Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.



reply to Lou Talman
Posted:
Jun 8, 1997 1:00 PM


Howard Swann gave me permission to forward this to the APCalc group.
*****************************************
Hello, Lou Talman;
David Klein forwarded a note you sent on calculus curriculum that mentioned my article in the MAA Monthly.
You state: ``The whole point of the criticism socalled "reformers" level at the MVT is that the standard presentation *isn't* a proof. ''
The points I attempted to make in the article are the following:
1. A glimpse of the power of mathematical proof is an exciting intellectual endeavor, worth any number of group `discoveries' based on putting numbers representing `real life problems' into a graphing calculator. By abandoning the attempt to do this, we discard one of the strongest arguments in favor of studying calculus.
However, in beginning multisemester calculus, selfcontained rigorous proofs are possible only when discussing limits at a point; `zooming' graphic calculators are alleged to eliminate the conceptual difficulties.
Yet, when placed in suitable historical context, we can make a discussion of the masterful pair <Dedekind's `real numbers', and Weierstrass' proof that guesses for limits are *unique* under the epsilondelta definition> fascinating indeed. For this pair provided the framework for humankind to finesse, (but not eliminate) the profound mysteries of human attempts at understanding infinity. This is precisely the point that Bertrand Russell made in the quotation in the Monthly article; I urge you to read his entire essay.
Today's `reformers' presumably assume that students are incapable of understanding the beauty and cleverness of these (admittedly difficult) ideas. Certainly the conventional presentation often falls short, yet my own experiments suggest that if the reformers got busy, they should be able to come up with a way to do this. One has only to see, for example, the current PBS programs on fractals and the Mandelbrodt set to realize how fascinating the question of what `on earth' `infinitesimally small' could mean. I invite you to `zoom' away on the example I cited in the Monthly article, and welcome the chance to send you the current `handout' I use when introducing the idea of limit if you are interested.
2. With regard to the MVT: I urged the following treatment in the Monthly article: (a) First a geometric argument; (b) A discussion of why the geometric argument is not a `proof' (c) A discussion of Rolle's theorem as a special case; a proof to be given here by CITING the extreme value theorem and celebrating it as a triumph of global results obtained from the pair <real numbers, Weierstrass' epsilons and deltas.> The mystery of `number' can easily be revisited here, for in what way is there a maximum for (x^22)^2, when no computer has ever seen \sqrt{2}? (e) A return to the usual proof of the MVT using Rolle's theorem; rigorous this time by CITING Rolle's theorem.
Although lengthy, such a sequence of arguments is of value, for it gives a paradigm for how we traditionally hope that most (nonmathematicsmajor) students should understand their calculus and postcalculus math courses. A clear distinction should be made between what is and is not a proof; important proofs should be given where accessible, which will usually entail CITING (and celebrating) some global result.
If this experience is not encountered occasionally, subsequent math courses (with their classical texts) will be extremely frustrating to students and they never get a real glimpse of the power of mathematical proof, although `reform' is at hand as new texts omitting any notion of mathematical proof proliferate and it is asserted that somehow the `meaning' of mathematics is made clear.
You continue with ``But, in point of fact (and Swann's mistaken "counterexamples" to the contrary notwithstanding..N.B. DO TELL ME THE DIFFICULTIES HERE!), no proof given in the traditional calculus sequence requires the full strength of the MVT. ...the Increasing Function Theorem...or ...the Racetrack Principle suffice in every case. ''
I return to the example of `arc length' cited in the Monthly article: The traditional argument is the following: The sum for approximating arc length of y = f(x) is \sum_{i=1}^n \sqrt{(x_{i}  x_{i1})^2 + (y(x_{i})  y(x_{i1}))^2}.
One then uses the MVT applied to assumed continuously differentiable f(x) to write y(x_{i})  y(x_{i1}) = f'(x_{i}^*)(x_{i}  x_{i1}) for some x_{i}^* between x_{i} and x_{i1}. This is then converted to a RIEMANN sum of form \sum_{i=1}^n \sqrt {1 + f'(x_{i}^*)^2} \delta x_{i} We then CITE Riemann's central theorem, crucial to calculus, that if a function is continuous, then Riemann sums converge to the integral no matter where the x_{i}^* is located and the proof is concluded.
I can find no attempt to formally state Riemann's theorem in the Harvard Calculus, for example; the argument for the formula for arc length is based on the skilled use of the symbol `~', which is no proof whatsovever.
How can one contrive any `proof' of the formula for arc length using a `Racetrack' argument?
Yours truly,
H. Swann



