The texts in the romancero viejo were probably intended to be sung, or at least singable, exactly like ballads in English. But what were the melodies like?
I must confess, I have no idea, though I have a few tantalizing clues from secondary and tertiary sources. This post is mostly here to solicit the input of people more knowledgeable about early music than I, and to arm them with what knowledge I already have access to in order to help with this question.
Luis Milán on Sixteenth-Century Performance Practice, by Luis Gásser (1996), contains a very brief discussion of romance melodies, which includes two examples written in a historical notation that I am unable to read (it appears to be a form of mensural notation but on staves of only three lines):
In sixteenth-century Spain various melodies were adopted for the singing of ballads; one melody could serve different poems and one poem could be sung with several melodies, as is nowadays the case with traditional poems-songs-dances like the jota. According to Francisco Salinas (1577), these melodies may well have been the very simple ones he described as "a melody with which all eight-syllable verses can be sung, [which is] often used in the narration of histories and stories," shown in example 1.
Salinas adds that with the melody of Conde Claros "the Spaniards use to sing with it many of the so-called romances." The melody is provided twice, the second time with the statement that this is the meter used for all the Spanish compositions in which old narrations or stories were sung by the ancients. The melody is shown in Example 2.
In addition to the simplicity of the melodic line, this is an early illustration of the combination of simple triple and compound duple time, often used by Milán in his songs and by his contemporary Juán del Encina. Milán could well have been reciting-singing his long poem series with a melody similar to those found in El Maestro or to the ones recorded by Salinas, and according to the descriptions of Galilei and Zarlino, in a practice that has reached our time without substantial changes, except those of context. To sum up, we know through the descriptions of El Cortesano the average duration of a particular program, and that expressive declamatory settings of epic, lyric, or love-poetry were part of this performing context. The examples found in Milán's work can be used in considering the proper performing context and style for similar pieces. (Gasser 1996, 32-33)Now, Luis Milán was a Valencian composer thoroughly and formally trained in the music theory of his era and actively employed in teaching it to others; in other words, he was not a folk musician, though he must necessarily have been familiar with the folk musical practice of his era. His compositions (which are the primary subject of Gásser's book) include some of the first pieces written explicitly for what we would now refer to as classical guitar (though Milán intended them to be played on the vihuela de mano, an instrument which is an intermediate link in the guitar's evolution from the lute). Milán did, however, compose at least one musical setting of a romance text, the romance de Durandarte ("ballad of Durendal," a fragmentary and confusing text clearly based on the Matter of France). Does this melody have anything at all in common with how a folk performer would have presented the same ballad, even in a courtly setting?
Spanish Wikipedia (which is often a useful starting place for research despite its notorious lack of reliability) has notations for two romance melodies linked at the bottom of its article on "Romance (poesía)", but no dates are provided for these. I have no idea whether these melodies are taken from sixteenth-century sources or whether they are collected by recent Spanish folklorists; in either case they provide only minimal information about how these melodies would be fitted to the lyric lines, and if they are recent I do not know how similar they are to golden age performance. (They obviously won't be identical, but they could be vaguely close.)
There are two fairly fundamental problems with all four of these melodies, however, as they are in all cases only sixteen notes long - perfect for a setting of two eight-syllable lines. (This feature is not shared with Milán's Durandarte, which may be a further clue about how it differs from the folk settings of romances in his day.) The first difficulty that presents is one of length and repetition; while a two-line melody repeated many times seems fine for something like the "romance del rey de Aragón" ("Ballad of the King of Aragon"), many romances are hundreds of lines long (in fact I'm working off and on at one, the "romance del conde Dirlos," which clocks in at just over 1600); constant repetition of such a brief melodic phrase seems like a surefire way to put your audience to sleep. The other problem is that of stresses; unlike French (which also often uses counted-syllable metric forms), Spanish most certainly does have stressed syllables and unstressed syallables, and it can feel awkward for a melody to stress a syllable which would not receive the accent in speech in favor of one which would. Since the position and even the number of stresses varies freely from one line of a romance to another, identical meter is probably inappropriate.
My tentative guess - and it's nothing more than that, and I would very much like the input of people who know things about music - is that the romances would have been presented in a recitative, chant-like manner, based on a repeated melody line but allowing the performer all the same tricks of vocal presentation available in spoken recitation of poetry to cause the notes and rhythms to vary; such an approach would enable the performer to put the stress where it belongs, as well as keeping an interested audience engaged by letting the natural variety of the poem show through. (Whether audiences could be held by a narrative poem of that length at all is a different question, but I'm going to with yes - a good performer can still draw a modest crowd today with a recitation of Beowulf, which typically takes over three hours.)
Does such a performance style make sense to people who are familiar with early music? If so, what does it sound like? I'd love to hear somebody more vocally competent than myself give it a shot (if you use an extant Spanish melodic phrase, I'd suggest either a Spanish text or my translation of "Ballad of the King of Aragon" on the grounds that it has lines which are reliably the right length), but I'd also love to hear alternate hypotheses about how a romance ought to be performed.
For the time being, in the absence of contrary knowledge, my preference is to treat the romancero viejo the way most literary scholars in the Spanish-speaking world do: as a collection of poems, rather than as songs. But it feels like some part of their authentic presentation is lost, and our experience today would be richer if it were to be restored.