. The proportion of plot unfolding units is a gross measure of how much of the goal plan is involved in developing the primary plot line. All of the goal plan units that functioned as non-attempt events, attempts, purposes, reactions, and inferred or explicit outcomes were included as plot unfolding units. As discussed with respect to the proportion of plot opening units, both the ratio of plot unfolding units to total plot units and the absolute number of plot unfolding units must be taken into consideration because of differences in narrative length. Question 2d: Search Coherence As mentioned above, the location of each attempt in a narrative was noted. The proportion of narrators encoding an attempt at each of these locations was calculated for each age group. Because attempts coded as Other could occur at ahy time in the narrative, they were included with whichever pictured location occurred on the same page, with the search by the same character. For example, if the boy searched for the frog under the bed, it was counted as an attempt at the boot location because it occurred in the same room and was made by the same character. The number of locations for a particular narrative did not necessarily correspond to the number of attempts, since more than one attempt could occur at the same location. Attempts at the first three locations, which motivated the rest of the narrative, were considered the beginning of the search because they occurred within the same physical space that the frog had escaped from. The next 52 six locations formed the middle of the search. Based on Trabasso et al.'s (1992) finding that most 4-year-olds abandoned the search after the first three locations, this portion of the search was where the most developmental change was expected. Attempts at the final two locations were considered the end of the search, because they were usually accompanied by successful outcomes for either the Find or Take goal plans. Question 2e: Plot Concluding Coherence Narration of the final, successful attempts to resolve the subordinate (Search/Find) and superordinate (Take) goal plans is a key component of telling a coherent narrative. Goal resolution completes the structure of the underlying goal plan of action and forms the end of the plot. Narrators could encode either the Find or Take outcome, or both. Previous research by Trabasso et al. (1992) indicates that narrators progress from encoding no goal resolution to encoding the resolution of the subordinate, then superordinate, goals and, finally, both the subordinate and the superordinate goals. Secondary Plot Coherence Question 3: Secondary Plot Coherence The secondary plot line can be coherent at both a local, episodic level and at a global level. Well-formed local episodes, which contain an explicitly stated goal, an attempt, and an outcome, are more coherent than episodes that have no clear purpose or outcome. The number of nonfrog G A O sequences was calculated for each child's narrative to determine if the older children were better able to encode this information. Since the number of nonfrog episodes varied widely across the narratives, the proportion of episodes that contained a G A O was also calculated for comparison. If the local episodes share constituents with other episodes, then they also contribute to the global 53 coherence of the plot line. For example, the dog's actions with regard to the bees and their beehive and the boy's climbing up on the rock can be developed into quite complex subplots to the main story. Primary Goal Plan Complexity Question 4a: Global Primary Goal Plan Complexity Two rough measures of the complexity of a narrative are: (1) its length, and (2) the number of utterances that are part of the primary goal plan. The overall length of the narrative was calculated according to the number of unified predicates the child produced. The length measure did not include utterances that were unintelligible or were related to topics outside the scope of the narrative task (e.g. negotiating behaviour). The number of goal plan clauses was calculated from the number of unified predicates that functioned in some way as part of the primary goal plan. Because KS's (3-year-old subject 2) narrative was substantially longer than any other 3-year-old's, the measures were also calculated without his narrative to reduce the effect of outliers on the group mean. A third measure of global goal plan complexity is the number of G A O sequences, which was also used as a plot coherence measure. The story elements and their number and arrangement constitute the structure and complexity of the narrative; inclusion of these same elements is necessary for coherence, as mentioned throughout this chapter. Because each sequence indicates a local episode, the number of G A O sequences is equal to the number of episodes in the primary goal plan. 54 Question 4b: Plot Opening Complexity As discussed earlier, the opening of the plot has several functions, including character introduction and initiating a goal plan. Hypothetically, an increase in the complexity of this portion of the narrative could either occur across all the plot opening unit categories or in some categories before others (e.g. introduction of the boy before introduction of the dog). Since some of the categories are more important to the coherence of the plot, the groups' performance for each of the plot opening categories was analysed separately to determine where any changes were occurring. Question 4c: Pattern of Initiating Events Once a relationship has been established between the boy and the frog, the narrator must convey the events that motivate the goal plans of the story. Trabasso et al. (1992) and Berman and Slobin (1994) found that developmental differences could be seen in the types of initiating events children chose to encode in their narratives, depending on (1) event saliency, (2) importance of the event to the establishment of the Search, Find, and Take goal plans, and (3) whether the picture depicted the event explicitly or it had to be inferred from the story context. Each of the initiating events was ranked according to the number of times it was included in a story by narrators from each of the two age groups. Each initiating event was categorized as either depicted in the book (P) or inferred from the context (I). Question 4d: Plot Unfolding Complexity If a story has an underlying hierarchical goal plan, then the narrator should encode attempts to achieve both the superordinate and the subordinate goals. The more attempts that are made to achieve a goal, the more complex the goal plan becomes. By attaching a purpose to an attempt, the narrator makes the underlying goal explicit to the listener, enhancing both the complexity, and 55 the coherence, of the local episode and the global plot line. Another way that a narrative can become more complex is if there is more than one plot line. A second plan of action is needed if there are two protagonists, which was the case for several of the narratives. In order to distinguish if one type of attempt increased before some other type of attempt, the total number of attempts was broken down according to protagonist and goal, either explicit (purposeful) or implicit. Because of the variability within the two groups, the narrators were later grouped into five categories, depending on the number of attempts they encoded. Question 4e: Plot Concluding Complexity Since the plot line of Frog, Where are You? is made up of three primary goal plans, the conclusion of the plot and resolution of the goal plans can vary in complexity depending on the outcomes the narrator encodes. Narratives were coded for whether an attempt was successful or a failure and whether a failure was explicitly stated (e.g."but there was no one there!") or could be inferred from the context (e.g. by a continued search for the frog). Only explicit statements of goal success were coded as Find or Take Outcomes (e.g. "He finded his frog," or "He took one home to his place"). The number of outcomes in each of these categories was calculated and the group means compared. Secondary Goal Plan Complexity Question 5a: Secondary Goal Plan Complexity A rough measure of the complexity of the nonfrog goal plans was taken by calculating the number of nonfrog goal plan units and nonfrog episodes, and the number of nonfrog goal plan units per episode. Their complexity is an indicator of the child's ability to narrate complete local goal plans. The group means for these measures was then calculated. However, the children in both age 56 groups had a wide range of scores for the number of nonfrog goal plans. As such, the group mean hid any of the patterns that could be seen in the distribution of scores. In order to resolve this problem, the children were divided up according to how many nonfrog episodes they included, and the distributions were compared. Question 5b: Pattern of Secondary Goal Plans As was mentioned with respect to attempts, some of the story's events are pictured, while others must be inferred from the preceding and following pages. In addition, some of the nonfrog plans may be more salient to a particular child or group of children. Therefore, some effect of event saliency, narrative importance, and picture abstraction might also be seen for the nonfrog goal plans. I would expect that older narrators would be less affected by saliency and picture abstraction because they can rely more heavily on their knowledge of goal plans and how to tell a good story than the 3- and 4-year-olds. To determine if there was a distinguishable pattern in the narration of nonfrog episodes, each narrative was coded for the particular nonfrog goal plans the child had chosen to include. The proportion of narrators from the two groups including each plan was then calculated, and the plans were ranked in the order of most to least frequent. Nongoal Plan Units Question 6: Nongoal Plan Units In young children's narratives, nongoal plan units have a negative impact on the complexity of narratives and do not serve to further the plot line. However, each type of nongoal plan unit has a slightly different effect. Descriptions have perhaps the clearest effect on goal plan complexity. They indicate that the narrator is having difficulty accessing the underlying goal plan and using it to incorporate the information into the narrative. 57 Berman and Slobin (1994) and Trabasso et al. (1992) found that description and naming were the most common kinds of nongoal plan utterances, particularly for the youngest subjects (3 to 5 years old). This finding may be due to the effect of adult storytelling style with this age group, in which children are frequently prompted to describe pictures and name objects. The effect of narrator strategies on narrative complexity is less clear. While they serve to maintain listener interest and may therefore aid story comprehension, they do not contribute to the complexity of the narrative and can reduce the coherence of the narrative if they are overused. Because of their different effects on the narrative, each type of nongoal plan unit was compared separately. The mean number of each type of nongoal plan unit was calculated as was the proportion it made up of all the plot units. Content-based Comparisons Because of the variability within the two subject groups and the overlap in their score distributions for many of the measures, the group differences were small or absent. In an attempt to capture the qualitative differences seen in the data, all of the children were placed into a single group for further analysis. A subgroup of the most important measures of complexity and coherence used previously was chosen. A group mean was taken for each measure and z-scores were calculated to describe individual children's performances. Of particular interest was the performance of the longitudinal subjects on these measures. Coherence Analysis The coherence measures selected were: 1. The proportion of goal plan units to number of plot units; 58 2. Initiating episode score; 3. The proportion of plot unfolding units to number of plot units. Narratives were ranked as high (;>+1SD), medium (+/-1SD), or low (<-1SD), based on a combined z-score for measures 1 and 2 and compared to the z-score for measure 3. Complexity Analysis The complexity measures selected were: 1. The number of initiating events; 2. The number of attempts; 3. The number of nonfrog episodes; 4. The number of nongoal plan units. Narratives were ranked as high (;>+1SD),medium (+/-1SD), or low (<-1SD), based on the combined z-score for measures 1 and 2. This ranking was compared to the z-scores obtained for measures 3 and 4. The complexity ranking was then compared to the coherence ranking for the same story. Narratives were divided into the four logically possible groups: coherent and complex, coherent but not complex, complex but not coherent, and neither coherent nor complex. 59 C H A P T E R T H R E E R E S U L T S Overview In this chapter I will present the results of the goal plan analyses of the 3- and 4-year-old children's narratives. The order of presentation follows that of the research questions outlined in Chapter One. The first results relate to age-based comparisons of the children's narratives. The second section presents content-based comparisons of all the children, regardless of age. Finally, where they are particularly illustrative, individual and longitudinal results are provided. The unit of analysis was the plot unit, as outlined in the Analysis section of Chapter Two, except where otherwise indicated. Age-based Comparisons Goal Plan Knowledge Question 1: Do preschool aged children have knowledge of underlying goal plans of action, as measured by: (a) the presence of goal-relevant attempts, purposes and outcomes, (b) encoding multiple attempts at a goal, (c) goal-attempt-outcome sequences, and (d) encoding sufficient information for the listener to infer superordinate (Take) and subordinate (Search and Find) goals? The individual results for the goal plan knowledge measures are shown in Tables 3.1a and b. These results show that all of the children demonstrated some knowledge of underlying goal plans of action, based on the information they chose to encode in their narratives. The group means and ranges were similar for all four measures. However, there was substantial individual variation in performance for the first three measures. At a minimum, the 3-year-olds' narratives had 5 principal 60 goal plan units, 1 renewed attempt to find the frog, and 3 local goal plans. These local episodes were always governed by at least 2 of the 3 global goal plans - to search for and find the frog (table 3.2). The minimum score for these measures amongst the 4-year-olds was slightly better, with 8 principal goal plan units, 2 renewed attempts to find the frog, and 3 local goal plans. Only one narrative lacked a Take goal plan (table 3.2). Most of the 3-year-olds demonstrated better knowledge of goal plans than the minimum performance. There were two whose narratives had 8 and 11 principal goal plan units and 4 and 5 G A O (goal-attempt-outcome) sequences, respectively. However, their scores for the number of renewed attempts remained quite low, scoring at the 4-year-olds' minimum level of 2. The other five demonstrated relatively complex knowledge of narrative structure, with 15 to 20 principal goal plan units, 5 to 10 renewed attempts to find the frog, and 6 to 9 G A O sequences. None of the 4-year-olds scored 1 S D below the group mean on more than one measure of goal plan knowledge. The poorest of the 4-year-olds' narratives fell into the mid-range of the 3-year-olds' narratives. Their narratives were less variable, with eight children encoding 8 to 13 principal goal plan units, 2 to 7 renewed attempts to find the frog, and between 3 and 6 G A O sequences. As seen for the 3-year-olds, a number of 4-year-olds demonstrated quite complex knowledge of underlying goal plans of action. Three of them had 18 to 24 principal goal plan units, 7 to 10 renewed attempts, and 7 to 9 G A O sequences. 61 Table 3.1a: 3-year-old subjects' performance across goal plan knowledge measures Child # Principal Goal Plan Units # Renewed Find Attempts # GAO Sequences # of Inferrable Goals 1 8 2 4 3 2 19 10 9 3 3 18 8 8 3 4 6 2 3 2 5 6 1 3 2 6 20 7 9 3 7 5 2 3 2 8 11 2 5 3 9 17 5 6 3 10 15 9 8 3 Group Mean Range: 12.5 5-20 4.8 1-10 5.8 3-9 2.7 2-3 Table 3.1b: 4-year-old subjects' performance across goal plan knowledge measures Child # Principal Goal Plan Units # Renewed Find Attempts # GAO Sequences # of Inferrable Goals 1 10 3 5 3 2 20 7 7 3 3 9 4 5 3 4 12 2 4 3 5 18 10 7 3 6 24 9 9 3 7 10 6 5 3 8 9 4 3 2 9 11 5 4 3 10 8 2 5 3 11 13 7 6 3 Group Mean Range: 13.1 8-24 5.4 2-10 5.5 3-9 2.9 2-3 62 Table 3.2: Number of narrators encoding take, find and search goals. Goal 3-year-olds (n=10) 4-year-olds (n=11) Take Frog Home 7 10 Find Frog 10 11 Search for Frog 10 11 The Development of Narrative Coherence Several measures were used to determine if the children's narratives became more coherent between ages three and four. These measures focussed on the development of global narrative coherence, and the development of local coherence for the plot opening, plot unfolding and plot concluding portions of the primary and secondary plot lines. Primary Plot Line Question 2a: How does the pattern of global narrative coherence change from ages three to four, as measured by a) the number of essential story elements, and b) the distribution of global goal plan, nonfrog goal plan, and nongoal plan units? All of the children scored well on the first measure of global coherence, inclusion of essential story elements. As seen in Tables 3.3a and b, all of the narrators initiated a goal plan of action, made an attempt to find the frog and encoded a successful outcome for the goal plan. A successful outcome for the Take goal was included by eight 3-year-olds and ten 4-year-olds. All but one narrator, a 3-year-old, renewed the search for the frog following a failure. Only three children from each group mentioned the relationship between the boy and the frog in their narratives. 63 Table 3.3a: Individual scores for 3-year-old subjects on essential story elements. Child Relationship Initiating Event Find Attempt Renewed Attempt Outcome Find/Take Total Score 1 - + + + ++ 4.0 2 - + + + ++ 4.0 3 - + + + ++ 4.0 4 + + + + + 4.5 5 + + + - ++ 4.0 6 - + + + ++ 4.0 7 + + + + + 4.5 8 - + + + ++ 4.0 9 - + + + ++ 4.0 10 - + + + ++ 4.0 Total 3 10 10 9 10/8 x = 4.1 Table 3.3b: Individual scores for 4-year-old subjects on essential story elements. Child Relationship Initiating Event Find Attempt Renewed Attempt Outcome Find/Take Total Score 1 - + + + ++ 4.0 2 - + + + ++ 4.0 3 + + + + ++ 5.0 4 - + + + ++ 4.0 5 - + + + ++ 4.0 6 - + + + ++ 4.0 7 - + + + ++ 4.0 8 - + + + + 3.5 9 + + + + ++ 5.0 10 + + + + ++ 5.0 11 - + + + ++ 4.0 Total 3 11- 11 11 11/10 x=4.2 64 Tables 3.4a and 3.4b show the individual and group scores for the second set of measures, those concerning the distribution of global goal plan, nonfrog goal plan, and nongoal plan units. The two groups had similar means and ranges for both the goal plan and the nonfrog goal plan units. However, the 3-year-olds encoded slightly more nongoal plan units than the 4-year-old children. On average, narrators from both groups encoded more global goal plan units (GPU) than nonfrog goal plan units (NFU). Nongoal plan units (NGPU) were encoded the least often. The range of scores for the three measures shows that individual narrators' performances were quite variable in terms of the absolute number/proportion of units in each category. For example, some narrators had fewer than 25% of their narrative clauses associated with the primary goal plans, while others encoded very low numbers of nonfrog or nongoal plan units. Despite this variability, six of the 3-year-olds and seven of the 4-year-olds fit the pattern predicted by the group means - high proportions of global goal plan units, intermediate numbers of nonfrog goal plan units and fewer nongoal plan units (GPU>NFU>NGPU). Of the 3-year-old narrators who differed from this pattern, one child had more nonfrog goal plan units than global goal plan units (NFU>GPU>NGPU), two children had higher proportions of nongoal plan units, with an intermediate proportion of global goal plan units (NGPU>GPU>NFU), and one child had a low number of global goal plan units relative to both the nonfrog goal plan and nongoal plan units (NGPU>NFU>GPU). Of the 4-year-old narratives that differed from the predicted pattern, three fit the second pattern (NFU>GPU>NGPU), and one child's narrative fit the last pattern (NGPU>NFU>GPU). 65 Table 3.4a: Proportion of goal plan, nonfrog goal plan and nongoal plan units for 3-year-old subjects. Subject Goal Plan Units Nonfrog Goal Plan Units Nongoal Plan Units x units % units x units %units x units % units 1 23 21 39 39 41 40 2 39 29 61 46 33 25 3 32 48 25 38 9 14 4 18 48 13 34 7 18 5 12 37 5 16 15 47 6 30 60 17 34 3 6 7 18 67 7 26 2 7 8 18 41 8 19 17 40 9 28 55 16 31 7 14 10 35 60 12 20 12 20 Group Mean Range: 25 12-39 47 21-67 20 5-61 30 15 16-46 2-46 23 6-47 Table 3.4b: Proportion of goal plan, nonfrog goal plan and nongoal plan units for 4-year-old subjects. Subject Goal Plan Units Nonfrog Goal Plan Units Nongoal Plan Units x units % units x units %units x units % units 1 25 54 16 35 5 11 2 43 52 34 41 6 7 3 22 37 29 48 9 15 4 18 51 10 29 7 20 5 37 44 46 55 1 1 6 32 54 14 23 14 23 7 23 57 13 33 4 10 8 21 26 23 30 34 44 9 32 59 19 35 3 6 10 22 70 7 23 2 7 11 31 45 30 43 8 12 Group Mean: Range: 28 18-43 50 27-71 22 36 9 7-46 23-55 1-34 15 1-44 66 Plot Opening Coherence Question 2b: Did the 4-year-old subjects encode more information relevant to the opening of the plot than the 3-year-old subjects? The individual narrators' scores and the group means for the three measures of plot opening coherence are presented in Tables 3.5a and b. The group means and ranges were similar across all three measures. Within the setting of the narrative, a higher number of 4-year-olds introduced all three characters (five compared to three), and they encoded an average of one more opening event. The temporal location of the narrative was given by eight of the 4-year-olds and five of the 3-year-olds. The scores for the initiating episodes increased an average of one event between the ages of three and four. The distribution of scores was also different, since four of the 3-year-olds encoded fewer than 3 of the events in the episode, compared to none of the 4-year-olds. The majority of 4-year-olds (7 children) encoded 4 out of 5 events in the episode. The boy waking up was encoded by only one 3-year-old compared to eight 4-year-olds. The temporal location of the episode was encoded by two 3-year-olds and four 4-year-olds. The boy's realization that the frog was gone was explicitly stated by three 3-year-olds and six of the 4-year-olds; it could be inferred for three of the 3-year-olds and five of the 4-year-olds. The fact that the frog was gone or that the jar was empty was explicitly stated by all of the narrators. All of the 4-year-olds and eight 3-year-olds encoded some reaction to the disappearance. An example of the typical initiating episode of a 3-year-old is taken from subject 9 (RS): 67 And then he goes to sleep with the dog, And he [: the frog]5 gets out. Frog where are you? The typical initiating episode for a 4-year-old required less inference by the listener and contained more marking of temporal and semantic relations. The following is an excerpt from subject 5 (RP): And they both are asleep, Then the frog tiptoes out! And then [:!], when they awake, they see that the frog is gone. He looks under [: the bed]. Scores for the number of plot opening units did not correlate well with the proportion of plot opening units, indicating considerable variation in the length of the narratives for both groups of children. Most narrators dedicated approximately one third of their narrative to the opening of the plot (six 3-year-olds and eight 4-year-olds). Only four children encoded substantially less than one third (three 3-year-olds and one 4-year-old). However, all four had at least 5 plot opening units, including the introduction of a character, and a score of 3 or higher on the initiating episode. 5 Square brackets are used to indicate my comments. 68 Table 3.5a: Plot opening coherence measures for 3-year-old narrators. Child Setting Initiating Episode # Plot Opening Units (%) 1 5 2 10(44) 2 5 4 10(26) 3 5 4 6(19) 4 6 2 9 (50) 5 1 3 4(33) 6 1 3 5(17) 7 3 2 5(28) 8 2 1 5(28) 9 3 3 6(21) 10 7 3 12 (34) Group x 3.8 3.0 7(29) Range: 1-7 1-4 4-12 (17-50%) Table 3.5b: Plot opening coherence measures for 4-year-old narrators. Child Setting Initiating Episode # Plot Opening Units (%) 1 4 4 8 (32) 2 8 4 13(30) 3 2 5 7(32) 4 1 3 5(28) 5 5 4 11 (30) 6 3 4 5(16) 7 3 5 7(30) 8 4 3 9(43) 9 8 4 14 (44) 10 5 4 11 (50) 11 5 5 10(32) Group x 4.4 4 9(33) Range: 1-8 3-5 5-14 (16-50%) 69 Plot Unfolding Coherence Question 2c: Did the 4-year-olds' narratives unfold in a more coherent manner than the 3-year-olds', as demonstrated by: a) more frequently sustaining the search for the frog until the end of the story, b) including more attempt-outcome pairs, and c) dedicating a larger proportion of the story to achieving the three goals? As shown in Tables 3.6a and b, the search was sustained in eight of the 4-year-olds' narratives and five of the 3-year-olds' narratives^ The group means and ranges for the number of attempt-outcome (AO) pairs and ratio of plot unfolding units to total plot units were very similar. However, the distribution of scores were slightly different between the two groups on both measures. For the A O pairs, four of the 3-year-olds encoded fewer than 5, compared to two 4-year-olds. Of those who did encode 5 or more pairs, three had between 5 and 7, with the remainder encoding either 9 or 10 pairs. Four of the remaining 4-year-olds had between 5 and 7 pairs, with the other five children encoding 8 or 9. Similar groupings can be seen for the overall number of plot unfolding units, which did not correlate well with the ratio of plot unfolding units. The 3-year-olds performed evenly over the range, with four children using less than 25% of all the plot units in the narrative as plot unfolding units, three using between 30 and 40%, and the remaining three encoding more than 40%. The 4-year-olds were less variable, with two children encoding 25% or less, seven children between 30 and 40%, and two encoding 40% or more of their narrative as plot unfolding units. 70 Table 3.6a: Plot unfolding coherence measures for 3-year-old narrators. Subject Sustained Search (Yes or No) # Attempt-Outcome Pairs Plot Unfolding Units/ Total Plot Units (%) 1 No 3 13/103 (13) 2 Yes 9 29/133 (22) 3 Yes 10 26/66 (39) 4 No 4 9/38 (24) 5 No 2 8/32 (25) 6 Yes 5 25/50 (50) 7 No 4 13/27 (48) 8 No 5 13/43 (30) 9 Yes 7 22/51 (43) 10 Yes 10 23/59 (39) Group x Range: ; 6.3 2-10 18/60 (33) 8-29(13-50) Table 3.6b: Plot unfolding coherence measures for 4-year-old narrators. Subject Sustained Search # Attempt-Outcome Pairs Plot Unfolding Units/ Total Plot Units (#) 1 Yes 5 17/46 (37) 2 Yes 9 30/83 (36) 3 Yes 7 15/60 (25) 4 No 4 13/35 (37) 5 Yes 8 26/84 (31) 6 Yes 9 27/60 (45) 7 Yes 8 16/40 (40) 8 No 4 12/78 (15) 9 Yes 9 18/54 (33) 10 No 5 11/31 (36) 11 Yes 7 21/69 (30) Group x Range: _ 6.8 4-9 19/58 (33) 11-30 (15-45) 71 Question 2d: Do the two groups show different patterns of goal plan continuation over the length of the story, as indicated by the locations of the attempts? Figure 3.1 shows the general pattern for the two groups for the 11 pictured locations. An initial search of the bedroom 6 (boot, jar) by either the boy, the dog, or both, was included by 80% of the 3-year-olds and 100% of the 4-year-olds. The 4-year-olds included most of their attempts at the beginning of the search (27 out of 81). The 3-year-olds focussed more on the middle portion of the boy's search for the frog, particularly the gopher hole and tree locations (30 out of 67). All of the children had at least one search-concluding attempt at either the log or the pond. The majority also had an attempt to take the frog home from the pond (80% of the 3-year-olds and 82% of the 4-year-olds). Boot Jar Window Field Hole Hive Tree Rock Cliff Log Pond Location Figure 3.1: Proportion of narrators encoding attempts at each location 6 Any attempt in that room was counted as a search of the bedroom. 72 Plot Concluding Coherence Measures Question 2e: Did more 4-year-olds than 3-year-olds encode a successful outcome for the superordinate goal plan? Table 3.7: Number of narrators encoding find or both find and take outcomes by age group. Plot Concluding Measures 3-year-olds (n=10) 4-year-olds (n=11) Find Outcome 10 11 Find and Take Outcomes 8 10 All of the narrators encoded the final outcome for the subordinate goal of finding the frog (see table 3.7). The superordinate goal plan of taking the frog home was concluded by 18 of the 21 narrators (ten 4-year-olds and eight 3-year-olds). None of the children encoded an outcome for the superordinate Take goal plan without an outcome for the subordinate Find goal plan as well. Secondary Plot Coherence Question 3: Did the 4-year-olds have more coherent secondary plot lines than the 3-year-olds, as measured by a) the number of nonfrog Goal-Attempt-Outcome sequences, b) the proportion of GAO 's to total nonfrog episodes, and c) the number of shared elements between nonfrog episodes? As shown in Table 3.8, all of the narrators had at least one coherent local episode. A small number of narrators in each group (three 3-year-olds and four 4-year-olds) had 4 or more of these episodes. The majority of nonfrog episodes in a narrative were based on G A O sequences for all but one 3-year-old child. One 4-year-old child encoded 2 complex episodes that contained several 73 G A O sequences, resulting in a percentage higher than 100. Aside from these two exceptions, the two groups had similar distributions of scores for this measure. On average, the 4-year-old narrators had a higher number of interepisodic ties than the 3-year-olds. The most common number of ties was 2 for the 3-year-olds and 4 for the 4-year-olds. However, there were still three 4-year-olds who encoded no ties between either other nonfrog episodes or an episode from the main plot line. Table 3.8: Individual and group results for secondary plot line coherence. 3-year- # G A O % G A O Ties 4-year- # G A O % G A O Ties olds Sequences NFE olds Sequences N F E 1 5 83 2 1 2 50 — 2 8 89 8 2 4 80 4 3 3 75 2 3 4 80 5 4 1 33 1 4 1 50 1 5 1 100 — 5 10 167* 10 6 2 67 2 6 3 100 4 7 1 50 ~ 7 3 100 — 8 2 100 • ~ 8 3 60 4 9 3 100 2 9 3 75 4 10 4 100 3 10 1 50 — - - - - 11 5 83 5 Mean: 3.0 81 2.2 Mean: 3.5 87 3.4 Range: 1-8 33-100 0-8 Range: 1-10 50-167* 0-10 * Subject 5 had a percentage higher than 100 because she had multiple G A O sequences within a single episode. 74 The Development of Narrative Complexity The measures relevant to the primary goal plans of finding the frog and taking him home are presented first. In the next section, the results for the nonfrog goal plans are presented. Finally, the results of the nongoal plan unit measures are given. Primary Goal Plan Complexity The analysis of the complexity of the children's goal plans was broken down according to the portion of the narrative being told. First, the results of the global measures of narrative complexity are presented. The remaining three sections present the results of the plot opening, plot unfolding, and plot concluding complexity measures, respectively. Global Measures of Narrative Complexity Question 4a: Did 4-year-olds produce either a) longer narratives, b) narratives containing more goal plan units, or c) more G A O sequences than younger subjects? As seen in Tables 3.9a and b, the average length of the children's narratives is remarkably similar, once subject 2 is excluded from the comparison. However, within both groups there was a wide range of narrative and goal plan complexity as measured by the number of unified predicates and the number of goal plan clauses, respectively. A high score on one of these two measures did not entail a high score on the other. This finding indicates that a long narrative does not necessarily mean a complex underlying primary goal plan. Some of the discrepancy may be accounted for by the complexity of the secondary plot line. On average, the 4-year-olds encoded more utterances that were related in some way to the primary plot line of the story than the 3-year-olds. A group of seven narrators encoded fewer than 5 G A O sequences, four 3-year-olds and three 4-year-olds. 75 Interestingly, the group of seven children who encoded more than 6 G A O sequences was made up of four 3-year-olds and three 4-year-olds. While the 3-year-olds encoded fewer local episodes overall, clearly some of them told narratives that were as complex as the best of the 4-year-olds' narratives. In addition, the global complexity of some of the 4-year-olds' narratives was similar to that seen for the lowest 3-year-olds. Table 3.9a: Global measures of narrative complexity for 3-year-old narrators. Subject # of Unified Predicates # of Goal Plan Clauses # of G A O Sequences 1 84 20 4 2 118 29 9 3 44 21 8 4 24 11 3 5 25 9 3 6 36 21 9 7 19 12 3 8 34 12 5 9 38 19 6 10 40 25 8 Group Mean: 46.2 18.0 5.8 Without subject 2 38.2 16.6 4.8 Range: 19-118 9-29 3-9 76 Table 3.9b: Global measures of narrative complexity for 4-year-old narrators. Subject # of Unified Predicates # of Goal Plan Clauses # of G A O Sequences 1 30 18 5 2 52 32 7 3 34 14 5 4 23 12 4 5 55 27 7 6 42 23 9 7 28 19 5 8 65 22 3 9 33 18 4 10 15 10 5 11 45 23 6 Group Mean: 38.4 19.8 5.5 Range: 15-65 10-32 3-9 Complexity of the Plot Opening Question 4b: Was the opening of the plot more complex in the 4-year-olds' stories than in the 3-year-olds' stories, as measured by a) the characters they introduce, b) the number of opening events they include, or c) the number of initiating events motivating the goal plan? The children's individual scores and the group scores for the plot opening coherence measures are shown in Tables 3.10a and b. As expected, the frog was the most frequently introduced character (nine 3-year-olds and ten 4-year-olds). The two groups differed substantially with respect to the introduction of the boy protagonist (five of the 3-year-olds compared to nine of the 4-year-olds). 77 However, this increase did not correspond to an increase in the number of 4-year-olds encoding a goal relationship. The dog was introduced by five of the children in each group. Narrators in both groups included the goal relationship about one third of the time. All three characters were introduced by three of the 3-year-olds and five of the 4-year-olds. Eight subjects from each group had at least 1 opening event with a mean of 1.4 opening events for the 3-year-olds and 1.6 for the 4-year-olds. The 4-year-olds included approximately 5 of the 6 main initiating events, almost 2 more events than the 3-year-olds. Figure 3.2 shows the distribution of scores for the two groups. 6 n • 3-year-olds — » — 4-year-olds Figure 3.2: Distribution of narrators by number of Initiating Events 78 Table 3.10a: Plot opening complexity of 3-year-olds' narratives. Subject Boy Introduced Frog Introduced Dog Introduced Relationship Stated # Opening Events # Initiating Events 1 Y Y Y N 2 5 2 Y Y Y N 2 5 3 Y Y N N 1 3 4 Y N Y Y 3 3 5 N Y N Y - 2 6 N Y N N - 4 7 N Y N Y 1 2 8 N Y N N 1 3 9 N Y Y N 1 3 10 Y Y Y N 3 4 Group x Range: 5 Yes 5 No 9 Yes 1 No 5 Yes 5 No 3 Yes 7 No 1.4 0-3 3.4 2-5 Table 3.10b: Plot opening complexity of 4-year-olds' narratives. Subject Boy Introduced Frog Introduced Dog Introduced Relationship Stated # Opening Events # Initiating Events 1 Y Y N N 1 5 2 Y Y N N 5 6 3 N Y N Y - 5 4 Y N N N - 4 5 Y Y Y N 2 6 6 N Y N N 1 3 7 Y Y N N - 5 8 Y Y Y N 1 5 9 Y Y Y Y 4 6 10 Y . Y Y Y 1 6 11 Y Y Y N 1 6 Group x Range: 9 Yes 2 No 10 Yes 1 No 5 Yes 6 No 3 Yes 8 No 1.6 0-5 5.2 3-6 79 Question 4c: Did the 3-year-olds and 4-year-olds differ in the particular initiating events they choose to include? The patterns of initiating events encoded by the children in each age group are shown in Table 3.11. Those events that are most crucial to the accurate inference of the Search, Find, and Take goal plans were the most frequently encoded by the narrators in both groups. Slightly fewer than half of the narrators from both groups encoded an initiating event other than the original 6 outlined by Trabasso et al. (1992). As expected, the boy's reaction to the escape was the least narrated initiating event, at 10% of the 3-year-olds and 27% of the 4-year-olds. However, the ranking of initiating events is slightly different in the two groups. All of the initiating events were encoded by a higher proportion of the 4-year-old subjects than the 3-year-old subjects. Both the 3- and 4-year-olds were slightly more consistent in including initiating events that were depicted in the story's illustrations, 59% and 58%, respectively. 80 Table 3.11 Rank order of initiating events for 3- and 4-year-old subjects. 3-year-olds 4-year-olds P/ l** Initiating Event # of Narrators P/l** Initiating # of Narrators (%) Event P Frog Leaves 8 P Frog Leaves 11 P Frog is Gone 8 P Frog is Gone 11 P Jar is Empty 5 P Jar is Empty 11 P/l Other Event* 5 I Boy Goes to 8 Sleep I Boy Goes to Sleep 2 I Boy Wakes 7 up I Boy Wakes up 2 P/l Other Event* 5 I Reaction to Escape 1 I Reaction to 3 Escape *As mentioned in Chapter 2, the Other category was added to the original coding schema in response to the large number of unique initiating events included by the subjects. " P = depicted; I = must be inferred Measures of Plot Unfolding Complexity Question 4d: Does the pattern of narrated attempts change between three and four years of age, as measured by: a) the number of attempts, b) the number of attempts made to achieve superordinate and subordinate goals, c) the number of purposeful attempts, and d) the number of attempts made by the boy versus those made by the dog? As shown in Table 3.12, the group means and range of scores for the two groups were very similar across all the measures. The majority of attempts were carried out by the boy alone, in order to resolve the Find goal. Roughly half of the attempts had an explicit purpose encoded for them. All but one child, a 4-year-old, encoded a purposeful attempt in their story. The majority of the children 81 included at least one attempt by the dog (50% of 3-year-olds and 64% of 4-year-olds) and many stories had a shared attempt between the two protagonists (40% of 3-year-olds and 73% of 4-year-olds). The dog's search for the frog was often abandoned after the characters searched the bedroom. There were 11 attempts to find the frog in the bedroom, compared to 6 attempts at other locations. The most common shared attempt occurred at the pond, with 10 of 14 attempts. The distribution of scores for the number of attempts encoded by each narrator was somewhat different between the two groups, as shown in Table 3.13. Of the ten 3-year-olds, five encoded fewer than half of the attempts and three encoded fewer than a third of the possible 12 attempts. Of the eleven 4-year-olds, four encoded fewer than 6 attempts but none encoded fewer than 4 out of 12. No narrative had fewer than 3 attempts. Table 3.12: Summary of mean scores for plot unfolding complexity measures. Complexity Measure 3-year-olds (n=10) 4-year-olds (n=11) Group Mean Group Mean Number of Attempts 6.6 (range 3-12) 7.4 (range 4-12) Search/Find Attempts 5.8 (range 2-10) 6.4 (range 3-11) Take Attempts 0.8 (range 0-2) 1.0 (range 0-2) Purposeful Attempts 3.2 (range 1-6) 2.8 (range (0-7) Attempts by Boy 5.6 (range 2-12) 5.5 (range 3-10) Attempts by Dog 0.7 (range 0-2) 0.9 (range 0-3) Shared Attempts 0.4 (range 0-1) 0.9 (range 0-3) 82 Table 3.13: Three and 4-year-old narrators grouped according to number of encoded attempts. Number of Attempts 3-year-olds (n=10) 4-year-olds (n=11) n 10 -12 3 2 n 8 - 9 1 3 n 6 - 7 1 2 n 4 - 5 2 4 n = 3 3 — Measures of Plot Concluding Complexity Question 4e: Is there a discernable difference in how the 3-year-old and 4-year-old groups encoded outcomes for either local or global goal plans? The individual and group scores for the plot concluding complexity measures are shown in Tables 3.14a and b. The two groups' means and ranges were similar for all the measures. Narrators in both groups encoded approximately the same total number of outcomes for global and local goal plans. For most local goal plan attempts, the children relied on the listener to infer the protagonist's failure from the narrative context. However, the 4-year-olds appeared to rely slightly more than the 3-year-olds on these implied fails. Of the eleven 4-year-olds' narratives, six contained 5 or more implied fails. Only three of the 3-year-olds' narratives contained that many implied fails. Very few narrators in either group encoded explicit fails (five 3-year-olds and three 4-year-olds). All of the narrators encoded a successful Find outcome, with three of the 4-year-olds separating this outcome into two parts: (1) discovering the group of frogs, and then (2) recognizing the original frog. As expected, the successful Take outcome was encoded less often than the Find outcome. 83 Table 3.14a: Individual and group plot concluding scores for 3-year-old narrators. Subject Total Outcomes Implied Fail Explicit Fail Successful 'Find' Successful 'Take' 1 3 1 - Y Y 2 9 6 1 Y Y 3 10 6 2 Y Y 4 4 3 - Y N 5 2 - - Y Y 6 9 3 4 Y Y 7 4 3 - Y N 8 5 2 1 Y Y 9 7 3 2 Y Y 10 10 8 - Y Y Group x Range: 6.3 2-10 3.5 0-8 1 0-4 11 Yes 0 No 8 Yes 2 No Table 3.14b: Individual and group plot concluding scores for 4-year-old narrators. Subject Total Outcomes Implied Fail Explicit Fail Successful 'Find' Successful 'Take' 1 5 3 - Y Y 2 9 5 2 Y Y 3 7 5 - Y Y 4 3 1 - Y Y 5 8 6 - Y Y 6 8 2 4 Y Y 7 8 6 - Y Y 8 3 2 -- Y N 9 8 6 - Y Y 10 5 2 1 Y Y 11 7 5 - Y Y Group x Range: 6.9 4-9 3.9 1-6 0.6 0-4 11 Yes 0 No 10 Yes 1 No 84 Nonfrog Goal Plans Question 5a: Is there a difference between the two groups in the complexity of their nonfrog goal plans, as measured by a) the number of nonfrog goal plan units, b) the number of nonfrog goal plans, and c) the mean number of units per goal plan? The individual and group results for these measures are shown in Tables 3.16a and b, below. The 4-year-old children produced slightly more nonfrog goal plan units than the 3-year-olds. This difference can be accounted for by the slight increase in the number of nonfrog goal plans. Narrators in both groups showed substantial variability in the number of nonfrog units and nonfrog goal plans they encoded in their narratives. As is inherent in the definition of a goal plan, no plan had fewer than three member units. A fairly narrow range of scores was seen for the number of units per plan for both groups, with only three 3-year-olds and two 4-year-olds averaging more than 6 units per plan. All but one narrator, a 3-year-old, had a secondary plot line with multiple episodes. In order to illustrate the pattern of secondary plot line complexity within each group, the narrators were grouped according to the number of nonfrog goal plans they encoded. The results are presented in Table 3.15, below. While a 3-year-old had the highest number of nonfrog goal plans overall, the majority of the 3-year-olds (70%) had 3 or less. In comparison, the majority of the 4-year-olds (64%) had 4 or more nonfrog goal plans and only two narrators had less than half of the 6 plans 7. 7 The six nonfrog episodes pictured in the book are: the dog falling out of the window, the gopher biting the boy, the dog chasing the bees, the owl flying out of the tree, the bees chasing the dog and the boy getting stuck on the deer's head. A category for 'other' nonfrog goal plans was added, as some of the narrators included goal plans other than these six. 85 Table 3.15: Three and 4-year-old narrators grouped according to number of nonfrog goal plans. # of Nonfrog Goal Plans 3-year-olds (n=10) 4-year-olds (n=11) More than 6 1 ~ 6 1 2 5 — 3 4 1 2 3 4 2 2 2 2 1 1 — Table 3.16a: Complexity of nonfrog goal plans for 3-year-old narrators. Subject Nonfrog goal plan units # Units/Plan # Plans 1 39 6.5 6 2 61 6.8 9 3 25 6.3 4 4 13 4.3 3 5 5 5.0 1 6 17 5.7 3 7 7 3.5 2 8 8 4.0 2 9 16 5.3 3 10 12 4.0 3 Group Mean: Range: 20.3 5-61 5.1 3.5-6.8 3.6 1-9 86 Table 3.16b: Complexity of nonfrog goal plans for 4-year-old narrators. Subject Nonfrog goal plan units # Units/Plan # Plans 1 16 4.0 4 2 34 6.8 5 3 29 5.8 5 4 10 5.0 2 5 46 7.7 6 6 14 4.7 3 7 13 4.3 3 8 23 4.6 5 9 19 4.8 4 10 7 3.5 2 11 30 5.0 6 Group Mean: 21.9 5.1 4.1 Range: 7-46 4-7.7 2-6 Question 5b: Is there a difference between the 3-year-old and 4-year-old groups in the pattern of nonfrog episodes that the children included in their stories? All of the nonfrog goal plans were encoded by a higher proportion of the 4-year-olds than the 3-year-olds, as shown in Table 3.17. The 3 most common plans were the same for both groups: (1) the boy being carried away to the pond by the deer, (2) the dog falling out of the window with the jar on his head, and (3) the bees chasing the dog because he broke their hive. Of the younger subjects, five had no episode in which the dog and the bees were involved in a chase of some sort. Only two of the 4-year-olds had no such goal plan. Both age groups tended to include more 87 pictured than inferred plans, 75% and 78% respectively. For the plans that were coded as Other, 6 were related to a pictured event (e.g. an episode about the little frogs), and 3 could be inferred from the story context (e.g. predicting what happened next after the boy and dog left the pond). Table 3.17: Rank order of 3- and 4-year-old narrators encoding nonfrog goal plans. Nonfrog goal plan P/l 3-year-olds (n=10) Nonfrog goal plan P/l 4-year-olds (n=11) Boy carried by deer P 1.00 Boy carried by deer P 1.00 Dog falls out window P .70 Dog falls out window P .91 Bees chase dog P .40 Bees chase dog P .55 Gopher bites boy I .40 Dog chases bees P .55 Owl flies out of tree I .30 Owl flies out of tree I .45 Other goal plan P/l .30 Gopher bites boy I .36 Dog chases bees P .30 Other goal plan P/l .36 Nongoal Plan Units Question 6: Did the 3-year-old and 4-year-old groups differ in the kinds of nongoal plan units they encoded? The 3-year-olds had more nongoal plan units of all four types than did the 4-year-olds. Units coded as narrator strategies were the most common nongoal plan type, with descriptions of events falling a distant second. Only the narrator strategy units made up a meaningful proportion of the story. At least one description of an event was encoded by 90% of 3-year-olds and 27% of the 4-year-olds. Descriptions of a state were encoded by the same number of 3-year-olds as Table 3.18: Mean number and proportion of nongoal plan units by type. 88 3-year-olds (n=10) 4-year-olds (n=11) Nongoal Plan Unit Type Units % Units % Description of an Event 2.2 4% .64 1.0% Description of a State 0.9 1% .45 0.8% Naming 2.0 3% .36 0.6% Narrator Strategy 10.0 17% 7.00 12.0% 4-year-olds, at four narrators per group. Eighty percent of the 3-year-olds named some object or nonprotagonist character in their stories, compared to 27% of the older subjects. All of the narrators used at least one rhetorical strategy to maintain listener attention in their narrative, such as making predictions about what would happen next, or calling attention to something in the picture. Content-based Comparisons In this section, I present the results of the content-based analyses done on the 3- and 4-year-old subjects. Although I have frequently mentioned the differences between the two age groups in the previous section, most of these differences were, in fact, quite small, and likely not statistically significant. However, I also had an impression that there were important qualitative differences among the narratives produced by the children in the two age groups. In an attempt to capture these differences, I decided to do further analyses on the content of the narratives, changing 'age' from an independent to a dependant variable. All of the children were placed in a single group to determine if a pattern of coherence and/or complexity development could be seen, regardless of age. A subset of the analyses done on an age group basis were chosen, and individual z-scores were calculated from the group mean for each measure. The z-scores for the two coherence 89 measures (initiating episode and proportion of goal plan units) and the two complexity measures (number of initiating events and number of attempts) that best captured the quality of the narrative were combined for a composite z-score. Each story was ranked for its coherence and complexity relative to the other children, based on these two composite z-scores. The narratives were categorized as High (z-score of +1 SD or higher), Medium (z-score between +/-1 SD), or Low (z-score at or below -1 SD), coherence and/or complexity. Subjects were renumbered as: 3+subject number and 4+subject number to facilitate comparison between subjects. Coherence Analysis The narratives with combined z-scores at, or above, +1 SD for the proportion of goal plan units and the initiating episode measures were classified as High coherence. These narratives contained both a higher than average percentage of goal plan units (45% or higher) and at least four of the five necessary components for the initiating episode. The Medium coherence narratives fit three patterns: 1) a higher percentage of goal plan units but a relatively low score for the coherence of the initiating episode, and 2) a lower than average percentage of goal plan units and a coherent initiating episode, or 3) an average score for both measures. The 3-year-olds in this group tended to fit the first pattern (four of five) while the 4-year-olds tended to fit the second (two of four). One narrator of each age fit the third pattern. Of the Low coherence narratives, three had below average scores on both the goal plan percentage and the initiating episode coherence measures, two had a fair percentage of goal plan units (approximately 40%) but a very limited initiating episode, and one had a coherent initiating episode but very few goal plan units overall. All of the children whose narratives were coded as most coherent were 4-year-olds, with a mean age of 57.3 months (4;9.10). The narratives that were of average coherence for the group were 90 told by five 3-year-olds and four 4-year-olds, with a mean age of 48.9 months (4;0.27). The least coherent narratives were told by five 3-year-olds and one 4-year-old, with a mean age of 43.7 months (3;7.21). The narrators were ranked as follows, in Table 3.19, based on their composite z-score for the percentage of goal plan units and initiating episode scores. Table 3.19: Coherence ranking of 3- and 4-year-old narrators combined. Coherence Level Subjects High Coherence 4-10, 4-7, 4-9, 4-11, 4-6/4-1 (Ranks 1-5.5) Medium Coherence 4-2, 4-3, 3-10/3-6/3-3, 4-5, 3-9/3-7, (Ranks 7-15) 4-4 Low Coherence 3-2, 3-5, 3-4, 4-8, 3-8, 3-1 (Ranks 16-21) Note: #/# indicates a tie. When the composite z-score was compared to the ratio of plot unfolding units to total number of plot units in the narrative, the two were found to correlate quite well (see Table 3.20). Narratives that were scored as either High, Medium, or Low coherence based on the percentage of goal plan units and the initiating episode coherence measures tended to be scored the same way on the percentage of plot unfolding units measure. Those narratives that changed coherence classification were generally nearer the cut-off for their original grouping and moved up or down one group. No narrative differed by two coherence levels between the two measures (i.e. went from High to Low or vice versa). Overall, the 3-year-olds tended to move up one level (five children) and the 4-year-olds tended to move down from the High category to Medium (five children) on the percentage of plot unfolding units measure. 91 The overall coherence of the longitudinal subjects' narratives increased between ages three and four. Subject 1 (BT) made the most improvement, encoding twice the number of events in the initiating episode, more than twice the percentage of goal plan units, and twice the number of plot unfolding units as he or she did as a 4-year-old. Likewise, subject 4 (DM) produced a narrative that was more coherent according to all three measures. Subjects 3 (SK) and 6 (DA) encoded slightly lower percentages of goal plan units and plot unfolding units than they had as 3-year-olds but had more coherent initiating episodes. Subject 10 (RL) improved for the first two measures but had a lower ratio of plot unfolding units compared to the total number of plot units. Table 3.20: Coherence scores for 3- and 4-year-old narrators as a whole. Subject % Goal Plan Units Initiating Episode Combined z-score Ranking (H/M/L)* PUU**/ Total Plot Un 1 21 2 -3.42 L .126 (L) 2 29 4 -0.95 L .218 (L) 3 48 4 0.51 M .394 (M) 4 48 2 -1.35 L .237 (L) 5 37 3 -1.26 L .250 (M) 6 60 3 0.51 M .500 (H) 7 67 2 0.12 M .482 (H) 8 41 1 -2.81 L .302 (M) 9 55 3 0.12 M .431 (H) 10 60 3 0.51 M .390 (M) 11 54 4 0.98 H .370 (M) 12 52 4 0.82 M .361 (M) 13 37 5 0.60 M .250 (M) 14 51 3 -0.18 M .371 (M) 15 44 4 0.21 M .310 (M) 16 54 4 0.98 H .450 (H) 17 57 5 2.14 H .400 (M) 18 26 3 -2.11 L .154 (L) 19 59 4 1.36 H .330 (M) 20 70 4 2.21 H .355 (M) 21 45 5 1.21 H .304 (M) Mean: 48.3 3.4 - .333 Range: 21-70 1-5 - - .126-.500 SD: 13 1.0 - - .100 *H=High (rounds to +1S.D.), M=Medium (rounds to between +/-1 S.D), L=Low (rounds to -1 S.D.). **PUU=Plot Unfolding Units. 93 Complexity Analysis The High complexity narratives had composite z-scores one standard deviation above the group mean for the number of initiating events and number of attempts measures. These narratives contained at least 5 initiating events and 7 attempts to resolve the primary goal plan. As with the Medium coherence narratives, the Medium complexity narratives fit 3 different patterns: (1) a high number of initiating events and fewer than 7 attempts (2) an average number of initiating events and attempts, and (3) fewer than 4 initiating events and more than 7 attempts. Narrators of both ages used each of these patterns. The Low complexity narratives had only 2 or 3 initiating events and four of the five had 4 or fewer attempts to resolve the goal plan. Only five 3-year-old narrators, with a mean age of 40.8 months (3;4.24) had Low complexity narratives according to these initial measures (see table 3.21). The majority of the 4-year-olds' narratives (seven of eleven) and three of the 3-year-olds' narratives were scored as Medium complexity, with a mean age of 52.5 months (4;4.15). The remaining children's narratives (two 3-year-olds and four 4-year-olds) were scored as High complexity, with a mean age of 51.2 months (4;3.6). The narrators were ranked as follows, based on their combined z-score for the number of initiating events and the number of attempts. Table 3.21: Complexity ranking of 3- and 4-year-old narrators combined. Complexity Level Subjects High Complexity 4-5,3-2,4-2/4-11,4-9,3-10 (Ranks 1-6) Medium Complexity 4-7, 3-6, 4-10, 4-6, 4-3, 3-3, (Ranks 7-16) 4-1/4-8, 3-1,4-4 Low Complexity 3-9, 3-8, 3-4, 3-5/3-7 (Ranks 17-21) Note: #/# indicates a tie. 94 Table 3.22: Complexity scores and ranking for 3- and 4-year-old narrators. Subject Initiating Events # of Attempts Combined z-score ComplexityR anking Nonfrog Goal Plans Nongoal pi units* 1 5 4 -0.42 M 6 (H) 41 (L) 2 5 13 2.41 H 9 (H) 33 (L) 3 3 10 -0.07 M 4 (M) 9(M) 4 3 3 -2.27 L 3(M) 7(M) 5 2 3 -3.04 L 1 (L) 15 (M) 6 4 9 0.38 M 3(M) 3(H) 7 2 3 -3.04 L 2(L) 2(H) 8 3 4 -1.96 L 2(L) 17 (M) 9 3 7 -1.02 L 3 (M) 7(M) 10 4 11 1.01 H 3 (M) 12 (M) 11 5 5 -0.11 M 4 (M) 5(H) 12 6 9 1.91 H 5 (M) 6(H) 13 5 6 0.21 M 5 (M) 9 (M) 14 4 5 -0.88 M 2(L) 7(M) 15 6 12 2.85 H 6(H) 1 (H) 16 3 11 0.24 M 3(M) 14 (M) 17 5 8 0.84 M 3 (M) 4(H) 18 5 5 0.11 M 5 (M) 34 (L) 19 6 7 1.14 H 4 (M) 3(H) 20 6 4 0.34 M 2(L) 2(H) 21 6 9 1.91 H 6(H) 8 (M) Mean: 4.3 7.0 _ - 3.9 11.4 Range: 2-6 3-13 - - 1-9 1-41 SD: 1.4 3.2 - - 1.9 11.3 *High>+0.5 SD, Medium=between +/-0.5 SD, Low<-0.5 SD. The criterion was changed because of the amount of variability in the scores. The majority of the narrators had similar complexity scores for the secondary plot line, as they did for the global goal plan (thirteen of twenty-one children). Of the eight narrators who scored 95 differently on the two measures, three had more complex secondary plot lines than primary plot lines (subjects 1, 4, and 9) and five had less complex secondary plot lines (subjects 10, 12, 14, 19, 20). Subjects 1, 2, 8, and 18 encoded a large number of nongoal plan units. Except for subject 8, who already had a low complexity narrative, these nongoal plan units had a negative effect on the overall complexity of the narrative. Encoding an average or low number of nongoal plan units can have different effects on the overall complexity of the narrative. It certainly does not increase the complexity, although such a score may be used to infer the narrator's level of goal plan knowledge, which in turn reflects goal plan complexity. Of the five narrators who had low complexity rankings, three also encoded fewer than 10 nongoal plan units. Therefore, the lack of nongoal plan units may be due to a general lack of any type of plot unit rather than a higher level of narrative complexity. The overall complexity of the narratives produced by longitudinal subjects 1 (BT), 3 (SK), and 4 (DM) increased from age three to four. The overall complexity of subjects 6 (DA) and 10's (RL) narratives decreased between the two time periods. However, only RL's narrative fell into a lower complexity category. The complexity of their secondary plot lines decreased in three instances (BT, DM, RL), remained the same for DA, and increased by 1 episode for SK. All of the longitudinal subjects except DA encoded the same number or fewer nongoal plan units. DA encoded 11 more nongoal plan units than he had as a 3-year-old. Comparison of Complexity and Coherence Rankings Four categories of narratives are logically possible from the complexity/coherence paradigm -coherent and complex, coherent but not complex, complex but not coherent, and neither complex nor coherent. Narratives that were categorized as High or Medium complexity were counted as 96 complex. Those narratives that were categorized as High or Medium coherence were counted as coherent. The narratives were then sorted into each of the four categories (see table 3.23). Table 3.23: Categorization of narratives by coherence and complexity. Coherent and Complex Complex, Not Coherent 3-3, 3-6, 3-10, 4-1,4-2, 4-3, 3-2, 4-8, 3-1 4-4, 4-5, 4-6, 4-7, 4-9, 4-10, 4-11 Coherent, Not Complex Not Complex, Not Coherent 3-7, 3-9 3-4, 3-5, 3-8 The majority of the children told relatively coherent and complex narratives. The narratives told by the 3-year-olds were divided quite evenly among the four categories. Half of them told narratives that were complex, and four 3-year-olds told coherent narratives. However, three children told narratives that were neither coherent nor complex. In contrast, only one 4-year-old, R C , told a narrative that was not both coherent and complex. No 4-year-old narrated a story that was neither coherent nor complex. The longitudinal subjects either remained in the coherent and complex category (subjects 3, 6, and 10) or moved up to that category (subjects 1 and 4) as 4-year-olds. 97 C H A P T E R FOUR DISCUSSION Overview In this chapter, I will discuss the results presented in Chapter Three with respect to previous research in the area of narrative development and the research questions posed. In particular, I will compare the results with those of Berman and Slobin (1994) and Trabasso and his colleagues (Trabasso & Nickels, 1992; Trabasso, Stein, Rodkin, Park Munger & Baughn, 1992; Trabasso & Rodkin, 1994). I present the general conclusions drawn from the results and their implications for future research are provided. Age-based Comparisons As mentioned in the previous chapters, the coherence and complexity of a narrative are inter-related. The development of narrative complexity involves encoding more goal plan relevant information, particularly in the plot unfolding portion of the narrative (e.g. more episodes, as marked by G A O sequences). The development of narrative coherence also requires that "more" of certain elements are encoded, but the elements are chosen because they are vital to the development of the plot and are closely tied to other elements of the story (e.g. goal statements remind the listener of the theme of the narrative and motivate renewed attempts; these attempts, in turn, maintain the plot line, and their outcomes motivate the protagonist's next action). The same elements are being counted in these two examples, however, the purposes for counting them are different -determining the number of episodes in the narrative, versus determining how well the narrator has motivated, pursued, and resolved the goal of the story. 98 Goal Plan Knowledge A narrator's knowledge about the underlying structure of a narrative determines how he or she encodes the content of the story. The result of this knowledge is a coherent narrative with at least the basic setting, initiating event, attempt, and outcome components of the plot. Its effects can be seen both at the local level, in the form of coherent local episodes, and the global level, through continuing the goal plan over multiple episodes and encoding multiple goals. Since Berman and Slobin (1994) and Trabasso et al. (1992) found that few, if any, of their 3- and 4-year-old subjects were able to structure their narratives at either a local or global level, some initial measures of this knowledge in my own subjects were necessary. Therefore, the first question was: Do preschool aged children have knowledge of underlying goal plans of action, as measured by: (a) the presence of goal-relevant attempts, purposes and outcomes, (b) encoding multiple attempts at a goal, (c) goal-attempt-outcome sequences, and (d) encoding sufficient information for the listener to infer superordinate (Take) and subordinate (Search and Find) goals? The children in both groups demonstrated that they had sufficient narrative competence to tell this story, although a wide range of abilities was evident, both between and within the two groups. All of the narrators satisfied Trabasso's criteria for allowing the inference of underlying goal plan knowledge, as outlined in Chapter Two. None of the 4-year-olds told as minimally structured a narrative as the three weakest 3-year-olds. However, the remainder of the 3-year-olds told narratives that were comparable to the 4-year-olds' narratives. Within this more knowledgeable group of 3-year-olds, there were five children whose performance was as strong as the best 4-year-olds. On the other hand, the 4-year-olds were more consistent in their ability to selectively encode information according to the global goal plan, as shown by: the number of narrators who encoded both the Find and Take goal plans, the tendency towards a higher number of principal goal plan units, and more frequent encoding of multiple episodes. Given Bamberg and Marchman's (1990) 99 findings on the importance of renewed attempts to the plot, the weaker 3-year-old narratives would be judged as having less developed, and less coherent plots. All of the narrators demonstrated better knowledge of goal plans of action than the subjects in the Trabasso and Berman and Slobin studies. Their 3-year-olds demonstrated little knowledge of goal plans of action, encoding mostly irrelevant states and actions and descriptions of the story context. None of their narratives was based on either a Find or a Take goal plan, and G A O sequences were rare (Trabasso & Nickels, 1992). In contrast, the 3-year-olds in the present study were found to be in a state of transition between knowledge of simple global goal plans (i.e. good local episodic structure and one global goal plan - Find) and knowledge of hierarchical goal plans (i.e. two global goal plans - Find and Take). The 4-year-olds in the Trabasso studies demonstrated slightly better goal plan knowledge than their 3-year-olds, with approximately 20% of them encoding a Find goal, but none encoding a Take goal. Local goal plans (GAO sequences) were still rare (Trabasso & Nickels, 1992). The performance of the subjects in my study was more consistent with the performance of the 5- and 9-year-olds in the Trabasso and Nickels study for these measures. The Development of Narrative Coherence The coherence of a narrative can develop either globally or locally, or both. Local coherence is achieved by creating semantic, temporal and causal ties among the various elements of the story. Global coherence requires the narrator to encode certain key components of the plot and avoid adding unnecessary information that is not tied to the rest of the narrative. The following questions related to primary plot coherence were asked: Question 2a: How does the pattern of global narrative coherence change from ages 3 to 4, as measured by a) the number of essential story elements, and b) the distribution of global goal plan, nonfrog goal plan, and nongoal plan units? 100 Question 2b: Did the 4-year-old subjects encode more information relevant to the opening of the plot than the 3-year-old subjects? Question 2c: Did the 4-year-olds' narratives unfold in a more coherent manner than the 3-year-olds', as demonstrated by: a) more frequently sustaining the search for the frog until the end of the story, b) including more attempt-outcome pairs, and c) dedicating a larger proportion of the story to achieving the three goals? Question 2d: Do the two groups show different patterns of goal plan continuation over the length of the story, as indicated by the locations of the attempts? Question 2e: Did more 4-year-olds than 3-year-olds encode a successful outcome for the superordinate goal plan? Global Narrative Coherence All of the children scored well on the Trabasso and Rodkin (1994) essential story elements criteria for global coherence, encoding no fewer than four of the six possible elements. Although the possession relation was encoded by only three children in each group, all of the other elements were included by a majority of the narrators in both groups. These scores indicate that all of the children's narratives had some degree of global coherence. In contrast, the 3-year-olds in the Trabasso et al. (1992) study encoded very few of the essential story elements (e.g. 17% encoded the possession relationship, and attempts and Find or Take outcomes were rare). Their 4-year-olds demonstrated some development on this measure, with higher proportions of narrators encoding each element. This development resulted in a typical 4-year-old narrative being described as an "action sequence that is relevant to the goal plan" (Trabasso & Rodkin, 1994; p. 147). However, their performance was still very different from that of both the 3- and the 4-year-olds in this study, whose narratives were much more structurally coherent, with multiple initiating events, attempts, and a successful outcome for at least one of the global goal plans (Find goal). 101 Trabasso and his colleagues claim that the relationship between the boy and the frog is of great importance to the coherence of the plot because it motivates the boy's desire to search for and retrieve the frog. However, as mentioned previously, very few of the children in this study encoded this component of the story. In fact, those children who did mention the relationship had the poorest scores for goal plan knowledge for their age groups, and had a range of scores for the proportion of their narrative encoded as goal plan units (two below the mean, two at the mean, and two above the mean). Clearly, this component is not predictive of the narration of a coherent narrative and can be eliminated in any future analyses of coherence as measured by goal plan knowledge. Overall, the 3- and 4-year-old groups both fit the mature pattern for the distribution of goal plan, nonfrog goal plan, and nongoal plan units. The majority of narrators encoded mostly goal plan units and fewer nonfrog goal plan or nongoal plan units. On average, the 4-year-olds encoded more information relevant to either a global or local goal plan than did the 3-year-olds. The 4-year-olds' narratives would therefore contain more semantic, causal and temporal ties between clauses and would likely be judged as more coherent than the 3-year-olds' stories. Individually, narrators differed from each other in terms of the exact number or proportion of units associated with each category. Part of this variability was due to differences in overall narrative length, but it also reflects some differences in the ability to selectively encode information relevant to the underlying goal plan. Those narrators who scored lower on the goal plan knowledge measures also tended to encode lower proportions of goal plan units relative to their peers. Two interesting exceptions to this pattern were subject 4-10 and subject 3-2. Subject 4-10 scored quite low, compared to his peers, on the goal plan knowledge measures, but encoded a high proportion of goal plan units (70%). In this case, the overall length of the narrative affected both 102 measures substantially. Because his narrative was quite short (31 plot units), the number of principal goal plan units was also quite low (8 units). However, the numbers of nonfrog goal plan units and nongoal plan units were also very low (7 and 2, respectively), resulting in a good score for global coherence. This child's narrative would likely be judged low in terms of complexity but globally quite coherent. Subject 3-2 was essentially the opposite case to 4-10. She scored high on the goal plan knowledge measures but encoded less than a third of her narrative as part of the global goal plan. The majority of her narrative consisted of nonfrog goal plan units. A possible conclusion is that she has good knowledge of local goal plans but more difficulty with global narrative structure. Her narrative would probably be judged as less coherent because its primary focus is the secondary, rather than the primary, plot line. Plot Opening Coherence The coherence of the opening of the plot is important because it allows the listener to correctly infer the goals of the story. The listener needs to know who is in the story, what context (temporal and physical) it occurs in, and what events take place to motivate the rest of the narrative. Although Berman and Slobin (1994) and Trabasso and his colleagues include the boy's relationship to the frog in their plot opening measures, for the reasons outlined above, with respect to global narrative coherence, it is eliminated from the current discussion. The number of plot opening units and the proportion of the goal plan that is dedicated to the opening of the plot were used as rough measures of how well the narrator had encoded this information. By analysing the setting and the initiating episode, the functions of the components they have included can be determined. 103 On average, the 4-year-olds encoded two more plot opening units than the 3-year-olds. Individually, they were more consistent than the 3-year-olds at encoding the key components of this category. Their understanding of what their listener needed to know about this first portion of the story was better developed than it was for several of the 3-year-olds. For example, while the majority of narrators in both groups introduced the frog, the 4-year-olds more frequently introduced both the dog and the boy, or all three of the characters involved. They also encoded the temporal location of the narrative more often (eight compared to five). The 4-year-olds' initiating episodes required less effort on the part of the listener to determine what was happening and how it related to previous and subsequent events. The individual variability in the number and proportion of plot opening units can be accounted for in three ways: (1) the children were unsure of what information to include at the beginning of a story (2) they did not understand the need for this information, or (3) the complexity of this portion of the narrative is determined, in part, by the person's narrative style. It seems likely that all three accounts are true to some extent for these young children. The third possibility is supported by the performance of the longitudinal subjects. Three of the subjects (subjects 1, 4, and 10 - BT, DM, and RL) encoded fewer plot opening units as 4-year-olds than as 3-year-olds. It seems unlikely that they forgot what their listener needed to know, especially since they all narrated better initiating episodes as 4-year-olds. In fact, the primary reason for the decrease was a reduced number of opening events in the setting. Opening events are sensitive to changes in personal storytelling style because they contribute to the coherence of the plot but are not mandatory. Therefore, their omission is of limited significance to the listener and has little effect on the coherence of this portion of the plot. 104 Both the 3- and the 4-year-olds performed better than predicted by Berman and Slobin's (1994) results for the same measure. The majority (seven children) of their 3-year-olds encoded no components for the initiating episode, compared to the 3 or more components encoded by six of the 3-year-olds in this research study. Similarly, their 4-year-old subjects usually encoded 2 components, compared to the 4 or more components encoded by nine of my 4-year-olds. Several of the children's initiating episodes were narrated using the "tightly packaged syntactic means of expression" (p.53) noted by Berman and Slobin for much older narrators. An example of this type of episode is taken from 4-year-old subject 1 (BT): When it was night... the boy looked in his bedroom with the frog. And then, when he was sleeping, the frog jumped out of the window. Then the boy woke up! And the frog was gone!! He searched and searched. Plot Unfolding Coherence For the plot to unfold in a coherent manner, the boy's search for the frog must continue until a frog is found and taken home by the boy and his dog. Each attempt to resolve these goal plans should be accompanied by either an explicit outcome or one that can be inferred from the subsequent narrative context. The end result of this sustained search is that a larger number of attempts and outcomes occur and a higher proportion of the narrative is dedicated to the unfolding of the plot. The majority of the 4-year-olds' unfolded their narratives in a more coherent manner than the 3-year-olds'. They more frequently sustained the search for the frog until the end of the story (eight compared to five) and tended to have more attempt-outcome pairs, although the group means were similar. These results support the notion that their goal plan knowledge is generally better established, and therefore, they know what plot elements to include and how to tie them to other elements of the story. 105 None of the narratives with fewer than 15 plot unfolding units, regardless of overall length, contained a sustained search, using the critrion outlined in Chapter Two. Relative to the adult target narrative, they would be judged as having a less coherent plot unfolding (subjects 3-1, 3-4, 3-5, 3-7, 3-8, 4-4, 4-8, 4-10). However, if the plot of the book is ignored for the moment, some of them might still be considered coherent, but lacking in relative complexity (subjects 3-7, 4-4, 4-10). These narratives were shorter than the mean but contained locally coherent episodes, both in the opening and the unfolding of the plot, and low proportions of nongoal plan utterances. In fact, the criterion used to determine if the plan was sustained may have been too strict, since the eight narratives that were scored as having incomplete goal plans had final outcomes for the Find goal and five had a final Take outcome. Although Trabasso and his colleagues do not describe their criterion for a sustained search, none of their 3-year-olds could have had sustained searches, since their narratives were not based on goal plans to begin with, and attempts were rare. Their 4-year-olds encoded more attempts than the 3-year-olds did at each of the locations, but only 25% encoded an attempt at the final location (Log) (Trabasso et al., 1992; p. 145, Trabasso & Rodkin, 1994; p. 100). Their performance decreased across the length of the narrative, particularly after the Room location. Differences in the expected pattern of attempts are discussed further below, with respect to search coherence. In my study, the coherence of the plot unfolding improved from ages 3 to 4 for two of the longitudinal subjects (BT and DA). One subject remained the same (DM), while the remaining two subjects' narratives decreased on this measure (SK and RL). Subject S K still told a coherent narrative, but encoded fewer attempts as a 4-year-old, which accounts for the decrease in attempt-outcome pairs. As mentioned previously, subject 4-10 (RL) told a much shorter narrative than the 106 average. In fact, it was half the length of his 3-year-old narrative. The decrease in the apparent coherence of his narrative is directly related to this decrease in length. Search Coherence In addition to whether or not the search was continued to the end of the narrative, it was important to determine whether there were differences in the attempt locations the children encoded. In particular, were there scenes in the story that the children had difficulty with, or more frequently encoded attempts for? Based on results from previous studies (Trabasso et al., 1992; Trabasso & Rodkin, 1994), I expected to find limited search coherence for both subject groups. The 3- and 4-year-old age groups each showed quite different patterns for their searches. One similarity between the two was that the vast majority of children encoded attempts at the beginning of the search and at the end of the search. The consistent inclusion of these two portions of the search indicates that the majority of the children in both groups understood their relative importance to the coherent development of the plot. The beginning of the search motivates the formation of the global Find goal plan, while the end of the search provides a resolution for that goal plan. However, the 4-year-olds demonstrated a better understanding of these roles than the 3-year-olds by encoding relatively more attempts at these locations. A second similarity was that very few narrators, in either the 3- or 4-year-old groups, encoded attempts at the Hive, Rock, or Cliff locations. These scenes were generally encoded as local, nonfrog goal plans. The decline in the number of narrators encoding attempts across the length of the narrative seen for the 4-year-olds in the Trabasso et al. study (1992) did not occur for either group of narrators in my study. The pattern of attempts in the middle portion of the search was a second difference between the two groups of children. The number of 4-year-olds encoding attempts dropped substantially after 107 the Hole location, while the number of 3-year-olds increased substantially for this location and the Tree location. The exact reason for this difference is unclear, but event salience, and the interaction of the secondary plot line (i.e. nonfrog episodes) are possible explanations. The outcomes of the boy's attempts at these locations are highly salient in that they are dramatic — the boy gets bitten, and he falls out of the tree. These dramatic events may have made the attempts more salient, and therefore more important, to the 3-year-olds, regardless of their relative importance to the plot. The 4-year-olds also viewed these events as important, but the majority of them focussed on the outcomes of the attempts and their effects on the boy, rather than their significance to the search. In support of this view, the Tree location was encoded as a nonfrog goal plan by three of the five 4-year-olds who did not encode an attempt at that location, while the other two described the actions of the dog in that scene. A similar interaction between the primary and secondary plot lines was found for the 3-year-olds with respect to pronominal reference (Gomme, 1994). She found that some of the children attempted to distinguish between whether the boy or the dog was the agent in the scene by altering the form of reference used - from the full noun to the pronoun he or it. In other cases, the children assigned names to the characters but quickly abandoned this strategy in favour of the full noun or the pronouns "he" and "it". The conclusion is that some of the children from both groups experienced difficulty integrating the two plot lines for the local and global goal plans. This difficulty might be due to limited resources for linguistic and cognitive processing, less developed knowledge of goal plans of action, or less complete knowledge of how to use linguistic forms to accomplish the integration. As mentioned above, the search patterns seen for the 3- and 4-year-olds in this study were substantially different from those obtained by Trabasso and his colleagues. The major source of 108 this discrepancy is that attempts were encoded at each of the locations by much higher proportions of narrators than reported by Trabasso and Rodkin (1994). In addition to the general increase in the proportion of narrators encoding attempts at each particular location, the two groups did not fit any of the attempt patterns outlined in the Trabasso study. Plot Concluding Coherence The narration of the outcomes for the Find and Take goal plans is a key component of telling a coherent Frog narrative. These outcomes provide resolutions to the goals established at the beginning of the narrative, bringing the story to an end. Based on Trabasso et al. (1992), I expected that few of the 3-year-olds and only slightly more 4-year-olds would encode the Find outcome. Only 4-year-olds were expected to encode a Take outcome. However, all of the narrators encoded a Find outcome and eighteen also encoded a Take outcome. As predicted by the number of Take goals that could be inferred at the beginning of the narrative, more of the 4-year-olds encoded the Take outcome (ten compared to eight 3-year-olds). The developmental progression seen did not fit Trabasso's predicted pattern of Find>Take>Find + Take, since no narrators encoded a Take outcome without a Find outcome. This result is not surprising, since a Take outcome entails a Find outcome. Secondary Plot Coherence Trabasso and Rodkin (1994) claim that goal plan knowledge and structure facilitates both local and global plot coherence. Given the results from the goal plan knowledge measures, the following question was asked to determine if there were changes in the coherence of the secondary plot line: Question 3: Did the 4-year-olds have more coherent secondary plot lines than the 3-year-olds, as measured by a) the number of nonfrog Goal-Attempt-Outcome sequences, b) the proportion of 109 G A O ' s to total nonfrog episodes, and c) the number of shared elements between nonfrog episodes? As a group, the 4-year-olds had more globally coherent secondary plot lines than the 3-year-olds due to an increased number of intra- and interepisodic ties. Individually, there was overlap between the two groups, with four of the 3-year-olds scoring well on all three measures and four 4-year-olds scoring poorly on all three measures of coherence. As expected, performance on the measures of local coherence was better than the global measures. All of the narrators had at least one locally coherent episode, and all but one child, a 3-year-old, narrated the majority of their nonfrog episodes in a coherent manner. Those children who scored poorly on the goal plan knowledge measures also scored poorly on the measures of nonfrog goal plan coherence. This finding supports Trabasso and Rodkin's claim that goal plan knowledge and structure facilitates both local and global plot coherence. As mentioned previously, episodes in the secondary plot line can either contribute to the coherence of the narrative as a whole, or detract from it. When an episode has a relational tie to either another nonfrog episode or a frog episode, it contributes to the overall coherence because it is integrated into the larger plot line of the narrative. If a nonfrog episode is unrelated to other events in the story, it detracts from the overall coherence of the narrative because it disrupts the thematic "flow" of the plot. This integration of episodes in the secondary plot line was found for four of the 3- year-olds and seven of the 4-year-olds. Berman and Slobin (1994) also found that their 3- and 4- year-old subjects could produce locally coherent episodes. However, these episodes were not well linked to either the secondary or the primary plot lines. 110 The Development of Narrative Complexity As mentioned with respect to narrative coherence, development can occur either globally, throughout the narrative, or for a particular portion of the plot. The changes seen in the complexity of the narrative are primarily due to changes in the individual's knowledge of goal plans of action. An increased sensitivity to the listener's needs was also found to have an effect on the relative complexity of the narrative. The following questions were asked to determine what changes in primary goal plan complexity occur between the ages of three and four: Question 4a: Did 4-year-olds produce either a) longer narratives, b) narratives containing more goal plan units, or c) more G A O sequences than younger subjects? Question 4b: Was the opening of the plot more complex in the 4-year-olds' stories than in the 3-year-olds' stories, as measured by a) the characters they introduce, b) the number of opening events they include, or c) the number of initiating events motivating the goal plan? Question 4c: Did the 3-year-olds and 4-year-olds differ in the particular initiating events they choose to include? Question 46: Does the pattern of narrated attempts change between 3 and 4 years of age, as measured by: a) the number of attempts, b) the number of attempts made to achieve superordinate and subordinate goals, c) the number of purposeful attempts, and d) the number of attempts made by the boy versus those attempts made by the dog? Question 4e: Is there a discernable difference in how the 3-year-old and 4-year-old groups encoded outcomes for either local or global goal plans? Global Goal Plan Complexity The overall complexity of a child's narrative is related to how long it is, how long the underlying goal plan is, and how many episodes it contains. The length of the narratives did not seem to change 111 substantially over the year. This measure was affected to a large extent by the narrative style the child chose. Some children told very brief narratives with little extraneous detail (15-35 unified predicates), while other children told rambling, detailed stories of up to 118 unified predicates. This influence of style was evident in the longitudinal subjects, four of whom had shorter narratives as 4-year-olds than as 3-year-olds. Based on their performance on other measures of complexity, their ability to produce complex narratives had not decreased; they simply told the story more concisely than before. In fact, this might be evidence that their knowledge of goal plans of action had improved, thereby improving their efficiency at telling a good story. The other two measures of global complexity support this conclusion, with the 4-year-olds encoding more goal plan units and more episodes than the 3-year-olds. However, as seen in some of the previous measures, not all of the 3-year-olds performed more poorly than the 4-year-olds. Six of the 3-year-olds told narratives that were as globally complex, or more so, than the average 4-year-old (subjects 3-1, 3-2, 3-3, 3-6, 3-9, 3-10). Subject 3-2 was particularly interesting because her narrative was so much longer than any of the others'. She clearly had a complex narrative, based on the criteria used in this study (e.g. encoding 9 episodes) but her knowledge of story structure may not be as well developed as these measures suggest. Approximately 90 of her utterances had nothing to do with the search for the frog. This high proportion of unrelated material is an indication that she had difficulty selecting what information to include, and may, therefore, have included everything that came to mind. In fact, some of the concise narratives were actually more structurally complex than some of the longer narratives. They dealt only with the global goal plan and a few local goal plans rather than using a great deal of description or rhetorical strategies. For this reason, and based on the longitudinal data, I concluded than overall narrative length has limited use as a complexity measure. 112 Instead, global narrative complexity due to better goal plan knowledge can be viewed as the result of the selective encoding of more of those events that are the most important to the development of the plot (e.g. attempts to find the frog) and their logically associated elements (e.g. purposes and outcomes). When narratives are complex in this manner, they are also more coherent, since the additional elements are integrated into the plot of the story. Plot Opening Complexity Increases in the introduction of the characters and the number of initiating events reflect increases in the complexity of the goal plan and have a positive effect on the coherence of the plot. Overall, the 4-year-olds demonstrated increases in both these components of the plot opening, particularly in the number of initiating events. As mentioned previously, the explicit statement of the relationship between the boy and the frog was not found to be predictive of either goal plan complexity or plot coherence. Similarly, opening events, while not harmful to the quality of the narrative, did not appear to be necessary components of a narrative, but rather stylistic devices. The number of initiating events showed the most development from age 3 to age 4. These findings were supported by the performance of the longitudinal subjects, discussed, in part, above. Eight of the 4-year-olds and three of the 3-year-olds demonstrated excellent knowledge of how to encode the components of the plot opening. By introducing both the boy and the frog, and encoding almost all of the initiating events, they established a well structured plot opening with which to motivate the rest of their narrative. Four of the longitudinal subjects showed either similar or increased complexity in their plot openings as 4-year-olds. The fifth subject (subject 6, DA), encoded fewer initiating events as a 4-year-old. The 3- and 4-year-olds showed a pattern for character introductions similar to that seen in the Trabasso et al. (1992) study -- frog>boy>dog, where the frog was most common and the dog was the least common. However, they encoded 113 higher numbers of initiating events than the Trabasso subjects (3.4 compared to 1.7 for the 3-year-olds, and 5.2 compared to 3.0 for the 4-year-olds). Pattern of Initiating Events Berman and Slobin (1994) and Trabasso et al. (1992) found developmental differences in the types of initiating events their subjects encoded. Given these findings and the increase in the number of initiating events from ages 3 to 4, the initiating events were analysed to determine if these subjects demonstrated developmental differences and how those differences could be characterized. These children showed results similar to the Berman and Trabasso studies in terms of event patterns. All of the 4-year-olds and eight of the 3-year-olds narrators demonstrated an awareness of the events' importance to the plot and the listener's ability to infer the correct goal plan. Events that were depicted explicitly in the book were more frequently encoded by all narrators, although the 4-year-olds encoded more events that had to be inferred from the picture context than the 3-year-olds. An interesting difference between the two groups occurred for the Boy Sleeps and the Boy Wakes events, which were encoded by substantially more 4-year-olds than 3-year-olds. A possible explanation is that the 4-year-olds had a better understanding of how the boy's actions allowed the frog to escape. In fact, one 4-year-old (subject 4-8) explicitly stated that he would not have allowed the frog to escape, and subjects 4-2 and 4-4 made comments at the end of their narratives about the boy letting the frog escape. Plot Unfolding Complexity As mentioned previously, a basic measure of narrative complexity is the number of episodes it contains. Each attempt is an episode within the larger goal plan. The more complex each of these 114 local episodes is, the more complex the global goal plan becomes. Similarly, the inclusion of a second protagonist, in this case the dog, necessitates the inclusion of a second, subordinate goal plan. As a group, the 4-year-olds had more episodes in their Find goal plan than the 3-year-olds. In addition, more of them encoded an attempt episode for a Take goal plan. However, seven of the 3-year-olds also had multiple episodes and Take goal plans. All of these narrators clearly have sufficient knowledge of goal plans of action to base their narratives on a hierarchical goal plan. Three of the longitudinal subjects encoded more attempts as 4-year-olds (BT, DM, DA). The two subjects who encoded fewer attempts (SK and RL) encoded a high number of attempts as 3-year-olds and had much shorter narratives as 4-year-olds. The pattern of attempts, when broken down by goal and protagonist, was the same for the two groups (from most common to least common) and fit the pattern predicted by the relative importance of the attempt to the plot: Boy Find attempts>Dog Find attempts>Shared Find attempts>Take attempts This result supports the idea that most of the children have sufficient knowledge of how to tell a good story that they can selectively encode attempts. Several narrators from both groups seemed to have a standard structure for relating an attempt that was used for much of the story, as shown in the examples below. This structure either did, or did not, include a purpose. The narrators may have used this standardized attempt format to reduce the planning load of storytelling. Examples: (No purpose) "He looked in the hole. Nobody there. He looked in the tree. Nobody there." (Purpose given) "He called, "Froggie? You in there?" Nope. He climbed the tree. "Froggie?" Nobody there." 115 As mentioned previously with respect to the coherence of the unfolding of the plot, there were substantially more attempts - and, hence, outcomes and purposes -- encoded by my subjects than by the Trabasso and Berman and Slobin subjects. Therefore, narratives in my study were also substantially more complex than expected. Trabasso et al. (1992) report only one purposeful attempt encoded by a 3-year-old, compared to an average of 3.2 for my 3-year-old subjects (range: 1-6). Similarly, their 4-year-old subjects rarely encoded purposes, compared to an average 2.8 purposes encoded by my 4-year-old subjects (range: 0-7). Plot Concluding Complexity The conclusion of the plot becomes more complex, depending on the number of outcomes and which goals are resolved. Failed outcomes can either be encoded explicitly by the narrator or the listener can infer them from the narrative context. Trabasso et al. (1992) found that, with age, narrators encoded more outcomes, particularly for explicit fails and successful Take outcomes. In my study, the 4-year-olds generally had more complex plot conclusions than the 3-year-olds, although four of the 3-year-olds' narratives had a large number of outcomes. The majority of the outcomes had to be inferred by the listener from the narrative context (e.g. from a subsequent attempt to find the frog) in narratives by both the 3- and the 4-year-olds. The number of narrators who encoded explicit fails dropped from five 3-year-olds to three 4-year-olds. Given the 4-year-olds' better understanding of their listeners' needs, they may be trying to avoid redundancy in their narratives by relying more on the narrative context. Eight 3-year-olds and ten 4-year-olds encoded Take outcomes, a surprising result given the previous reports in the literature. 116 Secondary Goal Plan Complexity The following questions were asked to determine if the 4-year-olds were able to produce more complex secondary plot lines and goal plans than the 3-year-olds: Question 5a: Is there a difference between the two groups in the complexity of their nonfrog goal plans, as measured by a) the number of nonfrog goal plan units, b) the number of nonfrog goal plans, and c) the mean number of units per goal plan? Question 5b: Is there a difference between the 3-year-old and 4-year-old groups in the pattern of nonfrog episodes that the children included in their stories? Goal Plan Complexity As mentioned previously, the complexity of the goal plan can increase locally, globally, or both, depending on what has changed. Increasing either the number of nonfrog units or the number of nonfrog goal plans changes the global complexity of the secondary goal plan. Increasing the number of units per goal plan increases the local complexity of the episode. Overall, the 4-year-olds had more episodes in their secondary goal plans. Of course, with this increase in the number of episodes came an increase in the number of nonfrog units. No differences were found between the two age groups for local goal plan complexity. The increase in the global complexity of the secondary plot line may be due to several factors. Based on the variable results from the longitudinal subjects, more than one of the following factors is involved: 1. Some children were clearly more interested in relating nonfrog events — such as the dog chasing the bees -- than others, so the increase may be due to stylistic, rather than developmental, differences. 2. The 4-year-olds may be better able to pursue divergent plot lines because the underlying structure of their narratives is more complex hierarchically. 117 3. Their better knowledge of goal plans of action may ease their planning load and facilitate the inclusion of more detail than the 3-year-olds can provide. 4. Since the 4-year-olds are less affected by differences in event saliency, they have more events to choose from. Pattern of Secondary Goal Plans Given that some effect of event saliency, narrative importance, and picture abstraction was found for the initiating events, I also investigated this possibility for secondary goal plans. In fact, effects similar to those seen for initiating events were found for these events. Concrete events that preceded important points in the narrative (e.g. landing in the pond so the frogs could be discovered, and the dog falling out the window so that the search could continue outside) were the most common events for both groups of children. The bee episode was highly salient to the children, if they understood the dangers of angering bees, which several seemed to, based on their comments. This finding agrees with Berman and Slobin's report of the effect of picture saliency on increased event selection by young children (1994, p. 61). In a narrative where these events were not integrated into the overall plot of the story, the inclusion of these salient, but irrelevant, events reflected negatively on the complexity of the secondary goal plan and the narrator's knowledge of goal plans of action. Nongoal plan Units In young children's narratives, the inclusion of nongoal plan units reflects the individual's difficulty with incorporating the content of the narrative into a goal plan. Based on the results presented in the various Trabasso studies and the Berman and Slobin book (1994), I expected to find that both the 3-year-olds' and the 4-year-olds' narratives consisted primarily of utterances that were not members of either a local or a global goal plan. The following question was asked to determine 118 where changes occurred between the two groups of children: Question 6: Did the 3-year-old and 4-year-old groups differ in the kinds of nongoal plan units they encoded? While the 3-year-olds did encode more nongoal plan units than the 4-year-olds, the number was much lower than expected from the Berman and Slobin and Trabasso studies. None of the narratives had more than eight plot units that were purely descriptive. One of the 3-year-olds (subject 3-1) named each character as it appeared in the story (e.g. "Look an owl!"), but these introductions were usually accompanied by the appropriate narration of an episode, as shown in the following example: "The boy climbed the tree. No frog. Look an owl! He scared the boy." The majority of the nongoal plan units encoded were rhetorical strategies. These strategies were used by all of the children to engage and maintain their listener's interest. Where they become problematic is when they are overused, decreasing the global coherence of the narrative. The general use of these strategies can be interpreted differently, depending on the motivation ascribed for their use. Is the child aware from previous experience that the listener's attention may wander from the story and that it is the narrator's responsibility to prevent that? In this case the use of such strategies would indicate a higher level of narrative development. Or, the child may view the task as a conversation in which the two parties take turns, particularly if their parents read stories in a question-answer, conversational style. In this second situation, the child's action indicates a lower level of narrative development, as suggested by Berman and Slobin (1994). Perhaps their use occurs when children are at some intermediate stage, where they understand their role as a narrator with respect to attention but not with respect to telling a coherent, well organized story. 119 Content-based Comparisons As mentioned in Chapter Three, the group differences found for the age-based comparisons were generally small and their interpretation complicated by the variability in the group samples. Therefore, content-based comparisons of the children's narratives were done in an attempt to capture the qualitative differences seen in the narratives. These comparisons were based on composite scores for both coherence and complexity, which were then compared to a few select measures of coherence and complexity. Coherence Analysis To determine if a developmental progression for narrative coherence could be seen across the twenty-one subjects, the children's narratives were ranked according to their coherence relative to the entire group. A progression was seen in that the most coherent narratives were told by 4-year-olds and the least coherent were told by 3-year-olds. This ranking confirms the previous conclusions that the 4-year-olds generally have a better understanding of what content they must include in their story and how to encode it. The 3-year-olds, however, are in the process of acquiring this knowledge. Half of them achieve moderate coherence, while the other half achieve only low coherence. This conclusion is supported by the performance of the longitudinal subjects, all of whom increased the coherence of their narratives by age 4. Complexity Analysis The narratives were also ranked for relative complexity, disregarding age of the narrator, to determine if a developmental progression could be seen. The results of this ranking were less clear than those for coherence. The most complex narratives were told by both 3- and 4-year-olds. However, only 3-year-olds told narratives of poor complexity, which supports the idea that the 4-year-olds have more complex underlying goal plans. It appears that the 3-year-olds valued 120 encoding more of the content of the story more highly than they valued telling a coherent story. The 4-year-olds attempted to do both, and largely succeeded. This supposition was confirmed in a comparison of the two rankings, discussed below. The variability in the complexity of the 4-year-old narratives was also seen in the longitudinal subjects. A possible explanation for this variability in the 4-year-olds is that those who had less well-established knowledge of story structure had to sacrifice the complexity of the narrative in order to make it coherent. Comparison of Complexity and Coherence Rankings When the qualities of complexity and coherence are considered together, four categories of narrative classification are logically possible. The story can be: 1) Coherent and Complex, 2) Coherent but not Complex, 3) Complex but not Coherent, or 4) Neither Coherent nor Complex. As predicted by the previous discussion', all but one of the 4-year-olds' narratives was classified as coherent and complex. The 3-year-olds, in contrast, were spread out quite evenly among the four categories. Using the assumption outlined in the Complexity section for the longitudinal subjects, the degrees of goal plan knowledge would be as follows: High Goal Plan Knowledge: Narrators whose stories were classified as both coherent and complex. High-Medium Goal Plan Knowledge: Narrators whose stories were classified as coherent, but not complex. Low-Medium Goal Plan Knowledge: Narrators whose stories were classified as complex, but not coherent. 121 Low Goal Plan Knowledge: Narrators whose stories were classified as neither coherent nor complex. The trade-off between coherence and complexity seen in this hierarchy of goal plan knowledge is in accord with Karmiloff-Smith's (1983) work on the interaction of top-down and bottom-up processing in various cognitive tasks. The children initially use whatever semantic content they can — a resource-intensive bottom-up planning process — to form their narratives. These narrators belong to the low goal plan knowledge category. At the next level, the narrator has acquired some knowledge of goal plans of action, which decreases the planning load and facilitates the inclusion of more story elements such as attempts.and outcomes. Although this is a top-down process, the narrator does not have sufficient knowledge to select elements based on their importance to the development of the plot. At the third level, the child becomes aware that some elements are more important to include than others. This increased awareness places more demand on the narrator because he or she must coordinate the use of both top-down and bottom-up planning processes. The result is a decrease in the complexity of the narrative, as linguistic and cognitive resources are reallocated. Once the narrator's knowledge of goal plans of action, listener needs, and plot coherence is better established, the complexity of the narrative will rise ~ resulting in narratives like those in the Coherent and Complex category. Since changes in linguistic, discourse, and cognitive skills continue through the school-age years, this trade-off between coherence and complexity in narratives would likely continue as well, until a mature knowledge of language and narrative structure was attained. Summary The 3-year-olds appear to be a group in transition from producing narratives characterized by local goal plans relating to both a single global frog goal and nonfrog goals to having hierarchical goal 122 plans of action (two global goals). They were able to use their developing knowledge of narrative structure and discourse to selectively encode the pictured information into their narratives. The 4-year-olds' narratives were less variable, and their better understanding of narrative structure was apparent, particularly on the coherence measures. The majority were able to initiate, sustain and resolve both the superordinate and subordinate goal plans of the story. Children from both groups were able to follow divergent plot lines between the boy and his dog as well as between the primary goals of the story and local nonfrog episodes.. The disparity between the results obtained in this study and those reported in the literature was surprising in its magnitude. Because the coding schema was closely based on that developed by Trabasso et al. (1992) and Berman and Slobin (1994), it seems unlikely that the analysis method is the major factor responsible for the differences seen between the two groups. The children in the current sample appear to have more developed knowledge of goal plans of action and how to use these plans to structure their narratives in a coherent manner than those in the Renner and Marchman database. Data Comparison Considerations Several differences between this study and those previously published may have an effect on the interpretation and comparison of the data. These differences include: familiarity with the researcher and her expectations for the child's participation, individual children's approach to the task, story familiarity, and the effect of small sample size. Trabasso and Rodkin (1994) reported that many of the 3-year-olds in the Marchman and Renner database were extremely reluctant to tell the story and tended to treat the task as an interactive 123 conversation rather than storytelling., Renner reported that she needed to interview sixty 3-year-olds in order to obtain 12 usable narratives (1988, p. 83, as cited in Berman & Slobin, 1994, p. 59), and that the 3-year-olds' narratives "were almost exclusively conversational" (as cited in Berman & Slobin, 1994, p. 60). Little, if any, reluctance in performing the task was noted for either the 3-year-olds or the 4-year-olds in the present study. Of the children interviewed, all but one of their narratives were used in the analysis. The one child was eliminated solely because she could not participate in the full study. A possible reason for this discrepancy is the time taken by the experimenter to familiarize each child with both herself and the task. As mentioned in Chapter Two, several hours were spent with each child in the classroom so that he or she was comfortable with accompanying the researcher to the "reading room." As for their approach to the task, the children did use numerous rhetorical and interactive devices, but these did not appear to affect the content of the narratives to the extent noted by Renner. The majority of the narratives analysed in this study were second tellings of the story. Most of the other studies that have used the Berman et al. (1986) protocol do not indicate whether first or second tellings were analysed. It is possible that either the children's familiarity or the influence of the parent's story may have allowed them to perform better on some measures. Gomme (1994) noted that the 3-year-olds' second tellings tended to include more clauses relevant to the advancement of the plot. However, when scored for primary plot components in order to select the narrative to be analysed, the two stories showed only minor differences. Differences between the children used in this study and those used in the Berman and Slobin and Trabasso studies are to be expected simply because of the small sample sizes being used. The effect of sample size is particularly notable for children of this age, since they are in the process of acquiring the skills we are interested in measuring and comparing. We would therefore expect 124 more variable performance, both within individual subjects and among children grouped solely by age. An example of this variability can be seen in subject 6 (DA), who told a much shorter and less complex narrative as a 4-year-old than as a 3-year-old. If no longitudinal data had been available for this subject, I might have concluded that he had a lower level of goal plan knowledge. However, given that his narrative was relatively coherent and he was capable of encoding a more complex narrative his performance can be interpreted as the result of either changes in personal narrative style or the importance of narrative coherence, or both. Two final variables were the socio-economic status of the children - who were all middle to upper middle class, and their experience at the university's child study center preschool. Both of these factors may have had significant effects on the children's familiarity with stories and storytelling, as well as on their general level of development. Directions for Future Research. The results of this study have several methodological implications for future research into narrative development. The collection of longitudinal data was found to be a useful tool for verifying observed differences in the two groups performances, particularly on measures that showed substantial variability. The continued collection of longitudinal narrative data from these and other subjects will greatly facilitate the investigation of the developmental patterns observed in this study. Further collection of narratives from children from more diverse backgrounds will reduce the possible effects of socio-economic and cultural differences on the data. A second methodological implication is the relative value of the variables and analyses used. As mentioned throughout this thesis, some measures were found to more predicitive of narrative coherence or complexity than others. In particular, the length of the narrative and the number and 125 proportion of plot opening, plot unfolding, and plot concluding units were found to be too heavily dependant upon narrative style to be useful for assessing narrative competence. Some of the measures used in previous work, by Trabasso and his colleagues, and Berman and Slobin, were also found to be problematic. The explicit statement of the relationship between the boy and the frog was not well correlated with either goal plan knowledge, narrative coherence or narrative complexity. Similarly, Trabasso and Rodkin's (1994) essential story elements did not differentiate between the High, Middle,,and Low coherence narratives as they were intended to. As a result of these various difficulties, these measures can be eliminated from future investigations of narrative coherence and complexity. A final methodological issue is the use of "more" to investigate the development of narrative complexity. While complexity is, to some extent, based on encoding a higher number of various kinds of story elements, focussing solely on this aspect of the issue can obscure some interesting observations. This problem is particularly apparent when the increases in complexity are used to infer better goal plan knowledge. If a narrator includes a great deal of detail in her story, as seen for some of my subjects, then the narrative is complex at the level of individual story elements. However, it does not necessarily reflect strong goal plan knowledge if the narrator was unable to selectively encode the information in the story. For this reason, investigators must be cautious in their interpretations of "more" in their subjects' narratives. By now there is a considerable amount of information about the development of specific indicators of narrative complexity, such as goal plan knowledge, reference, causal relations, and temporal marking. A new goal of research should be to determine how individual children's narratives coordinate these parameters at different developmental times. Just as some of the children in this study appeared, to some extent, to sacrifice complexity for plot coherence, other aspects of telling 126 a "good" story may interact in interesting ways, so that no single system appears to develop in a simple linear manner. BIBLIOGRAPHY 127 Applebee, A., (1978). The Child's Concept of Story. Chicago: University of Chicago Press. Bamberg, M., (1987). The Acquisition of Narratives. Berlin: Mouton de Gruyter. Bamberg, M. & Marchman, V. (1990). What holds a narrative together? The linguistic encoding of episode boundaries. Papers in Pragmatics, 4, 58-121. Bennett-Kastor, T., (1983). Noun phrases and coherence in child narrative. Journal of Child Language, 10; 135-149. Berman, R., (1988). On the ability to relate events in narrative. Discourse Processes, 11, 469-497. Berman, R. & Slobin, D., (1994). Narrative structure. In R.A. Berman & D.I. Slobin (Eds.), Different Ways of Relating Events in Narrative: A Cross-linguistic Developmental Study. Hillsdale, NJ : Erlbaum. Berman, R., Slobin, D., Bamberg, M., Dromi, E., Marchman, V., Neeman, Y., Renner, T., & Sebastien, E., (1986). Coding Manual: Temporality in Discourse (rev. ed.). Berkeley: University of California, Cognitive Science Program. Black, J . & Bower, G. , (1980). Episodes as chunks in narrative memory. Journal of Verbal Learning and Verbal Behaviour, 18(3), 309-318. Botvin, G . & Sutton-Smith, B., (1977). The.development of structural complexity in children's . fantasy narratives. Developmental Psychology, 13, 377-388. Brighouse, J . , (1990). Coast Salish Children's Narratives. Unpublished M.Sc. Thesis. University of British Columbia. . Gi l lam, R., (1989). An Investigation of the Oral Language, Reading, and Written Language Competencies of Language-learning Impaired and Normally Achieving School-age Children, Unpublished Ph.D. Thesis, Indiana University. Gomme, N., (1994). Preschoolers' Anaphoric Reference. Unpublished M.Sc. Thesis, University of British Columbia. Grice, H., (1975). Logic and conversation. In P. Cole & J.L. Morgan (Eds.), Syntax and Semantics. New York: Academic. Hedberg, N. & Stoel-Gammon, C , (1986). Narrative analysis: clinical procedures. Topics in Language Disorders, 7, 58-69. Johnson, N. & Mandler, J . , (1980). A tale of two strucutures: underlying and surface forms in stories. Poetics, 9, 51-86. Karmiloff-Smith, A., (1983). Language development as a problem solving process. In Papers and Reports on Child Language Development. 128 Kemper, S. & Edwards, L , (1986). Children's expression of causality and their construction of narratives. Topics in Language Disorders, 7, 11-20. Liles, B., (1985). Narrative ability in normal and language disordered children. Journal of . Speech anaI Hearing Research, 28, 123-133. Liles, B., (1987). Episode organization and cohesive conjunctives in narratives of children with and without language disorder. Journal of Speech and Hearing Research, 30, 185-196. Mandler, J . , (1982). An analysis of story grammars. In F. Klix, J . Hoffman, & E. Vad der Meer (Eds.), Cognitive Research in Psychology (pp. 124-141). Amsterdam: North-Holland. Mayer, M., (1969). Frog, Where are You?. Hong Kong: South China Printing Co. Merritt, D. & Liles, B., (1989). Narrative analysis: clinical applications of story generation and story retelling. Journal of Speech and Hearing Disorders, 54, 438-447. Orsolini, M., (1990). Episodic strucutre in children's fantasy narratives: "Breakthrough" to decontextualised discourse. Language and Cognitive Processes, 5, 53-79. Pritcher, E. & Prelinger, E., (1963). Children Tell Stories. New York: International Universities Press. Propp, V., (1968). The Morphology of the Folktale. Austin: University of Texas Press. (Originally published in Russian in 1928). Ripich, D. & Griffin, P., (1988). Narrative abilities of children with learning disabilities and nondisabled children: Story strucutre, cohesion, and propositions. Journal of Learning Disabilities, 21,165-173. Schank, R. & Abelson, R., (1977). Scripts, plans, goals, and understanding. Hillsdale, NJ : Erlbaum. Seidman, S. , Nelson, K., & Gruendel, J . , (1986). Make believe scripts. In K. Nelson (Ed) , Event Knowledge: Structure and Function in Development. Hillsdale, NJ: Erlbaum. Stein, N. & Glenn, C , (1979). An analysis of story comprehension in elementary shool children. In R.O. Freedle (Ed.), New Directions in Discourse Processing. Norwood, NJ : Ablex. Trabasso, T. & Rodkin, P., (1994). Knowledge of Goal/Plans: A conceptual basis for narrating Frog, where are you?. In R.A. Berman & D.I. Slobin (Eds.), Different Ways of Relating Events in Narrative: A Cross-linguistic Developmental Study. Hillsdale, NJ : Erlbaum. Trabasso, T. & Nickels, M., (1992). The development of goal plans of action in the narration of a picture story. Discourse Processes, 15, 249-275. Trabasso, T., Stein, N., Rodkin, P., Park-Munger, M., & Baughn, C , (1992). Knowledge of goals and plans in the on-line narration of events. Cognitive Development, 7, 133-170. 129 Trabasso, T., van den Broek, P., & Suh, S., (1989). Logical necessity and transitivity of causal relations in stories. Discourse Processes, 12, 1-25. Umiker-Sebeoki J . , (1979). Preschool children's intraconversational narratives. Journal of Child Language, 6, 91-109. vanDijk, T., (1980). Story comprehension: an introduction. Poetics, 9, 1-25. Vygotsky, L., (1934). Mind in Society: The Development of Higher Psychological Processes. Cambridge, MA: Harvard University Press. 130. APPENDIX A Frog, Where Are You? (Mayer, 1969): Picture by Picture Description 1. Boy, dog, and frog are in bedroom; boy and dog are watching frog who is in the jar. 2. Boy and dog are asleep in bed; frog is stepping out of jar. 3. Boy and dog are awake and look at the empty jar from the end of the bed. 4. Boy looks in one of his boots; dog sticks his head in the jar. 5. Boy and dog are at the window; boy is calling and dog has his head stuck in the jar. 6. Dog is falling from window ledge; boy is watching him fall. 7. Boy is down on ground below window holding dog; dog is licking boy's face and there is broken glass on the ground. 8. Boy is calling towards forest; dog is sniffing at a line of bees coming from a hive at the edge of the forest. 9. Boy is calling in a hole in the ground; dog is barking at the bee hive. 10. Boy is holding his nose as if in pain; a little animal is at hole entrance; dog is leaning against bee hive tree. 11. Bee hive is on the ground and the bees are swarming out; boy is up a tree looking in a hole. 12. Boy is on his back on the ground; an owl is at the entrance; the bees are chasing the dog, who has run past the boy. 13. Boy is holding his hand above his head as if to fend off the owl flying above him; dog is at the bottom of a large rock. 14. Boy is calling from the top of the rock; he is leaning on some things behind the rock that look like brancjes. 15. Boy is on top of a deer's head; dog is almost entirely behind the rock where the deer is. 16. Deer is running towards a cliff with the boy on his head; dog is running beside deer, watching boy. 17. Deer stops at edge of cliff; boy and dog fall over cliff towards a body of water. 18. Boy and dog splash into water. 19. Boy is sitting in the water with the dog on his head; boy has his hand cupped to his ear. 131 20. Boy is leaning against a log and saying "SHH!" to the dog beside him. 21. Boy and dog look over the log. 22. Boy and dog are on top of log; two adult frogs are on other side. 23. Nine baby frogs join the adult frogs (who are looking like proud parents). 24. The frogs are on top of the log facing the water; boy and dog are walking through the water, away from the log; boy has one baby frog in his hand and is looking back and waving at the frog family. (Used with permission from Brighouse, 1990.) 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0099021/manifest

Comment

Related Items