The Art of Decryption

A Heuristic Guide to Solving Monoalphabetic Substitution Ciphers

I’ve recently developed a renewed interest in cryptography. I decided that my skills could use some work, and so I set out to get some practice. I would grab chunks of text from documents and web pages and save them to text files. After a suitable period of time had passed, and I wasn’t likely to remember the original contents of the files, I would run them through a program that replaced all the letters with a shuffled alphabet, removed capitalization, and removed all non-alphabetic symbols (punctuation, spacing, numbers, etc.). I would then run the encrypted file through two frequency analysis program that gave me the counts of every 1, 2, and 3 letter sequence in the text, as well as any longer blocks of text that occurred more than once. Finally, with the frequency information in a spreadsheet, I would pull up a substitution program that would help me try out key values, showing me the ciphertext and the candidate plaintext simultaneously.

I started out with texts that were in the 600-1200 letter range, but recently I’ve been tackling much shorter texts in the 200-500 letter range. I’ve been temporarily stumped by some of them, and others, even short ones, sometimes just fall right into place. Actually, a longer body of text doesn’t always factor into easier decoding, although it does increase my confidence in the results of the frequency analysis. In the process of decoding these texts, I’ve developed some techniques that help to crack open the secrets held by the jumbles of letters before me.

This introduction serves to explain what conditions I was working under, and what kind of information I expect the reader to either have or be able to obtain. In summary, the initial information is a text file containing capital letters of ciphertext without spacing or punctuation. Analysis assumes that lists of 1, 2, and 3 letter combinations (letters, digrams, and trigrams respectively) is available and sorted by frequency, and that a similar list for plaintext is available.

Plaintext

For the record, my plaintext sample that I worked from was the complete Hitchhiker trilogy by Douglas Adams, in the original English with only a few spelling errors and some introductory material. While this text does contain a number of unusual letter combinations and some strange uses of letters (Adams would frequently make up words, and in doing so, he would favor the least frequent letters such as Q, Z, X, and J). Thanks to Zaphod Beeblebrox, the letters Z and X appear a bit more often than usual, as do the trigrams ZAP and BEE. It turns out that this has very little bearing on the relative frequencies of the letters. Q, Z, X, and J still appear in the bottom four spots, and E, T, A, O, I, N, H, S, R, D, L, and U top the list in that order. U doesn’t seem to be unfairly biased due to the British use of OU instead of O, and the most common digrams and trigrams appear to be roughly what they should be.

This is much better than a frequency analysis of the 9/11 Commission Report, which has far more of the letter Q than it should, and due to words like Al Qaeda and Iraq, it has a significant number of letters that follow Q instead of U (although it still seems that the letter most likely to appear second after a Q is E).

In my plaintext sample, the letters, in order of frequency, are E, T, A, O, I, N, H, S, R, D, L, U, W, C, G, M, F, Y, P, B, K, V, X, Z, J, and Q. There are a total of 1,152,456 letters in the sample, with E occurring 136,052 times and Q appearing 1,071 times. Total frequencies are listed in Table 1. The middle column contains the number of occurrences of the letter in the Guide, and the third column indicates what percent of the total that is.

E	136052	11.81
T	110940	9.63
A	94865	8.23
O	85077	7.38
I	81190	7.04
N	75116	6.52
H	74789	6.49
S	69698	6.05
R	64330	5.58
D	54228	4.71
L	49171	4.27
U	34741	3.01
W	27612	2.40
C	26937	2.34
G	26462	2.30
M	25103	2.18
F	24427	2.12
Y	24230	2.10
P	21582	1.87
B	17835	1.55
K	10891	0.95
V	10393	0.90
X	2210	0.19
Z	1829	0.16
J	1677	0.15
Q	1071	0.09

The 50 most frequent digrams and trigrams are listed in table 2. No frequency percentages are used. Instead, I have included the total number of instances of the letter combination. Note that these combinations can span words, since all non-alphabetic characters were removed prior to frequency analysis.

TH	38691	THE	22816
HE	33224	ING	10128
IN	22854	AND	8459
ER	17863	HER	5349
AN	16709	HAT	5328
ED	15562	THA	5178
RE	14310	ERE	5031
OU	12917	THI	4887
TO	12801	WAS	4694
NT	12757	DTH	4095
ND	12727	HIS	4032
AT	12708	HES	4006
ST	12629	TTH	3919
HA	12597	ETH	3872
ES	12570	ENT	3742
IT	12186	FOR	3716
NG	11833	YOU	3647
HI	11770	NTH	3559
EA	11193	RTH	3486
ON	11173	INT	3242
EN	11078	AID	3229
AR	10709	ART	3225
AS	10503	SAI	3180
ET	9639	ALL	3162
OR	9146	VER	3160
TI	9107	HIN	3059
TE	8756	EDT	3040
SA	8652	OTH	2804
IS	8626	OUT	2734
OF	8041	ITH	2689
WA	7969	GHT	2653
LE	7860	FTH	2611
SE	7772	EDA	2578
AL	7671	STH	2572
TT	7495	TER	2551
VE	7445	SHE	2533
LL	7162	OFT	2532
TA	7136	REA	2412
RO	6890	OME	2341
ME	6877	ONT	2309
NE	6740	HAD	2298
LY	6606	AST	2260
DT	6604	ONE	2153
RT	6591	TIN	2149
EL	6539	EST	2146
ID	6272	ESA	2139
LI	6245	NGT	2134
AI	6074	ION	2107
RA	6020	ATT	2098
SH	5937	EAR	2078

Some of these are biased by the text, but sometimes it’s merely important that a string appears on the list anywhere. Remember that there is no guarantee that any given digram or trigram will appear in a text. For example, I have decoded one ciphertext that did not contain the trigram THE. We will refer back to these tables.

Frequency Clues

Frequency analysis is the key to breaking a substitution cipher. This section will go over some of the most useful techniques for using the lengthy charts of gibberish that can be obtained through frequency analysis.

Accentuate the Positive:

Whenever you have added a new letter to the key, it is important to go and look for sections of text that can be part of larger words or phrases. Sometimes a lucky guess at some long word will yield several new letters and start an avalanche of new possibilities that lead to finishing up the decryption. Also, the deciphering endgame is entirely made up of looking through the text for spaces and filling in the blanks with appropriate letters. Every other trick in this paper is to help you get to the point where the remaining letters become obvious.

An example of this would be one time when I thought I had found a block that included _ISSI_N. A guess led to me finishing that as MISSION. This filled in a lot of gaps in the text, and soon helped me realize that it didn’t say MISSION, but rather MILLION.

Eliminate the Negative:

After each trial of a new key value, look through the text for strings that don’t make sense. Lots of consonants or vowels in a row, unlikely letter strings, and a general feeling of gibberish are good indications that you’ve made a mistake somewhere. Sometimes it’s a good idea to keep your error for a while to see if any of it reminds you of something else, but often it’s a good idea to get rid of bad key values if you can figure out which ones are likely to be bad.

Also, remember that sometimes strange letter combinations are perfectly legitimate. The Q in Iraq need not be followed by a U, and A can precede E. STS and TST are both legitimate combinations, but TLT is not as likely as LTL. Winnowing out unacceptable combinations is one area where a codebreaker must develop a feel for how the language works.

THE:

The first thing to do when staring at a new block of ciphertext is to find the most common trigram, THE. This is one area where I disagree with most other cryptographers, who recommend finding E first. Fundamentally, they may be about the same task, but my approach looks at more information up front.

THE can often be identified based on its influence on all three major frequency categories. In the list of trigrams, it is often at or near the top. The two digrams, TH and HE, should appear frequently, and usually more often than THE. Also, TH should occur more often than HE, but don’t get hung up on that. In the letter frequencies, both T and E should be near the top of the list (usually in the top two spots), and H should be about 1/4 to 1/2 of the way down the list.

If the most common trigram doesn’t seem to work, then work down the list of trigrams trying to find trigrams that match the right relative letter frequencies. It sometimes helps to limit your search to trigrams ending with the most common letter or two (especially when one letter occurs much more frequently than all of the others).

Another method that can help eliminate unlikely candidates is to look at the candidate TH digram and see how often the HT digram appears. While it is possible to have HT appear often in a text, it is significantly less common than TH in most texts. Also, HT is almost always preceded by G.

A word of warning should be given about other common words. In a specific short body of text, it is common to have a single word or variants of a word to appear many times. When this happens, trigrams from that word may dominate the list of trigrams. If some common trigram is always or almost always found either before or after another letter (or group of letters), then it is probably not THE. One amusing example of this was in a text where HER occurred most often, and it was represented by the ciphertext trigram HIM.

E:

Sometimes THE proves to be more elusive than you’re ready to deal with. When that happens, it is often possible to identify the letter E. It appears far more often than any letter in the English language, and is almost always on the top of the frequency charts. If one letter occurs a couple percent more often than the rest, then it’s probably E.

THAT:

Most bodies of text contain the string THAT. If you see one or more TH_T strings, then try out A in the middle. If this gives AE digrams, then you may want to question it, but remember that TH_T almost always requires a vowel, and that vowel is almost always A.

The ER/RE Effect:

Looking through the list of digrams, one might notice that ER occurs 4^th, and RE is 7^th on the list. This means that if the same digram appears frequently as both XY and YX, then it probably represents ER/RE (distinguishing which is E and which is R is easy, since E is more than twice as common as R).

Double Letters:

Only so many plaintext letters are likely to occur twice, like the Ts in the word leTTers. The most common doubles are TT, EE, LL, SS, and OO, with DD, FF, HH, MM, NN, PP, and RR being reasonably likely as well. If a letter other than T or E (which are hopefully identified by this point) that is doubled several times is likely to be L or S. If a doubled letter is in the top few spots in terms of frequency, then S and O are likely values. A, I, and U are almost never doubled, although it is not impossible (“spa activity”, “I inched…”, and “gnu udder” are examples).

Double Letters Special Case—IGGIs:

In “The Shadow of the Hegemon” by Orson Scott Card, a character deciphered a code by identifying that a string of four letters in the form XYYX meant IGGI, which was part of another character’s name. The rest of the text fell into place after that. These types of letter combinations are quite uncommon, and therefore of limited use in code breaking, but they can be useful at times. The most common IGGI patterns are ILLI, OTTO, ESSE, ETTE, HTTH, ALLA, OLLO, NOON, ITTI, ELLE, and ISSI. For the record, IGGI occurs twice in the Hitchhiker plaintext, and is 80^th on the list of IGGIs. Unless people are writing to you about a Wiggin, then it is unlikely that IGGI will appear in your plaintext.

QU:

It sometimes pays to look through the bottom few letters on the frequency list for possible translations of Q. If a very uncommon letter occurs more than once, and is always followed by the same letter, then the pair could be QU, especially if the third letter after it is E (which you should know by now). The letter U should be about halfway down the frequency list, so candidates for QU should be checked for a match with the appropriate frequencies. If there are letters that only occur once in the ciphertext, look for the letter that follows it and see if it could be Q. If the letter following a single letter isn’t in the top 10 on the frequency chart, then it is probably QU (J, X, and Z are almost always followed by a vowel, and typically that vowel is E).

Pick a Vowel:

If A, I, and O haven’t been found up to this point, then it may make sense to try them out, one at a time, in the top unused spots. As a general rule, when they come together, A comes after E, but I and O are more likely to come before E (I can go either way, but most of the time it doesn’t come after C or sounded like “ay”). This kind of guessing is tricky, and the only way to know if you have it right is to look at what it changes the plaintext to and see if it makes any sense.

Pick a Consonant:

When you’re taking wild stabs at the code, picking a vowel will often create strings of letters that have way too many vowels in a row. When that happens, replace one of the vowels with a consonant, and see what happens. While A, I, and O all occur frequently, and at about the same frequency, the consonants vary from T, which is very common, to Q, which is not at all common. So, when trying consonants, first use the most common ones like T (if you haven’t already found it), S, H (again, if needed), R, D, and L. Anything less common than that shouldn’t have to be found by this kind of haphazard method. Pay some attention to doubled letter frequencies when doing this, although most of the common consonants are likely to appear doubled.

Keep Trying:

If nothing seems to be working, then go back and try again. Look further down the lists for possible candidates. Try to come up with better explanations for strange letter patterns. If nothing else works, go away and come back later when you can look at it with a fresh eye.

Example Problem:

Consider the ciphertext:

UHRTQBJOXUPQOBQUOGKWHQLUXXWPQERLIGQOIROPU
ILDFKVQFUQTQUOGKHQRLIIHRIGKPRDSHIQLUWBJGU
OSHJGQFDOWXJLPWRUPHUWTJUBQVLQRMUOSXJLIHQJ
OFKIUGQUORGUODIQBHRICUIHLQEJLIQLWUYDWIPJO
IMOJCCHQOWJGQIUGQWUIUWRFUIIFQHRLPIJWCRFFJ
CVDIKJDYDWIWCRFFJCUIROPMQQESJUOSVQBRDWQIH
RIWCHRIKJDHRTQIJPJEJFKSLREHWSUTQOIJVJIHHQ
LXRIHQLROPHQLSLROPXRIHQLRLBHUQFDOWXJLPLRU
WQPOJUGGQPURIQBJOBQLOWIJUOTQWIUSRIJLWRDIH
JLUIUQWROOJDOBQPIDQWPRKREJFKSLREHJXYQWWUB
RWSLROPGJIHQLLDIHFDOWXJLPUOBFDPQPICJLQWEJ
OWQWIHRILRUWQPLQPXFRSWVDIBJDFPHRTQVQQOWIL
QWWLQFRIQP

Frequency analysis gives:

C	#	P	#	C	#	P	#	C	#	P	#	C	#	P	#
Q	55	E	136052	IH	11	TH	38691	HQL	6	THE	22816	FF	2	TT	7495
I	46	T	110940	RI	10	HE	33224	HRI	5	ING	10128	II	2	LL	7162
R	40	A	94865	HQ	9	IN	22854	ROP	5	AND	8459	QQ	2	EE	5676
J	36	O	85077	HR	9	ER	17863	IHQ	4	HER	5349	WW	2	SS	4485
U	36	I	81190	QL	9	AN	16709	SLR	4	HAT	5328	CC	1	OO	3730
W	36	N	75116	UO	9	ED	15562	XJL	4	THA	5178	GG	1	DD	1640
L	34	H	74789	JL	8	RE	14310	DOW	3	ERE	5031	HH	1	PP	1571
O	33	S	69698	LR	8	OU	12917	FDO	3	THI	4887	LL	1	RR	1563
H	28	R	64330	QW	8	TO	12801	HRT	3	WAS	4694	OO	1	FF	1277
P	25	D	54228	OW	7	NT	12757	IHR	3	DTH	4095	XX	1	NN	779
D	19	L	49171	QP	7	ND	12727	JLP	3	HIS	4032	AA	0	HH	774
F	18	U	34741	GQ	6	AT	12708	LRO	3	HES	4006	BB	0	MM	612
B	13	W	27612	LQ	6	ST	12629	OBQ	3	TTH	3919	DD	0	GG	570
G	13	C	26937	RO	6	HA	12597	OWX	3	ETH	3872	EE	0	CC	404
S	12	G	26462	TQ	6	ES	12570	QUO	3	ENT	3742	JJ	0	WW	325
X	11	M	25103	UI	6	IT	12186	RTQ	3	FOR	3716	KK	0	BB	285
K	10	F	24427	WI	6	NG	11833	UOS	3	YOU	3647	MM	0	YY	200
C	9	Y	24230	DI	5	HI	11770	WXJ	3	NTH	3559	NN	0	AA	157
E	8	P	21582	IJ	5	EA	11193	BJO	2	RTH	3486	PP	0	KK	139
T	7	B	17835	IQ	5	ON	11173	CRF	2	INT	3242	RR	0	ZZ	102
V	7	K	10891	IU	5	EN	11078	CUI	2	AID	3229	SS	0	UU	52
M	3	V	10393	JO	5	AR	10709	DIH	2	ART	3225	TT	0	II	38
Y	3	X	2210	OP	5	AS	10503	DWI	2	SAI	3180	UU	0	QQ	2
A	0	Z	1829	QO	5	ET	9639	EJF	2	ALL	3162	VV	0	JJ	0
N	0	J	1677	UW	5	OR	9146	FFJ	2	VER	3160	YY	0	VV	0
Z	0	Q	1071	BJ	4	TI	9107	FJC	2	HIN	3059	ZZ	0	XX	0

From this, it seems that Q=e, and IHQ=the. Everything seems to back this up, even with IHQ appearing 4th on the trigram list. Inputting it gives (with lowercase letters indicating plaintext)

UhRTeBJOXUPeOBeUOGKWheLUXXWPeERLtGeOtROPU
tLDFKVeFUeTeUOGKheRLtthRtGKPRDShteLUWBJGU
OShJGeFDOWXJLPWRUPhUWTJUBeVLeRMUOSXJLtheJ
OFKtUGeUORGUODteBhRtCUthLeEJLteLWUYDWtPJO
tMOJCCheOWJGetUGeWUtUWRFUttFehRLPtJWCRFFJ
CVDtKJDYDWtWCRFFJCUtROPMeeESJUOSVeBRDWeth
RtWChRtKJDhRTetJPJEJFKSLREhWSUTeOtJVJthhe
LXRtheLROPheLSLROPXRtheLRLBhUeFDOWXJLPLRU
WePOJUGGePURteBJOBeLOWtJUOTeWtUSRtJLWRDth
JLUtUeWROOJDOBePtDeWPRKREJFKSLREhJXYeWWUB
RWSLROPGJtheLLDthFDOWXJLPUOBFDPePtCJLeWEJ
OWeWthRtLRUWePLePXFRSWVDtBJDFPhRTeVeeOWtL
eWWLeFRteP

Near the middle of the second line, we see “thRt” which means that R=a.

UhaTeBJOXUPeOBeUOGKWheLUXXWPeEaLtGeOtaOPU
tLDFKVeFUeTeUOGKheaLtthatGKPaDShteLUWBJGU
OShJGeFDOWXJLPWaUPhUWTJUBeVLeaMUOSXJLtheJ
OFKtUGeUOaGUODteBhatCUthLeEJLteLWUYDWtPJO
tMOJCCheOWJGetUGeWUtUWaFUttFehaLPtJWCaFFJ
CVDtKJDYDWtWCaFFJCUtaOPMeeESJUOSVeBaDWeth
atWChatKJDhaTetJPJEJFKSLaEhWSUTeOtJVJthhe
LXatheLaOPheLSLaOPXatheLaLBhUeFDOWXJLPLaU
WePOJUGGePUateBJOBeLOWtJUOTeWtUSatJLWaDth
JLUtUeWaOOJDOBePtDeWPaKaEJFKSLaEhJXYeWWUB
aWSLaOPGJtheLLDthFDOWXJLPUOBFDPePtCJLeWEJ
OWeWthatLaUWePLePXFaSWVDtBJDFPhaTeVeeOWtL
eWWLeFateP

Everything’s looking good, but nothing else is sticking out at the moment. So, we move on to something else. Examining the frequency chart shows QL 5th on the list, and LQ 13th. From this, we guess that L=r.

UhaTeBJOXUPeOBeUOGKWherUXXWPeEartGeOtaOPU
trDFKVeFUeTeUOGKheartthatGKPaDShterUWBJGU
OShJGeFDOWXJrPWaUPhUWTJUBeVreaMUOSXJrtheJ
OFKtUGeUOaGUODteBhatCUthreEJrterWUYDWtPJO
tMOJCCheOWJGetUGeWUtUWaFUttFeharPtJWCaFFJ
CVDtKJDYDWtWCaFFJCUtaOPMeeESJUOSVeBaDWeth
atWChatKJDhaTetJPJEJFKSraEhWSUTeOtJVJthhe
rXatheraOPherSraOPXatherarBhUeFDOWXJrPraU
WePOJUGGePUateBJOBerOWtJUOTeWtUSatJrWaDth
JrUtUeWaOOJDOBePtDeWPaKaEJFKSraEhJXYeWWUB
aWSraOPGJtherrDthFDOWXJrPUOBFDPePtCJreWEJ
OWeWthatraUWePrePXFaSWVDtBJDFPhaTeVeeOWtr
eWWreFateP

Near the middle of this, we see “ttherXathera”, which could be “… her father …”. This fits in with the frequency of X, so we try X=f.

UhaTeBJOfUPeOBeUOGKWherUffWPeEartGeOtaOPU
trDFKVeFUeTeUOGKheartthatGKPaDShterUWBJGU
OShJGeFDOWfJrPWaUPhUWTJUBeVreaMUOSfJrtheJ
OFKtUGeUOaGUODteBhatCUthreEJrterWUYDWtPJO
tMOJCCheOWJGetUGeWUtUWaFUttFeharPtJWCaFFJ
CVDtKJDYDWtWCaFFJCUtaOPMeeESJUOSVeBaDWeth
atWChatKJDhaTetJPJEJFKSraEhWSUTeOtJVJthhe
rfatheraOPherSraOPfatherarBhUeFDOWfJrPraU
WePOJUGGePUateBJOBerOWtJUOTeWtUSatJrWaDth
JrUtUeWaOOJDOBePtDeWPaKaEJFKSraEhJfYeWWUB
aWSraOPGJtherrDthFDOWfJrPUOBFDPePtCJreWEJ
OWeWthatraUWePrePfFaSWVDtBJDFPhaTeVeeOWtr
eWWreFateP

At this point, more of it is starting to make sense. Since we have the context of a father, it is reasonable to expect that in the second line we have “PaDShter” meaning “daughter”. This almost looks like too good a match, especially since HT is almost always preceded by a G, so we try it out with P=d, D=u, and S=g.

UhaTeBJOfUdeOBeUOGKWherUffWdeEartGeOtaOdU
truFKVeFUeTeUOGKheartthatGKdaughterUWBJGU
OghJGeFuOWfJrdWaUdhUWTJUBeVreaMUOgfJrtheJ
OFKtUGeUOaGUOuteBhatCUthreEJrterWUYuWtdJO
tMOJCCheOWJGetUGeWUtUWaFUttFehardtJWCaFFJ
CVutKJuYuWtWCaFFJCUtaOdMeeEgJUOgVeBauWeth
atWChatKJuhaTetJdJEJFKgraEhWgUTeOtJVJthhe
rfatheraOdhergraOdfatherarBhUeFuOWfJrdraU
WedOJUGGedUateBJOBerOWtJUOTeWtUgatJrWauth
JrUtUeWaOOJuOBedtueWdaKaEJFKgraEhJfYeWWUB
aWgraOdGJtherruthFuOWfJrdUOBFudedtCJreWEJ
OWeWthatraUWedredfFagWVutBJuFdhaTeVeeOWtr
eWWreFated

At this point, we can see that O=n, and we can guess that F=l. We could also guess that “reEJrter” means “reporter”, but there’s no need to rush things.

UhaTeBJnfUdenBeUnGKWherUffWdeEartGentandU
trulKVelUeTeUnGKheartthatGKdaughterUWBJGU
nghJGelunWfJrdWaUdhUWTJUBeVreaMUngfJrtheJ
nlKtUGeUnaGUnuteBhatCUthreEJrterWUYuWtdJn
tMnJCChenWJGetUGeWUtUWalUttlehardtJWCallJ
CVutKJuYuWtWCallJCUtandMeeEgJUngVeBauWeth
atWChatKJuhaTetJdJEJlKgraEhWgUTentJVJthhe
rfatherandhergrandfatherarBhUelunWfJrdraU
WednJUGGedUateBJnBernWtJUnTeWtUgatJrWauth
JrUtUeWannJunBedtueWdaKaEJlKgraEhJfYeWWUB
aWgrandGJtherruthlunWfJrdUnBludedtCJreWEJ
nWeWthatraUWedredflagWVutBJuldhaTeVeenWtr
eWWrelated

“raUWedredflagW” and the blocks that follow it suggest that U=i, W=s, V=b, B=w, J=o, and T=v. “grandGJther” should be “grandmother”, giving G=m. We try these out:

ihavewonfidenweinmKsheriffsdeEartmentandi
trulKbelieveinmKheartthatmKdaughteriswomi
nghomelunsfordsaidhisvoiwebreaMingfortheo
nlKtimeinaminutewhatCithreEortersiYustdon
tMnoCChensometimesitisalittlehardtosCallo
CbutKouYustsCalloCitandMeeEgoingbewauseth
atsChatKouhavetodoEolKgraEhsgiventobothhe
rfatherandhergrandfatherarwhielunsfordrai
sednoimmediatewonwernstoinvestigatorsauth
oritiesannounwedtuesdaKaEolKgraEhofYessiw
asgrandmotherruthlunsfordinwludedtCoresEo
nsesthatraisedredflagsbutwouldhavebeenstr
essrelated

The first line now reads, “I have wonfidednwe in …” which makes me think that B=c. We also can see that E=p, K=y, M=k, C=w, and Y=j.

ihaveconfidenceinmysheriffsdepartmentandi
trulybelieveinmyheartthatmydaughteriscomi
nghomelunsfordsaidhisvoicebreakingfortheo
nlytimeinaminutechatwithreportersijustdon
tknowwhensometimesitisalittlehardtoswallo
wbutyoujustswallowitandkeepgoingbecauseth
atswhatyouhavetodopolygraphsgiventobothhe
rfatherandhergrandfatherarchielunsfordrai
sednoimmediateconcernstoinvestigatorsauth
oritiesannouncedtuesdayapolygraphofjessic
asgrandmotherruthlunsfordincludedtworespo
nsesthatraisedredflagsbutcouldhavebeenstr
essrelated

At this point, we’re done. The original text read,

“I have confidence in my sheriff's department, and I truly believe in my heart that my daughter is coming home,” Lunsford said, his voice breaking for the only time in a 10-minute chat with reporters. "I just don't know when. Sometimes it is a little hard to swallow, but you just swallow it and keep going because that's what you have to do."

Polygraphs given to both her father and her grandfather, Archie Lunsford, 72, raised no immediate concerns to investigators. Authorities announced Tuesday a polygraph of Jessica's grandmother, Ruth Lunsford, 73, included two responses that "raised red flags," but could have been stress-related.

(From a CNN article about then missing Florida girl Jessica Lunsford. For the record, I’m not morbid, but I typically grab paragraphs from CNN articles for my plaintexts.)