Final Report Summary - WMCIG (Inter- and Intra-Individual Differences in Knowledge-Based Inference Generation During Reading Comprehension: The Role of Working Memory Capacity and Utilization)
The present research project was designed to explore developmental, individual and motivational differences in various cognitive processes that support reading comprehension. Specifically, five text comprehension processes—predictive and bringing inference generation, text retention, text reactivation, and text and inference suppression—were examined in skilled comprehenders at different ages, in high- and low-span adult comprehenders, in three types of disabled comprehenders (children with dyslexia, children with specific reading comprehension deficit, and adolescents with attention disorder), and in high- and low-motivation adult comprehenders.
To address these goals, participants were asked to read 3-6 sentence narratives and named text or inference probes appeared in different positions within the texts. By comparing naming times of probes in inference-evoking and control texts (the latter did not suggest the probed inferences or not included the probed text information), the researcher examined the (re)activation (naming facilitation) or suppression (naming inhibition) of inference and text information during reading comprehension.
Results indicated that performance observed for high- and low-span groups differed only when participants were divided based on a listening-span test, but not when divided based on operation- or symmetry-span tests. High-listening-span participants generated predictive inferences faster than low-listening-span participants. Moreover, high-span participants generated more bridging inferences than low-span participants, presumably due to enhanced retention and reactivation of inference-evoking textual information. These findings support the inference activation, text retention, text reactivation, and discourse-domain-specific hypotheses of the role of working memory in inference generation.
This pattern of results was generally replicated with normal children at five age groups (fourth, fifth, sixth, eight-to-nine, ten-to-eleven grades). The only difference observed in the performance of the different age groups concerned text suppression. The researcher found that the ability to suppress text information which is temporarily less relevant for comprehension is improved with age.
The results with the disabled children showed impaired processing of both inferential and textual information. Specifically, the participants with ADHD and poor comprehenders exhibited difficulties to generate predictive and bridging inferences, due to deficient retention and reactivation of the critical text information. The results with the children with dyslexia were too noisy to establish firm conclusions, because their probe reading times were too long. Apparently, the probing method is less appropriate to study children with dyslexia.
Note that the adolescents with ADHD exhibited normal accuracy in answering inference and text questions presented after reading, although in a slower manner compared to control group. These findings suggest that readers with ADHD do not spontaneously generate inferences during reading, but can do so when they are prompted by questions presented immediately after few sentences. These conclusions have important implications about the development of comprehension strategies for readers with ADHD.
Finally, an effect of motivation on inferential and textual processing was found only as a function of working memory capacity. Low span participants showed elevated activation of inference and text information when they were motivated to do so. In contrast, high span readers reduced activation of inference and text information, particularly of bridging (backward) information, when they were actually motivated to generate those inferences. These puzzling findings required further research to illuminate this interesting interaction of motivation and WM. Nonetheless, the findings are encouraging in regard to the low-span group, suggesting that they can enhance the utilization of their working memory resources when they are encourged to generate inferences during reading comprehension.
In addition to the behavioral studies, this fellowship enabled the researcher to develop and test the computation model he designed during his post-doctoral studies. Specifically, in this project he extended the WM module of the model and implemented new features such as the simulation of suppression and reinstatement of information during reading. Using these new features, the researcher succeeded to simulate the majority of the data obtained with high- and low-span readers in the main behavioral study. These simulations further illuminated the role of WM in reading comprehension using a computational methodology.
Taken together, this project produced important theoretical knowledge about the role of WM in reading comprehension and its development during the years. In addition, it produced vital practical information about the text processing of low-span and disabled comprehenders, and the possible effect of motivation. Any future remediation programs that will be developed to support text processing by struggling comprehenders should consider the findings of this research.
It is worth noting that beyond scientific contribution, this research project has facilitated the integration of the researcher as a new faculty member in the Learning Disabilities division of the School of Education in Bar-Ilan University, Israel. Specifically, this research supported the development of his research skills in relevant areas (i.e. reading comprehension and learning disabilities) and his mentoring skills via the supervision of four graduate students participated in this project. Thus, the EU funding was extremely essential for the researcher's reintegration in an Israeli academic institute.
To address these goals, participants were asked to read 3-6 sentence narratives and named text or inference probes appeared in different positions within the texts. By comparing naming times of probes in inference-evoking and control texts (the latter did not suggest the probed inferences or not included the probed text information), the researcher examined the (re)activation (naming facilitation) or suppression (naming inhibition) of inference and text information during reading comprehension.
Results indicated that performance observed for high- and low-span groups differed only when participants were divided based on a listening-span test, but not when divided based on operation- or symmetry-span tests. High-listening-span participants generated predictive inferences faster than low-listening-span participants. Moreover, high-span participants generated more bridging inferences than low-span participants, presumably due to enhanced retention and reactivation of inference-evoking textual information. These findings support the inference activation, text retention, text reactivation, and discourse-domain-specific hypotheses of the role of working memory in inference generation.
This pattern of results was generally replicated with normal children at five age groups (fourth, fifth, sixth, eight-to-nine, ten-to-eleven grades). The only difference observed in the performance of the different age groups concerned text suppression. The researcher found that the ability to suppress text information which is temporarily less relevant for comprehension is improved with age.
The results with the disabled children showed impaired processing of both inferential and textual information. Specifically, the participants with ADHD and poor comprehenders exhibited difficulties to generate predictive and bridging inferences, due to deficient retention and reactivation of the critical text information. The results with the children with dyslexia were too noisy to establish firm conclusions, because their probe reading times were too long. Apparently, the probing method is less appropriate to study children with dyslexia.
Note that the adolescents with ADHD exhibited normal accuracy in answering inference and text questions presented after reading, although in a slower manner compared to control group. These findings suggest that readers with ADHD do not spontaneously generate inferences during reading, but can do so when they are prompted by questions presented immediately after few sentences. These conclusions have important implications about the development of comprehension strategies for readers with ADHD.
Finally, an effect of motivation on inferential and textual processing was found only as a function of working memory capacity. Low span participants showed elevated activation of inference and text information when they were motivated to do so. In contrast, high span readers reduced activation of inference and text information, particularly of bridging (backward) information, when they were actually motivated to generate those inferences. These puzzling findings required further research to illuminate this interesting interaction of motivation and WM. Nonetheless, the findings are encouraging in regard to the low-span group, suggesting that they can enhance the utilization of their working memory resources when they are encourged to generate inferences during reading comprehension.
In addition to the behavioral studies, this fellowship enabled the researcher to develop and test the computation model he designed during his post-doctoral studies. Specifically, in this project he extended the WM module of the model and implemented new features such as the simulation of suppression and reinstatement of information during reading. Using these new features, the researcher succeeded to simulate the majority of the data obtained with high- and low-span readers in the main behavioral study. These simulations further illuminated the role of WM in reading comprehension using a computational methodology.
Taken together, this project produced important theoretical knowledge about the role of WM in reading comprehension and its development during the years. In addition, it produced vital practical information about the text processing of low-span and disabled comprehenders, and the possible effect of motivation. Any future remediation programs that will be developed to support text processing by struggling comprehenders should consider the findings of this research.
It is worth noting that beyond scientific contribution, this research project has facilitated the integration of the researcher as a new faculty member in the Learning Disabilities division of the School of Education in Bar-Ilan University, Israel. Specifically, this research supported the development of his research skills in relevant areas (i.e. reading comprehension and learning disabilities) and his mentoring skills via the supervision of four graduate students participated in this project. Thus, the EU funding was extremely essential for the researcher's reintegration in an Israeli academic institute.