Doug Lemov's field notes

 

Recently, I shared an overview of the topics in the forthcoming book on reading I’m writing with Colleen Driggs and Erica Woolway: seven key principles of reading instruction that should inform what we do in k-12 classrooms. (Note we’ll be discussing these principles at our Nashville workshop Dec. 5  & 6)

Today, I’m going to share a bit more about the another of those principles, the idea that once students are fluent, background knowledge is the most important driver of understanding and comprehension. A common misunderstanding about reading comprehension is that it involves transferable skills like making inferences that once learned can be applied to other texts. Unfortunately there is little evidence that the skill translates and significant evidence that the skills happen naturally when readers have sufficient background knowledge to disambiguate texts.

One of the themes of Middlemarch, George Elliott’s classic 19^th century novel, is Mr. Casaubon’s fruitless pursuit of a concept he refers to as “the key to all mythologies.” A scholar, he imagines a single understanding that will illuminate the true meaning of every tale. He spends his life toiling at the task of finding this universal key.

It’s hopeless of course. The novel reveals his delusions. When he dies, his admiring wife, Dorothea, at last reads his papers and can see that the project was absurd from the start, but the reality proved all but impossible to acknowledge because the dream was so beautiful. It was a ‘chimera,’ something so alluring the believer desperately wants it to exist even when the facts are telling him it cannot be so.

The belief in transferable skills is perhaps the most common chimera among teachers of reading. Imagine a handful of universal tools we could teach students and in so doing allow them to understand every text they read. Who wouldn’t seek out “the key to all inferences,” for example, knowing that once mastered this skill would allow them to unlock what was unspoken in every story? Or the “key to main ideas’ which would allow our students after a bit of diligent study to grasp the gist of any passage we put in front of them for the rest of their lives. Who among us would not dream such a beautiful dream?

The problem of course is that for all the beauty of the dream, the evidence is squarely against it. While we make inferences constantly while reading, and while doing so clearly assists with comprehension, practicing strategies like making inferences doesn’t help much and there’s no evidence that the ability to make inferences well transfers from one book to another. The opposite in fact.

“People don’t decide that they’re going to make these inferences, the mind just makes them happen,” Daniel Willingham writes. This is perhaps one reason why “practice brings no benefit to reading-comprehension strategy use.[1]” Summarizing the finding of recent studies, he writes beyond a very small amount of introduction to the idea: “There was no evidence that increasing instructional time for comprehension strategies—even by 400 percent!—brought any benefit.[2]”

The reason for this is that our ability to inference is a function of our prior knowledge.

Here’s an example from a third-grade classroom we recently visited. The class was reading Charlotte’s Web when they came across this scene:

 

“But Charlotte,” said Wilbur, “I’m not terrific.”

“That doesn’t make a particle of difference,” replied Charlotte. “Not a particle. People believe almost anything they see in print. Does anybody know how to spell terrific?”

“I think,” said the gander, “It’s tee double ee double rr double rr double eye double see see see see see.”

“What kind of acrobat do you think I am?” said Charlotte in disgust.

 

The teacher paused and asked why Charlotte was disgusted. Two students responded. The first said because the gander always talked too much. The second because the gander always said everything three times. Both of which are true and both of which are wrong if the goal is to explain why Charlotte was disgusted.

Perhaps students didn’t understand how to infer a character’s point of view from her words. This would be the assumption in a lot of classrooms and the result would be a lesson (or a series of lessons) on the “skill” of inferencing.

But the source of the problem was revealed when the teacher asked: Who knows what an acrobat is?

There was a smattering of two or three hesitant hands. A boy who’d raised his responded: “It’s a little bit like a magician, I think.”

Charlotte, for those who haven’t read Charlotte’s Web, is disgusted because she intends to write the word in a spider web and the gander’s very long spelling of the word implies lots of work hanging precariously from a web for her. But if you don’t know what an acrobat is, you cannot know that. The problem was not a skill problem. Knowledge cues an inference and without it, no further explanation of how to make an inference or what an inference is will help much.

That the knowledge enables the inference is an inconvenient fact. It means we can’t just explain and practice and have students get better at inferencing. There is no Casaubon-like short cut. We instead have to go the long way around and make sure students have the background knowledge they need to make better sense of what they read.

As Dylan Wiliam writes in Creating the Schools Our Children Need, “The big mistake we have made in the United States is to assume that if we want students to be able to think, then our curriculum should give our students lots of practice thinking. This is a mistake because what our students need is more to think with.”

A classic study by Recht and Leslie and known as The Baseball Study[3] demonstrates this.

The authors divided 64 7^th and 8^th grade students into two groups based on their reading levels: weak readers and strong readers. But they also divided those groups again, based on whether the students knew a lot about baseball. Now they had four groups. Good readers who knew a lot about baseball; good readers who knew very little about baseball; weak readers who knew a lot about baseball; and weak readers who knew very little about baseball.

They gave them a passage to read. Here are the first few lines:

Churniak swings and hits a slow bouncing ball toward the shortstop. Haley comes in, fields it, and throws to first, but too late. Churniak is on first with a single…

After students read the passage, the researchers tested the four groups of students to see how much of the passage they understood.  Some of the results were exactly what you’d expect:

 

The students with high reading ability and strong knowledge of baseball had no trouble with the passage. They got almost all of the answers correct. By contrast the weak readers with weak knowledge of baseball really struggled. They got less than half the questions right, scoring little above the level you’d expect them to get if they were merely guessing.

The surprise was in the two middle groups. The students with low reading ability but strong knowledge of baseball did better than the high ability readers with little knowledge of baseball. Quite a lot better in fact- they were a few points behind the top group and scored almost 30% better than the students who were “better” readers but knew less about the topic of the passage.

It’s a study that has been repeated many times with students reading about different topics, and it demonstrates quite elegantly that you read well and successfully when and because you have background knowledge of what you are reading about. To return to Dylan Wiliam’s point: if we want students to understand and think more deeply about what they read we should focus on knowledge and not waste time trying to teach them abstract skills like making inferences or finding the main idea.

The reason why this is the case has to do with what you might call the inherent ambiguity of every text. As Daniel Willingham recently pointed out[4], every sentence is to some degree ambiguous. An author never tells you everything. If he or she did, reading would be incredibly tedious and meaning making would become nearly impossible.

Here’s an example. We’ve rewritten the previous sentence to eliminate ambiguity and ensure your accurate comprehension of our exact point regardless of background knowledge:

 An author of a book, article, poem, treatise or other example of a written text in English or any other language never tells you—in this case the reader but also people like the reader who might also be reading or be imagined to be reading said text—everything that he or she intends to communicate in that text because if he or she did reading would become incredibly tedious due to the overwhelming loads of marginally relevant information jammed into the sentence in order to clarify every possible misunderstanding or gap in perception and in the end every text would read like a dense contract between two massive corporate entities seeking to eliminate any possible gray area to their transaction.

Authors by necessity always make assumptions about what readers know and assume they will fill in gaps. This is always true, even when they aren’t deliberately leaving blanks and ambiguities for stylistic or artistic reasons. Understanding any text always involves “disambiguating” it.  

Here’s a very short text, the ambiguities of which make an interesting case study:

The wooden box was massive. She placed her bear on the ground. It was going to be hard to carry.

The ambiguities are probably not even apparent to you at first because you resolved them simply and easily with inferences you didn’t know you made. One ambiguity is that her “bear” is not a real bear but a teddy bear. There’s no way she’d be carrying a real bear. As a result you probably inferred that “she” was a child. Another ambiguity involves resolving what noun the pronoun “it” refers in the second sentence. Grammatically it’s just as plausible for it to refer to the teddy bear as the box. But you knew that it referred to the box. It doesn’t make sense for a teddy bear to be hard to carry but a wooden box, yes. Especially for a child. You disambiguated because you had knowledge–the weights of common things; that someone with a teddy bear is probably a child–the author assumed you would have.

But what if the author assumes you know something you don’t. Like in this sentence:

For pudding she allowed herself some cake.

If you’re a reader in England, where “pudding” means roughly the same thing as “dessert” does to an American, the sentence is easily disambiguated. She ordered cake after dinner. It was a small indulgence. This is implied by the phrase “she allowed herself.”

But if you lack that background knowledge, the sentence is nonsensical, even if you are a very good reader.

Your working memory is busy wrestling with what seems like a riddle—for pudding she had cake? How could she have cake for pudding? Does the author mean “instead of pudding” maybe? (If you are English imagine the sentence “For custard she had cake” to get the general sense for how an American might experience the text). You not only failed to understand the first part, you might not have even noticed or perceived the subtle implications of the word allowed which tells you quite a bit about the “she” in the sentence and the small indulgence which perhaps she wouldn’t ordinarily consider, of permitting herself a piece of cake: perhaps she is conscious of her weight. Perhaps she is conscious of money. When your working memory is overloaded wrestling with something you don’t immediately understand, your perceptiveness of other details is also degraded. Meaning is interrupted everywhere. There is no ‘story’ in the sentence for readers who don’t have background knowledge about what “pudding.”

Comprehension, this is to say, is knowledge-based. We make better inferences, we perceive more, we have working memory to think more deeply, when we know more about what the author assumes we know enough about. We can’t do those things when we don’t.

In a 2020 review of the literature, Reid and collegues “consistently found that higher levels of background knowledge enable children to better comprehend a text. Readers who have a strong knowledge of a particular topic, both in terms of quantity and quality of knowledge, are more able to comprehend a text than a similarly cohesive text for which they lack background knowledge. This was evident for both skilled and low skilled readers.”[5]

“Controlling for other factors, knowledge plays the largest role in comprehension. The more a reader knows about a topic, the more likely they are to successfully comprehend a text about it,” Literacy specialist Jennifer Walker of Youngstown State writes,[6] again summarizing the broader literature on the topic.

In fact, the connection between knowledge and all types of high-order thinking is clear if often overlooked. “Data from the last 30 years lead to a conclusion that is not scientifically challengeable: thinking well requires knowing facts…The very processes that teachers care about most—critical thinking processes like reasoning and problem solving [and reading!]—are intimately intertwined with factual knowledge that is in long-term memory,” Daniel Willingham writes. “Most people believe that thinking processes are akin to those of a calculator. A calculator has a set of procedures available (addition, multiplication, and so on) …. If you learn a new thinking operation (for example, [how to make inferences]), it seems like that operation should be applicable to all [settings]. The human mind does not work that way. The critical thinking processes are tied to the background knowledge.”

Somehow this fact does not seem to be getting through to schools. Perhaps it’s the chimerical nature of skills-based instruction. We just can’t let go of how beautiful it would be if we could just give students a super-skill, a universal key and the hours we spend chasing that dream is time taken from far more productive tasks.

Again, we don’t blame teachers for this. Even at the highest levels of policy, even in schools of education, the misconception prevails.

In his book Why Knowledge Matters E.D. Hirsch tells the story of France, which had among the best and most equitable school systems in Europe before the French replaced their knowledge-rich curriculum with a skills-intensive approach. Results declined steeply overall and gaps between rich and poor students expanded. Scotland followed suit a few years later. It “downgraded the status of knowledge and adopted a competence-based approach, emphasizing the development of transferable skills and interdisciplinary learning,” Sonia Sodha wrote in the Guardian[7].  The results? Scottish students lost an average of about 6 months progress in reading on the 2022 PISA while inequality increased. The lowest-status group fell twice as fast (a drop of 20 points) as in the highest-status group,”[8] Lyndsay Paterson, a professor of education at the University of Edinburgh observed.

You can find a thousand voices on the internet telling us to choose less knowledge and more skills, but even asking what the right balance is between skills and knowledge is the wrong question, Daisy Christodoulou has observed. It is like asking what the right balance is between ingredients and cake.he ingredients become the cake; the knowledge becomes the skill. If you want deeper thinking, if you want better reading, start by building students’ knowledge. You can do quite a bit of that if you don’t spend hours explaining what an inference is.

[1] https://www.ascd.org/el/articles/beyond-comprehension

[2] Increasing over a bare minimum of a few minutes, this is to say. In other words everything you need to know about what an inference is can be taught in less than a single lesson. Beyond that you are wasting your time.

[3] https://psycnet.apa.org/record/1988-24805-001. Also described here: https://www.coreknowledge.org/blog/baseball-experiment-two-wisconsin-researchers-discovered-comprehension-gap-knowledge-gap/

[5] https://www.tandfonline.com/doi/full/10.1080/02702711.2021.1888348

[6] https://education.ohio.gov/getattachment/Topics/Learning-in-Ohio/Literacy/Literacy-Academy/2023-Literacy-Academy/Language-Comprehension-Components-Necessary-for-Reading-Comprehension.pdf.aspx?lang=en-US. Walker refers to Cromley & Azevedo, 2007; Ozuru, Dempsey & McNamara, 2009 among other studies.

[7] https://www.theguardian.com/commentisfree/2023/dec/10/scottish-schools-have-tumbled-from-top-of-the-class-this-is-what-went-wrong

[8] https://reformscotland.com/2023/12/pisa-2022-in-scotland-declining-attainment-and-growing-social-inequality-lindsay-paterson/