Tuesday, October 14, 2014

Section 5-2: Types of Quadrilaterals (Day 48)

Section 5-2 of the U of Chicago text covers the various types of quadrilaterals. There are no theorems in this section, but just definitions. The concept of definition is important to the study of geometry, and in no lesson so far are definitions more prominent than in this lesson.

The lesson begins by defining parallelogram, rhombus, rectangle, and square. There's nothing wrong with any of those definitions. But then we reach a controversial definition -- that of trapezoid:

A quadrilateral is a trapezoid if and only if it has at least one pair of parallel sides.
(emphasis mine)

Just as with the definition of parallel back in Section 1-7, we have two extra words that distinguish this from a traditional definition of trapezoid -- "at least." In other textbooks, no parallelogram is a trapezoid, but in the U of Chicago text, every parallelogram is a trapezoid!

To understand what's going on here, let's go back to the first geometer who defined some of the terms in the quadrilateral hierarchy -- of course, I'm talking about Euclid:


Of quadrilateral figures, a square is that which is both equilateral and right-angled; an oblong that which is right-angled but not equilateral; a rhombus that which is equilateral but not right-angled; and a rhomboid that which has its opposite sides and angles equal to one another but is neither equilateral nor right-angled. And let quadrilaterals other than these be called trapezia.

Of course, the modern term for "oblong" is rectangle, and a "rhomboid" is now a parallelogram. The word "trapezia" is actually plural of "trapezium." In British English, a "trapezium" is what we Americans would call a trapezoid, but to Euclid, any quadrilateral that is not a parallelogram (or below on the quadrilateral hierarchy) is a "trapezium." But the important part here is that to Euclid, a square, for example, is neither a rectangle (oblong) nor a rhombus. He makes sure to say that a rectangle (oblong) is "not equilateral," and that a rhombus is "not right-angled." And of course, neither a rectangle nor a rhombus is a parallelogram (rhomboid).

These are called "exclusive" definitions. For Euclid, there was no quadrilateral hierarchy -- each class of quadrilaterals was disjoint from the others. But since the days of Euclid, more and more geometry texts have slowly added more "inclusive" definitions.

One of the first inclusive definitions I've seen was the definition of rectangle. It was mentioned on an episode of Square One TV, when a Pacman parody character named Mathman was supposed to eat rectangles, and then ate a square because "every square's a rectangle." I would provide a YouTube link, but I haven't found the link in years and it doesn't come up in a search. (I even remember someone posting in the comments that just as for me, his first encounter of the inclusive definition of rectangle was through watching that clip when it first aired so many years ago!)

But many of my family members were also teachers, and one relative gave me an old textbook that still mentioned some exclusive definitions. In particular, it declared that a square isn't a rhombus. A little later, my fifth grade teacher then taught the inclusive definition of rhombus. I then blurted out that a square isn't a rhombus, then actually brought the old text to school to prove it! She replied, "Wow!" but then, if I recalled correctly, told me that this definition was old, and that by the new definition, a square is a rhombus. And so all modern texts classify the square as both a rectangle and a rhombus, and that all of these are considered parallelograms.

So we see that there is a tendency for definitions to grow more inclusive as time goes on. (We see this happening in politics as well -- for example, the definition of marriage. But I digress.) And so we see the next natural step is for the parallelogram to be considered a trapezoid.

One of the first advocates I saw for an inclusive definition of trapezoid is the famous Princeton mathematician John H. Conway. He is best known for inventing the mathematical Game of Life, which has its own website:


But Conway also specializes in other fields of mathematics, such as geometry and group theory (which is, in some ways, the study of symmetry). Twelve years ago, he posted the following information about why he prefers inclusive definitions:


The preference for exclusive definitions arises, I think, from
what I call "the descriptive use". Of course, one wouldn't DESCRIBE
a square table as "rectangular", since that would wantonly use
a longer term to convey less information. So in descriptive uses,
there's a natural presumption that a table called "rectangular"
won't in fact be square - in other words, a natural presumption
that the terms will be used exclusively.

But the descriptive use is unimportant to geometry, where the
really important thing is the truth of theorems. This means that we
should use a term "A" to include "B" if all the identities that
hold for all "A"s will also hold for all "B"s (in the way that
the trapezoid area theorem holds for all parallelograms, for

You might worry about the consistency of switching to the
inclusive use while other people continue with the exclusive one.
But there can be no consistency with people who are inconsistent!
I've seen many geometry books that MAKE the exclusive definitions,
but none that manage to USE them consistently for more than a few

Indeed, Conway advocated taking it one step forward and actually abolishing the trapezoid and having only the isosceles trapezoid in the hierarchy! After all, there's not much one can say about a trapezoid that's not isosceles -- just look ahead to Section 5-5. There's only one theorem listed there about general trapezoids -- the Trapezoid Angle Theorem, and that's really just the Same-Side Interior Angle Consequence Theorem that can be proved without reference to trapezoids at all. All the other theorems in the lesson refer to isosceles trapezoids. In particular, the symmetry theorems in the lesson refer to isosceles trapezoids. (Recall that Conway's specializes in group theory -- which as I wrote above is the study of symmetry.) I suspect that the only reason that we have general trapezoids is that they are the simplest quadrilateral for which an area formula can be given.

This is now another digression from Common Core Geometry, so I'll just provide another link. Notice that here, Conway also proposes a hexagon hierarchy based on symmetry. There's also a pentagon hierarchy, but there are only three types of pentagons -- general, line-symmetric, and regular -- just as there are for triangles. It's easier to make figures with an even number of sides symmetric.


Another advocate of inclusive definitions is Mr. Chase, a high school math teacher from Maryland. I see that he is so passionate about the inclusive definition of trapezoid that he devoted three whole blog posts to why he hates the exclusive definition of trapezoid:




One reason Chase states for using inclusive definitions is that it simplifies proofs:

When proving that a quadrilateral is a trapezoid, one can stop after proving just two sides are parallel. But with the exclusive definition, in order to prove that a quadrilateral is a trapezoid, you would have to prove two sides are parallel AND the other two sides are not parallel.

Regarding some of the others to whom I refer regularly, Dr. Wu uses the inclusive definition:

"A quadrilateral with at least one pair of opposite sides that are parallel is called a trapezoid. A trapezoid with two pairs of parallel opposite sides is called a parallelogram."

while Dr. Mason uses the exclusive definition:

"A trapezoid is by definition a quadrilateral with precisely one pair of parallel sides."
(emphasis Dr. M's)

So which definition should I use for trapezoid? Well, this is a Common Core blog, so the definition favored by Common Core should have priority over all other definitions. The following is a link to the information that will appear on the PARCC End-of-Year Assessment for geometry:


And right there in the column under "Clarifications," it reads:

i) A trapezoid is defined as “A quadrilateral with at least one pair of parallel sides.”

And that plainly settles it. The PARCC Common Core assessment uses the inclusive definition of trapezoid, and so it's my duty on a Common Core blog to use the Common Core definition. Of course, we notice that this is the definition given by PARCC -- but so far I've seen no information on what definition Smarter Balanced is using. It would be tragic if PARCC were to use one definition and Smarter Balanced the other. But as I can't say anything about Smarter Balanced, I will use the only definition that's known to be on a Common Core test, and that's the inclusive definition. The fact that this definition is already used by the U of Chicago is icing on the cake.

There is one problem with the inclusive definition of trapezoid, and that's when we try to define isosceles trapezoid. The word isosceles suggests that, just as in an isosceles triangle, an isosceles trapezoid has two equal sides -- the sides adjacent to the (parallel) bases. But in a parallelogram, where either pair of opposite sides can be considered the bases -- the sides adjacent to these bases are also equal. This would make every parallelogram an isosceles trapezoid. But this isn't desirable -- an isosceles trapezoid has several properties that parallelograms in general lack. The diagonals of an isosceles trapezoid are equal, but those of a parallelogram in general aren't. But the diagonals of a rectangle are equal. So we'd like to consider rectangles, but not parallelograms in general, to be isosceles trapezoids.

This dilemma is mentioned in the comments at one of the Chase links. It's pointed out that there are two ways out of this mess -- we may either define isosceles trapezoid in terms of symmetry, as Conway does, or we can use the U of Chicago's definition:

"A trapezoid is isosceles if and only if it has a pair of base angles equal in measure."

Some don't like this definition, because it violates linguistic purity -- the word isosceles comes from Greek, and it means "equal legs," not "equal angles." But as it turns out, it's a small price to pay to make the quadrilateral hierarchy and other theorems work out. And besides, any geometer who calls a nine-sided polygon a nonagon should just shut up about linguistic purity!

There's one definition in this lesson that I've neglected to mention -- the kite. As it turns out, there are two definitions of kite, one exclusive and one inclusive. The inclusive definition makes every rhombus (and therefore every square) a kite. Naturally, those who prefer the exclusive definition of trapezoid, like Dr. M, also prefers the exclusive definition of kite, while others take inclusive definitions of both trapezoid and kite. (Wu is silent on this issue -- he doesn't mention kites on his site at all.)

The fact that exclusive definitions make proofs longer -- as mentioned by both Conway and Chase -- is noticeable when we look at Dr. M's lesson on kites (Lesson 6.6 on his site). Since he is using the exclusive definition of kite, Dr. M must make sure that every property that a kite has, such as having a pair of equal opposite angles, applies only to one pair of angles and not to the other -- otherwise the figure would be a parallelogram (indeed a rhombus) and not a kite. But if we were to use the inclusive definition, we don't need to fear that the figure is a rhombus because a rhombus is still considered to be a kite. Dr. M's lesson contains 15 PowerPoint pages, but we could cut out almost half of them simply by using the inclusive definition -- five pages of indirect "not the other" proofs, and two more pages to explain why the "not the other" proofs are needed!

So here is the U of Chicago definition of kite:

A quadrilateral is a kite if and only if it has two distinct pairs of consecutive sides of the same length.

(Notice that here distinct means that there are two different pairs with the same length, for a total of four sides -- not that the lengths themselves must be distinct.)

The text has the quadrilateral hierarchy, but with one link missing, from rhombus to parallelogram. It states that this will be proved in Lesson 5-4 (actually 5-6), since it uses the Alternate Interior Angles Test that doesn't appear until that lesson. But here on this blog, we already proved the AIA Test, so we can actually prove the entire hierarchy right now.

Now some might notice something here. Yesterday, I wrote that the first four sections of Chapter 5 don't require a Parallel Postulate. Yet today, I'm discussing quadrilaterals like rectangles and squares, and as it turns out, rectangles (and hence squares) don't even exist without a Parallel Postulate!

So what gives here? Actually, every statement proved in this lesson is still true in non-Euclidean geometry, even those like "every rectangle is a parallelogram." If rectangles don't exist, then the statement "every rectangle is a parallelogram" is vacuously true -- there exist zero rectangles, and all zero of them are parallelograms! (Similarly, all unicorns are white.) No statement about rectangles or squares made in this lesson actually requires any of them to exist -- only that if they exist, then they have these properties. Wu does the same trick on his page:

"A quadrilateral all of whose angles are right angles is called a rectangle. A rectangle all of whose sides are of the same length is called a square. Be aware that at this point, we do not know whether there
is a square or not, or worse, whether there is a rectangle or not."

Also, the statement that every rhombus is a parallelogram uses the Alternate Interior Angles Test, but it's the Parallel Consequences, not the Parallel Tests, that require a Parallel Postulate. It is not until Section 5-5, where we derive properties about trapezoids using the consequences of their parallel sides, that we'll need a Parallel Postulate.

Let's move on to the exercises. I decided to throw out the first eight questions since defining, drawing, and placing into a hierarchy the seven types of quadrilaterals fit better in the notes, not in the exercises that come after the notes. As for the other questions, it's interesting to point out how the answers might be different using inclusive/exclusive definitions, or Euclidean/non-Euclidean geometry.

I include the first three true/false questions, from 9-11. Question 9 is true, even in non-Euclidean geometry (where it's vacuously true) and Question 10 is always false. Question 11 is true, but becomes false if we use an exclusive definition of kite. (Notice that my exercises make no reference to trapezoids, but only kites, since kites are coming up sooner in Section 5-4.)

Then I skip to Question 20. If set A is the set of all rectangles and set B is the set of all rhombuses, then A intersect B is the set of all squares. This remains valid even in non-Euclidean geometry, since there set A, the set of all rectangles, becomes the empty set. Then A intersect B would also be the empty set, which equals the set of all squares, since that's the empty set as well.

But there is a similar intersection problem that's too advanced to be given here, and where the answer differs depending on what geometry one is using. The intersection of the set of all parallelograms and the set of all isosceles trapezoids is, in Euclidean geometry, the set of all rectangles. (One way to prove this is to note that isosceles trapezoids have equal diagonals, and -- though this isn't proved in the U of Chicago -- parallelograms with equal diagonals are rectangles.) Yet in hyperbolic geometry, there are no rectangles, but there exists figures that are both parallelograms and isosceles trapezoids -- in particular, the Saccheri quadrilateral is both. A hyperbolic geometer may miss the fact that a Saccheri quadrilateral is an isosceles trapezoid because he/she is using the exclusive definition, where a parallelogram can't be a trapezoid. But in some ways, a Saccheri quadrilateral is more like an Euclidean isosceles trapezoid than a Euclidean parallelogram, since the Saccheri and the isosceles trapezoid share the same type of symmetry line that the general parallelogram lacks.

In Question 21, we prove that NOPQ is a kite -- a proof that requires only four steps (since I always add a Given step to the three steps asked for in the book). But if we use the exclusive definition of kite, NOPQ might not be a kite because it could be a rhombus. Technically speaking, we can't prove that NOPQ is an exclusive kite unless we add another hypothesis, such as circles O and Q having unequal radii.

No comments:

Post a Comment