Research Article Comprehension and Inference: Relationships Between Oral and Written Modalities in Good and Poor Comprehenders During Adolescence Anna PotockiEl and Virginie Lavala Purpose: We investigated the relationships between text reading comprehension and oral idiom comprehension in adolescents. We also examined the more specific relationships between inference in text comprehension and inference in idiom comprehension. Method: We selected participants from an initial sample of 140 students aged 1315 years to form 2 groups, according to their decoding and reading comprehension abilities: 1 group of good ccmprehenders/good decoders (n = 49) and 1 group of less skilled ccmprehenders but with adequate decoding skills (n = 20). The reading comprehension task comprised both literal and inferential (textbased and knowledgebased) questions. These 2 groups were then compared on an idiom comprehension task. In this task, idioms were presented orally, and students were placed in omprehension skills are usually described as \"general\" abilities that are not specic to any particular modality of language presentation (e. g, written, oral, or picture form). As a result, readers exhibiting comprehension difficulties (i.e., less skilled ccmprehenders) are likely to perform poorly in all types of activities requiring the understanding of linguistic material, regardless of pre- sentation modality (e. g., Cain & Oakhill, 2007). The idea of \"modality-nonspecific\" comprehension decits has already been examined in the literature, but using either the same type of material (i.e., texts) presented in different modalities (i.e., written, oral, or picture form; see, e.g., Gernsbacher, 8Centre de Recherches sur la Cognition et l'Apprentissage, UMR 7295, CNRS, University of PoitiersUniversity Francois Rabelais of Tours, France Correspondence to Anna Potocki: anna.potocki@univpoitiers.fr Editor-in-Chief: Sean Redmond F'rlitnr' Megan Dunn Davisnn a situation that simulated a reallife oral interaction. The idioms were novel for the students (translated from a foreign language), either transparent or opaque, and presented either with a supportive context or without any context. Results: Good reading ccmprehenders outperformed less skilled ones on the idiom task. Both groups benefited from the supportive context, especially the good ccmprehenders. Knowledge-based inferences in written text comprehension were related to contextual inferences for opaque idioms, while semantic inferences for transparent idioms were related to literal text comprehension, but not to text- connecting inferences. Conclusion: These results are discussed both theoretically, in terms of crossmodal comprehension processes, and practically, in terms of implications for remediation. Varner, & Faust, 1990) or different types of material (e. g., texts vs. idioms) presented in the same modality (i.e., writ- ten format; see, e. g., Cain, Towse, & Knight, 2009). One of the objectives of this study was to take this research a step further by examining adolescents' comprehension of different types of linguistic material (texts vs. idioms) pre- sented in different modalities (written text vs. oral idioms). We reasoned that if comprehension processes are indeed general, then the processes (e.g., inferential) underlying the comprehension of linguistic utterances should be similar across modalities and types of material. This study therefore investigated this issue by comparing the inferences required to understand texts (text-connecting and knowledge-based inferences) and those required to understand idioms (seman- tic and contextual inferences). Furthermore, this research focused on adolescents in order to deepen our knowledge about the development of text and idiom comprehension skills in an age group that has rarely been investigated up to now. This study therefore had important implications both theoretically, in terms of advances in our knowledge of comprehension development, and practically, in terms by the two modalities. Similarly, Berl et al. (2010) compared (e.g., Levorato & Cacciari, 1992; Nippold, Moran, & to understand its meaning (Ackerman, 1982; Brinton, Fujiki, of the remediation of comprehension difficulties. functional magnetic resonance imaging activation in children Schwarz, 2001; Nippold & Rudzinski, 1993; Nippold & & Mackey, 1985; Cacciari & Levorato, 1989; Cain et al., Mage = 10 years) as they listened to a paragraph-length Taylor, 1995, 2002), participants (generally adults) were 2005, 2009; Gibbs, 1987, 1991; Hsieh & Hsu, 2010; Laval, Comprehension as a General Process story or silently read the same story. Again, although some asked to rate the idioms' familiarity on a scale, indicating 2003; Levorato & Cacciari, 1992, 1999). According to the areas were differently activated for reading or listening to whether they already knew the idioms and at what age global elaboration model (Levorato & Cacciari, 1995, 1999), According to the simple view of reading model stories, results also showed overlapping activation in regions they thought they had encountered them for the first time. the ability to understand idioms goes through different (Gough & Tunmer, 1986), reading comprehension is the generally dedicated to the comprehension of language (e.g., These studies indicated that the role of familiarity changes phases. Up to 6-7 years, children use a semantic strategy product of two main components: written word identifica- left superior temporal sulcus). with age, taking on added importance at the end of ado- and so are not affected by an inappropriate understanding tion and language comprehension. This model describes Although this view of a unitary comprehension pro- lescence. In 2005, Cain, Oakhill, and Lemmon (see also of the context. In the second phase, the context becomes the comprehension component as a general ability to under- cess has been well supported, it should nonetheless be noted Cain et al., 2009) came up with an original way of con- crucial, and children are able to search for relevant contex- stand language. The term general implies that similar pro- that most of the above-cited studies examined the impact of trolling for familiarity with their idiom comprehension tual elements. Then, at 10-11 years, semantic analysis once cesses are involved in language comprehension whatever the modality (listening vs. reading) on a single type of material, task. These authors divided idiomatic expressions into two more becomes predominant, and the context is no longer sensory modality of the text, sentence, or utterance. Thus, namely, texts (mostly narrative). We can therefore hypothe- categories: real expressions (i.e., exist in the children's native essential for idiom comprehension. However, other data both listening comprehension and reading comprehension size that the high correlations between reading and listen- language) and novel expressions (translations of idiomatic show that context continues to influence the understanding are thought to rely on the same general processes. As a re- ing comprehension performances and similar patterns of expressions from other European countries that have n of idiomatic expressions in adolescence, albeit to a lesser sult, once decoding skills have been mastered, performances neural activation during reading and listening comprehension equivalent in the participants' mother tongue). This distinct extent (Nippold & Martin, 1989; Qualls, O'Brien, Blood on listening and reading comprehension tasks should be are due to the involvement of the same type of comprehens tion allowed them to assess the role of other factors involved & Hammer, 2003). As mentioned above, context interacts comparable. Several studies have supported this hypothesis. sion process (e.g., knowledge about narrative structure). in idiom comprehension (see below) by controlling for with the idiom's degree of transparency, insofar as context is First, Gernsbacher et al. (1990) tested the hypothesis that a Thus, the purpose of this study was to explore the relationships the possible confounding effect of familiarity and prior particularly crucial for opaque idioms that do not allow for general comprehension skill factor transcends language between reading and listening comprehension for different knowledge. semantic analysis. The use of context to understand opaque comprehension by administering a multimedia comprehension types of linguistic utterances. More specifically, we com- A second factor that influences idiom comprehension idioms relies on inferential ability battery to 270 undergraduates. This battery (Gernsbacher pared written text comprehension performances with idiom is its degree of transparency. The transparency of an idiom & Varner, 1988) comprised six narratives: two in written comprehension performances in real-life oral communication corresponds to the proximity between the meaning of the form, two in oral form, and two in picture (i.e., nonverbal) expression and the words making it up (Caillies & Le Sourn- Relationships Between Reading Comprehension situations. form. When the authors compared the students' story Bissaoui, 2006). To understand a novel idiomatic expression, and Idiom Comprehension comprehension scores, they found very high correlations Factors Involved in Idiom Comprehension individuals can implement a semantic analysis strategy that Assuming that comprehension is a general process, (.72-.92) between performances across the three modalities. consists of relying on the meanings of the individual words another factor that can influence the comprehension of Results of a principal component analysis also clearly and Its Development in the expression to derive its idiomatic meaning (Caillies & idiomatic expressions, besides the nature of the idiom itself supported the hypothesis of a general comprehension factor. Idiomatic expressions are figurative statements with Le Sourn-Bissaoui, 2006; Gibbs, 1987, 1991, 1994; Nippold or the context in which it appears, is the children's level of Kendeou, Bohn-Gettler, White, and Van Den Broek (2008) conventional meanings that cannot be deduced from the & Duthie, 2003; Titone & Connine, 1994, 1999). However reading comprehension. Several studies have examined idiom found similar results when they followed two cohorts of literal meanings of the individual words making them up this type of semantic strategy is only relevant for transparent comprehension abilities in good and poor comprehenders of young children aged 4 and 6 years until they turned 6 and (Marquer, 1994; Tabossi, Fanari, & Wolf, 2008). For ex- idioms (i.e., idioms whose literal and idiomatic meaning written texts. Poor comprehenders are children who exhibit years, respectively. At each time point, the children's ample, it is not enough to know the meanings of the words are closely linked). For example, the expression throw money impaired comprehension skills despite possessing adequate narrative comprehension skills were assessed using oral, hot and potato to understand the meaning of the idiom a out the window can be analyzed word by word to deduce decoding abilities (e.g., Cain & Oakhill, 2006; Catts, Hogan, elevised, and written stories. Results demonstrated that the hot potato. Idiomatic expressions are generally specific to a its idiomatic meaning (spend money profligately). However, & Fey, 2003). These children represent 3%-10% of a school- children's narrative comprehension skills were significantly given language and can rarely be translated word for word this semantic analysis strategy is not appropriate for other age population (e.g., Aaron, Joshi, & Williams, 1999; Catts and highly correlated across the different media in both into another language without losing their meaning. Far opaque expressions. For example, it is not possible to deter- et al., 2003; Leach, Scarborough, & Rescorla, 2003; Torppa cohorts and at each time point. The authors therefore sup- from belonging to a formal language register, idioms are mine the meaning of the French expression to cut a flank et al., 2007). The text comprehension difficulties these ported the view of a generalized comprehension ability. extremely common in both written and spoken language steak (meaning to have a conversation) by analyzing the indi- children experience are generally observed in both the The relationships between written and oral compre- (Gibbs, 1994; Kerbel & Grunwell, 1997; Lundblom & vidual words, as its literal meaning bears no relation to its written and oral modalities (Kendeou et al., 2008; Megherbi hension have also been explored by contrasting different text Woods, 2012). For instance, around 10% of sentences in idiomatic one. Transparent idioms have generally been & Ehrlich, 2005; Megherbi, Seigneuric, & Ehrlich, 2006; genres. Diakidoy, Stylianou, Karefillidou, and Papageorgiou (American) reading books aimed at 8- to 12-year-old chil- found to be better understood than opaque ones (Caillies Paris & Paris, 2003). However, as mentioned earlier, in (2005) tested 612 children in Grades 2, 4, 6, and 8. The chil- fren contain idiomatic expressions (Nippold, 1991). Indeed, & Le Sourn-Bissaoui, 2006, 2008, 2013; Cain et al., 2005, most cases, their comprehension performances in different dren read and listened to two narratives and two expository according to Gibbs (1999), idioms are among the most 2009; Gibbs, 1987, 1991; Levorato & Cacciari, 1999; modalities have, up to now, been assessed using the same texts and completed corresponding comprehension tests. common forms of figurative language. Given the high fre- Nippold & Taylor, 1995, 2002). From a developmental type of material (i.e., texts, mostly narrative). In general, Results showed significant correlations between reading and quency of such expressions, it seems important that children point of view, researchers have also shown that the semantic their comprehension difficulties concern the ability to pro- listening comprehension scores for both narrative and ex- and adolescents are able to interpret them effectively. The analysis strategy is used from an early age (5 years) and duce inferences (Cain & Oakhill, 2006; Potocki, Ecalle, & pository texts at all grade levels. However, the relationship ability to understand idiomatic expressions certainly appear continues to develop until adulthood (Cacciari & Levorato, Magnan, 2013). Text comprehension relies on two types between listening and reading comprehension was weaker for essential for successful communication, and many researchers 1989; Caillies & Le Sourn-Bissaoui, 2006; Gibbs, 1987; of inferences (Cain & Oakhill, 1999; Potocki, Bouchafa, expository texts than for narratives. Finally, these empirical have therefore used idioms to examine the cognitive processes Levorato & Cacciari, 1992). Magnan, & Ecalle, 2014). The first of these, mostly referred findings have been reinforced by neuroimaging data. In underlying the comprehension of nonliteral language (e.g., For opaque idioms in particular, there is a third crucial o as text-connecting inferences, allows successive textual adults, Michael, Keller, Carpenter, and Just (2001) compared Caillies & Le Sourn-Bissaoui, 2008; Cain et al., 2009; Laval, factor: the presence of a context. Defined by the characteris- statements to be linked together. For example, they are the neural substrates of reading and listening comprehension 2003; Levorato & Cacciari, 1992, 1999). tics of the communication situation (e.g., location of the recruited to resolve anaphora across sentences. The second of sentences using functional magnetic resonance imaging. These studies have identified several factors that in- individuals in space and time, activity in which they are type, mostly referred to as knowledge-based inferences, allows Although they observed different degrees of activation for fluence the comprehension of idiomatic expressions. First, engaged at the time of interaction), this context has mostly the text's implicit information to be understood by integrat- reading versus listening in some areas, they also found evi- the more familiar an idiom is, the more likely individuals been operationalized by inserting idiomatic expressions into ing it with the reader's or listener's previous knowledge. dence of the partial sharing of neural systems (especially in are to know its meaning (Bonin, Meot, & Bugaiska, 2013 stories. From the age of 6 years, children seem able to rely For example, if the text contains the words sand, swimsuit, he left inferior frontal and left posterior temporal regions) Tabossi, Arduino, & Fanari, 2011). In most previous studies on the context in which an idiomatic expression is inserted or wave, these inferences are needed to understand that 3432 Journal of Speech, Language, and Hearing Research . Vol. 62 . 3431-3442 . September 2019 Potocki & Laval: Reading and Idiom Comprehension 3433the story is taking place at the beach. Knowledge-based inferences are generally the most difcult for children to generate (Cain & Oakhill, 1999; Casteel, 1993), but poor comprehenders may encounter difculties with both types of inferences, and some may even have problems with literal comprehension (Potocki et al., 2013). Some researchers have investigated whether the dif- culties experienwd in understanding written texts also lead to impaired idiom comprehension. Preliminary evidence for a relationship between text and idiom comprehension was supplied by Nippold et a1. (2001). In a population of young adolewents (MIKE = 12;4 bearsgmonthsD, these authors found that a good level of idiom comprehension was closely associ- ated with reading and listening comprehension skills In youn- ger children aged 779 years, Levorato, Nesi, and Canaan (2004) explored the ability of good and poor comprehenders to choose the appropriate meaning of idioms presented in context. Although performances were relatively high for both groups of children, they nonetheless found qualitative differences when they analyzed incorrect responses, with poor comprehenders selecting literal responses more oen, suggesting that they paid less attention to contextual infor- mation. In the wake of these studies, Cain et a1. (2005) specically examined the ability of good versus poor com- prehenders to perform the two types of idiomatic inferences (semantic vs, contextual) They adopted an innovative way of investigating the role of context in the comprehension of idioms, using novel idioms (i.e., idioms translated from foreign languages that had no equivalent in the children's native language) to control for the familiarity effect (see also Cain et al., 2009; Oakhill, Cain, & Nesi, 2016). They then compared 9-year-old children's ability to understand idioms that were transparent or opaque, or novel or real, and pre- sented either in a supportive context or without any context. The authors found that good and poor comprehenders differed in their ability to use the context to interpret novel opaque idioms (i.e., contextual inferences), but not to inter- pret novel transparent idioms in or out of context This effect was replicated and investigated further by Cain and Towse (2008). Using the same types of variables to differen- tiate between idioms (transparent vs. opaque, real vs. novel, with vsr without supportive context), these authors compared the abilities of good and poor comprehenders aged #10 years to understand idioms in a multiple-choice task. They found no differmoes between good and poor comprehendera' ability to understand transparent idioms, but good comprehenders outperformed poor ones in understanding opaque idioms presented in context. The authors therefore concluded that poor comprehenders were able to use semantic inferencing properly, and their idiom comprehension difculties arose from a specic decit in contextual inferencing. Although both groups had higher comprehension scores when the opa- que idioms were presented in context, the poor comprehenders still performed more poorly than the good comprehenders Taken together, these results support the View that comprehension is indeed a general process, as the difculty some children have understanding written texts is also observed when these children have to understand idiomsr However, one limitation that undermines this conclusion is that idiom comprehension was always assessed in the written modality in these studies (or by reading the idioms aloud to the children: see Cain A'L Towse, 2008). As a result, the conclusion of a general comprehension process has been drawn either by studies comparing comprehension skills across modalities but with the same type of material (i.e., texts) or by studies comparing comprehension abilities across different types of material (texts vs. idioms) but within the same modality (i.e., written modality). One my of testing the general character of comprehension skills further would therefore be to compare written text com prehension and the ability to understand oral idioms in a (simulated) reallife communication situation. Objectives of This Study and Hypotheses The main objective of this study was to confirm the general nature of comprehension skills by studying the rela- tionships between written (narrative) text and oral idiom comprehension abilities. To this end, we compared one group of good written text comprehenders and one group of poor comprehenders (but with adequate decoding skills) on an idiom comprehension task. We expected to observe a difference between these two groups in favor of the good comprehenders. Previous research has generally examined idiom and text comprehension skills within the same modal- ity (i.e., written) To overcome this limitation, in our study, the ' 'orns were presmted orally in a situation that simulated a real-life oral interaction, We then investigated the rela- tionships between the different types of oral and written inferences in order to explore whether the processes under- lying comprehension are indeed general. We assumed that the same kinds of inferences are involved when a reader has to link successive textual sentences (text-connecting inferences) as when a listener has to semantically analyze a transparent idiom (semantic inferences), as similar semantic analyses are likely to take place in both conditions. Similarly, we asumed that there is a specic link between inferences that require the reader's general knowledge to be added to the textual information (lmowledgeebased inferences) and those that rely on the context to elucidate opaque idioms (contextual inferences). In both situations, individuals need to integrate the linguistic statements with their own back- ground knowledge to produce the relevant inferences. Finally, since very few studies have explored adolescents' idiom comprehension skills, we examined the role of context in idiom comprehension in this specic population We expected to find that context continues to play an important role in idiom comprehension during adolescence, Method Participants The initial sample was composed of 140 adolescents aged 1215 years (M,Ape = 1315 years) attending different middle schools in urban or suburban areas in western France. All these adolescents were native French speakers. 3434 Journal oISpsech, Language, andHeari'ng Research - Vol, 62 - 34513442 - September 2019 According to their teachers' reports, none of them had linguistic, behavioral, or cognitive deficits, Two groups were then constituted on the basis of the students' 2 scores in reading comprehension and written word identification, The rst group (good comprehmders/good decoders, n = 49) was composed of children with z scores above 1 for reading comprehension and l for written word identification. The second group (poor comprehenders but with adequate decoding skills, it = 20) was composed of students with z scores below 1 for reading comprehension but above 1 for written word identication. The characteristics of these two groups are set out in Table L Material Written Word Identication The test La Pipe e1 12 Rat (Lefavrais, 1968) was used to assess adolescents' word identification skills This task features 486 words (50% animal names), each printed on a separate line on three pages. The students had to silently read as many words as possible and underline all the animal names. A total of 3 min were allotted to the task. The dependent variable ms the difference between the number of correct responses (animal names correctly underlined) and the number of false alarms (underlined words that were not animal names). Reading Comprehension Reading comprehension was assessed with two narra- tive texts, one short (\"lulie,\" 236 words; derived from Cain dc Oakhill, 1999; see Appmdix) and one long C'Ansaldo," 644 words; derived from Gemsbacher et al., 1990). A set of 12 multiple-choice questions was created for each text. Four of these questions referred to literal information explicitly stated in the text, four elicited text-connecting inferences, and four required knowledge-based inferences. The texts and questions had been pretested on a different population of 121 adults (university students), and several adjusunents were made to the task (questions rewarded) prior to using it with the adolescents based on the pretest scores. After silently reading each text, participants had to read and answer the 12 associated multiplechoioe questions. We calculated the total number of correct responses for each of these texts and summed the scores for each type of question (i.e., literal, text-connecting inferences and knowledge-based inferences). Cronbach's alphas were acceptable or nearly acceptable (or = .75 for \"Ansaldo\" and at = .67 for \"Julie\"). Idiom CompreheI-on Idiom comprehension was assesmd using Idiom experi- mental software (Hattouti, Gil, Almecija, & Laval, 2019; Laval, Gil, & Hattouti, 2016) with 12 novel idioms: six transparent and six opaque. Two other idioms (one trans- parent, one opaque) were used as examples at the beginning ofthe task. As in Cain et all (2005; see also Cain & Towse, 2008; Oakhill et al., 2016), these were novel idioms, as they had been translated from other European languages and had no equivalent in French. The idioms were presented using videos that simulated a reallife communication situation (i.e., in these videos, the male or female actor seemed to speak directly with the participant when pronouncing the id- iom). The idioms were presumed either within a supportive context that helped to place the use of the idiom within a spe- cic situation/story or without any context. An example il- lustrating the organization of each video is provided in Table 2 (video stimuli can also be shared upon request). The idiomatic expressions produwd by the actors were idencal (facial expressions, prosody, gaze, etc.) across conditions (i.e., in context and without context). In the supportive context, the entire video (Parts 175) was shown to participants. In the without-context condition, only Parts 4 and 5 were shown. After viewing each video, the students wrote their answers to the question \"What does [the actor] mean?" in a booklet. Their answers were then coded according to six categories based on the Classication developed by Cain et all (2005): idiomatic answer (awarded 2 points), partial idiomatic answer (awarded 1 point), answer without an idiomatic link (0 point), literal answer (0 point), repetition (0 point), and other for responses that did not t any of the above categories (0 point) For example. for the idiom \"to carry a torch\" (meaning \"to be in love\"; see Table 2), an idi- omatic answer would be \"she it in love": a partial idiomatic answer, \"she is happy"; an answer without an idiomatic link, \"she is smart"; and a literal answer, \"she cannot rec clearly." Tupi. 1. Means (standard deviations) ior each task for the two groups of good and poor comprehenders (N = 69). reek Good comprehenders (n - 4e) Poor comprehenders (n - 20) Written word idemican'on 99 (11.4) 92.7 (10.4) Total reading comprehension 21.2 (1.7) 14.4 (3.1)- Literal comprehension a6 (1) 5 (1 .4)- Text-connecting inierences 7.33 (0.75) 4.5 (1 .5)- Knowledge-based mierenues 7.31 (or) 4.95 (1 .3)- Transparent idioms without context s9 (1.9) 6.35 (2.2)- Transparent idioms in context 10.3 (1.5) 9.1 (2r6)' Opaque idioms without context 1.73 (1.6) 1 (1 r2) Opaque idioms in context 3.9a (1.9) 720 (2.5)' \"Signicant differences between the group at good and poor comprehenders. Potockl It Laval: Reading and idiom Compmnensim 3435 Table 2. Example of the composition of a video in context with the opaque idiom to carry a torch (to be in love). Figure 2. Interaction between group, idiom transparency, and context. Part Description Details Transparent idioms Opaque idioms Scene location A public building in the background and an esplanade, trees, and benches - still image (sepia filter) in the foreground. 12 12 Presentation of the supportive context Paul is sitting on a bench, reading. Marie is standing a few meters behind, 10 10 Good comprehenders - dynamic video with voiceover (sepia filter) typing on her smartphone. Paul turns, and Mary looks at him. Voiceover accompanying the scene: "Paul comments that Marie has Poor comprehenders changed lately. She keeps looking at him and constantly sends him text messages." 3 Course of action Paul turns and looks at Mary, who looks up at him. He looks at the camera - dynamic video (natural color) and leans forward. diom production Fixed shot of Paul facing the camera, who says "I think Marie carries a torch fixed shot of the actor with online voice for me." (natural color) Request for interpretation Image freezes on Paul's face. - still picture with voiceover (natural color) Voiceover: "What does Paul mean?" With context Without context With context Without context The maximum score for each experimental condition effect was significant, F(7, 268) = 122.8, p