Late Roman Unit Sizes

[Robert Vermaat]

That's an interesting point, yes. Terentius would seem to be the dux Armeniaca, and in the ND this official commands three legions (XV Apollinaris, XII Fulminata and I Pontica). The first two of these are old principiate formations, and formed the main military force of the old senatorial governors of Cappadocia (Arrian being the most famous one - both appear in his 'array against the Alans' - his presumed order of battle (see below) might suggest a good comparison for Terentius's expedition). The third is probably a tetrarchic foundation.

Ammianus gives a number for the army of Procopius in Persia during 363 AD as 30.000 (AM XXIII.3.5), though others set this at a lower number: 20.000 (Lib. Or. 18.214), 18.000 (Zosimus III.12.5) and 16.000 (Malalas Chron. XIII.21). No mention of the number of units involved though.
This is all much smaller than the main army in persia by 363 (65.000 - Zosimus III.13.1), but corresponds slightly to other such armies later:
25.000 - the army of Galerius in Persia, 298 AD (Festus Brev. XXV)
20.000 - the Roman army in Mesopotamia, 531 AD (Proc. Bella I.18.5)

So it seems that Ammianus is using 'legion' here to refer to a much smaller unit, perhaps a cohort or a pair of cohorts - thinking off the top of my head, if the two old legions were c.5000 men and the new one c.2000, that would maybe provide 12 units of c.1000 men each for Terentius's force - but it almost certainly isn't as simple as that!

Ammianus is unfortunately not the best source for terminology, as he uses both old and new words in his text. On the one hand we could se this as similar to his archaic use of 'gladius' (for a spatha) or 'Parthians' (for Sassanid Persians), but on the other hand the term 'legion' could aslo already have become a generalitic description of a Roman infantry unit, which even continues to the British monk Gildas in the early 6th century.