You are on page 1of 13

On the Pleasures of Constraints

(Guest lecture for the 100 Drawings class at Bennington College)

1. Constraints and good problems


Let me dive in with word puzzles, which are a pastime of mine. Constraints absolutely
make word puzzles more fun. For example, a famous word game is the “change-a-word,”
in which you are given two words and you have to get from one to the other by changing
one letter at a time. So for example you might have to get from the word HAND to the word
CUFF by changing one letter at a time. You might do this as follows:

HAND
HARD
BARD
BARN
BURN
TURN
TURF
TUFF1
CUFF

Change-a-words are sort of fun, but it has always annoyed me that there is no constraint
on the number of steps you can take to get from word A to word B. For example, we might
have meandered around like this:

1
A kind of volcanic stone

1
HAND
SAND
SAID
SAIL
SOIL
COIL
COIN
CORN
CORD
CARD
HARD
BARD
BARN
BURN
TURN
TURF
TUFF
CUFF

Admittedly, there is a pleasure in doing things this way, a certain paradoxical satisfi-
cation in not making progress, but I can’t help but wonder about the solutions that get us
from word A to word B by the shortest route possible. In my own version of this game, I
make this a requirement. So for example, given BASH and CONK, we have to solve it in just
four steps:

BASH
BASK
BANK
BONK
CONK

In a way, the four-step constraint has made the game easier, because, unlike the tra-
ditional version, there are only 24 paths from A to B, instead of some untold gigantic
number of paths from A to B. On the other hand, the four-step constraint makes the
problem harder, because you have to think harder before you take a step. I specifically
construct the games so that you have at least two choices at most of the decision points,
but only one of those choices eventually leads to the solution. And you have to choose the
steps carefully, because you won’t get to revisit a letter.
As you solve more and more of these, the method of their construction begins to dawn
on you, and that makes them even more enjoyable; you get the sense that you’re dueling
with someone on the other side of the paper. There is also a little art (of a rather degenerate

2
sort) in that I always make sure that word A and word B have some sort of conceptual
relationship to one another, here one of synonymy.
Actually, it occurs to me as I write this that another sort of masterful solution would
present the longest possible route from A to B! I wonder what kind of a game that would
make? Obviously, one would have to require that no word be repeated. . . .2
Before I lose myself in that question, let me move on to another example.
I think you will agree that it would be a pretty dumb puzzle if I just asked you to
write a sentence. Any sentence. End of puzzle.
But there are all kinds of fun puzzles that result when we put constraints on the
sentences we have to write. I wrote a little puzzle book on this theme, and here are some
of the puzzles in it:[11]
1. “Tractor Pull.” What is the longest sentence you can write in which every word is
one letter longer than the previous word?
Pretty good solution: I am not sure where marine pelican colonies typically congre-
gate.
2. “Alphabetical sentences.” A sentence is “alphabetical” if all of the letters appearing
in the sentence appear in alphabetical order. For example, the sentence “I mow” is
alphabetical. What is the longest alphabetical sentence you can write, in terms of
total letter count? How about in terms of the total number of words?
Pretty good solution: Beg him not.
3. “Parts is parts.” The parts of speech are noun, pronoun, adjective, verb, adverb,
preposition, article, conjunction, and interjection. What is the shortest sentence you
can write (in terms of total letter count) that contains at least one instance of every
part of speech? (Actually, let’s leave out interjections; all they do is make the answers
sound corny.)
Pretty good solution: Am I not sad and in a fix?
4. “Monotone sentences.” A sentence is “monotone” if all of its words have the same
number of letters. For example, “Buy now” (n = 3) and “Consume liquids” (n = 7)
are both monotone sentences. For each value of n in turn (n = 2, 3, 4, . . .), write the
longest monotone sentence you can. What is the highest value of n for which you
can write a monotone sentence of any length?
Pretty good solution for the case of ten-letter words: Headstrong schoolboys ruthlessly
vandalized clientless fruiterers exhibiting unsellable tangerines.
As these examples might suggest, people who love solving problems usually love design-
ing them as well. The person I turn to next, the great 17th Century French mathematician
Pierre de Fermat, was no exception. (Figure 1.)
2
And what two words A and B have the shortest longest-path between them?

3
Figure 1: Pierre de Fermat. Source: http://webpages.math.luc.edu/∼ajs/courses/
100summer2002/homework/hw2.html.

2. Fermat’s challenge
Back in Fermat’s time, mathematicians would often issue challenges to one another,
there being a special rivalry between England and the Continent at that time. Here is an
excerpt from one of Fermat’s challenges to the mathematicians of Europe.[10]

February 1657
There is scarcely any one that sets forth purely arithmetical questions, and
scarcely any one that understands them. Is it not because arithmetic has hereto-
fore been treated geometrically rather than arithmetically? This is certainly in-
timated by many works of ancient and modern writers, including Diophantus
himself. . . .
Therefore arithmetic claims for itself the theory of whole numbers as its whole
estate. Arithmeticians (“children of arithmetic”) should strive either to advance
or restore it, which was only imperfectly represented by Euclid in his Elements,
and moreover not sufficiently perfected by those who followed him. Perhaps it
lies concealed in those books of Diophantus which the damages done by time
have destroyed.
To these, in order to show them the light which may lead the way, I propose
the following theorem or problem to be either proved or solved. Moreover, if
they discover this, they will admit that questions of this sort are not inferior to
the more celebrated ones from geometry, either in subtlety or in difficulty or in
method of proof.
Given any number not a square, then there are an infinite number of squares
which, when multiplied into the given number, make a square when unity is
added.
EXAMPLE. Given 3, a non-square number; this number multiplied into the
square number 1, and 1 being added, produces 4, which is a square.

4
Moreover, the same 3 multiplied into the square 16, with 1 added makes 49,
which is a square.
And instead of 1 and 16, an infinite number of squares may be found showing
the same property; I demand, however, a general rule, any number being given
which is not a square.

In a letter on the same topic from the same month:[10]

What is for example the smallest square which multiplied by 61 with unity added,
makes a square?
Moreover, what is the smallest square which, when multiplied by 109 and with
unity added, makes a square?
If you do not give me the general solution, then give the particular solution for
these two numbers, which I have chosen small in order not to give too much
difficulty.

In modern terms, Fermat proposes to solve for x in the equation

nx2 + 1 = y 2 , (1)

where n is given. If we rearrange this and examine one of Fermat’s specific challenges, we
are trying to solve for x in the equation

y 2 − 61x2 = 1 . (2)

So what’s the big deal? This equation is actually quite easy to solve for x. We are
even allowed to let y be anything we want, because the problem is only asking for x. So
why not just take y to be, say, 8, and the equation becomes

64 − 61x2 = 1 . (3)

A high school student could solve this for x; we get


r
63
x= ≈ 1.016261 . (4)
61
Easy, right? But before we conclude that 17th Century mathematicians were far behind
today’s high school students, we should point out the catch. When Fermat talks about
“arithmetic,” he actually means that all of the numbers in the problem have to be whole
numbers. According to the rules of the game, the solution x = 1.016261 . . . doesn’t count.
So that’s the constraint. The equation is trivial to solve over the real numbers, but it
becomes a minor motherfucker to solve when the solution is required to be a whole number.
Let me show you a picture of what the problem has become when we impose the integer
constraint (see Figure 2).

5
50

40

30

20

10

1 5 10 15 20

Figure 2: The blue dots indicate integer values of x and y. Although any point on the red
curve solves the equation y 2 − 61x2 = 1, Fermat’s problem asks us to find a blue dot that
lies on the red curve.

6
Just for the sake of illustration, let’s replace Fermat’s number 61 with 83; this will make
the numbers more manageable. With 61 replaced by 83, the equation to solve becomes

y 2 − 83x2 = 1 . (5)

A solution to this equation is given by x = 9, y = 82, because

822 − 83 · 92 = 6724 − 83 · 81 = 6724 − 6723 = 1 . (6)

A picture of this solution is in Figure 3.

85

82

70

5 9 15

Figure 3: The integer solution (9, 82) of 83x2 + 1 = y 2 .

7
Now, returning to Fermat’s challenge of 61, a few mathematicians were in fact able to
solve the problem.[5] It turns out that the first blue dot intersected by the curve is the one
at

x = 226,153,980 (7)
y = 1,766,319,049 , (8)

a solution given by several mathematicians, including several English mathematicians, who


thus preserved English pride in the face of Fermat’s French taunting.
The solution is pictured in Figure 4.

1766319054

y 1766319049

1766319045

226153976 226153980 226153985

Figure 4: The numbers x = 226,153,980 and y = 1,766,319,049 solve Fermat’s challenge


y 2 − 61x2 = 1.

In this case, the whole-number constraint has made the problem very difficult, but
the upside in this case is that when you do finally exhibit the solution, it appears to be a
felicitous gift from God. Only in truly difficult problems can the solution be a masterstroke,

8
and so we should choose our problems to match our interests and abilities but also our
ambitions.
Constraints can make problems harder, but on the other hand, there is often no problem
in the first place when they are absent. There wasn’t any sentence puzzle at all, until we
specified the constraint which the sentences must satisfy. Likewise, “Find the largest even
number” is not even a math problem; it’s nonsense. There is no largest even number!
But if I say “Find the largest even number that cannot be written as a sum of two prime
numbers,” now we’ve got a problem we can sink our teeth into.[8]
We’ll move on now, but if you want to look into Fermat’s problem further yourself, you
should know that Eq. (1) is known as “Pell’s Equation,” because because Leonhard Euler
mistakenly referred to it that way once. But the first serious work on the equation was
actually done by the Indian mathematician Brahmagupta (598-670). Brahmagupta did not
give a complete solution to the problem, but he made substantial progress. For example,
it was actually Brahmagupta who first solved the case of 83 that we looked at above, and
it is incredible to think that this was a thousand years before Fermat was born.[6]

3. Physical constraints
Physics is a great subject for constraints. In a physics class, many of the homework
problems come with constraint conditions. For example, how fast can you drive over a
hill, if your car must remain in contact with the road the whole time? There are lots of
problems like this.
But the more interesting constraints operate on a deeper level. Many of the most
profound statements of physics take the form of constraints—limits on our ability to move,
certainly, but also limits on what we can even know, imagine, or decide. Here is a short
list of such principles, some of which will doubtless be familiar to you:

1. The universal speed limit: To the best of our knowledge, nothing in the universe can
go faster than the speed of light.[9]

2. Energy can be neither created nor destroyed. (It just moves from place to place and
converts from one form to another.)

3. The law of action-reaction: You can’t touch without being touched.[12]

4. The second law of thermodynamics: You can’t build a perpetual motion machine.[2]

5. The Heisenberg Uncertainty Principle: You can’t know the position and momentum
of an atom with complete precision at the same time.[1] , [13]

6. The Grandfather Paradox: You can’t go back in time and kill your grandfather before
he was born, thus creating a causal paradox.[7]

9
Some of these “no-can-do” principles are much more solidly grounded than others. For
example, if we were to discover a way to go faster than the speed of light, then I think it’s
safe to say that the entire present edifice of physics would come crashing down. Actually,
that would be kind of fun. But if somebody were to discover a way around the second
law of thermodynamics, then I’m afraid that not only physics, but all of science—indeed,
perhaps rationality itself—would crack wide open.[4] The stakes may be high for time
travel as well, but as to whether we can change the past and create causal paradoxes, the
arguments both ways on that are speculative.
Talking about physical laws as constraints puts an interesting spin on things, but it
does make the universe sound more negative than I think of it as being. Einstein’s God
begins to sound like a scold: You can’t do this. You can’t do that. I prefer to think that,
just as in Fermat’s problem, and just as in one of my humble word puzzles, the constraints
we are working under, in the universe we’ve been given, make for a pretty good game.

Jason Zimba
Bennington, Vermont

10
Postscript. Some puzzles you might enjoy:

1. I made eight changes in getting from HAND to CUFF. Is this the shortest solution
possible?

2. From my blog:[14][15][16]

(a) Call a word “unambiguous” if it has only one definition in your dictionary of
choice. Here are some examples of unambiguous words (according to the Amer-
ican Heritage Dictionary, 4th Edition): pub. . . schadenfreude . . . anyone . . . logy
. . . vim . . . withered. (There are more examples on my blog.) Can you think of
more unambiguous words?
From my blog: “I think it would be fun to assemble hundreds of these words and
then use them to write poetry or short fiction. Would the paucity of meaning
and the poverty of connotation lead to flat writing? Or would every word appear
to be, in virtue of its specificity, le mot juste?”
(b) Change-a-word: INCH → OBIT. (It is not easy.)
(c) Homophones are words such as MAIN and MANE that are spelled differently but
pronounced the same. Recently I decided that the coolest homophones are surely
those that contain their own homophones. (Maybe call them “isophones”?) An
example would be BUSS, which contains its homophone BUS. How many examples
like this can you think of?
(d) A quickie. Starting with the ten-letter word STREAMBEDS, remove a letter to yield
a nine-letter word. Then remove another letter to yield an eight-letter word.
Continue in this way until you have a one-letter word.
(e) The same as above, starting with INSOLATING.
(f) Say that a letter is “special” if it can be used together with the letters A, F, and
W to form a four-letter word. (You are allowed to rearrange the letters.) Find all
of the letters that are “special” in this way. Then rearrange the special letters
to form a new word.
(g) The same as above, using the letters I, C, and L in place of the letters A, F, and
W.
(h) (Some math knowledge is required for this one.) I’m thinking of a game with a
curious feature. The probability of winning the game once is exactly the same
as the probability of losing the game a million times in a row. Are the chances
of winning this game one in a million, greater than one in a million, or less than
one in a million?
(Math wizards may wish to show that if “a million” is replaced by a large number
N , then the probability of winning is given approximately by W (N )/N , where
W is the Lambert W-function.)

11
(i) This one is based on the Paradox of the Liar.[3] . The classic paradox is the sen-
tence This sentence is false. The sentence is paradoxical because we cannot
assign it a truth value that respects its meaning. Also well-known is the para-
doxical pair of sentences, (The next sentence is true, The previous sentence
is false.) Again, there is no way to assign truth values to the sentences in a
way that respects their meaning.
One day a couple of years ago, it occurred to me that these two famous examples
can be seen as instances of a general pattern. We consider an ordered n-tuple of
sentences (S1 , S2 , . . . , Sn ) in which each sentence asserts the truth or falsehood of
the next. The last sentence closes the loop by asserting the truth or falsehood of
the first. Under what circumstances is such a collection of sentences paradoxical?
As a warmup exercise, you might want to verify that the following pair of
sentences is not paradoxical: (The next sentence is false, The previous sen-
tence is false.) There is no paradox here because we can read the first sen-
tence as true and the second sentence as false; this reading respects the meaning
of the two sentences.

3. Find a positive whole number x and a positive whole number y for which y 2 −5x2 = 1.

References
[1] J. Bub, Intepreting the Quantum World, Cambridge University Press, 1997. A good
book by a colleague of mine. Some of it is technical, but you can skim to get the gist.

[2] A.S. Eddington, The Nature of the Physical World, Kessinger Publishing, 2005. ISBN
978-1417907182. This book reprints the Gifford Lectures given by the great astro-
physicist Arthur Eddington at the University of Edinburgh during January to March
1927.

[3] The Internet Encyclopedia of Philosophy, http://www.iep.utm.edu/p/par-liar.htm. I


usually cite the Stanford Online Encyclopedia of Philosophy, but in this case the IEP
seemed more appropriate.

[4] E.T. Jaynes, “Where do we stand on maximum entropy?” available at


http://bayes.wustl.edu/etj/node1.html.

[5] H.W. Lenstra, Jr., “Solving the Pell Equation,” Notices of the AMS 49(2), February
2002.

[6] The MacTutor History of Mathematics Archive, http://www-groups.dcs.


st-and.ac.uk/∼history/HistTopics/Pell.html.

12
[7] I. Novikov, “Can We Change the Past?” in S.W. Hawking (Ed.), The Future of Space-
time, Norton, 2002. ISBN 978-0393020229. (Plausible answer: No.)

[8] T. Olveira y Silva et al., Goldbach Conjecture Verification (ongoing); see


http://www.ieeta.pt/∼tos/goldbach.html.

[9] R. Penrose, The Emperor’s New Mind, Penguin, 1991. ISBN 978-0140145342. Penrose
was my former thesis advisor at Oxford. Here is what the Amazon review has to say
about the book: “The overview of relativity and quantum theory, written by a master,
is priceless and uncontroversial. The exploration of consciousness and AI, though, is
generally considered as resting on shakier ground.”

[10] E.D. Smith, A Source Book in Mathematics, Dover Publications, 1984. ISBN 978-
0486646909.

[11] J. Zimba, Word Puzzles for the Seriously Smart, 2006, vanity-published at
www.lulu.com.

[12] J. Zimba, Force and Motion: An Illustrated Guide to Newton’s Laws, Johns
Hopkins University Press, 2009. http://www.amazon.com/Force-Motion-Illustrated-
Guide-Newtons/dp/0801891604

[13] J. Zimba, “Simple realism and canonically conjugate observables in non-relativistic


quantum mechanics,” Found. Phys. Lett. 11(6), 1998.

[14] J. Zimba, http://jzimba.blogspot.com/2008/04/some-puzzles-for-fun.html.

[15] J. Zimba, http://jzimba.blogspot.com/2008/05/four-letter-words.html.

[16] J. Zimba, http://jzimba.blogspot.com/2007/08/flop-on-pop.html. This post com-


bines word puzzles with a strange piece of writing on the theme of physical constraints
(or the lack thereof).

13

You might also like