Abstract
About 10% of primary school children, who have had no history of disability, have unexpectedly poor reading and listening comprehension. The majority of children who have language impairment, autism, or who are deaf or hard of hearing have reading and listening comprehension deficits. This study was designed to investigate listening comprehension skills in typically developing children, between 7 and 10 years old, using an assessment that compared whole and segmented presentation of spoken passages to compare inference-making ability during and after text presentation. The intent was to determine whether segmented presentation benefited inference-making and whether the effects were consistent across the primary school age range or specific to younger children.
The ability to make inferences when listening or reading is critical to comprehension. In this study, the authors distinguished between two types of inferences: local and global coherence inferences (Graesser, Singer, & Trabasso, 1994). Local coherence inferences are made when listeners use information provided in the text to integrate the meanings of clauses and sentences within the text. Sometimes these are cued by pronouns, or can be made by appreciating the link between synonyms or category exemplars, for example, He noticed
The Study
Participants/Task
Sixty typically developing children in third and fifth grades from mainstream schools in England took part in the study.
Test materials consisted of six short stories similar to those used in school-based comprehension tasks. Materials were of appropriate interest (e.g., buying a pet, a birthday party, playing football) and language level for children across the age ranges. There were eight comprehension questions related to each story that assessed ability to establish local and global coherence (four questions for each type of inference). The questions were either presented at intervals during the story (segmented version) or at the end of the story (whole version). The order of the questions followed the order of information presented in the story.
There were six stories. For each story, there was a whole and segmented version (these were exactly the same in terms of story content and questions, only the format of administration differed). For the segmented presentation format, each story was split into five segments of similar length. Each segment was followed by one or two questions. The six stories were grouped into two sets of three stories (sets A and B). Each child completed both sets of stories, one in whole format and one in segmented format. The sets were counterbalanced. Children were not presented with any pictures to accompany the story. This was to prevent them from attempting to use clues in the images to answer the questions. The following is an example from part of one of the stories (L = local inference; G = global inference): Today was Grandma’s birthday. The family was getting ready for the party. Dad and Josh were putting up the party tent in the back garden. Mum told them to put on some suncream, so that they didn’t burn.
Q1: What were the family getting ready for? (L)
Q2: What was the weather like? (G)
Mum drove over to pick up Grandma, who lived an hour away. Mum told Sarah to keep an eye on the cake in the oven and to make some fruit juice. Sarah was slicing oranges when the knife slipped. Her finger was bleeding but she couldn’t find any plasters! Luckily, Brenda, their next-door neighbour, had some.
Q3: Why did Sarah need some plasters? (G)
Q4: Where did Sarah get the plasters? (L) (Freed & Cain, 2017, p. 105)
Children were seen individually by the researcher in a quiet room in school on two separate occasions (a minimum of 2 days apart). On each testing session (lasting approximately 30 min), the child was presented with one set of stories (set A/B). Each set consisted of three test stories, presented one at a time. In one of the sessions, the stories were presented in whole format and in the other they were presented in segmented format. The procedure (whole or segmented presentation) was explained to children at the beginning of the session and they were told they would be asked questions about the story either after the story (whole presentation) or at various points during the story (segmented presentation). Stories were prerecorded and presented using a PowerPoint file on a laptop computer. This ensured that all children received the same input (i.e., to control for pace of delivery, word stress, etc.).
Whole presentation procedure. The child listened to the first story. At the end of the story they answered the eight comprehension questions relating to the story. The remaining stories were then completed according to the same procedure.
Segmented presentation procedure. The child listened to the first story in a series of five segments. After each segment of the story, the child was asked one or two comprehension questions (eight in total for the story). The remaining stories were then completed according to the same procedure.
Prompts were used during the questions where necessary to encourage the child to expand on their answer. The prompt consisted of either a repetition of the question or encouragement for the child to be more specific in their answer. The results reported in the article are based in the child’s first answer (i.e., before prompting). Each inference question was coded as correct (2 points), partially correct (1 point), or incorrect (0 points) such that each story was scored out of a maximum of 8 points for local inferences and 8 points for global inferences.
Results
The fifth-grade students obtained significantly higher scores than the younger third-grade children.
Segmented presentation of the stories resulted in slightly higher scores than whole presentation, but the effect of presentation format did not reach significance.
Scores on the global coherence questions were significantly higher than the local coherence questions for both third- and fifth-grade students.
There was a significant interaction between question type and grade group. There was a bigger increase in scores for the local coherence items with increasing age compared with small increases in scores for global coherence items; the grade groups differed significantly on local coherence inferences but not on global coherence inferences.
There was a significant interaction between question type and presentation format. Scores for the local coherence questions were higher for the segmented compared with the whole presentation format, but the same was not true for global coherence inferences.
Discussion
There was an interaction between age and question type. The interaction arose because there were small increases in the ability to establish global coherence with increasing age, compared with bigger increases in scores for the local coherence items. Only the third-grade children benefited from segmentation. They found the local coherence inferences more difficult in the whole presentation format condition than in the segmented format. There are at least two possible reasons for this. First, segmentation may have directed attention to the key elements or relationships in the text, resulting in superior performance when questions are asked after each relevant segment. Second, segmentation may have supported the children’s performance on the local coherence questions because it alleviated memory demands; children were prompted to integrate the relevant information close to where it had been presented. The pause between the different segments of the story may have also provided the younger children with an opportunity to monitor coherence so far and attempt to fill in any missing information before presentation of the next segment. The authors attribute the absence of an effect of presentation format for the older children to their greater experience of with comprehension and expectation that stories should be coherent. Furthermore, the increasing cognitive and strategic resources that come with age may enable older children to make a greater range of inferences during story presentation.
For both groups of children, performance was stronger on global coherence questions compared with the local coherence questions. Currie and Cain (2015) proposed that children may show stronger performance for global coherence inferences because of key differences in the centrality of the information tapped by the two question types. Some of the global coherence inference questions related to story themes, for example, the setting or main character. For that reason, these inferences were essential to understand core details of the text. For the local coherence inferences, the links made between these locally adjacent elements in the text could be viewed as supporting more peripheral details in the text. This explanation is supported by research into the centrality effect, which shows that children are more likely to remember information that is central to the overall meaning of the story than they are to remember peripheral information (Miller & Keenan, 2009). If children are unable to make all the required inferences while listening to a story, a strategic technique to adopt is to focus on understanding the overall gist of the story.
Implications
Further research is needed to explore how children with identified language impairments respond to making inferences with whole or segmented texts. The results from this study have implications for supporting and assessing comprehension skills.
Supporting inference-making: The results suggest that segmented presentation of comprehension tasks results in better comprehension for some children. Segmented presentation also has ecological validity because it reflects a child’s real-life experiences as they will often pause while listening to a story to discuss what is happening with their caregiver/teacher. The segmented presentation format in this study could be viewed as a more “structured” version of shared book reading as set questions are asked after each segment of text.
Assessing inference-making. The authors note that standardized assessments of reading and listening comprehension typically follow a whole text presentation format and do not enable clinicians and teachers to assess the potential of inference-making ability during text presentation, even though this format is used informally to support the construction of meaning from text. In addition, standardized assessments do not systematically distinguish between local and global coherence inferences, even though these have been distinguished empirically (Cain & Oakhill, 2014; Currie & Cain, 2015) and may have different developmental trajectories, as suggested by this data. The results of this study showed that segmented presentation may be beneficial for some children, particularly for the generation of inferences to support local coherence. The results from the present study have implications for the design of comprehension assessments that can measure potential in addition to actual performance as well as for classroom practice supporting children with comprehension difficulties in school.