Clinical Question: Would young children demonstrate improvements in inferential question-answering after interactive book-reading intervention that targeted inferential questions in comparison to a similar intervention that targeted overall language ability?
Method: Systematic Review
Study Sources: ERIC, Academic Search Complete, ASHAWire
Search Terms: inferential questions OR inferences OR inferential language AND intervention OR instruction OR treatment OR language impairment
Number of Included Studies: 3
Primary Results: Children who participated in interactive reading interventions that targeted inferential language, including inferential question-answering, demonstrated improvements in literal and inferential language. No direct comparisons of interventions to target inferential language and interventions to target overall language were available.
Conclusions: Although there is substantial evidence to support interactive bookreading as an effective strategy for language in young children, few studies have examined teaching inferential language in this context. The small number of studies available suggests that strategies such as prompting for responses to questions, modeling appropriate responses, and including "think aloud" explanations of responses may be appropriate for teaching inferential question-answering in young children. There is a need for additional research to identify the best approaches to teach inferential language, including answering inferential questions about stories, to young children.
Kate is a speech-language pathologist who works in an elementary school serving many children from low-income families. As part of her practice, she works with kindergarten children, many who have limited oral language skills but have not been identified with language impairment. For these children, her primary service delivery model is classroom-based biweekly small-group sessions that focus on language skills in the context of interactive book-reading activities. Kate uses an interactive reading style to provide multiple opportunities for children to engage with the book and respond to questions. Kate bases her interactive reading approach on dialogic reading (Whitehurst et al., 1994). She selects age-appropriate books, including some of the same books that are used in the whole-group classroom instruction provided by the teacher. During the small-group sessions, Kate reads the storybook aloud, embeds multiple opportunities for children to respond to questions and make comments about the story, and uses recasts and expansions to teach vocabulary and other language targets. As part of the interactive book-reading, Kate targeted two primary types of story questions: literal questions and inferential questions. Literal questions are questions that the child can answer by recalling information from the story. For example, Kate asks children to answer questions like, Who was in this story? What was Ashley doing? What happened at the end of the story? For most literal questions, the answer to the question is directly stated in the story or shown in an illustration.
Kate also targeted inferential questions during her small-group sessions. Inferential questions require the child to make an inference about a character or event in the story. The child might be asked to make an inference about a character’s emotion, talk about the reasons for a character’s action, or make a prediction about events in the story. The necessary information to answer inferential questions is not usually directly stated in the text or shown in the pictures. Inferential questions are generally more abstract than literal questions and, often, the child needs to make a connection between the story and his or her own knowledge. Examples of inferential questions include, Why was Anna happy? Why did Theo help Ben? and What do you think will happen next?
At the end of the first semester, Kate carefully examines the data she collected during sessions and from brief progress-monitoring assessments. During sessions, Kate kept a quick tally of correct and incorrect responses to questions about the story for each child. Because she typically reads the same story repeatedly on consecutive days, she can monitor how each child is improving in his or her ability to answer the questions he or she has been practicing with the same story. In general, she notices that children improve across sessions with the same storybook, frequently learning to answer most of the literal questions and many of the inferential questions by the end of several readings. Kate conducted three brief progress-monitoring assessments during the semester. For these assessments, Kate reads a new story with each child individually and asks the child to respond to literal and inferential questions about the story. Because the children have not heard the stories or the questions before, progress-monitoring sessions tell Kate how children are improving in their overall ability to answer questions about stories. The data from the progress-monitoring sessions indicates that the children improved their ability to answer literal questions about stories. For example, one child improved from 30% correct at the first progress-monitoring assessment to 90% correct at the most recent time point. However, the data indicates that children still struggle with inferential questions. The same child averaged around 30-40% correct at all three progress-monitoring sessions.
During classroom reading activities, Kate observed that other children in the kindergarten class are able to answer these types of inferential questions. She believes that answering questions about stories is an important skill for the children she works with, because she knows that early language skills contribute to later reading comprehension ability. Reflecting on her interactive book-reading sessions, Kate noted that she frequently spent more time on, and provided more practice opportunities for, literal questions. She thought that increasing the focus on inferential questions might be one strategy to improve the kindergarten children’s ability to answer inferential questions. However, Kate hoped that she could find evidence to support other specific practices.
As part of a multi-tiered system of support (MTSS or Response to Intervention model), tiers of services are delivered to support the learning needs of all children. In many schools, one of the newer roles for the speech-language pathologist is to contribute to efforts to prevent later academic difficulties (Ehren, Montgomery, Rudebusch, & Whitmire, 2006). Special-education funds can be devoted to support children who may need additional support (i.e., Tier Two), but who have not yet been identified as needing special-education services. Kate is currently in this position. She collaborates with the kindergarten teacher to work with a group of children who do not meet benchmarks in the oral language area. Her goal is to support the learning needs of these children sufficiently so they make adequate progress.
Interactive Book-Reading Intervention
Interactive book-reading is an evidence-based practice with strong research support. In a meta-analyses of 31 studies of interactive book-reading with preschool and kindergarten children, Mol, Bus, and de Jong (2009) reported moderate effects on expressive and receptive vocabulary (overall d = .54). Interactive book-reading programs have been demonstrated to improve vocabulary knowledge of young children when administered individually or in small groups, implemented by parents or teachers, and with groups of children who are at risk due to socioeconomic factors or limited oral vocabulary (Dale, Crain-Thoreson, Notari-Syverson, & Cole, 1996; Hargrave & Sénéchal, 2000; Lonigan, Anthony, Bloomfield, Dyer, & Samwel, 1999; Whitehurst et al., 1994; Whitehurst et al., 1988). Although many studies of interactive book-reading have focused on vocabulary knowledge as an outcome, few studies have examined interactive book-reading as an approach to improve other oral language skills.
Kate’s interest in the inferential language abilities of the children she works with is well warranted. In preschool-age children, the ability to make inferences is related to comprehension of narratives (Kendeou, Bohn‐Gettler, White, & Van Den Broek, 2008; Lepola, Lynch, Laakkonen, Silvén, & Niemi, 2012) and in school-age children, the ability to make inferences is predictive of later reading comprehension (Cain, Oakhill, & Lemmon, 2004). For the children that Kate works with, her observations of limited inferential language are not surprising. Children with language difficulties are likely to struggle with both literal and inferential language (Bishop & Adams, 1992) and there is evidence inferential language may be particularly challenging (Blank, Rose, & Berlin, 2003). Children with language impairment perform poorly relative to typically developing peers on inferential tasks, and performance on these tasks is related to language comprehension (Adams, Clarke, & Haynes, 2009; Ford & Milsoky, 2003). Difficulties with inferential language and deficits in reading comprehension are interrelated (Cain et al., 2004; Cain & Oakhill, 1999).
In studies of parent–child interactions during storybook sharing activities, children who were exposed to more inferential language used more of it and improved in inferential language abilities (van Kleeck, Gillam, Hamilton, & McGrath, 1997). Importantly, children who participated in storybook sharing with inferential language had higher scores on measures of reading comprehension in the third grade than peers who were less engaged in storybook sharing (Serpell, Baker, & Sonnenschein, 2005). Van Kleeck (2008) argues that inference-making contributes to later text comprehension by encouraging children to make connections between information in the text and their own knowledge. Indeed, text comprehension strategies taught to older children frequently include strategies for generating inferences and making connections with background knowledge (see for a review: Rosenshine & Meister, 1994).
Kate used the PICO framework to develop a targeted, clinically-relevant research question, as suggested by the American Speech-Language-Hearing Association (ASHA). She defined (P) the population, (I) the intervention, (C) the comparison intervention, and (O) the intended outcome as:
- P - young children with or without language difficulties
- I - interactive book-reading intervention with a focus on inferential questions
- C - interactive storybook reading with a more general language focus
- O - improvements in inferential question-answering
Kate’s question was: Would young children demonstrate improvements in inferential question-answering after interactive book-reading intervention that targeted inferential questions in comparison to an interactive bookreading intervention that targeted overall language ability?
Search for the Evidence
Before she began to look for articles, Kate thought carefully about the criteria for her search. She wanted to find high-quality evidence so she decided to limit her search to articles published in peer-reviewed journals. She would include studies that used a number of study designs (experimental, quasi-experimental, single-case design). Although she was most interested in studies that included children around the kindergarten age, she thought that studies that included preschool-age or early school-age children could provide important information because the language abilities and intervention approaches appropriate for children in this age range would be similar. She decided that her primary inclusionary criteria would be (a) a published, peer-reviewed study of; (b) an interactive bookreading intervention that targeted inferential questions with; (c) preschool-age or early elementary-age children as participants; and (d) a measure of inferential language included as an outcome.
Kate began her search using two databases: ERIC and Academic Search Complete. Figure 1 provides details of the search process.
She conducted multiple searches using combinations of the keywords inferential questions, inferences, inferential language, AND intervention, instruction, treatment. Kate found that searches using more general search terms like story comprehension AND questions resulted in too many unrelated articles (638 articles in ERIC). Although Kate did not limit her search to articles focused on children with language difficulties, she also conducted searches with language impairment in combination with her other search terms. These searches resulted in over 160 articles, including many duplicates. Kate continued her search using the ASHAWire database but did not identify any additional articles. Her next step was to review the titles and abstracts to identify articles that were appropriate. Many of the articles identified in the search did not meet criteria (i.e., were not treatment articles, did not specifically target inferential language). Kate next reviewed the full text of the most relevant articles and searched the reference list of each of those articles to see if any additional articles were listed and located one additional article. During her search, she also located one review article (Hall, 2015) but no articles included in that paper met her search criteria.
Evaluating the Evidence
Summaries of Included Studies
Three studies met Kate’s original search criteria: van Kleeck, Vander Woude, and Hammett (2006), Desmarais, Nadeau, Trudeau, Filiatrault-Veilleux, & Maxés-Fournier (2013), and Bradshaw, Hoffman, and Norris (1998). See Table 1 for summaries of the three studies.
Van Kleeck et al. (2006) delivered an individualized interactive-reading intervention to 30 preschool children with language impairments. Participants attended Head Start centers and were randomly assigned to either the intervention condition or a no-intervention control group. In the intervention condition, children participated in brief (15 minute) sessions twice a week for eight weeks. Trained research assistants led children through scripted interactive reading activities using storybooks that had been modified for this purpose. The two storybooks, Mooncake (1983) and Skyfire (1984) by Frank Asch were selected for similarities in complexity and length. Three scripts for each book were developed based on previous study of parents and children reading the same two books (van Kleeck Kleeck et al., 1997). Each script included 25 questions, with 25% of them inferential. Samples from a script were included in an appendix (p. 95). For each question, the script included opportunities for the child to respond and prompts to guide the child to a correct response. For example, for the question How do you think Bear feels because his friend Little Bird is leaving?, a prompt might highlight key information in the story using a cloze procedure. Maybe he feels sad because he won't see (point to picture of Little Bird)_______. When children responded appropriately, the adult gave a confirmation, Yes. I think he feels sad because he won't see his friends for a long time. When children did not respond appropriately, a correct response and explanation was modeled by the adult. For example, the adult might say, Bear is sad because he won’t see his friend for a long time. Do you ever feel sad when you won’t see someone for a long time? The authors described the modeled responses and explanation as “thinking aloud,” a strategy that would help children understand the story and how to respond to questions.
At pre- and posttest, participants received a measure of receptive vocabulary, the Peabody Picture Vocabulary Test, Third Edition (PPVT™-III; Dunn & Dunn, 1997) and a measure of literal and inferential language, the Preschool Language Assessment Instrument (PLAI; Blank, Rose, & Berlin, 1978). The PLAI was scored to create two subscores, one for literal language and one for inferential language. Participants who received the intervention demonstrated greater gains than the control group in receptive vocabulary, literal language, and inferential language with medium to large effect sizes.
Desmarais et al. (2013) provided a similar intervention to 16 children (4- through 6-year-olds) with language impairments. The study used a multivariate repeated-measures design and there was no comparison group. Participants received 10 sessions (15 to 20 minutes each) of individual intervention provided by their speech-language pathologist. The authors stated that intervention procedures were modeled from van Kleek et al. (2006) and consisted of interactive book-reading with scripted literal and inferential questions. Two books were selected for length, illustrations, and narratives structure. Each book-reading session included 16 questions, eight of which were inferential, with a heirarchy of prompts to guide children to appropriate responses. Prompts began with a simpler form of the question. (i.e., rephrasing a question into a cloze statement), followed by a semantic cue (i.e., adding information from the story), and finally a phonemic cue (i.e., giving the first sound or syllable in the response). After prompting, children were asked to answer the question again. Outcome measures were the PLAI and a researcher-created measure of story comprehension.