Standard algorithms unteach place value

I found a page full of computations sitting around the house this evening. Naturally, I picked it up and gave it a look.

Griffin (10 years old, 5th grade) had been doing some multiplication in class today. Somehow his scratch paper ended up on our couch.

Here is one thing I saw.

Naturally I wanted to ask the boy about it. He consented.

Me: I see you were multiplying 37 by 22 here.

Griffin (10 years old): Yeah. But I got it wrong so I did it again with the lattice.

Me: How did you know you got it wrong?

G: I put it in the answer box and it was wrong.

It turns out they were doing some online exercises. There is an electronic scratchpad, which he found awkward to use with a mouse (duh), plus his teacher wanted to be able to see their work, so was encouraging paper and pencil work anyway.

I was really hoping he would say that 37 times 22 has to be a lot bigger than 202. Alas he did not.

Anyway, back to the conversation.

Me: OK. Now 37 times 2 isn’t 101. But let’s imagine that’s right for now. We’ll come back to that.

G: Wait. That’s supposed to be 37 times 2? I though you just multiplied that by that, and that by that.

He indicated 7 times 2, and then 3 times the same 2 as he spoke.

Me: Yes. But when you do that, you’ll get the same thing as 37 times 2.

A brief moment of silence hung between us.

Me: What is 37 times 2?

G: Well….74.

Let us pause to reflect here.

This boy can think about numbers. He got 37 times 2 faster in his head than I would have with pencil and paper. But when he uses the standard algorithm that all goes out the window in favor of the steps.

THE STEPS WIN, PEOPLE!

The steps trump thinking. The steps trump number sense.

The steps triumph over all.

Back to the conversation.

Me: Yes. 74. Good. I like that you thought that out. Let’s go back to imagining that 101 is right for a moment. Then the next thing you did was multiply 37 by this 2, right?

I gestured to the 2 in the tens place.

G: Yes.

Me: But that’s not really a 2.

G: Oh. Yeah.

Me: That’s a 20. Two tens.

G: Yeah.

Me: So it would be 101 tens.

G: Yeah.

I know this reads like I was dragging him through the line of reasoning, but I assure you that this is ground he knows well. I leading him along a well known path that he didn’t realize he was on, not dragging him trailing behind me through new territory. We had other things to discuss. Bedtime was approaching. We needed to move on.

Me: Now. We both know that 37 times 2 isn’t 101. Let’s look at how that goes. You multiplied 7 by 2, right?

G: Yup. That’s 14.

Me: So you write the 4 and carry the 1.

G: That’s what I did.

Me: mmmm?

G: Oh. I wrote the one

Me: and carried the 4. Yeah. If you had done it the other way around, you’d have the 4 there [indicating the units place], and then 3 times 2 plus 1.

G: Seven.

Me: Yeah. So there’s your 74.

This place value error was consistent in his work on this page.

Let me be clear: this error will be easy to fix. I have no fears that my boy will be unable to multiply in his adolescence or adult life. Indeed, once he knew that he had wrong answers (because the computer told him so), he went back to his favorite algorithm—the lattice—and got correct answers.

I am not worried about this boy. He is and he will be fine.

But I want to point out…I need to point out that this is exactly the outcome you should expect when you go about teaching standard algorithms.

If you wonder why your kids (whether your offspring, your students, or both) are not thinking about the math they are doing, it is because the algorithms we (you) teach them are designed so that people do not have to think. That is why they are efficient.

If you want kids who get right answers without thinking, then go ahead and keep focusing on those steps. Griffin gets right answer with the lattice algorithm, and I have every confidence that I can train him to get right answers with the standard algorithm too.

But we should not kid ourselves that we are teaching mathematical thinking along the way. Griffin turned off part of his brain (the part that gets 37 times 2 quickly) in order to follow a set of steps that didn’t make sense to him.

And we shouldn’t kid ourselves that this is only an issue in the elementary grades when kids are learning arithmetic.

Algebra. The quadratic formula is an algorithm. Every algebra student memorizes it. How it relates to inverses, though? Totally obfuscated. See, we don’t have kids find inverses of quadratics because those inverses are not functions; they are relations. If we did have kids find inverses of quadratics, they could think about the relationship between the quadratic formula:

and the formula for the inverse relation of the general form of a quadratic:

Calculus. So many formulas (algorithms) that force students not to think about the underlying relationships. If we wanted students to really think about rates of change (which are what Calculus is really about), we might have them develop a theory of secant lines and finite differences before we do limits and tangent lines. We might have Calculus students do tasks such as Sweet Tooth from Mathalicious (free throughout October!). There, students think about marginal enjoyment and total enjoyment.

On and on.

This is pervasive in mathematics teaching.

The results are mistaken for the content.

So we teach kids to get results. And we inadvertently teach them not to use what they know about the content—not to look for new things to know. Not to question or wonder or connect.

I’m telling you, though, that it doesn’t have to be this way.

Consider the case of Talking Math with Your Kids. There we have reports from around the country of parents and children talking about the ideas of mathematics, not the procedures.

Consider the case of Kristin (@MathMinds on Twitter), a fifth grade teacher, and her student “Billy”. Billy made an unusual claim about even and odd numbers. She followed up, she shared, we discussed on Twitter. Pretty soon, teachers around the country were engaged in thinking about whether Billy would call 3.0 even or odd.

But standard algorithms don’t teach any of that. They teach children to get answers. They teach children not to think.

I have read about it. I have thought about it. And tonight I saw it in my very own home.

Advertisement

The latest “Common Core” worksheet

You have seen this on Facebook.

My annotated version (Click to enlarge)

Ugh what a mess.

Please share the annotated version widely.

I’ll say what I have to say (comments closed) and move on. If you wish to discuss further, hit me up on Twitter or pingback to the blog. Want to talk in private? Click the About/Contact link up top.

Also, Justin Aion—middle school teacher extraordinaire—wrote up his views on the matter. You can read them over in his house.

Here goes…

The intended answer

Dear Jack,

You only subtracted 306 from 427, not 316. You need to subtract another 10 to get the correct answer of 111.

Sincerely,

Helpful student

The purpose of this task

I cannot say whether this was the right task for this child at this time because I do not know the child, the teacher or the classroom.

I can say the following:

• Analyzing errors is a useful way to encourage metacognition, which means thinking about your thinking. This is an important part of training our minds.
• The number line here is a representation of a certain kind of thinking—counting back. The number line is not the algorithm. The number line records Jack’s thinking. He counted back from 427 by hundreds. Then he counted back by ones. He skipped the tens. We can see this error because he recorded his thinking with a number line.
• Coincidentally, the calculation in question requires no regrouping (borrowing) in the standard algorithm, so the problem appears deceptively simple in its simplified version.
  
• This task is intended to help students connect the steps of the standard (simplified) algorithm with reasoning that is based on the values of the numbers involved. Why count back by three big jumps? Because you are subtracting 300-something. Why count back by six small jumps? Because you are subtracting something-something-6. Wait! What happened to the 1 in the tens place? Oops. Jack forgot it. That’s his mistake.

So what?

The Common Core State Standards do require students to use number lines more than is common practice in many present elementary curricula. When well executed, these number lines provide support for kids to express their mental math strategies.

No one is advocating that children need to draw a number line to compute multi-digit subtraction problems that they can quickly execute in other ways.

The Common Core State Standards dictate teaching the standard algorithms for all four arithmetic operations.

But the “Frustrated Parent” who signed that letter, and the many people with whom that letter resonated, seem not to understand that they themselves think the way Jack is trying to in this task.

Here is the test of that.

A task

What is 1001 minus 2?

You had better not be getting out paper and pencil for this. As an adult “with extensive study in differential equations,” you had better be able to do it as quickly as my 9-year old.

He knows with certainty that 1001 minus 2 is 999. But he does not know how to get the algorithm to make that happen.

If I have to choose one of those two—(1) Know the correct answer with certainty based on the values of the numbers involved, and (2) Get the correct answer using a particular algorithm, but needing paper and pencil to solve this and similar problems—I choose (1) every time.

But we don’t have to choose. We need to work on both.

That’s not Common Core.

That’s common sense.

[Comments closed]

Wrapping up the [#algorithmchat]

Intentions were good and initial interest was high for #algorithmchat.

And then people realized how incredibly boring algorithms are unless you’re really, really into them. And it was the end of the school year. Et cetera.

We did get a bit of Twitter back and forth, and Karl Fisch dove in twice. Which is awesome.

Let me know if I missed anything with the list below, which I believe to be the comprehensive collection of posts on the matter—in order by posting date.

Hopefully, this list will get others thinking and we’ll add to it. Find me on Twitter, or post your link in the comments.

May 6: Reading group. Overthinking My Teaching.

May 6: Algorithm nation. The Fischbowl.

May 12: Algorithms, quadcopters, and the CCSS-M. The Fischbowl.

May 15: Common numerator fraction division. Overthinking My Teaching.

May 22: What is “the standard algorithm”?. Overthinking My Teaching.

Various dates: Posts on algorithms. A collection of posts from David Wees: Thoughts from a Reflective Educator. [Technically, these were not written in response to our original article, but they certainly are on topic.]

Summertime (or anytime) reading recommendations

A friend asked for tips on getting started understanding some new domains in mathematics teaching the other day. An experienced high school teacher, he wants to know more about  elementary and middle school topics, especially fractions, place value and multiplication and division algorithms.

For obvious reasons (mainly that I won’t shut up about these topics), I was on his short list to ask for recommendations.

It occurred to me that others might be interested in this particular brain dump. So here it is, lightly edited. Enjoy.

—

Fractions. Entry level stuff on this is Connected Mathematics. In particular, Bits and Pieces 1, Bits and Pieces 2, and Comparing and Scaling. Any version of these units is fine. Work the problems from the student edition; have the teacher edition there for guidance.

I made major progress on understanding student thinking when I constrained myself to using only ideas that must have come earlier (i.e. in elementary school) and to those that had been previously developed. When I tried to appreciate the problems on their mathematical merit, or to build connections to my undergraduate mathematics knowledge, I didn’t make much progress that was useful to working with kids.

Then turn to Extending Children’s Mathematics (written by the Cognitively Guided Instruction team—CGI—and published by Heinemann). There is a lovely research perspective that should give you new ways to think about the CMP stuff.

More advanced perspectives are to be found in the work of the Rational Number Project (RNP), and there’s Susan Lamon’s book, Teaching Fractions and Ratios for Understanding. For contrast, read Hung Hsi Wu’s Math for Teachers curriculum. For extra credit, write a comparative analysis paper reconciling Wu’s work with CGI and with RNP; argue which has the greater influence on the Common Core fractions development.

Conspicuously absent from these recommendations is the “Essential Understandings” series from NCTM, published relatively recently. I find the writing style of these texts hard to process. Others may recommend them, and if so, perhaps you ought to take them more seriously than I have been able to.

Place value. There is an oldish JRME piece by Karen Fuson, the CGI folks and another research team about place value. It’s a seminal piece and totally worth your time. There is no one book I can recommend; my exploration of the conceptual landscape of place value has been idiosyncratic and informed more by small pieces of others’ research work combined with my own classroom experience and experiments. Most of that is documented on this blog.

The “Orpda” number system that Cady and Hopkins wrote about (and which I bastardized as “Ordpa”) was foundational to these explorations. Short, short article but the ideas opened a whole new space for me in thinking about what it means to learn place value.

The Young Mathematicians at Work book on number sense, addition and subtraction is pretty good. But those articles and the blog are better starting points.

Multiplication and division algorithms. I am trying to recall how I came to know the algorithms I know. I have to say that these steps I cannot really retrace.  I am loathe to recommend digging through Everyday Math for them, because things are so diffuse; it’s hard to get the right book in your hand in that curriculum to learn any one particular thing.

The Kamii piece I recommended a while back is good. It was published in the 1998 NCTM Yearbook on algorithms. Sybilla Beckmann’s Mathematics for Elementary Teachers book is good, too.

But looking back at my standard algorithm diatribe last week and trying to think about what small set of resources would prep someone else to build a similar case (or to counter it), I am less clear than I am about fractions or place value. I do not know what this says about my knowledge, nor about the topic.

What is “the standard algorithm”? [#algorithmchat]

Richard Skemp wrote, in “Relational Understanding and Instrumental Understanding,” about faux amis—those pesky words in other languages that look like words you are familiar with, but which mean something else entirely. Skemp argues that the word understand is like this—different people use it to mean completely different things. This leads to misunderstanding

And so I fear it is with the standard algorithm.

I have heard it said that the use of this phrase (repeatedly) in the Common Core State Standards was a compromise (although I cannot find a source for this—leave any breadcrumbs you can find in the comments, won’t you?) It would satisfy some parties who believe that the standard algorithm is an essential seawall against the encroaching fuzzy math tide, while leaving the precise nature of the standard algorithm unspecified would appease those who argue that alternative algorithms are helpful in developing and maintaining children’s number sense.

But if a compromise owes its precise nature to the fact that different parties will interpret the terms of the compromise differently, has there really been a compromise? Have we really made an agreement when we disagree about its meaning?

What is an algorithm?

Karen Fuson and Sybilla Beckmann, in their “Standard Algorithms in the Common Core State Standards” cite a CCSSM Progression document.

In mathematics, an algorithm is defined by its steps, and not by the way those steps are recorded in writing.

Hyman Bass, in his article from Teaching Children Mathematics, “Computational Fluency, Algorithms, and Mathematical Proficiency: One Mathematician’s Perspective” agrees.

An algorithm consists of a precisely specified sequence of steps that will lead to a complete solution for a certain class of computational problems.

So far, so good. We have accord on the meaning of algorithm.

What is the standard algorithm?

The definite article in the phrase the standard algorithm seems to be important to the alleged compromise I referred to.

Here, for example, is Hung-Hsi Wu on standard algorithms.

[T]he essence of all four standard algorithms is the reduction of any whole number computation to the computation of single-digit numbers.

Wu states the following steps for the standard algorithm (.pdf) for multidigit multiplication.

To compute say 826 × 73, take the digits of the second factor 73 individually, compute the two products with single digit multiplier— i.e., 826 × 3 and 826 × 7 — and, when adding them, shift the one involving the tens digit (i.e., 7) one digit to the left.

He explicitly allows for moving left-to-right, as well as inclusion of zeroes instead of shifting. But explicit attention to place value in the process of working the algorithm seems to be proscribed.

Contrast this with the following figure (click for full-size version) from Fuson and Beckmann.

This figure is labeled “Written methods for the standard multiplication algorithm , 2-digit x 2-digit”. Note in particular methods D (lower left) and F (upper right). Method D shows that we are thinking 6 x 9 tens as we work the algorithm. Method F suggests that we are thinking 6 x 90 as we work.

But wait. The lattice method is an example of the standard algorithm?

Recall that an algorithm is defined by its steps. In Wu’s standard algorithm, you may proceed from left to right, or from right to left; either is acceptable. The lattice has both left/right and up/down steps, and you may do the single digit multiplication steps in absolutely any order.

I cannot imagine that Wu would count the lattice as a standard algorithm, and I seriously doubt he would count partial products (method D) in that category.

All of this got me thinking about whether there are any non-standard algorithms for multi-digit multiplication in the viewpoint that Fuson and Beckmann present. Pretty much every multiplication algorithm I know is in that Fuson and Beckmann figure. Every one except the Russian Peasant Algorithm, that is.

an alternative

I have argued that the compromise of using the standard algorithm but not specifying the standard algorithm in the Common Core is problematic because different people mean different things by it. The lattice is explicitly counted in the standard algorithm by Fuson and Beckmann, but our agreement on what constitutes an algorithm (a precisely defined series of steps) implies that the lattice constitutes a different algorithm from (say) partial products. Both cannot be the standard algorithm.

But here is an alternative. What if Common Core, instead of using the language of the standard algorithm used the following construction: an algorithm based on place-value decomposition.

In this case, 5.NBT.B.5 would read:

Fluently multiply multi-digit whole numbers using an algorithm based on place-value decomposition.

This construction would seem to include all of the algorithms in Fuson and Beckmann’s figure; it would make clear that the Russian Peasant Algorithm does not count; and it would be more transparent than the standard algorithm.

Until and unless I receive cease-and-desist notifications, I will go ahead and use this version in everything I do.

For your convenience, I have rephrased the various citations below. You can thank me later.

4.NBT.B.4 Fluently add and subtract multi-digit whole numbers using an algorithm based on place-value decomposition.

5.NBT.B.5 Fluently multiply multi-digit whole numbers using an algorithm based on place-value decomposition.

6.NS.B.2 Fluently divide multi-digit numbers using an algorithm based on place-value decomposition.