Monday, 15 February 2016

Reading silently and reading out loud in antiquity

It used to be common wisdom that when the Greco-Romans read, they did so out loud. Libraries, it was imagined, must have been a hubbub of constant chatter, not the solemn silent spaces of today. As Bernard Knox puts it,
Greek literature, at least up to Thucydides, was intended for public delivery or performance, and from the early part of the fourth century B.C. on to the end of antiquity, rhetoric was the foundation and eloquence the aim of the educational process.
Woman reading a scroll (fresco from Pompeii)
Woman reading a scroll (fresco from Pompeii; Naples)
The traditional view was laid down by Eduard Norden in 1898, and Josef Balogh in 1927 (see below for full references). I confess I haven’t read Norden; Balogh is more standard anyway. Balogh didn’t take the extreme view that the ancients were incapable of reading silently, but he did conclude that silent reading was unusual enough to surprise anyone who saw it happening and to require explanation. He supplied a catalogue of supporting testimony, mostly from Roman-era sources.

Exhibit A for the reading-was-always-done-out-loud case has always been this passage in Augustine (ca. 400 CE; Confessions 6.3):
For I wasn’t able to ask (Ambrose) what I wanted, when I wanted... When he was reading, his eyes ran over the pages and his heart searched for the intent, but his voice and tongue were still. Often when we were present -- anyone was allowed to enter, and it wasn’t his custom to have visitors announced -- we watched him read silently, and never in any other way. We’d have to sit there in silence for a long time, then leave. For who would dare to annoy someone so absorbed?
We guessed that he wanted to have a break from the noise of other people’s business for that little time that he had for his own thoughts, and didn’t want to be disturbed. Perhaps he was also wary that if he should say out loud something difficult that he was reading, his listener would be interested and want to understand, and he would have to explain it... Though he may also have had a better motive for reading silently, namely to spare his voice, which wore out quickly. Whatever his reason was, a man like that must have had a good one. In any case I never had the opportunity to question him about what I wanted...
The idea is supposedly that Ambrose’s habit of reading silently was something out of the ordinary, so Augustine -- a very learned man himself -- feels the need to explain it. Another passage, in Plutarch, has been influential because of how it has affected some people’s picture of Julius Caesar (Life of Brutus 5):
...Cato and Caesar were standing together, and had opposing views. Just then Caesar was passed a little note from outside, and he read it in silence; but Cato shouted that Caesar was doing something strange, and receiving communications from the enemy. And the crowd went wild...
Supposedly the ‘something strange’ that Caesar was doing, reading silently, cast suspicion on him in Cato’s eyes.

This was the standard view of things for a long time. There were a few dissenting voices, like Knox, mentioned above. In the 1980s and 1990s, the mediaevalist Paul Saenger argued repeatedly that not only did the Greco-Romans invariably read out loud: reading out loud was a ‘physiological necessity’ (Saenger’s italics). It was impossible to read silently, Saenger argued, because ancient manuscripts had no spaces between words. Supposedly, empirical tests of modern people’s ability to read English text without spaces showed that they could not read as fluently as when reading text with spaces: even if they don’t read out loud, they supposedly have to do a lot of subvocalising -- talking in their head, or under their breath.

The turnaround came in 1997. That’s surprisingly recent. If you find a reputable book claiming that reading out loud was universal, check the date of the book: if it’s before 1997, that’s why. (That was the same year Saenger published a book-length summary of his arguments, so don’t be too hard on him: Saenger had no way of knowing what was coming.)

That year an important article by A. K. Gavrilov (subscription required) abruptly and completely overturned the old orthodoxy. The appendix of the article is where it’s all happening: there Gavrilov gives a tidy, straightforward catalogue of evidence both for and against silent reading. The catalogue doesn’t just undermine the old orthodoxy: it makes it blindingly obvious that silent reading wasn’t just an occasional thing, it was absolutely standard. References to silent reading are about three times as common as references that can be interpreted as supporting the reading-was-always-done-out-loud position, all the way from the 5th century BCE to late antiquity.

Here are a few samples:
  • (5th cent. BCE) Euripides, Iphigeneia at Aulis 34ff. and 107-108 -- Agamemnon reads and edits a letter repeatedly, but a slave standing next to him still doesn’t know its contents.
  • (5th cent. BCE) Euripides, Iphigeneia among the Taurians 760-3 -- ‘I shall tell you aloud everything written in the folds of the writing tablet, my friends, for safety. And if you keep the writing safe, it will continue to speak its contents silently.’
  • (5th cent. BCE) Aristophanes, Knights 115ff. -- Nikias gives Demosthenes a bundle of oracles to read, and Demosthenes is so absorbed in reading them that when Nikias asks him what they say, he just says ‘pour me another cup of wine’. A few lines later he paraphrases what’s in the text without quoting it.
  • (3rd cent. BCE) Herodas, Mime 4.21-4 -- ‘Who was the builder who made this stone, and brought it here and erected it?’ ‘Praxiteles’ sons; don’t you see the inscription, there, on the base?’
I omit the other 21 items. Perhaps the most striking piece of testimony is one Gavrilov misses: it appears in an afterword by Miles Burnyeat. The source is Ptolemy (2nd century CE; De iudicandi facultate 5.1-2, Burnyeat’s translation):
...[F]or judging a thing and discovering its nature, the internal logos [in this context, ‘faculty of reasoning’] of thought is sufficient: uttered logos makes no contribution here -- rather, its activity, like the exercise of our senses, disturbs and distracts one’s investigations. That is why it tends to be in states of peace and quiet that we discover the objects of our inquiry, and why we keep quiet when engaged in the readings themselves if we are concentrating hard on the texts before us. What talk is useful for, by contrast, is passing on the results of our inquiries to other people.
Woman reading a scroll (Attic vase, ca. 435 BCE)
Woman reading a scroll (Attic, ca. 440-430 BCE; Louvre CA 2220) (source:
Another important result from Gavrilov’s catalogue is that it shows an aural metaphor was routinely used for reading and writing, but without any particular implication of actual noise. For example, in Euripides’ Hippolytus, 856ff., Theseus opens a letter to ‘see’ what it 'speaks' to him; a few lines later, having read it silently, he cries out at its horrible contents, stating -- using the aural metaphor -- ‘it shouts, it shouts’. At 882 he cries, ‘I will no longer keep this unspeakable story within my mouth’s gate’. In Herodotus 1.123-125 Cyrus reads a letter: he is alone, so the references to what the letter ‘said’ and what Cyrus ‘heard’ are the aural metaphor again; and Herodotus 8.22 tells us what an inscription ‘said’. We have the same metaphor in English, of course: texts routinely ‘tell’ us things, and people ‘say’ things in writing.

This is important because most of the catalogue of evidence for reading out loud is premised on expressions of this kind that use the aural metaphor. Not only is there overwhelming evidence for reading silently: the aural metaphor disposes of much of the contrary evidence too!

The discussion since then has basically been about thinking about exactly how far the implications of Gavrilov’s findings extend. (Though here’s an amusing and embarrassing exception, a 2007 conference paper that ignores Gavrilov altogether, pretends there are only two pieces of testimony for silent reading, and assumes that Saenger ‘definitively disproves’ all contrary possibilities.) This 2000 article (subscription required) thinks about reading out loud as a social act: whether or not people could read silently, Greco-Roman culture was fundamentally an oral culture, and the social dimension was an intrinsic aspect of reading. This 2012 article (subscription required) accepts Saenger’s objection that the lack of spaces between words posed a problem for rapid reading, but suggests that the ancients managed by recognising common letter combinations at word breaks instead, based on a statistical analysis of letter-pairs in ancient Greek. And this 2015 article (subscription required) tries to write a history of the debate, and looks at what the debate tells us about the people who are arguing.

The lack of divisions between words in Greco-Roman texts does pose, if not actual problems, then at least questions. (I leave aside Latin inscriptions of the 1st centuries BCE and CE, which regularly use interpuncts, or mid-line dots, between words; ca. 100 CE it seems the Romans decided interpuncts didn’t add anything of value, and they stopped using them.) The 2012 article, above, is a worthy attempt to deal with the question of how exactly ancient readers managed to read efficiently.

It’s not clear to me that an explanation is actually needed. Saenger’s claim is that spaces are necessary for fluent reading without subvocalisation. But the only piece of experimental evidence he cites for this is a 1962 study which tested subjects’ reading ability in their native language and script, which they had been accustomed to reading with word divisions for between 10 and 25 years, and didn’t even test the effect of word divisions as Saenger claims it did. Ancient readers had a lot more experience in reading continuous text than modern English speakers do. How to design a better test? Where can we find a bunch of people who have decades of practice in reading a phonetic script without word breaks? (Anyone want to bring up a set of 30 or so children who only ever get to see continuous text and are never allowed to see a word break?) Unfortunately, so far as I can find out, all modern phonetic alphabets and syllabaries have spaces or dividers of some kind between words at least some of the time. (Saenger claims the Vai syllabary of Liberia does not; that may possibly have been the case once, but it isn’t now.) So it’s very hard to imagine an experiment that would actually offer a good parallel to the ancient situation.

Much of the testimony that supposedly supports the rarity of silent reading needs to be reinterpreted in light of the aural metaphor. Gavrilov’s interpretation of the Augustine passage is that Augustine resents Ambrose’s silence and absorption, and that he is not trying to find an explanation for Ambrose’s odd behaviour but trying to justify his friend to himself. In the passage about Cato and Caesar, the ‘something strange’ (δεινά) that Caesar was doing is a tendentious interpretation: a more natural translation of δεινά would be ‘something terrible’, meaning the fact that Caesar was supposedly corresponding with Rome’s enemies.

Perhaps it’s useful to compare another Plutarchan passage about Caesar (Life of Caesar 11.3):
In like manner we are told again that, in Spain, when he was at leisure and was reading from the history of Alexander, he was lost in thought for a long time, and then burst into tears. His friends were astonished, and asked the reason for his tears...
Here too, Caesar reads silently and people are startled. But in this case it’s perhaps more obvious that it’s not the reading silently that his friends are astonished at, but at his unexpected tears.

The opposing view would have to be that Caesar must have been reading out loud and then fell into thought, and that Plutarch simply happens to omit the fact that Caesar read the book out loud. Give me a break!

Woman reading a scroll
Reading as a social activity: woman reading a scroll
(British Museum E190)



  1. Hi, How could I contact you? I would like to ask permission to use our picture. Thanks!

    1. Hi you can use my real name at to get in touch. But I should warn you, I didn't create the images used in this post! I try to be more careful with permissions these days, but at the time I wrote this I ... wasn't.