In an earlier post I made a brief reference to an article about proving that I had just begun to read. I’ve now read it.

Patricio Herbst, “Engaging Students in Proving: A Double-Bind on the Teacher” in *Journal for Research in Mathematics Education*, Vol. 33, Number 3, May 2002

It’s available from JSTOR but I couldn’t find it for free.

This is a very thought-provoking article. I have to confess I also found it quite frustrating stylistically. I had to read the whole thing, underlining and scribbling notes in the margins, reread several sections multiple times, and then reread the whole thing start to finish to feel like I fully understood what the man is saying.

I am going to attempt to distill the arguments in a concise and readable way. Afterward, I’ll provide some commentary.

**Herbst’s argument**

Herbst analyzes a classroom episode he observed and videotaped, of an advanced 9th grade integrated math class, in which the class tried to create a two-column proof of a simple proposition about angles. He sees the episode as “a typical case of a teacher engaging students in proving.” (p. 185) The proof had been assigned as an exercise for homework the previous night and the students had struggled with it. The teacher asked the class to construct a proof together; they made a little progress but then floundered. They began, alternately, to offer unproductive variants on the premises, and to assert the conclusion without adequate justification. The teacher intervened and suggested the key ideas to make progress. She framed her interventions as advice for future situations. Afterward, the students complained that the proof had been too hard, and the teacher said something to the effect that she wouldn’t assign too many problems like that on a test.

Herbst argues that:

a) When the teacher and students undertook the proof exercise as a class, they entered into a sort of “contract” where they were each expected to do a certain kind of thing.

b) The specifics of this contract placed conflicting demands on the teacher, and also prevented the students from taking certain steps that were needed to complete the problem.

c) This explains why the teacher intervened to give the key ideas in the proof, why she framed her interventions as she did, and why afterward, the students saw the proof as particularly hard, and why the teacher granted this.

More broadly, Herbst is making a case that two-column proof exercises tend to be doomed to lead to this sort of situation. Even more broadly, that any task that attempts to isolate proof as the skill being learned is similarly doomed. In other words, that in learning proof, form can’t be separated from substance. Learning about proof has to be integrated with learning the actual mathematical content.

Here is how his argument goes:

1) In the class, “proof” was understood as “transforming premises step-by-step into a conclusion by a sequence of logical deductions, with a reason cited for each step in the sequence.” With this understanding, the teacher’s job on the given occasion was to provide the students an opportunity to “do a proof,” and the students’ job was to “do a proof.” Thus the teacher and students entered into a “contract” with the following clauses:

Teacher’s Job: I’ll give you a task in which the hard part, the part that requires your thinking, is organizing the information into a chain of logical deductions. The task will be fair, i.e. you should be able to do it.

Students’ Job: We’ll take the givens and transform them step-by-step into the conclusion.

2) In order to do her job according to the contract, the teacher picked a task with the following features:

*it explicitly asked for a proof

*the proposition to be proved was stated at the outset

*the “givens” were identified at the outset

*the number of concepts involved was small

*a diagram was provided with a labeling that supported reading it a certain way

(Here is the task: the problem; the diagram. This is on p. 183; the problem came from *Integrated Mathematics 2* by Rubinstein et. al., published 1995 by McDougal Littell.)

The task was designed to try to keep the students’ work limited to *formally organizing ideas into a proof*. Its design overtly (e.g. via the statement of what was to be proved) and covertly (e.g. via the labeling of the diagram) suggested the ideas that the students were supposed to use, so they wouldn’t be responsible for coming up with these ideas and could concentrate on the formal task of organizing them into a proof.

3) In spite of the way the task was set up, the students floundered when they tried to do their job. One way to explain this is that the intended proof had two points where it required students to do something other than transform the givens step-by-step. One was to glean information from the diagram that was not explicitly stated in the givens (specifically, that angle ABC is composed of, so its measure is the sum of the measures of, angles CBD, DBE, EBF, and FBA); another was to write down an equation based on this information and manipulate it algebraically. Students did not produce these moves; this could be a product of the fact that, adhering to the letter of the contract, they thought they were supposed to just keep looking at the givens and transforming them, so these more flexible, broader-scope moves didn’t occur to them.

4) When these difficulties emerged, the teacher was in a double bind. On the one hand, if she intervened and said “you may need to look at these broader types of moves,” she would be giving away too much information; if the students then succeeded in completing the proof, it wouldn’t feel like they really did it themselves (more or less, this is what actually happened). On the other hand, if she didn’t intervene and allowed the students to keep working as they had been, the proof might never get completed. They would probably be left with frustration and the impression that the task was too hard. It would seem as though the teacher hadn’t done her job according to the contract.

5) The teacher dealt with this situation by intervening and framing her intervention as “advice for future proving endeavors.” By doing this she transformed the activity from “an opportunity for students’ joint production of a proof” to “an opportunity to learn a strategy for future use in proving.” (p. 199) This validated the time spent, but the students still didn’t really do the proving.

Herbst makes the case that the dynamics at play here are going to be at play any time the formal skill of “doing proofs” becomes its own curricular goal. Here is his general argument:

6) As long as the goal is the formal skill of “doing proofs,” the teacher needs to create proof tasks where the only thing the student is asked to do is organize ideas into a logical proof. If the task also requires the student to creatively generate the *ideas* for the proof, then the task isn’t fair – a kid might be perfectly able to organize a logical argument, but still fail, because they couldn’t think of these ideas. So the teacher needs to relieve the student of the burden of generating *ideas* for the proof by overtly and covertly giving these ideas to the student ahead of time. (For example, by stating the proposition to be proved at the outset, and/or by providing a diagram labeled a certain way.)

7) But if the task is organized in this way, then students are likely to fail to pick up on the covertly-suggested ideas, or correctly use the overtly-suggested ideas, because they are under the impression that what they are supposed to do is logically transform the premises step-by-step into the conclusion, but really they are also supposed to pick up on all these secret cues the problem is giving them. The form of a proof inevitably depends too much on the specific content for the students to be able to produce it just by trying to transform premises. When this breakdown happens, the teacher is in a double bind: s/he can’t intervene in the middle of the process to direct students to the hidden ideas without making them feel like they didn’t really do the work; but s/he can’t let them just continue barking up the wrong tree till kingdom come.

Herbst concludes that learning the skill of “proof” can’t be separated out from the mathematical content being proven. “Any sort of tools and norms that teachers can use to engage students in proving must allow room for the teacher to negotiate with the class what counts as proof in the context of the investigation of specific, substantive questions.” (p. 200)

**My thoughts**

A) The classroom excerpts in the article felt extremely familiar (both from classes I’ve taught and watched). In particular, the way the initial progress ground to a halt and then students either just transformed premises unhelpfully or stated the conclusion without justification.

Also familiar from my own practice, teaching a “proof” unit in an Algebra I class in 2001-2004, was the dynamic where you search specifically for problems where the ideas in the proof aren’t too hard to come up with, so students can concentrate on the task of organizing the proof and not get tripped up on generating these ideas. And how problematic this becomes if the students don’t come up with the ideas even when you tried to make them easy. At a few points while reading I felt confused by Herbst’s contention that choosing an exercise that “made the substance of the expected proof available to students” (p. 191) was something the teacher was doing in order to fulfill her “contract” – I mean, it didn’t *work*, right? The kids flopped anyway! So what good is it doing the teacher? But then I realized that I knew exactly what he was talking about from my own experience trying to teach a “proof unit.” I remember thinking thoughts like, “this problem doesn’t require too many difficult insights, so it’ll be good for getting them to focus on the logical structure.” Of course they flopped anyway with me too, and so it didn’t do me any good. But the important thing is that I felt compelled by that thought while choosing the problem. The fact that I turned out to be totally wrong doesn’t mean that I hadn’t used that consideration. It amounted to me choosing problems specifically to avoid asking my kids to have to demonstrate any creative insight.

With hindsight this seems like an insane way to teach proof now. Talk about playing small. Coming up with creative ideas is what makes proving things fun. How dare I claim to be teaching proof and deny kids the opportunity to touch the most exciting part of it? Of course, the alternative, to ask kids to actually develop creative justifications, will never work either if proof is ghettoized into a unit of Algebra I or Geometry. (I’ve been doing math for 9 years without being asked to generate a single original idea, and now all of a sudden you want me to be *creative*? WTF?) Looking for proof should be a daily or weekly part of every single math class, all the way down. (But it shouldn’t be too formal too soon. More on this in (B).)

B) I buy Herbst’s case that the kids’ overly rigid understanding of what they were supposed to be doing (“transforming the premises step-by-step into the conclusion”) probably contributed to their difficulty. But I have a lot more to say about what was in their way. The real problem the kids were having is what I’ve talked about before: in the class session described in the article, there is absolutely no connection between the kids’ attempts at proof and their *actual, real-live sense of what they know and don’t know*. In fact, these two things have been forcibly, violently separated. *That’s* why the can’t do it. To get concrete, here’s an excerpt (Andie is the teacher):

“Andie then asked, ‘What are we trying to prove?’ – a question that students could answer but then were unable to offer further ideas. So Andie asked, ‘How are we going to get to *FBD*? Do we know anything about *FBD*?’ A student’s response, ‘Maybe it’s a right angle,’ led Andie to ask whether the students knew ‘what makes up *FBD*.’ The answer, ‘*FBE* plus *EBD*,’ justified by the ‘whole and parts postulate,’ started a small discussion as to whether they were then entitled to say ‘*ABF* plus *EBD* equals ninety degrees’ or whether they ‘[didn’t] know it’ yet.” (p. 184)

*ARE YOU KIDDING ME?* In this classroom, as described, “we don’t know that yet” clearly doesn’t mean we don’t know it yet. Everybody in the room in fact does know that ABF plus EBD is 90, because they were told to prove it (well, the equivalent), so it must be true. Not only that but it looks in the diagram like it’s true. When the students say “we don’t know that yet,” what they mean is “we don’t know if the mysterious authority that certifies that proofs are correct is going to allow us to claim that yet.” What they don’t realize (why should they? it’s been forcibly hidden from them) is that the only mysterious authority that should be involved is *their own sense of conviction*.

After all, the ultimate source of mathematical authority in *the world* is the collective conviction of mathematicians. That’s why our standards of rigor have changed so much over time. Calculus was practiced for well over a hundred years before Cauchy bothered to prove its central theorems from a (fairly) precise definition of limits. It wasn’t until the late nineteenth century that Dedekind saw a need to provide a rigorous construction of the reals from the rationals, in order to be able to prove theorems about the reals (such as the fact that they are complete).

Historically, what has pushed the mathematical community’s sense of rigor forward is not an insistence on greater rigor from an outside source, but the encounter with new ideas and examples that caused a crisis of knowledge. Hippasus’ proof that the diagonal of a square is incommensurate with its side, when all the Pythagoreans thought all lengths were commensurate, led to Eudoxus’ theory of proportions. The rigorization of calculus (beginning with Cauchy) was driven by the recognition that Fourier series didn’t behave the way anybody expected them to. (At least, according to this awesome book by David Bressoud.)

How people learn about proof and rigor is the same whether we’re talking about a class or about human civilization. We learn to prove by being challenged to convince ourselves of things. We learn rigor by encountering examples and ideas that throw our assumptions into question. If you believe something, but then somebody (like your teacher) whose authority you trust says to you “you don’t know that yet,” you are stuck. Your mathematical soul is at an impasse. You believe it but you believe somebody telling you you don’t know it yet. You get the message you can’t trust your own reasoning. And in this state you will never, ever produce a coherent proof. If there’s a gap or a flaw in your reasoning, then to grow what you need is to be shown this flaw on your own terms. To see an example that contradicts your assertion; or to hear a counterargument that debunks yours – but not from the mouth of an authority you would dare trust without thinking through it.

This is why insisting on too much formality too early is bad for people who are learning how to prove. The need for formal rigor has to be *earned* through crises of knowledge. If we want our students to develop an appreciation for a formal proof we have to show them counterexamples to arguments they produced less carefully. If someone is insisting on formality from you when you don’t have any reason to doubt something less carefully argued, you will get the idea that proof has nothing to do with what makes sense to you, what *you* find convincing. But you can’t produce a proof without being guided by this.

All of this adds up to the case I’ve made before, that saying “prove that such-and-such is true” is the wrong problem for students first encountering proof. The minute you say it, they know it’s true; and this gets in the way of their natural mathematical reasoning process giving them a readout on what’s true. Herbst is arguing that we go for problems like this because of the “contract” that says we are supposed to give them a chance to work on “proof” as opposed to something else. If they also have to figure out what is even true, that could feel like we’re asking them to do more than just prove something. The problem is that they will never learn how to prove something if we don’t ask them for more.

C) Herbst writes that, in the episode he analyzed, the “contract” forced the teacher to make sure the proof was completed. If the class had not been able to complete the proof, then the students might have interpreted their difficulty as meaning that the problem was too hard for them. (Of course this happened anyway.) Herbst says that the fear of this possibility would have made the teacher feel pressed to intervene. (p. 195)

I’m not sure that the teacher’s sense that she needed to intervene to move the process along was based on this consideration, but I suspect he’s right that it was forced by the implicit “contract.” Most math classes that I have observed or taught make sure any question the class works on jointly is resolved that period. In this way we create the expectation that if a class undertakes a proof (or any other juicy math task) together, the teacher will make sure it gets finished. This trains the kiddies to be uncomfortable with irresolution, and quite possibly to experience a class segment that ends on a note of irresolution as the teacher’s failure to do her “job.”

Of course, the discipline of mathematics is rife with irresolution. Questions posed, worked on, and not yet answered, are its lifeblood. They are *fertile*. Recently I’ve been reading about the history of algebraic number theory, and I learned that it was Ernst Kummer’s work on Fermat’s Last Theorem that led not only to the entire field of algebraic number theory, but to the notion of an ideal that is totally fundamental to ring theory.

I don’t have a straightforward conclusion to draw but given the fertility of irresolution in the history of mathematics, doesn’t it make sense to think about how to make our students more comfortable with it?

* * * * *

This is the first of what I hope will be a series of posts that engage with provocative pieces about proof. But I have to put this off for a month or two and switch gears. For the group theory class I’ve been teaching, I have been researching the history of the theory of equation solving, from the ancient Babylonians thru the birth of Galois theory. So my next post is going to be about two landmark algebra textbooks: Muhammad ibn Musa’s *Compendium on Calculating by Completion and Reduction* (the original Arabic name for which is totally *the origin of the word ‘algebra’* – how awesome is that?), and Girolamo Cardano’s *The Great Art* which is the first published discussion of the general solution of cubic equations. I’ve read the first of these, just begun the second, and am totally excited to tell you about them. (In a week or hopefully not more than two.)

I will absolutely stipulate to all of this:

But I need some SERIOUS training in how exactly one goes about teaching that way. I wasn’t taught that way and none of my colleagues teaches that way. Sometimes I feel like I get close, because I make the kids investigate and measure and conjecture (today, for example: median of a trapezoid), but then I stop before asking them to prove it. Or I do say something like “Why should that have to be true? Can we come up with some kind of explanation?” But they have no idea how to even start and it feels unfair and scary to ask them to. It would not occur to them to draw a picture and extend the legs and think about similar triangles, in a zillion years.

So, yeah, I’m totally incompetent. The end.

independence presumably has to be learned

with actual independent work. i had a bike

too big for me or maybe it was my older

sister’s when i was a little kid and couldn’t

ride yet. “training wheels”. no good.

anyway not for me. so what i can pedal.

still can’t balance the bike. lo and behold

i get a smaller bike… and my legs, like

honest abe would have it, are “long enough

to reach the ground”. bingo. i won’t fall

and i’m not faking it. a couple days later:

total confidence in riding. still bike-phobic

though. you have to *maintain* the damn

things and they’re easy to steal.

meanwhile kids’ll look like clever hans

for cues from the teacher to get out

of thinking. and we’ll go right on giving

’em because the frustration is too awful.

if god forbid i were to try to teach

theorem proving in a school setting

and god provide me a free hand

doing it, maybe i’d start… it looks

pretty promising… with “critical

thinking” like they used to say

when it meant something and look

for *flaws* in proofs. “what’s wrong,

oh class, with *this* one?” and they

find all kinds of logical (and calculational

and outright “conceptual”) errors.

which, when they begin with

baby-step idea-free proofs,

i can then say “oh, wait…

we agreed that *this kind of thing*

is *such-and-such error*;

how can we fix this?” etcetera.

of course i don’t expect a quick fix so please

don’t think i’m proposing one. students will

continue playing “school games” in school

*in the nature of the case* and no level

of participation… not even writing up

unnecessarily obscure descriptions of

the process in expensive journals…

will change this. most of those who

say they *want* to change it are con artists

or worse moreover. crypto-authoritarians.

what then are we to do.

I don’t know if I discussed this with you, but I’ve certainly mentioned it to others: Numbers are wonderful for proof.

4, 16, 36, 64, … hmm all perfect squares, and all divisible by 4. It’s true the first four times, must always be true, right? Can we call this a law?

Factors of 6 are 1, 2, 3, 6 and factors of 9 are 1, 3, 9…. Even numbers like 6 have an even number of factors (6 has four) and odd numbers like 9 have an odd number of factors (9 has three), right? Can we call this a law?

3 and 5 are “twin primes” and so are 11 and 13 and so are 17 and 19 and 29 and 31 and also 41 and 43. It looks like every pair of twin primes except 3 and 5 surround a multiple of 6. Can we call this a law?

All the primes are odd, except 2 (which is even, making the fact that it is prime extremely odd), right? Can we call this a law?

All the rows of pascals triangle look like powers of 11 (ex the 0th row is 1, the 1st is 11, the 2nd is 121, the 3rd is 1331), right? Can we call this a law?

Each row of pascals triangle adds up to twice the row above it (ex 8, 16, 32), right? Can we call this a law?

Kids can prove (and disprove these sorts of things. Much more fun if they all look likely… but there are sometimes things to be disproven…

I don’t mind helping formalize this stuff… and they tend not to mind either…

But we don’t start this process knowing that our potential law is true. That helps. Now, I generally like to resolve these the same period. But not always. And while I don’t usually offer them problems they can’t solve, I merrily discuss problems that I am stuck on (I often try the problems in Mathematics Magazine from the MAA for junior faculty… about twice as hard as I can really handle, but occasionally I can make progress… and I tell kids exactly that, too)

Anyhow, I can introduce quite a bit of formality, in the wake of a proving/disproving exercise, that the kiddies willingly tolerate, or actively participate in. What am I missing? Definitions, axioms, postulates, theorems (as reasons). IOW, I’m not bringing them close to 2-columns.

Let them do that (deadly boring piece) when they are solving equations.

Jonathan

Great stuff!

But, someone has to play math with them BEFORE they arrive in a proving course.

If proving starts at the beginning of the geometry course, it will be hard not to recreate the classroom you describe above (or skip the proving part of geometry altogether, which is really the only reason we give the topic its own course and a full year)

Jonathan

Yes, that’s just it! Natural, informal mathematical argumentation has to be part of your vocabulary before formal proof can make sense.

[…] is something I was repeatedly and inchoately hollering about five years […]