When Do Young Children Make Inferences from Prose?

Abstract

1980, 51, 906-908. The present study was an attempt todetermine whether young children generate inferences while they read or only later in responseto tasks which require inferences. Children from grades 2 and 6 read passages, each of whichcontained 2 premise statements which could be combined to form a simple transitive inference.Immediately after the presentation of each passage, the children were asked to read and rapidlyverify the truth value of statements based on information presented in individual sentences, oron information which could be inferred from a combination of 2 sentences. 2 indices of thetiming of inference making were used—1 based upon the position of premise statements in thepassages and the other based upon the probability of correctly verifying inferences as a func-tion of accuracy on the component premises. These measures indicated that children from bothgrades generated inferences while they were reading.Adults easily and regularly make infer-ences which are based on and go beyond thefactual information contained in texts (Kintsch1974), but almost nothing is known about thedevelopment of this important reading compre-hension skill in children. Although several re-cent lines of investigation indicate that youngchildren have basic inference-making abilities(Bryant & Trabasso 1971; Kail, Chi, Ingram, D Paris & Carter 1973; Stein &Clenn, Note 1), we do not know whether chil-dren use these abilities while reading or onlylater in response to tasks which require infer-ences. Therefore, the purpose of this study wasto determine whether or not children makeinferences while they read.The basic procedure was as follows. Fif-teen middle-class children from grade 2 and 20from grade 6 read out loud 10-sentence pas-sages which were projected on a screen, onesentence at a time, at the child's own pace.After completing each story, sentences basedon the story were presented, one sentence at atime. The child read each sentence out loudand then rapidly pushed one of two buttons toindicate whether or not it was true. A photo-cell triggered by light from the projector, a pro-jector controller, and two millisecond timersenabled us to keep track of the total readingtime for each story and the time it took to readand respond to each verification sentence. Wetherefor had an overall measure of readingspeed and measures of the speed and accuracywith which information from the stories wasverified. Some of the verification sentences re-quired an integration of information from twosentences in the story. The children's perfor-mance on these sentences served as an indexof their inference-making abilities.Two approaches were used to determinewhen inferences were made. The first involvedvarying the positions of the two critical sen-tences in each passage which, when combined,form the inference. These two sentences ap-peared either in passage positions 2 and 7 or inpositions 6 and 7. If an inference based onthese two sentences is made while reading,there should be little effect of the position ofthe first critical sentence (position 2 or 6) oninference verification latency. That is, in eithertype of passage, an inference made immediate-ly after reading sentence 7 should result in amemory representation whose accessibility isindependent of the accessibility of the compo-nent sentences upon which it is based. This isconsistent with the text processing model ofThis research was supported by a grant from the Wisconsin Research and DevelopmentCenter Royalty Fund. Requests for reprints should be sent to Fred Danner, Department ofEducational Psychology, University of Kentucky, Lexington, Kentucky 40506.