Tuesday, 7 March 2023

Homer's metre 2. Hermann’s Bridge

Hermann’s Bridge is a metrical feature in Homeric verse. Often it’s regarded as advanced — for specialists only. Well, hear this. If you’re educating someone about Homeric verse, you cannot neglect something as fundamental as Hermann’s Bridge is.

How fundamental? More fundamental than having a dactyl or spondee at the start of the line, that’s how fundamental.

1. Structure of the hexameter | 2. Hermann’s Bridge | 3. Catalogue and glossary
Gottfried Hermann, 1772–1848 (unknown artist)

Here’s how it’s traditionally defined.

[T]here cannot be a word end between the two short elements of the second half of the fourth foot.
De Decker 2017: 60

Here’s what this means in the twelve-beat rhythm we looked at last time. In the musical notation, Hermann’s Bridge is marked by the dotted line; in the metrical notation, it’s marked by the bracket. The point is that it’s always mid-word, never between words.

— ⏕ — ⏕ — ⏕ — ͜⏕ — ⏑⏑ — ×

A bridge is any position in a poetic rhythm where word break doesn’t usually happen. This particular bridge was observed over 200 years ago by Gottfried Hermann (1805: 692–693). Hermann’s Bridge is by far the strictest of its kind.

98% of Homeric lines have a word break in the third foot. Does that sound like a lot? Well, over 99.94% of lines observe Hermann’s Bridge.

This is no incidental feature. What it is, is an absolutely central feature of the Homeric hexameter. It isn’t any more ‘advanced’ than the mid-line caesura, it’s exactly the same kind of phenomenon. Caesuras happen because prosodic units have typical rhythms; bridges happen because there are some rhythms that prosodic units just don’t use.

The sample space

According to the definitions I’ll adopt below, there are fourteen violations of Hermann’s Bridge in Homer. That is, 0.050% of all 27,803 lines.

Note. The definitions adopted below are, briefly: (a) A bridge is mid-prosodic unit, rather than mid-word. (b) Monosyllabic words, including enclitics, don’t cause a violation.

Is that a fair claim, though? After all, a violation can only ever happen in lines with a dactylic fourth foot. Should we stipulate that our sample space should be confined to those lines?

It isn’t really necessary, but fine, let’s do that. How many lines have a dactylic fourth foot? Oswald (2014: 421) claims that it’s ‘19 out of 20’ lines, but that’s just wildly wrong: a glance at the text of Homer will show that.

Schoisswohl’s and Papakitsos’ tool Dactylo is able to perform automated scansion of over 96% of all Homeric lines. Their figures indicate the following (Schoisswohl and Papakitsos 2020: 170–171):

  Iliad Odyssey
full corpus, including unscanned lines (reported) 15,693 lines 12,110
unscanned lines, % of full corpus (reported) 3.7% 3.5%
scanned lines, % (calculated as proportion of sum of all reported figures) ~96.302% ~96.505%
lines scanned, actual (calculated) ~15,113 ~11,687
dactylic fourth foot, % of full corpus (sum of reported figures) 68.287% 67.548%
dactylic fourth foot, % of lines scanned (calculated) ~70.909% ~69.994%
dactylic fourth foot, actual (calculated) ~10,716 ~8,180

So about 70–71% of Homeric lines have a dactylic fourth foot. These figures give a total of about 18,896 eligible lines. (The real number of lines with a dactylic fourth foot will be slightly higher.)

With fourteen violations of Hermann’s Bridge, and 18,896 lines that could possibly have a violation, the proportion with violations rises from 0.050% to 0.074%. So in this restricted sample space, only 99.93% of dactylic fourth feet observe Hermann’s Bridge. (Only!)

Hermann’s Bridge is still stricter than starting the line with a long syllable. There are a bit over 50 lines in Homer that start with a short syllable. There are only 14 violations of Hermann’s Bridge.

Note. Homer tolerates a short initial syllable for certain words at line-beginning: amphibrachs with ἕως (20×) and τέως (1×); tribrachs with κλύθι/κλύτε (17×), ἴομεν (6×), and φίλε (3×); baccheics with ἐπεί δή/ἐπειδή (6×). This gives a total of 53: so at least 0.19% of Homer has a line-initial short. These are the examples cited by West (1967: 135–139); there may be others. Incidentally, these parallels strongly imply that ἕως should not be understood as preserving the trochaic rhythm of pre-Homeric *ἧος/ἆος.
Wrong bridge: Hermann Bridge, Missouri, demolished 2008. (Source: BridgeHunter.com)

Colon break, not word break

Previously, in part 1, I said caesuras are a side-effect of a deeper phenomenon. The same phenomenon is responsible for bridges. Colon breaks are the gaps between the trees: prosodic units, or cola, are the trees themselves. A bridge is a position that is normally full of tree.

So I’m adopting a refinement here. Word breaks aren’t as important as colon-breaks. A bridge is a position that is normally mid-colon. It’s a constraint on colon shape.

It would be perverse, after all, to imagine an archaic oral poet adopting a rule of the form ‘I will avoid ending a word at such-and-such a position.’ There are plenty of modern scholars who take the stance that a caesura isn’t the same thing as a ‘sense break’ (see Janse 2020: 10–13 for examples). A metrical principle operating on ‘caesura’ in that sense would be arbitrary.

(If you do want to base theories on caesuras defined as ‘word breaks’, say what you mean! Say ‘word break’, not ‘caesura’. Then it’ll be easier to see how absurd the theory is!)

Hermann’s Bridge is anything but arbitrary. When a violation happens, it isn’t because the poet has broken an abstract rule. It’s because there’s something uncommon about the prosodic units in that line.

When Hermann’s Bridge is violated, the real nature of the anomaly is that there’s a line-end colon shaped ⏑ — ⏑⏑ — ×.

More often, violations have been described in terms of the preceding colon (e.g. West 1982: 37–38). It so happens that all Homeric lines that violate Hermann’s Bridge have a mid-line caesura, so we can also say that in each line, the violation occurs because of a mid-line colon shaped ⏓ — ⏑.

Put this way, Hermann’s Bridge turns out to be equivalent to Wernicke’s Law. Wernicke’s Law is that when a word ends at the end of a spondaic fourth foot, that word is naturally spondaic: that is, its last syllable will be naturally long, without having to be lengthened by two consonants. If a word ending in the natural rhythm — ⏑ is put in the fourth foot, then it violates either Hermann’s Bridge (if its last syllable remains short) or Wernicke’s Law (if it is long by position).

So don’t think of this in terms of ‘trochaic caesura in the fourth foot’. That’s the wrong mental model. The true nature of Hermann’s Bridge (and, mutandis mutatis, Wernicke’s Law) is this:

It is extremely rare to find a word shaped ⏓ — ⏑ after the mid-line caesura.
(And ditto for ⏑ — ⏑⏑ — × at line-end.)

What makes this particular position so special? A trochaic phrase-end is apparently just fine in other positions of the line. Could it be that it’s really the line-end colon shape that matters after all, not the mid-line colon?

Maybe that’s part of the answer. But it’s independently known that a given word tends to stick like glue to certain positions in the line. O’Neill (1942) found that 90% of words in Homer appear in only a third of the positions where they could be rhythmical. Forstall and Scheirer (2012) have examined words with the rhythm — — in the Iliad and classified them into seven groups based on which position they ‘prefer’, and have found that each group is far stricter than even O’Neill realised. There is a strong likelihood that similar reasoning underlies the ‘specialness’ of the eighth beat.

Frequency profiles from Forstall and Scheirer (2012) for each of the seven classes of words with rhythm — —. Class 1 comprises words that appear most frequently at the first beat (annotated as position ‘110’), class 2 comprises words that appear most frequently at the second beat, and so on. Class 8 comprises words that are statistically ‘most distant’ from the nuclei of each of the first seven classes; the frequency profile for class 8 is similar to that of all classes combined (bottom right; I have adjusted the scale of the combined data to match that of the other figures). Note that (a) some possible positions have no affiliated words: there are no spondaic words that regularly begin on the 3rd, 5th, 9th, or 10th beats. (b) Some classes are more mobile than others: words that are attached to positions before the fourth foot may frequently also be transposed to line-end. (c) Class 6 bridges the position of the ‘bucolic caesura’, contraindicating Witte’s theory that the hexameter emerged from a combination of a tetrameter + adoneus.

This definition reduces the number of major violations

Reframing the bridge in terms of colon-shape means that the following lines can’t be regarded as major violations. I think they shouldn’t be regarded as violations at all. (Word breaks at Hermann’s Bridge are marked by a colon.)

  • Iliad 9.482 ... πολλοῖσιν : ἐπὶ κτεάτεσσι
  • Theogony 23 ... Ἑλικῶνος : ὕπο ζαθέοιο
  • Iliad 16.627 = Odyssey 17.381 ... καὶ ἐσθλὸς : ἐὼν ἀγορεύεις
  • Odyssey 5.272 ... καὶ ὀψὲ : δύοντα Βοώτην

Not all word breaks are equal. Yes, these lines have word breaks at the bridge, but there’s definitely no colon break or phrase break. The first two are prepositional phrases that can’t be separated; the last two are tightly linked participle phrases, with a participle and its complement spanning Hermann’s Bridge.

Note that this doesn’t reduce the importance of Hermann’s Bridge. On the contrary, it means that it’s even stricter.

With some word breaks, it’s a matter of judgement whether we should regard them as being mid-phrase or not.

  • Iliad 1.168: ἔρχομ’ ἔχων ἐπὶ νῆας, ἐπεί κε : κάμω πολεμίζων
  • Iliad 23.76: νίσομαι ἐξ Ἀΐδαο, ἐπήν με : πυρὸς λελάχητε
  • Odyssey 5.400: ἀλλ’ ὅτε τόσσον ἀπῆν ὅσσον τε : γέγωνε βοήσας

Here, the relatives introducing the second half of the line — ἐπεί, ἐπήν, ὅσσον — are linked to the following clauses, to an extent. But they’re not parts of a larger discrete phrase: their role is conjunctive — like syntactical mortar, rather than part of a brick.

In spite of that, these ones aren’t violations. The word break is caused by a one-syllable word, and it’s traditionally understood that one-syllable words don’t violate bridges.

Some violations are unquestionably more serious than others. Part 3 will give a catalogue and taxonomy of violations, in five categories, and give a glossary of technical terms.


  • de Decker, F. 2017. ‘Ὅμηρον ἐξ Ὁμήρου σαφηνίζειν: an analysis of augment use in Iliad 1.’ Journal of Indo-European studies 45: 58–171. [Universiteit Gent]
  • Forstall, C. W.; Scheirer, W. J. 2012. ‘Revealing hidden patterns in the meter of Homer’s Iliad.’ 2012 Chicago Colloquium on Digital Humanities and Computer Science, University of Chicago, 17–19 November 2012. [abstract | poster]
  • Hermann, G. 1805. Orphica. Leipzig. [Internet Archive]
  • Janse, M. 2020. ‘Phrasing Homer: a cognitive-linguistic approach to Homeric versification.’ Symbolae Osloenses 94: 2–32. [DOI]
  • O’Neill, E. G. 1942. ‘The localization of metrical wordtypes in the Greek hexameter.’ Yale classical studies 8: 105–178.
  • Oswald, S. 2014. ‘Metrical laws.’ In: Giannakis, Georgios K. (ed.) Encyclopedia of ancient Greek language and linguistics. Volume 2. G–O. Leiden. 419–423.
  • Schoisswohl, O.; Papakitsos, E. C. 2020. ‘Automated metric profiling and comparison of ancient Greek epics in hexameter.’ Linguistik online 103.3: 157–177. [DOI]
  • West, M. L. 1967. ‘Epica.’ Glotta 44: 135–148.
  • —— 1982. Greek metre. Oxford.

No comments:

Post a Comment