I

As has been often observed, certain metaphors widely used in humanities and social sciences contribute to the amenity of scholarly discourse and, in a way, also serve as useful shortcuts in communications among those happy few that have been properly initiated into a given specialty. On the negative side of the ledger stands the stark fact that such metaphors, and similes as well, have, as a rule, been hastily and inadequately defined and thus easily lend themselves to ambiguity or obfuscation. Abundant examples of such usage have been, over the years, contributed by almost every discipline; certain instances or trends of figurative formulae can be illustrated with several disciplines; and there has been no dearth of loan translations, especially from one Western language serving as a research tool into another. One readily thinks of such cases as current vs. stagnancy, congealment as against thawing, blossoming beside flowering, withering in rivalry with waning and decay, awakening alongside dormancy, zenith in contradiction to nadir, and many more. As a matter of fact, a special inquiry into this dangerously attractive vocabulary and phraseology is overdue.¹

In glottodiachrony, there has been going on, for several centuries now, a good deal of thinking and discussion about the most effective genetic grouping of languages in families (sometimes referred to as stocks).² Given the age-old tradition of genealogical trees, or family trees (Ger. Stammbäume), institutionalized before any curiosity about languages had jelled, today’s student of historical linguistics actually has to cope with two similes superimposed on each other: first, supposedly cognate languages and/or dialects arrayed in a system of relationships borrowed from the familiar pattern of human (or, by way of exception, animal) kinship (“pedigree”): hence “ancestral (or parent) language,” “daughter languages,” and the like; and second, this already complicated (and, as scholars now realize, controversial) arrangement projected onto the model of a plant, with roots; a trunk or a stem; limbs, branches, and sub- branches; and the like. I might add that pioneers of the generation of Bopp and Pott went much farther in their whimsical or deliberately playful handling of such labels than even the boldest advocate of a pictorially saturated style would care to do at present, unhesitatingly calling certain languages “fickle paramours” or “concubines,” and so forth.

It is common knowledge that slightly over a century ago—to be specific, in 1872—Johannes Schmidt, a Berlin Indo-Europeanist initially influenced by August Schleicher but later able to stand on his own feet, proposed a counter model, usually referred to as the wave theory, because in his favorite imagery he operated with innovations originating at one given spot of a major territory, then spreading like the waves on a lake or pond produced by the sharp impact of some object on the surface of the water.³ The wave theory, essentially identical with the diffusionist projection made prominent by turn-of-the-century dialect geographers in Europe and by the Boas school of cultural anthropology in this country, is nowadays almost obligatorily pitted against the older family-tree theory in standard textbooks of linguistics, in introductions into historical linguistics, and in the rapidly accumulating accounts of the history of linguistics.⁴ This very dichotomy has thus gradually become part and parcel of Establishmentarian doctrine (and, incidentally, a well-nigh predictable examination question). The most authoritative opinion, over the last half-century, has been that the two models, family-tree and undulatory projection, were about equally respectable, as regards their ranking. This conciliatory view could easily be misinterpreted as a veiled admission of a dangerous statement, if it were indeed true that these two leading schools of thought were mutually incompatible.⁵ Not unexpectedly, the last few decades have witnessed sporadic attempts to demonstrate that the alleged polar opposition between the two was something of an exaggeration if not a downright myth; that, with a measure of flexibility, both approaches could be brought to bear on linguistic material, as circumstances counseled: e.g. either on divergently shaped slices of such material, or even on essentially the same pile of material viewed at discrepant angles or at different stages of growth.⁶ Aside from such soothing, mediatory voices one could hear others, which argued, e.g., that Schmidt, for all his originality, hardly came up with a “creatio ex nihilo” in 1872. To cite just one possible predecessor, who, as if to make any claims of his more valid, at one time was likewise exposed to the teaching of Schleicher in Jena, Hugo Schuchardt, through use of effectively drawn, contrasting diagrams in Vol. I (1866) of his revised and expanded Bonn doctoral dissertation, Der Vokalismus des Vulgärlateins, anticipated the appeal to wave-like diffusion as a defensible alternative to the model of “ramifications” symbolic of organic growth.⁷

But while these and similarly slanted debates have been going on, in various quarters, and while, in narrower groups of experts, several schemes have been successively proposed for the spatio-temporal classification of Romance and Germanic languages,⁸ certain relevant issues either have not been raised at all or have not been tackled with sufficient strength and, shall we say, articulateness, which of course need not have bordered on stridency. Because, in analytical discourse, one constantly runs the risk of becoming a prisoner of one’s own metaphors and similes, spokesmen for the family-tree allegory might have obligingly enlightened their readers on the sort of configuration they had in mind for their branches and roots, to say nothing of fruits and foliage. In the majority of graphic representations that one is apt to come across the branches evoked appear to have grown away not only from the respective trunks, but also from one another, an image of dispersal that strikingly dramatizes the forces of disruption and divergence rather than those—equally important on balance—of convergence and coalescence in the evolution of languages. Moreover, one is left wondering whether such—not so marginal—processes as the intertwining of branches at their far ends (or at some intermediate points), or even their occasional concrescence and eventual merger could not have been graphically suggested, with a spark of imagerial or pictorial imagination.

Aside from this insufficient quota of attention reserved for the limbs of a tree, as well as for its branches, twigs, and offshoots, there remains the even more vexing issue whether ramification and undulation (or radiation) really exhaust the range of possibilities in schematizing the proliferation of languages. The noted Leipzig and Chicago scholar Walther von Wartburg fell prey to the, I dare say, classic vision of a Romance family sharply split into a Western and an Eastern branch, a pattern clearly influenced by geographic considerations of the kind which once presided over the thinking of tone-setting Indo-Europeanists as well. However, for all his learning, Wartburg did not succeed in squeezing into this Procrustean bed Sardinia, that middle-sized Mediterranean island whose most characteristic dialect communities, he freely granted, sided now with the West, now with the East.⁹ Wartburg, a hard-working, competent practitioner but a fairly weak, easily vulnerable theorist, apparently was unaware of the serious implications of his virtual admission of defeat. An apple, for instance, cannot be cavalierly supposed to “jump” from one branch, or limb, of an apple tree to another; to justify such leaps (i.e. shifts of allegiance) one must first vindicate a radically different model, providing for the possibility of such erratic moves. Perhaps an atoll can drift with relative ease from one island or archipelago to another, but I am unprepared to press this issue, on account of my limited familiarity with coral reefs.

Among European Romanists of the next generation (i.e. those who were at their best and most active toward the middle of this century) Heinrich Lausberg certainly deserves a hearing before our forum. Being most adept, at that stage of his zigzagging development, at manipulating, as an analyst, phonological data through early exposure to Prague School concepts, Lausberg managed to eliminate the heavy geographical mortgage inherited from Wartburg by distinguishing, fundamentally, between an archaic and an innovatory category of daughter language. Geographically, most of the former, in the case of Romance, stretched to the East, while the majority of the latter extended to the West; but this tendential areal split was henceforth reduced to a mere concomitant, so that the discovery of a sub-Pyrenean zone in which the intervocalic surds remained unvoiced, counter to these consonants’ “typically Western” record, which was one of voicing and, here and there, subsequent spirantization (or even total loss, as frequently in French), no longer needed to cause embarrassment.¹⁰ Scholars were thus relieved of any need to justify an exception from a postulated norm, facing as they did a mere peculiarity or idiosyncrasy worthy, at most, of a brief mention or crisp comment. This constituted, unquestionably, a major step forward.

Even Lausberg’s novel sophistication, however, left certain vital things unexplained. In addition to being distinctly stronger in the phonological domain than in morpho-syntax or in the lexico-etymological realm,¹¹ the former Münster scholar also acquired a far more finely nuanced view of the Apennine than of, say, the Iberian Peninsula. The interplay of these two proclivities, plus the fact that, as a relative beginner, Lausberg, to a more exclusive degree than his mentor Gerhard Rohlfs,¹² concentrated, with excellent results, on fieldwork in Southern Italy, tended to make his influence discernibly stronger among Italianists, particularly those concerned with the vicissitudes of sounds and sound systems, than among Hispanists and Occitanists.

It is, for this very reason of imbalance, from the slightly neglected Spanish side that I intend to reopen the discussion, placing at its center certain facts for the most part known but, traditionally, swept under the rug; and groping, almost simultaneously, for some theoretical model not yet sufficiently experimented with.

II

Among the prime classifiers of Romance languages, it is customary to include, on the side of vowels, the given tongue’s acceptance of, as against its aloofness from, the tendency to merge Lat. ē and ĭ, as well as Lat. ō and ŭ; and, on the side of the consonants, its attitude toward voicing of intervocalic surds; i.e. toward the challenge of shifting p to b, t to d, and, as a rule, except before front vowels, c, q /k/ to g. A superadded salient peculiarity of Romance phonology is the by no means obvious fact that the s, a rarely found voiceless sibilant between vowels in Classical Latin, as in casa ‘cabin’, rosa ‘rose’, tended to join /p/, /t/, /k/ wherever these became subject to voicing. One may argue, to be sure, that certain other features of local sound developments could yield equally valid dialect classifiers; the range of metaphony, the presence (and, if so, specific configuration) of diphthongs, and the comportment of lengthened “geminate” consonants, not only occlusives, would qualify as examples in point. For the sake of argument, however, one may, by way of initial experiment, confine oneself to the two above-mentioned features, with heightened but not exclusive attention to the state of affairs in the Iberian Peninsula.

As was first established in the opening years of this century through fieldwork professionally conducted by Paris-trained J. Saroïhandy, and as has been since confirmed by a string of investigations carried out by highly competent, reliable observers,¹³ ancestral /p/, /t/, /k/ have remained unvoiced in Upper Aragon, a mountainous, once fairly isolated region, whose speech, typically for such terrain, abounds in conspicuous peculiarities, many of them, but by no means all, plain archaisms. One finds a parallel situation to the north of the Pyrenees, in Béam.¹⁴ while the raw, unequivocal facts have thus been satisfactorily established, one can lean at first blush toward interpreting them in one of several defensible ways. (Some of these analyses may yet turn out to be mutually compatible or complementary.)

There will be those inclined to blame the Roman set-lers’ contacts with speakers of related or unrelated languages (varieties of Ibero-Basque?) who preceded them as inhabitants of the Peninsula for such regional scotching effect on a broad tendency which elsewhere in the Peninsula led to a powerful breakthrough. A variation on this theme is the hypothesis that within the ranks of the veterans of the legions assigned estates in the Peninsula many, perhaps the majority to begin with, had come with the command of a regionally differentiated (e.g. Oscanized) Latin.

Analysts temperamentally inclined to brush off as unrealistic any such assumption of sub-, ad-, or super-stratal influence will, foreseeably, gravitate toward an internal (i.e. purely Romance), perhaps structurally flavored solution. In so doing, however, they will face another dilemma. Either they will incline toward classing the widespread voicing as an innovation which evolved spontaneously over the entire Peninsula except for certain pockets of resistance (in that event the exact focal point remains to be determined) in undeniable harmony but only in loose actual contact with parallel developments in adjoining territories, e.g. in Southern France (again, except for Béarn) . The implication is that those “pockets,” i.e. stretches of relatively inaccessible (e.g. mountainous) territory, universally known as being protective of residual cultural features, have, in this particular case, once more (obligingly for today’s historical linguist) preserved the original, state of affairs. In case one adopts this belief, then such characteristic Upper Aragonese forms as mica ‘bit, crumb’, from Lat. mica, and dital ‘thimble’, from Lat. digitāle, rather than their familiar Castilian and Portuguese counterparts, namely miga and dedal, may well—indeed, must—at some remote point, minor details apart, have been dominant all over the Peninsula. If this was so, then the widely upheld classification of Hispano-Romance as, having been from the start, part of, or as a gradually emerging “branch” of, Western Romance would have to be immediately abandoned as grossly misleading.

Alternatively, scholars are free to operate with the hypothesis of a Second Wave of Latinity that struck the Peninsula from without, the tacit implication being that the First Wave coincided with the original occupation of the territory by the Romans, a slow, long-drawn-out process stretching from the Second Punic War until, approximately, the reign of Emperor Augustus. The First Wave was rather sharply delimited by ancient historians, some of them near-contemporaries of the events chronicled; the Second Wave involved a sort of gradual, nearly imperceptible percolation, devoid of memorable, attention-catching highlights. The focal point of the First Wave, which basically corresponded to the later Republican period in Roman history, was the city of Rome itself, the urbs, with some allowance to be made for the presumed “Oscan accent,” we recall, of the legionaries recruited in Southern Italy. With the decline of Rome as the standard-setting social and cultural center under the emperors—a downward curve or spiral that picked up momentum from the third to the fifth century—, it was Lugdunum, i.e. present-day Lyon, located at the strategic confluence of the Rhône and the Saône rivers, that emerged as the prime, if not the sole, candidate for leadership in matters of norm, taste, and decorum throughout Late Antiquity, as it was lived out by the Romanized portion of the Empire, not least on account of the city’s enviably celebrated schools. As a matter of fact, the entire area between Marseille (the ancient Massilia) and Lyon shone forth as one of the few remaining paragons of sophisticated Graeco-Roman civilization, with a network of military, political, demographic, economic, and socio-cultural routes spreading from Southern France, including Toulouse, right into Spain; not exactly across the Pyrenees (migrants, invaders, would-be settlers, and, at a later date, crusaders knew how to by-pass the steep, inhospitable mountain range) but to the west or the east of them, as was dramatized in the early fifth century by the itinerary of the Visigoths who, under pressure from the Franks, poured from Southwestern Gaul into Hispania. But long before the spectacular peak of the Völkerwanderung, a kind of strong, persistent cultural infiltration of Spain by Lugdunum and the Narbonensis, with all sorts of novelties and innovations raying out, in the last analysis, from Lyon, may be plausibly supposed to have been in lively operation. Consequently, when we invoke the influx of early Gallicisms into preliterary or medieval Spanish and Portuguese, we must aim to distinguish between three successive processes: the cultural pressure of Southern Gaul on Hispania here referred to, culminating in the Visigothic episode; the Carolingian expansion, responsible for the crystallization of Catalan; and the distinctly later contacts associated with the Cluny Reform, the pilgrimages to Santiago, and the Crusades. It is this last-mentioned phase which is usually taken into account by students of a thin layer of loanwords, under the conventional rubric of Gallicisms and Provençalisms.¹⁵ However, anyone concerned with phonology and grammar must take all three stages into consideration.

The Latinity filtering into those provinces of the crumbling Empire that came under the cultural sway of Southern Gaul, along the Toulouse-Lyon axis, starting with the decline of the pagan era, must have been in many respects significantly different from the older city dialect of Rome. Among other distinctive marks, it was doubtless completely free from any residue of an Oscan streak, for geographic and historical reasons alike (Osco-Umbrian having itself by then become extinct). On circumstantial evidence, it is, to say the least, plausible to hypothesize that the “Latinitas Lugdunensis” favored the voicing of /p/, /t/, /k/, and /s/ between vowels (or, on a somewhat smaller scale, between vowel and /r/, as in patre, mātre, frātre)—witness the unequivocal testimony of Gallo-Romance to this effect—and that the innovative fashion, after spilling over into Spain at chosen points of entry, overran certain regions, without, however, completely dislodging the older, deeply entrenched mode of pronunciation, where it could effectively ward off the invasion—primarily to the south of the Pyrenees, in the Navarro-Aragonese zone (which, we recall, was geophysically shielded), and, by way of last redoubt, in the Alto Aragón. If this reconstruction of events is accurate, then most of Spain, to the extent that it had been weaned away from allegiance to Basque (or conjectural kindred tongues), now began to show a rapidly dwindling pattern of relationship between Latin and vernacular voiceless obstruents reminiscent of South-Central Italian and Rumanian.

Playing on Wartburg’s favorite terminology, or rather imagery, but in a way basically detrimental to the cause he himself championed, one can thus declare that Hispano-Romance, after having at the outset—ironically in view of its geographic position—been anything but “Western,” now, at long last, indeed began to become partially “Westernized.”

This idea, heterodox as it may sound at first, is not, strictly, new; in the mid-twenties, it was broached by Meyer-Lübke, who, however, with his staunch advocacy of a “Südromania,” ran into stiff resistance on the part of Menéndez Pidal, a formidable opponent indeed, the specific point at issue having been the apparently equivocal, hence controversial, representation of the consonants under study in Arabic script, within the context of Mozarabic texts.¹⁶ Over the last two decades, closer examination of poetic texts from Southern Spain, traceable, for the most part, to the late tenth, eleventh, and early twelfth centuries, as well as of telltale toponyms scattered over the southern, especially the southeastern, half—not only the southern tip—of the Peninsula and over the adjacent islands have convinced seasoned and many-sided specialists, such as the ranking explorer of Valencian M. Sanchis Guarner, that voiced and voiceless pronunciations may long have coexisted, presumably with tendential social differentiation, over a hazily delimited South-Central territory, but that Granada, Murcia, Valencia, and far-off Mallorca, in the Balearic archipelago, for a while remained almost completely intact from exposure to the oozing of sonantization (1961:291-342, especially 319-21). In this country Robert A. Hall, in the mid-seventies, resolutely sided with this opinion on, at least, two occasions;¹⁷ thirty years ago a more timid Lausberg had still made a point to echo Menéndez Pidal¹ s skepticism.¹⁸

But surely, if our conjecture of two successive and, in the end, conflicting waves is to hold water, there must have been other vestiges, at least stray, random vestiges, left, aside from the solid block of present-day Upper Aragonese and adjoining Béarnais evidence and from the debris of Mozarabic usage strewn over medieval Muslim Spain. What about present-day Peninsular dialects, so much more finely nuanced and differentiated than their, by and large, leveled overseas counterparts and spoken in a country suffering from no dearth of mountains and wastelands? With a measure of perseverance, pertinent dialectal illustrations from modern or recent sources can indeed be cited, and I shall gladly produce and discuss at length at least one such proof.

Included among my juvenilia is a rather extended study of the vicissitudes of the Latin verb quatere ‘to shake’ in Hispano-Romance, with constant side-glances at dialect usage (1946:104-59). Quateve itself failed to survive, having yielded ground to its iterative-intensive companion form quassāre,¹⁹ but such satellites of quatere (all of them involving the allomorph *-cutere) as perouteve ‘to strike, hit, beat’ (witness Eng. percussion), reoutere ‘to strike backwards, cause to rebound’, and suooutere ‘to fling aloft, toss up’ were all represented through well-shaped descendants in older Spanish: peroodir, veoodir, sa -or seoodir (with vowel dissimilation) . Later the -codir segment, for reasons understandable in a verb, was replaced by -cudir, while prefix change accounts for the eventual transmutation of recudir into acudir, paralleled by an appropriate semantic adjustment.

There was one complication of which I was fully aware thirty-five years ago, but which I could not at that point satisfactorily explain away (1946:149-54, 158). In addition to a profusion of dialectal forms in -cudir and of a whole palette of préfixai and suffixal offshoots therefrom: Ast. recudir ‘to dry (by exuding moisture)’, Arag. (Huesca) despercudir ‘to wash, cleanse’, Leon. (Cespedosa de Tormes) espercudir ‘to remain clean’, E.-And., also Mex. and Cub. empercudir ‘to sully’ (laundry), with the grotesque by-form epencudir ‘to do a poor job of washing’, once more observed in Southern Spain, one discovers, again and again, rival forms ending in -cutir, e.g. empercutir in nearby Salamanca. Moreover, one detects a semantically many-faceted form cutir, bare of any prefix and, importantly, lacking any such counterpart as *cudir. It has been identified in Murcia and Asturias, and Cervantes upon occasion had recourse to it four centuries ago. In Costa Rica, semantically unstable cutir has been confused with curtir ‘to tan’; in Ecuador, tundir ‘to beat, spank’ and cutir coalesced into cutundir ‘to beat up, whip, harass’.

Transparently, cutir cannot go back in a straight line to quatiō, -ere; just how ancestral qua- has developed under varying accentual conditions is illustrated with quátt (u) or ‘four’ > cuatro and quàdrāgintā ‘forty’ > OSp. quaraénta > Mod. cuarenta vs. quatt (u) órdeci (m) ‘fourteen’ > OSp. eatorze, Mod. catoree. Cutir therefore must have cut loose from the -cutere group, either at the time of the First Wave, or later, in a secluded area where the regime of that wave was destined to continue for a while. But why, one may ask at this juncture, was the t preserved after the speakers had switched to the norm of the Second Wave, replacing, as scholars incline to extrapolate, *capu by cabo ‘end, head’ (from Lat. caput), *totu by todo ‘whole’ (from tōtu), etc.? A clear answer can, for once, be provided: Wholesale substitutions of this sort were impossible, because, in countless cases hazily familiar to untutored speakers, a surd was shared by Wave One and Wave Two; e.g., ‘to bind’ may well have been atare (from Lat. aptāre ‘to fit’) and a ‘cat’ would equally well have been gato (from Late Lat. c-/g-attu) for followers of both norms, because in neither instance was an underlying intervocalic surd involved in the first place. So speakers were necessarily wary in making the change; even those eager to execute the leap from eapu to eabo presumably hesitated to switch from cutir to _*cudir, a form which, through a whim of circumstances, had failed to cut loose from per-cudir, etc. and which they, consequently, could not hear from the lips of more prestigious speakers. But once cutir was tolerated beside percudir and the latter’s numerous fellow satellites, compromise forms on the order of percutir were apt at any moment to come into existence; they may, but need not, be particularly old— they are latently ever-present, as it were.

I have examined this arresting case under a microscopic lens, on account of its symptomatic importance (and, let me confess, to poke fun at my own helplessness thirty-five years ago). I understand that a major study of similar (if not exactly analogous) instances of the widespread preservation of surds at the level of Peninsular rustic speech is in the making (or has been completed or is in press); what is more, it is gratifying enough, from the pen of a talented native investigator.²⁰

I do not rule out the concomitancy of other factors in regard to the ensemble of processes here examined, but I question their comparative weight. Thus, Lat. intervocalic -d- is known to have disappeared from Galician-Portuguese, in the company of -I- and -n-, for reasons not easily traceable to the Latin starting point; and its incipient, or just imminent, loss could have produced a vacuum and, ultimately, “pulled” t in the direction of d, with the further possibility that p and /k/, on account of their inherent affinity, may have joined it on its way to voicing. In this manner the contrast between Western /b/, /d/, /g/ and Eastern (Aragonese, Old Valencian) /p/, /t/, /k/ in characteristic cognates might have been somewhat reinforced.²¹ However, an appeal to the subsidiary agency of this force is not, strictly, indispensable in any effort to account for the forms on record.

The two conclusions to be drawn from this preliminary balance sheet are, first, that Hispano-Romance, on the strength of one important criterion, appears as an aggregate not of “Western,” but of, preponderantly, “Westernized” Romance dialects. If we agree that a gradual transfer from one group to another is, at least, a strong possibility, then the family-tree projection, which resorts for its symbolization to branches whose course obviously cannot be reversed at some intermediate point, ceases to be truly helpful. But its polar opposite, the unqualified wave theory, might preclude any orderly classification, too, and even erode the usefulness of such a concept as “Western(ized).” What emerges as a realistic alternative to the classic dichotomy, at least for certain concrete historical situations, is a neatly drawn two-or three-wave projection. In this revised diffusionist view, each successive major wave may force us to redraw the grouping of languages or dialects within a family.

III

On the second major problem here selected for scrutiny there exists, once more, a slender corpus of writings. Disappointingly, however, its specific repercussions in Hispano-Romance, here viewed as crucial, have not yet been recognized as a vital issue left in abeyance, yet worthy of a fresh frontal attack. Moreover, the close connection, implied by the structure of this paper, between the preceding question and the one now moving into the center of attention has not, generally, been granted. We shall thus, in certain respects, be treading virgin ground.

The problem now on our agenda has a bearing on the Romance vicissitudes of Latin vowels. Specifically, it concerns the tendency, in certain (in part neatly identifiable) varieties of provincial Latin, of two symmetrically distributed ancestral pairs, namely ĭ and ē, as well as ŭ and ō, to coalesce each, rather early, into a single phoneme, namely ẹ and ọ, respectively—with con-trastable degrees of closure of the mouth significantly replacing previous quantitative distinctions. Those familiar with the less widely studied Romance languages and with dialect usage are aware of the fact that the parallelism between the two pairs is incomplete: In Rumanian, above all, ĭ is indeed allowed to become e (thus, vĭrĭde ‘green’ changes to verde, and dĭgĭtu ‘finger’ to deget) and, as a result, to merge with the outgrowth of parental ē (thus, mese ‘tables’ [pl.] perpetuates mēnsae), unless secondary metaphonic disturbances are superimposed on this model, a shift which may in turn entail tertiary adjustments (as in the sg. masă ‘table’, via *measă). Conversely, ŭ normally follows a different, less radical course, by preserving its articulatory height, but losing in the process its distinctive quantity, with the consequence that ŭ and ū locally converge. Thus, lup ‘wolf’, from lŭpu, and urs ‘bear’, from ŭrsu, join not only mult ‘much’, from mŭltu, but also un ‘one’, from ūnu, and mut ‘dumb’, from mŭtu, over against, say, nepot ‘nephew’, from nepōt-e, *-u. (There is a handful of exceptions, notably toamna ‘fall, autumn’, from autumna [aetās], via *tomnă, allowing us to guess that an incipient balancing process was brusquely interrupted.) Typologically, East Lucanian joins Rumanian in this salient feature of asymmetry;²² this dialect was not so long ago discovered and surveyed in the Southern Italian province of Basilicata, wedged in between Calabria and Apulia.

Reverting now to the focus of events, without losing sight of goings-on in this “Eastern” side arena, we may briefly remind ourselves of the distressing fact that the seemingly symmetric conflation of two of the front vowels and two of the back vowels shared by the culturally most prestigious Romance languages—Portuguese, Spanish, and Catalan included—has long exerted such fascination in the minds of comparativists that, in manual upon manual of historical grammar, surely strung over more than one century, these two vocalic mergers ended up by being projected onto the temporal (and hierarchical) level of an ill-defined “Vulgar Latin,” a term traditionally used very loosely indeed, as Paul M. Lloyd not long ago pointedly reminded his European and American readers (1979: 110-22). Telling examples of ĭ either becoming e or at least passing through that stage on its way, for instance, toward some falling diphthong and of ŭ similarly either ending up as ọ or, transcending that stage, headed for, let us say, ou, can be counted by the scores, if not by the hundreds. As a matter of fact, on account of the reputed triviality of this point, authors of historical grammars rarely bother to toss off more than just two or three examples of the norm, concentrating thenceforth on the alleged exceptions.

Let me, for once, control this tradition of impulsive stinginess and provide an assortment of illustrations from Old Spanish, confining myself to the etymologically most transparent and grammatically least complicated cases, with consistent attention to the stressed syllable, protected from erosion, regardless whether that syllable was free or checked, and irrespective of the form classes involved :

1. atrevo ‘I dare’ ˂ attrib (u)ō ‘I assign [to myself the right]’, ceja ‘bow’ ˂ cilia (pl.) ‘eyelids’, cerca ‘close’ ˂ circa ‘around’, cesta ‘(large) basket’ ˂ cista ‘basket’, dedo ‘finger’ ˂ digitu, ella ‘she’ ˂ ilia ‘that one’ (f.), ˂ ende ‘thence’ ˂ inde, essa ‘that there’ (f.) ˂ ipsa ‘same, herself’, esta ‘this here’ (f.) ˂ ista, maestro ‘teacher, master’ ˂ magistru, menos ‘less’ ˂ minus, meto ‘I put, place’ ˂ mittō ‘I send’, mermo ‘I diminish’ *minimō, pera ‘pear’ ˂ pira ‘pears’, saeta ‘arrow’ ˂ sagitta, seco ‘dry’ ˂ siccu, selva ‘forest’ ˂ silva.

Contrastive examples, showing that e could be equally well arrived at from ē (including oe and, less consistently, ae) as a starting point in classical usage, include such familiar pairs as:

arena ‘sand’ ˂ (h)arēna, cesped ‘lawn’ ˂ caespite ‘turf, sod’ fezes (Mod. heces) ‘dregs, lees’ ˂ faecēs, peso ‘weight’ ˂ pē(n)su, pena ‘pair, torment’ poena ‘punishment’, and veno (later vino) ‘he came’ ˂ vēnit.

2. acoso ‘I pursue’ ˂ cursō ‘I run hither and thither, to and fro’, (en) sosso ‘tasteless’ ˂ insulsu, fondo ‘bottom’ ˂ fundu, gordo ‘thick, fat’ ˂ gurdu, lobo ‘wolf’ ˂ lupu, lodo ‘dirt, mud’ ˂ lutu, lomo ‘back, ridge, loin’ ˂ lumbu, mondo ‘pure’ ˂ mundu, olmo ‘elm tree’ ˂ ulmu, os(s)o ‘bear’ ˂ u(r)su, rompo ‘I break’ ˂ rumpō, sordo ‘deaf’ ˂ surdu, tomo ‘I take’ ˂ (aes) tumō ‘I assess,²³ tronco ‘trunk, log’ ˂ truncu.

Once more, it is easy to array complementary examples tending to demonstrate that o, analogously placed in other words of the same period and environment, is as smoothly traceable to ancestral ō: calor ‘warmth, heat’ ˂ calore, color ‘hue’ ˂ colōre, corono ‘crown’ ˂ corōna ‘garland, wreath’, fermoso (Mod. hermoso) ‘handsome, beautiful’ ˂ fōrmōsu ‘shapely’, forma (Mod. horma) ‘mold, shoe-tree’ fōrma ’shape’, olla ‘pot’ ōlla (∿aula), roe ‘he (she, it) gnaws ˂ rōdit, todo ‘all, whole, every’ ˂ tōtu.

Discoveries of this sort, which can be effortlessly extended to other Hispano-Romance dialects and which, via Catalan, lead to parallel findings in Old Provençal and beyond, have convinced several generations of Romanists that, at least so far as the concrete case of these four vowels is concerned, Spanish and Portuguese pertain at present, and have at all times pertained, to the core group of Romance languages and dialects. Lausberg, in his innovative mid-century manual,²⁴ carefully circumscribed the territory occupied by the aggregate of these tone-setting languages, crediting to it a total of eight major, median, or minor chunks: (1) all of Central Italy, (2) the northern edge of Southern Italy, (3) all of Northern Italy, (4) Dalmatia, (5) Istria, (6) the three islets of Raeto-Romania, (7) all of Gallo-Romania, (8) all of Hispano-Romania (short of the Basque zone, by implication), i.e. grosso modo, almost the entire extent of the Iberian Peninsula, of France, of Continental Italy, and adjacent areas to the North and to the East; plus fully three quarters of Peninsular Italy. Even in strictly spatial terms the territory thus staked out is extraordinarily impressive; add to its sheer size the further circumstance that such cultures as blossomed in Portugal, Spain, Catalonia, Provence, Northern France, and Central Italy produced some of the finest art and most exquisite literature that medieval Europe exhibited, and you will readily understand, without of course necessarily endorsing such heresy, how the trailblazers of Romance linguistics en bloc, and an embarrassingly high number of the most competent among their followers, down to the mid-twentieth century, could have virtually overlooked, or cavalierly brushed off as marginal curiosities or scattered instances of unique complication, any such traces of alternative developments as were apt to show up in their data collections.²⁵

Yet, upon more leisurely inspection, important and copiously represented alternatives do turn out to exist, on such a scale, indeed, that they no longer can be written off with impunity as inconsequential exceptions. Mention has already been made of a predominantly asymmetric Rumanian model (see note 22) and of a closely allied East Lucanian type;²⁶ there is also on record a separate, erratic Sicilian pattern, abounding in i’s and u’s²⁷ and conceivably generated through protracted contacts with a Greek-speaking population in that island. For our immediate purposes, this lateral aberrancy is irrelevant, and I think we are free to disregard it. The fourth and last pattern, however, which has of late been referred to as “archaic,” potentially constitutes very fertile ground for the Hispanist. It is the contention of this paper that this “archaic” system once struck root in the Iberian Peninsula, a conjecture that chronologically makes sense, against the background of ancient historiographie records; was later, to an appreciable extent, dislodged from that territory by the very same invader we already encountered on our previous foray into the domain of consonantism, namely the Latinity of Lugdunum and the nearby Narbonensis; but nevertheless has left a wealth of relics in Standard Spanish and Portuguese, as well as in the dialects—vestiges which, inexplicably, have either been consistently neglected or, equally dangerous, been noticed, yet hastily moved out of the way, through misinterpretation. But before thus plunging into the state of affairs in ancient Spain, we shall be well advised to examine, at least cursorily, what has already been securely established about the manifestations of the “archaic” vowel system in other corners of ROMANIA, where discoveries to this effect have fallen short of triggering heated controversy.

What encourages us to bracket together, under a single formulaic label, developments observed at widely scattered points of the far-flung Republic and, ultimately, Empire, is the local speakers’ recurrent disinclination to allow ĭ and ē, ŭ and ō to coalesce into a single unit-phoneme. To be sure, mergers did occur in the “archaic” zones, in step with the abolishment of distinctive vowel quantity; but the partners on these occasions were ī and ĭ, also ū and ŭ, which converged through neutralization of vowel length into phonemically indifferent i and u, respectively, to the strict exclusion of e’s and o’s from this process of regrouping.

Toward the end of the last century it was already known that the evolution of Latin sounds had taken this sharply profiled course in the better specimens of Sardinian dialect speech and within the thick Latin layer of the Albanian lexicon; the latter feat we owe mainly to the efforts of Gustav Meyer—witness that pioneering Balkanologist’ s contribution to the original opening volume of Gröber’s Grundriss (1888:§19, as against §14).

The patterns of those earliest identifications prepared scholars to expect additional discoveries at geographically distant, rather than adjoining points; in smaller, residual Romance language-and-culture areas (typical “Rückzugsgebiete”); plus, not one whit less noteworthy, in lexical residues of Latin absorbed—via bilingual contexts—by other, sometimes distantly (if at all) related languages which ultimately became dominant as “superstrata” in territories once held by the Romans and militarily, socially, or economically controlled by them to astonishingly varying degrees.

As a rule, such prognosis has turned out to be essentially correct. The generally up-to-date handbooks by Vidos and Lausberg, without fully coinciding in the bits of information they convey, provide, in capsulized form, the necessary briefing on what has been ascertained on African Latinity, through intensive sifting of the Berber vocabulary, and what, independently, has been established about the Latin elements of Basque, as distinct from its Spanish (i.e. Navarrese and Old Castilian) components, which slowly percolated at discernibly later dates. But this is not all. Lausberg is particularly authoritative on a long badly neglected zone in Southern Italy, an old “Sprachlandschaft” which straddles two modern provinces, comprising portions of Southern Lucania (Basilicata) and the adjacent northern rim of Calabria.²⁸ Vidos is, inevitably, less original on this score but, as if by compensation, refers his readers to the debris of Latin in the Serbocroat speech of the Adriatic island of Kres (It. Cherso) and to the Latin inscriptions of Pannonia, neither source adduced by Lausberg. Both treat in approximately equal detail—possibly because either depends on the pioneering research of M. L. Wagner—the evidence culled from South Corsican as well as from three core dialects of Sardinia. The two linguists happen to be in sharp disagreement on the relevant finer nuances of the submerged circum-Adriatic Latinity of Istria and Dalmatia.²⁹ Regrettably, neither makes the slightest allusion to the highly pertinent Latin deposits in the British Isles, obliquely observable through the prism of Cymric (Middle Welsh), on which enlightening data have of late been assembled, and/or insightful analyses proffered, by Kenneth Jackson and Harald Haarmann, in the Old World, as well as—in briefer critical appraisals—by Madison S. Beeler and Michael R. Dilts on this side of the Atlantic.³⁰ This point-by-point comparison of the positions of just two experts could, of course, be advantageously replaced by a polyphonic chorus of voices, were fairly recent syntheses by C. Tagliavini, G. Devoto, and other Romanists properly taken into account.³¹

Space is unavailable for a more than cursory illustration of the point here made—a sketch barely sufficient to prepare us for a fairly searching inspection of the far more recalcitrant Hispanic material, which alone, in the end, may enable us to draw any theoretical conclusions worth pondering. To start almost simultaneously at two ends of the line: Here is a sampling of the cross-temporal equations established as early, I repeat, as 1888 by G. Meyer between Albanian and Latin, with special reference to the preservation of ancestral ŭ as u, and not by any chance as o (note his use of e̥ in lieu of presently favored ë) :

buke̥ ‘bread’ ˂ bucca, lit. ‘inflated cheek’ [engaged in chewing], funt ‘ground’ ˂ fundu, fure̥ ‘oven’ ˂ furnu (with gender change), furke̥ ‘distaff, hay- or pitchfork’ ˂ furca, gune̥ ‘overcoat’ ˂ Late Lat. gunna (a transparent borrowing), gušt ‘August’ (Au)gustu, gute̥ ‘gout’ ˂gutta lit. ‘drop’, kulm ‘ridge of a roof’ ˂ culmen ∿ columen (perhaps confused with culmus ‘stalk, stem’), k’un ‘wedge’cŭneu, kut ‘ell, yard’ ˂ cubitu lit. ‘elbow’, l’ufte̥ ‘struggle, fight’luctḁ, l’undre̥ ‘boat, ferry’ lunter (cf. Rum. luntre), mušt ‘must, cider, jui̊ce’ ˂ mustu, ngušt ‘narrow’ ˂ angustu, pe̥lumbe̥ ‘dove, pigeon’ ˂ palumba, pl’ump ‘lead’ ˂ plumbu, pus ‘weli̊’ ˂ puteu, etc.³²

Turning our attention next to recent statements on British Latin, or Romano-Brittonic, to the extent that its mosaic can be reconstructed from older and more modern Welsh, we find such testimonies to archaic modes of pronunciation as fide ‘faith’ > ffydd [fɪđ], pice ‘pitch’ > [pʉg] / [pɪg] / and plumbu ‘lead’ > plwm [plum], all of which takes us away, so far as the timbre of the stressed back vowel is concerned, from the prototypes of Fr. foi, poix, and plomb, even if we strive to pronounce these latter words in medieval fashion. Conversely, Welsh plwm approximates with astonishing neatness Alb. pl’ump or plump, obviously, again only with respect to the accented vowel. Then, too, both languages, despite the respectable distance between them whichever yardstick one cares to use for measurement, are in substantial harmony as to the separate treatment they mete out to ancestral ō: They concur in dragging it away from ŭ. Thus, corōna, which has already crossed our path, emerges in Welsh as corun ‘top, crown of head’ (i.e. with u rather than w), and in Albanian, depending on the dialect chosen, as (metathe-sized) kunorë or (Tosk) kurorë (i.e. with o, not u) . Contrast this state of affairs with the tendential convergence of the two vowels in the French cognates: plomb /plõ/, originally /plõ/, alongside couronne /kuron/, originally /kurõnə/.

One can easily obtain specimens very similar to those just cited, in regard to the feature selected, by sifting even small quantities of Sardinian records, or relevant data extracted from any other of the peripheral residues of Latin inventoried above and already investigated by thoroughly dependable scholars.

The crucial question now, at long last, before us is: What bearing do all these arresting discoveries, made little by little in outlying lexical deposits of Latin, conceivably have on research traditionally conducted in Hispanic headquarters? The answer, couched in cautious terms, is: The consequences for the classification of Hispano-Romance within the confines of the larger family may be unforeseeable, if only one is predisposed to cast a fresh, uninhibited look at the evidence. Even if material extracted from such mines of information as Albanian and Lucanian may seem farfetched, there surely is no excuse for neglecting the unequivocal testimony of Basque, which coincides with those of the remote witnesses.

Consider the following fairly clear-cut instances of vacillation between o and u as between Spanish and Portuguese, or as between two consecutive phases of either language, or else as between two members of the same lexical family viewed in synchronic perspective:

OSp. cobd-, cod-icia ∿ cudicia, Ptg. cobiça ‘greed’, from cupiditās, through substitution of the learned suffix -icia, semilearned -iça ˂ -itia (here cited on account of the loss of cupidus ‘longing, fond’); Sp. corto, Ptg. curto ‘short’ ˂ curtu (but Sp. Ptg. cortar ‘to cut’); Sp. corvo beside curvo ‘arched, curved, bent’, Ptg. curvo ˂ curvu ‘crooked’; OSp. dubda ∿ dobda (dolda), Ptg. dúvida ‘doubt’, from dubdar (Mod. dudar), etc. ˂ dubitāre ‘to doubt’; OSp. duz/duoe beside doz/doce, Ptg. doce ‘sweet’ ˂ dulce; OSp. enxundia (Mod. enj-), Ptg. enxúndia ‘animal (e.g. pork, chicken) fat’ ˂ axungia; OSp. fondo (> Mod. hondo) ‘deep’, fondo (Old and Mod.) ‘bottom, rear’, Ptg. fundo ‘deep’ (adj.), ‘bottom’ (subst.) ˂ fundu ‘bottom, foundation’ alongside profundu ‘deep, vast’; Sp. gola ‘gullet’ beside gula ‘gluttony’ ˂ gula ‘throat, gullet’, fig. ‘gluttony’ (Sp. goloso ‘sweet-toothed’, ‘gourmand’, cf. golosina ‘delicacy, tidbit’, contrasts with guloso ‘gluttonous’; less frequent Ptg. gola, gula are likely to be Castilianisms, and only guloso is on record); Sp. gusto, Ptg. gôsto ‘relish, enjoyment, pleasure’ ˂ gustu ‘taste’; OSp. os(s)o (Mod. oso), OPtg. usso (Mod. urso) ‘bear’ ˂ ursu; Sp. plomo, Ptg. chumbo ‘lead’ ˂ plumbu ; Sp. punto, Ptg. ponto ‘point’ ˂ punctu ‘small hole’; Sp. Ptg. sombra ‘shade, shadow’, but Sp. umbrio ‘shady’, umbría ‘shady place’, from umbra (+ so- ˂ sub- ‘under’? or through semantic polarization with sol ‘sun’?); Sp. sollozo, Ptg. soluço ‘I sob’ ˂ singultiō (via *sub-gluttiō); Sp. sordo, Ptg. surdo ‘deaf’ ˂ surdu ; Sp. suma, Ptg. so-, su-ma ‘sum’ ˂ summa (but Sp. asomar, Ptg. assomar ‘to emerge, loom, begin to show’, with the implication ‘at the hilltop’); Sp. tronco ∿ trunco’, Ptg. trunco ‘I cut off, lop’ ˂ truncō, but Sp. Ptg. tronco ‘tree trunk, torso’ ˂ truncu, lit. ‘maimed, mangled’.³³

This is by no means a homogeneous group; in a few instances rival explanations are admissible, through appeal to learnèd transmission; in other cases multiple causation may be involved. (Certain exhibits of apparently similar distribution have been removed, where their irrelevance became immediately clear; thus Sp. oost-a, -e, -o vs. Ptg. cust-a go back to a verb costar/custar, based on cōnstāre, which is of no use in any discussion of Lat. ŭ, although the split illustrates one direction of subsequent spread of the wave-ring.³⁴

If our broad-gauged conjecture of an overlay of native “archaic” u ˂ ŭ by a foreign more “advanced” o ˂ ŭ raying out from Southern France is at all tenable, then the next phase must have been one of intensive dialect mixture, a state of temporary “muddiness” of contour out of which a new order or equilibrium was bound to emerge. It is a fair assumption, the theoretical basis for which was tentatively laid in the mid-sixties,³⁵ that the norms presiding over such reordering of an imbroglio are very narrow, requiring at every step the analyst’s close attention to fine details of each relevant word’s individual biography, which may show a very zigzagging course. One can fully expect that, in such no doubt slow filtering process, small molecules of lexical units, held together by bonds of form and/or meaning, will tend to exhibit parallel directions of development; in such a context, non-phonological factors, e.g. formal polarization of semantic opposites, avoidance of bothersome homonymy, mutual attraction exercised by synonyms and near-synonyms, availability of potential substitutes, préexistence of Provençal-Hispanic cognates, are apt, in the aggregate, to play a conspicuous role; while certain minor trends of sound development, e.g. the suspected influence of surrounding phonemes on the chosen feature, may indeed, but need not, “make sense” in strictly phonological terms, being often attributable to all sorts of associative phenomena so elusive as to appear to be bordering on chance.

Consider, by way of illustration, Sp. cumbre ‘height, peak, summit’, which has traditionally—and, up to a certain point, cogently—been traced to ancestral culmen (var. colŭmen) ‘top, summit, ridge’.³⁶ This derivation involves or implies a number of difficulties, of which the preservation of ŭ as u, counter to what is commonly held to be the local “law” or “rule,” is merely one; the most outstanding among the others are: (1) the loss of medial preconsonantal l (in harmony with alteru ‘the other’ > otro ‘another’, but against altu ‘high, tall, deep’ > alto, which in turn is contradicted by ot-ero ‘hillock, knoll’); and (2) the hypothesized survival of a parental formation consisting of -men attached to a consonantal root, the tendency having actually been to eliminate such words (blurred derivatives) from living speech (cf. the extinction of ag-men ‘crowd, army, band’, ful-men ‘lightning, thunderbolt’, seg-men ‘cutting, shred’, tegmen [beside -ĭmen, -ŭmen] ‘cover’, and of many others). Hardly a difficulty, but a peculiarity worth watching, is the change of gender undergone by the supposed base: culmen was neuter and as such was normally headed for the status of a masculine in any branch of Romance, as is indeed true of its Portuguese product cume; but Sp. cumbre has at all times been feminine.

How are these various threads to be most persuasively connected? To begin with, Portuguese also uses the oft-cited word gume (m.) ‘sharpness’, which, despite the rarity of the apheresis demanded by this genetic hypothesis, has unanimously been traced to acūmen ‘sharpness’, i.e. to the abstract flanking the past-participial adj. acutus ‘sharp’. The apheresis must have postdated the voicing of /k/. Since ‘sharpness’ and ‘height’ do display an imagerial overlap (remember the contour of a broken cliff, a projecting rock, a mountain peak, and the like), a gradual rapprochement between acūmen and cŭlmen is conceivable. It would help to explain the prevalence of the u (the normal outcome almost everywhere of Lat. ŭ), where speakers had a choice between the two back vowels, and might account for the striking apheresis in Ptg. gume alongside agudo ‘sharp’ (a discrepancy so far left unexplained). Moreover, the discovery of scattered vestiges of acūmen in Gallo-Romance (inventoried in W. von Wartburg’s FEW) makes its extended survival in yet other corners of Romance territory fairly plausible. Could Sp. cumbre represent a blend of aeūmine and culmine, one that has inherited its shape chiefly from the former and its meaning from the latter? Note the contrast, as regards the crucial vowel and the nuclear consonant cluster, to Sp. colmo, Leon, cuelmo ‘thatch’ ˂ cŭlmu ‘stalk, stem, straw’ (a reliable etymology, despite Corominas’ capricious disagreement).

One can think of an alternative to this supposition of a straight lexical blend. Cumbre is not alone in having undergone the aforementioned gender switch: several Spanish words in -(u)mbre (also, secondarily and with less consistency, other items that end in -ambre, -imbre, and -iembre) are feminine, even though their Galician-Portuguese counterparts, less surprisingly, are masculine, while the Latin prototypes were neuter: lumbre/lume ‘light, fire’ and legumbre/legume ‘vegetable, green’ are examples in point, involving the ancestral segment -ūmine/ -ūmen. The pressure of these and similar words (to whose category aforementioned acūmen also belongs) might have sufficed (1) to push speakers in the direction of u rather than o; (2) to prevail on them to drop the unstable l; and (3) to dangle before their mouths and ears the advantage of the feminine gender in this particular context.

Exiguous as the extant literature on this point of gender reshuffling happens to be,³⁷ enough has been established to justify the remark that the conflation on Spanish soil of the three suffixes (1) -(t)ŭdō, -inis,(2) -ŭgō, -inis, and (3) -ŭmen, -minis (all three, significantly, lending services to derive now abstracts, now mass nouns from various classes of primitives) must have produced a real turmoil in the central section of the Peninsula, since (1) and (2) were feminine, while (3), confusingly, was neuter. The resulting blend, -(ed)umbre,turned out to be feminine; witness (1) muchedumbre ‘multitude’, (2) herrumbre ‘iron rust’, and (3) legumbre. Portuguese kept these three derivational suffixes apart (as -idão, orig. -idõe; -ugem; and -ume), assigning the masculine gender—the expected substitute—to the erstwhile neuter and leaving the other two in the original feminine gender slot.

Interestingly, there developed before long, presumably stimulated by the bold leap executed en bloc by the contingent of -umbre words, the broader tendency to relegate nearly all nouns in -re to the feminine in Spanish, whichever their specific antecedents. Hence the well-known, striking series landre ‘small tumor’ glāns, -ndis (∿*-inis), mugre ‘dirt, filth’ mūcōre ‘mould, mustiness’ (with stress shift), podre ‘pus, corruption’ (m. ∿ f.) ˂ pŭter, -tris ‘mouldering, decaying’, postre (as in the phrase a la postre ‘finally’, which distinctly outranks al postre; but [m.] postre ‘dessert’) from post ‘after-(ward)’, sangre ‘blood’ ˂ sanguine, and ubre ‘teat, udder’(n.) ŭbere ‘breast that gives suck’. Here in Portuguese, assuming this language offers a pendant, that counterpart will tend to be masculine: witness podre, sangue, and úb(e)re. Now, some of the most sharply profiled words in this series combine the swing of the pendulum toward the feminine with a preference for the u vowel, as a rule traceable to w. Azufre (originally aç-) ‘brimstone’, from Gr.-Lat. sŭlphure —conceivably transmitted through Arabic—may initially have been ambi-generic, like OSp. açúcar ‘sugar’, árvor(e) > árbol ‘tree’, all of them ushered in by the definite article el(l), indeterminate as to gender in words so structured (cf. Mod. el águila ‘the eagle’, el agua ‘the water’). Preponderantly, though far from mandatorily, the word-final segment -re, in Spanish nouns, at a certain point began to show a strong affinity with the feminine gender and a marked compatibility with the stressed vowel u, so that, even if (a)cūmine, counter to probability, had been unavailable on Spanish soil to tilt the vocalic nucleus of cŭlmine toward a u, the combined pressure of lumbre, legumbre, etc., as well as of mugre and ubre, might still have sufficed to perform this service. (Perhaps not coincidentally, ŭtere ‘goatskin wine bag’ wc.s allowed to retain its masculine gender when speakers opted in favor of odre, to the exclusion of equally imaginable *udre.)

In such a welter of confusion it was to be expected that certain words containing o traceable to a source other than ŭ (e.g. to ō or au), should eventually have been dragged into the rivalry between o and u and, by way of hypercorrection or factional overassertion, have been, at first tentatively, and, then definitively, assigned to the u camp. Thus, the familiar calendar unit Octōbre, judging from the state of affairs in extra-Peninsular languages, should have been reflected by vernacular formations containing stressed ō rather than u, as was clearly not the case with OSp. Oohubre (> Mod. Octubre, partially Latinized), Ptg. oit-, out-ubro; also, the word for ‘knot’, from nōdu, perfectly regular in Portuguese (nó, contracted from earlier noo), might in Spanish, ideally, have been _*nodo rather than nudo, var. ñudo. Cat. uytubre and nu speak in unison with Spanish. Against the voices of earlier scholars, who were willing to operate with putative *Octobrius, Menendez Pidal, down to the latest revision of his handbook of historical grammar (1941:§2), appealed for an explanation of the aberrant vowel to Oscan or Oscanized Latin, whose ŭ sometimes matched Lat. ō, and cited even ancient epigraphic evidence to this effect. Perhaps so; but, on balance, it appears simpler to invoke false correction, a factor reinforced in Proto-or Old Spanish by the affinity of -re with nuclear u in the case of Oohubre. As for nudo, it is safest to start with the verb; innōdāre ‘to knot’ > *eñ-, *añ-odar could easily have been influenced by (d)esnu (d)ar‘to bare, strip naked’ despite the almost forbidding semantic distance—a factor apparently less than significant in this connection (witness the impact of siega ‘he mows’ on riega ‘he waters’ and pliega ‘he folds’; or that of acuesta ‘he lays down, puts to bed’, from accŏstat, on cuesta ‘it costs’, from cō(n)stat). It is in this context that aforementioned Ptg. eus ta belatedly acquires a limited measure of relevancy.

Earlier generations of scholars were of course by no means unaware of the wide margin of doubt attendant upon the development of ŭ in Luso-and Hispano-Romance, but insisted, or rather took it for granted, that the pattern lŭpu > lobo represented the norm. The problem for the spokesmen for that school of thought amounted, consequently, to academic exercises in explaining away the unaccom-modatingly numerous “exceptions,” by grouping them, first in classes and, next, wherever this was feasible, in subclasses. Since “exceptions,” by definition, were embarrassing to historical linguists of neogrammatical persuasion, not a few counterexamples were simply omitted from mention; in this respect Menéndez Pidal’s Cid grammar, with its goal of exhaustive treatment of all and any aberrancies, stands in glorious isolation. Then again, few practitioners of that approach were candid enough to admit the self-contradictory pattern of the Peninsular distribution of ŭ, now as u now as o, and the damaging effect of this record of apparent inconsistency on the strict doctrine unerlying their analyses; for sheer candor, Meyer-Liibke, as a comparatist, deserves recognition (1890:§147).

Typically, after having tacitly removed exotic words (such as açúcar ‘sugar’ ˂ Ar. sukkar) from further consideration, the authors of turn-of-the-century and even later historical grammars would break down the residue of rebellious cases into two categories:

1.Learnèd (or partially learnèd) words, i.e. lexical items exhibiting u in lieu of o through deliberate imitation of the given prototype. On this front, the principal disadvantage has been the (continued) unavailability of any theoretically underpinned study of early “cultismos”;

2.Vernacular words influenced (“deflected”) in their growth by special, superadded phonic conditions. These lend themselves, as a rule, to further bi-or tri-furcation, even though opinions of experts on particulars have not always coincided; what brackets the various consecutive pronouncements is the analysts’ shared belief in a sort of subjacent zigzag movement: Classical ŭ is assumed to have been, initially, transformed into /o/ in these exceptional instances, as in all other words (barring, at most, a few isolated “cultismos” of early vintage), only to revert, with a later swing of the pendulum, to /u/, under pressure from this special—as it were, newly emergent—factor. Now, zigzag movements are known to have indeed occurred, if only at rare intervals; thus, the s of casa ‘cabin’ (a) was voiceless in Antiquity; then (b) became voiced in Hispano-Romance and in some other descendants of provincial Latin; and (c), more than a millennium later, was devoiced in Spanish (though not in Portuguese), and thus once more became /s/. While such concatenations of events are not unparalleled, then, their clustering on the scale visualized by most pioneers of the o ∿ u controversy is unprecedented (and unrealistic, to boot).

One potential alternative has been lost sight of, in the process, which acquires weight if one prefers to operate with the hypothesis of two successive and, in the end, conflicting waves of Latinity. Assuming, as we have done, that the first wave favored retention of u as u and the second its lowering to o—could it not be argued that certain sharply silhouetted environments (hiatus, the mediate or immediate vicinity of a palatal consonant, the adjacency of some neatly profiled consonant cluster, and the like) might have prevented the onrushing /o/ wave from overrunning certain (structural) islands and islets of resistance to innovation—a reconstruction that would, as a result, eliminate any need for having recourse to a zigzag movement?

While representatives of the school of thought here impugned have learned to distinguish sharply between the two channels of transmission (1) and (2), most if not all of them have overlooked the marginal possibility of interaction of forces. Thus, at first glance Sp. dulce ‘sweet’ clashes so sharply with Ptg. doce, Fr. doux (f. douce), It. dolce, etc., that one is tempted to declare it a form artificially transplanted by Church Latin onto Spanish soil. This procedure presupposes the lumping together of all its distinctive features. Yet the example of Rum. dulce, which in every single particular conforms to standard sound correspondences, and yet very much (except for its assibilated c /č/) sounds and even looks like its Latin model, should caution us against hasty conclusions. The u of the Spanish congener could, after all, have been imported by the first (“archaic”) wave; the preservation of l before consonant in a primary cluster, while avowedly erratic, is far from unprecedented; finally, the medieval record discloses such MS variants as duc(e), doc(e), e.g. in Juan Ruiz. One could extrapolate from this array of circumstances a more delicately nuanced biography of ancestral dulce in early Spanish. The dispersal of variants is proof that duç(e), doc(e), and perhaps even dulc(e) could very well have sprung into existence without much assistance from Church Latin. Nevertheless, the undeniable fact that, wherever the speech community had an option (e.g. in the choice of the stressed vowel; in the salvaging, as against omission, of /l/ before the sibilant; and, above all, in the restoration of the final vowel—note the striking contrast to hoz ‘sickle’ ˂ falce and to haz ‘sheaf’ fasce), the feature préexistent in the Latin prototype was consistently preferred—makes it highly probable that the growing influence of that model interfered, to some extent, with the pattern of spontaneous, uninhibited growth observable elsewhere.³⁸

Let us now see, by means of quick spot checks, how earlier authorities, over a period of slightly more than eighty years (1888-1969), have judged the issue of learnèd transmission of ŭ as u rather than o; for the sake of fairness, appeal will be made, in chronological order, to spokesmen for German, Austrian, German Swiss, Spanish, and North American scholarship.

1. G. Baist (1888:§23) peremptorily declared orudo ‘raw’ and ludo ‘game’ learned, on the strength of the preservation of -d- ; then argued that nōdu ‘knot’, after the loss of its central pillar (witness Ptg. nó), had the voiced dental restored through learned pressure, and in the process coincidentally adopted the vowel of crudo and ludo. (In criticism it suffices to remark that the two Peninsular languages steered dissimilar courses with respect to Lat. -d-; that crúo and crudo coexisted in Old Spanish; and that, judging from its record, ludo, in lieu of juego ‘game play’, indeed was a “cultismo,¹¹ but certainly neither crudo nor nudo.)
2. A still young W. Meyer-Lübke (1890: §147) decreed that Sp. (also Ptg.) cruz, pulpa, surco (Ptg. sulco), yugo,and bulto owed their deviation from the straight path to their status as Latinisms. Anticipating doubts as to how words pertaining to the humble rural vocabulary could possibly qualify for such a lofty status, he remarked parenthetically that the genuinely “popular” word for ‘furrow’ in Portuguese was not suloo, but rega (he supplied no such vernacular equivalent for Spanish); similarly, he felt, Ptg. canga ‘yoke for oxen’ represented the vernacular counterpart of yugo. (This statement contains a few glaring errors, in part factual; Ptg. rega—far more commonly rego— designates an ‘irrigation ditch’ rather than a ‘furrow’; Ptg. jugo is used as a term of rural carpentry, in addition to denoting a ‘yoke’, and thus cannot be cavalierly written off; Sp. pulpa ‘pulp’, on record since 1400, is learnèd for Meyer-Liibke and semilearnèd for Corominas [1967:482a] but its interpretation as vernacular could equally well be defended; does not the sheer spelling of bulto, in preference to vulto, separate it from the contingent of straight “cultismos”? But over against such slips, note Meyer-Liibke ‘ s mastery in bracketing coñuseo ‘with us’ with conmigo ‘with me’, and his commendable candor in admitting his inability to account for Sp. nunea ‘never’ and juneo ‘reed’ beside troneo ‘trunk’; for Sp. cumbre/ Ptg. cume —where he hesitatingly suspects the complicity of the -I-[see above and below]; for the unbridgeable gap separating Ptg. chumbo from Sp. plomo ‘lead’ and from Ptg. lombo ‘animal’, ‘back, loin’ as well. Also included in this penumbra are Ptg. curto, curvo, custa, surdo, urso [as against tordo ‘thrush’, shared by both languages, incidentally; and as against Sp. oso]; Sp. duda ‘doubt’, if not Ptg. dúvida; and Sp. nudo ‘knot’.)
3. R. Menéndez Pidal, in his first pronouncement on phonology (1904:§14.1, n.l), drew a sharp dividing line between (α) learnèd púrpura ’purple’, número ‘number’, mundo, cruz, bul(d)a ‘bull’, lucro ‘gain, profit’ and (β) vernacular pórpola, nombre ‘noun’, mondo ‘clean’, bolla ‘tax on manufacture of playing cards’, and logro ‘attainment, gain’, ‘usury’. (The reader wishes he were told more about the contrasting ambits of such authentic doublets as púrpura and pórpola, luoro and logro; mention of nombre—unlike the situation in French—involves a downright mistake, since the Spanish noun is descended from nōmine rather than from numeru.)
4. The same scholar, acting a few years later as an ex-egete of the Cid epic (1908:147), reiterated his previous comment on cruz and mundo, but added that mondo, in Old Spanish, served not only as a qualifier (‘clean’) but also, intermittently, as a designation of ‘world’ (in imitation, one is tempted to ask, of an Old French model? Cf. also It. mondo).
5. A. Zauner, in his revised grammar of Old Spanish (1921: §10), declared cruz and mundo Latinisms, confessing his inability (despite Meyer-Lŭbke’s helpful hint dropped over thirty years before!) to cope with coñusco ‘with us’ and convusoo ‘with you’.
6. J. Huber, another student of Meyer-Lübke’s, but one of somewhat later vintage, gratuitously complicated matters (1933:§95, n.2) by discriminating between “ecclesiastic terms” (including cruz and culpa ‘fault, guilt’) and “book terms” (e.g. mundo and segundo ‘second’) in Old Portuguese. This last-mentioned remark certainly should have entailed a brief elaboration.
7. With Edwin B. Williams (§1938:§§6, 38.IB) not only another Lusophile, but also a representative of the tradition of American research entered the arena. He avowedly lumped together into a single pile all such words as seemed to him to testify in favor of learnèd or partially learnèd transmission: cruz, culpa, curto, curvo, fundo (adj. and noun, one supposes), furto ‘theft’, mundo, segundo, sulco, surdo, plus OPtg. usso ‘bear’ > Mod. urso. In this list only the inclusion of the progeny of fŭrtum is innovative; but note that in Classical Latin fūr ‘thief’ (and its subfamily) clashed with fŭrtum, the way nīvit ‘it is snowing’ competed with nĭx, -vis ‘snow’. Spanish hurto, of course, followed a parallel trajectory. More important than the enrichment of the inventory, however, was Williams’ decision to cite the minority view of two fellow scholars who were willing to reckon with the preservation of vestiges of parental Ŭ as u, even if he, Williams, did not see fit to endorse their opinion: the American pioneer Edwin H. Tuttle, plus P. Fouché (a point expatiated upon below).
8. Finally, H. Lausberg (1969:§183) reverted, somewhat anticlimactically, to a view championed long before his time: He declared Sp. Ptg. cruz, Sp. dulce, and Ptg. surdo “Buchwörter”—as if a word for ‘deaf’ could have been drawn so easily from that category. Unwittingly, by aligning cruz with its cognates transmitted, upon his own admission, through vernacular channels—Sard, rughe, Rum. cruce, OProv. crotz, Fr. croix, Surs, crusch, It. croce, Vegl. [i.e. N.-Dalm.] kŕauk—he dramatized the isolation into which his avenue of approach, one ventures to think gratuitously, had pushed Hisp. cruz.

To round out this phase of the discussion: What argument forces scholars to drive a wedge between the Luso-Hispanic and the general Romance transmission of crŭce? In the related case of lūce ‘light’ (Class, lūx, lūcis),there simply are no internal criteria—phonological, morphological, or semantic—for distinguishing between word-of-mouth and learnèd channels of transmission, as regards Sp. Ptg. luz and Tusc. luce, even though fragments of the external record of the word (e.g. date and place of appearance, genres of the relevant texts, bits of biographic information on the authors involved, and the like) may swing the pendulum of probabilistic analysis one way or the other. With respect to cruz, the single internal criterion so far adduced has been, precisely, ŭ > u. But once scholars can be persuaded, on the basis of their experience with hurto, surco, surdo, and the like, to accept ŭ > u as a residual reflex of the archaic vowel system in the Peninsula (“First Wave”) the last barrier to joint classification of cruz, croix, croce, etc. will fall.

From these counterproposals it must not be inferred that Hisp. u traceable to ŭ should under no circumstances be charged to a Latinizing tendency: diluvio ‘deluge’ and estudio ‘study’, e.g., to cite the two key formations alleged by P. Fouché (1929:16), display sufficient independent phonic and semantic features to allow the observer to declare them “cultismos.” One simply ought to be extra-cautious in classifying a given word as a Latinism on the sole grounds of its exhibiting such a seemingly erratic vowel correspondence as u ˂ ŭ.

On the emergence of u in the vernacular as a representative of parental u for “non-cultural reasons,” there exists a dense network of opinions, only a few of which can, almost randomly, be selected for our survey of earlier pronouncements. It will be remembered, from our preliminary statement, that a strong majority of scholars has opted, over the years, for the assumption of a zigzag movement: ŭ > o > u, though there has been no consensus on certain details. The principal spokesman for this dominant school of thought was, indisputably, for many decades, Menéndez Pidal, whose ceaseless concern with this problem can be followed by observing the constant revisions made in the successive editions of the Manuel de gramática. Along this meandering path, he remained convinced that Lat. ŭ and ō had become firmly and inextricably amalgamated into ọ at the Vulgar Latin stage, so that, wherever the Hispanic record showed vernacular u for expected o, he endeavored to find a reason for the deviation (a hypothesis which, on top of other obvious disadvantages, also involved the given word’s embarrassing return to its starting point in Classical Latin).

In his Cid grammar (1908:147-9) Menéndez Pidal saw the problem of “u in lieu of o” dissolve into a mosaic of individual lexical issues: an inflectional pattern explained conuvo ‘he learned’ from conocer; a suffix change accounted for rancura instead of rancor; pressure of the semantic opposite suso ‘upward’ justified the rise of yuso ‘downward’ beside etymologically more “correct,” if sparsely documented, yoso (a split with counterparts in Gallo-Romance); an innovation at the level of Vulgar Latin had transmuted ōstium ‘exit’ into *ūstium (hence OSp. uço); Portuguese and Catalan congeners with -u- flank OSp. nunquas ‘never’, while Old French and Old Provençal here opted for -o-. The author expressly admits his inability to justify dubda ‘doubt’ beside infrequent dobda, dolda, and documents the wavering connusco ∿ connosco ‘with us’, convusco ∿ convosco ‘with you’ leaving it, however, unexplained. The only phonological conditioning factors that he recognized at this stage of his growth are environmental: (1) Before a palatal consonant traceable in part to /k/, /g/, or /l/ u was favored over o, hence OSp. conducho ‘supplied, provisions’ ˂ conductu, puño ‘fist’ ˂ pugnu [a graphy which modern Latinists, however, would tend to interpret as /pŭŋnu/], uña ‘nail’ ungula ‘(little) nail’; (2) l before consonants [whether preserved or becoming submerged in the process] also favors the return to u, witness azufre, buitre ‘vulture’ vol-, vulture, cumbre, and dulce [the last item is thus separated, by implication, from the contingent of Latin-isms].

From these, all told, modest beginnings Menéndez Pidal advanced very far by the time he revised his handbook for the last time (1941 :§14.2) . He attributed the lion’s share of changes of *ọ > u, including those involving ọ ˂ ŭ and thus, strictly, subject to the alternative interpretation of retention of u, to metaphony (his own term for this process was “inflexión”), whose agency, real or reputed, had meanwhile become dramatized by M. Křepinský’s dissertation and the comments V. García de Diego had attached to its translation (1923).

Using his newly devised scheme of the “four consecutive yods,” Menendez Pidal now argued that, of the consonants following upon the critical vowel (ọ), (1) ç and z lacked the power to raise it to u, hence pozo ‘well’ ˂ pŭteu; (2) n, to the extent that it reflected older /nj/, did raise it, hence cuña [∿ cuño] ‘wedge’ cŭneu, puño [see above], and the suffix -uño, as against otoño ˂ autumnu; (3) y or i, to the extent that they perpetuated /j/ wholly or in part, showed inconsistency, witness especially rubio ‘blond’ ∿ (dial.) ruyo (toponyms in Soria, Burgos, Avila) ∿ royo (toponyms in Castile, Aragon, Andalusia) ˂ rŭbeu ‘reddish’—there is a thin connecting line from here to the manifestations or metaphony in conjugation (huyo ‘I flee’ ˂ fŭgiō); and (4) ch /č/, whether from -ct- or -ult-, on an overwhelming scale converted ọ into u: aguaducho ‘aqueduct’ ˂ aqua(e)dŭctu, lucha ‘struggle’ ˂ lŭcta, trucha ‘trout’ trŭcta; escucho [OSp. ascucho] ‘I listen’ a(u)scǔltō, mucho ‘much, very’ ˂ multu. As a sort of unaccountable residue Menéndez Pidal listed cases like cumbre ˂ cŭlmine, empujo ‘push’ impŭlsu, azufre ˂ sŭlphure vs. (en)soso ‘tasteless’ insŭlsu; and poso ‘sediment’ pŭlsu. After a long excursus on the vicissitudes of the diphthong oi > ue, the author briefly reverted to the ọ > u issue at the very end of the far too long paragraph, citing examples of raising caused by -io ˂ -ĭdu, as in dial, rucio ‘dew’ rōscidu, turbio [OSp. turvio] ‘muddy, troubled, confused’ ˂ tŭrbidu, all of them involving a Latin vowel secondarily transformed into a Romance semiconsonant, as a result of syllabic contraction.

Many Hispanists have expressed ideas similar to those launched by Menéndez Pidal; there would be little point in itemizing minor variations.³⁹ However, an assortment of major divergences invites a short digest.

The original version of Baist’s grammatical sketch (1888:§23), which antedates Menéndez Pidal’s earliest gropings by a wide margin, may appropriately lead off this survey. Baist assumed that ọ changed to u in hiatus, citing suyo ‘his, hers, its, theirs’ and tuyo ‘thy, thine’, from suu, tuu, and charging the intercalation of the /j/ glide to the influence of cuyo ‘whose’ cūiu(s). He distinguished carefully between the different effects of ñ₁ (˂ gn, /nj/) and ñ₂ (˂ nn, mn), and in this framework effectively contrasted cuño ‘wedge’ ˂ cŭneu with coño ‘cunt’ cŭnnu, without capitalizing on the possible desirability, for speakers, to keep apart as neatly as possible a neutral word and one struck by sexual taboo; he segregated into a separate category such items as punto ‘point’ ˂ pŭnctu, junto ‘joint’ iŭnctu, unto ‘ointed’ ŭnctu, preguntar ‘to ask’ (without here giving away his etymological credo), and nunca ‘never’ ˂ numqua(m), recognizing throughout the agency of the same sound law: ọ > u before -nt- (nct) that is observable in Italian; finally, he maintained that a preceding /j/ acted similarly to a following /j/, in words like yugo ‘joke’, junco ‘reed’ (from iŭgu, iŭncu)—an idea taken up, half a century later, by E. B. Williams, this time in reference to Portuguese (1938:§38, I.B.), with special application to chumbo ‘I lead’ ˂ plumbu and justo ‘just, fair’ ˂ iŭstu (this argument, however, collapses under its own weight, since Latinists now interpret the last-cited word as iūstus; see Ernout and Meillet’s authoritative dictionary, s.v. iūs [1959-60: 3292b]).

Singularly unhelpful was A. Zauner’s approach on several occasions; where the going was difficult, he invented Latin prototypes to fit them into his pigeonholes, dreaming up *gūstus ‘taste’ in lieu of gŭstus (1900:47); or expressed himself in an arcane, ambiguous way to gloss over complications—e.g. apropos conducho (1921 : § 13.4).

J. D. M. Ford (1911:xvii-xix), like Baist, reckoned with the possible effect on the vowel either of the preceding or of the following consonant; he specifically excluded from the range of the latter influences cl [i.e. c’l-], li̯, and ti̯—a gambit which allowed him to reconcile the word-final segment -ucho with the suffix -ojo ˂ -uc’lu. Ford, a staunch neogrammarian, was quite explicit in adhering to the zigzag schema in most instances; e.g., his trajectory for the line fŭgiō ‘I flee’ > Mod. Sp. huyo included, as the intermediate stage, V. Lat. *foyo; similarly, he theorized that pŭnetu, to reach the modern form punto, had to pass through the intermediate phase *pọñto (here he correctly observed that ñ before the dental was depalatalized); also, tŭrbidu, he argued, could not have reached the present-day stage of turbio without first becoming *tọrbi̯o (a more accurate notation might have been *tọrvi̯o). In a few cases, however, Ford, for some unstated reason, bypassed the o stage; thus, he allowed buitre to have evolved from vŭltŭre via *vultve; intercalated *trui̯ta, rather than *troi̯ta, between trŭeta and trucha; etc. The resulting confusion could not have been worse, exposing the absurdity of the “zigzag theory” if carried to an extreme.

During the long period of ascendancy of comparative “Romanistik” and of the Madrid School of Hispanic philology, it was a foregone conclusion that, minor variations of opinion apart, the “zigzag theory” represented a perfectly safe reconstruction of events. What contributed most to its solid entrenchment was the fact that, common-sensical as an (almost intuitive) opposition to it would have been, and effective as sharp disagreement with it might have been if persuasively presented, the few dissenting voices came from relative outsiders—maverick scholars removed from the mainstream of advances and singularly awkward, in addition to being unauthoritative, in stating their dissent.

The first outspoken dissenter was Edwin H. Tuttle, at that time connected with Yale who, in a crisply worded note on the Romanic vowel system (1914:347-53), after laying down his principle of visualizing three “periods” by linking them to languages and territories (“Sardic,” “Rumanian,” and a combination of Italian with Western Romance), categorically declared: “It is commonly but wrongly held that Italian [i.e. Tuscan] and Western Romanic represent the third-period vowel-system only. They show many traces of earlier conditions, and their history cannot be understood if we ignore this fact. The evidence of Sardic and Rumanian is not isolated: in the other Romance languages palatal influence often formed close i from Ĭ, and close u directly from Ŭ.” Read as a programmatic statement, these words have a promising enough ring. Unfortunately, the flaws in the execution of this self-imposed program turned out to be overwhelming. Even if one discounts numerous eccentricities of terminology and idiosyncrasies of phonetic notation, the very model with which Tuttle elected to operate seems bizarre in retrospect. Demographic shifts and the power of cultural (including glottal) diffusion were wholly absent from his schema, so that, when he assigned Ptg. cegonha ‘stork’ ˂ cicōnea to the “third period,” but cunha ‘wedge’ ˂ *cŭnea [in lieu of cŭneu ‘corner’] and junge ‘he joins’ ˂ iŭngit to recognizably earlier strata, no independent reason for this disparity became entirely clear at any point; as a matter of fact, there was an element of circularity in his reasoning.

Fifteen years later, P. Fouché came up with a real package of studies^rolled into a single monograph, titled—hazily enough—“Études de philologie hispanique” (1929: 1-171). Essentially, he combined very extensive reviews of the revised fifth edition (1925) of Menéndez Pidal’s Manual de gramática histórica, a classic, and of Meyer-Lübke’s highly controversial Das Katalanische: Seine Stellung zum Spanischen und Provenzalischen (of that same year) with a string of independent explorations, assigning pride of place to the relation of ō to ŭ. So far, so good; and one is indeed encouraged to read on as one stumbles, right at the outset, across this provocative statement: “On confond d’ordinaire l’évolution de ces deux voyelles, qu’il s’agisse d’ailleurs du français, du vieux provençal ou du castillan. A tort, car les résultats, s’ils concordent souvent, peuvent ne pas être les mêmes” (2). But in the end the reader’s expectations are not fulfilled; after briefly arousing them once more, apropos of Sp. puño ‘fist’ ˂ pŭgnu, muñe ‘he summons’ ˂ mŭngit ‘he blows the nose’, uñe ‘he yokes’ iŭngit, and ducho ‘skillful expert’ dŭctu [but others, I hasten to add, have favored the alternative etymon dŏctu], Fouché deftly polarizes the explicative possibilities (“L’u de ces formes provient-il de la fermeture d’un ancien ọ issu de u latin, ou au contraire continue-t-il un ŭ latin maintenu à cause de la palatale suivante?” [12]), then immediately lets his readers down by declaring: “A notre avis ni l’une ni l’autre de ces deux hypothèses n’est exacte.” Throughout, Fouché refrains from assigning to the forms he cites definite places along the coordinates of time and space; he even chides his countryman Georges Millardet for aiming at a chronological stratification. Fouché’s sole tool is his constant appeal to articulatory phonetics, which he applies, like a juggler, chiefly to undocumented intermediate stages, and there is no dearth of “howlers,” as when he apodictically declares rubio ‘blond’ [OSp. ruvio] a form “évidemment savante” (11). Small wonder Menéndez Pidal had no trouble brushing off rather cavalierly the objections raised by scholars as poorly equipped for jousting as were Tuttle and Fouché (also, before them, Cornu)—who otherwise might have been formidable adversaries.

It would have been a pleasure to be able to credit a major breakthrough to modernistically slanted linguistics. But does the record bear out such optimism? Take the revised third edition of E. Alarcos Llorach’s immensely successful Fonología española, which devotes considerable space (§§142-5) to the prehistory of Spanish vocalism (1961:206-19). The author briefly acquaints his readers with the Sardinian vowel system in synchronic and dia-chronic projection, but mainly as a foil to the alternative solution adopted by the “latín vulgar del Occidente” —a concept he adopts from the 1949 version of the Essai pour une histoire structurale du phonétisme français, by Haudricourt and Juilland, two authors who have had no second-hand, let alone first-hand, experience with Hispanic research. Hidden away, as inconspicuously as possible, close to the end of a footnote (213 rcl6), one finds cursory mention of the fact that in Basque borrowings from Latin ē and ĭ were still distinguished, with a reference to J. Caro Baroja’s monograph bearing on the history of Basque (1946:39-44). There is not a hint of the parallel, at least equally relevant, discrimination, in Basque, between ŭ and ō in Latin loanwords. Above all, a splendid opportunity has been passed up to voice the suspicion that we might here have a priceless clue to the vowel system presiding over the earliest Latinity of Spain.

Conceivably, Alarcos Llorach’s gaffe is atypical, as was before Haudricourt and Juilland’s hasty generalization.⁴⁰ A more sober, if less friendly, conclusion to be drawn from these facts would be that excessive schemati-zation, which underlies much of modernist research in linguistics, simply does not lend itself to the fine-tuning needed in the meticulous unravelling of historical clues.

What makes it desirable to reopen the discussion at this point, with entirely different chances of success, is the unforeseen coincidence of three events: (1) the discovery of a dozen spots scattered on the geographic map where Lat. ŭ has been preserved as u (Albania, Basque Provinces, North Africa, pre-Anglo-Saxon England, etc.); (2) the gradually accumulating evidence of the protracted preservation of intervocalic surds in Mozarabic, Valen-cian, Aragonese, etc. (see above), a finding which flatly contradicts all earlier tentative classifications of Hispano-Romance; and (3) the collection of examples, undertaken in this paper, purporting to illustrate the preservation of ŭ in contexts where neither a rash appeal to learnèd transmission would make sense, nor the agency of any particular phonetic environment (a preceding or a following palatal consonant, hiatus, and the like) can be plausibly suspected: Ptg. curto and surdo, OPtg. usso, Sp. hurto, Ptg. sulco beside Sp. surco, to cite but a few from among those already discussed at length. If the history of Lat. /p/, /t/, /k/ on Peninsular soil shows an older pattern of conservative transmission eroded through the invasion, or contagion, of a rival trans-Pyrenean system, more “advanced,” then a similar preponderant displacement of an indigenous Latinity by a more appealing or prestigious regional variety (probably traceable to Lyon and Toulouse) may be chiefly responsible for the blurring of the original, pristine diachronic contour ŭ > u, as familiar from Sardinian, Rumanian, etc. The theoretical acceptance of this possibility will force us to take a realistic revisionist attitude toward several contentions of transmission through learned channels and, above all, toward the—all too frequently invoked—”zigzag movement” ŭ > o > u.

The dialect mixture that ensued, and may have stretched from the third to the sixth century (or thereabouts), led, as was to be foreseen, to the constitution of small nuclei of words tied together by formal and/or semantic resemblances, in which either u or o was allowed to prevail; in some instances phonetic circumstances may have facilitated the choice. Hence the noted affinity of -ñ and -ch radicals, perhaps also of j- and y- radicals, for u, which has long, and deservedly, been in the focus of discussion, on account of its transparent phonetic implications; hence also, apparently devoid of such implications at first glance, the mutual attraction between nuclear u and -rt, -rd, -rc and similarly structured consonant clusters, as evidenced by Ptg. curto and surdo, by Sp. surco (Ptg. sulco), by OSp. Ptg. furto (Mod. Sp. hurto), and the like. Even here one cannot completely exclude the factor of phonetic reverberations: given the oft-demonstrated paradigmatic connection of l, r, and n in Hispano-Romance,⁴¹ one could convincingly argue that the commonness of -nt- (as in junto, punto) and in -nd- (as in segundo) could have sporadically entailed the prevalence of u before l and r followed by t, d, or some other obstruent.

IV

The problem not yet properly broached, except for a few anticipatory hints, is whether the vowel system of the oldest Latinity imported into the Peninsula—and still hazily recognizable despite subsequent overlays—in addition to keeping ŭ and ō apart also continued to distinguish between ĭ and ē; that is, whether it corresponded to the widely scattered “Archaic Type” perhaps best-known from Sardinia’s core dialects (merger of ŭ and ū, as well as of ĭ and ī), or to the developmentally somewhat more advanced, but territorially limited, transitional “Rumanian Type” (key words lup but verde). In a nutshell, the evidence so far marshalled is inconclusive, and any decision invites a much larger volume of preliminary fine-meshed sifting.

A few exploratory appraisals of individual, almost randomly selected case histories will highlight the degree of the difficulty involved. From the earliest texts extant Spanish copiously used the adjective firme ‘steady, hard, staunch’, which on two scores clashes with its reputed prototype fĭrmus, -a, very neatly reflected by Fr. ferme, Prov. ferm, It. fermo, etc.: Its i agrees, to be sure, with the vowel of Sard, firmu (REW₃§3320), but runs afoul of the consensus of all other cognates, exhibiting -e-; and the final vowels of the Latin base should, normally, have yielded -o/-a rather than -e. To explain away the latter complication, one could with relative impunity have recourse to one of several auxiliary conjectures, none of which, however, is truly satisfactory. Thus, one might argue, with J. Cornu (1884:289), that the adv. firmē, I suppose through frequent use in notarial texts, secondarily assumed the function of an adjective—except that the parental -ē adverbs show a singularly meager representation in the daughter languages (one such rare instance is Ptg. entregue ‘delivered, busy, absorbed’, from integrē), a fact long known to scholars, which Keith E. Karlsson’s 1980 dissertation has fully borne out. One superadded circumstance that might have accounted for erratic -u > -e would have been an indirect channel of transmission, namely through Gallo-Romance, except that Old French, Old Provençal (and even Basque, in a contiguous territory) in unison show here nuclear -e-, in preference to -i-. The third (and, to my knowledge, last) avenue of escape from the predicament is the assumption that colloquial Latin minted firmis alongside Class, firmus. Such a hypothesis, indeed, has much to recommend it; there exist precedents for the rise of such adjectival by-forms, moving in either direction from the standard, as when ācer, ācris ‘sour’ was gradually dislodged, in the vernaculars, by ācru, ācra (cf. OSp. Ptg. agro; see DELL₄, 6a). Above all, the form firmis actually occurs in an occasional text tinged by folk speech, e.g. the earliest Bible translation (Itala); if so, its coinage may have been stimulated by the préexistence of the near-synonym fortis or the antonym infirmis (Rönsch 1875:274), whose leadership, in this context, has in turn been attributed to the semantic vicinity of dēbilis ‘weak’, fragilis ‘frail’, lit. ‘breakable’, gracilis ‘thin, slim’, and the like. In any event, the adv. firmiter is on record in Classical Latin. But, once that much has been granted, how does one justify the survival, on Spanish-Portuguese soil, of enfermo ‘sick’ ˂ infirmu, with preservation of -nf- (in lieu of its alternative reduction to -ff-) best understood as pointing to the long survival of a morphemic boundary (cf. D. Catalân 1968:410-35)? Was there a time—say, ca. 500 A.D.—when adj. firme and *fermu coexisted in the Peninsula? Interestingly, traces of fermo have been detected in Old Portuguese (see J. Leite de Vasconcelos 1898:422 and unsigned review in Rom. 28.485 [1899]).

As one turns his attention from the ending to the word’s radical, the imbroglio tends to get worse. To be sure, one could charge the preservation of ĭ as a high front vowel to learnèd transmission and, in so doing, place heavy emphasis on the notarial meaning of the corresponding verb (Sp. Ptg. firmar ‘to sign’, firma ‘signature’ have an isolated counterpart in Old Italian, but the semantic development elsewhere was incomparably more lively and, above all, more concrete, as is shown most dramatically, but not exclusively, by Fr. fermer ‘to shut, close’ [Meyer-Lŭbke 1935:§§3318, 3320]? on the Latin prelude see V. J. Fahrenschon 1938). The Latinizing trend might have been reinforced by the “abstract” verbs affirmōre and confirmāre handed down at this level to Hispano-Romance (while French opposes pictorially saturated affermir to pale af-, con-firmer, and all three to en-, ren-fermer, modeled after fermer; Italian, as usual, opts for af-, con-fermare). But, assuming this high degree of learnèdness is plausible in reference to the Peninsula, why, one is left wondering, was firmitāte rejected in that territory—it could easily have been adopted as *firmidad, which would have formed a pendant to vernacular OFr. ferté, OProv. fermetat ‘fortress’—for the sake of innovative firmeza? Similarly, why have recourse to self-contradictory firmedumbre instead of *firmitud, from abundantly recorded firmitūdō? Clearly, the derivational suffixes -eza and -edumbre would match firm- far more snugly if only the i were not prejudged as a symptom of learned transmission. Even these stumbling blocks, however, do not exhaust the range of legitimate doubts. Sporadic epigraphic evidence (CIL 4.175, 6.1248, 5230), which the DÉLL₄ (237a), reports without endorsing it, shows that stone-engravers, in painstakingly recording FIRM-, had recourse to the “i longa.” But Ernout (1959-60) hastens to reject this bold interpretation of the word as fīrmu solely on the basis of (admittedly conflicting) Romance reflexes, an argument which smacks of circularity. There is a slim chance, then, that fĭrm- and fīrm- could have coexisted for a while, at least in certain territories, with incipient mutual semantic alienation. Alternatively, it could be maintained that firm- > Hisp. firm- represents one fairly isolated trace of the Archaic Vowel System, which at this point in time, space, and structure resisted the tide of the ĭ > e overlay by reason of (1) secondary support received from learned sources, and (2) special privileges accorded the following -rm- cluster (for details see below).

An equally intricate situation confronts the student of a whole cluster of derivational suffixes: -itia and -itiēs, appealed to for the formation of adjectival abstracts (hence functionally comparable to Eng. -th and -ness), as well as -ĭcius beside -īcius, used in minting relational adjectives, with special reference to the expanded, chain-like forms -ātīcius and -ārĭcius. There may be some advantage to starting out with the latter facet of the problem.

As has been established in searching monographs by generations of front-line Latinists (particularly by E. Wölfflin and Manu Leumann), speakers of the parent language had at their ready disposal the twin suffixes -īcius and -ĭcius, marked by inchoate dif ferentiation.⁴² As compound suffixes gained in popularity through fusion, the interfixai element -āt- was paired off with -īcius, while its counterpart -ār- was assigned to -icius, with the result that Old French displays a wealth of derivatives in -eiz (originally -ediz) and another contingent of, rather technical, qualifiers in -erez.⁴³ The older stages of Italian (its dialects included) show a virtually parallel split between (1) -aticcio, -atizzo and (2) -areccio or -ereccio, as has been convincingly demonstrated by G. Rōhlfs (1954:§§1038-9).

In Spanish and Portuguese -āriciu cast off -ariço/ -arizo (later -erizo), while -āticiu yielded -adiço/ -adizo. (To disencumber the complex problem, suffice it to observe that the shift -arizo > -erizo is due to contamination by -ero ˂ -āriu, as when cabrarizo. and cabero both meant ‘goatherd’, porcarizo and porquero were interchangeably used for ‘swineherd’; the local voicing of ç to z, which drove a wedge between the two languages at issue, constitutes a separate problem.⁴⁴) As one inspects such slivers of material as Ptg. achadiço ‘easy to find’, Sp. antojadizo ‘fickle, capricious’, encontradizo ‘bobbing up all the time’, tornadizo ‘changeable, renegade, turncoat’ (-izo having consistently been the medieval spelling), placing them alongside the above-mentioned slices -ariço/-erizo (hortaliza ‘vegetable’, dissimilated from -riza, may here be thrown in for good measure), one recognizes that Hispano-Romance has been steering a radically different course, generalizing the i at the expense of the e. From this divergent state of affairs the Hispanist is at liberty to draw three different preliminary conclusions, for one of which he will have to opt in the end:

1. In the provincial Latinity of Spain—i.e. at the temporal level of Antiquity— -īcius may have overpowered -ĭcius (in terms of morphological “economy,” or tightening of resources).
2. -iço/-izo could have displaced *-eço/*-ezo at the Proto-Hispanic stage, much as later -illo did -iello, -ito did -uelo, etc.—i having been one of the three favorites, in the company of a and u, among those vowels called upon to herald, or usher in, a derivational suffix, often to the detriment of -e-, -ie-, and -ue-.
3. The development ĭ > i may have been “regular,” representing a remnant of the Archaic Vowel System, protected from the tidal overlay (ĭ > e) by the vogue of i-dominated suffixes.

The situation is surpassed in sheer intricacy by the development of -itia beside -itiēs in Hispano-Romance adjectival abstracts.⁴⁵ At first glance it seems insufficient to state that -ez (as in niñez ‘childhood’, vejez ‘old age’, borrachez in rivalry with -era ‘drunkenness’) reflects -ǐtie; that -eza (as in nobleza ‘nobility’, pobreza ‘poverty’, and riqueza ‘wealth’) perpetuates -ǐtia; that the replacement of -tāte (as in vetustāte, nobilitāte, paupertāte) is a trivial matter of suffix change; that -icia (as in malicia ‘evil, trickiness’, OSp. trist-icia ∿ -eza ‘sadness’) testifies to learnèd transmission, as does post-medieval -icie (calvicie ‘baldness’, planicie ‘plateau’). Upon closer examination, however, things turn out to be less simple and straightforward. To begin with, Old Galician-Portuguese offers two mutually competing equivalents of Sp. -ez, namely -ece and -ice (velh-ece, -ice ‘old age’), of which the second won out. Then again, the counterpart of Sp. -icia is Ptg. -iça, of distinctly less learnèd appearance, especially in lexical items like maiça ‘malice’, which visibly participated in characteristic sound changes, such as the loss of intervocalic -l-. Semantically, the words in -ice are at the opposite pole from any display of refinement and pretentiousness, referring as they do not only to age levels (e.g. meninice ‘childhood’), but also to a profusion of patterns of comic or reprehensible behavior.

Under these circumstances, should Ptg. -ez be written off as an intruder from Spanish? Should -ice and -iça be recognized as vernacular descendants of -ĭtie and -ĭtia, respectively—merely protected from the “overlay” by a Latinizing tendency (whereas -ícia and ície, traceable to modern times, could then pass off as genuine “cultismos,” suggested by Latin models and precedents in neighboring Castile)? Were -ece and -ice once regionally separated? And is -eza essentially a Provençalism, with Old Spanish pressure acting as a reinforcement?⁴⁶

Problems of this type have infrequently come up for discussion in the past and, if so, certainly not within the context here chosen. The Establishment has recognized rather numerous instances of Class. ĭ as i (as against the norm, exemplified by pĭlu ‘hair’ > pelo, pĭra ‘pears’ > pera ‘pear’, ĭnter ‘between, among’ > entre) and has assigned them to certain categories. After subtraction of transparently learnèd words, too trivial for their taste to invite sustained attention (e.g. libro ‘book’ ˂ lĭbru), spokesmen for this approach discovered a residue, traceable to the effects of metaphony, which Menéndez Pidal, in his revised handbook (1941:§11), subdivided as follows :

Since once more, a zigzag movement is implicitly posited here, it stands to reason that vindēmia ‘vintage’ > Sp. vendimia, ec(c)lēsia ‘church’ (lit. ‘assembly’) > OSp. eglisa (e)grija, vēnī ‘I came’ > vine, Iber. Garsea (anthrop.) > Garoía should have been conjoined with the above cases. As a matter of fact, the author threw in for good measure, without forewarning, even a few instances of /ɛ/ > /ẹ/ > /i/, as with tĕpidu ‘lukewarm’ > Leon, tebio > Sp. tibio, or mĕa ‘my, mine’ (f. sg.) > */mea/ > mía. These are, of course, examples of a double gambit.

Despite the ingeniousness he displayed throughout, Menéndez Pidal ran into serious difficulties with the method he adopted. Unable to reconcile mĭliu ‘millet’ > mijo with (pl.) cilia ‘eyelids’ > (sg.) ceja ‘eyelid’, he reconstructed [*] mīliu, appealing to a lexical blend with mīl[l]e, [pl. mīlia] ‘a thousand’; the incompatibility of (pl.) sĭgna ‘signs’ > (sg.) seña ‘mark, token, watchword’ and tĭnea ‘gnawing worm’ > tiña ‘ringworm, beehive spider’ prompted him to toy with [*] tīnea. For mancipiu ‘property’, ‘slave, servant’ (lit. ‘taking by hand’), which clashed with sēpia ‘cuttle-fish’ > jibia, he substituted *mancipu as the alleged base of mancebo ‘lad, youth, adolescent’, oblivious of the fact that the reconstructed form, inescapably stressed on the a, would have produced an utterly different descendant. As a result, the entire paragraph gives the impression of atypical untidiness.⁴⁷ In a footnote, there is a fleeting, evasively polite reference to E. H. Tuttle’s aformentioned 1914 note, but from that hint the inexperienced reader cannot possibly guess that the American pioneer, far from upholding the Spanish scholar’s ideas, was, as in the case of ŭ, diametrically opposed to them.

Similar breakdowns can be culled from numerous textbooks and reference books, starting with G. Baist’s sketch of Spanish historical phonology (1888); as a matter of fact, Menéndez Pidal’s analysis may, on balance, emerge as the most forceful presentation of an inherently vulnerable approach. If carried to its extreme, the “zigzag theory” may border on the grotesque. Suppose one takes seriously the standard formulations of post-Classical metaphony by word-prefinal /w/; it would follow that, lĭngua ‘tongue’ (Stage 1) became /lengwa/ in (Common?) Vulgar Latin (Stage 2); reverted to /lingwa/ in Proto-Spanish through the agency /w/, except for the loss of distinctive vowel quantity (Stage 3), judging from Ast. llingua and Gal.-Ptg. língua; with the pendulum swinging back to lengua (Stage 4) at an undisclosed juncture, an assumption one is virtually forced to make to justify the form familiar from Modern Spanish? Or is the new lengua essentially identical with Stage 2 /lẹngwa/, coming from a territory that had throughout remained sheltered from the effects of metaphony (Stage 3)? If so, how can that territory be geographically circumscribed, and what was the reason for its impact, in this instance, on Standard Spanish? Sp. mengua ‘dearth’ beside Ptg. míngua obviously invite a parallel scrutiny, except that in their case no comparably reliable starting point is available to today’s analyst (the Classical verb having been, I repeat, mĭnuere rather than *mĭnuāre).

But once we are compelled to operate with such assumptions as dialect cleavage and eventual dialect mixture and reconciliation within the traditional framework, why not go one step farther and admit, at least on the theoretical plane, that we might be facing a situation familiar to us from earlier concern with the vicissitudes of the ŭ; Colloquial Hispano-Latin may initially have maintained the Classical distinction between ĭ and ē, capping it with the new tendency to merge instead ĭ and 5; but an overlay of dialect speech affected by the newly prominent Lyon-Toulouse vowel system would have blurred the picture—indeed, far more radically so than with ŭ, perhaps as a result of the superadded power of metaphony.

For the time being, then, the ĭ - ē issue must remain in abeyance. When the debate is reopened someday, it will be useful to take into account not only syntagmatic, but also paradigmatic conditioning, which has so far been consistently neglected. Thus, if -nt- (not unlike the situation in Tuscan) turns out to have had, somewhere along the line, a raising effect on the preceding vowels e and o (or, alternatively, if its contiguity has helped to preserve otherwise endangered i and u), it will be legitimate to wonder whether—given the general affinity of n with l and r in Hispano-Romance, lt and rt could not have acted similarly. For lt a positive answer has already been provided, by implication: punta ‘sharp end, tip’ and yunta ‘yoke’ would simply be bracketed non-randomly with mucho ‘much’ mŭltu and purches ’porridge, pap’ pŭltēs under a common denominator. As regards the front vowels, Ptg. Sp. pinga ‘it drips, leaks’ ˂ *pĕndicat might be allied with (so far unexplained) pírtiga ‘pole, rod, staff’ (OSp. piértega) ˂ pertica and even with Ptg. irto ‘stiff’ ˂ *ĕrctu, in lieu of ērectu (cf. Sp. yerto); indeed, elusive Ptg. irmão ‘brother’ ˂ germānu (beside OSp. yer-, er-mano, Mod. hermano) might begin to fall into place.⁴⁸

V

In addition to the three situations here examined in considerable detail, there possibly exist as many as half a dozen others in Hispanic phonology that give the impression of inviting a similar analysis. These additional contexts all involve contradictions well-known to scholars but so far left unsolved or, at least, not fully resolved; the margin of doubt would, one feels, recede if one were to allow for the superimposition of one wave of Latinity (akin to, or connected with, Gallo-Romance) upon an earlier wave traceable to the late Republican (or call it Gracchan, or Punic Wars) era.

Easily the most thoroughly, if still insufficiently, investigated of these additional complications bears on the transmission of the syllable-initial (in most instances word-initial clusters) /kl/, /pl/, /fl/ of the parent-language, with some attention to /bl/ and /gl/.⁴⁹By briefly disregarding certain crosscurrents that muddied the emerging picture, one can conveniently start from the classic opposition between the Spanish and the Portuguese outcomes, ll /λ/ as against ch- /š/ traceable to older /č/ (dialectally preserved to this day). The standard examples are clāve ‘key’ > Sp. llave vs. Ptg. chave; flamma ‘flame’ > Sp. llama vs. Ptg. chama; plōrare ‘to cry aloud, wail’ > Sp. llorar, Ptg. chorar. The correspondences are not always that neat; plumbu ‘lead’ yielded chumbo in the West, but “semilearnèd” plomo in the Center. The vernacular development is sometimes observable only in toponymy: witness Sp. Lloredo, Ptg. Choredo ˂ flōre ‘flower’ + -ētu (plantation, grove, garden, and orchard suffix), as against Sp. Ptg. flor, a learnèd form appropriate to a generic label (cf. animal, bestia).

The relation of /č/ to /λ/, though seemingly easy to describe in geographic terms, fails to lend itself to any smooth interpretation on other levels of analysis. The difficulty is exacerbated by certain apparently isolated word histories. Thus, Sp. choza ‘hut, cabin, lodge’—a term of primitive pastoral culture—seems to involve the Western, rather than the expected Central, sound development if it reflects the semantically self-evident base (pl.) plŭtea ‘sheds, parapets, penthouses’, which satisfies other phonetic requirements. Chopo ‘black poplar’ can be persuasively traced to pōpulus via the reconstruction *plōppus—admittedly bold, but very effectively supported by Tusc. pioppo [pjop:o]; the unwelcome vicinity of populus ‘people’ could have triggered the chain of events conducive to *plōppus in the first place. OPtg. chus ‘more’ plūs causes little surprise, but the emergence of that same form in far-off La Rioja, where a sub-dialect of Navarrese (i.e. of an Eastern Spanish dialect), on the testimony of that early thirteenth century poet, Gonzalo de Berceo, is noteworthy; apropos these bits of evidence observe that plūs, in the Peninsula, was a relic word doomed to extinction, as a result of the prevalence of its virtual synonym magis (> Sp. más, Ptg. mais). Most important of all, where cl-, etc. were syllable-, but not word-initial, Spanish, on a par with Old Portuguese, displays /č/ spelled ch. Thus, Sp. conchabar ‘to join, unite, hire’ (surrounded by several satellites: conchabanza ‘comfort, [coll.] ganging-up’; conchabo ‘hiring, work’, conchabero ‘pieceworker’, a few of them confined to dialect speech) snugly fits conclāvāre ‘to nail together’, marginally recorded in Graceo-Latin glosses.⁵⁰ Add to this example such familiar illustrations as Sp. Ptg. ancho ‘broad, wide’ ˂ amplis ‘spacious’, OSp. finchar (> hinchar), Ptg. inchar ‘too well’ inflāre and OSp. fenchir (> henchir), Ptg. encher ‘to fill, stuff’ ˂ implēre. There exist, I repeat, a number of troublesome side developments, as when Sp. chubazo ‘shower’ is best understood as a “Western” word, closer to Ptg. chuva ˂ chuiva than to Sp. lluvia (humidity comes from the Atlantic coast!); or as when near-homonyms can be more sharply distinguished by discriminatory treatment of the characteristic consonant group (Sp. clavo ‘nail’ ˂ clāvu is thus protected against confusion with llave ‘nail’ clāve, while clavel ‘carnation’ looks like an imported Provençalism); or as when, through dissimilation of palatal consonants or consonant clusters, playa ‘beach’ was preferred to *llaya and flojo ˂ OSp. floxo /flos̆o/ ‘loose, slack’ flŭxu ‘flow’ was favored over *lloxo /λos̆o/. But, if one disregards such minutiae, the general impression that one gains is that /č/ as an outcome of /kl/, etc. represents an old, autochthonous development in Spain and Portugal alike, faithfully preserved (gradual deaf-frication apart) in the latter territory, whereas in Spain, with the exception of certain noteworthy relics like choza, chopo, obs. chus, and conchabar, ancho, hinchar, henchir, and with the further exception of instances where, under learnèd or phonetic pressure, the original consonant clusters were restored, a novel mode of pronunciation, namely /λ/, was introduced. Philological and dialectological evidence shows that this /λ/ was preceded by /k/, /f/, and /p/, converging dyadic clusters from which the opening component was peeled off, at the preliterary stage. These clusters readily lend themselves to interpretation as palatalized trans-Pyrenean prongs of Occ. cl-, fl-, and pl-.

One situation of extraordinary difficulty in Spanish—in fact, known for bristling with contradictions not yet accounted for—is the development of word-medial l before consonants. The intricacy, to be sure, has its roots in Latin, where the preceding vowel would color the pronunciation of the lateral; but in several Romance languages—preeminently in Florentine Tuscan—standard /l/ was before long generalized through all sorts of leveling and tightening processes. Elsewhere ancestral l, in this position, would tend to become r.

Old Spanish shows the familiar split between (1) preservation of the l and (2) its very early (preliterary) transmutation into /w/, with all sorts of consequences flowing from this initial gambit (mainly, crystallization of a falling diphthong through contact with the preceding vowel and eventual fusion of that diphthong into a monophthong in Spanish, if not in Portuguese). The -ǔlt- segment stands apart, having yielded -uit- (e.g. buytre ‘vulture’, muyt ‘much’ > Mod. muy beside mucho) . Except for this minor sideline there appears to be no way of isolating the syntagmatic factor that might have presided over the individual choices, and certain word families were actually torn apart by conflicting developments of their members, not to mention the occasional rise of doublets. The traditional explanation, transparently lame, involves an appeal to lexical stratigraphy: It has been argued that, aside from the consistent survival of l in learnèd words, it was also maintained in careful, formal discourse in the vernacular (“upper layer,” “Oberschicht”), but was allowed to become a semivowel and eventually to disappear in low-level parlance (highly informal or rural). This attempt to have recourse to stylistic or social registers as prime determinants is visibly forced in most instances; to what layer should one credit a “neutral” word like olmo ˂ ŭlmu ‘elm tree’? Why does pulvus ‘dust, powder’ survive as polvo(s) in Spanish, but as pó in Portuguese? Another dimension of the problem is the ever-present possibility of the I having been secondarily reintroduced. Thus, Sp. dulce ‘sweet’ can be misleading, if inferences from it were to be made to early evolutionary stages: Ptg. doce, OSp. doc(e), duc(e)—which have already come up for cursory mention in a different context—clearly show the real sequence of events.

If only we knew for sure how the /l/ was pronounced in the Lyon-Toulouse area between, say, the third and the sixth century, then it might, under certain circumstances, be permissible to appeal once more to the hypothesis of a confluence of two currents of Latinity in the Iberian Peninsula in late Antiquity. Unfortunately, our sole direct evidence for the provincial Latin of Southern France is the corpus of Old Provençal texts, none of them predating the tenth century; that is to say, we have no reliable, immediate insight into goings-on between the disintegration of Latin and the jelling of a newly constituted local vernacular. And that vernacular is of, practically, no help to the Hispanist, because here he encounters essentially the same dichotomy as in the Peninsula, especially before t, d, n, and s. To quote such a seasoned expert as O. Schultz-Gora: “... kann es bestehen, es kann sich jedoch ... erweichen und tut es hōufig” (1911: § 96). For fine-tuning through deliberate appeal to an assortment of qualifying adverbs (‘stets’, ‘fast stets’, ‘seltener’, ‘kaum je’, and the like) one can profitably consult the parallel, more circumstantial analysis supplied by C. Appel (1918:79). The one feature distinguishing Old Provençal from its cognates is the abundance and diversity of falling diphthongs thus arrived at: au, ou, eu, iu. Significantly, the sheer number of doublets is staggering: aut ∿ alt ‘high, tall’, ausar ∿ alsar ‘to raise, lift’, beutat ∿ beltat ‘beauty’, coutel ∿ coltel ‘knife’, piusela ∿ pulsela ‘maiden’, viutat ∿ viltat ‘villainy’, etc.

Nothing of that sort is familiar to the Hispanist from his own corpus. He may be baffled by the cooccurrence of alto, altear ‘to rise, stand out’, on the one hand, and, on the other, ot-ero ‘hillock, knoll’, ot-ear ‘to survey, look down upon, watch, keep an eye on’ (via *out-), but does not expect to stumble over either ⁎outoor ⁎alt ero. The most cogent reason for the extremely large quota of unpatterned wavering observable in tenth to twelfth century Provençal texts would be the sheer recency of the tendential vocalization of I before consonant; on that assumption, there simply would not have been a sufficient number of generations of speakers to segregate the scores of chaotic doublets, semantically still undifferentiated. If this tendency indeed began to operate as late as the seventh, eighth, or ninth century (i.e. during the preliterary period), one can extrapolate for the preceding centuries, i.e. for the Lyon-Toulouse Latinity at its peak, the solid preservation of l before consonant: alt, beltat, etc.

One could then argue that this original pronunciation, inferrable but undocumented, invaded, or percolated into the Iberian Peninsula toward the close of Antiquity, in alliance with (p)ll- ˂ pl- (etc.), o ˂ ŭ, and the other phenomena already examined. In a few ecclesiastic words, the “firming” could have been reinforced by parallel pressure culturally enforced, as was probably the case with altar ‘altar’.

The last major point to be summarily examined, against the possibility of yield of fresh information on the two Waves of Latinity, is the compression of the structure of tri-and polysyllabic words, in regard to the loss of weakly stressed vowels. The two extremes on the scale of possibilities are (1) Sardinian, South Italian, and Rumanian which—certain narrowly circumscribed special cases apart—as a rule do not syncopate such extravulnerable vowels; and (2) Northern French, which already at its medieval stage and, a fortiori, during subsequent phases, energetically lopped off such vowels, typically to the exclusion of staunchly resistant /a/. The most characteristic positions affected by this trend were (a) the intertonic, antepenultimate vowel in tetrasyllabic words, ; and (β) the penultimate vowel in proparoxytonic words (comprising three or more syllables), . Closely allied situations arose in the case of unstressed and word-final vowels, in lexical units bracketing two or more syllables (“apocope”); and in word-initial vowels heralding words of virtually any length (“apheresis”). Although sporadic attempts have been made to study, say, apheresis or apocope in isolation, probings of this sort, however sophisticated their execution, have at all times been doomed to produce severely limited results, since it is precisely the interplay of such tendential compressions that promises to yield the sought-for pattern.

By way of preliminary remark, a cautionary statement on Portuguese will prove useful: Through radical erosion of its inherited consonant structure (loss of Lat. d, l, and n between vowels) plus certain side effects of nasalization, not to mention the familiar “blurring” or “skipping” of weakly stressed vocalic peaks in Lisbonese, Portuguese at present gives the general impression of a heavily eroded language, abounding in words marked by radical syllabic contraction: Contrast Ptg. nu ‘bare ’(˂ nūdu) with Sp. des-nudo, Ptg. vir ‘to come’ (˂ venīre) with Sp. venir, Ptg. rir ‘to laugh’ (rīdēre) with Sp. reír, Ptg. réis ˂ rēgāles, lit. ‘royal’ (as in milreis, Brazilian money unit) with Sp. reales. But this characteristic image must not be projected onto the level of Old Portuguese, still less on the plateau of Proto-Portuguese, where an utterly different ensemble of conditions prevailed. Just what the pristine state of affairs may have been in the West is best disclosed by such “undisturbed” vestiges as amêndoa ‘almond’ (Gr.-Lat. amygd-ala, -ula) beside Sp. almendra, or táboa ‘board’ (˂ tabula) beside Sp. tabla, or âncora ‘anchor’ (˂ ancora) beside Sp. ancla, or OPtg. côvedo, -ado ‘elbow’ (cubitu), later replaced by cotovelo, alongside OSp. cobdo ˃ codo. With respect to the word-final vowel in certain environments one also notices the split between the West, which has preserved -e on a generous scale (witness bondade ‘goodness’ ˂ bonitāte, cidade ‘city, town’ ˂ cīvitāte ‘citizenry, citizenship’, as against Sp. bondad, ciudad, orig. cibdad; and árvore ‘tree’ ˂ arbore, as against Sp. árbol, OSp. árvol ; also cárcere ‘jail, prison ’ ˂ carcere, as against Sp. cárcel) . Finally, from Lat. pignerārī, *-orārī ‘to take as a pawn’ there branched off penhorar ‘to pawn, pledge’ in the West, but OSp. pendrar ˃ Mod. prendar in the Center. (True, this is contradicted by Ptg. honrar ‘to honor’ ˂ honōrāre—could it be an adaptation of OSp. onrrar, ondrar?)

In the tenth century glosses traceable to the Navar-rese area, i.e. reflecting folk speech to the northeast of Castile, one finds such forms as babtizare, bebetura (flanking the graphy ueuetura), collitura, desposatos, duplicaot, nafregata, side by side with “straight” Latinisms (benedictione, debiles, junctatione, manducaret, nominatus, rapinaret, salutare), whose evidential value may be distinctly smaller, since they seem to involve incrustations of bits of Church Latin. By confining oneself, then, to genuine specimens of vernacular speech, one recognizes a degree of faithful preservation of ancestral syllabic structure reminiscent of Proto-Portuguese and, even more so, of Sardinian, South Italian, and Rumanian. These, one begins to suspect at this juncture, must or could all be representatives of the First Wave of Latinity.

Over against Proto-Portuguese and Archaic Navarrese stands Old Castilian, a dialect marked by very bold omissions of weakly stressed vowels and, as a result, by the rise of innovative dyadic and triadic consonant clusters alien to the parent language, and, not infrequently, displaying an assortment of buffer consonants: Low Lat. carricāre, lit. ‘to transport by cart’, yielded cargar in the Center (as against Ptg. carregar) ; recitāre ‘to recite’ (prayers) led to rezar ‘to pray’ via rezdar; *īlicīna (an elaboration on Class. īlex, -icis ‘oak tree’) gave rise to Ilzina, later Enzina, etc.; dēbile ‘weak’ led to deble (preserved in en-deble, its prefixal expansion due to coexistence with enfermo ‘sick’). The issue before us is whether such syllabic contractions, coincident with vocalic syncopes, involve a purely local process or testify to an overlay whose ultimate focus can be traced to Southern France.

The word-FINAL segments of Old Provençal lexical units are, of course, not at all comparable to what one expects to find in Old Spanish even at the height of the predominance of apocope, on account of (1) the very heavy loss of final vowels in Provençal (except for -a, and for -eused as a supportive minimum vowel) and (2) the simplification of consonant clusters secondarily placed in this position (e.g. -mentu > -ment > -men), so that the separate Latin words for ‘world’ and ‘mount(ain)’ tend to collide in the language of the troubadours (mon1 and mon2), whereas Medieval Spanish sharply discriminates between the two (mundo vs. mon-t, -te; Mon- only as an ingredient entering into orographic or toponymie compounds). But word -MEDIALLY Old Provençal, while differing from Old Spanish in untold details, exhibits a picture strikingly similar to that of its congener: ‘to accuse, defy, challenge’, for example, was reptar in both languages, derived from parental reputāre: the title ‘count’ (lit. ‘ [the king’s] companion’) was comte in Old Provençal and cuende or conde in Old Spanish; the finite form ‘he sows’, from Lat. semĭnat (culled from the paradigm of an -āre verb), would be semna in Old Provençal and siemna, eventually siembra in Spanish, as against Ptg. semeia; for ‘tame’ one finds duendo in Spanish (also, correspondingly, dondo to the west of Castile) and domde to the north of the Pyrenees; for ‘elbow’, OProv. cobde looks like a good match for OSp. cobdo, from cŭbitu, while OPtg. côvedo went off in a different direction; for ‘head, leader’, a word based on capitellu (a diminutive, in turn, of caput),one encounters OProv. capdel, which agrees in its treatment of the nuclear segment with OSp. cabdiello (> Mod. caudillo); membrar, an impersonal verb for ‘to remember’ in the Cid epic—a text which also used the epithet membrado ‘clever, judicious’, lit. ‘one endowed with a good memory’—reappears, again with an impersonal construction in Old Provençal, the common sources of the congeners being memorāre. To be sure, the lexical isoglosses of the two languages do not consistently coincide, and in contexts where borrowings occurred they did not necessarily take place along lines inviting comparison. Thus, for ‘folly, madness’ Old Provençal writers used either follia or foldat (foudat), and it is, through a caprice of circumstances, the former that penetrated into Spanish, much as its Old French equivalent (folie) was absorbed by English (folly); but had the rival form foldat infiltrated into Spanish, it would not have been cacophonous in its new environment, judging from such local abstracts as maldat ‘wickedness’ and even fealdad, initially ‘fealty, loyalty’ (later, through secondary association with feo, ‘ugliness’) and frialdad ‘coldness, coolness’. Only under special sets of circumstances did the two languages in question part company on this score, as when ‘master’ yielded dompne in Provençal, yet dueño in Spanish, from dŏminu as the common source.

Given this degree of affinity, which could be effortlessly documented with many more equally telling examples, we have before us the classic dilemma of diachrony: Should we opt for monogenesis or polygenesis? If the hypothesis of monogenesis and subsequent diffusion should win out in the future, then the direction of the spread, on almost crushing circumstantial evidence, could only have been in the direction from Southern France to the north of the Iberian Peninsula along routes that flanked the Navarrese area, which could thus foreseeably turn into a conservative zone (witness the late tenth century glosses), along with far-off Portugal. But the basic dilemma, pending more thorough inquiry, for a while must remain in abeyance.

Of the various points of diachronic phonology that have been here in part examined under a powerful lens, in part merely sketched out, not all have yielded equally clear, unambiguous information. But even if it has been established in as few as three cases—e.g. in reference to the widespread survival of the voiceless intervocalic occlusives, of the Latin ŭ preserved as u rather than o, and of /č/ as a strongly represented alternative to the more familiar /λ/ as product of /kl/, /pl/, and /fl/— that we have before us, in each instance, two competing developments of a comparable degree of regularity (or, if you wish, legitimacy) rather than a single noun surrounded by a suspiciously high number of “exceptions”, then the need has arisen for pausing and reflecting about the theoretical basis underlying our analyses.

Without engaging in unwarranted, premature generalizations, we can simply state that the family-tree model, whatever the wisdom of its use in other contexts, simply does not apply to the Iberian Peninsula, where no branchlike break-off from a common trunk has ever occurred. Neither does it make much sense to operate with a kaleidoscopic sequence of separate areas for each successive change, positing ever new focal points, by joining the extreme wing of diffusionists. It suffices to posit a single displacement of the major center of innovations, a center which can and ideally should be determined, and defended against possible criticism, by purely linguistic analysis, but also one which, preferably, falls into the broader patterns of general history, including its cultural facets. While it is true that, under certain dramatic circumstances, the forces of general history and those propelling language change may be at loggerheads, or totally independent of each other, it is equally true (and not necessarily “unexciting”) that philological evidence and historical records, in the majority of cases, tend to support each other, as of course the past century knew all along—in fact, took for granted.

It is a fact that, down to the second century A.D., while the Republic and the early Empire (the prineipatus) were steadily on the upswing, Rome City acted as the indisputable center of an enormous, close-knit Western and Central Mediterranean unit—administrative, military, political, economic, cultural, and demographic—keeping under tight control the divergent idiosyncrasies of the individual provinces. Because of the relatively early date of its conquest and settlement; because of its comparative distance from the metropolis; and, not least, in harmony with a certain intrinsic conservatism observable over a period of two millennia, the oldest layer of Hispano-Romance (“First Wave”) can very well have adopted an archaic vowel system and, for a while, preserved intact its intervocalic surds, to cite just two critical features. The much closer cohesion of the Iberian Peninsula with Southern Gaul, starting with the third century and coincident with the cultural hegemony of Lyon, was reinforced by the gradual retrenchment of communication lanes with the Empire’s rapidly aging and decaying historic capital. This secondary rapprochement with Southern Gaul laid Spain and Portugal open to new influences (“Second Wave”), which, however, did not completely overlay the older structure, leading instead to all sorts of minor compromises, on a tactical rather than strategic scale. The closer proximity to Gaul of Spain rather than of the Atlantic Coast may have codetermined the survival of /č/ in Galician-Portuguese, whereas speakers of Proto-Spanish lost no time in switching to /kλ/, etc., and eventually to /λ/—in word-initial, but, significantly, not in medial syllable-initial position, a niche which afforded far more effective protection of the original state of affairs.

Of course, there were at all times other forces at work, in addition to this secondary rapprochement alluded to; the erosion of certain intervocalic consonants in the extreme West is just one case in point. Such concomitants easily influenced the local reactions to the influx of Lyonnais Latinity, codetermining the areas of speakers’ surrender or resistance to it.

In hastening to credit Hispano-Romance to a Western variety of provincial Latin, earlier generations of scholars were unduly impressed, first, by gross geographic conditions; and, second, by such a conspicuous feature as the uniform preservation of -s throughout the Middle Ages (and, for a while, also of -t). Viewed in the perspective here advocated, this temporary survival of certain final consonants, through articulatory “firming,” is also attributable to the secondary sharing of certain features by Gallo-and Hispano-Romance (under the leadership of the former), except that this time a process of preservation, rather than of innovation, is involved.

NOTES

*The opening sections of this paper were initially presented at the Houston Convention (December 1980) of the Modern Language Association. Other segments were consolidated into the text of a semiformal lecture (“Old and new ideas about the classification of the Romance languages”), which was offered—from March through May 1981—at a variety of campuses: Texas (Austin), Illinois (Urbana-Champaign), Northwestern, Brown, Harvard, Oxford, and Liverpool; in most of these places it elicited a stimulating, if short, discussion. I am grateful to the organizers of those meetings and to the discussants; also to Giulio C. Lepschy (London and Reading), who—independently—read the typescript and commented on it; and to Nicoline Ambrose, a young linguistic scholar in her own right, who lent the first typographic interpretation to this piece.

1.One scholar particularly concerned with this side issue was the Italian Slavist R. Poggioli, who during his Harvard years became an original theorist of literature.

2.For a succinct presentation of archaic family trees, in reference to what later became known as Indo-European, or Indo-Germanic, one may turn to George J. Metcalf’s essay, “The Indo-European hypothesis in the sixteenth and seventeenth centuries” (in Hymes 1974:233-57). We owe a slightly more detailed, painstakingly accurate picture to Anna Morpurgo-Davies.

3.Schmidt’s slender volume, of 1872 vintage, initially produced a light impact on the progress of scholarship. Approximately a half-century later, H. Pedersen, in his incisive appraisal of that pamphlet (see the English trans., 1931:314-8), remarked that the discovery and decipherment of, above all, Tokharian and, secondarily, Hittite, gave a fresh impetus to Schmidt’s pioneering ideas about “dialect waves.” In examining the division of the Italo-Celtic, Hittite, and Tokharian group of languages into a western, middle (the one most subject to change), and eastern section, Pedersen was reminded of the similarly patterned division, in Antiquity, of the Southern Greek dialect group.

4.The pitting of the family-tree projection against the model of the wave theory has gradually become a hallmark of all comprehensive handbooks of linguistics and especially of all introductions to historical linguistics. This preoccupation was not yet clearly delineated in the classic syntheses by Sapir (1921; see 1949:205 ff.), who took up the issue of morphological diffusion without any explicit reference to its limited compatibility with the configuration of family trees; and by Vendryes (1921:349-66; but the writing was actually concluded by 1914), who invoked only a single comparative method. The picture changed radically with L. Bloomfield’s thoroughly revised textbook (1933:316-8). The transition to the modern era was marked by the appearance of W. P. Lehmann’s textbook (1962: 138-42). L. Hjelmslev, as often, stood completely apart in his Sproget (1963: written ca. 1943) in deliberately disregarding or at least deemphasizing the wave theory.

5.Thus, Bloomfield’s surprisingly evenhanded approach may have its ultimate roots in his inability either to reconcile, in a bold synthesis, the rigid tenets of the neogrammatical approach with the more flexible implications of dialect geography and of diffusionism, or resolutely to rule out the one in favor of the other, as I have tried to demonstrate on another occasion (1967: 137-49).

6.Testifying to such a conciliatory perspective and to a commonsensical attitude was, among other pronouncements, the short paper by E. Pulgram (1953: 67-72). For a bird’s-eye view of the entire discussion see my own earlier attempt at a preliminary synthesis, with special reference to Romance (1978:467-500).

7.See (1866:1.76-103), esp. 76, 79, 82, and 94; also (1868: 3.27-57). Nevertheless, it took Schuchardt several years to yank himself completely loose from the straitjacket of neogrammatical doctrine; see my short comment to this effect (1980:93 ff.).

8.For the Romance side see the article identified toward the end of note 6. One guide to the Germanic side of the issue is W. P. Lehmann’s contribution (“The grouping of the Germanic languages”) to a miscellany (Ancient Indo-European dialects) relevant in its entirety to the question here discussed (see Birnbaum and Puhvel 1966:13-27).

9.The best known and most technical presentations of W. von Wartburg’s “theory” (to use the latter tag very loosely) are traceable to the years 1936 and 1950. There is no need here to refer to a profusion of satellite studies from his pen. The “theory” has provoked a deluge of criticism, including two qualifications or disagreements of my own (1972:863-8, and 1976b: 1.27-47).

10.Interestingly, although Lausberg’s thinking about such problems reached its peak in the mid-fifties, it goes back in a straight line to his brilliant doctoral dissertation on Lucanian dialects (1939).

11.One may go even one step farther and assert that Lausberg’s ideas on the configurations of vowel systems, as expressed in articles and other writings, left by far the strongest impact. See, e.g. the enthusiastic reactions of A. Kuhn, characteristic of the climate of opinion thirty years ago (1951:54, 86, 96 ff., etc.). Even though Lausberg himself at present (i.e. in retrospect) feels that his volume on inflection represents his single weightiest contribution within his far-flung magnum opus (personal communication, April 1981), the fact remains that he has never mustered the strength to attack, within the domain of morphology, the adjacent provinces of derivation and composition, let alone the territory of syntax.

12.The appearance of Rohlfs’ monograph on Gascon (1935; rev. 1970) at once placed the author in a strategically very advantageous position: It not only afforded him instantaneous insight, as the subtitle programmatically announced, into problems left pending of Pyrenean linguistics, but also allowed him to counterbalance his already established expertise in South Italian with a comparable command of Gallo- and Hispano-Romance. The 1935 book was preceded by a string of articles on Basco-Latin and Pyrenean themes.

13.Saroihandy briefly noted the. phenomenon in his original 1898 and 1901 research fellowship reports, then studied it in more searching detail in his cartographically supported 1913 article. Data culled from Kuhn’s Habilitationsschrift (1935a:§22) can be supported or supplemented with the help of a companion article (1935b; see 565 on the reflexes of spatha) .

14.See Elcock, loc. cit., and Rohlfs (1970:§§444-8), with a digest of opinions advanced by Elcock, Gavel, Jungemann, Martinet, Menéndez Pidal, Ronjat, and Saroïhandy. After briefly warming up to the hypothesis of Basque influence, and after praising Elcock’s efforts directed toward data gathering on both slopes of the Pyrenees, Rohlfs finally (§448) espouses Elcock’s explicative thesis, as elaborated upon by F. H. Jungemann (1956:243): It suffices to assume the absence of any Celto-Latin bilingualism from the Beam and Upper Aragon territory, permeated as that area was by a Basque-Aquitanian strain, to capture the factor which accounts for the preservation of Latin intervocalic surds. For a balanced assessment of Jungemann’s approach see, in turn, Blaylock’s pithy review article (1960:414-9).

15.For the latest panoramic view see all of T. E. Hope’s chapter contributed to Posner and Green (1980:241-87, and especially 253-8) with many useful bibliographic hints. For sheer precisionist workmanship and innovative insightful analysis Anita Katz Levy’s piece and Suzanne Fleischman’s slightly later book-length monograph, both devoted to the multipronged diffusion of derivational suffixes (-él and -atge, respectively) from the same focal area (principally Southern France and secondarily Northern France) deserve special commendation.

16.Both scholars, significantly, included in their purviews the vicissitudes of ancestral intervocalic -f-, as in the reflexes of dēfēnsa ‘protected terrain’, aedificāre ‘to build’. Meyer-Lübke argued that in the territory seized by the Arabs, by the time of its occupation, the surds assuredly had not yet been voiced; in all likelihood, he added, the same holds for the North. If so, the voicing of the surds in Spain occurred distinctly later than in Gaul (1924:1-32). Menéndez Pidal countered that, judging from Mozarabic and Arabic texts alike, the process of voicing had already invaded the entire Peninsula by the year 700, but that in the South the surds had not yet been entirely dislodged, and that in the written language (except for -f- > -v-) the more conservative variants, deemed more refined, were favored for cultural reasons; the focal point of sonantization was Leon, the headquarters of the Reconquest (1926:247-65 = §45). In the revised third edition of his masterpiece (1940:240-59 = §§45-6) the author thoroughly recast his interpretation, leaning heavily on a newly published paper by A. Tovar, slanted in the direction of epigraphy: In Roman inscriptions of the West, where a Celtic population prevailed, the surds occurring in the local pantheon were modified, while being kept unaltered in inscriptions traceable to the East and the South (1948:279 ff.). This almost sounds like a prelude to the thinking of M. Sanchis Guarner (1960).

17.(1974:133) and (1975:530-5, at 533), where he chastises H. Lausberg for having sided, early in 1948, with Menéndez Pidal’s rather than with Meyer-Lübke’s interpretation of the state of affairs in ancient Andalusia; see the following note.

18.(1948a:13). What the reader would like to find out most is Lausberg’s later definitive response to the challenge of the ambiguous Old Andalusian evidence, rather than his initial reaction. Also, Hall, in positing Gallo-Romance pressure on Hispano-Romance on such a sweeping scale, was conceivably unaware of the rival explanation proffered by Tovar in 1948 and endorsed by Menéndez Pidal two years later, to the effect that Proto-Galician-Portuguese and the adjoining dialects of Western Spain in their primitive form, reflecting a Celtic substratum long ago hypothesized, on independent grounds, could easily have exerted that pressure on Castilian first and on the fringe area of Navarro-Aragonese next. Actually, the two suppositions are not mutually exclusive: The East Central part of the Peninsula was in a sort of squeeze, with convergent waves raying out from the West and from beyond the Pyrenees, across the originally non-Celtic South of Gaul.

19.Q(u)assare is rather neatly represented in French (casser ‘to break’), but seems to have merged with Gr.-Lat. campsāre, originally a nautical term (‘to avoid, shun, circumnavigate’), to produce cansar ‘to wear out, tire’ beside deseansar ‘to rest’, whereas It. (s)cansare ‘to escape from’, refl. ‘to move aside’ represents a direct continuation of the Graeco-Latin base. The corresponding noun, Sp. cansancio (OSp. cansacio), Ptg. cansaço ‘fatigue’, preserved in the nominative case possibly because a medical, hence semilearnèd term was involved, still recognizably mirrors the contour of quassātiō ‘break’. For full details see my earlier microscopic inspection of the record (1955:225-76).

20.I am referring here to the research in progress of my U.C. Davis colleague Máximo Torreblanca.

21.This tentative analysis reflects, of course, the general thinking of André Martinet and of his closest followers, as of the mid-fifties. The “chef d’école,” incidentally, busied himself, at least, twice with the voicing of intermediate surds, and was fully prepared to make an allowance for the possible agency of Celtic, though not as the sole factor, and not entirely to the exclusion of alternative explanations (1952:192-217; 1954:257-96 [translated, revised, and expanded]). Here are some excerpts from his concluding remarks in the 1952 piece: “We have refrained from categorically rejecting the assumption that Celtic ‘lenition’ and Western Romance consonantal development, resulted from parallel evolution [i.e. A. Meillet’s ‘développement indépendant parallèle’] determined by structural analogy; but it must be clear that there exist potent arguments in favor of determining Western Romance development as ultimately due to Celtic influence. In such cases, it is usual to speak of a substratum—a term which would be just as good as any other if it had not been extensively abused ...” My single mild disagreement with Martinet here is his continued readiness to operate with the concept of “Western Romance” adopted from W. von Wartburg, the advice to the contrary implied in the research conducted by Tovar and Menéndez Pidal just a few years before, which recognized the long-drawn-out survival of surds in the East and the South of the Peninsula apparently had not yet seeped through. See also Chap. 6 in Jungemann (1956), who likewise gravitated toward the assumption of Celtic substratum, drawing neither Blaylock’s applause nor his disapproval (1960:415). I have not yet familiarized myself with G. Hilty’s presumably relevant study bearing on the Mozarabic outcomes of the Latin intervocalic voiceless obstruents (1979:145-60).

22.On this latter point see especially Lausberg (1969:160). Characteristic examples, picked from Daco-Rumanian, are cerc ‘circle’ ˂ cĭreu, lemn ‘wood’ ˂ lĭgnu, (m. sg.) negru ‘black’ ˂ nigru, plec ‘I depart’ ˂ plĭcō lit. ‘I fold’ (the tents), verde ‘green’ ˂ vĭrĭde, throughout with stressed /e/, beside cruce ‘cross’ ˂ crŭce, furcă ‘hay fork’ ˂ fŭrca, mulge ‘(s)he milks’ ˂ mŭlget, nucă ‘walnut’ ˂ nŭx -cis + -a, (m. sg.) surd ‘deaf’ ˂ sŭrdu. East Lucanian chimes in, with cruce, furea, munge, nuce, surda, except that in the front-vowel series the e from ĭ tends to become /ε/, particularly in checked syllable: Ięnga ‘tongue’ ˂ lĭngua, vęrde ˂ vĭrĭde. The state of affairs in Rumanian has been known for, at least, one century, but the older scholars refrained from drawing any sweeping conclusions from the essential assymmetry between the courses followed by ĭ and ŭ, respectively; e.g. Tiktin, in his experimental sketch: “Die rumänische Sprache” §§15, 18, 19, 23 (cf. Grōber 1888: 443-50); also, with discernibly greater assurance, in his book-length treatment (1904:§§22, 35-9). The development has been slightly obscured by the fact that, in the wake of a secondary crosscurrent, ē̆ and ō̆, in certain environments (e.g. before a nasal consonant), have tended in Rumanian to be raised to i and u, respectively. In the unsatisfactory perspective chosen by C. Tagliavini for his Lautlehre (1938:1-57), an approach which stresses change but neglects preservation, the transmutation of i into e appears blurred (45) and the survival of ŭ as u is wholly camouflaged.

23.For a vindication of this hypothesis see the excursus appended to my earlier inquiry into the lexical polarization aestimāre ‘to assess, esteem’ vs. blasphēmāre ‘to curse’ (1976:115-7).

24.(1969:156). The volume goes back to the year 1956, and could thus be put to use by Vidos (passim).

25.The following examples are cited, on purpose, almost randomly, without any serious effort to be consistent about the choice of the original or of the revised latest editions available, but with a measure of attention to the various matrixes (national cultures) involved in the gestation of the books: G. Baist, in Gröber (1888: §§20, 23), with lapidary simplicity (e.g., “ĭ und ē mit oe fallen in ẹ frŭh und unbedingt zusammen”, “ŭ und ō fallen in ọ zusammen”); B. Wiese (1904:§§19, 26); C. H. Grandgent (1905:§§33), formulaically; e.g. “Cl. L. ō, ŭ V.L. ọ > Prov. ọ”; A. Zauner (1908 : §§9, 11), again sweeçingly (“Betontes asp. e beruht: [a] auf vulgl. e (= klass. ē und ĭ, etc.), although the same author previously set apart the course taken by Sardinian and Rumanian (1900:§22; 1905:§18). J.D.M. Ford (1911:xiii, xvii); F. Hanssen (1913:§47): “... ă y ă, y casi siempre también ĭ y ē, ŭ y ō dan un mismo resultado en todas las lengues românicas”; E. B. Williams (1938, 1962, 1968:§6), who brackets, e.g. Class, ŭ and o under V. Lat. /ọ/; F. Brunot and C. Bruneau (1949:33), who credit / nọ́ké/, from nŭce, and /kọrréré/, from cŭrrere, to “roman commun”; P. Bec (1963:26): “Pas de diphtongaison des voyelles du latin vulgaire [ẹ́, ọ́] fermées = lat. class, ē, ĭ, ō, ŭ.”

Only a minority of scholars have adopted an attitude of caution and restraint, as when H. Suchier (1904-06, 1912:§8), in Gröber (1929), pointedly limited his remark to Old French and Old Provençal and added the qualification that occurrence in checked stressed syllable was initially a vital constraint; or as when É. Bourciez (1930: 148) specifically excluded from further consideration Sardinian, South Corsican, and Rumanian; or, to switch attention to a much younger scholar, as when E. Alarcos Llorach (1961:206-19)—a victim, like his constant mentor Martinet, of the fascination exuded by Wartburg’s dazzling model—operated explicitly with Western Vulgar Latin.

Interestingly, Meyer-Lübke, at the outset, not only correctly recognized that the merger of ĭ and ē may, by a significant margin, have preceded the amalgam of ŭ and ō, but, less cogently, argued that in Rumanian, pervasively, and in Spanish, under certain conditions, there later occurred a secondary rapproachement—culminating in conflation—of ŭ and ū, ŏ and ọ̄; see Gröber (1888:360). The classic formulation of Meyer-Lübke^fs subsequent thinking is found in the various editions of his Einführung (e.g. 1901:§85, 1909:§§97-98), where, even though he leaned toward labeling ŭ > ọ and ĭ > ẹ as pre-Romance shifts, he remained noncommittal vis-à-vis the dilemma of single sweeping change vs. parallel independent developments. This thinking, in cruder, simplified form, became tone-setting among his followers (e.g. Wiese 1908:§19). The total exclusion of ĭ > ẹ and ŭ > ọ from the approximately chronological tables of sound shifts appended to Meyer-Lübke’s historical grammar of French (1913:266 ff.) —a table serving the needs of relative rather than absolute chronology—may testify to his growing belief in pre-Romance dates for both processes. The entire discussion eventually rests on the immense collection of materials laboriously assembled and tentatively sifted by H. Schuchardt (1867-68:2.1-191, 3.163-226).

26.On the state of affairs in nearby Northern Calabria see R. Stefanini’s insightful review (1968:65-7) of Rensch’s Beiträge (1964). The latter’s spadework is now supplemented by his own similarly tilted dialect atlas (1973).

27.According to Lausberg (1969:§162) the “Sicilian” system encompasses also Calabria (except for a patch of territory in the extreme north) and Southern Apulia (beyond Brindisi). The author reckons with the agency of Greek adstratum and substratum. For a mature reflection on the motley tapestry of dialects in Southern Italy, with heightened attention to stressed vowels, see Devoto (1975:146-8).

28.Cf. 1969:§§158-60. Lausberg associated this “archaic system” principally with Sardinian, Southern Lucanian, and with African Latinity, throwing in for good measure—in the text, if not in the subtitle—the evidence of Latin words in Basque. For Sardinian he could fall back on the entire oeuvre of M. L. Wagner; the single most relevant (also, most handy and most up-to-date) piece being the separate volume on historical phonology, a very elaborate survey (1941:§14), plus his subsequent attempt at a synthesis, couched in semitechnical terms (1951:310).

29.Vidos’ rather meticulous documentation, which clearly outweighs his slightly pedestrian discussion—how can one continue to invoke a law or a rule in the face of so many exceptions?—(1959: 193-5), owes much to the writings of W. Meyer-Lübke, M. G. Bartoli, E. Bourciez, G. Rohlfs, H. Lausberg, E. Gamillscheg; he is particularly beholden to C. Tagliavini’s well-known textbook (1952:151, 153, 189), which, among other clues, digested for him a difficult-of-access source on the Latinity of Pannonia (A. Luzsensky’s 1933 Hungarian grammar of Latin inscriptions unearthed in that province of the Empire). For Rumanian he depended on the guidance of S. Puçcariu (1937:28). Wagner’s expertise was appealed to not only for Sardinian, but, equally important, for scattered remnants of Latin collected in North Africa (1936:29, 32). Rohlfs, who in the first place had inspired Lausberg’s dissertation, became the logical authority for the establishment of a bridge between Sardinia and South Italy—chiefly through his 1937 piece (32 ff.), whose writing coincided rather neatly with the approaching completion of Lausberg’s work on his own dissertation.

In regard to the issue of continued distinction between, versus merger of, ē and ĭ, ō and ŭ, in Vegliote (i.e. in Northern Dalmatia), M. G. Bartoli’s verdict was not to be trusted; see the authoritative qualification by Ž. Muljačić (1971:405), who in turn credits the pivotal correction to R. L. Hadlich (1965:405).

30.See Jackson (1953:86 ff., 107 ff.); Haarmann (1970:135); Beeler (1975:629); Dilts (1977:296). The discussion started out much earlier, with J. Loth’s trail-blazing 1882 monograph, which established not only that ĭ and ē, ŭ and ō were differently pronounced, judging from the reflexes of Latin words borrowed into Cymric, but that ĭ and ĭ, ŭ and ū (plus, even more remarkably, ă and ā) were separately treated. From here Meyer-Lübke took over (1901:§86; 1909:§95), combining as best he could the Celtic with the Germanic evidence (Old High German, Old English). He drew a sharp line of distinction between the Cymric reflexes, presupposing a Latinity with all quantitative vowel distinctions left intact, and the Corso-Sardinian reflexes^ resting on a Latinity drifting toward certain convergences (ī = ĭ, ū = ŭ). In E. Gamillscheg’ s circumstantial review of the Brunot-Bruneau manual one also finds a hint of the residual distinction between Lat. ā and ă in Celtic (1952:420); but he unwisely projected this discrimination onto the level of “Vulgar Latin” (left temporally and territorially undefined) and uncautiously equated it, if only by implication, with the contrast Old Provençal orthoepists would make between a larc and a estreit.

31.Much of Tagliavini’s thinking has been perpetuated by Vidos. Devoto, in his concluding summation (1974:146-8), confined himself to epitomizing the views of Rohlfs, Lausberg, and Hall. The latter’s investigations stretch from an article-length synthesis (1950:6-27) to a book-length panorama (1974); for a critique of the latter see the detailed review article by Thomas J. Walsh (1980:64-77). Although Hall’s successive researches contain some feeble attempts at self-improvement, in regard to details, he has been, I believe,—unlike myself—a consistently staunch supporter of the concept of “Western Romance” as a sub-branch of “’Italo’-Western Romance.”

In the Classification of Romance Languages appended to P. Bee’s attempt, ten years or so ago, at emulating Gröber’s Grundriß (2.472 ff.), one finds, unoriginally enough, Sardinian subclassed with South-Central Italian, then assigned to Eastern Romance; while Spanish and Portuguese, bracketed under Ibero-Romance, enter into Western Romance.

32.In his exhaustive review of H. Haarmann’s Der lateinische Lehnwortschatz im Romanischen (1972)—for which he could afford to fall back on landmark studies by E. Çabej (1962), H. Michăescu (1966), and E. P. Hamp (1972)—Michael R. Dilts made it clear that Romano-Albanian shared the phonemic merger of Lat. ĭ and ē with Continental Western Romance, but not with Romano-Brittonic; conversely, it joined Romano-Brittonic in opposing the fashion of fusing ŭ and ō peculiar to Continental Western Romance (1977:296). All of which, one might add, is reminiscent of the state of affairs in Rumanian.

33.Of these words, or word-families, I have in the past studied two under a microscopic lens, namely sombra/umbrío (1968:268-79) and du(l)ce/doce (1977:24-45), without, however, paying heed as much as I should have done to the partial retention of Lat. u. Moreover, the paper on the adjective for ‘sweet’ might have gained from exploration of one sorely neglected dimension: the general decay of such Old Spanish qualifying adjectives as tended to become monosyllabic (duz/doz, vil, and rez from older rafez > rahez). The one exception from this rule is the—fairly late—Gallicism gris ‘gray’, superimposed on native ceniciento lit. ‘ashen’, cf. Ptg. cinzento.

34.One nevertheless faintly recognizes a side-connection with the intricate case of Sp. gusto, Ptg. gôsto ‘taste, pleasure’ ˂ gŭstu beside Sp. gustar, Ptg. gostar ‘to taste, enjoy’ ˂ gŭstāre, except that here it is the West that favors the o and the Center that has settled for an u, counter to distributional pattern of the vowels in the case of Sp. costar, Ptg. custar. The link becomes even more elusive when one reminds himself that the forms of Old and Classical Spanish were still gosto and gostar; and that neither agosto ‘August’ (from Augustu) nor angosto ‘narrow’ (from angustu) display any wavering. Note the frequency of the spelling custume in older Portuguese.

35.See the excursus on primary vs. secondary causation appended to my 1955 LSA Presidential Address (1967b:242-5).

36.There seems to be a consensus of opinion among experts, including the Royal Spanish Academy, about the descent of Sp. cumbre from culmĭne, an oblique case (= ablative) form of cŭlmen. While Meyer-Lübke, in his comparative phonology (1890: §147), candidly admitted that the u ˂ ŭ was unexplained—except possibly as a vestige of the vanished l—he had no qualms, in the original version of his dictionary (1911 [-20]:§2376), to class Sp. cumbre, Ptg. cume with Rum. culme ‘top, roof-tree’, Mac.-Rum. culmu ‘gable’, It. colmo ‘ridge of the roof’, along with It. colmare, Cat. Sp. Ptg. colmar ‘to fill to the brim’, with the truncated past participle colmo ‘full’—involving the past authority of Diez and Gröber as having voiced this idea in unison. This analysis obviously left the irreducible contrast between -mbr- and -lm- in Spanish accounted for (in Italian, it was arguable that the verb had branched off the noun colmo ‘ridge, brim’ rather than perpetuating ancestral culmināre) .

Despite this transparent difficulty, and the additional embarrassment at having to separate genetically two verbs so similar in appearance, and virtually identical in semantic load, as Sp. colmar and Fr. combler, Prov. comolar ˂ cumulāre ‘to pile up’ (Meyer-Lübke, 1911[-20]:§2389), visibly akin to cumulus ‘pile, heap’ (which Meyer-Lübke saw reflected in Ptg. combro ∿ cômoro ‘elevation, steep bank’), the support for the derivation of cumbre from cŭlmine continued unabated, even though J. Corominas, for example, tends to class the verb colmar and the noun colmo, both initially recorded in the late fifteenth century, with cumulāre and cumulu, respectively (1967:159b).

The etymological problem is further complicated by the genetic uncertainty surrounding Sp. colmena, Ptg. colmeia ‘beehive’, and by the intrusion of the family of columna ‘column’, via its diminutive offshoot columella (‘little column’), Late Lat. -ellu ‘eye tooth, canine tooth’ as colmillo ‘tusk’. It would seem, then, that colm-, suggestive of ‘protrusion’, was fed by at least two separate sources, cŭmŭl- and colŭm-, but that, despite semantic affinities and striking formal resemblances, cŭlmine > cumbre steered its separate course. As if this imbroglio were insufficient, there enters into the picture the word biography of cŭlmus ‘blade, stalk, straw’, which, normally developed in Portuguese (colmo ‘straw used for the roof’), is so troublesome, on account of its diphthong, in Asturo-Leonese (cuelmo‘bundle of rye straw’) as to have pushed J. Corominas into a blind alley, seducing him to posit an identical Celtic reconstructed base both for cuelmo and colmena (1967:159b, and on earlier occasions), without winning much support from such an expert in Peninsular dialectology as V. Garcia de Diego (1955:707b). Add to this knot, increasingly difficult to untie, the facts that (a) columbrar ‘to glimpse, descry’ (documented not before the mid-sixteenth century, as a slang word) calls to mind Mod. Dial, acumbrar ‘id.’ and may represent—so Corominas conjectures—culmināre (reinterpreted as ‘to view from a summit’) conflated with alumbrar lit. ‘to enlighten’ illūmināre, and that (b) Ptg. gume, as against cume, invites association with (a)cūmen, and the lexical tangle, clamoring for a monographic inquiry, will impress itself on one in all its complexity.

37.On decades of earlier dissections of Sp. -(ed)umbre, Ptg. -(id)ão see Martha E. Schaffer’s recent paper (1981:37-62), a thoroughly revised chapter of her 1980 Berkeley dissertation. Hanssen unhesitatingly made the suffix -(ed)umbve, based on a blend of -(it)ūdine (abstracts) and -ūmine (mass nouns), responsible for the gender adopted by primitives and blurred compounds like cumbre, legumbre, and lumbre (1913:§459). The same consideration would hold for the choice of u in preference to o.

38.The problems of preference granted to Latinisms, choice of gender, continued vitality of derivational suffixes, cohesion of vocalic gamuts, and influence of leader words on other members of the given series are so tightly interwoven as to invite a far more thorough discussion than is compatible with the main thrust of this paper. Thus it is arguable that speakers adopted the learnèd form for pulpa ‘pulp’, pulpo ‘octopus’, and pulpejo ‘soft flesh’ from (a) Lat. pulpa ‘lean meat, flesh’ and (b) the zoonym pulpus, traceable to the Hellenism, polypus, in an effort to avert an unwelcome homonymie collision with the descendants of palpāre ‘to stroke, touch softly’, preserved in the Peninsula (cf. Ptg. poupar). Bulto ‘bulk’, ‘bundle, package’, ‘form, body, shadow’—a straight Latinism except for the trivial substitution of b- for v- —steers clear of any conflict with the (religiously significant) family of voto ‘vow’, etc., an advantage that would not have accrued to the “ideal” outcome *voto˂ vŭltu; etc. Then again, the fact that derivatives in -ambre, except for raigambre ‘intertwined roots’, fig. ‘deep-rootedness’, tend to be masculine (e.g. osambre ‘skeleton, bones’, pelambre ‘batch of hides to be fleshed’, ‘hair stripped from skins’), whereas those in -umbre are mostly feminine, testifies first and foremost to the decay of the -āmen/-īmen/-ūmen gamut; secondarily, it catapults OSp. (ell) enxambre ‘swarm’, (ell) estambre ‘worsted, woolen yard’, figuratively, ‘thread, course’, into a position of prominence, since the definite article ell, here ambigeneric, tended to be interpreted by the speech community as being masculine. (Note that mimbre ‘osier, wicker’ ˂ vīmine continues to be ambigeneric.) Then again, such lexical items, invariably of rural background, as involve a switch from preconsonantal l to r after ŭ (surco ‘furrow, wrinkle’ ˂ sŭlou [more faithfully preserved in Portuguese], urce ‘heath’ [phy.]) clamor for a separate investigation.

39.Menéndez Pidal’s advantage over his contemporaries consisted in his joint discussion of the entire issue, which they, conversely, tended to fragment. Thus, Hanssen discussed separately the trajectories of (a) cuno ‘stamp, die, mark’ (cf. cuña ‘wedge’) ˂ cŭneu, uña ‘nail’ ˂ ŭng(u)la, punto ‘point’ ˂ pŭnctu, puño ‘fist’ ˂ pŭgnu; (b) ducho ‘skillful, expert’ ˂ dŭotu; and (c) puches (m.) ‘porridge, gruel’ ˂ puttēs, as if these evolutionary threads were not closely interwoven (1913:§§55, 87, 89). In the process, to be sure he made valuable incidental observations, e.g. on the tendential dialectal reduction of ui to u (Old Nav. empuyssa beside OSp. empuxa ‘he pushes’ ˂ impŭlsat); but he altogether omitted important data for the sake of decorum (coño ‘cunt’ ˂cŭnnu); failed to mention alternative etymological proposals (as when ducho, OSp. var. duecho had been authoritatively traced to dŏctu ‘learnèd’); hesitated to draw obvious chronological conclusions from the contrast cuño vs. coño (/nj/ became /ŋ/ at a distinctly earlier date than /n:/); and, above all, accepted, if only by implication, the zigzag movement ŭ > *o > u in the biographies of, say, punto, puño, uña (plus) ĭ > e > i in cinta ‘ribbon, tape, band’ ˂ cĭncta) without so much as considering the possibility of selective preservation of ŭ as u and of ĭ as i.

40.More in a spirit of self-discipline than in one of optimism I have attempted to determine what help, if any, can be derived from avant-garde research in phonology. James W. Harris, for one, to cite a spokesman for the orthodox transformational approach, discusses competently the morphophonemic alternations u ∿ o, i ∿ e, as visible in the paradigms of dormir ‘to sleep’ and pedir ‘to ask’, viewed in synchronic perspective; as a matter of personal preference, I suppose, he trespasses on the domain of diachrony only with respect to certain categories of consonants. My experience with Joan B[ybee] Hooper’s Introduction (1976) has not been any better, despite the wide margin of attention she pays to Spanish.

41.Aside from knowledge gained from general linguistics, the specifically close ties among /l/, /n/, and /r/ in the development from provincial Latin into Hispano-Romance could be demonstrated with such facts as (1) the failure of the respective geminates to be simply shortened—the treatment undergone by all other consonants in medial position; (2) the evolution of word-initial r- (and, in a smaller territory, of l- and n- as well) in the same direction as word-medial -rr-, -11-, -nn-; (3) the vogue of buffer consonants intercalated between word-medial, particularly radical-final, n, l, r and some other consonant, in words like saldré ‘I’ll leave’ (lit. ‘I’ll hop out’), from salir ; tendré ‘I’ll have or hold’, from tener; (4) the strong representation of the clusters -Iz-, -nz-, and -rz-, side by side with -lç-, -nç-, and -rç-, and the ability of the former series to capture the descendants of -lǵ-, -nǵ-, and -rǵ-, under circumstances described by me on several occasions (1968b:21-64; in greater detail: 1977:33-75; with a qualification: 1982:247-66).

42.For a critical digest of Wölfflin’s and Leumann’s analyses, as well as of several other relevant contributions (by G. Cohn, F. T. Cooper, and G. N. Olcott, among others), see my contribution to the Kurt Baldinger testimonial (1979:361-74).

43.The key study, conducted with exemplary care, came from A. Thomas (1904:62-110). Later E. Gamillscheg took his cue from his former teacher’s monograph (1921:1-60).

44.A brief discussion of the agentives in -ariço, -erizo is included in my 1956 review article dedicated to Delmira Maçãs’ dissertation, Os animais na língua portuguesa. On the difficult problem of the seriatim voicing of ç to z in certain Proto-Spanish derivatives, to the strict exclusion of their Portuguese counterparts, see my exposition of a new hypothesis (1971:1-52).

45.The information available on the distribution of -ez and -eza in Spanish, at different cut-off points (Medieval, Golden Age, modern literary usage) is scarcely more than an incision performed on the tip of an iceberg (1966:341-3).

46.Phonologically, the relation of -ez to -ece reminds one strongly of Sp. -d as the counterpart of Ptg. -de, as in ciudad (originally cibdad) vs. cidade ‘city’ ˂ cīvitāte lit. ‘citizenship’ (to the exclusion of cases like Sp. merced vs. Ptg. mercê ‘mercy, grace’ ˂ mercēde, which involve ancestral -d-, rather then -t-, in radical-final position). Morphologically and lexically, Sp. -icia vs. Ptg. -iça calls to mind other instances of derivational suffixes transmitted through a learned conduit in the Center, but not in the West; witness (O)Ptg. -ença beside Sp. -encia ˂ -entia, except for scattered relics of older usage, such as OSp. simiença ‘seed’ ˂ sementia.

47.A hypothetical base which clearly yields no satisfactory results is untenable, even if Corominas lends it his support (1967: 377a). The latter, in addition to listing the fem. manceba ‘lass’, records as suffixal derivatives only mancebía, the verb amancebarse, plus the verbal abstract amancebamiento. Actually, Medieval Spanish had a whole corolla of offshoots: not only mancebía but also mancebez, plus the diminutives and hypocoristics mancebillo, -illa. Manceb-ía, -illa, and -illo could very smoothly have branched off *mancebio ˂ mancipiu, and under their joint pressure, on the analogy of neuvo ∿ novillo, and the like, mancebo could secondarily have been extracted and substituted for isolated, erratically built *mancebio. Cf. such germane cases as limpi-o ‘clean, neat’ alongside limp-ito; agri-o ‘sour’ alongside agr-ete ‘sourish’, agr-illa = acedera ‘sorrel’, agr-ura ‘sourness, acerbity’’ (even though this cluster of derivatives, strictly speaking, is based on OSp. agro).

48.These words are, on the whole, singularly recalcitrant to analysis; despite an ever present margin of doubt, one suspects that the pressure of certain adjacent consonant groups (nt, rc, rm, rt) acted, at least, as a coconditioning factor. Ptg. irto, Sp. yerto transparently continue ancestral *erctu (in lieu of well-attested ērēctu), a “strong”, i.e. rhizotonic past participle accompanying the paradigm of amply documented ērigere ‘to set oneself up, rise’, ‘to raise, build’ on its way to *ergere, which in turn is cogently explained as echoing: pĕrgere ‘to go on, continue, proceed’ and pŏrgere ‘to spread or stretch out’ (∿ porrigere), as well as sŭrgere ‘to raise up’ (∿ surrigere). The brevity of the nuclear vowel in pĕrgere, pŏrgere, sŭrgere spread to *ĕrgĕre. The ĕ thus obtained triggered a rising diphthong in the past participle yerto, with collateral support from the present indicative forms yergo, yergues, ... and from subjective yerga, ..., preferred to irgo, irga. The latter forms, mandatory in the West, coincide with the Ptg. past participle irto. Dial. Sp. erecho (var. arrecho) could then be classed as perpetuating the nonsyncopated parental form, with gradual switch from e- ˂ ē- to a- ˂ ad-.

OSp. evmano, in lieu of arch, yermano (preserved in anthroponymy) falls into place—except for the erratic mute h- —as yet another instance of enforced monophthongization of pretonic /je/ (cf. enero ‘January’ in lieu of *yenero, from lēn(u)āriu, and enzía ‘gum’, instead of *yenzía, from gingīva). But Ptg. irmão, instead of foreseeable *germião ˂ germānu, is baffling. An old intruder from the Central dialects, with ye-, alien to the West, reduced to i? The coexistence of the obsolete kinship terms Sp. cormano, Ptg. cormãno may have played a minor role.

Ptg. pirtiga, on balance, causes less surprise than does Sp. pértiga, as a replacement of OSp. piértega, a perfect reflex of pĕrtĭca. Can one posit the collateral influence of pertig-al (orig. -egal) ‘rod, staff’ and pertiguero ‘verger’ (-ería ‘office of verger’) as a joint agent—the way mancebillo, etc. has been credited with deflecting *mancebio ˂ mancipiu (see above, note 47) from its straight course? In this context note that Sp. pérdida ‘loss’ conflicts with pierdo ‘I lose’ ˂ perdō; one may invoke either the growing rarity of rising diphthongs in proparoxytones (despite tuétano ‘marrow’ and the like), or the derivative’s alliance, through semantic polarization, with búsqueda (∿ busca) ‘search’, which of course has an unchangeable monophthong in the stressed syllable. Conceivably, the latter factor, plus the pressure of perdigal, etc., transmuted piérdega into Mod. pérdiga. For Ptg. pírdiga, however, the contiguity of -rd remains by far the best explanation.

49.On the difficult problem of the five Latin word-initial consonant clusters including l as their second component see, by way of preliminary orientation, my own attempt at a provisional balance sheet: (1963:144-73), (1964:1-33), plus, among the more elaborate reactions to it—by no means consistently friendly—, the reviews by K. Baldinger, H. Meier, J. Simon, and H. G. Tuchel.

50.Other examples are ancho ‘broad’ ˂ amplu ‘wide’ and chumaceva ‘pillow’, for the most part figuratively: (mach.) ‘pillow block, journal bearing’, (naut.) ‘row lock, oarlock’, ‘strip of wood through which tholepins are driven’, from OSp. chumazo ‘feather pillow’ ˂ plūmāceu ‘feathery’.

I am now prepared to go one step farther than in the mid-sixties by arguing that the word-medial outcome /č/ in -Cc(u)lu, -Ct(u)lu (and their feminine counterparts), significantly shared by Spanish and Portuguese, pertains to the very same layer—the First Wave of Latinity—as the /č/ in chopo, choza, chubasco, and chumacera. Illustrations can be culled from any representative historical grammar (e.g. Menéndez Pidal 1941:§61.2): cercha ‘segment (of rim of wheel), rib (of center of arch)’ ˂ circulu (with gender change); concha ‘(tortoise) shell’ ˂ conchula; macho1 ‘male’ ˂ masculu (beside OSp. maslo reminiscent of muslo ‘thigh’ ˂ mūsculu, lit. ‘little mouse’); macho2 ‘sledge hammer’ ˂ mar-culu, -tulu; mancha ‘blot’ ˂ macula (with echoing of word-initial nasal); sacho ‘weeding tool’ sarculu. The survival or disappearance of the opening consonant of the cluster (e.g. /r/ or /s/) is as unimportant in this context as is the sporadic insertion of a nasal echoing word-initial m-. The sole assumption that must be made in bracketing cl- with -c(u)lu, etc., is a conspicuously early date for the syncope of ŭ in the penultimate.

It would seem, then, that at a certain juncture the use of, say, (p)ll- and, later, of 11- in lieu of indigenous ch- /č/ became fashionable, inviting speakers in the Center who until then would render Class, plēnu ‘full’ by *cheno (cf. Ptg. cheio) to switch instead to lleno. Very unpretentious country folk, who could not have cared less about speech fashions, persisted in using ch- /č/ in rustic speech at its humblest (hence choza, chumazo, chubasco, etc.). Also, speakers’ alertness being at its peak in word-initial segments and at its ebb word-medially, ancho, conchabar, plus cercha, concha, macho, etc. were somehow overlooked by the fashion-conscious and left unchanged, often receiving additional protection from the solid embedment of /č/ in such sequences as -nch- and -rch-.

REFERENCES

Alcarcos Llorach, Emilio. 1961. Fonología española. Rev. 3rd ed. Madrid: Gredos. [A 4th ed. dates from 1968.]

Appel, Carl. 1918. Provenzalische Lautlehre. Leipzig: 0. R. Reisland.

Baist, Gottfried. 1888. Die spanische Sprache. Grundriß der romanischen Philologie I, ed. by G. Gröber, 689-714. [Rev. 2nd ed., 1904-06:878-915.]

Baldinger, Kurt. 1969. Review of Malkiel (1963-64). Zeitschrift für romanische Philologie 85:512-6.

Bartoli, Matteo Giulio. 1906. Das Dalmatische: Altromanische Sprachreste von Veglia bis Ragusa und ihre Stellung in der apennino-balkanischen Romania. Schriften der Balkankommission, Linguistische Abteilung 4-5. Wien: Akademie der Wissenschaften (A. Hölder).

Bec, Pierre. 1963. La langue occitane. Que sais-je, No. 1059. Paris: Presses universitaires de France.

——. 1970-71. Manuel pratique de philologie romane. Connaissance des langues 5-6 (dir. Henri Hierche). Paris: A. & J. Picard.

Beeler, Madison S. 1975. Review of Harald Haarmann (1970). Romance Philology 28:4.626-30.

Birnbaum, Henrik, and Jaan Puhvel. (eds.) 1966. Ancient Indo-European dialects. Proceedings of the Conference on Indo-European linguistics held at the University of California, Los Angeles, April 25-27, 1963. Berkeley £ Los Angeles: University of California Press.

Blaylock, Curtis. 1960. Substratum theory applied to Hispano-Romance. Romance Philology 13:4.414-9.

Bloomfield, Leonard. 1933. Language. New York: H. Holt & Co.

Bourciez, Edouard. 1930. Éléments de linguistique romane. Rev. 3rd ed. Paris: Klincksieck. [4th ed., 1946, is a mere reprint; 5th ed., 1967, involves revision by author’s son.]

Brunot, Ferdinand, and Charles Bruneau. 1949. Précis de grammaire historique de la langue française. Rev. 3rd ed. Paris: Masson et Cie.

Caro Baroja, Julio. 1946. Materiales para una historia de la lengua vasca en su relacíon con la latina. Acta Salmanticensia, Filosofía y Letras, 1:3. Salamanca: La Universidad.

Catalán Menéndez-Pidal,Diego.1968.La pronunciacíon [ihante] por /if-frante/ en la Rioja del siglo XIII. Romance Philology 21:4.410-35.

Cornu, Jules. 1884. Remarques sur les voyelles toniques. Romania 13:285-314.

Corominas, Juan. 1967. Breve diccionario etimológico de la lengue castellana. Rev. 2nd ed. Madrid: Gredos. [A revised 3rd ed. dates from 1973.]

Corpus Inscriptionum Latinarum. 1893. Berlin: Academia Litterarum regia Borussica.

Çabej, E. 1962. Zur Charakteristik der lateinischen Lehnwōrter im Albanischen. Revue linguistique 7:161-99.

Devoto, Giacomo. 1974. Il linguaggio d’Italia. Storia e strut-ture linguistiche italiane dalla preistoria ai nostri giorni. Milano: Rizzoli.

Dilts, Michael R. 1977. Peripheral Latinity in Albanian. Romance Philology 31:2.283-98.

Elcock, W[illiam] D[enis]. 1938. De quelques affinités phonétiques entre l’aragonais et le béarnais. Paris: Droz.

Ernout, A., and Ϯ A. Meillet. 1959-60. Dictionnaire étymologique de la langue latine. Histoire des mots. Rev. 4th ed. Paris: C. Klincksieck.

Fahrenschon, Johann. 1938. Firrnus, Geschichte der Bedeutungen dieses Wortes und seiner Ableitungen in den romanischen Sprachen. Dissertation, Mŭnchen.

Fleischman, Suzanne. 1977. Cultural and linguistic factors in word formation: An integrated approaoh to the development of the suffix -AGE. University of California Publications in Linguistics 88. Berkeley: University of California Press.

Ford, Jeremiah D. M. 1911 (and later printings). Old Spanish readings, selected on the basis of critically edited texts. Rev. 2nd ed. Boston: Ginn & Co.

Fouché, Pierre. 1929. Etudes de philologie hispanique. Revue hispanique 77:1-171.

Garnillscheg, Ernst. 1921. Grundzüge der galloromanischen Wortbildung. E. G. and Leo Spitzer, Beiträge zur romanischen Wortbildungslehre 1-60. Biblioteca dell'”Archivum Romanicum” 2:1. Genève: Leo Olschki.

——. 1952. Review of Ferdinand Brunot & Charles Bruneau, Précis de grammaire historique de la langue française. Zeitschrift für romanische Philologie 68:424-49.

Garcia de Diego, Vicente. 1955. Diccionario etimológico español e hispánico. Madrid: S.A.E.T.A.

Grandgent, C. H. 1905. An outline of the phonology and morphology of Old Provençal. Rev. ed. Boston: Heath.

Gröber, Gustav (ed.) 1888. Grundriß der romanischen Philologie, 1. Straßburg: Trübner. [Rev. 2nd ed., 1904-06.]

Haarmann, Harald. 1970. Der lateinische Lehnwortschatz im Kymrischen. Romanische Versuche und Vorarbeiten 36. Bonn: Romanisches Seminar.

——. 1972. Der lateinische Lehnwortschatz im Albanischen. Hamburger Philologische Studien 19. Hamburg: Helmut Buske.

Hadlich, Roger L. 1965. The phonological history of Vegliote. University of North Carolina Studies in the Romance Languages and Literatures 52. Chapel Hill.

Hall, Robert A. (Jr.) 1950. The reconstruction of Proto-Romance. Language 26:6-27.

——. 1974. External history of the Romance languages (= Comparative Romance Grammar I). New York: American Elsevier Publishing Co., Inc.

——. 1975. La non-lenizione nella Romània occidentale. Romance Philology 28:4.530-5.

Hamp, Eric, P. 1972. Albanian. Current trends in linguistics, ed. by Thomas A. Sebeok, 9.1626-92. The Hague: Mouton.

Hanssen, Federico. 1913. Gramática histórica de la lengua castellana. Halle: Niemeyer.

Harris, James W. 1969. Spanish phonology. Cambridge, MA: M.I.T. Press.

Haudricourt, A. G., and Alphonse Juilland. 1949. Essai pour une histoire structurale du phonétisme français. Paris: C. Klincksieck. [Rev. 2nd ed. The Hague: Mouton, 1970-71.]

Hilty, Gerold. 1979. Das Schicksal der lateinischen intervokalischen Verschlußlaute -p-, -t-, -k- im Mozarabischen. Festschrift Kurt Baldinger zum 60. Geburtstag, ed. by Manfred Höf1er, Henric Vernay, Lothar Wolf, 145-60. Tübingen: Niemeyer.

Hjelmslev, Louis. 1963. Sproget. En introduction. København: Berlingske forlag.

——. 1970. Language. An Introduction. Trans, by Francis J. Whitfield. Madison: The University of Wisconsin Press.

Hooper, Joan Bfybee]. 1976. An introduction to natural generative phonology. New York: Academic Press.

Hope, Thomas E. 1980. Inter-language influences. Posner and Green 241-87.

Huber, Joseph. 1933. Altportugiesisches Elementarbuch. Sammlung romanischer Elementar- und Handbücher 1-8. Heidelberg: C. Winter.

Hymes, Dell (ed.) 1974. Studies in the history of linguistics: Traditions and paradigms. Bloomington: Indiana University Press.

Jackson, Kenneth. 1953. Language and history in early Britain a chronological survey of the Brittonic languages. Cambridge, MA: Harvard University Press.

Jungemann, Frederick H. 1956. La teoría del sustrato y los dialectos hispano-romances y gascones. Trans, by Emilio Alarcos Llorach. Biblioteca Románica Hispánica 1:7. Madrid: Gredos.

Karlsson, Keith E. 1981. Syntax and affixation: The evolution of MEUTE in Latin and Romance. Suppl. 81 to Zeitschrift für romanische Philologie. Tübingen: May Niemeyer. [Title of original 1980 Berkeley dissertation: The adverbial suffix -MENTE; its rise in Late Latin and evolution in Romance.]

Křepinský, Maximilian, and Vicente Garcia de Diego. 1923. Inflexión de las vocales en español. Suppl. 3 to Revista de Filología Española. Madrid: Centro de Estudios Históricas. [The Czech original dates from 1918.]

Kuhn, Alwin. 1935a. Der hocharagonesische Dialekt. Revue de linguistique romane 11:1-312. [Also, separately, 1936, Paris: Champion.]

——. 1935b. Studien zum Wortschatz von Hocharagon. Zeitschrift für romanische Philologie 55:561-634.

——. 1951. Romanische Philologie 1: Die romanischen Sprachen. Wissenschaftliche Forschungsberichte, Geisteswissenschaftliche Reihe 8. Bern: A. Francke.

Lausberg, Heinrich. 1939. Die Mundarten Südlukaniens. Suppl. 90 to Zeitschrift für romanische Philologie. Halle: M. Niemeyer.

——. 1947. Zum romanischen Vokalismus. Romanische Forschungen 60:295-307.

——. 1948. Zur Stellung Italiens in der Romania. Romanische Forschungen 61:320-3.

——. 1956. Romanische Sprachwissenschaft, vols. 1-3:1. Sammlung Göschen 128/128a, 250, 1199. Berlin: W. de Gruyter. [The revised 2nd ed. of Vol. I appeared in 1963; the revised 3rd ed. in 1969.]

——. 1965. 1965. Lingüística románica. Biblioteca románica hispánica 3:12. Madrid: Gredos.

Lehmann, Winfred P. 1962. Historical linguistics: An introduction. New York: Holt, Rinehart, and Winston. [Rev. ed., 1973.]

Leite de Vasconcelos, José. 1898. Notas filológicas, II. Revue hispanique 5:417-29.

Leumann, Manu. 1917-18. Die Adjektiva auf -ǏCIUS. Glotta 9:129-68.

Levy, Anita Katz. 1965-67. Contrastive development[s] in Hispano-Romance of borrowed Gallo-Romance suffixes. Romance Philology 18:399-429, 20:296-320.

Lloyd, Paul M. 1979. On the definition of “Vulgar Latin”; the eternal return. Neuphilologische Mitteilungen 80:110-22.

Loth, J[oseph]. 1892. Les mots latins dans les langues brittoniques: gallois, armoricain, comique; phonétique et commentaire, avec une introduction sur la romanisation de l’Ile de Bretagne. Paris: E. Bouillon.

Luzensky, A. 1933. A Pannoniai latin feliratok nyelötona. Egyetemes philologiai közlöny 57 (Budapest).

Malkiel, Yakov. 1946. The word-family of Old Spanish recudir. Hispanic Review 14:104-59.

——. 1955. En torno a la etimología y evolución de cansar, canso, cansa(n)cio. Nueva Revista de Filología Hispánica 9:225-76.

——. 1956. Studies in Spanish and Portuguese animal names. Hispanic Review 24:115-43, 207-31.

——. 1963-64. The interlocking of narrow sound change, broad phonological pattern, level of transmission, areal configuration, sound symbolism: Diachronic studies in the Hispano-Latin consonant clusters CL-, PL-, FL-, Arehivum Linguisticum 15: 144-73, 16:1-33.

——. 1966. Genetic analysis of word formation. Current trends in linguistics, ed. by Thomas A. Sebeok, 3: Theoretical foundations, 305-64. The Hague: Mouton.

——. 1967a. Each word has a history of its own. Glossa 1:137-49.

——. 1967b. Linguistics as a genetic science. Language 43:223-45.

——. 1968a. Identification of origin and justification of spread in etymological analysis. Romance Philology 22:3.259-80.

——. 1968b. The inflectional paradigm as an occasional determinant of sound change. Directions for historical linguistics: A symposium, ed. by Winfred P. Lehmann and Yakov Malkiel, 21-64. Austin: University of Texas Press.

——. 1971. Derivational transparency as an occasional codeter-minant of sound change: A new causal ingredient in the distribution of -Ç- and -Z- in ancient Hispano-Romance. Romance Philology 25:1.1-52.

——. 1972. Comparative Romance linguistics. Current trends in linguistics, ed. by Thomas A. Sebeok, 9: Linguistics in Western Europe, 835-925. The Hague: Mouton.

——. 1975[-77]. En torno al cultismo medieval. Los descendientes hispánicos de DULCIS. Nueva Revista de Filología Hispánica 24: 1.24-45 [Homenaje a Raimundo Lida].

——. 1976a. The etymology of Hispanic tomar. Romance Philology 30:1.115-17 [“Jean Frappier Memorial”].

——. 1976b. Critères pour l’étude de la fragmentation du latin. Atti del XIV Congresso Internazionale di Linguistioa e Filologia Romanza (Napoli, 15-20 Aprile 1974; 3 vols., ed. Alberto Vàrvaro) 1:27-47. Napoli: Galtrano, Macchiaroli, and Amsterdam: John Benjamins B.V.

——. 1977. Il paradigma flessivo come fattore determinante di mutamenti fonetici. Nuove tendenze delta linguistica storioa, ed. by W. P. Lehmann and Y. Malkiel, 33-75. Trans, by Ruggero Stefanini. Bologna: il Mulino.

——. 1978. The classification of the Romance languages. Romance Philology 31:3.467-500.

——. 1979. The prelude to the Old French “frequentative action nouns” in -ëiz. Festschrift Kurt Baldinger zum 60. Geburtstag,361-74. Tübingen: Max Niemeyer.

——. 1980. Was Schuchardt ever a neogrammarian? Romance Philology 34:1.93 ff.

——. 1982. Interplay of sounds and forms in the shaping of three Old Spanish medial consonant clusters. Hispanic Review 50:247-66.

Martinet, André. 1952. Celtic lenition and Western Romance consonants Language 28:192-217.

——. 1954. Économie des changements phonétiques. Bibliotheca Romanica 1:10. Berne: A. Francke.

Meier, Harry. 1969. Review of Malkiel (1963-64). Archiv für das Studium der neueren Sprachen 24:385-90.

Menéndez Pidal, Ramón. 1904. Manual (elemental) de gramática histórica española. Madrid. [Also, revised 5th ed., 1925; revised 6th ed., 1941.]

——. 1908-11. Ed. Cantar de Mio Cid, Vol. 1: Gramática. Madrid.

——. 1926. Orígenes del español. Estado lingüístico de la Península ibérioa hasta el siglo XI. Suppl. 1 to Revista de Filología Española. Madrid: Centro de Estudios Históricos. [Rev. 2nd ed., 1929; rev. 3rd ed., 1950.]

Meyer, Gustav. 1888. Die lateinischen Elemente im Albanesischen. G. Gröber, Grundriß 1:804-21. [Rev. 2nd ed. (1904-06): 1038-57.]

Meyer [-Lübke], Wilhelm. 1888. Die lateinische Sprache in den romanischen Ländern. In Gröber (1888:351-82).

——. 1890. Grammatik der romanischen Sprachen, 1: Lautlehre. Leipzig: Reisland.

——. 1901. Einführung in das Studium der romanischen Sprachwissenschaft. Sammlung romanischer Elementar- und Handbücher 1:1. Heidelberg: C. Winter. [Rev. 2nd ed., 1909; rev. 3rd ed., 1920.]

——. 1911[-20]. Romanisches etymologisches Wörterbuch [= REW]. Heidelberg: C. Winter.

——. 1913. Historische Grammatik der franzosischen Sprache, 1: Laut- und Flexionslehre. [Rev. 2nd and 3rd eds.] Sammlung romanischer Elementar- und Handbücher 1:2. Heidelberg: C. Winter.

——. 1924. La sonorización de las sordas intervocálicas latinas en español. Revista de Filología Española 11:1-32.

——. 1925. Das Katalanische. Seine Stellung zum Spanischen und Provenzalischen. Heidelberg: Carl Winter.

Mihǎescu, H. 1966. Les éléments latins de la langue albanaise. RESEE 4:5-33, 323-53.

Morpurgo-Davies, Anna. 1975. Language classification in the 19th century. Current trends in linguistics, ed. by Thomas A. Sebeok 13:1.607-716. The Hague: Mouton.

Muljačić, Žarko. 1971. Dalmate. Pierre Bee, Manuel pratique de philologie romane 2:393-416. Paris: A. & J. Picard.

Pedersen, Holger. 1924. Sprogvidenskaben i det Nittende Aarhundrede; Metoder og Resultater. København: Gyldendalske Boghandel.

——. 1931. The discovery of language. Linguistic science in the nineteenth century. Trans, by John Webster Spargo. Bloomington: Indiana University Press.

Posner, Rebecca, and John N. Green (eds.) 1981. Trends in Romance linguistics and philology. Vol. 1: Comparative and historical Romance linguistics. The Hague: Mouton.

Pulgram, Ernst. 1953. Family tree, wave theory, and dialectology. Orbis 2:67-72.

Puscariu, Sextil. 1937. Études de linguistique roumaine, traduites du roumain ... Cluj-Bucureşti: Imprimerie nationala.

Rensch, Karl-Heinz. 1964. Beiträge zur kenntnis nordkalabrischer Mundarten. Forschungen zur romanischen Philologie 14 (dir. H. Lausberg).

——. 1973. Nordkalabrischer Sprachatlas, an Hand der Parabel vom verlorenen Sohn. The Hague: Mouton.

Rönsch, Hermann. 1875. Itala und Vulgata. Das Sprachidiom der urchristlichen Itala und der katholischen Vulgata unter Berücksichtigung der römischen Volkssprache durch Beispiele erläutert. Rev. 2nd ed. Marburg: N. G. Elwert [original ed.: 1869].

Rohlfs, Gerhard. 1931. Beiträge zur Kenntnis der Pyrenäenmundarten. Revue de linguistique romane 7:119-69.

——. 1935. Le gascon; études de philologie pyrénéenne. Suppl. 85 to Zeitschrift fur romanische Philologie. Halle: M. Niemeyer. [Rev. 2nd ed.: Tübingen: M. Niemeyer, & Pau: Marrimpouey Jeune, 1970.]

——. 1937. Sprachliche Berührungen zwischen Sardinien und Süditalien. Donum natalicium K. Jaberg. Zürich & Leipzig: M. Niehans.

——. 1954. Historische Grammatik der italienischen Sprache und ihrer Mundarten. Vol. 3 (Syntax und Wortbildung). Bern: Francke.

Sanchis Guarner. Manuel. 1960. El mozárabe peninsular. Enciclopedia lingüistica hispánica, dir. M. Alvar et al., 1 (Antecedentes, onomástica) 291-342. Madrid: Consejo Superior de Investigaciones Cientificas.

Sapir, Edward. 1921. Language. An introduction to the study of speech. New York: Harcourt, Brace, & World. [The page references here are to the 1949 paperback.]

Saroïhandy, J[ules]. 1898, 1901. Mission ... en Espagne. Annuaire de l’École pratique des Hautes-Études 85-94; 106-18.

——. 1913. Vestiges de phonetique ibérienne en territoire roman. Revue internationale des études basques 7:475-97.

Schaffer, Martha E. 1981. Portuguese -idão, Spanish -(ed)umbre,and their Romance cognates. Romance Philology 35:1.37-62 [Henry R. Lang Memorial/Frank M. Chambers Testimonial].

Schmidt, Johannes. 1872. Die Verwandtschaftsverhältnisse der indogermanischen Sprachen. Weimar: H. Böhlau.

Schuchardt, Hugo. 1866-68. Der Vokalismus des Vulgärlateins. 3 vols. Leipzig: B. G. Teubner.

Schultz-Gora, 0[skar]. 1911. Altprovenzalisohes Elementarbuch. Rev. 2nd ed. Sammlung romanischer Elementarbücher 1:3. Heidelberg: Carl Winter. [Rev. 4th ed.: 1924.]

Simon, J. 1969. Review of Malkiel (1963-64). Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 22: 412-15.

Stefanini, Ruggero. 1968. Review of Karl-Heinz Rensch, Beiträge zur Kenntnis nordkalabrischer Mundarten (1964). Romance Philology 22:1.65-7.

Suchier, Hermann. 1904-06. Die französische und provenzalische Sprache und ihre Mundarten, nach ihrer historischen Entwicklung dargestellt. Gröber (1:712-840). [Reprinted in 1912.]

Tagliavini, Carlo. 1952. Le origini delle lingue neolatine, introduzione alla filologia romanza. Rev. 3rd ed. Bologna: Riccardo Pàtron.

Tagliavini, Karl [sic]. 1938. Rumänische Konversations-Grammatik. Heidelberg: Julius Groos.

Thomas, Antoine. 1904. Le suffixe -aricius. Nouveaux essais de philologie française 62-110. Paris: Librairie Émile Bouillon.

Tiktin, H. 1888. Die rumänische Sprache. In G. Gröber (1888: 438-60).

——. 1905. Rumänisches Elementarbuch. Sammlung romanischer Elementarbücher 1:6, dir. W. Meyer-Lübke. Heidelberg: C. Winter.

Tovar, Antonio. 1948. La sonorización y caída de las intervocálicas, y los estratos indoeuropeos en Hispania. Boletín de la R. Academia Española 28:265-80.

Tuche1, H. G. 1969. Review of Malkiel (1963-64). Romanische Forschungen 81:617-21.

Tuttle, Edwin H. 1914. The Roman vowel system. Modern Philology 11:347-53.

Vendryes, J[oseph]. 1921. Le langage: Introduction linguistique à l’histoire. L’évolution de l’humanité (dir. H. Berr). Paris: La Renaissance du Livre.

Vidos, B. E. 1959. Manuale di linguistica romanza. 1st ed. italiana, completamente aggiornata dall’autore. Trans, by G. Francescato. Firenze: Olschki.

Wagner, Max Leopold. 1936. Restos de latinidad en el norte de África. Biblioteca Gerai da Universidade, Nos. 45-46 (Coimbra).

——. 1941. Historische Lautlehre des Sardischen. Suppl. 93 to Zeitschrift für romanische Philologie. Halle: Niemeyer.

——. 1951. La lingua sarda: storia, spirito e forma. Bern: A. Francke.

Walsh, Thomas J. 1980. Hazards in the reconstruction of Proto-Romance. Romance Philology 34:1.64-77.

Wartburg, Walther von. 1928-. Französisches etymologisches Wörterbuch [= FEW]. Eine Darstellung des galloromanischen Sprachschatzes. Bonn: F. Klopp.

——. 1936. Die Ausgliederung der romanischen Sprachen. Zeitschrift für romanische Philologie 56:1-48 [with seven maps].

——. 1950. Die Ausgliederung der romanischen Sprachräume. Bibliotheca Romanica 1:8. Bern: A. Francke.

Wiese, Berthold. 1904. Altitalienisches Elementarbuch. Sammlung romanischer Elementarbücher 1:4 (dir. W. Meyer-Lübke). Heidelberg: C. Winter.

Williams, Edwin B. 1963. From Latin to Portuguese. Historical phonology and morphology of the Portuguese language. Rev. 2nd. ed. Philadelphia: University of Pennsylvania Press. [Reprinted in 1968.]

Wölfflin, E. 1888. Die Adjektiva auf -icius. Archiv für lateinische Lexikographie und Grammatik 5:415-37.

Zauner, Adolf. 1900. Romanische Sprachwissenschaft. Leipzig: G. J. Göschen. [Also revised 2nd ed. in 2 vols., 1905.]

——. 1908. Altspanisches Elementarbuch. Sammlung romanischer Elementar- und Handbücher (dir. W. Meyer-Lübke). Heidelberg: C. Winter. [Also revised 2nd ed., 1921.]