Modern translations are often accused of having “deleted” or “removed” Matthew 17:21 from the New Testament. The assumption in this accusation, of course, is that this verse must be original and therefore it was willfully removed. Modern translators, however, point out that this verse is not present in the earliest manuscripts we possess. They contend, then, that they have not deleted anything. Rather, later scribes added these words, and they are not part of what Matthew actually wrote. Simply pointing out that the KJV has the verse and that many later translations don’t is not proof that the verse is an original part of Matthew’s gospel. It may be that modern translators have not removed the verse, but rather that Scribes inserted the words and the KJV translators (possessing only the text of a few late manuscripts) kept the inserted words without realizing it. To decide between the two, we must consider the data.

The verse(s) in question

The verse under discussion here traditionally reads in older English translations:

“Howbeit this kind goeth not out but by prayer and fasting,” (Matthew 17:21, KJV).

Most modern translations lack this sentence, as does the Greek text upon which they are based. The discussion is closely linked to the parallel verse in Mark:

“And he said unto them, This kind can come forth by nothing, but by prayer and fasting,” (Mark 9:29, KJV).

This verse is present in all translations, though here too there is a textual variant. Many modern translations simply read:

“And He said to them, ‘This kind cannot come out by anything but prayer,'” (Mark 9:29, NASB).

Thus, a similar verse to Matthew 17:21 is universally accepted to be an original part of Mark’s gospel, though the reference to fasting is in question even here. While our main focus will be on the verse in Matthew, Mark’s version will necessarily play a role in the discussion.

The evidence for Matthew 17:21

Let’s begin by looking at the manuscript evidence and related physical data in favor of Matthew 17:21. The verse is found in the majority of Greek manuscripts. Now, the majority of Greek manuscripts are from a rather late date and mostly from just one regional area, so this kind of information can be misleading, but it is certainly worth noting that most medieval Greek manuscripts have the verse. The verse is also present throughout the Latin tradition, thus it is found in the vast majority of Latin manuscripts as well, both early and late. The verse is found in some important early Greek codices: Codex W (late fourth/early fifth century), Codex C and D (both fifth century), Codex E and L (both eighth century), Codex G (ninth century), and others. While the verse was not originally in the fourth-century Codex Sinaiticus, nor was it added by the early first corrector, a later scribe did add the verse into that manuscript as well, perhaps as early as the fifth or sixth century and probably no later than the seventh. The verse is found in some Syriac and Bohairic Coptic manuscripts as well and is the reading in the Armenian and Slavonic copies. It is also found in Origen of Alexandria’s commentary on Matthew,¹Origen’s Commentary on Matthew, Book 8, Chapter 7 which dates back to the third century, and in Chrysostom’s Homilies in the late fourth century.²John Chrysostom, Homilies on Matthew, Homily 57 Augustine (early fifth century) and several other Latin fathers likewise cite the verse as occurring in Matthew,³Augustine, Harmony of the Gospels, Book 2, Chapter 63 though such cases ultimately only tell us what we already know, i.e. that the verse was part of the Latin manuscript tradition.⁴Other early writers mention the words of the verse but do not specify if they are quoting them from Matthew or Mark, thus they cannot be appealed to as support for Matthew 17:21 At any rate, the verse was clearly widespread at a fairly early time in Church history and cannot simply be written off.

The evidence against Matthew 17:21

If manuscripts with the verse are early and diverse, so too are the manuscripts that lack it. The verse is absent in Codex B and in the original hand of Codex Sinaiticus.⁵It was not added in by the early first corrector either, but rather by the later second corrector no earlier than the fifth or sixth century and possibly as late as the seventh These early fourth-century manuscripts are the oldest copies we have of this passage, and they do not contain verse 21. The verse is also absent in Uncial 0281 (seventh century), Codex Θ and minuscules 33 and 892⁶The verse was added to this manuscript in the hand of a later scribe, but was not present in the original (ninth century). It is also present in other later manuscripts on into the middle ages. While the vast majority of the Latin tradition favors the verse, there are two Old Latin manuscripts that do not contain it: it^e (fifth century) and it^ff1 (eighth century). The verse is also missing in the earliest Syriac manuscripts: The Sinaitic Palimpsest (fourth century) and Curetonian Gospels (fifth century), as well as the later Palestinian Syriac. Most of the Coptic manuscripts also lack the verse, including early middle Egyptian manuscripts like Schøyen 2650 (early fourth century), all the Sahidic manuscripts, such as Bodmer XIX (fourth or fifth century), and some of the Bohairic.⁷Later ancient translations like the Georgian and Ethiopic are split, some manuscripts containing the verse and others lacking it. Additionally, the fourth-century Christian writer Eusebius also places this verse only in Mark 9:29 and not in Matthew. Thus, while the vast majority of late sources support the inclusion of the verse (though a minority tradition of its absence in Matthew persists even in those later generations), our early sources are rather split on the matter. Whatever happened to Matthew 17, an addition or a subtraction, it must have happened very early.

Possible explanations for the presence or absence of Matthew 17:21

There are only four possible ways that this situation could have arisen:

1. Matthew 17:21 is original and one or more scribes mistakenly omitted it.
2. Matthew 17:21 is original and one or more scribes intentionally deleted it.
3. Matthew 17:21 is not original and one or more scribes mistakenly added it.
4. Matthew 17:21 is not original and one or more scribes intentionally added it.

Either way, the original reading was preserved throughout the ages. Whether in the majority or the minority tradition, whether with the verse or without it, one of these options is what Matthew wrote. Nothing has been lost here. Further, even if the verse is not original in Matthew, most or all of it is original in Mark, and thus, either way, we have inspired testimony that Jesus did say at least most of what Matthew 17:21 says. As far as teaching goes, the only thing at stake in this discussion is whether or not fasting is required along with prayer in order to cast out certain demons. Obviously, this is by no means a central issue of the faith, and most King James Only churches I have attended are not actively trying to cast out demons, with or without fasting. Still, it is legitimate to be concerned with every word that God has revealed. We do not want to deny as Scripture anything that God truly inspired, nor do we want to call Scripture words of men which were added by a later scribe. The manuscript data is not, by itself, enough to settle the question. We have to consider the likelihood of our four possibilities.

Accidental omission

The first option is that a scribe (or even more than one scribe on different occasions) accidentally skipped or omitted this sentence while copying Matthew’s gospel. Others copied this scribe’s manuscript without noticing the error, and this led to the minority tradition in the manuscripts. This explanation is possible. When someone is hand-copying a document, simple mistakes like skipped words, phrases, or even skipping an entire line certainly happen. Most scholars, however, doubt that this is the case here. There are typical reasons to explain most scribal errors of this sort. When going back and forth between the page one is copying and the page one is writing on, one’s eyes might accidentally jump to the same word in a different place on the page, causing one to skip a block of text. Similarly, one might accidentally jump to a nearby word with a similar ending and copy on from there, having mistakenly skipped the words in between. And, of course, it is common to accidentally skip a line when working down the page. So, when there is text in one manuscript that is absent in another manuscript, scholars look for these kinds of features to see if it is easily explained by a common scribal mistake. In the case of Matthew 17:21, most scholars agree that none of the usual causes of accidental omission apply here, so there is no clear line of evidence to say that this is an accidental omission.

Of course, anyone who has done a great deal of hand-copying or transcribing knows that sometimes our mistakes are not as clean and logical as this. Sometimes we skip words or phrases and there is no clear or obvious reason why it happened. We didn’t mean to, and none of the usual explanations apply, we just seem to have randomly messed up. Thus, an accidental omission is still possible, but there are reasons to think it unlikely. First, these kinds of “random” mistakes are rarely clean and precise and usually result in a nonsense reading that is obviously a mistake. Here, if the verse is original, we are talking about the removal of one exact clause from beginning to end with no additional words removed and no part of the clause left behind. It is, of course, possible that this could happen by pure chance as a random error, but it is not probable.

Intentional deletion

This brings us to the second option. Maybe some scribe removed the phrase on purpose, perhaps to further some theological agenda, or because he had some erroneous reason for thinking it was not original. While these kinds of explanations make good headlines, we should always be careful about running to them too quickly. The fact of the matter is, scribes were, for the most part, dedicated to faithfully preserving the document in front of them, especially those who held the document to be sacred. The vast majority of textual variants can be explained by honest mistakes, mental lapses, misreading, or errors in judgment. They are accidents. We need not and should not rush to the conclusion of scribes willfully altering the texts they copied however they saw fit. This simply doesn’t fit the facts, especially among the New Testament manuscripts.

Further, while we can come up with all kinds of imaginative stories about why a scribe might do just about anything, there is no obvious reason why any scribe would desire to remove this particular verse. The power and importance of fasting became more emphasized as we move forward through church history, not less. The only early Christian documents that ever say any negative thing about fasting are not talking about all fasting but only about Christians who participate in traditional Jewish fasts (such as the Day of Atonement or the Jewish weekly fast days). Some later Christians saw this as a dangerous compromise and opposed it, but those same Christians firmly believed in Christian fasting traditions. Nothing in Matthew 17:21 supports Jewish weekly or annual fasts, so this is not a plausible reason why a scribe might have removed the verse. Indeed, there is no evidence-based reason why a scribe might willfully delete the verse, only imaginative speculations about why someone theoretically could want to do so. This is why Bruce Mezger concluded:

“Since there is no good reason why the passage, if originally present in Matthew, should have been omitted in a wide variety of witnesses, and since copyists frequently inserted material derived from another Gospel, it appears that most manuscripts have been assimilated to the parallel in Mk.9.29″⁸Bruce Metzger, A Textual Commentary of the New Testament, 2nd edition (German Bible Society, 1994) 43

Accidental addition

This raises our third possibility: could a scribe have accidentally added the phrase into Matthew? At first glance, this seems unlikely. How does a scribe accidentally add an entire phrase that was not in the text they were looking at? Isn’t that the kind of thing that, if it happens, has to be on purpose? Actually, no. In cases where a scribe regularly copies multiple documents with similar passages, the scribe may become so familiar with the one that he accidentally adds material from memory into the other. This is especially possible among the synoptic gospels (Matthew, Mark, and Luke) who tell so many of the same stories, sometimes even in similar words, but who also each contain information not found in the others. Thus, as Metzger points out, Matthew 17:21 could simply be an example of a scribe writing in from memory familiar content from Mark 9:29. He may not have even realized he had done it! My wife, who once memorized several of Paul’s letters in their entirety, often made this exact kind of mistake when reciting similar passages in, for example, 1 and 2 Thessalonians. Familiar content from one was accidentally transferred to the other without any intentional editing. Interestingly, many of the mistakes she made in reciting often had parallels in the manuscript tradition. Scribes made these kinds of mistakes too!

It’s also possible that a scribe put some paraphrased information from Mark in the margins of his copy of Matthew as an explanatory note and that a later scribe mistakenly copied it into the main text. As D. A. Carson notes:

“Occasionally, honest errors of judgment have led to the introduction of an error. For example, if a scribe accidentally left out a line or a few words, the corrector might put them in the margin. The next scribe who came along and copied this manuscript might reinsert the words into the text in the wrong place. Alternatively, the marginal note might have been a scribe’s comment rather than an integral part of the text; but the scribe who copied that manuscript might well have inserted the note into the new copy he was writing, thus adding something to the text of scripture that should not be there. No malice was involved, no intentional corruption of the text – just an error in judgment.”⁹D. A. Carson, The King James Version Debate: A Plea for Realism (Baker Books, 1979) 22-23

These are exactly the kinds of mistakes that led to many of the textual variants we have between our manuscripts, and Matthew 17:21 may well be such a case. It is, at least, a highly plausible explanation.

Intentional addition

As we noted above, the idea of scribes willfully changing the text to suit their own purposes should not be entertained lightly. This would have been much rarer than modern sensationalists would love to have us believe. Scribes were not often given to change the text simply because they thought it should say something else. We would need to have good reason based on hard evidence to conclude that a scribe would add a verse like this into the text. One might argue, for example, that a scribe added this material from Mark 9 to intentionally harmonize the two texts and thus undermine accusations of contradiction between the gospels. But, without any source to indicate that this passage was being used at the time as an example of contradiction by opponents of Christianity or some other hard line of evidence, this remains pure conjecture. Matthew and Mark do not actually contradict one another here, even without Matthew 17:21. We have no particular reason to think that any scribe would have been concerned about this. Thus, barring further evidence, willful addition does not appear to be our best option, especially given the plausibility of a purely accidental addition.

The complication of Mark 9

Of course, all of this thus far assumes that Mark 9:29 is basically parallel to Matthew 17:21. As we have noted above, there is a complication in this. Mark 9:29 itself has a notable textual variant where, in some early manuscripts and thus in most modern translations, Mark mentions only prayer and not fasting. The words “and fasting” are missing in three early Greek copies: Codex B, Sinaiticus (both fourth century), and Uncial 0274 (fifth century). They are also missing in the early Latin manuscript, Codex Bobiensis (late fourth/early fifth century). Likewise, the earliest citation by a church father, Clement of Alexandria (late second or early third century),¹⁰Eclogae Propheticae 15:1 lacks the reference to fasting. Less significantly, a later Georgian translation also lacks it. This is a very small group of witnesses, but all but the Georgian translation represent some of our earliest copies of this text and thus should not be ignored simply for being in the minority.

Yet, the manuscripts containing “and fasting” are quite impressive: P⁴⁵ (third century), our earliest manuscript for this section of Mark, is fragmentary and missing the place on the page where “and fasting” would be. However, when one measures the size of the letters and space on each line and estimates which text would better fit the missing space, it seems that “and fasting” was probably there. The words are certainly found in Codex W (late fourth or early fifth century), Codex A (early fifth century), The second corrector of Codex Sinaiticus (fifth, sixth, or seventh century), Codex C and D (both fifth century), Codex N and Σ (sixth century), Codex E (eighth century) Codex L (eighth century), Codex G (ninth century), numerous other significant Uncial texts, and all of the many later Medieval manuscripts. Likewise, every Old Latin manuscript other than Codex Bobiensis supports the reading “and fasting.” These include such notables as it^a (fourth century), it^b and it^ff2 (both fifth century), it^d (sixth century), and many others. The later Latin vulgate tradition also contained “and fasting,” as did the Coptic in both the Sahidic and Bohairic dialects. The words are also found in the later Gothic and Slovenian manuscripts. There is also Georgian manuscript evidence for the words, thus making the Georgian tradition split on the matter.

The Syriac is an interesting case, but one that overall favors the reference to fasting. Some later Syriac manuscripts match the reading “prayer and fasting” exactly. Most of the Syriac texts, including all of the earliest manuscripts, have a similar but reversed reading of “fasting and prayer.” This reversed reading is also found in some Bohairic manuscripts and in the Armenian and Ethiopic manuscripts. While this variation may raise some questions, it seems most reasonable to say that these favor the presence of “fasting” in the Greek texts from which they were translated.

As far as early Christian writings go, there are no relevant examples of an explicit citation of Mark containing fasting, but there are more ambiguous references that are worth noting. Several writers cite Jesus as having said “prayer and fasting” without reference to which gospel they got the words from. Thus, it seems likely that they either had a copy of Matthew that contained 17:21 or they had a copy of Mark that contained “and fasting” at 9:29. Either way, their voice is relevant here. Some of the earliest examples include:

“according to the teaching of our Lord, who hath said:  This kind goeth not out but by fasting and prayer, offered unceasingly and with an earnest mind,” (Pseudo-Clementine Epistles Concerning Virginity, First Epistle, Chapter 12).

“This kind of devils is not cast out but by prayer and fasting,” (Ambrose, Letter to the Church at Vercallae, Section 15).

Of course, these citations are all late enough that, if “and fasting” is a later addition and not original to Mark, the addition would likely have already occurred by the time these men were writing. Either way, What is most important for our purposes here, however, is that, if Mark 9:29 is the source of Matthew 17:21, it seems to demand that Mark 9:29 already contained the words “and fasting” by the time Matthew 17:21 emerged. And, as we noted above, the verse in Matthew, if it is an addition, arose very early on. How would scribes have become so habitually committed to the reading “prayer and fasting” in Mark that they transferred it to Matthew’s text if Mark 9:29 itself had only recently gained the words “and fasting”? This is, indeed, a serious complication! Yet, it is not one without an answer.

Prayer, fasting, and the early church

Fasting became increasingly important in the early church. In one second-century document, for example, the writer stated that “Fasting is better than prayer.”¹¹Pseudo Clement, 2 Clement, Chapter 16. Indeed, we have clear examples where fasting was directly added into citations of New Testament passages about prayer in very early Christian documents. For example, the Didache, a writing from the late first or early second century AD, reads:

“The teaching of these words is this: Bless those who curse you, and pray for your enemies, and fast for those who persecute you. For what credit is it if you love those who love you? Do not even the Gentiles do the same?” (Didache, Chapter 1).

This is clearly an allusion to Jesus’ words:

“But I say unto you, Love your enemies, bless them that curse you, do good to them that hate you, and pray for them which despitefully use you, and persecute you…For if ye love them which love you, what reward have ye? do not even the publicans the same?” (Matthew 5:44, 46, KJV).

But, where Jesus said to pray for those who persecute you, the author of the Didache said that you should fast for those who persecute you. This is, of course, not the willful alteration of actual biblical manuscripts, but rather an interpretive paraphrase of Jesus’ words. Fasting was so important and tied so closely to intercessory prayer that the author clearly assumed that Jesus’ command to pray included fasting. Similarly, Tertullian, the Latin father of the late second and early third century wrote of Jesus that:

“He taught likewise that fasts are to be the weapons for battling with the more direful demons,” (Tertullian, On Fasting, Chapter 8).

This would seem, at first glance, to imply that his copy of either Matthew or Mark included the words about fasting, but this is not necessarily the case. In the same chapter, Tertullian also writes:

“Finally, granting that upon the centurion Cornelius, even before baptism, the honorable gift of the Holy Spirit, together with the gift of prophecy besides, had hastened to descend, we see that his fasts had been heard,” (Tertullian, On Fasting, Chapter 8).

Here, Tertullian references Acts 10 and the story of Cornelius. He claims that “his fasts had been heard.” But is that what the text says? The angel actually says to Cornelius:

“And said, Cornelius, thy prayer is heard, and thine alms are had in remembrance in the sight of God,” (Acts 10:31, KJV).

Previously, in verse 30, it does mention that Cornelius had been fasting, but this is stated in passing. Nowhere does the text actually emphasize the fasting, though it repeatedly emphasizes Cornelius’ prayer and that God was pleased in it. By Tertullian’s day, however, fasting was crucial. Thus, when the angel says “thy prayer is heard,” Tertullian interprets that to mean “thy fasting is heard,” and builds his teaching about fasting upon this assumption. Thus, we see a very early interpretive pattern that adds or changes citations of the text to include fasting where only prayer was originally mentioned. These authors may not have even done so willfully. In their minds, they tied prayer and fasting together, fasting being the greater. They may well have written “prayer” as “prayer and fasting” without even realizing they did it simply because it is the way they understood the meaning and intention of the text. In such a church culture, a scribe might have done precisely the same thing, either accidentally or with the intent of clarifying the meaning of the text, just as we see in these early citations. This is not mere conjecture but rather based on the actual evidence of early citations in these known Christian writings from precisely the time period in the earliest centuries where these variants must have arisen!

An example

We see an example of something like this elsewhere in Scripture. Note the similar problem we find between the KJV and modern translations in 1 Corinthians:

“Defraud ye not one the other, except it be with consent for a time, that ye may give yourselves to fasting and prayer; and come together again, that Satan tempt you not for your incontinency,” (1 Corinthians 7:5, KJV).

“Stop depriving one another, except by agreement for a time, so that you may devote yourselves to prayer, and come together again so that Satan will not tempt you because of your lack of self-control,” (1 Corinthians 7:5, NASB).

In this case, the evidence in favor of the modern reading without “fasting and” is overwhelming. P⁴⁶ (late second or early third century), not only the earliest manuscript of this passage but one of the earliest manuscripts of Paul’s letters in existence, only mentions prayer without reference to fasting. The later P¹¹ is damaged at this point on the page, but based on the available space in the missing section and the scribe’s handwriting, it seems the words were missing in this papyrus as well. Codex B and the original hand of Sinaiticus (both early fourth century) also lack the reference to fasting,¹²Once again, the second corrector some few centuries later added the words in different handwriting. as do Codex A, C, and D (fifth century), Codex F, G, P, Ψ (ninth century), Minuscule 33 (ninth century), and numerous later manuscripts. The entire Latin tradition, both in the Old Latin and the Vulgate, lack “fasting and,” as do the Coptic dialects. The Armenian and Ethiopic manuscripts lack the words, as do some of the Georgian manuscripts. Clement of Alexandria (late second/early third century), Tertullian (late second/early third century),¹³Tertullian, On Exhortation to Chastity, Chapter 10 Origen (third century),¹⁴Origen’s Commentary on Matthew, Book 14, Chapter 2 Cyprian (third century), Methodius (third century), Dionysius of Alexandria (third century)¹⁵The Epistle to Bishop Basilides, Canon 3 Augustine (late fourth/early fifth century),¹⁶Augustine, On Original Sin, Chapter 43 Jerome, (late fourth/early fifth century),¹⁷Jerome, First Letter to Pammachius, Section 15 (Letter 48 in the NPNF series, part 2, volume 6) Gregory the Great (sixth century)¹⁸Gregory the Great, Book of Pastoral Rule, Part 3, Chapter 27 and numerous other early Christian writers quote this passage, all lacking the words “and fasting.” Interestingly, a document known as the “Testaments of the Twelve Patriarchs” may also lend its voice. Originally a pre-Christian Jewish document (fragments of it were found among the Dead Sea Scrolls, though they differ somewhat from our later versions), most scholars believe that it was edited by a Christian scribe in the second century AD. At one point it reads:

“For there is a season for a man to embrace his wife, and a season to abstain therefrom for his prayer,” (Testaments of the Twelve Patriarchs: Testament of Naphtali).

Most scholars believe that this was brought in from 1 Corinthians 7:5. Others have argued these words are original and that Paul was alluding to this earlier Jewish tradition in his own words. Either way, this provides additional, if a bit more circumstantial, evidence that the verse originally lacked the reference to fasting.

On the other side, “fasting” was added to Codex Sinaiticus by the second corrector (fifth, sixth, or seventh century). It is present in Codex L (8th century), Codex K (9th century), Uncial 0150 (9th Century), and the majority of medieval Greek manuscripts. The old Syriac manuscripts were gospel texts and thus lack Paul’s letters, but the later Syriac tradition affirms the reference to fasting here. The later Gothic, Slovenian, and some Georgian manuscripts contain “fasting.” No Latin manuscripts contain it, but one late Latin translation of Origen adds the reference. The earliest place we find the “fasting ” reference in this verse is in citations in other writings. The 4th-century poet Ephraem the Syrian cites it this way in Syriac, and John Chrysostom (late fourth/early fifth century) and Theodoret (fifth century) cite it in Greek with the fasting reference. Of course, such citations may have been interpretive additions, as we saw in the Didache and Tertullian earlier. Thus, the earliest certain reference we have of this actually being part of the text is not until the second corrector of Sinaiticus who could be as late as the seventh century. In all the time before this, we have numerous examples from wide geographical regions and going back to remarkably early dates that all agree that “fasting” was not originally in the text. If this could happen in 1 Corinthians in the seventh century and still eventually become the medieval majority reading, there is no reason to think that such an interpretive scribal error could not have occurred in the second century in Mark 9 and come to garner just as much influence, also giving rise early on to the parallel reading in Matthew 17.

Pulling it all together: What do we make of Matthew 17:21?

So where does all this leave us? Our most plausible explanation for Matthew 17:21 seems to be that it is not the original text but was brought over from Mark 9:29 by an early scribe. Mark 9:29 may well have originally contained the words “and fasting,” which would make the KJV correct in that verse against most modern translations. However, even here, early Christian citations of other passages provide a quite plausible way that “and fasting” may have come into Mark at an early date and still provided the reading for Matthew 17. At any rate, all of this drew our attention to 1 Corinthians 7:5, where the modern translations are certainly correct over against the KJV, thus refuting the KJV Only position regardless.

Still, considering the compelling manuscript data on both sides of our main gospel texts here, it is not impossible that Matthew 17:21 and the long form of Mark 9:29 might both be original. Thus, the wisest course seems to be that chosen by the NASB and NKJV translators, who retain both verses in the main text but include a footnote alerting the reader to these questions.