How Should We Teach Making Inferences

Making inferences is an important part of reading comprehension. An author does not necessarily put everything in print for everyone to read. Yes, you may infer at this moment and extrapolate that doing so requires too many words and simply leads to a wordy or long-winded article. Readers prefer concise sentences. Consequently, as early as the elementary years, students need to be taught to make inferences. Inferences should not be equated to guessing or reading too much between the lines. Extending too much from what is either directly stated or implied in text can be misleading when it comes to reading non fiction especially science literature. Inferences must be based on what one actually has read.

Testing for reading comprehension often falls under three categories (Factual Recall, Bridging Inference, Pragmatic Inference) as decsribed in a previous post on this blog, Stress, Working Memory, and Reading Comprehension. As an example, given the following passage:
"The waiter dropped a plate. He quickly went to get a dustpan and broom."
Here are possible questions for each of the three categories.
  • Factual Recall: "Who dropped a plate?" or "What did the waiter drop?"
  • Bridging inference: "Who is 'He' in the second sentence?"
  • Pragmatic inference: "Where did the plate land, and did it make a mess?"
"Bridging" is basically text-based (T) while "pragmatic" is elaborative (E) as it draws from either one's experience or general knowledge. Text-based inferences require a reader to see the connection between sentences (either with the use of pronouns or connecting the ideas between sentences even when words like "therefore" or "because" are missing). Elaborative inference, on the other hand, embellishes what is written. For example, reading the phrase "the sky is blue" can create an image on the reader's mind that "it is simply a gorgeous sunny day".

Being able to make inferences, both text-based and elaborative, is necessary for good reading comprehension. Teaching students to make inferences can be challenging. First, teachers sometimes cannot tell the difference between making inferences and reading too much between the lines. Making inferences is not so much predicting how the story will end in a fictional piece. Making inferences is about making sense, not making predictions. Second, seeing how one sentence leads to the next one is indeed a step ahead of both vocabulary and parsing a single sentence. Third, elaborative inference depends on one's background knowledge. Still, there are interventions designed to improve a student's inference skills and a recent metastudy by Elleman shows how effective these interventions are. Interventions that are included in this study teach students the following:
  • locate relevant information in text to generate an inference
  • integrate information across text
  • provide evidence in the text of their answers to inferential questions
  • activate and integrate background knowledge withinformation in the text
The effect sizes, Hedges' g (measured in terms of a standard deviation), are quite remarkable especially for less-skilled readers:

Above copied from
Examining the Impact of Inference Instruction on the Literal and Inferential Comprehension of Skilled and Less Skilled Readers: A Meta-Analytic Review.
Elleman, Amy M.
Journal of Educational Psychology, Feb 13 , 2017, No Pagination Specified. http://dx.doi.org/10.1037/edu0000180

The studies included in this pair are summarized in the following table:
Above copied from
Examining the Impact of Inference Instruction on the Literal and Inferential Comprehension of Skilled and Less Skilled Readers: A Meta-Analytic Review.
Elleman, Amy M.
Journal of Educational Psychology, Feb 13 , 2017, No Pagination Specified. http://dx.doi.org/10.1037/edu0000180
The studies that show large effect sizes have been encircled in red. Judging from the treatment hours for each of these highly effective interventions, one cannot really see a correlation between time spent and effect size. The study that has the largest effect size (Carnine, Stevens, Clements, & Kameenui (1982) 2.24) takes only two hours while interventions with 15 and 16 hours produce only 1.59 and 1.16, respectively. This perhaps shows that it only takes a short time to improve a student's skills in making inferences. This improvement is more likely to be due to understanding how pronouns work and how sentences get connected to each other (text-based inference). The lack of further improvement with increasing instructional time probably lies in elaborative inferencing for this requires knowledge, which obviously requires much more than 15 hours to grow.




Comments