Understanding the glue of narrative structure, childrens`s on- and off- line inferences about...

Cognitive Development 22 (2007) 323–340

Understanding the glue of narrative structure: Children’son- and off-line inferences about characters’ goals

Julie S. Lynch a,∗, Paul van den Broek b

a Department of Psychology, Saginaw Valley State University, University Center, MI 48710, United Statesb Department of Educational Psychology, University of Minnesota, United States

Abstract

Because characters’ goals play a key role in the structure of narratives, the ability to make inferencesabout goals is essential to narrative comprehension. Despite their importance, no previous studies haveexamined the process by which children make these goal inferences. In the current study, we examined 6-and 8-year-old children’s goal inference making processes through think-aloud protocols. We also examinedthe product of comprehension, the mental representation of text, through free recall and comprehensionquestions. The results revealed that children of both ages regularly made appropriate goal inferences whilelistening to narratives. In addition, the number of goal inferences predicted children’s recall of the stories.Thus, children as young as 6 years old are sensitive to the vital role of characters’ goals in narrative structure,and they can engage in sophisticated cognitive processing while they listen to narratives to form coherentmental representations of them.© 2007 Elsevier Inc. All rights reserved.

Keywords: Narrative comprehension; Discourse Processing

Characters’ goals provide an organizing principle for narratives. They are the “glue” that holdsevents together to make a coherent story (Bower & Rinck, 1998, p. 112). Goals provide reasonsfor a character’s actions throughout a narrative, and a character’s success or failure in achievinghis or her primary goal marks the natural conclusion of a story. Because of their centrality, readersor listeners of a narrative must be able to infer the connections between goals and other elementsof a story to fully comprehend it.

Despite its importance, there are few studies in which children’s ability to make goal inferenceshas been directly investigated. Moreover, these studies have used off-line measures of comprehen-sion. These measures probe children’s understanding only after they have read or heard a narrative,

∗ Corresponding author. Tel.: +1 989 964 4351.E-mail address: [email protected] (J.S. Lynch).

0885-2014/$ – see front matter © 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.cogdev.2007.02.002

mailto:[email protected]

dx.doi.org/10.1016/j.cogdev.2007.02.002

324 J.S. Lynch, P. van den Broek / Cognitive Development 22 (2007) 323–340

assessing if inferences are a part the product of comprehension, the mental representation of thetext. No previous studies, to our knowledge, have used on-line comprehension measures. Theseare measures that examine inference making while the comprehender listens to or reads a text,thus assessing the initial process of comprehension. Accordingly, the purpose of this investigationis to examine children’s developing ability to make inferences about characters’ goals on-line aswell as off-line. This will provide a more thorough understanding of children’s goal comprehen-sion, giving insight into how this sophisticated, yet essential, aspect of narrative comprehensiondevelops.

1. Narrative structure

Characters’ goals provide coherence both within and between narrative episodes. A character’sgoal is conceptually connected to most of the elements of a single episode (Mandler & Johnson,1977; Stein & Glenn, 1979). In a typical episode, the main character’s goal will motivate most ofhis or her actions, and these efforts will result in either success or failure in achieving the goal,marking the end of the episode. Furthermore, goals often provide coherence in multi-episodicnarratives by establishing a hierarchy among episodes (Johnson & Mandler, 1980). A charactermay fail to achieve his or her goal or may foresee difficulties in fulfilling the goal and then establisha subgoal in order to achieve the primary, or superordinate, goal. The subgoal, in turn, motivatesthe actions and eventually the outcome of another episode.

A typical hierarchical structure is illustrated in the narrative The Little Boy (refer to Table 1).In response to losing a pet frog, the protagonist’s superordinate goal becomes finding this frog(Event #13). This goal motivates his actions in the first episode of the story. When the little boyfails to achieve his main goal, he establishes a number of subgoals including wanting to see ifanything is in the tree (Event #28) and wanting to find the source of a sound he heard (Event #60).The boy achieving his superordinate goal, finding the frog, indicates the end of the narrative.

Key to narrative comprehension is the ability to make inferences to connect goals with actionswithin an episode and to connect hierarchically related goals with each other. Making such infer-ences on-line supports the construction of a coherent mental representation of text, a representationthat facilitates recall and learning (e.g., Langston & Trabasso, 1998; van den Broek, 1990). To makean inference linking two parts of narrative, such as a superordinate goal and subgoal, both mustbe in working memory simultaneously (e.g., Kintsch & van Dijk, 1978; van den Broek & Lorch,1993). Thus, goal inferences generally require recalling goals from previous points in the text ormaintaining a goal in working memory while reading, both potentially resource-demanding andeffortful processes. Furthermore, inferences linking subgoals and actions to superordinate goalsare often not necessary to maintain a sense of local coherence in a narrative, an understanding ofthe most immediate causes of events. Thus, it has been suggested that even adult readers may notregularly make goal inferences on-line (McKoon & Ratcliff, 1992).

1.1. The development of goal comprehension and inference making

Despite the potential challenges to making these inferences, there is consistent empirical evi-dence that adult readers regularly make goal inferences on-line (Albrecht & Myers, 1995; Albrecht& O’Brien, 1995; Bloom, Fletcher, van den Broek, Reitz, & Shapiro, 1990; Dopkins, Klin, &Myers, 1993; Graesser, Burtus, & Magliano, 1995; Long & Golding, 1993; Long, Golding, &Graesser, 1992; Suh & Trabasso, 1993; van den Broek & Lorch, 1993). As a result, goals arecentral in mature comprehenders’ mental representations of narratives. They recall characters’

J.S. Lynch, P. van den Broek / Cognitive Development 22 (2007) 323–340 325

Table 1Goal-directed version of The Little Boy narrative

Event no. Text

1 Once there was a boy2 The boy had a pet dog and a pet frog.3 The boy loved to be with his pets.4 He especially liked to play with them in the woods on sunny days.5 The dog even slept on the bed with the boy6 while the frog slept in a jar right by them.7 One night while the boy and his dog were sleeping8 his frog climbed out of its jar9 and ran away

10 But the boy and the dog didn’t wake up.11 The next morning the boy couldn’t find his frog.12 The boy was very worried.13 He wanted to find his frog.14 He looked all around his room.15 He looked under the bed and on top of the dresser.16 The dog sniffed in the closet and toy chest.

Probe Point 117 Finally, the boy looked inside his boots.18 but his frog wasn’t there.19 He didn’t know where else in the room to look.20 The boy tossed his boots under the bed21 and put on his shoes.22 The boy and his dog went outside to look in their yard.23 The sun was shining bright and24 it was very hot outside.25 The boy looked all around the yard26 He looked around the sandbox and the swing set.27 Then he came to a big, old tree.28 He wanted to see if anything was in the tree.

Probe Point 229 The boy climbed up onto the biggest branch30 and looked into a big hole in the tree.31 He peered down into the darkness.32 Suddenly, an owl flew out of the hole in the tree!33 The boy was so frightened,34 that he fell out of the tree.35 He got up quickly and36 he and the dog ran away from the owl.37 They ran a long way out of the yard and into the woods.38 Finally, they stopped to catch their breath.39 and the boy patted his dog on the head.40 Thankfully, the owl had stopped following them.41 They had come to a big rock42 The boy wanted to keep looking at what was around him,43 so he climbed onto the rock

Probe Point 344 and, for balance, leaned against a branch.45 But then the branch began to move,46 The branch was really a deer!47 The deer stood up and


Table 1 (Continued )

Event no. Text

48 lifted the boy onto its head, into the air.49 The deer was surprised.50 It had never had a little boy on its head before,51 and so it ran into the woods with the boy on its head.

Probe Point 452 The boy was so frightened he closed his eyes53 and held on for dear life.54 His loyal dog followed behind.55 Suddenly, the deer stopped at a high cliff and56 sent the boy sailing through the air.57 The boy and his dog landed in a pond.58 They were now very wet and muddy.59 Suddenly, the boy heard the sound of an animal.60 The boy wanted to find out what was making that sound.

Probe Point 561 He told his dog to listen62 and he sat very still to try to hear the noise again.63 After a little while, the boy began to look around him.64 He searched through the cattails and lily pads,65 but he didn’t see anything unusual.66 Then, the noise came again.67 The boy and dog realized that it was coming from behind a log.68 The dog, who was walking right behind the boy, stopped69 and sniffed at the log.70 The boy climbed over the log and71 there sat his frog.72 He croaked and smiled at the boy and dog.73 The boy was very excited to find the frog.

Probe Point 674 The boy and his dog took the frog home with them.75 Now, they all could play together again.

goals more often and rate goals as more important than other aspects of the narrative structure,such as characters’ actions (Mandler & Johnson, 1977; Goldman & Varnhagen, 1986; van denBroek, Lorch, & Thurlow, 1996). Further, they remember superordinate goals and episodes betterthan more subordinate ones (Black & Bower, 1980; Goldman & Varnhagen, 1986; Kintsch, 1977;Meyer & McConkie, 1973).

There is mixed evidence regarding the extent to which children incorporate goals into theirmental representations of narratives and regarding their understanding of the pivotal role of goalsin narrative structure. On the one hand, children as young as 3 and 4 years old are often able topoint to an appropriate goal when asked direct questions about the reasons for characters’ actions(Trabasso, Stein, Rodkin, Munger, & Baugh, 1992; Wenner, 2004). Preschoolers also show theirsensitivity to goals by remembering narratives with coherent, hierarchical goal structures to agreater extent than narratives without coherent structures, particularly when the narratives arerelatively long and challenging. In these cases especially, the goal structure may provide thesupport young children need to build an adequate mental representation of the narrative (Wenner,2004). By the time children are 5 years old, they are able to tell stories with some elements of


the goal structure including explicit goal statements, attempts directed towards fulfilling goals,and goal outcomes (Trabasso & Nickels, 1992; Trabasso & Stein, 1997; Trabasso et al., 1992). Inaddition, children as young as 8 years old are able to identify the superordinate goal, as opposedto subgoals, as the best statement to capture the main idea or theme of a narrative (van den Broek,Lynch, Naslund, Ievers-Landis, & Verduin, 2003).

On the other hand, there is evidence that children’s understanding of goals in narratives islimited in the early elementary school years. Unlike adults, children do not tend to include goalswhen freely recalling narratives (Stein & Glenn, 1979; Trabasso et al., 1984; van den Broek et al.,1996). Rather than recalling internal states of characters, such as goals, elementary-school-agedchildren tend to spontaneously recall more concrete aspects of a narrative such as actions (e.g.,Goldman & Varnhagen, 1986; Stein & Glenn, 1979; Trabasso et al., 1992; van den Broek et al.,1996). In addition, younger children may not be fully sensitive to hierarchical relations amonggoals and episodes. Eight-year-old readers’ ratings of statement importance are not influenced byhierarchical connections, whereas those of children 11 years old and older are (van den Broek,1989). Furthermore, 8-year-old children do not consistently recall statements on higher levels ofthe narrative structure better than statements on lower levels, as 10-year-old children and adultsdo (Goldman & Varnhagen, 1986; cf., van den Broek et al., 1996).

In contrast to their off-line comprehension, children’s on-line comprehension of goals hasbeen left unexplored. In fact, children’s on-line processes in general are only beginning to beinvestigated. The limited evidence we have suggests that children are actively engaged in a varietyof comprehension processes as they read, hear, or watch narratives. Children as young as 9 years oldmake at least some types of inferences on-line, such as those that connect causes to their immediateeffects (Casteel, 1993; Danner & Mathews, 1980). Children of this age also regularly engage inother important comprehension processes on-line, such as making predictions and maintainingrelevant information in working memory (Trabasso & Magliano, 1996; Trabasso, Suh, Payton, &Jain, 1995). Importantly, there is also evidence that children’s on-line comprehension processesare related to their off-line representations of a narrative, as is predicted by theoretical accountsof narrative comprehension (i.e., Langston & Trabasso, 1998; van den Broek, 1990). The numberof causal inferences and predictions children make on-line positively correlate with the numberof sentences they recall from a narrative (Trabasso & Magliano, 1996; Trabasso et al., 1995).

1.2. Current study

The current study focuses on goal inference making during the early elementary-school years(at 6 and 8 years old), a potential time of significant growth in goal inference making skills. Six-year-old children have a basic understanding of goals, as shown by their performance in a numberof on-line tasks (e.g., Goldman & Varnhagen, 1986; Trabasso & Nickels, 1992). Although theirunderstanding of goals may still be limited, by the age of 8 years they seem to be on the vergeof a more sophisticated understanding of goals (e.g., van den Broek, 1989). In choosing this agegroup, we are also extending downward the research on specific on-line comprehension processeswhich have rarely been explored in children younger than 9 years old.

The narratives used in this study are of central importance to its purpose. Two versions of twonarratives were constructed. One, the goal-directed version (see Table 1), has a typical hierar-chical goal structure in which a superordinate goal motivated several subordinate goals and theirrespective episodes. The other version, the non-goal-directed version (see Table 2), incorporatesthe same goals and episodes as the goal-directed narrative, but the goals are not hierarchicallyrelated. Instead, one goal is fulfilled before the other one is undertaken, creating a structure in


Table 2Non-goal-directed version of The Little Boy narrative

Event no. Text

1 Once there was a boy2 The boy had a pet dog and a pet frog.3 The boy loved to be with his pets.4 He especially liked to play with them in the woods on sunny days.5 The dog even slept on the bed with the boy6 while the frog slept in a jar right by them.7 One night while the boy and his dog were sleeping8 his frog climbed out of its jar9 and ran away

10 But the boy and the dog didn’t wake up.11 The next morning the boy couldn’t find his frog.12 The boy was very worried.13 He wanted to find his frog.14 He looked all around his room.15 He looked under the bed and on top of the dresser.16 The dog sniffed in the closet and toy chest.

Probe Point 117 Finally, the boy looked inside his boots.18 There he found his frog19 and he put the frog safely back in the jar.20 The boy tossed his boots under the bed21 and put on his shoes.22 The boy and his dog went outside to look in their yard.23 The sun was shining bright and24 it was very hot outside.25 The boy looked all around the yard26 He looked around the sandbox and the swing set.27 Then he came to a big, old tree.28 He wanted to see if anything was in the tree.

Probe Point 229 The boy climbed up onto the biggest branch30 and looked into a big hole in the tree.31 He peered down into the darkness.32 Suddenly, an owl flew out of the hole in the tree!33 The boy was so frightened,34 that he fell out of the tree.35 He got up quickly and36 he and the dog ran away from the owl.37 They ran a long way out of the yard and into the woods.38 Finally, they stopped to catch their breath.39 and the boy patted his dog on the head.40 Thankfully, the owl had stopped following them.41 They had come to a big rock42 The boy wanted to keep looking at what was around him,43 so he climbed onto the rock

Probe Point 344 and, for balance, leaned against a branch.45 But then the branch began to move,46 The branch was really a deer!47 The deer stood up and48 lifted the boy onto its head, into the air.


Table 2 (Continued )

Event no. Text

49 The deer was surprised.50 It had never had a little boy on its head before,51 and so it ran into the woods with the boy on its head.

Probe Point 452 The boy was so frightened he closed his eyes53 and held on for dear life.54 His loyal dog followed behind.55 Suddenly, the deer stopped at a high cliff and56 sent the boy sailing through the air.57 The boy and his dog landed in a pond.58 They were now very wet and muddy.59 Suddenly, the boy heard the sound of an animal.60 The boy wanted to find out what was making that sound.

Probe Point 561 He told his dog to listen62 and he sat very still to try to hear the noise again.63 After a little while, the boy began to look around him.64 He searched through the cattails and lily pads,65 but he didn’t see anything unusual.66 Then, the noise came again.67 The boy and dog realized that it was coming from behind a log.68 The dog, who was walking right behind the boy, stopped69 and sniffed at the log.70 The boy climbed over the log and71 there sat a duck.72 He quacked and smiled at the boy and dog.73 The boy was very excited to find the duck.

Probe Point 674 The boy and his dog took the duck home with them.75 Now, they all could play together.

which the goals are not causally related to each other. Because the content of the two versionsof each narrative is almost identical, any differences in comprehension between the two versionsshould be due to the goal structure rather than the specific content of the narratives (Goldman &Varnhagen, 1986; Suh & Trabasso, 1993; Wenner, 2004).

We explored goal inference making through the use of think-aloud protocols. The childrenwere asked to verbalize their thoughts at certain points as they listened to the narrative. Becauseno specific questions or prompts are provided in a think-aloud protocol, the responses representauthentic reflections of the participants’ current mental processes. This method has been shown toprovide valid evidence about a variety of conscious, effortful mental processes (Ericson & Simon,1994; Magliano, 1999), including inference making (e.g., Long & Bourg, 1996; Whitney & Budd,1996). Although children as young as 9 years old can successfully complete think-aloud protocols(Trabasso & Magliano, 1996; Trabasso et al., 1995), to our knowledge think-aloud protocolshave not been used previously with children as young as 6 years old in studies of narrativecomprehension. Therefore, a secondary purpose of this study is to ascertain if a think-aloudprocedure can be a useful measure of younger children’s comprehension processes.


We examine four specific hypotheses in this study. First, we predict that children will makeappropriate on-line goal inferences, as evidenced by making more goal inferences while thinkingaloud in the goal-directed version than in the non-goal-directed version of the narratives. Thiswould indicate children in early elementary school make these resource-demanding inferencesspontaneously as they build a mental representation of the narrative. Consequently, our secondhypothesis is that children will incorporate goals into their mental representations of narratives, asshown in their off-line comprehension. Specifically, we predict children will correctly answer off-line questions in the goal-directed narratives, and that they will recall the goal-directed narrativesto a greater degree than the non-goal directed narratives. Third, given the complex nature ofgoal inferences, we expect that on-line comprehension of goals will develop through the earlyelementary-school years. Relative to the younger children, we predict that the older children willmake on-line goal inferences more consistently as well as recall more of the narratives and answermore question correctly. Finally, we expect that the number of goal inferences a child makes in thegoal-directed narratives will predict the number of statements the child recalls from that narrative,providing evidence that these inferences contribute to forming a coherent representation of thetext.

2. Method

2.1. Participants

Twenty-five 6-year-old children and twenty-four 8-year-old children participated in the study.The data from one 6-year-old girl were eliminated because she had been exposed to one of thenarratives before the session. The data from one 8-year-old boy was also eliminated because hefailed to complete the tasks in the session. Thus, the final sample included twenty-four 6-year-olds (11 female) (mean age = 6.71 years, range: 6.38–6.81 years) and twenty-three 8-year-olds(11 female) (mean age = 8.57 years, range: 8.19–9.03 years). The participants were recruitedfrom a pool of families who had previously volunteered for research studies. The children wereprimarily from middle-class homes. Each child received a small, age appropriate toy for theirparticipation.

2.2. Materials

Two narratives were used. One narrative, The Little Boy, was based on the book, Frog WhereAre You? (Mayer, 1969, adapted by Wenner, 1999, 2004). The goal and non-goal-directed versionsof this narrative, as described previously, are found in Tables 1 and 2 respectively. The secondnarrative, The Little Girl (Murachver, Pipe, Gordon, Owens, & Fivish, 1996, adapted by Wenner,1999, 2004) was the same length as The Little Boy, and had a parallel goal structure, with thecharacter’s main goal being to find a lost ball. A non-goal-directed version of this narrative wasconstructed in the same manner as described previously.

Based on the goal structure of these narratives, we choose six points at which to probe forgoal inference making, particularly inferences about the characters’ superordinate goal (refer toTables 1 and 2). Examining children’s goal inferences at these six points in both narratives allowedus to trace the patterns of children’s inference making. Probe Point 1 was chosen as a place wherea superordinate goal inference could be made in both versions. At the five remaining points, weexpected goal inferences to be made in the goal-directed but not in the non-goal-directed version.The possible exception was Probe Point 4 where a goal inference was not required to maintain a


Table 3Off-line questions for The Little Boy narrative

1. Why was the boy very worried at the beginning of the story? (Plot)2. What did the boy do to his boots after looking in them? (Plot)3. Why did the boy look into the hole in the tree? (Action)4. Why did the boy want to see if anything was in the tree? (Subgoal)5. Why did the boy want to climb on the rock and see around him? (Subgoal)6. How did the boy and his dog get into the water? (Plot)7. Why did the boy tell his dog to listen while they were in the pond? (Action)8. Why did the boy want to find out what the sound in the pond was? (Subgoal)9. What did the boy find behind the log? (Plot)

sense of coherence in either version. In the goal-directed version, however, a goal inference couldbe made at this point if the child was maintaining the superordinate goal in working memory.

Nine off-line comprehension questions were written for each story, and correct answers weredetermined based on the causal structure. The off-line questions for The Little Boy are listed inTable 3. Among these questions, three directly addressed reasons for characters subordinate goals(subgoal questions). In the goal-directed version of the narratives, the correct answer to thesequestions was the superordinate goal. In the non-goal-directed version, there was no clear answerbased on the narrative structure. Two questions addressed causes of characters’ actions (actionquestions). In the goal-directed version, these questions could be answered by pointing to eitherthe superordinate goal or the most immediate subordinate goal. In the non-goal-directed version,the correct answer was the most immediate “subordinate” goal.1 The four remaining questionswere about basic plot information (plot questions). These questions addressed actions or eventsthat were not directly motivated by the goal, and the correct answers to these questions were thesame in the goal- and non-goal-directed versions.

A series of simple line-drawn pictures was made to accompany each narrative. Copies of apicture of a cartoon character thinking (i.e., having a “thought-bubble” over his head) were placedamong the story pictures as cues for the think-aloud task. As part of their training, children listenedto a short, nine-line narrative, The King Story (Mandler & Johnson, 1977).

To examine any influence of general verbal ability, the children completed the Peabody PictureVocabulary Test–Revised (PPVT-R) (Dunn & Dunn, 1997). The children also completed a measureof working memory (Siegel, 1994), which was not related to the hypotheses addressed here.

2.3. Procedure

Each child was seen individually in a university laboratory setting by the first author. First, halfof the children completed the PPVT and the other half completed the working memory measure.The children were then introduced to the think-aloud task.

Because think-aloud protocols had not been used previously with children this young, weused a thorough training procedure to help the children understand the task. To introduce thethink-aloud task, each child was shown the picture of the cartoon character with the “thoughtbubble” and asked what the character was doing. After the child responded that the character was

1 These goals were the subordinate goals from the goal-directed version. However, because the non-goal-directed versiondid not have a hierarchical structure, they are not technically subordinate in that version.


thinking,2 the experimenter read the following instructions:

Yes, this person is thinking, just like we are always thinking. Usually we just think insideour heads without saying what we are thinking. But today, we are going to play a gamewhere you get to say just what you are thinking. We are going to listen to some stories, andevery time we see this man, we are going to stop and you will say what you are thinking.Let’s try.

The children then practiced the procedure with the short training story. There were three pointsat which the children were prompted to think aloud in this story. If a child did not respond, aslightly more specific question was asked (“What are you thinking about [the main character]?”).If the child did not respond to the more specific prompt, the experimenter modeled thinking-aloud.Importantly, the modeling did not include a goal inference. All of the children produced a responseto the think-aloud task at one or more points in the practice story.

After the training procedure, the children completed the think-aloud task with one of theexperimental narratives. The order of the narrative (The Little Boy and The Little Girl) as well asof the version (goal- and non-goal-directed) was counterbalanced with the restriction that eachchild listened to both narratives, one in the goal-directed version and one in the non-goal-directedversion. The stories were presented on a cassette tape. At the six predetermined probe points,the children saw the cartoon character and were prompted to say aloud what they were thinking.After the children finished expressing their thoughts, the experimenter gave non-specific prompts(“Is there anything else you’re thinking?”) until the children indicated that they were done. If achild did not respond to the first prompt, the experimenter asked a more specific question (“Whatare you thinking about the [main character]?”). If the child did not respond to the more specificprompt, the story was continued.

When the story was over, the children were immediately asked to recall the story. Generalprompts were used until the children indicated they could not remember any additional informa-tion. The children then completed the comprehension questions. The children were given a shortbreak before completing the remaining tasks, either the PPVT or the working memory measure(whichever was not completed earlier) and the other narrative with the accompanying think-aloud,free recall and question tasks. The experimental sessions lasted between 45 and 75 minutes. Afterthe sessions, the think-aloud protocols as well as the children’s free recall of the narratives andanswers to questions were transcribed verbatim from videotapes and audiotapes.

2.4. Coding

2.4.1. Think-aloud protocolsThe purpose of the think-aloud coding scheme was to identify goal inferences, particularly

those involving the superordinate goals. Based on previous research (Suh & Trabasso, 1993),goal inferences were identified as either a statement that included an explicit statement of thesuperordinate goal (e.g., “He is trying to find the frog” or “She wants to get her ball back”) or astatement that contained an intentional action in connection to the goal object (e.g., “I think shecould borrow a ball”).

2 If the children said that the character was doing something other than thinking, such as reading, the experimenterasked if the character was doing anything else. After this second prompt, all of the children were able to answer that thecharacter was thinking.


The think-aloud protocols were parsed into verb phrases and coded by two raters. Twentypercent of the experimental sessions were coded by both raters to refine the scheme. An additional20% of the sessions were coded by both raters and were used to measure inter-rater agreement.The overall inter-rater agreement was 94% (range: 87.5–100%). Any disagreements between theraters were resolved through discussion.

2.4.2. Free recall and answers to questionsChildren’s free recall protocols were parsed into verb phrases. The quantity of recall was

measured by adding the statements each child made that directly matched an event in the storyand the statements that were a direct inference or summary based on the narrative.

Children’s answers to off-line questions, whether in response to the goal-directed or non-goal-directed version, were coded according to the “correct” answers in the goal-directed narrative.Specifically, the children’s answers to the subgoal questions (i.e., those questions addressingreasons for subgoals) were coded as correct if the child mentioned the superordinate goal. Actionquestions were coded for either mentions of an appropriate subgoal or of the superordinate goal.Plot questions were coded as correct or incorrect based on the basic plot information.

The two raters who coded the think-aloud protocols also coded the free recall and answersto questions. Again, both raters coded 20% of the sessions together. Another 20% were used tocalculate inter-rater agreement. Inter-rater agreement was 85% (range: 63.2–100%) for the free-recall protocols3 and 95% (range: 83.3–100%) for the answers to questions. All disagreementswere resolved through discussion.

3. Results

Preliminary analyses revealed that there were no main effects of presentation order of thenarratives and no interactions with this variable. Effect sizes (η2) are reported for each significantfinding.

3.1. On-line inference making

3.1.1. Total number of goal inferencesWe first examined the influence of goal structure, age and probe point on the total number of

goal inference children made. We conducted a 2 (version: goal-directed, non-goal-directed) × 2(age: 6 years, 8 years) × 6 (probe point) mixed analyses of variance (ANOVA) with version andprobe points as a within subjects variables. There was a significant main effect of story version,F(1, 45) = 18.14, p < .001, η2 = .287. As predicted, the children produced more goal inferenceswhile thinking aloud during the goal-directed version of the narrative (M = 1.87, S.D. = 1.71,median = 1.0, range: 0–6) than during the non-goal-directed version (M = .787, S.D. = .999,median = 1.0, range: 0–4).

There was no significant main effect of age on the total number of goal inferences or anysignificant interactions with this variable. Thus, children of both ages showed evidence ofappropriate goal inference making, drawing more goal inferences in the goal-directed version

3 Inter-rater agreement for free recall protocols indicates the rate at which the coders agreed on the exact statement fromthe narrative the child’s utterance matched or agreed if the statement was an inference or summary. It was not based onagreement on the overall quantity of recall.


Fig. 1. Mean number of goal inferences made at each probe point in the goal-directed and non-goal-directed versions ofthe narratives.

than the non-goal-directed version of the narratives. There was a main effect of probe point,F(5, 225) = 7.20, p < .001, η2 = .138. The interaction of version and probe point was marginallysignificant, F(5, 225) = 2.21, p = .055, η2 = .047.

3.1.2. Goal inference patternsBecause the different patterns of goal inferences in the two versions across probe points were of

theoretical interest (rather than the main effect of probe points), we conducted follow-up analyseson the interaction of version and probe point. Because we had a priori hypotheses, one-tailedmatched-pairs t-tests with an appropriate Bonferroni adjustment were conducted to examine thedifferences between the two versions at each probe point.

The mean numbers of goal inferences at each probe point in both versions are shown in Fig. 1.There were no significant differences between the goal-directed and non-goal-directed versionsat Probe Point 1, as expected. The differences between version reached statistical significance atProbe Point 4, t(46) = 2.87, p = .003 and Probe Point 6, t(46) = 3.92, p < .001. At Probe Point 4,children made an average of .234 (S.D. = .56) goal inferences in the goal-directed version andno goal inferences in the non-goal-directed version. At Probe Point 6, children made an averageof .511 (S.D. = .688) goal inferences in the goal-directed version and .085 (S.D. = .351) goalinferences in the non-goal-directed version.

3.2. Off-line measures of comprehension

Data from one 6-year-old child had to be eliminated from the analyses of off-line comprehen-sion (comprehension questions and recall) because his verbal responses were unintelligible onvideo and cassette tapes. For the same reason, data from another 6-year-old child were eliminatedfrom some of the analyses.

3.2.1. QuestionsA 2 (version: goal-directed, non-goal-directed) × 2 (age: 6 years, 8 years) mixed ANOVA with

version as a within subjects variable was conducted for each of the three types of questions. Therewere no main effects of age or interactions between version and age in any of these analyses.


The first type of question addressed reasons for subgoals. As expected, children cited the super-ordinate goal as an answer to the subgoal questions for the goal-directed versions (M = 84.14%,S.D. = 21.70%) more than for the non-goal-directed versions (M = 12.96%, S.D. = 27.35%),F(1, 43) = 177.92, p < .001, η2 = .805.

In response to the second type of question, action questions, children could logically point toeither the most recent subgoal or the superordinate goal for the goal-directed versions, whereasthey could only logically point to the most immediate subgoal for the non-goal-directed versions.Superordinate goal and subgoal answers to action questions were analyzed in separate ANOVAs.Children answered action questions with the superordinate goal more often for the goal-directedversions (M = 55.56%, S.D. = 36.47%) than for the non-goal-directed versions (M = 10.64%,S.D. = 25.40%), F(1, 44) = 52.83, p < .001, η2 = .546. Children answered these questions withthe subgoals more often for the non-goal-directed versions (M = 50.00%, S.D. = 39.00%) thanfor the goal-directed versions (M = 30.42%, S.D. = 33.75%), F(1, 43) = 7.97, p = .007, η2 = .153.Therefore, children’s answers to action questions conformed to the narrative structure.

Children’s answers to plot questions did not differ between the goal-directed (M = 91.22%,S.D. = 14.02%) and non-goal-directed (M = 89.17%, S.D. = 13.45%) versions of the story. Thus, asexpected, there were no significant differences in the understanding non-goal-related informationbetween the two versions.

3.2.2. Free recallA 2 (version: goal-directed, non-goal directed) × 2 (age: 6 years, 8 years) mixed ANOVA with

version as a within subjects variable was conducted on the children’s recall of the narratives.There were no significant differences between the goal-directed (M = 13.89, S.D. = 11.38) andnon-goal-directed versions (M = 15.46, S.D. = 10.44), no differences between age groups and nointeraction between age and version. Thus, contrary to predictions, the children did not recallmore of the goal-directed versions than the non-goal-directed versions.

3.3. Relation between on- and off-line processes

We conducted a regression analysis to examine the relation between number of goal inferenceschildren made while thinking aloud about the goal-directed versions and their recall of thoseversions. To control for any influence of verbal skill, the children’s vocabulary scores were enteredinto the regression first. Even after the influence of verbal skill was taken into account, the numberof goal inferences significantly predicted recall, F change (1, 43) = 5.160, p = .028, also refer toTable 4. We conducted a parallel analysis for the non-goal-directed versions. Interestingly, thenumber of goal inferences made in the non-goal-directed versions also significantly predicted

Table 4Summary of regression analysis for variables predicting recall of the goal-directed narrative

Variable B S.E. B β

Model 1PPVT .189 .084 .320

Model 2PPVT .178 .081 .301Number of goal inferences 2.076 .914 .311


Table 5Summary of regression analysis for variables predicting recall of the non-goal-directed narrative

Variable B S.E. B �

Model 1PPVT .006 .080 .118

Model 2PPVT .007 .077 .130Number of goal inferences 3.355 1.483 .321

children’s recall of that narrative after the influence of verbal skill was taken into account, Fchange (1, 44) = 5.12, p = .029, also refer to Table 5.

3.4. Performance on the think-aloud task

A secondary purpose of this study was to investigate if think-aloud protocols are a viablemeasure of younger children’s on-line comprehension processes. Think-aloud protocols wouldnot be a feasible method if young children generated few or no responses. Therefore, we comparedthe 6- and 8-year-old children’s quantity and rate of responses to the think aloud task. Overall, thechildren produced an average of 2.40 utterances for every probe point. A t-test revealed that therewere no significant differences between the age groups. The children provided some response atalmost every probe point (94.2%), with no significant difference between the two age groups intheir response rate. These findings suggest that both the younger children and their older peersresponded well to the task.

4. Discussion

The primary purpose of this study was to explore children’s developing ability to make goalinferences while comprehending narrative texts. In support of our first hypothesis, significantdifferences in goal inference making between the goal- and non-goal-directed narrative versionsdemonstrated that children made appropriate goal inferences as they proceeded through the nar-rative. In the goal-directed narratives, even the younger children regularly made inferences toconnect characters’ subgoals and actions to superordinate goals and, thus, establish coherencewithin and between episodes. To our knowledge, this is the first demonstration that children makegoal inferences on-line. Although evidence from off-line measures suggested that children cometo include goals in their representations of text, the process by which children make those infer-ences was not known. It was not clear whether they regularly make goal inferences spontaneouslywhile listening to a story or only later, when prompted to recall the narrative or answer questionsabout it.

The texts used in this study were relatively coherent, tightly connected around a limited set ofgoals. As a result of this and of the fact that subjects were not specifically asked to generate goalinferences, the frequency of goal inferences was not very high at any given point and the differencein their generation was not significant at all probe points. There were significant differences at twoof the predicted probe points, including the point at which a goal inference was not necessary tomaintain local coherence in the narrative (Probe Point 4), indicating that children could maintainthe main goal in working memory as they proceeded through the narrative.


In support of our second hypothesis, the children’s responses to the comprehension questionsconfirmed that they built mental representations of narratives that include superordinate goals.They generally pointed to the superordinate goal as the reason for actions in the goal-directedversions of the narratives rather than a more immediate subgoal. This mirrors previous findingswith adult readers (Long & Golding, 1993) and suggests that the children had a firm understandingof the importance of the superordinate goal in these narratives. However, there were no differencesin the overall amount of recall of the goal- and non-goal-directed narratives. Although this findingwent against our predictions, it is consistent with other studies that have found no differences inrecall of story versions when texts are relatively short (Goldman & Varnhagen, 1986; Wenner,1999, 2004). The children in this study were likely able to build a sufficient representation of thenon-goal-directed story to support immediate recall, even without the aid of a coherent narrativestructure. To further examine the influence of goal structure on recall, future research couldexamine both immediate and delayed recall as well as the overall quality of recall.

Contrary to the prediction in our third hypothesis, there were no overall differences betweenage groups in this experiment. The fact that the younger children performed no differently fromolder children speaks to the impressive narrative comprehension abilities of the younger children,and indicates that the origins of goal inference making are to be found earlier in development.The findings do not imply that goal comprehension abilities remain static in the early elementaryschool years. Given different texts or tasks, developmental trends within this age range may wellbe uncovered.

As predicted in our last hypothesis, there was a significant relation between the number ofon-line goal inferences made in the goal-directed versions and recall of that narrative, and thisrelation held above and beyond the influence of basic verbal skill. This points to the importance ofgoal inferences in narrative comprehension processes, as goal inferences alone predicted narrativerecall. It also extends downward theoretical concepts of comprehension processes. On- and off-line comprehension processes are integrated even early in their development, forming a coherentoverall process of active understanding.

Although we intended to use the non-goal-directed narrative simply for experimental control,the children’s responses to these narratives proved to be informative. First it was surprisingthat the number of goal inferences made predicted recall in this narrative. This is consistentwith evidence from this study and others that children are cognitively active as they listen tonarrative. Goal inferences, though technically incorrect, may have been attempts to form coherentrepresentations of the non-goal-directed narratives. Therefore, children who made more of theseinferences also may have constructed a more complete mental representation of the narrative that,in turn, supported greater recall.

Second, in their responses to the non-goal-directed narratives, the children showed that theyunderstood that stories “should” include a superordinate goal. During the think-aloud protocol,more than a quarter of the children expressed some type of confusion about the characters’ moti-vations in this narrative and another quarter inferred that the protagonist did have a superordinategoal although the text did not support that inference. Some of these children assumed that themain character had not really found the goal object (the frog or ball) and others made up goalsto explain the character’s actions. Similarly, in response to the questions about subgoals in thesenarratives, more than half of the children inferred connections to some type of primary goal.

From a methodological point of view, the current study indicated that children as young as 6years old are very capable of performing a think-aloud task, whereas previous text comprehensionstudies had not used this measure in children younger than 9 years old. Children in both age groupsin the current study made meaningful responses at almost every probe point. This is encouraging


in that it suggests that this method can be used effectively with young children. It is a potentiallyversatile and powerful method of examining the content of mental processing for both researchand instructional purposes (e.g., Loxterman, Beck, & McKeown, 1994; Magliano & Millis, 2003;McNamara, 2004; Trabasso & Magliano, 1996).

Ideally, think-aloud protocols would not be relied upon as the sole measure of on-line pro-cessing. Think-aloud protocols may not be a “pure” on-line measure because they allow timefor comprehenders to reflect on their thoughts. Therefore they should be verified with other on-line measures, such as reading time (e.g., Magliano, 1999; Magliano & Graesser, 1991; van denBroek, Fletcher, & Risden, 1993). Because most other measures of on-line comprehension relyon reading fluency, they have generally not been used with younger age groups. However, futurestudies should examine if these measures can be modified for use with younger children.

In addition, future studies should use narratives that vary in length and complexity. The use ofqualitatively different narratives would likely reveal more detail about the development of goalinference making. Future research should also explore possible mechanisms of development ingoal inference making, such as the role of literacy experience.

This study’s findings not only provide valuable theoretical insight into children’s developingcomprehension processes, but may also have practical implications for education. The ability tomake inferences is a critical factor in reading comprehension. In fact, young children’s narrativecomprehension processes, including their inference making, predict later reading comprehensionabilities (e.g., Paris & Paris, 2003; Kremer, Lynch, Kendeou, Butler, & van den Broek, 2002).Given this and the fact that characters’ goals are particularly important to the overall narrativestructure, goal inferences should be considered in early reading instruction. The results of thisstudy reveal that children as young as 6 years old have the ability to make these inferences, butit is likely that this ability could be further developed in terms of the range of circumstances inwhich they are made and automaticity of the process. Through a variety of means, including theeducational use of think-aloud protocols, instructors of young children can help them make goalinferences consistently in stories they hear, preparing them to fully grasp the structure of storiesthey will eventually read themselves (refer to Kendeou et al., 2006).

Acknowledgements

This research was supported by grants from the Center for the Improvement of ReadingAchievement (CIERA), the Center for Cognitive Sciences, University of Minnesota, and fromthe National Institute of Child Health and Human Development (HD-07151). Paul van den Broekreceived support through the Golestan and Lorentz fellowships from the Netherlands Institutefor Advanced Study and from the Guy Bond Endowment for Reading Research, University ofMinnesota.

References

Albrecht, J. E., & Myers, J. L. (1995). The role of context in accessing distant information during reading. Journal ofExperimental Psychology: Learning, Memory and Cognition, 21, 1459–1468.

Albrecht, J. E., & O’Brien, E. J. (1995). Goal processing and the maintenance of global coherence. In R. F. Lorch & E.J. O’Brien (Eds.), Sources of coherence in reading (pp. 263–278). Hillsdale, NJ: Lawrence Erlbaum.

Black, J. B., & Bower, G. H. (1980). Story understanding as problem solving. Poetics, 9, 223–250.Bloom, C. P., Fletcher, C. R., van den Broek, P., Reitz, L., & Shapiro, B. P. (1990). An on-line assessment of causal

reasoning during comprehension. Memory and Cognition, 18, 65–71.


Bower, G. H., & Rinck, M. (1998). Goals as generators of activation in narrative understanding. In S. R. Goldman, A.C. Graesser, & P. van den Broek (Eds.), Narrative comprehension, causality, and coherence: Essays in honor of TomTrabasso (pp. 111–134). Mahwah, NJ: Lawrence Erlbaum.

Casteel, M. A. (1993). Effects of inference necessity and reading goal on children’s inferential generation. DevelopmentalPsychology, 29, 346–357.

Danner, F. W., & Mathews, S. R. (1980). When do young children make inferences from prose? Child Development, 51,906–908.

Dunn, L. M., & Dunn, L. M. (1997). Peabody Picture Vocabulary Test–III. American Guidance Services: Circle Pines,MN.

Dopkins, S., Klin, C., & Myers, J. L. (1993). Accessibility of information about goals during processing of narrative texts.Journal of Experimental Psychology: Learning, Memory and Cognition, 19, 70–80.

Ericson, K. A., & Simon, H. A. (1994). Verbal reports as data. Psychological Review, 87, 215–251.Goldman, S., & Varnhagen, C. K. (1986). Memory for embedded and sequential story structures. Journal of Memory and

Language, 25, 401–418.Graesser, A. C., Bertus, E. L., & Magliano, J. P. (1995). Inference generation during the comprehension of narrative

text. In R. F. Lorch & E. J. O’Brien (Eds.), Sources of coherence in reading (pp. 295–320). Hillsdale, NJ: LawrenceErlbaum.

Johnson, N. S., & Mandler, J. M. (1980). A tale of two structures: Underlying and surface forms in stories. Poetics, 9,51–86.

Kendeou, P., Lynch, J. S., van den Broek, P., Espin, C., White, M., & Kremer, K. E. (2006). Developing successful readers:Building early comprehension skills through television viewing and listening. Early Childhood Education Journal,33, 91–98.

Kintsch, W. (1977). On comprehending stories. In M. A. Just & P. A. Carpenter (Eds.), Cognitive processes in compre-hension (pp. 33–61). Hillsdale, NJ: Lawrence Erlbaum.

Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text comprehension and production. Psychological Review, 85,363–394.

Kremer, K. E., Lynch, J. S., Kendeou, P., Butler, J., & van den Broek, P. (2002, April). The role of early narrativeunderstanding in predicting future reading comprehension. Paper presented at the annual meeting of the AmericanEducational Research Association, New Orleans, LA.

Langston, M. C., & Trabasso, T. (1998). Identifying causal connections and modeling integration of narrative discourse.In H. van Oostendorp & S. R. Goldman (Eds.), The construction of mental representations during reading (pp. 29–69).Mahwah, NJ: Lawrence Erlbaum.

Long, D. L., & Bourg, T. (1996). Thinking aloud: Telling a story about a story. Discourse Processes, 21, 329–339.

Long, D. L., & Golding, J. M. (1993). Superordinate goal inferences: Are they automatically generated during compre-hension? Discourse Processes, 16, 55–73.

Long, D. L., Golding, J. M., & Graesser, A. C. (1993). A test of the on-line status of goal-related inferences. Journal ofMemory and Language, 31, 634–647.

Loxterman, J. A., Beck, I. L., & McKeown, M. G. (1994). The effects of thinking aloud during reading on students’comprehension of more or less coherent text. Reading Research Quarterly, 29, 352–367.

Magliano, J. P. (1999). Revealing inference processes during text comprehension. In S. R. Goldman, A. C. Graesser, & P.W. van den Broek (Eds.), Narrative comprehension, causality, and coherence: Essays in honor of Tom Trabasso (pp.55–75). Mahwah, NJ: Lawrence Erlbaum.

Magliano, J. P., & Graesser, A. C. (1991). A three-pronged method for studying inference generation in literary text.Poetics, 20, 193–232.

Magliano, J. P., & Millis, K. K. (2003). Assessing reading skill with a think-aloud procedure and latent semantic analysis.Cognition and Instruction, 21, 251–283.

Mandler, J. M., & Johnson, N. S. (1977). Remembrance of things parsed: Story structure and recall. Cognitive Psychology,9, 111–151.

McKoon, T., & Ratcliff, R. (1992). Inferences during reading. Psychological Review, 99, 440–466.McNamara, D. S. (2004). SERT: Self-explanation reading training. Discourse Processes, 38, 1–30.Mayer, M. (1969). Frog, where are you? New York: Dial Press.Meyer, B. J. F., & McConkie, G. W. (1973). What is recalled after hearing a passage? Journal of Educational Psychology,

65, 109–117.Murachver, T., Pipe, M. E., Gordon, R., Owens, J. L., & Fivish, R. (1996). Do, show, and tell: Children’s even memories

acquired through directed experience, observation, and stories. Child Development, 67, 3029–3044.


Paris, A. H., & Paris, S. G. (2003). Assessing narrative comprehension in young children. Reading Research Quarterly,38, 36–76.

Siegel, L. (1994). Working memory and reading: A life-span perspective. International Journal of Behavioral Develop-ment, 17, 109–124.

Stein, N. L., & Glenn, C. G. (1979). An analysis of story comprehension in elementary school children. In R. O. Freedle(Ed.), New directions in discourse processing (pp. 53–120). Norwood, NJ: Ablex.

Suh, S., & Trabasso, T. (1993). Inferences during reading: Converging evidence from discourse analysis, talk-aloudprotocols, and recognition priming. Journal of Memory and Language, 32, 279–300.

Trabasso, T., & Magliano, J. P. (1996). How do children understand what they read and what can we do to help them? InM. F. Graves, P. van den Broek, & B. M. Taylor (Eds.), The first R: Every child’s right to read (pp. 160–187). NewYork: Teachers College Press.

Trabasso, T., & Nickels, M. (1992). The development of goal plans of action in the narration of a picture story. DiscourseProcesses, 15, 249–275.

Trabasso, T., Secco, T., & van den Broek, P. W. (1984). Causal cohesion and story coherence. In H. Mandl, N. L. Stein,& T. Trabasso (Eds.), Learning and comprehension of text (pp. 83–111). Hillsdale, NJ: Lawrence Erlbaum.

Trabasso, T., & Stein, N. L. (1997). Narrating, remembering and representing event sequences. In P. van den Broek, P. J.Bauer, & T. Bourg (Eds.), Developmental spans in event comprehension and representations: Bridging fictional andactual events (pp. 237–270). Mahwah, NJ: Lawrence Erlbaum.

Trabasso, T., Stein, N. L., Rodkin, P. C., Munger, M. P., & Baugh, C. R. (1992). Knowledge of goals and plans in theon-line narration of events. Cognitive Development, 7, 133–170.

Trabasso, T., Suh, S., Payton, P., & Jain, R. (1995). Explanatory inferences and other strategies during comprehension andtheir effect on recall. In R. F. Lorch & E. J. O’Brien (Eds.), Sources of coherence in reading (pp. 219–239). Hillsdale,NJ: Lawrence Erlbaum.

van den Broek, P. (1989). Causal reasoning and inference making in judging the importance of story statements. ChildDevelopment, 60, 286–297.

van den Broek, P. (1990). The causal inference maker: Towards a process model of inference generation in text com-prehension. In D. A. Balota, G. B. Flores d’Arcais, & K. Rayner (Eds.), Comprehension in reading (pp. 423–446).Hillsdale, NJ: Lawrence Erlbaum.

van den Broek, P., Fletcher, C. R., & Risden, K. (1993). Investigations of inferential processes in reading: A theoreticaland methodological integration. Discourse Processes, 16, 169–180.

van den Broek, P., & Lorch, R. F. (1993). Network representations of causal relations in memory for narrative texts:Evidence from primed recognition. Discourse Processes, 16, 75–98.

van den Broek, P., Lorch, E. P., & Thurlow, R. (1996). Children’s and adults’ memory for television stories: The role ofcausal factors, story-grammar categories, and hierarchical level. Child Development, 67, 3010–3028.

van den Broek, P., Lynch, J. S., Naslund, J., Ievers-Landis, C. E., & Verduin, K. (2003). The development of comprehensionof main ideas in narratives: Evidence from the selection of titles. Journal of Educational Psychology, 95, 707–718.

Wenner, J. A. (1999). Panning for goals: Preschoolers’ comprehension of goal structure in narratives. Unpublisheddoctoral dissertation, University of Minnesota, Minneapolis.

Wenner, J. A. (2004). Preschoolers’ comprehension of goal structure in narratives. Memory, 12, 193–202.Whitney, P., & Budd, D. (1996). Thinking-aloud protocols and the study of comprehension. Discourse Processes, 21,

341–351.

Understanding the glue of narrative structure, childrens`s on- and off- line inferences about...

Documents

Transcript of Understanding the glue of narrative structure, childrens`s on- and off- line inferences about...