67

BASE 27: THE KEY TO A NEW GEMATRIA

LEE SAL LO WS
Nijmegen, The Netherl a nds

"You never learn anything, you only get used to it". - Laurent Sik16ssy

Seotivigesimal Notation

The method of representing numbers by means of place-value notations that differ in their
radix or base, sucb as binary (base 2), octal (base 8), decimal (base 10), and so on, is
nowadays widely familiar. We remind ourselves that the numeral "123" interpreted in
base b means 1 x/T + 2xb 1 + 3 xbo, the case b= 10 then yielding 1 x 102 + 2 X 10 1 +
3 x 100 = 100 + 20 + 3 = 123, or one bundred twenty-three in our conventional decimal
notation. In each case the total number of distinct symbols or digits called for is simply
the same as the base indicated: two symbols for binary, eight for octal, etc. Standard
practice is to use the required number of decimal digits (0-9) for all bases up to ten and to
supplement these with letters of the alphabet for higber bases. Thus, in addition to 0-9,
hexadecimal (base 16) employs A for 10, B for 11, etc., up to F for 16, to complete its
sixteen digits, a hexadecimal numeral such as AB9F then indicating lOx 163 + 11 X 162 +
9x 161 + 16x 16° (= 43936 in decimal). This recurrent use of the familiar decimal digits
is a convenience for us decimal-oriented users, yet inessential. Our choice of what sym­
bols to use as digits in hexadecimal or in any other system is entirely free. Sixteen runes
(ranked in an agreed order) would serve equally. Or the first sixteen letters of the
alphabet, say. From the standpoint of logology bowever, to neglect the remaining ten
letters would seem a job half done.

An obvious idea then is a number system using base 26, with its digits comprising A to Z
in their natural order. This implies a notation in which every natural number is repres­
.ented by a unique string of letters, while every string of letters could be interpreted as a
unique natural number; an intriguing prospect offering scope for developments beyond
reach of traditional gematria. However, there is a problem. For in that case, as first digit
the letter A will represent zero, which is the place-holder or empty position indicator, a
key element in any positional number system. The trouble is that this not only conflicts
with the more natural A = 1, B= 2, etc. , the behavior of zero is peculiar in that leading
zeroes may be appended to or deleted from any numeral without affecting its value . Thus
in decimal notation, 12, 012, 0012, ... all stand for twelve. It seems arbitrary that A
should be singled out to exhibit an aberrant property not shared by the other letters. Still
more serious, the numerical value of string ABC will then be the same as that of BC,
while AARON will equal RON (!), and so on. Our hope of a system in which every
distinct letter string stands for a unique number is thus not realized in this scheme. Even
reordering the digits -which is undesirable- merely shifts the problem to a new letter.
 68

Having pondered this predicament on and off over a period of years, a solution recently The logoh
occurred to me. It is a simple step, bot one impossible to make as long as base 26 is the extends it
starting point - a central assumption difficult to shake off. The answer is to introduce an guages ust
extra symbol for zero. I shall use n _ " , the (bold) underscore , whicb is a near approach to such, ever
that typographical cousin of zero, the blank. With the addition of this extra character we word, is a.
arrive thus at a base 21 or septivigesimal (SV) system whose digits comprise the under­ For instan
score and the upper case letters, so that _ = 0, A = l , B= 2, ... Z= 26. Henceforth, if the PRIME, \1
underscore is a leading cbaracter it may be dropped or ignored; elsewhere it must be properties
retained. Note that use of the blank itself would have left any trailing zeroes invisible and (essentially,
thus undetectable , a fatal flaw . On the other band, there is nothing to prevent us from be were tlJ
interpreting underscores as blanks, and vice versa, should we so cboose. Tbe advantage of as blanks
doing so will emerge. finite set[!
BEAST,
The SV notation for numbers beyond 26 is then best illustrated through visualizing a suit­ punCtuatiOl
able odometer, which is a perfect model of a positional number system, the Number made to a uniq\l
Flesh, as it were. Its simple mechanism is familiar to us in the tape counter or car mile­ nothing sh
ometer: a row of rotatable discs viewed edge-on, each bearing the ten digits evenly inscr­ (texts) are
ibed around their periphery. In an odometer designed for SV, however, 0-9 are replaced
by _ -Z, the discs then rotating in 27 rather than 10 steps. Each successive input or event
to be counted advances the rightmost disc by one digit. With the completion of a cycle its Adding am
left hand neigbbor is advanced one digit also. And so on with the remaining discs. Each
new input thus gives rise to a new combination of digits in the viewing aperture. Starting If every w
from the reset or zero position, which is a row of underscores, the SV notation for a in SV nota
given number n is then found on the readout after entering n successive inputs: that yield,

n: 0, 1, 2, 26, 27, 28, 29, 53,54,55,56, 728,729,730 , 731, We need ,

SV: _, A, B, Z, A_, AA, AB, AZ,B_,BA ,BB, ZZ,L,A-A,A_B , contractiOI:
which are
The sequence speaks for itself. As the odometer advances, the readout progresses odometer :
systematically through every possible combination of digits, all the I-digit strings followed steps must
by all the 2-digit strings, etc., the first yielding a numeral for one, the second a numeral another wi
for two, and so on. Our goal is achieved: every distinct string of teners is now a unique which are.
code or label for a distinct positive integer, while leading zeroes may be added or deleted sing the 0
without affecting any letter; the behavior of A is no longer aberrant. At the same time, DISK, as 1
the codes for many integers include underscores, a fact which need cause no concern and
will even prove useful in a moment. Furthermore, negative integers, fractions, real num­ The above
bers, and so on, can all be represented by bringing in minus signs, "decimal points" (unit from SV i:
points, or separators as they are better called), and other signs, in the usual way. Aside into SV . 1
from its offbeat radix and digit symbols SV is an entirely conventional number system. tiplication

To convert a letter string to its decimal equivalent is easy. Take CAT. Noting C=3,
A=I, T= 20, we write: CAT = 3x2'? + I x27 1 + 20x 2J> = 2187 +27+ 20 = 2234.
The reverse process of deriving the SV notation for a given number calls for successive
We procC1
divisions by 27. Consider 74417. Then 74417 -;- 27 = 2756 with remainder 5. E=5 is
tables shOl
then the final digit of the SV numeral. The process is then repeated with the previous
quotient: 2756 -;- 27 = 102 with remainder 2. B=2, becomes the next to last digit. Then
Next, L +
E and car
102 -;- 27 = 3 with remainder 21. U =21 precedes B. The end is reached when the quoti­
ent falls below 27, its value then standing for the leading digit: C=3 in our example. So,
ONLY +
tion, mutt
744 17 = CUBE. Test yourself: what is 1492 in SV? See Answers and Solutions.
checked ir
 69
i)D recently The logological implications of SV now begin to unfold. For as the sequence above
e 26 is the extends it will come to include every English word, along with all the words of those lan­
uroduce an guages using the Roman alphabet. Every word is an integer written in SV notation, and as
tpproach to such, every word is greater or lesser than any other given word, is an odd or an even
taracter we word, is a composite or a prime word, is a perfect square or cube or whatever, and so on.
the under­ For instance, the smallest English word that is prime is AN, whicb is 41 . Another is
:>rth, if the PRIME, which is 8864267. Henceforth it becomes natural to identify a word with the
it must be properties of the integer it represents. Likewise, a dictionary now reveals itself as a list of
visible and (essentially random) integers whose lexicographical ordering is different to what it would
~t us from be were they ordered by magnitude. Moreover, the advantage of interpreting underscores
Ivantage of as blanks now appears. For then the series of natural numbers must also include every
finite sequence of words, such as WORD _ WAYS, THE _ NUMBER_ OF _ THE_
BEAST, and LNTERPRET _ UNDERSCORES_AS_ BLANKS. That is, neglecting
~ing a suit­
punctuation and other non-alphabetic signs, every sentence, indeed every text corresponds
nber made to a unique integer, also. The ramifications of all this in relation to logology become

r car mile­ nothing short of momentous in the realization that from here on relations among words

eoly inscr­ (texts) are expressible in mathematical equations.

'e replaced
Jt or event
a cycle its Addin~ and Multiplyin~ Words'
~scs. Each
e. Starting If every word is a number then two words can be added together and their sum expressed
ltion for a in SV notation. Chances are the result is a meaningless string of letters. Might cases exist
that yield a wor([l There are thousands.

~ We need a name for integers whose representation in SV is a word. I propose wine, a

contraction of word-integer. Consider the addition of two wints, 399749 and 85817,
~ which are TIllN and DISK, respectively. Starting from zero and advancing our SV
progresses odometer 399749 steps the readout will be THIN. Advancing its thin discs a further 85817
s followed steps must show a result that matches the decimal number 399749 + 85817 = 485566. It is
a numeral another wint: XRAY. That is, 485566 = 399749 + 85817 or XRAY = THIN + DISK,
v a unique which are just equivalent statements in different notations . Starting with XRAY and rever­
or deleted sing the odometer by 399749 backward steps thus produces XRAY - THlN, which is
:ame time, DISK, as we could have predicted from elementary arithmetic.
)Dcem and
real num- The above suggests two approaches to calculating with SV numerals. One is to convert
lints" (unit from SV into decimal, perform computations as usual, and then translate the answer back
,ay. Aside into SV. The other is to work directly in SV itself, a method requiring addition and mul­
ystem. tiplication tables, as provided on the next page. Consider an example:
ONLY
ing C=3, RE ST +
) = 2234. AFTER
successive
5. E=5 is We proceed exactly as with ordinary sums, beginning with the right-hand digits. The
~ previous
tables show Y + T = AR. Thus, write down the unit's R and carry the twenty-seven's A.
ligit. Then Next, L + S = AD, to which must be added the carry: AD + A = AE. Write down the
the quoti­ E and carry the A. Then N + E + A (the second carry) = T. Lastly, 0 + R = AF.
tmple. So, ONLY + REST = AFTER you have mastered this example. The procedure for subtrac­
tion, multiplication, and (long) division is entirely analogous. The following sums can be
checked in the same way:
 70

ANTI + BULK = DI ET
BEST + POLE =
RU DY
DU ET + GORE =
BUL L
Either of tI
HAV E + GIFT = OKAY
BEST + HOL E =
JUDY
AC ID + CASH =
DEAL

CHAT + TALK = WIND

BEDS + JOYS =
LUCK
AIDS + KEYS =
LOCK
which of t
IDEA + KEGS = TILT
AVOW + JI LT =
LEAP
FULL + GOAT =
NIN E
language,
FEAR + OMEN = URGE
AVOW + WILL =
YEAH
FUCK + PUNT =
W ORD
enough to
AMEN + CUBE = EGGS
EDEN + FATE =
KEYS
FLAW + MASK =
SMUG
handle tar
EGGS + GULP = MATH
CUBE + GOLF =
KINK
INCH + MA IL =
VOLT
2 14748364
CHEF + EGGS = HOLY
CUBE + POLO =
TINT
SEND + MORE = AEUEI
ENQUIRY
Semantic cohesion lends charm to some of these examples. If you are anti-bulk then diet; directly in
to have a gift is okay; cJuu. talk, and wind are synonyms; Rudy and Judy doubtless know Incidentall:
that the jays of sound sleep in comfortable beds is a matter of luck; SEND + MORE is not glance at a
equal to MONEY, contrary to cryptaritbmic superstition. But how were these specimens
derived? Doing sums in SV would be a lot easier if only we knew our SV tables by beart. Question:
Yet even then calculating by band would remain tedious and slow. Prohibitively slow. answer del
Using a computer for the job is not merely quicker, its ability to perfonn fast searches bers contai
through a data base makes it a virtually indispensible tool for research in this field. The letter resul
above examples, for instance, are among more than 7000 discovered by a simple program letter strinl
that, using a stored list of words, took each 4-1etter pair in turn, calculated their sum and words stor
then checked to see if it was present in the list. This process need not be limited to 4­ letter strinj
letter words of course, or just pairs, or addition only. Equivalent

No. of

AB C D E F G H I J K L M K 0 P Q R 5 T 0 V W x Y Z + I Percent
No. of

A A BC D E F G H I J K L MN 0 P Q R S T 0 V Wx Y Z A A
B B D D E F G H I J K L H N 0 P Q R S TU V W x Y Z A. AA B So wherea
C C F I F G H I J K L MN 0 P Q R S TU V W x Y Z A. AA AS C
D D H LP H I J KL H N0 P Q R 5 T 0 V W x Y Z A. AA AS AC D
around one
E E JO T Y J K L H K 0 P Q R S T 0 V W x Y Z A. AA AS AC AD E despite the]
F F L R XAC AI L HK0 P Q R STU V W x Y Z A. AA AS AC AD AE F bottom lin(
G G K 0 AA AH AO AV K 0 P Q R 5 T U V W x Y Z A. AA AB AC AD I.E AF G try. Nonel!
H H P X AE AM AU BE BJ P Q R 5 T 0 V W x Y Z A. AA AS AC AD AE AI AG H combine \\
I IRA. AI AR B. BI BR C. R S '1' 0 V Ii X Y Z A. AA AB AC AD AE AI AG All I much rarer
J J T AC AM Ali BF BP BZ CI CS T 0 V W x Y Z A. AA AS AC AD AE AF AG AH AI J
K K V AI AQ BA BL BW CG CR 08 DII v W x Y Z A.. AA AB 1.C AD AE !F AG AH AI AJ K
incongruity
L L XAI AU BF BR CC co D. DL DX EI X Y Z A. AA AS AC AD AE AF AG AH AI AJ AK L struggles tc
H H Z AL AY BK BX CJ CW 01 DV EH EU FG Z A. AA AB AC AD AE AF AG All AI AJ AK AL M
K N M WBB~OCOO OO ~U ~ IT IT ~ AAAB ~ADAE !F ~ AH il AJ AK AL AM N AMEND 1
o 0 AC AR BF au CI CX DL E. EO FC FR GF GU HI AC AD AE AF AG AH AI !J AX AL AM AN 0 ATONE 1
P P AE AO BJ BZ co DD DT EI~Y FK GC GS HH HX II! AE AF AG AH AI AJ AK AL AM AN AO P BYRON i
Q Q~ UM ~m ~D~ rn ITOO ~ ~ U~ ~ ~ AH il !JAKAL~ ANW~ Q CRIES -+
R R AI B. BR CI D. DR EI F. FR GI H. HR II J. JR KI L. AI AJ AK AL AM AN 1.0 ~ AQ R FIFTH -+
S S AK Be BV eN DF DY EQ FI GA GT HL ID Iii JO KG KZ LR MJ AK AL AM AN AO ~ AQ AR 5 HOVEl -+
T T AMM~~~U~FR ~~nw~ ~ ~ ~ ~ ~ w AMANW~~ U ~ T
o 0 AO Bl OC CX DR EL FF G. GO HO II JC JX KR LL KF N. NO 00 PI AO ~ AQ U AS AT 0 ~
V V~~CG W ~~FK~ ~ ~ru~~u ~~ ~~~~~ ~UASH~ v R
Ii W ASOO a~ ~~"~ D U~mnW~ ~ ill~~~~~ AS H ~~ W A
x X~~CO~nITGC L nru JR mLL~MOC~n~~wnill ~~u x C
Y Y Ali au ~ DQ EO FM GK HI IG JE KC LA LZ HX KV OT PR QP RK 5L TJ OR VF WD AW AX Y H
Z Z AY BX CW DV EO IT G5 HR IQ ~ KO LK MM NL OK PJ QI RH SG TF OE VD we XB YA AY Z ij

x I ABC D E F G H I J K L H K 0 P Q R S T 0 V W x Y Z I
Addition and multiplication tables for $V numbers. Ex: P+Q=AF, PXQ=JB
 71
Either of the two methods described above may be implemented in a program of this kind,
which of them is best? Decimal operations are a built-in fearure of every programming
language, while routines for converting from SV to decimal and back again are simple
enough to write, so this would seem easier and fast. So it is -provided your software can
handle large enough integers. In Pascal , for instance , the largest integer allowed is
2147483647, which is ENQWLTJ, so the highest wint you could process would be
:I ENQUlRY; a similar limit is met when using a pocket calculator. Routines for calculating
:hen diet; directly in SV evade this problem but are more difficult to write and slower in execution.
ess know Incidentally, programmers will find "@ " preferable to "_" in representing zero, as a
[ill is not glance at an ASCII code chart will make clear.
pecimens
Iby heart. Question: Given two wints a and b, what is the likelihood that (a + b) is a wint? The
!ly slow. answer depends on its length, and hence on that of a and b. The sum of two n-digit num­
searches bers contains n or n+ 1 digits. Suppose we add two smallish 3-letter wints to produce a 3­
ield. The letter result and ignore cases that contain underscores. The total number of possible 3­
program letter strings is 26J = 17,576. How many of these are words? A modest lexicon of 25,000
sum and words stored on my hard disk reveals 747 items of 3 letters. The chance of a random 3­
ted to 4­ letter string forming a word is thus something like 747 in 17,576 or about 4 .25 per cent.
Equivalent figures for some other string lengths are as follows:

No. of digits 1 2 3 4 5 6 7 I
Percent words 7.7% 13.4% 4.25% 0.46% 0.026% 0.001% 0.00005%
~ No. of words 2 91 747 2140 3091 3791 4043
-- - - - - -- - -- - ~ ------ - - - - -- '-- ­
A
B So whereas the chance of two smallish 3-letter words adding to form another word is
C
around one in twenty, for two smallish 7-letter words it is only about one in two million,
D
E despite there being over five times more 7-1etter words than 3-1etter words, as seen in the
F bottom line above. On the other hand, more 7-1etter words mean more candidate pairs to
G try. Nonetheless, solution counts diminish rapidly beyond 4-letter words. Instances that
H combine words whose meanings make them in some sense apposite then become very
I much rarer. A random selection of 5 and 6-1etter word sums follow . Sometimes the sheer
J
incongruity of the word combinations produces quite a comical effect as the brain
K
L struggles to invent scenarios to .. explain" the equations.
M
N AMEND + BEING = CROAK ADAPT + DIODE = EMPTY ANGER + MANIA = NOUNS
o ATONE + EXTRA = GRIEF AUDIO + JOKED = LIONS BEAST + MADLY = OFFER
p BYRON + FEMME = IDEAS CACTI + EXILE = HYMEN COMIC + LUMPY = PIZZA
Q CRIES + DREAM = HINGE DWARF + MINUS = REPLY EXIST + QUOTE = WRYLY
R FIFTH + NINTH = TRUMP GREAT + MEDAL = TWICE HAPPY + ISLAM = QUARK
S HOVEL + VENOM = ACUITY WALTZ + WASTE = ASCEND ODDLY + POKED = ADSORB
T
U ANCHOR + PROVEN = RESCUE ANTICS + REVEAL = STONED
V FITFUL + JESUIT = POMADE ACROSS + RANDOM = SE ETHE
W FORMAT + LYRICS = SNIVEL ACROSS + RANCID = SEESAW
x CLOSER + RAPING = UNEASY ELICIT + LOVERS = RADIAL
y HI LTON + MEEKLY = UNREAL GAUCHE + HAIRDO = OCCULT
z REGAIN + WINKLE = ANNULUS SADIST + THIEVE = ALIMONY

I
The focus here on word pairs of equal length is explicable. Consider AFFAIR + CUB =
AFFECT. Leading digits in the longer word survive in the total, a result less pleasing
than when a contrasting word is produced. The smaUer the difference in lengths the less
glaring is this effect. Single letter differences hardly show up, and sometimes not at all:
 72

BO WS + TOTEM = TRIBE COIN + HALVE = HEADS BUMS + SQUAT = STOOL

has identi
CATS + DOGMA = DR IFT LUST + COY LY = DATER GRIT + SAUC E = SI LLY
square on
MIR E + SLICK = SYRUP RIDE + ORGAN = PIPES JEST + DEATH = DOGMA
AM2. Thi!
HOLDS + SEPTUM= SNEEZE DIANA + SPIDER = STRESS can any n
DOWRY + DECEI T = DISBAR NOBLE + KNI GHT = LAXITY if lengths
QUART + ARMP IT = BIGRAM MOANS + BANANA = BOBCAT the square
After (a + b), what about (a xb)? The product of two n-digit numbers contains 2n or 2n-l triangle ju
digits. This is n digits greater than in their sum. The likelihood of (axb) forming a wint BEG. In f
is thus drastically reduced, a fact reflected in the extreme rarity of instances. The problem have fow
of leading digits surviving in the result disappears with products so that word lengths are looked at
not important. Even so, a program that ran through my lexicon and tested every single required I
pair (a ,b) formable from words using 6 or less letters discovered fewer than than 100 adth = r
cases for which (a xb) is a wint. A few of the better examples, by which I mean those MON squ
using no single letter words, acronyms, proper names, etc., follow. The funereal tone in which ex~
the right hand group (dig trench .. bury body .. rugged cojfin.. cram corpses) is pure chance! area is als
The six s1
IS x GEM = BOXED JOB x JI G = DAEMON DIG x DE FT = REVOKE + (DEE
BY x SAD = BASIS TAB x HOP = FIXATE HAW x BODY = TRENCH improve c
BE x FRET - NOUNS BAY x BREW = ENOUGH FRY x BURY = RUGGED Howabot
AT x KING = STOVE ARE x BUMS = DRI VEN FRY x NIT = COFFI N word? Thj
BE x DYKE = JUNKY TO x ZINC = TARTAR ZIP x CRAM = CORPSES
The equations looked at so far are of simplest type: y = a+b and y = abo Substitutions It is time
and transposals lead to more intricate forms; e.g, by combining BExFRET= NOUNS relation eJ
with FRET = SLID-LUCK and NOUNS = ANGER + MANIA, we get BE =(ANGER + That's rig
MANIA)/(SLID-LUCK). And so on to any degree of complexity . Any degree of com­ result of (
plexity? Peering beyond these elementary expressions, at this point in our progress all 0, and .. .
kinds of possibilities begin to suggest themselves. To employ a metaphor, until now we
have been examining the contours of an unfamiliar object recovered from the ground.
There are suggestions it may contain explosive material. Now we have realized the thing
is ticking. This brings us to the Big Bang.

Ten samp
The New Gematria encourage
FOUR +
On my bookshelf is The Penguin Dictionary of Mathematics. I open it at random and
Can it be
select an entry at will. The item reads, "GCD. Abbreviation for greatest common
divisor." Of course: the GCD of 12 and 18 is 6, for instance, because 6 is the highest
= ONE.
that worlc
number that divides both 12 and 18. GCDs for larger numbers can be found using a
simple process called Euclid's algorithm. But if every word is a number then every pair (FOR
(or more) of words must have an GCD as welL Could some of these GCDs themselves be ( EIG:
words? A program similar to the one described above but now incorporating Euclid's (N IN
algorithm pours out examples by the score: The GCD of JET and PIE is BAD, of BELCH
and DRYLY is LAW, of FEEDS and EARLY is ADD, etc. The GCD of STEWS and It would
PENNY is GCD. Verification of these instances is tedious but straightforward, easier if difficultie;
you convert to decimal first. Easier still performed on your Pc. single-leVI

Re-opeoing the Dictionary elsewhere I land by chance on Pythagoras Theorem. Let's What has
see, could a right-angled triangle have sides whose lengths are wints? That is, can three alights on
words be found to satisfy d = ~ + C ? A program that took pairs of words, calculated main diag
the square root of the sum of their squares and then checked the result against a word list wints! Ca
 73

TOOL has identified a cheerful fact about the

IL LY square on ooe hypotenuse: AN 2 = r + BOY
IOGMA AMI . This is the sole instance discovered; Area - CLOSET
can any reader track down a second? But
if lengths can be wints, so can areas. Take III
the square on the side of length AM in the Mt:j:i:b~'§0Bj~~B%SJl
rz or 2n-l triangle just mentioned; its area is AMl =
IIlg a wint BEG. In fact this is the only such square I
: problem have found, although the products we
~ngths are looked at earlier yield rectangles with the
ery single required property; e.g., If length x bre­
than 100 adth = JIG x JOB, then area = DAE­
lean those MON square units. Turning instead to triangles in general, anned with Heron's formula,
al tone in which expresses area in terms of 3 sides, I went fishing for wint-edged specimens whose
: chance! area is also a wint, but met with no success. Not to be outdone I tried a different figure.
The six sides of the L-shape above are all wints, while its area is given by (BOY x EGG)
KE + (DEE x SAW) = (BOY x ll..L) + (DEE x PLY) = CLOSET. Can any reader
CH improve on this, perhaps by finding a different shape that uses a more apt set of words?
ED How about a wint-sided figure whose area is equal to AREA, or some other appropriate
IN word? The possible variations are endless.
SES
)stitutions It is time for another lucky dip in the Dictionary. I find: "Congruence modulo n. A
=NOUNS relation expressing the fact that two integers differ by a multiple of a chosen number n. "
lNGER + That's right: we write a mod n == b, (spoken, "a modulo n is congruent to b") when the
: of com­ result of dividing a by n leaves a remainder of b. Hence: 10 mod 4 == 2, and 9 mod 3 ==
:>gress all 0, and ..... well, you can see it coming:
I now we
: ground. CIRCLE mod LOVE ~ FLAG MUSES mod BLUE = ARMY
the thing SATAN mod HATE ~ BOOK HANGS mod GREEN ~ JIBE
OUTCRY mod THIS 2 FOXY MODULO mod ARTHUR e DEAF
LEGATO mod THAT = SNOW PERFORM mod MODULO = CROCUS
Ten samples taken at random from among xillions the computer finds. This abundance
encourages specialization. Consider: (FOUR x SIX) mod TEN == ART, and (ZERO +
FOUR + EIGHT + TWELVE) mod SIX == GOD. Almost every wint is a number-word.
ldom and
Can it be done using number-words only? It can: (ELEVEN x NINETY SIX) mod TEN
common
== ONE. But alas, (11 x 96) mod 10 == 6 ( .. I)! What we really seek is an expression
Ie highest
that works on both levels of interpretation. Three (non-trivial) instances are as follows:
I using a
very pair (FORTYTWO + FORTYNINE)
mod SIX == ONE and (42+49) mod 6 == 1
selves be (EIGHTYFIVE- EIGHTYTW
O)
mod TWO !!! ONE and (85-82) mod 2 !!! 1
Euclid's (NINETYN INE - TWELVE)
mod FIVE e TWO and (99-12) mod 5 ~ 2
f BELCH
EWS and It would be nice to find two-level equaIions rather than congruences . However, the
easier if difficulties involved are horrendous. To date I have not succeeded in fmding even a
single-level equation. Immortal fame awaits the first logologist to succeed .

lm . Let's What has the Dictionary in store for us next? I pick a new page at random and my glance
can three alights on "Magic square. A square array of numbers whose sum in any row, column, or
:alculated main diagonal is the same." The contingency this suggests is exciting: a magic array using
word list winls! Can it really be done? A program for seeking squares formed from any set of
 74

special numbers is not difficult to write. Contrary to intuition tighter internal constraints English Sf:
make it harder to find wints that will satisfy 3 x 3 squares than it is for larger types. The asked of ,
rhyming word triples that appear in solutions reflect these closer restrictions : Perhaps th
be .. _PH
HOG BIG HUG 6244 1708 6406 LASH ARK LATE NAL_ .. ,
FUG FOG FIG = 4948 4786 4624 HATE HASH HARK
DIG JUG DOG 3166 7864 3328 DARK PATE DASH Another il
The magic constants here are SRU = 14358 (left) and XECX = 476142 (right). The expansion
challenge this suggests is obvious: can a square can be found whose constant total is itself
in which 1
a wint? Moving on to 4 x4 squares the problem has yielded to attack: Dictionary
dividing 1
division n
EWE HO E TOG LOW DOT RAJ TIT PAX repeat: C :
NUN RIP FIN HAW PIT TAX FOR PAL which is ~
PIN RAW DUN HIP TAR RIP NAP FOX
JOG BOW OWE ROE RAP BOX RAT TIN division i:
repetend i~
magic constant = ATOM magic constant = BEAT
FAR FOB OAT BUS DIM OWE TUG RAP
GAY JUG FAT FOG RIG TAP DOT PAY
HUN MAR ON HAM THE TIP NAP DI D
HOB BE HUE MAT PAP BUD SPY TOW
magic constant = ACME magi c constant = BEAN
Here the top squares have a (hidden) "graeco-Jatio" structure, which entails that the four
words in each quadrant and the four corner words of each 3 x 3 subsquare also total to
ATOM and BEAT. Evidently the potential of SV magic squares to combine both
mathematical and logo logical properties provides enormous scope for future investigations.
For the time being, however, I propose a specific challenge: can any reader discover a What has 1
square with MAGIC as its constant total? namely : 11
above but
Meanwhile, turning again to the Dictionary my eye falls on: "Irrational number. A real portion of
number that cannot be written as an integer or as a quotient of two integers." The classic tiplication
example is the diagonal of a unit square: ./2, or l.4142135 .. , the dots indicating an
endless string of digits. But can irrationals be expressed in SV? Of course they can, SV is Before lea'
just another number system. Using a recursive formula for finding the square root of any 5, is a pri
number k, such as, X.+ I = %(x. + klx.), in which X l is set to an initial guess, while Xz, tional mat
X3 , •• are successively closer approximations, the SV expansion of ,/2, or as we should now repeating ~
say , ./B, turns out to be A.KDYZAWWXDUAFBSVOTSTLYCBVUNUTPBFGSSEW .. in triguing
(as I trust A.K. Dewdney wilt note with interest). The deci mal interpretation of this is reveals the
then lx 27° + ll x27- 1 + 4x 27-2 + 25 x 27-3 + ... , (since A= l, K = ll, D =4, etc.)
= 1.4142135 .... Note that although we call the latter a "decimal" , the term becomes a
confusing misnomer when applied to SV expansions like A. KDYZA ... , i.e., SV numerals
that include separators or unit (not "decimal"!) points.
In fact 0"
Looking over this result we find that the fIrst English word to occur in the string is NUT,
obviously
closely followed by SEW. As it happens this initial fragment contains no underscores.
property 0
However, if ,/2 is a so-called "normal" number, as mathematicians believe, then its
cyclic nUIT
infInite (strictly, chaotic) sequence will eventually include every possible pattern of n
new cyclic
digits, for every n. An odd-seeming yet perfectly serious question is thus: What is the fIrst
 75
t1 constraints English sentence to occur in the septivigesimal expansion of ,/2 ? And the same can be
I( types. The asked of all similar numbers, including say, 7r , e, and Euler' s constant, y. Who knows?
Perhaps the fIrst sentence to appear in 7r is .. _ GOD _ EXISTS _ .. , while that in e might
be .. _ PROVE _ IT _ .. , and that in y might be " _ THIS _ NUMBER _IS _ IRRATIO­
lATE NAL_ .. , which would be fun since nobody yet knows whether this is true or not.
'ARK
'ASH Another intriguing number is phi (4)), the golden ratio, which equals 112(1 + J5 ). Its SV
(right). The expansion begins thus: A.PRNTPFCUCRKDYGRYLLCQNBIG_OVQTRTLY _IKM ... ,
total is itself in which BIG is the fu-st word to occur. This is an interesting case to consider since my
Dictionary of Mathematics has now fallen open at: "Reciprocal . The number produced by
dividing 1 by a given number." The reciprocal of C is thus lIC or Ale. Simple long
division reveals the answer: C into A won't go, so write" _." , append a zero to A and

~
x repeat: C = 3 into A_ =27 goes [ = 9 times exactly. Thus the reciprocal of C is _.I,
Al which is 9 x 27-1 = 9/27 = % = 0.3333 ... , in decimal . In fact the result of any long
OX
IN division is always a recurring quotient, although we tend to overlook this when the
repetend is made up of zeroes only. The reciprocals of the alphabet illustrate this point:
~T
A/A = A. ___ .. A/ N = _.AYAYAYAYA ..

~
p AlB = _.MMMMMM .. A/ D = _.AUPEJUPEJ ..
~Y
A/e = _. L __ .. A/P = _.AROEAROEA ..
[D A/ D = _.FTFTFT.. A/ Q = _.AOWVFINGYKCDTQLSAOWV ..
)1'1
A/E = _.EJUPEJUPEJ . . A/R = _.AMMMMMMMM ..
A/F = _.DMMMMM . . A/S = _.AKIYOQAKI YOQ ..
A/G = _.CWCWCW .. A/ T = _.AILDAILDA ..
~N A/H = _.CJCJCJ . . A/ U = _.AGSGSGSGS ..
A/ I = _.C ___ .. A/V = _.A FC RKAFCRK . .
nat the four A/ J = _.BRXHBRXHBR . . A/W = _.ADRUCNBIJOGADRU . .
dso total to
A/K = _.BLGIVBLGIV . . A/X = _.ACJCJCJCJ ..
A/L = _.BFTFTFT.. A/Y = _.ABDHQGOCFLYXVRISKWTNABD..
mbine both A/M = _.BBBBBBB .. A/l = _.AAAAAAAAA ..

lestigations .

. discover a What has all this to do with 4>1 Simply this: the reciprocal of 4> has a fascinating property ,

namely : 1/4> = 1-4>. That is, 114> = _ .PRNTPFCUCRKD .. , which is the same string as
above but minus its leading A! Some readers may like to test this by taking a small
.b er. A real portion of 4>, say A.PRNT, and dividing it into A by long division (using the mul­
The classic tiplication table as an aid). The exercise requires patience but is instructive.
idicating an
can, SV is Before leaving these reciprocals take another look at liE = _ .EJUPEJUP .. . Now E, or
root of any 5 , is a prime number, while the length of the repetend, EJUP, is 4. Students of recrea­
S, while x2 , tional math may recall that when the reciprocal of a prime p has a period of p-l the
should now repeating sequence forms a so-called cyclic number. A cyclic number of n digits has the
:FGSSEW .. intriguing property that when multipJied by any number from 1 to n, the resulting product
[l of this is reveals the self-same n digits arranged in the same cyclic order. Thus:
D=4, etc .) Ax EJUP = EJUP
becomes a Bx EJUP = JUPE
V numerals Cx EJUP = PEJU
Dx EJUP = UPEJ
In fact cyclic numeral would be a better term for these curiosities since their periods
[Ig is NUT,
obviously depend on the radix of the number system in use. Cyclicity is thus not a
nderscores . property of numbers but only of numbers-written-in-a-certain-base. EJUP is the smallest
e, then its
cyclic numeral in base 27, i.e., SV . After liE, the next prime reciprocal to produce a
attern of n new cyclic is 1117 = 1IQ = _ .AOWVFlNGYKCDTQLSAOW .. , with period 16. Better
: is the first
 76

yet, though, 1129 = IIAB , which produces the next case, results in a pangram of period Once an e
28: _. _ YCSNXEOVIHJFMZAWGLBUKDQRPTM_ ye. .. M alone occurs twice. So submit, is
if c is the integer represented by these 28 digits (27 discounting the leading _), we have : as number:
appeared e
A x c = _YCSNXEOVIHJFMZAWGLBUKDQRPTM ox C = MZAWGLBUKDQRPTM_YCSNXEOVIHJF that calculi
B x c = AWGLBUKDQRPTM_YCSNX EOVIHJFMZ Px c = NXEOVIHJFMZAWGLB UKDQRPTM
_YCS at length:
C x c = BUKDQRPTM_YCSNXEOVIHJFMZAWGL Qx c = OVI HJFMZAWGLBUKDQRPTM_YCSNXE
o x c = CSNXEOV IHJFMZAWGLBUKDQRPTM_Y Rx c = PTM_YCSNXEOVIHJFM ZAWGLB UKDQR now, the a
E x c = DQRPTM_YCSNXEOVI HJFMZAWGLBUK Sx c = QRPTM_YCSNXEOVIHJFMZAW GL BUKD
F x c = EOVIHJFMZAWGLBUKDQRPTM_YCSNX Tx c = RPTM_YCSNXEOVIHJFMZAWGLBUKDQ One altern
G x c = FMZAWGLBUKDQRPTM_YCSNXEOV IHJ Ux c = SNXEOVIHJFMZAWGLBUKDQRPTM_YC
H x c = GLBUKDQRPTM_YCSNXEOVIHJFMZAW Vx c = TM_YCSNXEOV IHJFMZAWGLBUKDQRP Planck In
I x c = HJFMZAWGLBUKDQRPTM_YCSNXEOVI Wx c = UKOQRPTM_YCSNXEOVIHJFMZAWG LB mention. 1
J x c = IHJFMZAWGLBUKDQRPTM_YCSNXEOV Xx c = VIHJFMZAWG LBUKDQRPTM_YCSNXEO That's all.
K x c = JFMZAWGLBUKDQRPTM_YCSNXEOVIH Yx c = WGL BUKOQRPTM_YCSNXEOVIHJFMZA price, for
L x c = KDQRPTM_YCSNXEOV IHJFMZAWGLBU Zx c = XEOV IHJ FMZAWGL BUKDQRPTM_YCSN
Mx c = LBUKOQRPTM_YCSNXEOVIHJFMZAWG A_ x c = YCSNXEOVIHJFMZAWG LBUKDQRPTM_ because e;ll
N x c = M_YCSNX EOVIHJFMZAWGLBUKDQRPT AA x c = ZAWG LBUKDQRP TM _YCSNXEOVIHJFM this notatie
integral d(
Note how the initial digits of successive products run through the alphabet! For an insight to work w
into this, as well as for much more on the lore of cyclic numerals, see Martin Gardner's more typo
uniformly marvellous Mathematical Circus (Pelican) .
do we stOi
* differ frO[]
* * * *
Life is short and so is space. The DictiOfUlry has played its part; its message should be Here and
clear. We began with a novel system for writing numbers with letters; every word was 1992, ChI
revealed as a unique number; elementary arithmetic became applicable to words. BOOM! letters in t
A Big Bang marks the explosion into being of an entire universe of mathematical word­ outer won
play . Almost anywhere we look in a dictionary of mathematics, a new variation, a new TIG is the
topic, a new area for research, opens up. Thus far we have not even scratched the particular
surface. Just think for a moment of the subjects not even touched on here: logarithms, metic pro:
trigonometry , progressions (complete: AM, BOB, EAR, .. ), group theory, probability, BINARY
calculus, polynomials (solve: It + NOWx - SAYING = _ ; x is a wint) , complex mean of I
numbers, self-descriptive numbers (remember GOdel coding?), and on and on . . Our new mention b
gematria is no mere rich vein but a vast mine as endless and ramified as mathematics actually SI
itself; it can never be exhausted. (digit ProI

Sematria
Sematria wordplay
underrated
"Septivigesimal gematria" is a jawbreaker; sematria is more musical, and even fits fairly getting th
well etymologically: serna (urjpa)= sign + matria (paTp ia) = measuring, Le., measure Limitation
expressed in signs. Sematria grew out of my dissatisfaction with standard gematria, a glad to el
system whose central feature is really its greatest weakness: different words are equaled. anyone ill
But every word is a unique ordering of letters. What then can we expect of a numerology wonder in
that is insensitive to letter order? It seemed to me that if we are going to have a system
that identifies words with numbers then distinct words must correlate with distinct values.
A new approach was evidently needed. To accept A=l, B=2, .. as a starting point was of
course natural, but thereafter a place-value system would be demanded. The trouble was
that my back-of-the-envelope try-outs would always run into the problem created by zero.
The place-holder was vital, I knew that. A to Z were already spoken for. On intuitive * Ross Ecklr
grounds the blank ought to stand for zero; it just seemed right. But the blank was invisible which zer(
work whicl
and that would prove ruinous! The riddle looked insoluble . And then the penny dropped.
 77
1m of period Once an extra symbol for zero was brought in everything fell into place. The result, 1
cs twice. SO submit, is less an invention than a discovery. It is not just a system for interpreting words
I), we have: as numbers, it is the system: simple, natural , inevitable; the only wonder is that it has not
appeared earlier*. But that is explicable, perhaps. For simple as it is, there is no denying
5NXEOVIHJF that calculating on paper in SV is plain arduous . Pretty excruciating in fact, if carried out
)QRPTM_ YC S
DTM_YCSNX£
at length: try seekipg two words whose sum is a word by hand. No, it has taken until
~WGLBUKDQR now, the age of the personal computer, before sematria could come into its own .

~ZAWG LB U K D

~A WG L BUKDQ

.~DQ RP TM_ YC
One alternative to SV (suggested to me, aptly enough, in the parking lot of the Max­
~G LB UKD QRP Planck Institute for Psycho-Linguistics in Nijmegen, by Doug Hofstadter) deserves
~JFMZAWGLB mention. It is base 26 but without zero: A = l, B = 2, ... Z = 26, AA = 27, AB = 28 , etc.
rM_YCSNXEO That's all. Et voilA: every letter string is a unique integer! However, loss of zero exerts a
:OV IHJFMZA
lRPTM_YCSN price, for although we can still add and multiply, we cannot, in general, divide. This is
IUKDQRPTM_ because expressions like AlB have no solution : the smallest number that can be written in
lXEOVI HJFM this notation is A, or .A if we bring in the unit point. The resulting system is a so-called
integral domain, whereas SV has the structure known as a field, a property that allows us
Ir an insight
to work with any real number. Of course the present system could be extended to include
[} Gardner's
more typographical signs: _ = 0, A =l , .. , Z = 26, != 27, ?= 28, $= 29, .. , but then where
do we stop? Base 36? Base 50? Base 100? Without a uniform standard word values will
differ from system to system. The alphabet furnishes our only natural standard.

~ should be Here and there SV can illuminate an existing topic. Writing in Word Ways for February
, word was 1992, Christopher McManus introduced halfway words, which are word trios in which
is. BOOM! letters in the centre word lie midway in the alphabet between corresponding letters in the
ltieal word­ outer words, e.g., AGE - llG - SKI. It is a neat invention. However, now look at it thus:
ion, a new llG is the average or arithmetic mean of AGE and SKI, because Ih(AGE + SKl) = llG, a
ratched the particular instance of Ih(a+ b) = C, with a, b and C wints. Halfway trios thus form arith­
logarithms, metic progressions. The converse however need not be true: CAUSES is the mean of
probability , BINARY and CUBISM, CIGARS is the mean of LIES and FRAUDS, GHOST is the
), complex mean of FAITH and HOUSE, and HINDOO is the mean of LOOP and PROTON (to
.. Our new mention but a few), none of which conform to McManus' defmition. So halfway trios are
Ilathematics actually special instances of arithmetic means, a result of their definition in terms of form
(digit properties) rather than content (numerical value).

Sematria offers enormous opportunities for research in Dew realms of mathematical

wordplay but the difficulties entailed in tracking down wonhwhile finds should not be
underrated. Programming computers to perform arithmetic with words is fairly easy,
1 fits fairly getting them to recognize and hunt down fruity correlations is something else again .
. , measure Limitations of space have made it impossible to discuss algorithms here, but I shall be
~ematria , a glad to exchange ideas, great or small, on programs or other aspects of sematria with
re equaled. anyone interested. Meanwhile I can only encourage others in their explorations and
lumeroiogy wonder into what strange worlds this new development may in time lead us.
e a system
nct values.
Jint was of
rouble was
!d by zero.
In intuitive * Ross Eckler informs me that credit is due to Philip Cohen for first intrOducing a base 26 system (in
1S invisible which zero was A), in Word Ways for February 1977. I regret having been unaware of Cohen's pioneering
y dropped. work which might otherwise have been discussed here.