§ 31. To assign the different hemistichs of a poem to these various types we have to follow as a regulating principle the natural word accent and syntactical accent of each sentence. In some cases the similarity or relation with one another of the types renders it a matter of difficulty to determine exactly to what particular type a hemistich may belong. Systematic investigations as to the principles which govern the combinations of the five types in pairs to form the long line have not yet been made. From such observations as have been made it would appear that by preference hemistichs of different rhythmical structure (ascending and descending) were combined with a view to avoid a monotonous likeness between the two halves of the verse.[79].
§ 32. The combination of two hemistichs so as to form a long line is effected by means of alliteration, one at least of the two fully accented syllables being the bearer of an alliterative sound. In no case is an unaccented syllable or even a syllable with a secondary accent allowed to take part in the alliteration. This fact, that secondarily accented syllables are debarred from alliterating, is another proof that it is better to look on them as belonging to the thesis rather than to the arsis of the verse.
The Principles of Alliteration.
§ 33. Quality of the Alliteration. It is an all but invariable rule that the correspondence of sounds must be exact and not merely approximate. A g must alliterate to a g, not to a c, a d to a d, not to a t, and so on. There is, however, one remarkable exception, namely, that no distinction is made between the guttural c (as in cūðe) and the palatal c (as in cēosan), nor between the guttural g (as in god) and the palatal g (as in gierede), not even when the latter represents Germanic j (as in geong, gēar). With exceptions hereafter to be noted, a consonant followed by a vowel may alliterate with itself followed by another consonant: thus cūðe alliterates not only with words like cyning, but with words like cræft, cwellan; and hūs alliterates not only with heofon but with hlēapan, hnǣgan, &c. The fact that different vowels, as ī, ū, and æ in īsig ond ūtfūs æðelinges fær Beow. 33, alliterate together is only an apparent exception to the strictness of the rule, as it is really the glottal catch or spiritus lenis[80] before all vowels which alliterates here. Wherever a vowel seems to alliterate with an h we are justified in assuming a corruption of the text, as in óretmecgas æfter hǽleðum frægn Beow. 332, where Grein improves both sense and metre by substituting æðelum for hæleðum; other examples are Beow. 499, 1542, 2095, 2930. In some cases where foreign names beginning with h occur we occasionally find instances of this inexact alliteration, as Hólofernus únlyfigendes Jud. 180 and 7, 21, 46, contrasted with Hólofernus hógedon āninga 250; in later works as in Ælfric’s Metrical Homilies we find alliteration of h with a vowel not only in foreign names but with native words, as
and he ǣ́fre his fýrde þam hǣ́lende betǣ́hte. Ælfr. Judges[81] 417.
and h before consonants (viz. r, l, w) is disregarded as
and hē hig āhrédde of þām rḗðan þḗowte.Ælfr. Judges 16.
on hwám his stréngð wæs and his wúndorlī̀ce míht.ibid. 306.