The University of Texas at Austin; College of Liberal Arts
Hans C. Boas, Director :: PCL 5.556, 1 University Station S5490 :: Austin, TX 78712 :: 512-471-4566
LRC Links: Home | About | Books Online | EIEOL | IE Doc. Center | IE Lexicon | IE Maps | IE Texts | Pub. Indices | SiteMap

Directions for Historical Linguistics:
A Symposium

Winfred P. Lehmann and Yakov Malkiel, eds.

The Inflectional Paradigm as an Occasional
Determinant of Sound Change

Yakov Malkiel

University of California, Berkeley


Introduction: Some Prospects of Research in General Diachronic Linguistics

Through the last two or three decades the balance sheet for diachronic linguistics has been predominantly negative. One notes no dearth of competent and even very respectable monographs written along traditional lines, inquiries which have clarified a number of important issues left previously in abeyance; such benefit as has accrued to general linguistics from these studies, however, has been meager, not necessarily because these inquiries displayed little imagination, still less because the data collected and distilled were incapable of stirring any fruitful discussion, but chiefly because the new and deeper insights, real or potential, which investigations of this kind could have afforded were buried underneath a mass of material of limited appeal, except to narrow specialists. To produce a truly beneficial large-scale effect, the matters of general import and broad applicability with which such studies are fortunately interspersed should have been, somewhere along the line, separated, explicated, possibly restated, and certainly made available, in a different context, on a different level of discourse, and through a different medium.

True, the picture confronting us is not all gray, let alone black; one discerns a few bright spots. Major advances have been scored either in terms of enhanced generality through cross-cultural, cross-temporal, cross-spatial analyses providing us with models and formulas, or in terms of more sophisticated, more experimental, less conventional analyses of unique historical situations high-lighted with exemplary finesse in terms of accomplishments which invite emulation. Along an independent axis, a third measurable token of progress has been the awakening of fresh interest among laymen and students, and the ability of several scholars to satisfy this hunger without resorting to rote and repetition.

To briefly illustrate these points: On the positive side of the ledger, we now have at our disposal, by way of broadly slanted methodological distillation, such tools as Uriel Weinreich's Languages in Contact (1952), André Martinet's Économie des changements phonétiques (1955), and Henry M. Hoenigswald's Language Change and Linguistic Reconstruction (1960). Among such trail-blazing monographs as exemplify rather than codify refined methods one may adduce, on the one hand, a number of weighty studies from the seasoned pen of Jerzy Kuryłowicz, bridging L'accentuation dans les langues indo-européennes (1952; 2d ed., 1958), L'apophonie en indo-européen (1956), and The Inflectional Categories of Indo-European (1964), with incisive prongs into comparative Semitics; and, on the other, Émile Benveniste's masterly Études sur la langue ossète (1959) and Hittite et indo-européen (1962), planned as inducements to reappraise the comparative method. Finally, the progressive trend in teaching is signaled by Winfred P. Lehmann's Historical Linguistics: An Introduction (1962), a book very much in demand among our tyros, and quite properly so.

Heartening as such symptoms are, it remains a fact that these advances have remained, by and large, unco-ordinated — involving virtuoso performances by brilliant, dauntless individuals rather than large-scale break-throughs achieved by teams of workers or schools of thought. Even more sobering is the realization that in some instances, after an energetic start, the probing has not been pushed to the hilt, counter to prognosis. To cite just one case of a job left unfinished: After Martinet, twelve years ago, concluded his stimulating Économie, significantly subtitled Traité de phonologie diachronique, one might have expected him to advance along at least one of the two roads conducive to beckoning goals: either the exploration of the principle of economy extended to linguistic domains other than that of phonology (say, inflection or syntax), or, alternatively, deeper probings, within the unvarying realm of sound development, into such discrete forces (the separate drives toward clarity and expressivism, among others) as are locked in ceaseless struggle with the trend toward parsimoniousness and symmetry. These hopes, one regrets to add, have not so far been fulfilled.

Yet economy of sound development, for all its overriding importance to the speakers and, if you wish, its esthetic appeal to the analyst, is just one of perhaps as many as five or even ten vital factors cutting across the growth patterns of individual languages, all of which clamor for dissection by experts combining the technical command of minute details with a detached view ranging, ideally, over numerous language families. Scholars meeting this dual qualification are, of course, few and far between; conceivably it is, more than anything else, this discouraging scarcity of leaders both well informed and farsighted, on whom the practitioners of our exacting discipline could depend for initiative and continued guidance, which is at the bottom of most of our current frustrations.

More cannot be undertaken here than to identify with varying degrees of brevity fifteen or so very pressing problems — broad-gauged and fundamental — whose firmer grasp would be of great assistance in grappling with a profusion of narrower, specific questions. These problems may, at this preliminary stage, be strung in noncommittal, haphazard sequence, as they come to mind. The structuring of these problems into anything approximating a single, coherent edifice of knowledge would mark the final stage of a venture upon which we are just now entering.

(1) Conclusions from Fluidity of Usage to the Date of Upheaval. On the whole, phonetic, morphological, and syntactic variants are known to survive side by side for an astonishing length of time (witness, for one example, the Standard Italian conjugation). Nevertheless, one might, under certain conditions, draw inferences from the intensity and the distribution of linguistic flux to its temporal distance from the event which set it in motion in the first place. Thus, in a language exhibiting, on the whole, rigidified verbal paradigms, individual spots of softness in the system of the verb might, if cautiously interpreted, yield clues to points of relative, if not of absolute, chronology.

(2) Conclusions as to Earlier Stages from the Comparison of Trajectories of Rival Forms. The systematic exploitation of the record of rival forms was initiated, but hardly brought to a successful conclusion, by the Italian neolinguists, who focused their attention excessively on the configuration of geographic areas. Tentative projection into depth might also be made through comparison of frequency of use at carefully chosen cutoff points. Suppose forms a and b are near-synonyms in 1950, with a occurring many times more often than b; suppose further that a, by 1750, was used only twice as frequently as b, and that by 1600 the two words enjoyed parity. Could one, barring external interferences, extrapolate from this gradual expansion of a at the cost of b the original prevalence of b over a and even its ultimate anteriority?

(3) Questioning of Specious Regularity. The older technique of dividing an inventory of historically analyzed forms into a “regular” and an “irregular” column (and, in the process, of almost mechanically attributing irregularity either to borrowing or to analogical pressure) has long been open to question. For one thing, that approach did not sufficiently take into account the speakers' active participation in the changes (folk etymology, camouflaging of borrowings); for another, the “exceptions” were embarrassingly numerous. There now emerge additional possibilities of exposing, through microscopic inspection, instances of apparent regularity as so many belated corrections of an earlier deviation. Thus, one (minor) analogical pressure within a verbal paradigm can deflect a few of its members from the expected path, onto which they may, three or four generations later, be pushed again through a stronger pressure exerted from a different direction. How can such adjustments (backspins) be detected where the early record is fragmentary or seems irretrievably lost? How many exceptions to sound laws may, on the assumption of such a zigzagging state of affairs, be reinterpreted as due to over- or undercorrection? How, above all, are we to explain the paradox that the (supposedly) stronger influence asserting itself in the end did not at once block the agency of its weaker counterpart, but tolerated it for a while before vigorously counteracting? (Cf. Point (17) below.)

(4) Paradigmatic Resistance to Sound Change. One of the most trivial sequences of events in language history involves a sound change (or a cluster, or else a chain, of such changes) which leads to such violent disruption of a close-knit morphological system (e.g., a declensional or a conjugational paradigm) as to entail a number of remedial adjustments, the obvious alternative being the abandonment of the impaired system itself. In some instances the succession of observable happenings seems to be inverted: an impending sound change that threatens to isolate a word from a morphologically meaningful context is, in the first place, discernibly delayed or completely warded off. The details of this shunting-off and its psychological roots remain to be determined.

(5) The Paradigm as a Stimulus for a Sound Change. It is unlikely that numerous sound changes are produced by morphological conditions; and it is further improbable that the few which may have been so produced display a conspicuously wide range. However, consider this possibility: a less than very common consonant cluster may for some reason occur pre-eminently in a characteristic group of verb forms and may thus become subject to the impact of a demonstrably influential leader verb. Once the outcome of this cluster has been affected through this primary analogical influence, at first strictly to the extent that it represents a recurrent segment of sharply delineated verb forms, further interplay of secondary analogical pressures, particularly at the point of transition from one generation to another, may extend the modification in question to nouns, adjectives, and other parts of speech. It is this particular hypothesis, applicable, above all, to some such situations as standard diachronic phonemics is unable to cope with satisfactorily, which will be examined in the main portion of the present paper.

(6) Multiple versus Simple Causation in Language Change. While it is well known that students of explicative historical linguistics tend to lean either in the direction of “substratum theories” (external influence) or in the direction of “structural modifications” (internal influence), little attention has been paid to the wisdom of positing, under certain conditions, the agency of complex, as against simple, causation, which might bridge the resultant gap. Such fundamental possibilities as habitual complementarity, bare compatibility, or mutual exclusiveness of potentially concurrent factors deserve systematic exploration.

(7) Primary versus Secondary Determinants of Language Change. Whereas the preceding point concerns simultaneously interacting forces, there exists an equally pressing need for the study of typical concatenations of such forces along the temporal axis. Mention has already been made of the (analogical) inflectional adjustments which are likely to occur in the wake of powerful sound shifts; in this context the two sets of changes, though diverse in essence, are of comparable magnitude. The situation is different if two dialects are sharply polarized, in their respective core regions, by, say, the limitation to either of one neatly profiled sound change, while the interjacent, transitional zone shows signs of confusion through osmosis. Gradually, very minor differentiating factors (neighboring sounds, number of syllables, place of stress, width of semantic scope, social or stylistic level) will allow speakers of that area — provided they lean neither toward bolder leveling nor toward the indefinite toleration of semantically unnuanced doublets — to break up the sum total of words potentially affected by the critical sound law into small groups, each typically dominated by a leader word. In such a progression of events the original sound change marks a primary determinant, while the adventitious differentiating factors must rank as secondary determinants.

(8) Interference of Nonlinguistic Factors with Language History. There has never been any doubt — least of all among lexicologists — of the important role played by certain overt external factors (say, invasions, seasonal migrations, commercial traffic, intermarriage) on the vicissitudes of languages. Far more difficult of even oblique observation, especially at past stages, are certain hidden external factors relating to the structure of the given society: the degree of cohesion of its members, the identifiable strata that can be set off on the scales of education or prestige, and the intensity of the imitative drives as against the resistance to such temptations, among others. In languages where “progressive” and “conservative,” “aristocratic” and “rustic” variants are suggested by differences in form, no truly satisfactory interpretation is conceivable without an equal share of attention granted to the social matrix. The techniques of diachronically slanted sociolinguistics remain to be established.

(9) Possible Relation of Lexical Frequency or of Incidence to “Regularity” of Sound Change. Any simple direct ratio of frequency to regularity must obviously be ruled out: When, through the syncope of an unstressed vowel, there arose, in proto-Spanish, the dyadic cluster -xr- [šr], in all likelihood uniquely found in two tenses of a single verb (exir ‘to go out’), it was perfectly predictable that the short-range solution would consist in the intercalation of a -t-: yxtré ‘I shall go out,’ on the analogy of quite common triadic clusters encountered in comparable contexts, such as -m(b)r-, -l(d)r-, -n(d)r-, -ç(t)r-. On the other hand, where a very infrequent cluster fails to match a broader pattern and each word involved seems to follow its own course, sometimes casting off three or more variants, the decision as to which variant of which word represents the norm is nearly impossible. Where lexical infrequency and relative isolation in the phonological system happen to coincide, regularity is bound to be at its lowest. Assuming that a given parent language has five word-initial consonant clusters involving l as their second ingredient, the fact that, by virtue of distinctive features, bl- and gl- (or kl- and pl-) support each other would tend to make their outcome in some daughter languages more regular than that of fl-, a group difficult to pair off; the (independently) weaker lexical representation of fl- in the chosen ancestral language would make prognosis of its subsequent course even more hazardous.

(10) “Strong” versus “Weak” Phonological Change. Scholars have learned to distinguish the pure phonological change (to be known as ph. ch.) from other modifications of sounds undergone — often incidentally — in some such context as borrowing, blend, saltatory or sporadic shift (as a rule, hazily delimited along the axes of time and space), or analogical adjustment. The distinction is useful and deserves retention, but does not exhaust the problem. There is some point in examining and even computing separately the total outcome of a phoneme (or family, or succession, of phonemes), regardless of the specific category of change at issue. The aggregate of these modifications, both “pure” and “impure,” of the sound in question may, for the sake of tidiness, be marked by capital letters: PH. CH. Let us assume that in the transition from Stage X to Stage Y every p was shifted to b, while the original b in 60 percent of the ascertainable cases yielded v and in 40 percent disappeared. Whichever explanation for the discrepancy we may offer, there is merit in discriminating between p > b as an eloquent example of a strong PH. CH., and b > v ~ zero as an equally cogent illustration of a weak PH. CH. The weakness of a PH. CH. may have a direct bearing on the infiltration of sporadic changes, lexical blends, and other modifications; it may also serve as an index of dialect mixture.

(11) Reformulation of the Age-and-Area Hypothesis. In the radical form in which “areal” or “spatial” linguistics was propounded by a few Italian extremists as a downright substitute for the comparative method, it led to a violent overreaction in many quarters, much as has, through its comparable exclusivism, glottochronology and lexicostatistics on this side of the Atlantic a decade or so later. But surely a cautious study of the extent to which temporal inferences can be drawn from present-day territorial distribution is, per se, a perfectly legitimate undertaking. It might be particularly rewarding for diachronically oriented linguists to compare notes on this subject with geologists, paleontologists, and paleobotanists, on the one hand, and with sociologists, anthropologists and f olklorists, on the other.

(12) Allowances for the Purposefulness of Language Change. Any involvement of assumptions redolent of “teleology” is bound to produce unusually sharp divisions among linguists — divisions along philosophical lines, accompanied by very strong emotional orchestration. It will be remembered that on this point L. Bloomfield, as the author of two book reviews, clashed violently with Jespersen (whom he otherwise greatly admired), in an attitude which did not prevent him from accepting (1933), tacitly and practically without qualification, some of Gilliéron's most celebrated etymological findings based squarely on the assumption of the speakers' free and active participation in the shaping of their speech (avoidance of homonymy, striving toward clarity, preference for prestige forms, indulgence in humor). The level of consciousness and the degree of volition mark the overlap of linguistics and psychology and raise problems which cannot be shirked or shrugged off by the diachronicist — indeed, which every generation of linguists must pose and seek to answer anew. Perhaps it would be helpful to separate the broad and highly explosive issue of “progress in language” from less controversial appeals to speakers' (semi)conscious involvements in narrower decisions.

(13) Complexity of Conditions Presiding over a Sound Change. One conceivably useful typology of sound changes might be organized around the number of isolable factors controlling each. Thus, a classic contrast between Old French (except for one dialect, Walloon) and Old Spanish, viewed in their simultaneous estrangement from the common parental language, consists in the fact that the diphthongization of Lat. Ĕ and Ŏ hinges in the former on three conditions: (a) word stress, (b) brevity of the vowel at the Latin stage, (c) openness of the syllable, but only on the first two in the latter. A parallel example is provided by the varying comportment, in French but not Spanish, of stressed vowels before nasals as against all other consonants. In general, Hispano-Latin sound changes are characterized by their simplicity, by Romance standards at least, while the Gallo-Latin equivalents strike one by their complexity. Granted the validity of this contrastive general characterization, it is noteworthy that a few isolated sound changes in Spanish depend on the concurrence of several factors; this is particularly true of the monophthongization of ue to e — as in fr(u)ente ‘forehead,’ fl(u)eco ‘fringe, ragged edge,’ where f-, qua labial, and -r-/-l-, qua liquids, seem to exert separate pressures — and of -ie- to -i-, as in avi(e)spa ‘wasp,’ mi(e)rlo ‘blackbird,’ pri(e)sa ‘haste,’ pri(e)sco ‘kind of peach,’ ri(e)stra ‘string,’ si(e)glo ‘world, century,’ and vi(é)spera ‘eve’, where ill-defined random combinations of (a) s-, especially before consonant or word-initially, (b) r and (c) l, in varying position vis-à-vis the diphthong, and (d) a preceding labial consonant (m, p, v) seem to produce the effect at issue. What conclusions — chronological, territorial, social, structural — can be drawn from such an atypical complexity of strictly phonological factors? (For a partial answer, see Section VII, below.)

(14) Clarity as a Driving Force, Viewed in Isolation and in Counterpoise to Economy. One shortcoming of the Gilliéronian school, known more advantageously for its detective flair than for any propensity toward theorizing, was its failure to integrate the much-vaunted avoidance of homonymic conflicts with the general dynamics of language change; add to this limitation the excessive stress on lexical — as against, say, phrasal — homonymy, and the neglect of semanticosyntactic ambiguity, along with polysemy. The entire problem, once broadened, may be reopened, and could be dramatized by gauging the impact of clarity against the power of economy matched as two rival prime determinants of language change.

(15) Reconciliation of the Family Tree and the Undulatory Projections of Language Change. Despite the wealth of scattered writings on this topic, there still exists no authoritative balance sheet, as regards either the theory or the recommended procedure.

(16) Onomatopoeia, Expressivism, Sound Symbolism, and Departure from Saussure's Classic Postulate of “L'Arbitraire du Signe.” This entire interlocking area clamors for fresh re-examination. Whereas the topic of phonic evocation of acoustic realities (noises produced in inanimate nature, animal cries, modulations of the human voice) may have been excessively — and at times inexpertly — labored, the no less attractive subject of phonic suggestion of events rhythmically segmented but inaudible (ripples on the surface of a pond, vibration, trepidation) remains a stretch of terra incognita. There arises the possibility of elevating playfulness to the same kind of pedestal as economy and clarity, particularly on the strength of linguistic developments where emotional coloration has been achieved at the expense of economy, or of clarity, or of both.

(17) Varying Intensity, as against Regularity, of Sound Changes. Within the large corpus of writings on the character of sound changes, heavy emphasis has traditionally been placed on the problem of regularity of recurrence, at the expense of the equally weighty and equally legitimate problem of a sound change's variable intensity at different periods of its ordinarily protracted agency. When we affirm that a sound change in Language A was operative over a stretch of, say, four centuries at least, we can tentatively measure its duration through several mutually complementary techniques. We can seldom pinpoint the actual start, but we are free to observe that the earliest records of the shift, within the purview of Language A, fall into the first century of its suspected activity. At the opposite end of the line, certain borrowings from Language B into A, on which the shift under study has left its imprint, cannot, for cultural or historical reasons, antedate the fourth century of its existence. The concluding segment of the line will have the benefit of an extra-neat “break” if it can be demonstrated that the next wave of borrowings no longer participated in the sound shift at issue. Without impugning the validity of this series of familiar arguments we can still wonder whether the intensity of the shift was really invariable over the entire period of four hundred years, or whether — to indulge in just one flight of the imagination — it quickly reached its all-time peak, then gradually lost momentum, tapered off, and, in the end, became extinct. We can expect to gauge the changing degrees of this intensity by pitting Sound Change X against other, contrastable forces in language growth. Let us assume that some such neatly delineated sound shift as diphthongization (X) clashes, at various successive steps, with rival forces — for instance, dissimilation in contact (Y) or at a distance (Y′), or else metaphony (Z). Let us further suppose that diphthongization succeeds in overcoming the resistance of these opposing forces, overwhelmingly in the first and barely in the second century of its existence, but fails to break their deadlock in the two concluding centuries of its four-hundred-year span (cf. Point (3), above). Granted the comparability of all the conditions involved, would it not be cogent to argue that the impact of the sound change under scrutiny has gradually weakened? Might not this hypothesis, in turn, help us to understand an otherwise paradoxical situation, namely, why a given change is allowed to occur at a certain moment, yet may, at a later juncture, be canceled by a league of the very same forces which could not, at the outset, prevent it from taking place, although they were already present and geared for action?


Exemplification of Point (5) through a Glance at Lat. -RǴ-, -LǴ-, and -NǴ- in Paleo-Romance Perspective. The General Picture

Advances in scientific insights have frequently been scored through scrupulous re-examination of all the loose ends left over after the successive or combined application of previously established methods. To formulate and exemplify one assumption seldom made in Romance quarters (and possibly elsewhere), namely that morphological conditions — specifically, characteristic or recurrent features of the verbal paradigm — may, in the last analysis, be held responsible for a baffling sound shift (see Point (5), above), the highly idiosyncratic Old Spanish development of Lat. -RGei-, LGei-, and -NGei- will here be thrown open for discussion. Section II is designed to furnish a bird's-eye view of the entire problem, to set off the erratic, self-contradictory trajectory of the three clusters in Spanish from their normal, perfectly transparent evolution in most if not all cognate languages (Old Provençal being another “outsider”), and to sketch in a very schematic causal explanation of the deviation thus isolated. In subsequent sections the varying degrees of anomaly in Old Spanish and Old Provençal will be gauged within the framework of the hypothesis advocated, attention will be given to the peculiar emergency situation — the détresse phonologique — which, in the first place, makes morphological interference with sound change more readily understandable, and a few tentative generalizations will be drawn concerning detection and diagnosis of such suspected interferences. Only by way of afterthought will a sampling of rival explanations be listed and discussed, as succinctly as possible, in an effort to show the source of their inadequacy.

In its narrowest projection and at first deliberately viewed out of context, the problem at hand can be described in the following terms:

The Old Spanish development of the Latin medial consonant groups (a) -RG-, (b) -LG-, and (c) -NG- before front vowels is puzzling within the edifice of Hispano-Latin phonology. An additional difficulty accrues to the outcome of the -NǴ- group inasmuch as, apart from the particular transmutation which places it in the neighborhood of the group listed under (a) and (b), its record shows traces of growth in two other directions; no neat hierarchy between these three slants (α, β, γ) has so far been established.

The parallelism between -RǴ-, -LǴ-, and -NǴ- consists in the shift of palatalized g () to z, a unit phoneme pronounced with or without affrication, depending on period, locus, and position within the word: [dz] or [z]. Examples from medieval sources (except for an occasional modern dialect form) include:

These latter outcomes, -nn- (i.e., ñ) and -n-, show such a degree of fluctuation before ie that, at least in this position, they seem to represent either two prongs or, still better, two phases of a single development, with -ñié- doomed, or tending, to yield ground to -ñe-.

Very frequently within the ranks of Group (c), and only by way of exception elsewhere, one observes the wavering between the three possible treatments: (α) change of ǵ to z, (β) merger of ǵ with the preceding consonant, conducive to the crystallization of a new consonant, and (γ) loss of ǵ without any concurrent change in the preceding consonant. Perhaps not insignificantly, the high rate of fluctuation in Group (c) coincides with the predominance of verbal families in that group, while the lower rate of fluctuation in Group (a) matches the nominal character of many, possibly most, of its constituents; Group (b) is too exiguous to invite any parallel comment on the spread of form classes. Thus, the rival reflexes of FRANGERE ‘to break, shatter, dash to pieces’ in the oldest texts were franzer and frañer, of which the latter before long emerged as the stronger. In the germane cases of CINGERE ‘to gird, surround’ and TANGERE ‘to touch, seize, strike, play (an instrument),’ the victory of -ñer, -ñir over *-zer precedes the dawn of vernacular literature, and there is no lack of passages in archaic texts where, before -ié-, n replaces ñ: cinientes (Alexandre), taniendo (Ruiz). The reverse outcome of the competition between variants can be tidily documented in at least one instance: from IUNGERE ‘to join, yoke’ standard Spanish has inherited uncir (< -zir), while uñir ‘to yoke,’ along with its offshoot uñidura ‘yoking,’ is at present relegated to dialect speech (Extremadura, León, Salamanca, Valladolid, Zamora). Finally, a sharp split within a single family has occurred in the case of *RING-ERE, Class. Lat. ‘to show the teeth, snarl, growl’ > reñir ‘to quarrel’ as against the corresponding noun renzilla ‘quarrel’; at older stages the latter, confusingly enough, was flanked by the var. reñ-illa.

It must be understood that in the medieval paradigm of verbs traceable to prototypes in -NGERE there survived certain forms, directly representative of the sequences -NGŌ or -NGA-, in which -ng- was temporarily left intact. Thus, the present indicative of the descendant of FRANGERE ran: 1 frango, 2-3 franze(s), 4 franzemos or frañemos; and the subjunctive was franga(s). The situation was further complicated by the accompanying gradual absorption of the given -er (<-ĒRE, -ĔRE) verbs by the vigorously expanding -ir (<-ĪRE) class, which, quite unlike the -er class, exercised a powerful metaphonic influence on the radical (hence mod. riño ~ reñimos). Add to this, as a separate factor of intricacy, the postmedieval unvoicing of z [dz] to ç [ts], later deaffricated to [θ] or [s], according to the region involved. It was the speakers' reaction to multidimensional complexities of this kind which, in the end, provoked widespread leveling, in the course of which the remaining -ng- forms were wiped out in favor of -ñ- or, occasionally, of -nz-, -nç-: tango > taño ‘I touch,’ unga > unza ‘let me (or him, her) yoke.’2

The one nontrivial instance of wavering outside the domain of -NǴ- involves ARGENTU ‘silver,’ which yielded either ariento (cf. top. [Leon.] Arintero, [Gal.] Arenteiro < ARGENTĀRIU) or arzinto (vivo) = Hisp.-Ar. azogue ‘quicksilver’; the -rz- form was favored, though not to the point of exclusiveness, indubitably in Mozarabic and possibly in Aragonese. Characteristically, the derivative ARGENTEU ‘made of, covered with, silver; of the color of silver’ — later substantiated in the vernacular: ‘small weight, small coin’ — followed a radically different course. Its Hispanic product, whose inventory shows a good deal of fluctuation in regard to minor features, was in the main arien-zo, -ço (with the satellite formation aren-, aran-çada *‘merchandise, terrain worth that coin’). Clearly, the alternative outcome *arzenço, involving an affricated dental spirant at the start of two consecutive syllables, would, if it ever came into existence (say, in Mozarabic), have succumbed to the powerful dissimilatory trend.3

The transmutation of -NǴ- into [ɲ] and its further occasional shift to [n] before -ié- pose no serious problem within the total context of Romance phonology.4 On the other hand, the emergence of -z- as the second ingredient of the cluster exposed to fronting is conspicuous in the extreme, whichever measuring rod the analyst favors to gauge the degree of irregularity.

For one thing, the broad trend in Romance has been to treat the second element of a dyadic medial group, especially if it was a stop, like its word-initial counterpart, leaving them both preferably intact; compare, with respect to t, APTĀRE ‘to fit, adapt, get ready’ > Sp. atar ‘to tie’ beside TŌTU ‘whole’ > todo; where apparent exceptions are detectable (as with MULTU ‘much’ > mucho, STRICTU ‘tight’ > estrecho ‘narrow’), the erratic form is usually at least two steps removed from the base (cf. the intermediate stages preserved in Ptg. muito, estreito). The fact that, in the case under investigation, l, n, and r happen to constitute the first ingredient of such a cluster does not, of itself, lead one to expect any departure from the general tendency, compare the unvarying comportment of so capricious a phoneme as D in DOMINU ‘master’ > dueño, on the one hand, and in CAL(I)-DĀRIA ‘kettle, stove’ > caldera, SOL(I)DU ‘solid (substance)’ > sueldo ‘copper coin, soldier's pay,’ VENDERE ‘to sell’ > vender, MORDERE ‘to bite’ > morder, on the other. Quite a few departures from this standard of preservation are on record, true, but they usually either (a) have a bearing on the first element of the cluster, which in the process may altogether disappear, (α) leaving traces (ALTERU ‘one of two’ > otro ‘other’ via outro, cf. Portuguese) or (β) failing to do so (MORSU ‘bite, biting’ > mueso — as in atar, above) or, if they affect the second element at all, (b) entail a minor adjustment (BARBA ‘beard’ > OSp. barva, HERBA ‘grass’ > OSp. yerva; the adjustment is here defensibly called minor against the background of the well-known Hispano-Latin b ~ v fluctuation). Since word-initially neither Ǵ- nor the consonantized I of, say, IĀ-, IĒ-NUĀRIU ‘January’ with which Ǵ- tended to coalesce in late provincial Latin ever yielded /z/, but rather disappeared — compare GELĀRE ‘to freeze’ > elar, GERMĀNU ‘(half-)brother’ > ermano, IĒNUĀRIU > enero5 — the three clusters -nz-, -lz-, and -rz- must, on this score of broad patterning, rank as highly idiosyncratic.

This judgment is unlikely to become subject to revision if we include in our field of observation a few cognate languages. In closely related Portuguese, for instance, events took a markedly different, far less astonishing course. Except for a dwindling lexical nucleus which stands apart (GERMĀNU > irmão, [gloss] IECUĀRIA ‘giblets,’ lit. ‘dish made of liver’ [IECUR] > iguaria ‘tidbit,’ cf. OSp. yegüería),6 the reflex of Ǵ- and J alike was /ž/, presumably affricated at the earliest stage: GELĀRE > gear (cf. geada ‘frost’), GENERU ‘son-in-law’ > genro, GENUCULU ‘(little) knee’ > jẽolho > joelho, IACĒRE ‘to lie, rest’ > jazer, IĀNUĀRIU > janeiro (beside IĀN-ELLA [dim. of IĀNUA ‘outer door, entrance’] > janela ‘window’), IĒIŪNĀRE ‘to fast’ > jejuar, DIĀRIA [dj-, j-] ‘daily (quota of work)’ > geira ‘yoke of land,’ lit. ‘day's plowing of a yoke of oxen.’ Word-medially, after a consonant, the same /ž/ appears, even following an N, especially in the ranks of the -NGERE verbs: SPARGERE > espargir, orig. -er, TANGERE > tanger ‘to ring (bells), play (an instrument), goad (a herd),’ LONGĒ > longe; characteristically, despite the obvious temptation to dissimilate the two ǵ's appearing at the start of two successive syllables, speakers of Portuguese have allowed GINGIVA to survive in its pristine shape almost unaltered, as gengiva.7

On the Italian side, to briefly adduce the Tuscan evidence: The situation in and around Florence is very much as in Portugal, except that the affrication of // has here been more faithfully preserved. Once again, we encounter the expected parallelism of the two positions: on the one hand, GENERU > genero, GENUCULU > ginocchio, IĒNUĀRIU > gennaio; on the other SPARGERE > spargere, SURGERE > sorgere,8 TERGERE > tergere, MULGĒRE > mungere ‘to milk, extract, sponge on,’ EXPINGERE ‘to push out’ (from PANGERE ‘to fix, drive in, undertake, agree upon’) > spingere ‘to push, thrust’; and, eloquently paired off in a single word, in defiance of the ubiquitously latent dissimilatory trend, we find Gi and -Gi in GINGĪVA > gengiva.9

French offers an entirely different picture, inasmuch as the early syncope of the weakest vowel here produced a series of unattractive consonant clusters which speakers later smoothed over by intercalating homorganic “buffer consonants,” particularly -d-: IUNGERE > *[džoɲrə] > joindre [žwε̃drə], PLANGERE ‘to beat one's breast’ (in token of pain or mourning) *[plaɲrə] > plaindre ‘to lament.’ It will be remembered, from a celebrated article by J. Gilliéron, that MULGĒRE was on its way to becoming moudre, but swerved from the straight path, because speakers shied away from the threat of this verb's coexistence with the homonymic progeny of MOLERE ‘to grind.’ The trajectory of -RǴ- is exemplified by SURGERE ‘to rise up’ > sourdre. Word-initially the outcome of Ǵ-, J- was [], eventually deaffricated: GENERU > gendre, IĀNUĀRIU > janvier, DIURNU ‘daily’ > jour ‘day.’ Because it is hazardous to reconstruct the preliterary form that underlies, say, sourdre, it seems wisest to withhold judgment on the extent to which proto-French may have abandoned the broad phonological trend in Romance, a trend so clearly discernible in Portuguese and Tuscan.

Old Provençal, in striking contrast to Old French even though on a distinctly more modest scale than Old Spanish, did occasionally exhibit the change of postconsonantal -Ǵ- to -z-, alongside the rivaling shift to /ž/ (spelled g). Thus, (a) ARGENTU here yielded argen; (b) BUR-GĒNSE cast off both bor-ges and -zes; (c) FUL-GUR, *-GER ‘lightning’ plus the verbs SPARGERE and SURGERE emerged on the local scene as fólzer, espárzer, and sórzer, respectively. The prevalence of verbs in Group (c), as against the predominantly nominal character of Groups (a) and (b), seems, to say the least, noteworthy. It is the first of these three developments which matches most neatly the local course of word-initial Ǵ-: GENTE ‘clan, breed, people’ > gen, Gr.-Lat. GYRĀRE ‘to turn, revolve’ > girar. As regards -NǴ-, however, Provençal displays the same development as Spanish in lueñe < LONGĒ: compare fránher < FRANGERE and plánher < PLANGERE (with the digraph -nh- signaling [ɲ]).

This partial coincidence of native strata in Old Spanish and Old Provençal — two languages geographically noncontiguous, each boasting a lexicon interspersed with easily recognizable mutual borrowings — is a matter of considerable importance. Whichever explanation of the rise of Cons. + z we may favor, the hypothesis endorsed must snugly fit two different contexts.

To revert to Spanish: given all these complications, it is baffling that an expert of the stature of R. Menéndez Pidal should — as late as 1950 (after abjuring a different interpretation, long championed by himself in the wake of others) — have referred to an evolutión perfectamente comprensible. Quite the contrary: phonologically the development represents a genuine crux. It is only after running into a blind alley in our strictly phonological operations that we can, with a clean conscience, solicit help from a neighboring subdiscipline.

In this impasse the history of Spanish verbal inflection seems to offer the missing link. Every scrap of evidence points to the powerful analogical influence exerted, in the two moods of the present tense, by DĪCŌ, -ĔRE, a verb whose paradigm is characterized by the neatly patterned interchange of -g- and -z-: (ind.) digo, dize(s)..., (subj.) diga(s). The influence here posited is visible in the attraction OSp. dizer exerted on fazer < FACIŌ, -ERE: while Portuguese (1 faço, 2 fazes,... 7 faça) and Italian (1 faccio,... 5 facete,... 7 faccia) preserve quite faithfully — after one makes the necessary phonological allowances — the configuration of the Latin prototypal forms, Old Spanish tramples upon the tradition by substituting 1 fago for *faço and 7 faga for *faça, thus giving rise to two perfectly parallel series, one regular: digo, diz(es),... diga, the other decidedly analogical: fago, faz(es)... faga. But the influence of the voiced velar, which acts as the central pillar of digo, diga (meanwhile reinforced by fago, faga), went much farther: the -g- infiltrated oya < AUDIA(M) and, on the dialect level (sporadically also in Golden Age texts), huya < FUGIA(M), leading to oiga and huiga, and made its pressure felt in many other ways; compare the subjunctives (obs.) fierga, ponga, salga, tenga, (obs.) tuelga, valga, venga — clearly echoing di-ga, ha-ga, also archaic franga, tanga — beside the corresponding infinitives ferir, poner, salir, tener, toller (> tullir), valer, venir.10

One need not, of course, hold DĪC-Ō, -AM > Sp. dig-o, -a alone responsible for the propagation of the velar. In Tuscan, where DĪCŌ yields dico and FACIŌ survives as faccio, the crystallization of pongo, tengo, and vengo must, for instance, be attributed exclusively to the combined weight of the FINGŌ, PANGŌ, PINGŌ, TANGŌ, TING(U)Ō... series, which also happens to be lexically more deeply entrenched in the Appenine Peninsula (cf. spingere). In the light of this evidence, it is wisest to postulate for proto-Spanish the combined pressure of digo and the gradually receding -ngo verbs. Once this assumption has been made, only one step separates us from the further conjecture that the present-tense paradigm of dezir, with its characteristic alternation of -g- and -z- (digo, dizes, diga), may have left an imprint of this alternation on, say, frañer, producing fran-go, -zer... alongside differently leveled frañ-o, -e(s)... and in preference to traditional frango, frañes, which for a while had proved immune to deflection. The novelty of the assumption consists solely in hypothesizing the analogical spread, from an established focus of diffusion, not of the characteristic g viewed in isolation, but of the striking alternation g:z taken en bloc. Given the well-known cohesion of all -er and -ir verbs in Spanish (a residual or closed series), it will further be readily granted that the innovation of -nz-, sparked by the contact of dezir/fazer with the -ñer verbs, could have leaped to other verbs of the same conjugation classes (-er/-ir), inviting the formation, under comparable conditions, of -lz- and -rz- clusters. The concluding and, theoretically, most noteworthy — if problematic — step in the chain of events here visualized would then have been the diffusion of -lz-, -nz-, -rz- to nonverbal components of the lexicon, culminating in the crystallization of a “sound law” unattached to any particular form-class.


Three Additional Considerations

These sweeping contentions are certainly in need, first, of a measure of qualification and, second, of corroborative support. Our first check must be on the comportment of Old Provençal, whose partial affinity to Old Spanish has been insufficiently stressed in the past.11

In the Provençal deposit of DĪCERE the alternation of radical-final -g and -z is not as sharply silhouetted as in Old Spanish, in consequence of the deeper erosion of word-final vowels — witness the 1 pres. ind. dic (< *digo), the inf. dir(e), the 2 pres. ind. ditz, as against OSp. digo, di- or de-zir, and dizes, respectively. But enough of the alternation remains (pres. subj. diga, 4-6 pres. ind. diz-em, -etz, -on, impf. ind. di- or de-zia, ger. dizen) to give substance to the surmise that some influence — distinctly weaker than in Old Spanish — could have acted on certain -GERE verbs.12 These expectations are fully borne out by the rather tidy record available for inspection. While the -NGERE verbs here stand consistently apart (cénher < CINGERE, destrénher ‘to harass’ < DIS + STRINGERE, esténher ‘to quench’ < EXTING(U)ERE, fénher ‘to feign’ < FINGERE, fránher < FRANGERE, plánher < PLANGERE), the -RGERE verbs supply a few telling examples: sorzia < SURGĒBAT and sórzer < SURGERE, in their relation to sorga < SURGAT, may have been proportionately arrived at on the model of dezia: diga;13 the same supposition holds true for esparzer. The point to remember is the correlation, gratifying from the vantage point of Spanish, between (a) the narrower range, in Old Provençal, of the /g/ ~ /z/ alternation in the paradigm here assumed to have started the entire movement, namely DĪCERE; and (b) the narrower scope, to the north of the Pyrenees, of the shift -Ǵ- > -z-, the dual restriction (unknown to Spanish) being phonological: only -RǴ- is affected, and morphological: verb forms almost exclusively are involved (though nominal borzes beside borges is on record). Another consideration — this time within the confines of Spanish — on which the acceptance of the hypothesis here championed may be made contingent, is whether a certain internal difficulty in the expected sound development could have produced an emergency — or, to speak with Gilliéron and his followers, a détresse — which would make the course taken by the community's tone-setting speakers more readily understandable in retrospect. Through a noteworthy coincidence, the development of word-initial Ǵ-, J- from Latin to Old Spanish, an evolution which, on the strength of a broad phonological trend, might very well have matched the controversial changes undergone by Ǵ, J word-medially after consonant, is itself far from transparent. If analogues may be borrowed from other Romance languages, the likeliest reflexes of Ǵ- and J- would have been /j/ or /ž/. Traces of both these developments are on record, with /j/ < J- surviving especially before central and back vowels. It is not yet quite clear how they can be most persuasively hierarchized; compare IUNCTA > junta ‘junction, seam, joint, gasket’ beside yunta ‘yoke (of animals),’ the chances being that /ž/ points toward a more aristocratic pronunciation (cf. refined joven beside racy moço ‘young’). Total loss of the initial consonant in this environment is rare and almost invariably attributable to dissimilation of palatals at the start of successive syllables (cf. ayuno ‘fast’ < IĒIŪNU, unc-ir ‘to yoke’ < IUNǴ-). Before a front vowel the loss of the consonant is, conversely, the norm: GERMĀNU > OSp. ermano, while its occasional preservation ranks as a clue to learned transmission: general (as against truly vernacular Ptg. gèral). If we assume that *yermano, *yenzía ‘gum’ were the proto-Castilian forms, it can be argued — and the argument is not new — that the reduction of *ye, that is, , to e was the direct consequence of the close association of the ie diphthong with the stressed syllable.14 Given the demonstrably late date of the diphthongization Ĕ > ie, the resultant monophthongization ye- > e- in pretonic syllable, through overreaction, would have to be assigned to a still later date. Wavering between *ye- and e- in words like (y)ermano, (y)enzía could have produced the setting for the sporadic elimination of (y)- from other contexts where it was bothersome, as in (y)unzir — with dissimilation, sporadic or saltatory shift, and regular sound change working hand in hand, as they often do.15

If we now ask ourselves just what events could be posited as ideally normal on the assumption that at the outset /j/ also developed from Ǵ medially after consonant, the answer is that -LǴ- might have cast off */lj/ and, by the same token, -NǴ- should have led to */nj/ and -RǴ- to */rj/. For -NǴ- this line of reasoning assuredly holds, with /ɲ/, as in LONGĒ > lueñe, actually emerging as the predictable next step. For -LǴ- a parallel sequence of events, conducive to /λ/, would be self-explanatory, but the material available is too scant to yield any useful information on the rate of recurrence of this evolution. In the case of /rj/, however, a development essentially metathetic, that is, quite at variance with that of /nj/, might be anticipated if one were to take one's cue from VARIU ‘variegated, spotted, changeable’ > ve(i)ro (as in the furrier's term vero), -ĀRIU > -e(i)ro, and the like. The single and very cautious assumption, then, that Ǵ-, J- tended to yield in Castilian /j/ rather than /ž/ as in most cognate languages (including Portuguese and Catalan at its flanks) accounts for the genesis of major disruptions both word-initially (especially through the late recoil from *ye- in pretonic position) and word-medially after consonant (particularly through the varying impact of /j/ on the preceding /n/ and /r/); also — a signal additional obstacle — for the discrepancy between the evolutionary trends in these two positions ordinarily known, at least in Romance, for their marked mutual affinity.16

To sum up: There very definitely did arise a state of acute phonological emergency in proto-Spanish with regard to Ǵ-, J-, a situation favoring, even inviting, a kind of external intervention as one avenue of escape from excessive structural fragmentation. Under these circumstances, contagious effect of DĪC-Ō, -ĔRE and of the -NGŌ, -NGERE verbs, that is to say, the extension, by force of analogy and by the interplay of the proportion g:z, of an inflectional pattern to the point of its elevation to the rank of a sound correspondence, gains in plausibility. One could speak of a transition from the morphophonemic to the phonemic level.

Interestingly, there exists an additional instance of the substitution of z in Old Spanish for *j traceable to Ǵ; again the process awaits clarification. This time, an isolated (if important and semantically many-faceted) word is involved: rezio ‘strong, thick, coarse, harsh, hard’ (mod. recio), in whose prototype the -Ǵ- was surrounded by front vowels: RIGIDUS ‘stiff, rough.’ If we ask ourselves: What should the outcome of RIGIDU have been? and if we fall back on such classic correspondences as FRĪGIDU ‘cold’ > fri(d)o (via *friyo), LEGERE ‘to pick, read’ > leer, MAGIS ‘more’ > ma(i)s, MAGISTRU ‘master, teacher’ > maestro, NĀVIGĀRE ‘to sail’ > navear,17 the answer to our question will be *reyio (or at most *reido), a locally most unappealing form by virtue of both the falling diphthong ei, barely tolerated by Old Spanish (cf. the Catalanism pleito ‘lawsuit’ beside native plazdo < PLACITU, also disyllabic reÿ ‘king,’ trisyllabic reína ‘queen’), and the latent instability of -io, a tidy suffixal marker of the far-flung adjectival series moored to -IDU:18 limp-io ‘clear,’ suz-io ‘dirty,’ teb-, tib-io ‘lukewarm.’ Small wonder that on the local scene so infelicitous an outcome of RIGIDU exerted no attraction on the nearly parallel product of FRĪGIDU, in eloquent contrast to the events that occurred in French (where OFr. freit, froit presuppose the influence of roit, the common ancestor of mod. raide and roide) and in Italian (where freddo points to an *-ĬGD- rather than -ĪGD- segment). One is tempted to go one step further: Far from exerting pressure on FRĪGIDU, the hypothetical proto-Spanish malformation that sprang from RĬGIDU could easily have succumbed to the influence of some other word, preferably an adjective forming part of the same close-knit series. Granted the likelihood of such a course of events, SŪCIDU ‘juicy, sappy’ > OSp. suzio might be credited with providing a model for the replacement by rezio of some such unviable form as *reyio. There was no dearth of semantic contacts between the two adjectives here suspected of interaction, though one has the impression that these affinities were too weak to have of themselves provided the initial stimulus. Nevertheless such links could easily have sufficed to translate into action the speakers' recoil from *reyio; that they are not merely a figment of one analyst's imagination follows from the reverse pressure plausibly exercised in Portuguese by the representative of RIGIDU > rijo on SŪCIDU > sujo.19 Rezio differs, then, from enzía, terzer, both in phonological detail and in the choice of specific analogical models; it shares with the words here under consideration the secondary character of the -z-, its source in the parent language, and the avoidance of */j/ in certain contexts.


The Mechanism of the Sound Changes and Their “Inner Diffusion”

It was the assumption in Section II that the striking relation diG-o, -a : diZ-e(s), within the contagious paradigm of a verb independently known for its aggressiveness, could have supplied a handy model for franG-o, -a : franZ-e(s) in rivalry with, and eventually as a substitute for, the older frañ-e(s). Similarly, esparG-o, -a : esparZ-e(s) may have replaced an older ineffectual *esparye(s). The reason for the vulnerability of the latter form can be thus pinpointed: Had it been an isolated word, unfettered by any paradigmatic ties, the natural course for the speakers to follow would have been to transmute it into *espaire(s); compare the unimpeded development of the common suffix -ĀRIU > -airo > -e(i)ro. But since the word at issue was the captive of its membership in a verbal alliance, this smooth road was blocked by the menace of the genesis of two allomorphs such as esparg- ~ *espair-, paired off in a pattern unprecedented in Hispano-Romance conjugation. With /j/ thus immobilized, through paradigmatic tightness and cohesion, in its phonologically no longer comfortable position after, rather than before, the /r/, its analogical replacement by /rz/, suggested by digo ~ dizes, must have come to speakers as a welcome alternative, as an actual relief.

Outside the domain of conjugation, but still within the realm of morphology, we discover derivational relationships that may have favored similar solutions. Granted that BURGĒNSE in all likelihood initially yielded some such product as *boryés, the phonologically tempting further advance to *boirés (cf. CORIU ‘leather’ > coiro [as in Ptg.] > cuero) was scotched by the coexistence — and continued semantic proximity — of the primitive borgo. A very opportune spark could at that stage of latent tension have flown from the SPARGERE to the BURGUS family, placing *boryés/borzés alongside *es-parye(s)/esparzes.20

Once the exclusive link to the verb was loosened or relinquished, any late and unremovable /rj/, /lj/ would tend to cast off /rz/, /lz/, while the temporary rivalry of doublets involving /rj/ beside /rz/ would, like any state of fluctuation, provoke the agency of codeterminants on a liberal scale. Thus, at the phase */arjiλa/, the descendant of ARGILLA could have profited from the ubiquitously latent dissimilatory trend in changing to arzilla and thus eschewing the sequence /rj . . . λ/, obnoxious through its excess of palatality. The final phase would be marked by the analogical penetration of -rz-, -lz-, -nz-, riding the crest of a vogue, even into nooks and crannies of the lexicon where speakers could derive no perceptible benefit from its prevalence over rival forms.


Some Procedural Remarks

If the concrete cases here inspected under microscope constitute a fair sample of the problems falling under the rubric “Morphological Interference with Sound Change,” a few tentative generalizations suggest themselves at this point. Diagnostically, it seems wisest to start from the axiomatic assumption that most sound changes can be accounted for in terms of a phonological system's internal balance and economy. It is only where explanations of this order fail, without undue stretching, to do justice to ascertainable facts that the analyst is well advised to try out a “second string” of possible factors of causation. This strategy has traditionally been adhered to where suspicion of borrowing from an adjoining language was at issue (lateral pressure); some scholars have further applied it to assessments of the plausibility of substratum influence (vertical pressure). It seems theoretically defensible and also feasible to attach to this second string the possibility, rare in some languages yet conceivably frequent in others, of inflectional pressure.

In many cases the first step will be the detection of an embarrassing discrepancy between (1) a broad phonological trend (such as the prevalent coincidence, in Romance, of consonantal development [a] word-initially and [b] medially after consonant) and (2) a specific sound correspondence or a set of such narrower correspondences (such as the triple shift characteristic of Old Spanish: -NǴ- > -nz-, -LǴ- > -lz-, and -RǴ- > -rz-). Where substratum influence and lateral borrowing cannot be invoked, the probability of a starting point concealed inside an inflectional paradigm increases very sharply. If two or more noncontiguous cognate languages show varying dosages of the same treatment, and if this significant differential can be persuasively harmonized with the discrepancies between the inflectional paradigms thought to have set these developments in motion, so much the better: witness, in our special problem, the understandably unequal scopes of adventitious z in Old Spanish and Old Provençal. The concluding step in the operation is to ask oneself what ideal narrow developments — only marginally on record, if at all — can be extrapolated from the known broad phonological trend. If these theoretically normal developments abounded in stumbling blocks for the speakers, the spread of the innovation (in this instance, of the z) from the paradigm of the verb, via certain proportional relationships fostered by fluctuation, to the core of the system of sound correspondences ceases to be erratic. By way of afterthought, one may then examine some evolutions which give the impression of embodying parallels. Even if this first impression of close resemblance evaporates upon closer inspection, as was true of rezio < RIGIDU, certain recurrent isolable features within each ensemble of circumstances — for example, the substitution of analogical /z/ for */j/ in both rezio and enzía < GINGĪVA — may be worthy of sustained attention.


A Bird's-Eye View of Earlier Analyses

A retrospective survey of earlier opinions is appended here, less for the sake of its intrinsic significance than par acquit de conscience, as it were. Incidental references to morphological pressures as a possible or probable molder of sound changes are, of course, scattered over many linguistic writings, old and new.21 Also, suspicion of an analogical spread of OSp. -zir from a few key verbs to other infinitives (or rather paradigms) is not entirely new,22 but its original advocate stopped short of taking the vitally important second step, namely that of concluding that the analogical -z-, after its diffusion over certain verbal paradigms, might ultimately have wormed its way into the strictly phonological domain. Yet, on the whole, the discussion has been distinctly barren, in part as a result of several recalcitrant etymologies, which acted as so many stumbling blocks,23 in part because scholars could afford either to disregard with impunity the rise of -lz-, -nz-, -rz- altogether (on account of the meager lexical representation of these clusters?)24 or, misled by their atomistic attitude, to state it as a bare fact, left unintegrated and unexplained.25

Among the explicative conjectures the following deserve to be set off:

(a) An appeal to consonant dissimilation involving the assumed unvoicing of the second member of a set in the case of GINGĪVA > Fr. gencive, Sp. encía (Meyer-Lübke, 1890).26 The same scholar on that occasion attributed sencillo to the influence of SINCĒRU and, possibly, to the coexistence of uncir/uñir, wondering whether the discrepancy separating arienzo (Ǵ > 0) from esparcer, ercer, arcén, and ancilla (Ǵ > θ) should be traced to varying vowel qualities or to a dissimilatory trend. Any such explanation, accounting for each word individually and abounding in doubts and alternatives, is unacceptable in the context of phonology; the intrinsic difficulty was here compounded by the author's inability to segregate medieval from modern graphies and forms.

(b) The contrast riño ‘I quarrel’ / riña ‘feud’ (n.): rencilla ‘bicker’ (n.) seems to point to word stress as the differentiating feature, at least in the ranks of the reñir < RINGĪ family. From the observation of this one alliance of forms Menéndez Pidal for a while gained — and conveyed to others — the impression that -NǴ- > -nz- and the two related developments (-lz-, -rz-) crystallized only in pretonic position;27 by the mid-20's, having become aware of his error, he candidly retracted the conjecture with a bow to a dissenter (A. Castro, see below under [d]).

(c) The intermediate stage between -RǴ- and -rz- must have been */rž/: a sequence pieced together out of whole cloth first by G. Baist28 and later by J. D. M. Ford in Old Spanish Readings (p. xxxviii). Was it suggested by the evolutionary trend in Portuguese?

(d) The difference between, on the one hand, QUĪNGENTŌS > quiñentos (coll. Sp., Jud.-Sp., Ptg.) and, on the other, the words displaying -nz-, -rz- (common starting point: -NG- > -NǴ- > *-ÑǴ-) involves, first and foremost, syllable juncture: thus opined A. Castro in a substantial book review (RFE, II [1915], 181). This is a sufficiently accurate description of the process at issue, but hardly a causal explanation of the differentiation.

(e) A richly nuanced development confronts the observer, within which certain strains can be neatly isolated. The most noteworthy of these streaks, the one involving the rise of -nz- and -rz-, allegedly shows a “perfectly understandable” shift Ǵ > z (see (c) above), running parallel to the more familiar shift > ç, that is, /c/ or [ts]. This has been Menéndez Pidal's view since the mid-20's,29 presented contagiously enough to have gained adherents30 and formulated on the basis of a copious array of data tidily transcribed, dated, and localized. The documentation is priceless and unassailable, but the conjecture seems vulnerable: If it is true that -Ǵ- > -z- (say, in ARGILLA > arzilla) acts as a voiced counterpart of -Ḱ- > -ç- (say, in VINCIT > vençe), it is equally true that the latter result concomitantly honors the broad trend equating the growth of word-initial and -medial postconsonantal occlusives (cf. CAELU > çielo), while z fails to straddle the two developments: One encounters not the slightest trace of GERMĀNU > *zermano.31 One additional flaw in Menéndez Pidal's reasoning: Since the assibilation of occurred early and that of Ǵ, which he posits, would have followed suit at once, it is inexplicable how this allegedly natural z could have dissolved in the following /je/, as in ARGENTEU > arienzo. Only on the assumption that there at first lingered on a /j/ : /arjenço/, which merged with the demonstrably late diphthong ie < ę < Ĕ, while succumbing in different contexts to an analogical z spreading from the most contagious of all verbal paradigms (di-, de-zir), do we fully reconcile plausible temporal sequences with narrow and broad phonological developments.

(f) At this point one wishes it were possible to report some breakthrough scored by the application of structural analysis; yet such, disappointingly enough, is not the case. E. Alarcos Llorach's paper, offering an advantageous bird's-eye view through joint consideration of Ge,i, J, DJ, GJ (and, as a foil, of Ke,i, TJ, KJ), has the merit of conjoining, for the first time, the processes under study, especially as observable between vowels, with the treatment of geminate occlusives and of -LL-, -NN-, -RR-, along imaginative lines suggested by A. Martinet;32 while some facets of the evolution, as a result of this choice of focus, stand out in gratifyingly sharper relief, the idiosyncratic Castilian treatment of -LǴ-, -NǴ-, -RǴ-, which does not smoothly lend itself to the approach, dwindles into insignificance, being relegated to a brief and inconclusive footnote.33 B. Pottier's passing mention of arcilla and encía is not at all helpful, and one is shocked to see him conjure up, however cautiously in comparison with V. García de Diego, the ghost of accentual interference which Menéndez Pidal had, one hoped, at long last allayed in 1926.34 H. Lausberg, through the subtle overtones of his wording, draws the well-versed reader's attention to the diachronically erratic character of OSp. esparzer, senziello, and unzir, but is discreet enough to withhold any comment that might firmly commit him to some narrow-gauged causal explanation.35 This triple restraint, which borders on unadmitted failure, is perhaps not entirely coincidental: It marks the limit, in explicative matters, of diachronic phonemics as it developed in the quarter century 1940-1965. Like the advocates of any legitimate and provocative method, the practitioners of diachronic phonemics, after elegantly solving a number of problems previously rebellious to analysis, have left in their wake an embarrassingly copious residue of unanswered questions. To come to grips with these residual difficulties we must make bold to shift certain emphases without further delay.


One Tentative Parallel

Were the case here examined at length a severely isolated instance of the presumptive concatenation of circumstances, the credibility of the entire argument would be reduced to the narrowest of margins. Fortunately, striking parallels can be marshalled; in fact, one need not go far afield to discover analogous sequences of events. The chief reason for calling the counterpart selected for mention here “tentative” is the less heavy documentation with which, for expedience's sake, the new hypothesis will be surrounded.36

The conditions under which Lat. Ĕ and Ŏ diphthongized to ie and ue shortly before the curtain went up on the recorded history of Spanish are well known and relatively simple to describe. A less easy task is the statement of those circumstances whose interplay, at a distinctly later date, allowed ie and ue to be monophthongized to i and e, respectively — in a radically asymmetric pattern.37 The monophthongization occurred only in a pitifully small minority of cases, especially as regards ue > e, and was slow in attracting scholarly attention, the upshot of this delay being the absence of any workable consensus on the explanatory level.

Let us first state the raw facts. The rising diphthong ue — from the outset one of the most characteristic features of the Spanish sound system — yields e in just a few cases which seem to demand a straight phonological explanation: the geographic name Bur-(u)eva, mod. La Bureba (Prov. of Burgos), plus cul(u)ebra ‘snake’ < COL-ŬBRA, *-Ŏ́BRA, fl(u)eco ‘fringe, flounce, reveled edge’ < FLOCCU, and fr(u)ente ‘forehead’ < FRONTE. In a few additional instances alternative causes come to mind; specific sound conditions may here have acted as mere concomitants: estantigua ‘procession of hobgoblins; bugbear, badly dressed person, scarecrow’ < OSp. huest(e) antigua, lit. ‘old army’ < HOSTE ANTĪQUA (either loss of primary stress through composition, or the agency of taboo, or the combination of these forces, may have acted as the prime mover); estera ‘mat(ting)’ < STŎREA ‘rush mat’ (contamination with the common suffix -era < -ĀRIA, which in this particular context showed marked functional affinity); compare the parallel shift of Berceo's suffix -duero < -TŌRIU to -dero;38 pestorejo ‘back of the neck,’ a masculine companion piece to the combination of POST ‘after, behind’ and AURICULA, the diminutive of AURIS ‘ear’ (vowel dissimilation may have been at work, brought to bear on *postorejo, compare redondo ‘round’ < ROTUNDU; in this eventuality there arose no diphthong to begin with).

The ranks of Old Spanish words containing i traceable to ie are more sizable, and the diphthongal forms are, for the most part, documented. Typical representatives include: aprisco (v.) ‘I gather the sheep in the fold,’ (n.) ‘sheep-fold,’ from *APPRESSICĀRE; avispa ‘wasp’ < VESPA (conceivably with an a- borrowed from abeja ‘bee’ < API-CULA ‘little bee’); mirl-o, rarely -a ‘blackbird’ < MERUL-A, rarely -U; níspero ‘medlar-tree’ and níspola ‘medlar-fruit’ < MESPILU beside-A; pingo ‘I drip’ < *PENDICŌ; prisa ‘haste’ < OSp. pries(s)a < PRESSA ‘pressure’; prisco ‘kind of peach’ < PERSICU, lit. ‘Persian apple’; remilgo (refl. v.) ‘I am affectedly nice, squeamish,’ (n.) ‘affected gravity, prudery,’ related to MEL ‘honey’; ri(e)stra ‘string, row, file’ < RESTE, with hypercharacterization of gender and secondary -r- after -st-; si(e)glo ‘world’ < SAECULU (presumably through the instrumentality of OFr. siegle, as against mod. siècle); vi(é)spera(s) ‘eve’ < VESPERA (HŌRA) ‘evening.’39 To this larger contingent of examples one must add numerous instances of -iello, -iella, in the overwhelming majority of cases a suffixal element, originally diminutive (a value not infrequently blurred at the Romance stage; cf. Cast-illa, OSp. -iella < CASTELLA, lit. ‘encampments, castles’); at rare intervals the concluding segment of a root morpheme, as in silla ‘chair (lit. stool), saddle’ < OSp. siella < SELLA. Finally, the reduction of ie to i occurred in a number of rather obscure medieval proper names, opportunely assembled by R. Menéndez Pidal in 1926: Liestra ~ Listra, Asieso ~ Asiso, Agierbe ~ Agirbe (g = /j/), Xavierre ~ Scavierri ~ Exavirr, beside mod. Javierre, Espierre, Espierlo.

This is not the place to review the existing, woefully scattered literature on this two-edged problem.

In analyzing the reduction of ue to e one has the impression of weakness and sporadicity, qualities which, in turn, are determined by the number of conditions that must be met before the change can eventuate. Aside from the primary stress falling on the critical syllable, there must be present, in close vicinity, a bilabial consonant plus a liquid (r or l). The accumulation of so many conditioning factors reminds one of Rumanian or Old French historical phonology, rather than of Spanish, known for its bold and clear-cut architecture, devoid of all manner of “Schnörkel.” Small wonder that the shift was easily blocked. Thus, OSp. vuestro ‘your(s),’ beside informal vuesso, met all the requisite conditions, yet failed to cast off *vestro, *vesso, presumably because the parallelism with nuestro ‘our’ sufficed to counterbalance the tendential sound shift. Yet Latin for centuries tolerated NOSTER beside VESTER.

If, in passing on to -ie- > -i-, we take as our prime classifier not some specific neighboring sound, but — on a more abstract plane of reasoning — the number of identifiable conditioning factors, we must draw a sharp line between, on the one hand, dim. -i(e)llo, -i(e)lla and, on the other, all the remaining examples. In the case of the diminutives, the single source of causation seems, at first glance, to be the [λ] of -i(e)llo and its feminine counterpart; at any rate, the phonic, syllabic, and accentual configuration of the root morpheme and the choice of -o or -a as the final vowel seem to have exerted not the slightest influence on the long-drawn-out fluctuation between diphthong and monophthong.40 The situation is radically different with the other members of the group: Here, as previously with the handful of -ue- > -e- cases, one observes, typically, the simultaneous agency of two separate factors, usually the appearance of (a) an adjacent or, at least, not too far removed l or r (better still, R and r) and (b) the characteristic Castilian apico-alveolar /s/ which, lying as it does between [s] and [š], gives the acoustic impression of a palatal consonant [ś]; almost invariably the s is either word-initial or represents the first ingredient of a medial consonant duster. Examples include aprisco, prisa, ristra, siglo, vísper(a)s. One finds certain variations; thus, it would seem that the co-occurrence of r and l makes the intervention of s dispensable, witness mirlo, remilgo. Because bilabials (b, p, m) happen to figure prominently in several of these examples, it is not impossible that in the end they too acted, upon occasion, as a secondary conditioning factor; compare avispa, where the s was a concomitant and the analogy of víspera(s) might have been operative, *pirtega (if this reconstruction mediates between OSp. piértega and mod. pértiga), which paired off p- with -r-, and pingo, where these supporting circumstances were absent unless one credits the nasal with the power to substitute, at intervals, for either liquid. Generally speaking, aprisco, pingo, and remilgo stand apart insofar as they are members of verbal paradigms rich in arrhizotonic forms; their -i- may have first crystallized in pre-tonic syllables, as a reflex of e before such clusters as -ng-, -lg-, -sc-, a circumstance which potentially detracts from the weight of their evidence.

There is no strict parallelism, then, between the respective scopes of ue > e and ie > i, quite apart from the conspicuous asymmetry of these reductions within the system. The retrenchment ie > i affects a palpably larger number of lexical items, some of them verbal, involves a richer interplay of pairs of conditioning factors, and, in one crucially important instance — be it only on account of its exceptionally high incidence (-illo, -illa) — seems to hinge on a single circumstance.

The temporal sweep of -ie- > -i- is also noteworthy. The trend remained in operation long enough for OSp. sieglo, clearly a borrowing from Old French (much as OPtg. segre is a Provençalism), to have yielded siglo. On the other hand, Menéndez Pidal, through scrupulous sifting of dated documentary evidence, established the fact that in the suffix -iello, -iella monophthongization occurred at a remarkably early date, though it took speakers of Spanish centuries to rid themselves entirely of the receding -ie- variants (which have entrenched themselves to this day in certain conservative Asturo-Leonese dialects).41 Taking into account all these circumstances of spread and impact, one is inclined to argue, first, that the shift -ie- > -i-, at least in its earliest manifestations, must have preceded, by a margin of centuries, the shift -ue- > -e-, almost — but not entirely — parallel; and, second, that the stimulus for the monophthongization of -ie- may very well have come from -i(e)llo, -i(e)lla, judging from chronological evidence.

These temporal considerations may be reinforced by phonological analysis. To the best of our knowledge, OSp. -ie- was pronounced /je/, an assumption which makes its occasional reduction to -e-, as in dix(i)eron ‘they said’ (cf. fn. 37, above), perfectly plausible, while its alternative reduction to i, under a different set of circumstances, is anything but readily understandable in phonic context. The difficulty would disappear if we were to appeal to analogy: Old Spanish pitted in stiff competition a variety of diminutive suffixes, esp. -iello, -ino, -ico, -ito, and -uelo. It would seem that at a certain juncture stressed i emerged as a characteristic marker of diminution; that in the wake of this process the discrepant suffix -uelo, once very abundantly represented, was relegated to the background; and that -iello temporarily survived the onslaught of its competitors at the price of exchanging its uncharacteristic -ie- for the highly suggestive -i-.42

One distinct advantage of the conjecture made here is that it is apt to account, as was no previous hypothesis, for the conspicuous rôle played by /l/, /r/, /R/, and /ś/ in the limited monophthongization of -ie-. If one recalls that, by virtue of its distinctive features, the /λ/ of -iello, -iella was linked, on the one hand, to /l/ and, somewhat more closely, to /r/ and /R/ and, on the other, to /ś/ — in view of the latter's well-known palatality —, one can readily see why the conjunctions of at least two such phonemes partially resembling the /λ/ could sporadically produce a similar effect in transmuting the -ie- into -i-.

The labials and labiodentals were not originally endowed with comparable force, witness fiesta ‘festival’ < FESTA and inhiesto < OSp. infiesto ‘steep’ < ĪNFESTU ‘menacing, hostile,’ which, in their resistance to change, behave exactly like tiesto ‘flowerpot’ < TESTU ‘pot lid’ and siesta ‘hottest part of the day, after-dinner nap’ < SEXTA (HŌRA), while the combined pressure of labial and liquid on -ie- may produce more impressive results: mirlo, pingo. However, when, through the agency of symmetry, speakers felt tempted to match the occasional monophthongization of -ie- with a similar treatment of -ue-, the most appropriate environmental feature that, in alliance with l or r, could be charged with controlling the shift -ue- > -e- was a /b/, a /v/, or an /f/. Hence Bureba, culebra, fleco, and frente.

Of the two relevant phonic ingredients of /λ/ it was thus the link to /l/ and /r/ that turned out to be the stronger, both within the ranks of -ie- words and, secondarily, in the far smaller group of -ue- words. Palatality did attract /ś/ into this process, but endowed it with much less power than that invested in /l/ and /r/; note the contrast between ristra and siesta. The contact with labials and labiovelars may have been accidental at the outset; at least, one can account for the monophthong of aprisco, mirlo, prisa, and vispera without any direct appeal to the presence of m, p, or v in the respective words. But whether or not merely coincidental in these particular instances,43 the labials and labiovelars in the end became prime conditioners (though at no time sole conditioners) after the transfer of the mechanism of monophthongization from the front to the back of the mouth cavity.

One side issue worth pondering is the relative weakness of monophthongization where ascending diphthongs are involved; contrast the meager yield of this process with the rich results of the earlier monophthongization, again in Spanish, of ei to e and of ou to o. This weakness of the trend has allowed conjugational analogy to override it effortlessly, with the result that the reduction of ie to i and of ue to e operates almost exclusively in the category of nouns, including onomastic items. It might be rewarding to discover whether such sound changes as have been sparked by analogy rooted in inflectional or derivational conditions lack, as a general principle, the strength commonly found in sound changes produced by purely phonic factors.

At this latitude, a comparison of the two major problems so far examined — the one involving three medial consonant clusters (-RǴ-, -LǴ-, -NǴ-) in their transition from Latin to Romance, the other bearing on the partial monophthongization of two rising diphthongs (ié, ué) at the concluding stage of Old Spanish — may be particularly rewarding. What the seemingly quite disparate problems share is the fact that at first glance they appear to be purely phonological in content, but as soon as the analyst turns his attention to the matter of causation, strictly phonological conditions fail to provide any adequate, truly dependable clue. This qualification does not exclude the possibility that such conditions played a subordinate role in the two nearly parallel processes; but the single most plausible factor of causation turns out, on both occasions, to have been analogy, operating through two adjoining provinces of morphology. In the case of the three confederated consonant groups the agency that transmitted the impact was a set of crucially important, rather similar verbal paradigms; in the case of the secondary monophthong, the stimulus came from the close-knit alliance of diminutive suffixes, most of them marked by the same characteristic (if you wish, “expressive”) vowel. The local intensity of analogical interference, as distinct from its original direction, was determined, in the case of -RǴ-, -LǴ-, and -NǴ-, by a number of interconnected phonic conditions. The transfer of monophthongization from ie > i, where a primary cause was at work (association of -iello with -ito, -ico, -ino), to > e, where the application of no such direct pressure was discernible, must have been made in response to a deeply ingrained craving for symmetry.

Romance linguistics boasts a rich reservoir of problems that can potentially benefit the general methodology of diachronic analysis of language. Some of these controversial problems have been solved, though even then the reward for any scrupulous re-examination is bound to include deeper insights and improved formulations. Others — in fact, counter to widespread belief, very many — await definitive clarification. Perhaps Romance scholars, who have been somewhat negligent of this challenging commitment, should educate themselves to give priority not to those questions which relate to residual gaps in their grasp of the particular fabric of Romance culture, but to those, more urgent, which may enrich our general understanding of the anatomy of language change.44


I owe a number of useful, if minor, suggestions to several scholars who attended the Texas conference, in particular J. Kuryłowicz and U. Weinreich, and also, among my own students, J. R. Craddock. A few weeks after its original offering the paper was presented orally to a group of linguists at UCLA, where it benefited from a second discussion sparked by queries from W. Bright and R. P. Stockwell. The distribution of the preprint netted a few interesting epistolary reactions, notably those of E. García (Columbia) and R. T. Harms (Texas); the latter suggested a restatement, in transformational terms, of the nuclear section of the paper. On the Berkeley campus R. Stefanini has drawn my attention to several Italian parallels and near-parallels. Thus, such northeastern dialects as Veronese and Trentino exemplify the spread of the voiced velar from di-go, -ga, ‘I say, may say,’ producing fa-go, -ga, pa-go, -ga, da-go, -ga, sta-go, -ga, which match stand. It. faccio (fo) ‘I make,’ vo ‘I go,’ do ‘I give,’ sto ‘I stand,’ and the corresponding sets of pres. subj. forms, thus corroborating the kernel of my conjecture. Also, borg-ese and borgh-ese ‘burgher’ coexisted in Old Italian, as they still do in anthroponymy. To Diego Catalán Menéndez Pidal, almost my nextdoor neighbor in the year 1965-1966, I am indebted for comments on some finer points of the Hispano-Romance material. Marilyn May Vihman gave me, as on many similar occasions, the benefit of her incisive stylistic criticism.


1 The data adduced at this preliminary stage have been culled chiefly from R. Menéndez Pidal's writings: Manual de gramática histórica española, 6th ed. (Madrid: Espasa-Calpe, 1941), § 47:2b, and Orígenes del español, 3d ed. (Madrid: Espasa-Calpe, 1950; Obras, Vol. VIII), § 49, with a measure of extra attention to the tidiness of Latin bases. The author also cites the top. Castil Anzul (between Antequera and Aguilar, i.e., in Andalusia, not too far from Málaga and Cordova), tracing it to ANGĒLLAS; if this is so, has there occurred any secondary association with azul ‘blue’?

2 Outside the verbal paradigm the leveling could, under certain circumstances, proceed in the opposite direction; cf. burgués, which hugs more closely the radical of burg-o than do the older forms burz-és (vernacular) and burgés (certainly an old Gallicism rather than a “cultismo,” as J. Corominas, DCELC, I, 548a, rashly assumes).

3 This dissimilatory treatment calls to mind the preservation of the initial cluster in Sp. clavija ‘pin, peg,’ playa ‘beach,’ and the atypical shift FL- > l (FLACCIDU ‘flabby, weak’ > Sp. lacio ‘withered, faded’), as explained in my article “The Interlocking of Narrow Sound Shift, Broad Phonological Pattern, ...,” ArL, XV (1964), 144-173; XVI (1965), 1-33. See also J. R. Craddock's parallel comment (“A Critique of Recent Studies in Romance Diminutives,” RPh, XIX [1965-1966], 286-325, esp. 315-318) on F. González Ollé's discovery of an important pattern in the distribution of -uelo and -i(e)llo.

4 One is free to interpret the shift -ñie- > -nie- as an alternative to the widespread reduction of -ié- to -e- after a palatal consonant, in conjugation and in suffixal derivation: dix(i)eron ‘they said,’ amarill-ento ‘yellowish’ beside gras-iento ‘greasy.’ These two processes are quite different, with regard to the attendant circumstances, from the tendential monophthongization of -ie- to -i- and of -ue- to -e- in the transition from Old to Modern Spanish. (See Section VII, below, and the corresponding footnotes.)

5 There are a few details to observe regarding the adjacent vowel: IĒIŪNU yielded ayuno (cf. also desayuno ‘breakfast’) rather than *eyuno by virtue of the vacillation between IĒ- and IĀ- which has left vestiges in Latin literature and may have been more sharply profiled in provincial speech, in all likelihood; the loss of I- was extra early here, as in uncir < IUNGERE, by way of recoil from medial -I-. GENUCULU > OSp. inojo ‘knee’ recalls, in its raising of e to i, the case of GERMĀNU > Ptg. irmão; the var. finojo owes its f- either to confusion with fe-, fi-nojo ‘fennel’ < FĒNUCULU or to the transfer of f- from the verb figuring in the stereotyped phrase fincar los inojos ‘to kneel down,’ on the strength of the partial identity of fin- and in-. (For further elaboration see “La F inicial adventicia en español antiguo,” RLiR, XVIII [1954], 161-191, esp. 185-190.)

6 On iguaria see my article in Language, XX (1944), 108-130, and the postscript, ibid., XXI (1945), 264-265. Despite the sporadic criticism that these writings aroused, I am still convinced that the solution advocated was correct.

7 For simplicity's sake, in dealing with Portuguese and French data I disregard the fact that, at a certain evolutionary phase, syllable-final n began to dissolve, with varying speed and results, into the preceding vowel.

8 The difficulty is compounded by the fact that certain Latin bases have been locally transmitted in different strata; thus, being vernacular, Fr. sourdre and It. sorgere are endowed with evidential force which is absent from Sp. surgir, visibly a crude Latinism.

9 At first blush the segment -nc- of Fr. gencive < GINGĪVA is reminiscent of its OSp. near-equivalent -nz- in enzía. But the resemblance may be specious; in all likelihood the OFr. var. gengive, at present peculiar to the southern belt of Oïl dialects (see E. Gamillscheg, Etymologisches Wörterbuch der französischen Sprache [Heidelberg: C. Winter, 1928], p. 465a, with clues to further literature), represents the original regular outcome, while the innovation gencive may be due to a lexical blend triggered by consonant dissimilation, or to the dissimilatory trend pure and simple.

10 This situation can again be contrastively dramatized through reference to the conservative state of affairs in Portuguese, where one finds tenha < TENEAM, venha < VENIAM (and, in their wake, ponha despite PŌNAM), also valha < VALEAM and saia ‘I may go out’ < SALIAM ‘I may jump.’ The analogical spread of the velar in Spanish involved, of course, a long-drawn-out process, and one is free to cite instances of late contamination such as oiga and huiga or, for that matter, caiga < ca(y)a < CADAM and traiga < tra(y)a ‘I may bring’ < TRAHAM ‘I may drag, pull along’ only on the explicit assumption that they represent distant reverberations of the same shift — observable at an early date in faga and fierga — which is here held indirectly responsible for the rise of unorganic -lz-, -nz-, and -rz-.

11 Surprisingly, W. Meyer-Lübke, in re-examining this entire issue toward the very end of his career (“Zur Geschichte von lat. Ge Gi und J im Romanischen,” VRom, I [1936], 1-31), paid proper attention to the Provençal (esp. p. 28, with a reference to C. Appel's Lautlehre) but not to the hastily sketched Spanish facet of the elusive problem.

12 In citing Old Provençal forms I am guided by authoritative handbooks such as O. Schultz-Gora, Altprovenzalisches Elementarbuch.

13 Additional models available in Old Provençal, though not in Old Spanish, included iaz-er ‘to lie’ < IACĒRE vs. p. ptc. iag-ut, lez-er ‘to be permitted’ < LICĒRE vs. p. ptc. leg-ut, nózer ‘to damage’ < NOCĒRE vs. p. ptc. nog-ut, plaz-er ‘to please’ < PLACĒRE vs. p. ptc. plag-ut, etc. The characteristic -g- ingredient of the past participles in question has been extracted from the corresponding preterites, in which it reflects ancestral -U̯Ī.

14 Characteristically, Old Spanish preferred otri and otre ‘somebody else’ to otrie, nadi (or ninguno) ‘nobody’ to nadie; alguien, introduced into the literary language from the western dialects (where it bordered on deeply entrenched OGal.-Ptg. alguén), was originally stressed on the e, as is known from rhymes.

15 Compare my observations in “Linguistics as a Genetic Science” (in Lang., XLIII [1967], 223-245) on R. Posner's and K. Togeby's mutually exclusive views on the agency of consonant dissimilation.

16 Inexplicably, J. D. M. Ford, Old Spanish Readings (Boston: Ginn & Co., 1911 and later printings), p. xxxviii, toyed with the idea that, after n and r, /j/ first became , then dz; for details, see the appended historique du problème. For one attempt to arrive at a total view of Ge,i, in Peninsular perspective and with a structuralist slant, see E. Alarcos Llorach, Arch., IV (1954), 330-342; cf. K. Baldinger, La formación de los dominios lingüísticos en la Península Ibérica (Madrid: Editorial Gredos, 1963), p. 24, and my own comment, infra.

17 See R. Menéndez Pidal, Manual de gramática histórica española, 6th ed., § 413.

18 For a panoramic view of this problem, which awaits monographic treatment, see my article “Multiple versus Simple Causation in Linguistic Change,” in the Festschrift To Honor Ramon Jakobson: Essays on the Occasion of His Seventieth Birthday (The Hague: Mouton, 1967), II, 1228-1246.

19 Since intervocalic G before vowel disappears in Portuguese no less consistently than in Spanish (LEGERE > ler), it is advisable to start from the stage *rigju which must have been reached on the local scene very early, given the extra-quick extinction of -D- along the Atlantic Coast. The (convergent) groups -dj- and -gj- tended to yield /ž/ in the West, but /j/ or zero in the Center unless the Center opted for a learned form; contrast Ptg. -ejar with Sp. -ear < -IDJĀRE, Ptg. enveja ‘envy’ < INVIDIA with Sp. envidia.

20 I realize that OProv. bor-zés beside -gés poses slightly different problems in the absence of any visible or inferable /j/. Could it be that the preservation of the r, vitally needed in a family headed by borc < *borgo, was better ensured in the embedment of the /rz/ than of the // cluster? Another problem, lexical in nature, must temporarily be left in abeyance: While OSp. OPtg. borgés is clearly of Gallo-Romance provenience judging by its form (cf. Fr. bour-geois, orig. -geis) — as were later to be in semantic content the analogically reshaped burgu-és ‘middle-class citizen’ and -esía — the exact relation (borrowing?) of OSp. bor-, bur-zés to OProv. borzés awaits definitive clarification, on the scale of a full-fledged word biography. The preliminary analysis here subscribed to reckons with the autochthonous status of OSp. borzes, thus implying some kind of polygenesis of the -rz- < -RǴ- cluster in the two languages.

21 Thus, Ford, Old Spanish Readings, p. xxxiv, remarking on the unexpected loss of -g- before a and o in certain verb forms (humear ‘to smoke, steam’ < FŪMIGĀRE, liar ‘to tie’ < LIGĀRE, lidiar ‘to tie’ < LĪTIGĀRE, rumiar ‘to ruminate’ < RŪMIGĀRE: also 1 sg. lío, lidio, etc.), wondered “whether the loss of the g did not commence in them with the forms of the verb whose ending began with e: LIGENT > líen.” On the more modern concept (H. Lausberg, R. L. Politzer) of détresse morphologique, see C. Blaylock, RPh, XVIII (1964-1965), 267. Morphological conditions as possible determinants of sound changes are very clearly isolated by R. Posner in her critique of O. Nandriş's Rumanian studies and in her skillful arbitration of the recent controversy between L. Romeo and F. Schürr anent Romance diphthongization (see RPh, XIX [1965-1966], 450-459, esp. 454 and 457ff.).

22 O. J. Tallgren[-Tuulio], Estudios sobre la “Gaya” de [Pero Guillén de] Segovia (Helsinki, 1907), p. 83, §§ 23f. Strictly speaking, Tallgren posited the joint influence of (a) dezir, -duzir ‘to lead,’ luzir ‘to shine,’ nuzir ‘to harm’ < NOCĒRE, plus arrezir ‘to grow stiff with cold’ (based on rezio) and (b) esparzir, estarzir < EXTERGERE, unzir, also sar-, sur-zir ‘to darn’ = mod. zurcir (cf. Ptg. serzir) on certain infinitives which might otherwise have yielded -cir; but since, of the two groups implicated, only (a) is entirely transparent in its phonological behavior, one may modify his hypothesis to the effect that Group (b) was itself deflected from its initially autonomous course by Group (a) before the two groups, joining forces, brought their combined pressure to bear on other verbs in the fifteenth century.

23 Thus, the equation senzi(e)llo < SING-ULU, *-ELLU, introduced by J. Cornu (1880) in lieu of Diez's untenable *SIMPLICELLU, caused considerable embarrassment to the pioneers; the background of arcén ‘border, edge, brim’ (in particular, its conceivable relation to a by-form of AGGER ‘rampart’) provoked hot discussion (Diez, S. Bugge, and others); the wisdom of operating with *RICIDU or *RECIDU, in preference to RIGIDU, was seriously weighed (e.g., by E. Gorra in 1898 and by Tallgren in 1907), inviting an occasional side glance at Alb. rekethe, among others.

24 This statement applies to A. Horning (1883), to J. Saroïhandy (1902), and to such other students of Old Spanish sibilants as moved in too narrow a groove.

25 Compare the relevant treatises of C. Michaëlis [de Vasconcelos] (1876), P. Förster (1880), R. J. Cuervo (1895). A. Zauner, in Altspanisches Elementarbuch (Heidelberg: C. Winter, 1908), though perfectly aware of the coexistence of (a) estreñer ‘to tighten’ < STRINGERE, lueñe, tañer (§ 50) and (b) enzía, senziello beside esparzer and burzés (§ 71), was worried, not about this major discepancy, but about a minor detail: the appearance in a few texts of ç for z (erçer, onçeja and, conversely, arzón, arienzo), a point no longer regarded as quite so troublesome after the publication of Menéndez Pidal's Orígenes del español (see the introductory chapter on spelling).

26 Grammatik der romanischen Sprachen (Leipzig, 1890), § 499.

27 Manual (elemental) de gramática histórica española, 2d ed. (Madrid: V. Suárez, 1905; 3d ed., 1914; 4th ed., 1918), § 47.2 b; retracted in Orígenes del español, § 49.3, in the light of the newly discovered var. reñilla. Compare the cautious acquiescence of F. Hanssen, Spanische Grammatik auf historischer Grundlage (Halle: M. Niemeyer, 1910), § 19.8, and his even greater restraint in Gramática histórica de la lengua castellana (Halle: M. Niemeyer, 1913), § 134, as against the unqualified acceptance by V. García de Diego, Elementos de gramática histórica española (Burgos: Tipografía de “El Monte Carmelo,” 1914), p. 44, a passage criticized by A. Castro one year later (see below); one may add, in vain, since the erroneous appeal to the stress still haunts the incorrigible author's Gramática histórica española (Madrid: Editorial Gredos, 1951), § 19.5, where he stubbornly parts company with Menéndez Pidal's second thinking. Menéndez Pidal's initial hypothesis fits into a pattern of conjectures suggested, at the turn of the century, by the unwarranted reverberations, in Romance, of Karl Verner's law (1876); see my note: “Quelques fausses applications de la loi de Verner aux faits romans,” in the A. Burger Testimonial Issue of the Cahiers Ferdinand de Saussure, XXIII (1966), 75-87.

28 “Die arabischen Laute im Spanischen,” RF, IV (1891), 401-403.

29 Orígenes del español (Madrid: Imprenta de la librería y casa editorial Hernando (s.a.), 1926; 2d ed., 1929; 3d ed., 1950), § 49.4.

30 Note the inordinately heavy reliance on Menéndez Pidal in G. B. Pellegrini, Grammatica storica spagnola (Bari: “Leonardo da Vinci” editrice, 1950), §§ 73.3 and 74.7.

31 On Ptg. iguaria ‘tidbit,’ OJud.-Sp. yegüería ‘mess, dish’ < (gloss) IĔCUĀRIA ‘giblets,’ also on Ptg. irmão = OSp. ermano < GERMĀNU and on the proper name Elvira see my papers in Lang., XX (1944), 108-130 and XXI (1945), 264-265, as well as R. Lapesa's comment in Asturiano y provenzal en el Fuero de Avilés (Salamanca: Universidad de Salamanca, 1948; Acta Salm., II:4), pp. 31-33, and E. Alarcos Llorach's reaction to these statements toward the end of the article listed in the following footnote.

32 “Resultados de Ge,i en la Península,” Arch., IV (1954; Miscelánea . . . Amado Alonso), 330-342. This paper clears up two points: (a) the mutual resemblance of Portuguese and Catalan and their joint departure from Castilian — a state of affairs corroborating Menéndez Pidal's general findings of 1926; (b) the marked inner consistency (i.e., regularity) of the Peninsula's lateral languages as against the extraordinary complexity of the centrally located language. Particularly praiseworthy is the distinction drawn between */j/ and */jj/ through inner reconstruction, in the absence of reliable graphemic evidence. The author is laudably cautious in assessing the Mozarabic material, seen chiefly through the prism of toponyms, and contributes a lexical vignette on the transmission of GYPSU.

33 See p. 338 n. 24, where esparzer, arzilla, enzía, on the one hand, and orçuelo ‘sty’ < HORDEOLU, vergüença ‘shame’ < VERĒCUNDIA, on the other, are subsumed under a single head, despite the clear indication, through the z:ç contrast, that two entirely different processes are involved. I have found little additional enlightenment in the author's Fonología española (3d ed. [Madrid: Editorial Gredos, 1961], pp. 229-232, 234, 244, and esp. 251-256), where Alarcos exacerbates his error by conjoining ARGILLA > OSp. arzilla and RADIU, *-A ‘ray’ > raça ‘light stripe, crack, slit, cleft.’ The book must be judged in the light of D. Catalán's strictures in RPh, XVIII (1964-1965), 178-191. Alcarcos disregards my study of VERĒCUNDIA (SP, 1944) and of -DJ- > -ç- (UCPL, XI [1954], passim). The recently published 4th ed. of the Fonología has not yet become available to me.

34 Introduction à l'étude de la philologie hispanique, I (Paris: B. Pottier, I960), § 75: tañer, uña, and § 119f.: arcilla, encía, (y)ermano (note the “Remarque”).

35 Romanische Sprachwissenschaft, II: Konsonantismus (Berlin: de Gruyter, 1956), §§ 410, 417.

36 For a fuller treatment of this multipronged issue see my article “Diphthongization, Monophthongization, Metaphony” in Yuen Ren Chao Testimonial Issue of Lang. XLII (1966), 430-472, and my forthcoming contribution to the Mélanges Jean Frappier.

37 We shall not here be concerned with the rival monophthongization > é after a palatal consonant, as in OSp. dix(i)eron ‘they said,’ mod. dijeron, beside tejieron ‘they wove’ (analogically remodeled), or as in amari-llento ‘yellowish’ alongside gris-iento ‘grayish’ (suff. -iento < -ENTU).

38 The fact that in STŎREA the expected *ue would have been a product of Ŏ + (attracted) j rather than of Ŏ alone and would have presupposed an older stage *oi (cf. CORIU ‘leather’ > Ptg. coiro beside Sp. cuero, more advanced) does not make the example any less relevant.

39 It is not impossible that mod. pértiga ‘long rod or pole,’ a notorious crux, has evolved from far more transparent OSp. piértega < PERTICA through the intermediate stage pírtega, with subsequent vowel metathesis. In this event the equally baffling Ptg. pírtiga ‘pole, shaft’ could conceivably be explained as a word transplanted from Spanish soil (e.g., by seasonal workers using this tool in harvesting fruit). In my aforecited article (Language, 1966) an attempt is made to demonstrate that the familiar conjugational series, yxo ‘I go out’ < EXEŌ beside inf. exir, sigo ‘I follow’ < SEQUOR beside seguir, sirvo ‘I serve’ < SERVIŌ beside servir, visto ‘I clothe, don’ < VESTIŌ beside vestir, involves at least as strong a dosage of monophthongization of older -ie- forms as it does of metaphony.

40 It is difficult to decide to what, if any, extent the s of siella was responsible for the word's transmutation into silla. It is hazardous to separate si(e)lla from si(e)glo; but note that siervo ‘serf’ < SERVU successfully withstood the pressure, while sirvo ‘I serve’ < SERVIŌ owes its i mainly, if not exclusively, to metaphonic influence. In isolation the s- of sie- certainly lacked the force to raise the front vowel, as is exemplified by siete ‘seven’ < SEPTE(M).

41 Orígenes del español see 3d ed., § 27, with a masterly summary and a refreshing nuancing of earlier methodology in subsection 5. The author discriminates sharply between the genesis of i < ie and its subsequent standardization. Isolated instances of the secondary monophthong are traceable to notarial documents of the tenth and eleventh centuries, from Old Castile; but when the literary language crystallized in the following century, it was Leon. -iello rather than Cast. -illo that initially won out in the unified scripta, while diversified dialects each followed its own course. Only in the fourteenth century did writers and refined speakers reverse themselves, in favor of -illo. Aragonese and Mozarabic speech communities adopted the monophthong more slowly and, above all, more hesitantly than did their Castillian counterparts. In modern Asturo-Leonese -iello predominates to this day, though certain subdialects use instead -iecho, -ietsu, -ichu, and (in the extreme West) -ello.

42 For a shrewd and conscientious appraisal of current inquiries, including a searching analysis of the monographs by Bengt Hasselrot and F. González Ollé, see J. R. Craddock, “A Critique of Recent Studies in Romance Diminutives,” RPh, XIX (1965-1966), 286-325.

43 It is worth noting that in Rumanian word-initial /j/ and labials (or labiovelars) often produce the same effect on the stem vowel: cf. iapă ‘mare’ < EQUA, pl. iepe, and pară (f. sg.) ‘pear’ < PIRA (n. pl.), pl. pere.

44 At proof I can, in some instances, supply fuller bibliographic references and, in others, add the barest minimum of fresh supplementary information. Ad. p. 25: concerning points (1) and (2), see my article “Range of Variation as a Clue to Dating,” to appear in the May, 1968, issue of RPh. Ad. p. 27: Point (6) has been made the specific object of a searching inquiry in my article identified in n. 18, below. Ad. p. 28: for further details see Lang., XLIII (1967), 242-245. Ad. p. 36: on metaphony see pp. 56-58, below, and the latest literature (1966) there adduced. Ad. pp. 55-63: for a far more elaborate and finely nuanced account of the problem here merely sketched in rough outline, see my paper “Le nivellement morphologique comme point de départ d'une loi phonétique: La monophtongaison occasionnelle de ie et ue en ancien espagnol,” to appear in the Mélanges Jean Frappier, a newly announced venture of the Parisian Klincksieck firm. Ad. pp. 58-59: OSp. piértega > mod. pértiga poses many additional problems, of which I grew aware after concluding this paper; a few of them are taken up in my article “Latin pedica, *pēnsum, and pertica in Hispano-Romance,” to appear in the Swedish miscellany Mélanges Alf Lombard. Ad. p. 60: the situation in Asturian is further complicated by the sporadic infiltration of Cast. -illo, disguised as -illu; see my forthcoming paper “Patterns of Derivational Affixation in the Cabraniego Dialect of East-Central Asturian,” to appear as part of a volume planned for the University of California Publications in Linguistics series. Ad. p. 63: I expect to show in the nearest future that at least two more phenomena usually tagged phonological, (a) the so-called parasitic /j/ of Asturo-Leonese and adjoining dialects (e.g., muriu ‘wall’ as against Sp. muro) and (b) the Castilian shift /s/ to /θ/ before consonant (as in bizco ‘squint-eyed,’ mezclar ‘to mix’), likewise have their roots in morphological conditions.