Blogs About Reading

Shanahan on Literacy

Timothy Shanahan

Literacy expert Timothy Shanahan shares best practices for teaching reading and writing. Dr. Shanahan is an internationally recognized professor of urban education and reading researcher who has extensive experience with children in inner-city schools and children with special needs. All posts are reprinted with permission from Shanahan on Literacy.

What Does It Take to Teach Inferencing?

August 9, 2021

Teacher question: I am reaching out to see if you can clarify for me and possibly point me in the direction of a resource(s) where I can read more about the differences between predicting, inferring, and drawing conclusions. Our curriculum was developed in house and is very skill/strategy based.

In Virginia, our state tests operationalize reading in the following way:

  • predicting is making an informed guess about what happens next using text evidence and schema
  • inferring is reading "between the lines" to a given point in text using text and schema to understand what is happening in the text
  • drawing conclusions is projecting forward using text and schema

Shanahan's response:

I’m happy to distinguish these three concepts, but it won’t help.

It won’t help you teach better.

It won’t help your students read better.

You are interests in making sure you are appropriately teaching these reading skills so that your students will comprehend well. But the problem is teaching those skills aren’t likely to do that.

Only one of these has a clear research record (inferencing) and that operationalization is gobbledygook.

Written messages — texts — are not so complete or explicit to allow readers to make full sense of them without filling some gaps or making some connections. Authors don’t tell everything. They imply an awful lot. Inferences are used to make sense of those implications.

But inferences are complicated.

There are lots of ways to characterize inferences … as demonstrated by the late Tom Trabasso.

There are, for instance, forward and backward inferences. Predictions and drawing conclusions are usually examples of forward inferencing. Readers draw a forward inference based on the textual information provided up to that point in the text. Backward inferences require that the gap be filled by information an author hasn’t yet revealed. Forward inferencing requires that you remember enough information the author provided so that when something is lacking you can fill the gap. You can’t possibly know where the author is going to leave gaps, so the more coherent and complete your memory is the smoother things are going to go. Backward inferencing is more complicated because you must spot the gap and realize that you don’t know how to fill it; then you must be vigilant for the needed info when it arises.

Another way to think about inferences has to do with the source of the information needed for filling the gap or making the connection. The critical information may have been provided by the author earlier in the text or it may come from readers’ own prior knowledge. Frequently, the information may come from both (and perhaps it will be necessary to coordinate information from more than one bit of the text).

Here’s a fairly simple example:

Mary and John went to the movies.

He asked her if she wanted popcorn.

Figuring out who asked who seems rather straightforward, but how do you go about it? You must first recognize that it isn’t entirely clear who was doing the talking, and then you must draw on your knowledge of the world. In our culture, people named John are usually boys so John must have asked Mary about the popcorn. This required information from working memory (sentence 1) and the reader’s world knowledge.

A third way to think about inferences is the functions that they fill. The John and Mary sentence would be an example of inferences used for text connection or slot filling. This is how we make cohesive links; these inferences fill a linguistic function, connecting chains of synonyms that operate across a text.

But inferences play other functions, too.

Look at this example:

He plunked down $10.00 at the window.

She tried to give him $5.00 but he refused to take it.

So, when they got inside, she bought him a large bag of popcorn.

To grasp this requires an inference that they must be at the movies. Making that inference transforms this into a scene that most of us could visualize.

This kind of inference allows the reader to create a model of what is going on. Other inferences have other purposes (e.g., explanatory, predictive, associative).

What’s my point?

Only to show the difficulties and complications inherent in the inferencing process during reading.

Given the different ways inferences work — requiring recognition of different kinds of omissions or implications, depending on information drawn from different sources (e.g., working memory, background knowledge), accomplishing different interpretive goals, and operating in different ways — what are the chances that you could teach students to infer effectively?

One wonders whether there is any sense to distinguishing inferences, predictions, and conclusion drawing. Do these odd distinctions (without psychological reality) help or hinder kids? My bet is that they learn to ignore such nonsensical teaching.

Nevertheless, many experts claim that asking students inferential questions improves their reading comprehension. Research hasn’t been especially supportive of those claims, however. Although we can ask questions that require inferences, that doesn’t mean answering them would have any discernible impact on future inferencing. Inferences are too complex and multi-faceted and texts too diverse to allow such teaching to help.

It’s not that questions cannot or do not shape readers’ attention during reading. Just that inferences by the descriptions provided couldn’t possibly do this.

An example from the research is a study done by Reynolds and Anderson. They found that if they asked questions about the quantities expressed in a text, readers started paying more attention to the quantitative info during future readings.

That makes sense.

I remember blowing a midterm in a history class. I hadn’t recognized that the prof wanted the dates of the events we’d studied. My attention shifted to the dates after that, and my final exam went much better. Questions tip kids off as to what information is important and studies show that such guidance is effective because it gets the readers to spend more time on certain information.

That can only work in those cases in which the information is recognizable. As a reader, I had no trouble recognizing the historical dates, so devoting more attention to them was a breeze. Telling kids to pay attention to inferences, to practice answering drawing conclusion questions, or making predictions during reading aren’t likely to provide much of a learning payoff since this guidance fails to direct students to specific workable actions that they can take during reading.

It is possible to improve reading comprehension through some kinds of inference training, but not the general kind emphasized in state standards.

For instance, when reading fiction, it is important to consider the characters’ motivations. What do they want? What are their goals? Why are they taking those actions?

Students can be taught about motivation and the importance of this can be stressed instructionally. Students can learn to ask themselves why characters do what they do and to pay attention to how they know that. If the author tells such information explicitly, they should pay particular attention to that. If the author doesn’t reveal what is driving a character’s action then an inference is needed.

Likewise, students can be taught to do something similar with causation when reading science. In this case, there are even signal words that can tip a reader off to causes and effects (e.g., because, so, consequently, therefore, thus, since) and distinguishing words that refer to other kinds of relationships (e.g., correlated, similar to, associated). Again, the reader needs to learn that such text demands special attention to cause and effect, whether the author states it explicitly or depends upon the reader to recognize that relationship.

Those earlier-mentioned lexical or cohesive inferences (e.g., repetitions, synonyms, pronouns) can be taught, too. Teaching students to track a character or idea across a text to support comprehension is doable and it is successful at improving students’ ability to comprehend. Learning to connect pronouns, even in the earliest grades, can have real payoffs.

Essentially, I’m suggesting treating inferencing not as a skill but as a strategy for intentionally making sense of text. Students who take specific steps to consciously identify motivations, causes, and linguistic connections are likely to comprehend better. Those taught to “infer” by the definition you provided would not.

Reading comprehension instruction should focus on guiding students to think actively about the ideas in text. Previewing predicting, self-questioning, visualizing, rereading, identifying text structure have all been found to be beneficial because they prescribe actions that encourage students to spend more time thinking about the ideas in texts. Teaching students to identify certain kinds of information when they read — information that is actually identifiable in lots of texts — can help as well (that’s where inference training succeeds).

(Prediction is an odd duck. I would recommend it as a strategy for motivating oneself through a text, but not one that I would want to question students on to evaluate their prediction ability. I can’t think of many texts in which predictions are truly important to gaining meaning. And, with regard to drawing conclusions, no studies of inferencing or comprehension have pursued that colloquial terminology).


Baumann, J. F. (1986). Teaching third-grade students to comprehend anaphoric relationships: The application of a direct instruction model. Reading Research Quarterly, 21(1), 70-90.

Cain, K., Oakhill, J., & Lemmon, K. (2004). Individual differences in the inference of word meanings from contexts: The influence of reading comprehension, vocabulary knowledge, and memory capacity. Journal of Educational Psychology, 96, 671-681.

Elleman, A.M. (2017). Examining the impact of inference instruction on the literal and inferential comprehension of skilled and less skilled readers: A meta-analytic review. Journal of Educational Psychology, 109(6), 761–781.

García-Madruga, J. A., Elosúa, M. R., Gil, L., Gómez-Veiga, I., Vila, J. Ó., Orjales, I., Duque, G. (2013). Reading comprehension and working memory's executive processes: An intervention study in primary school students. Reading Research Quarterly, 48(2), 155-174.

Hamada, A. (2015). Effects of forward and backward elaboration on lexical inferences: Evidence from a semantic relatedness judgement task. Reading in a Foreign Language, 27(1), 1-21.

Magliano, J.P., Trabasso, T., & Graesser, A.C. (1999). Strategic processing during comprehension. Journal of Educational Psychology, 91(4), 615-629.

Reutzel, D.R., & Hollingsworth, P.M. (1991). Reading comprehension skills: Testing the distinctiveness hypothesis. Reading Research and Instruction, 30(2), 32-46.

Reynolds, R.E., & Anderson, R.C. (1982). Influence of questions on the allocation of attention during reading. Journal of Educational Psychology, 74(5), 623-632.

Trabasso, T., & Magliano, J.P. (1996). Conscious understanding during comprehension. Discourse Processes, 21, 255-287.

See comments here >

Add comment

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
"When I say to a parent, "read to a child", I don't want it to sound like medicine. I want it to sound like chocolate. " — Mem Fox