Thursday, 24 December 2015

Christmas and its supposed pagan links

Christmas ... blends elements including both the feast of the Saturnalia and the birthday of Mithra.
-- Encyclopaedia Britannica, 'Roman religion'
Every year around Christmas, the world's mythbusters love to remind everyone that Christmas is really a pagan festival, and that Christians plagiarised it by enforcing their own beliefs on a pre-existing set of customs. The choice of festival varies: sometimes it's Saturnalia, sometimes the winter solstice, sometimes the birthday of Mithras, sometimes a Roman civic festival of Sol Invictus ('the unconquered sun'). Often it's a combo: for example, you may hear that Saturnalia was a solstice festival that the Christians corrupted.

Many Christians seem happy to accept this picture of things too, minus the hostile spin. After all there doesn't need to be anything very sinister about using a well-known festival to hold a celebration of your own. It's happened to Christian festivals too: St Valentine's Day was a Christian feast day for a legendary martyr of the late 3rd century, now it's mainly a day for celebrating romantic love. No evil conspiracy, and hardly anyone actually minds. Why not adapt a pre-existing festival for new purposes?

Anyway, soon the mythbuster-busters come out to play too. They'll dutifully point out:
  • There's no evidence to link Christmas to Saturnalia, and in fact Saturnalia continued to be celebrated by Christians, alongside Christmas for at least a century and probably longer.
  • Sol Invictus (a) was a specific local cult associated with a specific temple in Rome, (b) there were numerous sun cults in Rome and Sol Invictus was just one of them, and (c) the festival of Sol Invictus on 25 December isn't attested any earlier than Christmas and is very likely the later of the two.
  • Not only did Mithras' birthday have no influence on Christian thinking: Mithras didn't even have a birthday. He emerged from a rock.
  • The solstice is not on 25 December, and any Greco-Roman armed with a gnomon would be able to tell that reasonably accurately.
In the early stages of researching this post I was firmly on the side of the mythbuster-busters. That's mostly still true: Christmas has absolutely nothing to do with Mithras or Saturnalia.

But not entirely. There is a bit of mythbuster-buster-busting to do here. The most important point, as it turns out, is that there almost certainly is a link between Christmas and the solstice. That, in turn, creates a fairly strong presumption that there is a link between the date of Christmas and the tradition of linking Jesus with sun imagery. But as we shall see, it does not follow that there was any kind of plagiarisation between Jesus and Sol Invictus, in either direction.

Let's do the mythbuster-busting first, then we'll get onto the mythbuster-buster-busting. (This'll be a long post, in honour of the season: I'll change gear down after the new year.)

1. Saturnalia

Saturnalia was a Roman festival of Saturn on 17 December. Celebrations continued after Saturnalia for several days, finishing sometime between 19 and 23 December depending on which century we're talking about. It may originally have been a farming festival; the evidence is unclear. It was associated with an overturning of various regulations and social norms: free citizens often wore a freedman's hat, the pilleus, and played gambling games that were normally illegal; slaves would dine alongside their masters, or even act the part of master. People would exchange gifts of candles and clay figurines.

There's no link to Christmas. The date? No: they're not on the same date. (Hey, what's eight days between friends? Well, it's the difference between corroboration and no corroboration.) The fact that Saturnalia might have had some chance to have an influence on Christmas isn't evidence that it did.

Christmas trees? Yule logs? Holly and ivy? No, none of those come from Saturnalia. Individual customs may have pagan origins in some cases -- the Yule log may possibly be based on a late mediaeval Anglo-French or South Slavic custom -- but that has nothing to do with the claim that Christmas is based on Saturnalia.

Gift-giving? No: the magi may have given tribute to Jesus (traditionally at Epiphany, 6 January), but the modern custom of gift-giving at Christmas only goes back to the 16th century. That is when Luther introduced the Christkind in an attempt to discourage veneration of St Nicholas, who was associated with gift-giving on his feast day of 6 December. In late mediaeval Germany gift-giving had also been associated with the feast of the Holy Innocents, on 28 December. It's possible that Christmas charity from aristocrats to the poor goes back a bit further. But there's certainly no evidence to suggest continuity all the way back to when Saturnalia was still being celebrated.

And that brings us to the kicker: Saturnalia was still being celebrated, by Christians to boot, at least as late as the 5th century, alongside Christmas. By that time it was no longer a festival in honour of Saturn, in much the same way that Christmas has little to do with Jesus for modern atheists. But it'd be bizarre to conclude that Christmas, which was certainly being celebrated in the 4th century, was based on another festival that continued to co-exist with it for at least 100 years (and probably longer).

2. Mithras

Mithraism was a popular cult in the western Roman world in the first few centuries CE. Very little textual evidence about Mithraism survives: we have to rely heavily on the archaeology of mithraea, underground churches of Mithras. And interpretation of that evidence is very often uncertain and controversial. Popular accounts are even worse -- more games of speculation than anything else.

Roger Pearse has written some good online catalogues -- here, and here -- of rebuttals to the claim that Mithraism and Christianity had anything to do with each other. To sum up:
  • No, Mithras didn't have a virgin birth, and he wasn't born on 25 December. Mithras wasn't even born: he came out of a rock.
  • No, three wise men did not visit Mithras at his birth. (a) He didn't have a birth. (b) This idea is entirely derived from a single 1864 book that seems to have mistaken Mithras' attendants Cautes and Cautopates for 'wise men'. There isn't a shred of ancient evidence for it.
  • There is no good evidence of a concept of salvation through blood in Mithraism. This idea was based on a single inscription, but its reading isn't remotely clear-cut and it requires very heavy supplementation to get it to mean anything at all. (Here's what the inscription looked like in 1930; here's the most recent professional sketch of it. Go on, you try deciphering it.)
  • No, Mithras didn't die on a cross. That's completely made up.
  • Mithras didn't have twelve disciples, get buried and rise three days later, have a festival that coincided with Easter, and he didn't get called a "good shepherd", "the way, the truth, and the light", "logos", "redeemer", or "Messiah". All of these things were made up by Dorothy Murdock, writing under the pseudonym "Acharya S", in a series of books starting in 1999.
There's no connection at all. Let's just drop this one: it's silly, and life is short.

3. How far back does Christmas = 25 December go?

The earliest explicit evidence for the 25 December date of Christmas is in a document known as the 'Chronography of 354', a compilation of histories and calendars compiled in the year 354. One calendar, a catalogue of martyrs' feast days, has the following as its first entry:
VIII kal. Ian. -- natus Christus in Betleem Iudeae.
Eight days before the kalends of January (i.e. 25 December): Christ born in Bethlehem in Judaea.
But 354 isn't when Christmas was invented, it represents the earliest explicit attestation of Christ's birth being celebrated on 25 December. The likelihood is that the choice of date was earlier; though it's arguable whether that means a couple of decades earlier, or a couple of centuries earlier.
Digression #1: false leads. A number of online sources claim that Pope Julius I instituted 25 December as the date of the nativity in the year 350. This is untrue. The source usually cited is 'Manual of Liturgical History, vol. 2 (1955)', a book which on further inspection turns out not to exist. Wikipedia instead cites the History Channel -- and we all know how reliable they are. Other sources cite Julius Africanus as evidence that the date was fixed by the early 3rd century; but the surviving fragments of Africanus contain no reference to this (see S. Hijmans, 'Sol Invictus, the winter solstice, and the origins of Christmas', Mouseion 3 (2003): 377-98, at 377 n. 3; cf. Wallraff's 2007 edition of the Africanus fragments).
There are a few earlier sources that can potentially be taken as pointing towards celebration of Christ's birth -- or the Incarnation, depending on the theological flavour of the source ­­-- at an earlier date. They are:

  1. Clement of Alexandria, late 2nd/early 3rd cent., reports (Stromateis 1.21.145-6) that some people assigned Christ's birth to 'the 25th day of Pachon', and others to 'the 24th or 25th day of Pharmouthi'. These are dates in the Alexandrian calendar. When you convert them to the Julian calendar they're nowhere near December: they come out as 20 May and 19/20 April, respectively. However, the focus on the 25th (or 24th) day of the month is a tad conspicuous. Still, that could just be coincidence. Moreover, there are alternative interpretations of Clement's dates: see S. K. Roll, Towards the Origins of Christmas (1995), pp. 77-9.
  2. A reference to Christ's birth on 25 December appears in an early 3rd­ century writer, Hippolytus of Rome (Commentary on Daniel 4.23.3), but the reference is useless: there's no doubt that it is an interpolation of mediaeval date. (I give the text of the interpolation at the end of this post; see also Roll, pp. 79-81, esp. p. 80 with n. 106.)
  3. It was traditional among ancient Judaeo-Christian writers to treat prophets and saints as having the same date for their birthday and death-day. A modification to this appears starting in Clement, who reinterpreted the word 'birth' (genesis) as referring instead to conception (Strom. 3.12.83.2). As a result, in this typological thinking, the death-day coincided with the day of conception and the birthday fell exactly nine months after the death-day. Now, by the 2nd century, Christians were celebrating Jesus' death and resurrection on Pascha, 14 Nisan in the Hebrew calendar, or alternatively on 16 Nisan, corresponding to Good Friday and Easter Sunday respectively; the late 2nd and early 3rd centuries saw a controversy over which date was more important, the Quartodeciman controversy. Pascha shifts around each year, since the Hebrew calendar is lunar. If it were believed that 14 Nisan fell on 25 March in the year of Jesus' death, typological thinking would consequently put his genesis (conception) on the same date, and his birth nine months later on 25 December. This is exactly what the spurious reference in Hippolytus, above, claims; again, see the end of this post for the text. More details in Roll, pp. 79-80.
  4. An anonymous text dating to 243, the pseudo-Cyprianic De Pascha computus or 'computations concerning Pascha', at sections 18-23 puts the creation of the cosmos on 25 March; the creation of the sun on 28 March (the fourth day of creation); and therefore also the genesis of Christ, the 'sun of righteousness' (a phrase from Malachi 4:2 which according to Christian tradition refers to Jesus), on 28 March. The author claims divine guidance for these dates. If by nativitas the author means 'conception' -- as indeed seems likely, with the parallel of Clement's use of genesis -- that would then translate to a birthday of 25 December or 28 December, depending on how much of the author's argument you reject.
The first two of these earlier references are uncertain and wrong, respectively, but the third and fourth are decidedly possible. The upshot is that 354 is the earliest uncontroversial date for an assignment of Jesus' birth to 25 December; but the date may well have been already decided by the Quartodecimans in the mid-2nd century, 200 years earlier.

4. The winter solstice

The winter solstice does not fall on 25 December. The ancients were capable of identifying the solstices to within a day or so using gnomon measurements; and it seems to have been absolutely routine to do so.

However, there's more to this one than meets the eye. The Julian calendar was not as well calibrated as the modern Gregorian calendar, so the date of the solstice drifted with respect to the calendar date. Around the time of Jesus' birth, the solstice was on 23 December; by the time of the Chronography of 354, it was 20 December. (In 1580, just before the Gregorian calendar was introduced, it had drifted all the way back to 11 December!)

And yet Roman writers regularly quote the date of the solstice as the 25th of December. Here's Columella, a 1st century CE agricultural writer, on the subject (De re rustica 9.14.12):
ab occasu Vergiliarum ad brumam, quae fere conficitur circa VIII kalendas Ianuarii in octava parte Capricorni ...
From the setting of the Pleiades to midwinter, which occurs roughly around the 8th day before the kalends of January (i.e. 25 December), at 8° in Capricorn ...
We find similar wording in Pliny the Elder, also 1st cent. CE (Natural history 18.221):
... omnesque eae differentiae fiunt in octavis partibus signorum, bruma Capricorni a. d. VIII kal. Ian. fere.
... and all these changes occur at 8° in the (zodiacal) signs, the winter solstice in Capricorn on roughly the 8th day before the kalends of January (i.e. 25 December).
Note the parallels: the use of fere 'roughly', and the reference to 8° in Capricorn (most Babylonian-Greek astronomers reckoned the equinoxes and solstices as occurring when the sun was in the centre of each sign, i.e. at the 8° mark within that sign). Servius, in the 4th century, quotes the date by itself without the other parallels (commentary on Aeneid 7.720).

(Here, by the way, are links to published editions: Columella, and Pliny. The dating of the solstice to the 25th obviously worried the translators of these editions: they both mistranslate the date. This doesn't reflect any ambiguity over how to interpret Roman dates: there's absolutely no doubt that a. d. VIII kal. Ian. means 25 December. The translators here are either slipping up or being dishonest.)
Digression #2: why do Roman writers report the date of the solstice as 25 December? The textual parallels between Columella and Pliny, and the fact that the date was already wrong by their time, suggest they are both quoting an older source. A few pages earlier Pliny discusses three treatises by Sosigenes, who designed the Julian calendar (Nat. hist. 18.212). Sosigenes can't be the ultimate origin of the date either: in his time, in 46 BCE, the solstice had already drifted to 23 December. But he's fairly likely to be Columella's and Pliny's immediate source. Sosigenes must surely have been aware that the 25 December date was already inaccurate -- hence the fere 'roughly' in both Columella and Pliny. (It's likely that Sosigenes wrote in Greek, but fere could easily be a translation of φαύλως. And it's not actually impossible that he might have written in Latin, working for a Roman audience as he was. [Additional note, 19/1/16: the fact that the sources that give the solstice as 25 December are in Latin, and that there are none in Greek, also tends to suggest that Sosigenes wrote in Latin. Pliny is our only source for Sosigenes.])
R. Hannah, Greek and Roman Calendars (2005) p. 151, suggests Hipparchus, in the 2nd century BCE, as the ultimate source. Pliny's discussion of the periods between equinoxes and solstices in the surrounding passage appears to be taken from Hipparchus' figures (94½ days from vernal equinox to summer solstice, etc.; see O. Neugebauer, History of ancient mathematical astronomy vol. 1 pp. 57-8 and p. 307, citing testimony from Ptolemy's Almagest). However, Hipparchus still isn't early enough to get a solstice on 25 December. The last time the solstice actually fell on 25 December, using the retrojected Julian calendar, was in 230 214 BCE. And Columella states explicitly that he is departing from Hipparchus' practice: Hipparchus treated the equinoxes and solstices as occurring at the start of each zodiacal sign rather than at the mid-point; Columella/Sosigenes chooses instead to follow the more traditional practice of Eudoxus (early 4th cent. BCE) and Meton (5th cent. BCE), which was derived from Babylonian astronomy. And at the time of Meton and Eudoxus, the solstice was indeed usually on 25 December (Julian calendar, retrojected).
So Eudoxus must be the ultimate source of the date. The probable timeline is:
  • 4th cent. BCE: Eudoxus quotes the date of the solstice as 25 December. (He didn't use the Julian calendar, of course, but the date was later converted into the Julian calendar; we know the same conversion happened with Hipparchus' observations. See Neugebauer, p. 276, for examples of Ptolemy converting dates quoted by Hipparchus.)
  • 2nd cent. BCE: Hipparchus, as Columella and other ancient sources tell us, treats the solstices and equinoxes as occurring at the beginning of each zodiacal sign, not at the 8° mark as in Babylonian astronomy.
  • ca. 46 BCE: Sosigenes quotes Eudoxus' date. He realises that it's no longer accurate, so he qualifies it as approximate (fere); he also mentions the sun being at the 8° mark in Capricorn, following traditional Babylonian-Metonic practice rather than Hipparchus.
  • 1st cent. CE: Columella and Pliny quote from Sosigenes' account; Columella also mentions Hipparchus' alternate practice.
  • 4th cent. CE: Servius mentions 25 Dec. as the date of the solstice. By his time it's wrong by five days, but nonetheless a traditional piece of common wisdom.
Even if you don't buy my argument in the digression above about the ultimate origin of the date, Columella and Pliny leave no doubt that 25 December was popularly regarded as the solstice.

Saturnalia, as we saw, is off by eight days: that's mere coincidence. But when Christmas falls exactly on the traditional date of the solstice, it's a bit more of a stretch to say it's coincidence.

5. Jesus and solstice festivals?

Sol Invictus. Many modern readers see a link between Christmas and the festival of Sol Invictus, 'the unconquered sun'. This is often combined with a claim that (a) this equates to a celebration of Mithras and/or his birthday; or (b) Christmas is derived from the cult of the emperor, who in certain periods was often depicted with sun iconography.

The ultimate basis for the idea is a section of the Chronology of 354 known as the Philocalian calendar, which lists off civic festivals and other occasions in the city of Rome. In December it lists the following item:
VIII -- N̅·INVICTI·C̅M̅·XXX
8th (day before kalends of January, i.e. 25 December) -- festival of the Unconquered: thirty games ordered
Here N̅ stands for natalis ('festival') and C̅M̅ stands for circenses missi ('games ordered').

The confusion with Jesus is encouraged by the fact that in Latin natalis usually means 'birthday': so you will often hear that the birth of Christ was chosen to coincide with the 'birth' of Sol Invictus. That is not the meaning of natalis in the Chronology of 354: there it refers to any kind of celebration. Compare for example the entries for 25 January, a natalis for the arrival of the annual papyrus shipment; 21 April, a natalis of the city. (See also M. Salzman, On Roman Time (California, 1990) p. 119 n. 12, p. 126 n. 22).

There are two other equally important objections to the idea that Christmas was derived from the festival of Sol Invictus.
  1. Priority. The festival is first attested in the Chronology of 354, exactly the same document that gives us the first attestation of Christmas on 25 December. There's no basis for ascribing priority to one or the other. The Invictus celebration may be linked to an important temple of the Sun god founded by emperor Aurelian in 294 CE, but that's no more solid than the arguments I offered above for earlier allusions to Christmas.
  2. Lots of sun cults. Sol Invictus was just one of many aspects of the sun that received cult in Rome. In the 354 Chronology alone we find three other festivals of the sun: 6 June, the crowning of the Colossus; 28 August, the [natalis] Solis et Lunae; 19-22 October, the ludi Solis. Out of the four, the June festival is the one most likely to be linked to the imperial cult; either the August or October festivals stand just as much a chance as the December one of being linked to Aurelian's temple. S. E. Hijmans, Sol: the Sun in the Art and Religions of Rome (diss. Groningen, 2009) pp. 483-7 lists off three other temples to Sol within Rome; at p. 485 n. 17 he opts for 19 October as the festival of Aurelian's sun-cult.
And, in case it needs to be pointed out, the 25 December festival has nothing to do with Mithras. Mithras was depicted with sun iconography too, and was regularly called Sol Invictus Mithras or Deus Sol Mithras, but Mithraism was a mystery religion practised underground all over the empire; the 25 December festival was an official civic festival observed at a specific temple in Rome by the city of Rome.

Brumalia. Contrary to popular belief, solstice festivals were not a dime a dozen in antiquity: we have evidence of midsummer festivals from the Greek world, since that was the new year in Greek calendars; but not for the Romans. Brumalia is a rarity, as a clear case of a genuine winter solstice festival. In fact we're not even told explicitly that it was a solstice festival: we have to infer that, from the fact that bruma is the Latin for 'winter solstice', derived from an archaic word meaning 'shortest (day)'.

Christmas isn't based on Brumalia either. Brumalia isn't of any great antiquity: it's first attested in Tertullian (On idolatry 10.3), in the late 2nd century CE. It's fairly obscure, and we don't know very much about it: we do know it involved athletic games and gift-giving (but as we saw above, gift-giving wasn't introduced to Christmas until the modern era). Like Saturnalia, it continued to be celebrated long after Christmas was in place, and by Christians: in the 6th century, emperor Justinian organised a civic festival of Brumalia from which a celebratory oration by one Choricius of Gaza survives (Oration 13 ed. Foerster and Richtsteig; not in Boissonade's 1849 edition); by that time it had absorbed some elements of Saturnalia, both of them pagan festivals.

The solstice as a purely astronomical event. So it seems likely that Christmas is related to the solstice. But there's no particular reason to see it as linked to any pagan festival: we've firmly discounted Saturnalia, Mithras, and Brumalia, and there's only a potential case for Sol Invictus.

The most likely situation is that both Christmas and the Invictus celebration were assigned to 25 December because that was the traditional date of the solstice. That is, one wasn't borrowed from the other in any sense: they were sibling festivals, cognate rather than derivative. Note that we have evidence for the traditional date of the solstice 300 years before our earliest explicit evidence for the date of either festival. For Christians the motivation would be that early Christians already thought of Jesus as the 'sun of righteousness' (Malachi 4:2), as we saw above.

However, note well the discussion above of the relationship between Jesus' death-date and birthday: the death-date was primary, and the birthday calculated from that. That is, the important thing about 25 December wasn't that it was the traditional date of the solstice: it was exactly nine months after the traditional date of the vernal equinox, on 25 March. If that is indeed the basis for the date of Christmas, it's not because Jesus was understood as the sun returning light after a season of darkness: it's because his death and resurrection were believed to be represented by the centre-point of the sun's waxing, and therefore central in a cosmic sense too -- in the middle of life's journey, so to speak.


Further reading

Steven Hijmans, 'Sol Invictus, the winter solstice, and the origins of Christmas', Mouseion 3 (2003): 377-98.
Steven Hijmans, Sol: the Sun in the Art and Religions of Rome (diss. Groningen, 2009), esp. chs. 5 and 9.
Alden A. Mosshammer, The Easter Computus and the Origins of the Christian Era (Oxford, 2008).
Susan K. Roll, Towards the Origins of Christmas (Kok Pharos, 1995).

Endnote: pseudo-Hippolytus on the date of Christmas and Easter

Hippolytus, Commentary on Daniel 4.23.3 (M. Lefèvre, Hippolyte: Commentaire sur Daniel, Sources chrétiennes vol. 14, Paris: Éditions du Cerf, 1947):
ἡ γὰρ πρώτη παρουσία τοῦ κυρίου ἡμῶν ἡ ἔνσαρκος, ἐν ᾗ γεγέννηται ἐν Βηθλεέμ, ἐγένετο πρὸ ὀκτὼ καλανδῶν ἰανουαρίων, ἡμέρᾳ τετράδι, βασιλεύοντος Αὐγούστου τεσσαρακοστὸν καὶ δεύτερον ἔτος, ἀπὸ δὲ Ἀδὰμ πεντακισχιλιοστῷ καὶ πεντακοσιοστῷ ἔτει· ἔπαθεν δὲ τριακοστῷ τρίτῳ ἔτει πρὸ ὀκτὼ καλανδῶν ἀπριλίων, ἡμέρᾳ παρασκευῇ, ὀκτωκαιδεκάτῳ ἔτει Τιβερίου Καίσαρος, ὑπατεύοντος Ῥούφου καὶ Ῥουβελλίωνος.
For the first advent of the lord among us in the flesh, when he was born in Bethlehem, took place on the 8th day before the kalends of January (i.e. 25 December), a Wednesday (lit. 'the fourth day'), in the 42nd year of Augustus' reign, in the 5500th year from Adam; and he suffered in his 33rd year, on the 8th day before the kalends of April (i.e. 25 March), a Friday (lit. 'day of preparation'), in the 18th year of Tiberius Caesar's reign, in the consulship of Rufus and Rubellio.
Even if this were authentic, it would be either terribly corrupt or terribly incompetent. The year numbers make no sense. Augustus ruled only 40 years, so the birth date makes no sense unless the author thought Augustus' reign began in 44 BCE, immediately after Julius Caesar's death: in that case he would mean 1 BCE; that year 25 December fell on Saturday by modern reckoning, or within a day or two of that by Roman reckoning (Roman reckoning was not fully in synch with modern reckoning until 8 CE): certainly not a Wednesday. On the death date: the 18th year of Tiberius' reign began in September 31 CE, but the consulship of 'Rufus and Rubellio' (errors for C. Fufius Geminus and L. Rubellius Geminus) was in 29 CE. March 29 CE was in the 15th year of Tiberius' reign. Yet the consuls' names imply that the writer means the year 29 CE, and 25 March did fall on a Friday that year. A suggestion appears dotted around a few early sources that the sky darkening at Jesus' death (Mark 15:33) may be linked to a solar eclipse over Jersualem in 29 CE -- but the eclipse was in November of that year, not March/April. [Note, 19/1/16: just as a by-the-way, some modern enthusiasts instead link the report of darkness at Jesus' death to a solar eclipse in March 33 ... but that eclipse was over the southern Indian Ocean, and was certainly not visible in Jerusalem!]

Incidentally, we do not have adequate evidence to tell in what year 14 Nisan (Hebrew) might have been considered to correspond to 25 March (Julian). Modern reckoning is no use, as the Hebrew calendar at the time was observational; modern reckoning developed over the course of the 1st millennium CE.

[Note, 19/1/16: I've altered the wording and flow in a few passages to make my writing slightly less contorted. I've left annotations in the places where I've changed the content.]

Monday, 14 December 2015

The library of Alexandria and the loss of knowledge

Myth: the burning of the library of Alexandria was "the most destructive fire in the history of human culture".

Alexandria was the chief city of Hellenistic and Roman Egypt, and the most important cultural powerhouse of the ancient Mediterranean. The quotation above comes from this History Channel clip about its famous library, or rather libraries.

The narrator goes on (at the 1 min. 39 sec. mark):
In the battle that followed, Caesar ordered his soldiers to burn the Egyptian fleets lying in the harbour. The fire quickly spread from the waterfront to the great library. The flames consumed a large part of the library's collection, marking the single greatest loss of knowledge in history.
Some historians speculate that the fire set civilisation back by a thousand years. Who knows, if the great library of Alexandria hadn't burned, Columbus may not have sailed to the New World. He might have gone to the moon!
Recently a new library was built in Alexandria, but it can never replace the ancient collection burnt in the fire. It contained rare manuscripts, the comedies of Aristotle, and more than 200 plays by Aeschylus and Euripides -- classic works forever lost.
Doctor sitting reading by an armarium holding books
(early 4th cent. CE; Met. Mus. Art, New York)
This snippet ranges from absurd to outright false. (Let's do the easy bits right away: Aristotle didn't write comedies, and Aeschylus and Euripides wrote a combined total of about 170 plays.) The only bit that has any basis in reality is the first line, about Caesar burning the Ptolemaic fleet. Everything else is untrue, without any room for doubt on the point.

It's not like the History Channel is conveying an isolated opinion, by the way. It is really widely believed. Here's a full-length documentary that makes similar claims; the Wikipedia article on the subject refers to "the incalculable loss of ancient works"; Joel Levy's 2006 book Lost Histories calls it "the day that history lost its memory"; online forums frequently get questions about just how big a disaster it was.

Important point: I'm not talking today about the historical circumstances of the library's destruction. There certainly was a major fire in 47 BCE, and there may have been other important moments of destruction in later centuries. We're not here to pin down when it disappeared, or who's to blame: this is about the historical significance of the library's loss.

Several kinds of misconception feed into this myth.
  1. Misconceptions about the role of libraries in the ancient world.
  2. Misconceptions about what kinds of books the Alexandrian library actually held.
  3. Misconceptions about the actual causes for the loss of texts from antiquity.

1. The role of libraries

If the loss of the library was "the single greatest loss of knowledge" in history, that would mean the books destroyed were the only existing copies of those books.

Suppose -- heaven forfend -- that the British Library burned down tomorrow, or the Library of Congress. What kind of a loss would it be? In cultural terms, and purely in monetary terms, it would be catastrophic: millions of manuscripts, autographs, and rare and unique items would be lost, and the cost of replacing the printed collection would be vast.

But barely a scrap of actual knowledge would be lost. Ismail Kadare's novels would survive. The Thirty Years War would not be forgotten. Aeroplanes and computers would not become treasured relics, never to be recreated.

This is because there are lots and lots and lots of repositories of information in the world. And exactly the same was true in Greco-Roman antiquity. There were hundreds of libraries of Greek and Latin texts dotted around the Mediterranean. Alexandria was the biggest, but it was just one fish in a sea of libraries. There were also important centres at Pergamon, Athens, Rome, Constantinople, and many important private collections. Roman aristocrats founded many libraries in the early Principate; clubs and gymnasia in Greece were also centres of learning, with their own libraries, and we have inscriptions cataloguing regular deposits of books in their collections. Caesar's fire did not stop Athenaeus and Julius Africanus from being profoundly well-read more than two centuries later, and the likes of Pliny the Elder and Pausanias did their research privately or in Athens, not in Alexandria.

A fanciful depiction of the library in the Serapeion at Alexandria (Agora, 2009).
(Who's in charge of this mess? The scrolls don't even have labels!)
The book trade thrived and had mass audiences. The literacy rate was higher than many modern people would naively expect: nowhere near modern First World levels, to be sure, but there was a big market for things like popular romances, basic reference books, and how-to manuals. Literacy was certainly not limited to a small elite class: almost anyone could scrawl graffiti on a wall without much education. Cicero refers to the publishing business on a scale that, for the time, we may as well consider industrial (Q.fr. 3.6.6; Att. 12.6a). Books travelled from city to city easily: Pliny the Younger is delighted to hear that his own books were on sale at shops in Lyon (Letters 9.11.2). Book prices in 1st century CE Rome ranged from 6 sestertii for a cheap knockoff (Martial 1.66; one or two days' labourer's wage) to 5 denarii for a deluxe edition (Martial 1.117; = 30 sestertii). The amounts don't translate well into modern terms, but they're comparable to the prices of university textbooks: not chicken feed, but certainly not just for the elite either. To save costs further, publishers could recycle used papyrus (Catullus 22.5), or customers could commission copies made on the back of something else.

This last point is directly tied to one important function of ancient libraries. As well as being reading rooms, they were also scribal centres that bypassed the book trade. People could commission a scribe to go and make a copy of a book, and it seems this was a pretty economical thing to do. (Remember copyright is irrelevant in a society where reproduction is labour-intensive.) A beautiful example is the sole surviving copy of Aristotle's Constitution of the Athenians. An estate owner living near Hermopolis, Egypt, recycled four scrolls of his farm and business records by commissioning scribes to make a copy of some fairly high-powered intellectual works on the back, including Aristotle's book. (It's not very likely that the copying was done at Alexandria, about 200 km away.) The economics of the situation are telling: the owner was willing to hire professional scribes, but not to pay for clean papyrus. In other words, scribes were cheap.

It is unlikely that more than a handful of texts of any consequence were lost in the fire of 47 BCE, for the simple reason that anything important certainly existed in many copies, in libraries and private collections, all over the Mediterranean.

2. The books in the library

In the popular imagination, the library held all manner of arcane knowledge lost in the mists of time -- Babylonian mathematical treatises, dictionaries of Linear A, diplomatic correspondence between Egypt and Atlantis, the history of Göbekli Tepe, that kind of thing.

Illustrated edition of a poem about Herakles, probably for a popular audience:
Herakles' fight with the Nemean lion (P.Oxy. 2331, 3rd century)
In reality it was not a repository of records left by the Pharaohs of the Old Kingdom. It was a Greek library, of Greek texts, for Greek people, founded around 300 BCE. One late source tells us that there were accessions of Egyptian, Chaldaean, and Roman books, but they were invariably translated into Greek (Syncellus, Chronographia 516,6-10). We don't know if the originals would have been preserved too; it doesn't seem likely that they were prized.

We have a very good idea of the kinds of things that were in the library. This is because surviving books routinely cite and discuss other books, including ones that have been lost. Many important pieces of modern research revolve around gathering together the fragments that we obtain this way: the most important such collection, the Fragmente der griechischen Historiker, lists over 1000 lost authors -- and that's just in the genres of history and geography. In some cases we know a huge amount about these books; in other cases we know only titles. But it's more than enough to tell us that what we are missing is, essentially, pretty similar to what survived via the mediaeval manuscript tradition.

The thing that we're really missing out on is the colossal book-writing spree that Greek thinkers all round the Mediterranean went on in the late 4th to 1st centuries BCE: we have comparatively few intact books from that period -- we have Aristotle, Euclid, and Dionysius of Halicarnassus, but we're missing out on the likes of Eratosthenes, Hipparchus, and Crates.

3. How ancient texts actually got lost

The destruction of a library is a terrible thing, but it's a drop in the ocean. The disappearance of Greco-Roman texts is a story about culture and economics, not a timeline of specific events. Left to themselves, books vanish over time without any need for someone stepping in to destroy them. Poor storage, poor longevity in the materials, environmental factors, and human agency all hasten that natural decay, but that decay will happen anyway. Over a thousand years, that's plenty to ensure the demise of nearly every book in existence.

Sure, it would be nice if the library of Alexandria had survived to the present day. But that means positing a miracle. No ancient library has survived to the present. Even if the Alexandrian library had survived the fires, eventually it would have gone the same way as the Palatine library in Rome -- which suffered its own series of catastrophic fires (the History Channel never talks about those) -- and the libraries of Pergamon, Tralles, Athens, and so on. Aristotle's Constitution of the Athenians, mentioned above, is a truly extraordinary case: only a handful of texts have survived by being preserved on an intact ancient papyrus.

Codex and scroll (Pompeii, before 79 CE):
left, woman with note-taking codex (wax on wood);
right, man with commercial scroll (with titulus)
Books survive if many different people ensure that they're copied. And the people of the past who did that copying weren't operating with any top-down, organised plan; they weren't members of a worldwide Book Preservation Society. They were independent institutions and individuals living in many different places and many different centuries, and their efforts just happen to have the fortunate combined effect that many texts have survived to the present.

Texts were disappearing long before Rome fell. The 2nd century CE is when we really start to notice extant sources treating old texts as things they haven't personally read -- they only have second- or third-hand information. In other words, that's when texts start vanishing en masse. J. O. Ward, cited above, points out that many oratorical speeches from Cicero's time were already obscure in Tacitus' time. We have no evidence of any of the Epic Cycle surviving beyond the 2nd century. (Some of them did survive that long: so however they were lost, it had nothing to do with events in Alexandria.) Not a single ancient writer ever cites book 2 of Aristotle's Poetics, other than Aristotle himself: it was never as popular as the similar material in his On poets (also lost), which was intended for a wider audience, and about which we hear a great deal. Poetics book 2 may well have disappeared within a century of being written.

The 2nd-3rd centuries were also the time of a massive technical migration: from scroll to codex. ('Codex' is the word for a modern-style book, with pages sewn together at the spine.) The very biggest hurdle for the survival of books is nothing to do with libraries burning, or fictional stories about religious zealots destroying pagan books. It's about a format shift.

We first begin to hear about commercial use of codices by ancient booksellers in the 1st century CE poet Martial, who is impressed after seeing a codex edition of Ovid's Metamorphoses with the entire text in one volume (Epigrams 14.192):
haec tibi multiplici quae structa est massa tabella,
   carmina Nasonis quinque decemque gerit. 
Look at this bulk! It's built out of many­-layered leaves,
   and holds fifteen books of Naso's poem.
Lionel Casson's Libraries in the Ancient World (2001), pp. 127-8, reports the following proportions in Egyptian papyrus finds:

Scroll Codex
1st-2nd centuries CE 98.5% 1.5%
ca. 300 CE 50% 50%
ca. 400 CE 20% 80%
ca. 500 CE 10% 90%

An armarium for codices: the real reason
for the loss of Greco-Roman texts
(Codex Amiatinus, early 8th cent. CE)
A format shift doesn't only attach an extra cost to the survival of any text, it also attaches a time­-limit. If the storage units in your library are armaria for codices, scrolls that haven't been transferred by the deadline will simply not get stored in the library. In addition, ancient and mediaeval codices were normally stored flat on their backs -- not on end, as in modern bookshelves -- and they couldn't be piled high, if they could be piled at all. So even though a codex could hold a lot more text than a scroll, codices took up more space for the same amount of text!

Scrolls were effectively a self-destruct timer. A book published in scroll form might survive a century or three after 300 CE; but if it hadn't been copied into a codex by that date, the text was basically doomed.

Wars and fires don't help of course, but those are pretty minor things in comparison to a format shift that affected all books.

So don't lament for the library of Alexandria: celebrate it for what it was. It's an important chapter in the story of the development of knowledge. But in the story of the loss of knowledge, it barely warrants even a footnote.

Some other popular sources do a perfectly decent job with this topic: Carl Sagan's TV series Cosmos is a bit notorious for being unreliable on history, but it's on relatively steady ground here (1980; episode 1, "The shores of the cosmic ocean") --
Each of those million volumes which once existed in this library were handwritten on papyrus manuscript scrolls. What happened to all those books? 
The classical civilisation that created them disintegrated. The library itself was destroyed. Only a small fraction of the works survived. And as for the rest, we're left only with pathetic scattered fragments.
This could be a lot worse. It's not flawless: elsewhere Sagan implies that figures like Aristarchus of Samos and Archimedes had something to do with Alexandria, when there's no evidence they ever even visited the city. But he's absolutely right to emphasise the demise of the civilisations that created libraries, or rather their governments -- the Ptolemies in Alexandria, the Attalids in Pergamon -- and not any single moment of destruction. If Caesar's Alexandrian War caused a loss of knowledge at all, it wasn't because of a fire: it was because he effectively ended the Ptolemaic dynasty, which had been supporting the library's operations for 250 years. If the Ptolemies had still reigned in 300 CE, it's likely that more work would have been put into preservation efforts.