Download as pdf or txt
Download as pdf or txt
You are on page 1of 34

This article was downloaded by: [Paul O'Neill]

On: 10 July 2012, At: 15:21


Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Bulletin of Spanish Studies: Hispanic


Studies and Researches on Spain,
Portugal and Latin America
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/cbhs20

New Perspectives on the Effects of Yod


in Ibero-Romance
a
Paul O'Neill
a
University of Sheffield

Version of record first published: 04 Jul 2012

To cite this article: Paul O'Neill (2012): New Perspectives on the Effects of Yod in Ibero-Romance,
Bulletin of Spanish Studies: Hispanic Studies and Researches on Spain, Portugal and Latin America,
89:5, 665-697

To link to this article: http://dx.doi.org/10.1080/14753820.2012.696175

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-


conditions

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation
that the contents will be complete or accurate or up to date. The accuracy of any
instructions, formulae, and drug doses should be independently verified with primary
sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
Bulletin of Spanish Studies, Volume LXXXIX, Number 5, 2012

New Perspectives on the Effects of


Yod in Ibero-Romance
PAUL O’NEILL
University of Sheffield

1 Introduction
During the passage of Latin to Romance, and as a result of the change in
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

the nature of the accent the hiatus between vowels collapsed. When this
breakdown of hiatus coincided with a sequence of two vowels, the first being
a front vowel and the second any unstressed vowel, the result was that the
first vocalic element became the palatal glide [j], known as yod. This sound,
in the spoken language of the late empire, triggered in turn a number of
phonological changes in either the preceding consonant (usually through a
process of assimilation in which the consonant became palatalized and also
could have its manner of articulation affected), or the vowel in the preceding
syllable (via metaphony and/or metathesis). The effects in Spanish of yod
on the preceding consonant and vowel are of a complex and varied nature.1
In Table 1 there is listed, in summary, the changes which took place in the
consonant due to a following yod, in its passage from Latin to old Spanish,
alongside the attested presence or lack of metaphonic effects on the
preceding vowels.2

1 Cf. Ramón Menéndez Pidal, Manual de gramática histórica española (Madrid:


Espasa-Calpe, 1941), 290!91; Ralph Penny, A History of the Spanish Language (Cambridge:
Cambridge U. P., 2002), 62!65.
2 There are two question marks in this table which require explanation. The first
corresponds to the context of the consonant /n/ " [j] when preceded by the vowel /&/ and the
second to the consonants /g/ or /d/ " [j] when preceded by /o/. For both these contexts the
metaphonic effects of yod are unclear. In the first context the form sueño B SOMNIU would
suggest that the yod did not have a metaphonic effect upon the following vowel; this form,
however, could be derived from the form SOMNU in which the yod is not present (cf. Italian
sonno). For the second context, Penny notes the lack of available examples and that the verbal
form huyo B FUGIŌ, which is very often proposed in support of the effects of metaphony, is
problematic, due to the tendency of the verbs in -ir to display a high vowel in the root (cf.
Penny, A History of the Spanish Language, 50).

ISSN 1475-3820 print/ISSN 1478-3428 online/12/05/000665-33


# Bulletin of Spanish Studies. http://dx.doi.org/10.1080/14753820.2012.696175
666
BSS, LXXXIX (2012)
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

Table 1
Cons. Osp. Msp. /a/ /o/ /&/ /e/ /o/ Examples

/t/, /k/, "consonant #/ts/ /u/ no no no no no MARTIU !marzo CALCEA !calza


/t:/, /k:/ "vowel #/dz/ /u/ PUTEU !pozo ĒRĪCIU !erizo
/l/ /a/ /x/ no no yes no no ALIU!ajo; MELIŌRE !mejor; FOLIA !hoja;
CILIA !ceja; CUSCULIU !coscojo
/n/ /D/ /D/ no yes ? no yes ARĀNEA!araña; INGENIU !engeño,
LIGNU !leña; PUGNU !puño
/d/, /g/ "back vowel / / /5/ no yes yes no ? RADIĀRE !rayar; HODIE !hoy; CORRIGIA !correa;
"front vowel Ø Ø VIRIDIA !berza
"consonant /ts/ /u/
/b/, /m/ no change [bj], [mj] no yes yes yes yes RABIA!rabia; SUBERBIA !soberbia; NOVIU!novio,
VINDĒMIA !vindimia; RUBEU !rubio
/p/, /s/, / / metathesis of yod and /p/, /s/, / / yes yes yes no yes BĀSIU!beso; CERESIA !cereza; CERVĒSIA !cerveza;
metaphonic effect on vowel MANSIŌNE !mesón

PAUL O’NEILL
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 667

The examples given are all taken from the nominal paradigm, since within
the verbal paradigm there is some debate with regard to the effects
of yod. When the yod formed part of the lexical root of the verb the
developments are in line with that attested for nouns, e.g., the verb rayar
‘to scratch’ BRADIĀRE in which the root of the verb is invariable throughout
the paradigm. However, when yod formed part of the desinence its effects
vary from being sporadic or irregular to non-existent. The aim of this article is
to reassess the accepted beliefs regarding the effects of yod, or lack of
them, in the verbal paradigm of Spanish. Before entering into a full-scale
discussion on this matter, the distribution of desinential yod in Latin ought to
be detailed. For this it is necessary to discuss the conjugational classes of
Latin, since the reflexes of yod are intimately linked to the conjugation of
the verb.
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

1.1 The Latin Verb Conjugations, the Sound [j] and Its Effects on
Spanish Verbs
Latin verbs are considered to fall into four different conjugations which
determined the inflectional endings of each different verb form; thus, the
infinitives for the different four conjugations respectively end in -ĀRE,
-ĒRE, -ĔRE and -ĪRE. The only tenses which are of interest for the present
discussion are those of the present. For first conjugation verbs the present
indicative was formed in the first person singular by the concatenation of
-Ō- directly following the lexical root; in the other inflectional forms of this
tense the root was followed by the conjugational vowel A/Ā and the markers
of person and number. In the present subjunctive, no conjugation marker
was present and the root was simply followed by the vowel E/Ē and the
markers of person and number. The verbs of both the -ĒRE and -ĪRE
classes differed from this schema in that their conjugational vowel was
present directly after the lexical root in all forms. Compared with first
conjugation verbs, this meant that the conjugation vowel occurred in all of
the present subjunctive before the vocalic marker of subjunctive and also in
the first person singular present indicative before the final vowel -Ō-. Thus,
in these cells the unstressed conjugation vowel would become the sound [j].
This is illustrated in Table 2 for the verbs CANTĀRE ‘sing’, TIMĒRE ‘fear’,
METĪR(I) ‘measure’, where the cells in which the sound [j] is predicted are
shaded.
668 BSS, LXXXIX (2012) PAUL O’NEILL

Table 2
1st conjugation 2nd conjugation 4th conjugation

Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive

1SG CANTŌ CANTEM TIMEŌ TIMEAM MĒTIŌ MĒTIAM


2SG CANTĀS CANTĒS TIMĒS TIMEĀS MĒTĪS MĒTIĀS
3SG CANTAT CANTET TIMET TIMEAT MĒTIT MĒTIAT
1PL CANTĀMUS CANTĒMUS TIMĒMUS TIMEĀMUS MĒTĪMUS MĒTIĀMUS
2PL CANTĀTIS CANTĒTIS TIMĒTIS TIMEĀTIS MĒTĪTIS MĒTIĀTIS
3PL CANTANT CANTENT TIMENT TIMEANT *[metent]3 MĒTIANT

The third conjugation verbs (those with the infinitive ending -ĔRE) could be
divided into two subclasses:

(a) Verbs whose present subjunctive forms were marked by the vowels
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

I"A/Ā before the markers of person and number and by the ending
-IŌ for the first person singular present indicative.
(b) Verbs whose present subjunctive forms were marked only by the
vowel A/Ā before the markers of person and number and the form
of the first person singular present indicative ended only in -Ō,
immediately after the lexical root.

I shall refer to those verbs of the type (a), which are sometimes called
fifth conjugation verbs, as verbs of the 3a conjugation; and those of type (b)
verbs of the 3b conjugation. An example of the former is the verb RECIPERE
‘recover’, and of the latter is INCENDERE ‘set alight’, as illustrated in
Table 3.

Table 3
3a conjugation 3b conjugation

Indicative Subjunctive Indicative Subjunctive

1SG RECIPIŌ RECIPIAM INCENDŌ INCENDAM


2SG RECIPĪS RECIPIĀS INCENDIS INCENDĀS
3SG RECIPIT RECIPIAT INCENDIT INCENDAT
1PL RECIPĪMUS RECIPIĀMUS INCENDIMUS INCENDĀMUS
2PL RECIPĪTIS RECIPIĀTIS INCENDITIS INCENDĀTIS
3PL *[rekipent] RECIPIANT *[iEkendent] INCENDANT

3 This form is reconstructed from the proto Ibero-Romance forms with the desinence
-ENT (cf. Classical Latin desinence -UNT/-IUNT). It is hypothesized that the great majority of
Romance variants, with the notable exception of standard Italian, must be derived from such a
desinence, presumably through analogy with the other present indicative forms and in
particular the third person singular form.
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 669

It is apparent that all the verbs except the first conjugation and the 3b
conjugation display a desinence which contains unstressed E or I in hiatus
precisely in the same cells: the first person singular present indicative and
all of the present subjunctive. I shall refer to these cells (which have been
shaded in the tables above) as the L-pattern cells, for reasons which I shall
explain at a later stage. As stated, this unstressed vowel common to these
cells would have become the sound [j] in Romance and, accordingly, all verbs
of the second, fourth and 3a conjugations ought to display root allomorphs
in accordance with the sound changes in Table 1 precisely in the L-pattern
cells. Such developments, however, have clearly not taken place since an
application of the rules in Table 1, along with other regular sound changes to
the Latin forms in Tables 2 and 3 would yield the paradigms in Tables 4
and 5, which, with the exception of the reflexes of those verbs which did not
contain yod, CANTĀRE and INCENDERE, are all unattested.
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

Table 4
1st conjugation 2nd conjugation 4th conjugation

Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive

1SG canto cante timio timia mezo meza


2SG cantas cantes temes timias medes mezas
3SG canta cante teme timia mede meza
1PL cantamos cantemos tememos timiamos medemos mezamos
2PL cantáis cantéis teméis timiáis medéis mezáis
3PL cantan canten temen timian meden mezan

Table 5
3a conjugation 3b conjugation

Indicative Subjunctive Indicative Subjunctive

1SG recibio recibia enciendo encienda


2SG recebes recibias enciendes enciendas
3SG recebe recibia enciende encienda
1PL recebemos recibiamos encendemos encendamos
2PL recebéis recibiáis encendéis encendáis
3PL receben recibian encienden enciendan

Here the yod either (a) remains in the desinences and has a metaphonic
effect on the preceding vowel or (b) is absorbed by the consonant, altering its
pronunciation.
On account of the lack of attested effects of yod in the verb it has been
hypothesized that the etymological yod must have been eliminated at some
670 BSS, LXXXIX (2012) PAUL O’NEILL

point in the passage from Latin to old Spanish. The motivation for this loss of
-I- in late Latin is analogy with (a) Latin first and third conjugation verbs
which lacked this sound and (b) the other forms of the present tense of the
verbs which etymologically displayed this sound in the L-pattern cells but
whose other more frequent forms of the present indicative lacked it.4 This is
the accepted opinion of the L-pattern reflexes of Latin FACIŌ ‘I do’ and IACEŌ
‘I lie’, whose early attested Castilian forms are fago and yago for the first
person singular present indicative and faga, yaga for the present subjunctive;
this is instead of the expected faço, iaço, faça, iaça, which would suggest the
loss of yod at an early stage resulting in the preliterary forms [ ako] and [fako],
from which the attested forms can be derived via regular sound change.5
This loss of yod, however, is best exemplified by the Spanish -er verbs,
which contain reflexes of both the Latin 3a and 3b conjugations; the former
being etymologically expected to contain yod. In these verbs the glide is
considered to have been eliminated totally, since not only are there no
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

attestations in Castilian -er verbs of its effect on the preceding consonants


(respondo BRESPONDEŌ,6 muerdo BMORDEŌ, suelo BSOLEŌ, duelo
BDOLEŌ, hueloBOLEŌ, puedo B*[p&djo]), but it fails to raise the pre-
ceding vowel metaphonically (see previous examples and muevo BMOVEŌ,
temo BTIMEŌ, debo BDĒBEO).7
The Spanish !ir verb class is different, since a number of these verbs
do have indications of some preliminary effects of yod. In these verbs the
situation is complex since there is a total lack of evidence of the effects
of yod on the consonants (mido BMĒTIO, recudo BRECUTIŌ, sacudo
BSUCCUTIŌ); but the presence of a high vowel in a number of inflec-
tional forms for the verbs of this conjugation (see previous examples and
reciba BRECIPIAM, concibamos BCONCIPIĀMUS, hinchan BIMPLEANT,
cumplo BCOMPLEŌ), would suggest the existence of the sound yod. The
general opinion is, therefore, that [j] was present at an early stage in the
L-pattern cells and thus could exercise a metaphonic effect upon the pre-
ceding vowel but was deleted before it could exercise its influence on the
consonants.
The effects of yod on the previous consonant are not non-existent, and are
attested by a number of lexemes8 whose root, in Latin, ended either in a
voiced velar or dental consonant (/d/ or /g/), and, in these cases, the glide

4 Cf. Penny, A History of the Spanish Language, 174.


5 Cf. R. Menéndez Pidal, Manual de gramática, 292!93; Penny, A History of the
Spanish Language, 174.
6 The lack of diphthongization in the rhizotonic forms of this lexeme has been
attributed to the vowel being in a closed syllable with a nasal consonant in the coda; cf.
also EXPENDŌ ! expendo (cf. Marı́a Jesús López Bobo, El vocalismo radical átono en la
conjugación castellana: etapa medieval y clásica [Oviedo: Univ. de Oviedo, 1998], 33).
7 Cf. López Bobo, El vocalismo, 39; Menéndez Pidal, Manual de gramática, 292; Penny,
A History of the Spanish Language, 174.
8 AUDIŌ ‘I hear’, VIDEŌ ‘I see’ and RĪDEO ‘I laugh’, SEDEŌ ‘I sit’ and FUGIŌ ‘I flee’.
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 671

produced the expected results upon the preceding consonant. In Old Spanish
these verbs displayed a root ending in the palatal fricative consonant [ ] (the
regular outcome of /d/"[j] and /g/"[j] (see Table 1) precisely in the L-pattern
cells. This is illustrated in Table 6 for the old Spanish present tense
paradigms of the verbs oı́r ‘hear’, ver ‘see’ and reı́r ‘laugh’ (AUDEAM !
[odja] ![o a]!oya; VIDEAM ![vedja] ![ve a] !veya; RĪDEAM ![ridja] !
[ri a] !riya):

Table 6
Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive

1SG oyo oya veyo veya riyo riya


2SG oes oyas vees veyas ries riyas
3SG oe oya vee veya rie riya
1PL oimos oyamos veemos veyamos riemos riyamos
2PL oides oyades veedes veyades riedes riyades
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

3PL oen oyan veen veyan rien riyan

The effects of yod are also attested in the verb caber ‘fit’ and, in this case,
the particular allomorphy has persisted into modern Spanish in which the
root allomorph quep- is restricted to the L-pattern cells, as demonstrated in
Table 7.

Table 7
Classical Latin Modern Spanish

Indicative Subjunctive Indicative Subjunctive

1pl. CAPIŌ CAPIAM quepo quepo


2pl. CAPIS CAPIĀS cabes quepas
3pl. CAPIT CAPIAT cabe quepa
1pl. CAPIMUS CAPIĀMUS cabemos quepamos
2pl. CAPITIS CAPIĀTIS cabéis quepáis
3pl. *[kapent] CAPIANT caben quepan

In the present case the allomorph quep- was the result of the metathesis of
the glide to the previous syllable, the absorption of the glide by the vowel and
the concomitant change in the vowel quality: CAPIAM [kapjam] !
[kaipa] !quepa. This same process is also the cause of the present sub-
junctive form sepa BSEPIAM of the verb saber ‘to know’. However, in this
verb the first person singular present indicative form is sé, most probably
through analogy with he of haber ‘have’.
Thus, the story which is related in the manuals of the history of Spanish
is that yod was deleted in -er verbs, survived long enough in -ir verbs to have
a metaphonic effect upon the vowels of the preceding consonant and was
entirely maintained in a selection of verbs ending in Latin /d/ and /g/ and
672 BSS, LXXXIX (2012) PAUL O’NEILL

also saber, caber and possibly haber,9 whence the attested phonological
developments.10
The principal problem with this widely accepted account is that it is
based on retrospective observation and thus the criterion for deciding
between the different possibilities (deletion, partial maintenance, full
maintenance) is circular, in that it is assigned retrospectively on the basis
of the modern-day/old Spanish forms. Thus, for the Latin consonant
-V-"[j], the glide is considered to have disappeared in MOVEĀMUS
! !11 [mobamos] !movamos, whilst it metaphonically raised the vowel
in FERVEĀMUS !hirvamos. The explanations are even more ad hoc
with Latin D"[j], since it is considered to have been eliminated in
MORDEŌ !![m& do] !muerdo, had a metaphonic effect on the pre-
ceding vowel only in ŌRDIO(R) ![o djo] !urdo, and had an assimilatory
effect on the preceding consonant in RĪDEŌ![ridjo] ![ri o] !riyo !rı́o).
The fact that all these verbs, except urdir ‘warp thread’, have their etyma
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

in the same Latin conjugation is ignored, since the guiding principle


regarding the loss of yod depends largely on the conjugation into which the
verb passes in Spanish. This, to my mind, is to reverse the relationship
between cause and effect and suggests that the consequences of yod are a
matter of morphology and analogical levelling associated with conjugational
class and not a mere question of retention or deletion of the sound. Thus,
whilst it is feasible that yod could have been lost in the transition from Latin
to Spanish,12 its noted effects on the sounds /d/ and /g/ in the verb along with
the lexemes saber and caber, and the metaphonic effects upon -ir verbs,
suggest that yod had much more of an effect upon verbs than the grammar

9 Both Penny, A History of the Spanish Language and Menéndez Pidal, Manual de
gramática defend the sound change /b/"[j] ! [ ] and partly ascribe the Spanish reflexes of
Latin HABERE to this change. Outside the verbal paradigm, however, such a phonological
rule is only attested in the toponym Peñarroya B PINNA RUBEA (cf. Penny, A History of the
Spanish Language, 65). Malkiel notes that ‘Scholars are agreed on the absence of a direct link
from HABEAT to Ptg. [# Portuguese] haja, and Sp.[anish] haya. but there is no unanimity of
opinion on the best possible formulation of the state of affairs’ (cf. Yakov Malkiel, ‘New
Problems in Romance Interfixation’, Romance Philology, 27 [1973], 304!55 [n. 59]). He
suggests that the voiced bilabial may have been lost, and then suggests a possible analogy
with the regular old Spanish and old Portuguese continuants of SEDEAT ! seja and seya
respectively.
10 Cf. Menéndez Pidal, Manual de gramática, 292!93; Penny, A History of the Spanish
Language, 174.
11 I use double arrows to represent analogical change and a single arrow to represent
regular sound change.
12 Especially considering that even in Latin there was some variation as to whether the
yod was present or not in a number of lexemes (thus FERVŌ - FERVĒŌ, ŌLO ! OLEŌ; cf.
Menéndez Pidal, Manual de gramática, 292) and that in Romance this tendency could have
been enhanced (cf. TONDEŌ vs. TONDŌ, SORBEŌ vs. SORBŌ, PENDEŌ vs PENDŌ; cf.
Thomas Montgomery, ‘Sources of Vocalic Correspondences of Stems and Endings in the
Spanish Verb’, Hispanic Linguistics, 2 [1985], 99!114 [pp. 110!11]).
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 673

books propose, and that this question is in need of more detailed research.
In what follows I shall draw upon data within Spanish but primarily from
closely related Romance varieties to suggest that yod had a similar effect on
verbs as on nouns.

2 A Comparative Approach
In the previous section the contradictions and stipulations regarding the
presence or lack of yod and its concomitant effects upon both the preceding
vowels and consonants were highlighted. I would suggest that in the
spoken Latin of the Iberian Peninsula the etymological yod had equal
effects upon verbs as upon nouns, as the principle of the regularity of sound
change would lead us to expect,13 and thus there was a large number of
verbs which displayed allomorphy precisely in the L-pattern cells.
Corroboration for this hypothesis comes from a wide number of other
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

Romance varieties14 but in particular from the neighbouring Romance


variety, Portuguese, which contains verbs which display a special type of
allomorphy exclusively in the L-pattern cells, and whose forms are in
accordance with the presence and effects of an etymological yod. These
verbs are given in Table 8.

Table 8
Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive

1pl. faço faça meço meça valho valha


2pl. fazes faças medes meças vales valhas
3pl. faz faça mede meça vale valha
1pl. fazemos façamos medimos meçamos valemos valhamos
2pl. fazeis façais medis meçais valeis valhais
3pl. fazem façam medem meçam valem valham
Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive
1pl. venho venha vejo veja caibo caiba
2pl. vens venhas vês vejas cabes caibas
3pl. vem venha vê veja cabe caiba
1pl. vimos venhamos vemos vejamos cabemos caibamos
2pl. vindes venhais vedes vejais cabeis caibais
3pl. vêm venham vêem vejam cabem caibam

13 But Maiden gives examples of verbs behaving differently from nouns in respect of
regular sound change (Martin Maiden, Interactive Morphonology: Metaphony in Italy
[London: Routledge, 1991]).
14 Cf. Martin Maiden, ‘Morphophonological Innovation’, in The Cambridge History of
the Romance Languages, ed. Martin Maiden, John Charles Smith and Adam Ledgeway
(Cambridge: Cambridge U. P., 2011), 216!67.
674 BSS, LXXXIX (2012) PAUL O’NEILL

The interesting point about these verbs is that they display the Portuguese
results of the expected outcome of all the different types of consonantal
development of yod, as listed in Table 1. Thus /k/"[j] is attested in
FACIŌ !faço, and also the old form iaço BIACEŌ; /t/"[j] in METIŌ !
meço; /l/"[j] in valho BVALEŌ;15 /n/"[j] in tenhoBTENEŌ; /d/ "[j] in vejo
BVIDEŌ; /p/"[j] in CAPIŌ !caibo. Old Spanish contains remnants of the
latter development only for /p/"[j] in CAPIŌ !quepo and /d/"[j] (veya
BVIDEAM, seya BSEDEAM, oya BAUDIAM, riya BRĪDEAM) and /g/"[j]
(fuya BFUGIAM). However, on the basis of the comparative forms I would
suggest that in pre-literary Spanish the glide could also have had the predicted
effect upon the sounds /t k n l/. This would have produced a pattern of
allomorphy exclusively in the L-pattern cells for a number of verbs. An
example of such allomorphy is given in Table 9 for the supposed pre-literary
Spanish present tense forms of the Latin verbs METIŌ(R) ‘I measure’, FACIŌ
‘I do’, IACEŌ ‘I lie’, TENEŌ ‘I have’, SALIŌ ‘I leave’, VALEŌ ‘I am worth’.16
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

Table 9
Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive
' ' ' ' ' '
1SG [motso] [motsa] [fats
'
o] [fatsa] [ ats
'
o] [ atsa]
' ' '
2SG [motes] [motsas] [fadz
'
es] [fatsa] [ adz
'
es] [ atsas]
' ' '
3SG [mote] [motsa] [fadz
'
e] [fatsa] [ adz
'
e] [ atsa]
' ' '
1PL [metimos] [metsamos] [fadz
'
emos] [fatsamos] [ adz
'
emos] [ atsamos]
' ' '
2PL [metites] [metsates] [fadz
'
etes] [fatsates] [ adz
'
etes] [ atsates]
' ' '
3PL [moten] [motsan] [fadzen] [fatsan] [ adzen] [ atsan]
Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive
1SG [ eDo] [ eDa] [sa^o] [sa^a] [ba^o] [ba^a]
2SG [tones] [ eDas] [sales] [sa^as] [bales] [ba^as]
3SG [tone] [ eDa] [sale] [sa^a] [bale] [ba^a]
1PL [tenemos] [ eDamos] [salimos] [sa^amos] [balimos] [ba^amos]
2PL [tenetes] [ eDates] [salites] [sa^ates] [balites] [ba^ates]
3PL [tonen] [ eDan] [salen] [sa^an] [balen] [ba^an]

The interesting thing to note about these verbs is that, with the exception of
the verb medir ‘measure’, their modern forms all display an unexpected velar
allomorph precisely in the cells of the paradigm in which the regular effects
of sound change would have predicted the palatal/affricate pronunciations
(hago, yago/yazgo/yazco, tengo, salgo, valgo). This velar allomorph is also

15 Even though, following this argumentation, one would expect the reflexes of both
SALIŌ and DOLEANT to be salho, dolham instead of the attested saio and doam, Malkiel has
suggested that the latter could have been made regular through levelling (cf. Malkiel, ‘New
Problems in Romance Interfixation’, 325). Regarding the former, sallo is attested in old
Galician-Portuguese and also in varieties of Asturian-Leonese.
16 These are the respective etyma of the Spanish verbs medir ‘to measure’, hacer ‘to do’,
yacer ‘to lie postrate’, tener ‘to have’, salir ‘to leave’, valer ‘to be worth’.
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 675

attested in other Romance varieties, e.g., Italian, and its distribution in this
language is precisely in accordance with the cells in which yod was predicted
in the desinence.17 With regard to the origin and the extension of the velar
allomorph in Italian, Maiden notes that the velar first spread to forms which
in old Tuscan displayed palatal consonants as a consequence of yod.18 This is
illustrated in Table 10 in which the first person singular and the third person
singular present indicative of a number of verbs in both old Tuscan and
standard Italian are given and which display how the velar allomorph came
to replace an irregular allomorph caused by yod.

Table 10
Old Tuscan Modern Italian

verb 1SG PRS IND 3SG PRS IND 1SG PRS IND 3SG PRS IND

venire ‘to come’ vegno viene vengo viene


Downloaded by [Paul O'Neill] at 15:21 10 July 2012

valere ‘to be worth’ vaglio vale valgo vale


sedere ‘to sit’ seggio siede seggo19 siede

It ought to be noted that in old Spanish all the verbs which preserved20 the
allomorphy characterized by the palatal consonant / / at the end of
the root in the L-pattern cells, due in the most part to the effects of an
etymological desinential yod, were also subject to the extension of the velar
allomorph (see caya !caiga, oya !oiga, traya !traiga, roya!roiga, vaya !
vaiga, haya !haiga, destruya !destruiga, fuya !fuiga).21 Such coincidences
would suggest that the palatal/affricate allomorphs in Table 9 pre-date the
appearance of the velar allomorphs fago, yago, tengo, salgo and valgo. This
point is actually conceded for the sequences /n/"[j] by both Penny and
Menéndez Pidal.22 While they justify this on the basis of the Galician-
Portuguese forms teño-tenho, veño-venho and the old Italian forms tenho,
venho, the same consideration of Galician-Portuguese and old Italian forms
is not extended to /l/"[j] nor to /t/ or /k/"[j]. With regard to the latter, both

17 In Italian, in addition to the L-pattern cells, the yod was also present in the third
person plural present indicative (cf. footnote 3).
18 Cf. Martin Maiden, ‘Irregularity as a Determinant of Morphological Change’,
Journal of Linguistics, 28 (1992), 285!312; Martin Maiden, ‘Di nuovo sulle alternanze
velari nel verbo italiano e spagnolo’, Cuadernos de Filologı́a Italiana, 8 (2001), 39!61.
19 Whilst seggo is still attested in Italian and in earlier times was quite widely attested,
there is a tendency of speakers of contemporary Italian to use the form siedo.
20 The sound / / was only maintained after a non front vowel (fuya, oya, caya, traya);
when following front vowels it was subject to assimilation and deletion (seya ! sea, veya !
vea, riya ! rı́a).
21 In the modern language the velar has been lost from the forms vaiga, haiga,
destruiga, fuiga (cf. vaya, haya, destruya, huya).
22 Menéndez Pidal, Manual de gramática; Penny, A History of the Spanish Language.
676 BSS, LXXXIX (2012) PAUL O’NEILL

authors acknowledge the existence of the Portuguese form faço and jaço from
old Portuguese, but their claim is that the yod in such verbs and contexts in
Castilian was eliminated.23
I would suggest that such comparative evidence, along with the devel-
opment of the velar allomorphs in Spanish, whose extension in Italian was
intimately linked with those verbs which displayed yod-induced allomorphy,
is sufficient to assume that the effects of yod in pre-literary Spanish were
in accordance with those of regular sound change, as shown in Table 1. Such
a conclusion is also supported by attested Spanish verbs which display
invariable root allomorphs with palatal consonants, so that the explanation
for such consonantal forms can only be that they originated from the effects
of yod (e.g., puñirBPUNIŌ and muñirBMONEŌ).
The effects of yod which have been examined thus far are those in which
the glide has merged with the preceding consonant, thus changing its
pronunciation. For the sounds /p b m r s/ (see [Table 1]) the consonants are
left unchanged and the only expected development is either the maintenance
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

of the glide in situ or the metathesis of the yod, usually accompanied in both
cases with a concomitant change in quality of the preceding vowel due to
metaphony. Not only is evidence for the effects of metathesis present for the
allomorphs quep-, sep- of the verbs saber ‘know’ and caber ‘fit’ but the form
mojra from morir ‘die’ is also attested in a document from Burgos of 1217,
which is similar to old Portuguese and dialectal Galician moira and also old
French muire and Balearic Catalan muirga.24 There are no attestations in
Spanish to show that the glide remains in the desinences, not even after
voiced bilabial consonants. In what follows, I provide evidence from other
Ibero-Romance varieties which suggest that desinential yod was maintained
precisely in the L- pattern cells. This fact not only strengthens the hypothesis
that yod has had similar effects upon verbs in pre-literary Spanish but also is
of particular importance in explaining the effects of metaphony of yod upon
-ir verbs, as opposed to its effects upon the consonant.

2.1 Evidence for the Maintenance of Desinential YOD and


Its Metaphonic Effects
Remnants of desinential [j] in -er and -ir verbs can be found in the
varieties both to the west and east of the area in which Castilian was first
spoken. The best example is in the speech of El Bierzo in which the
desinential yod is present in all the L-pattern cells for -er and -ir verbs,
although these forms do alternate with desinences without yod.25 This is

23 Menéndez Pidal, Manual de gramática; Penny, A History of the Spanish Language,


174.
24 Cf. Ramón Menéndez Pidal, Documentos lingüı́sticos de España (Madrid: Imprenta
Nacional, 1919), 215.
25 Cf. Manuel Gutiérrez Tuñón, El habla de El Bierzo (León) (Ponferrada: Instituto de
Estudios Bercianos, 1986), 109!12.
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 677

illustrated in Table 11 for the -er verb correr ‘run’ and the -ir verb partir
‘leave’.

Table 11
Indicative Subjunctive Indicative Subjunctive

1SG corro ! corrio corra ! corria parto - partio parta ! partia


2SG corres corras - corrias partes partas ! partias
3SG corre corra ! corria parte parta ! partia
1PL corremos corramos - corriamos partimos partamos ! partiamos
2PL correis corrais - corriais partı́s partais - partiais
3PL corren corran ! corrian parten partan ! partian

In the majority of localities the desinential [j] is only present in the


first person plural and second person plural present subjunctive forms of
both -er and -ir verbs, as the different Asturian26 and Aragonese27 speaking
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

varieties testify. Also worthy of note is the fact that in certain varieties
of Asturian (Sobrescobio and Cabranes) the [j] is restricted to the subjunctive
cells only of -er and -ir verbs and in one variety (San Ciprián) it
has even extended into -ar verbs, to become a marker of the present
subjunctive.28
There is reason to believe that the [j] in all of these forms is not a new
development but related to the etymological [j] of spoken Latin, which was
once present in all of the L-pattern cells, given that in those varieties in
which it appears exclusively in the present subjunctive it prevents diph-
thongization, as can be appreciated in Table 12, in which the rhizotonic
forms of both third person singular present indicative and subjunctive are
given for the verbs correr ‘run’, vender ‘sell’, goler ‘smell’ and morrer ‘die’.
The lack of diphthongization in these varieties exclusively in the present

26 Valledor (Celso Muñiz, El habla del Valledor: estudio descriptivo del gallego de
Allande [Amsterdam: Academische Pers, 1978], 331!32); Alto Aller (Lorenzo Rodrı́guez
Castellano, La variedad dialectal del Alto Aller [Oviedo: Instituto de Estudios Asturianos,
1952]); Sayago (Julio Borrego Nieto, Norma y dialecto en el Sayagués actual [Salamanca:
Univ. de Salamanca, 1983], 25); Bimenes (Marı́a T. Cristina Garcı́a Álvarez, ‘Morfologı́a
verbal del bable de Bimenes’, Archivum, 10 [1960], 405!25 [p. 416]); Sobrescobio (Marı́a
Victoria Conde Sáiz, El habla de Sobrescobio [Mieres del Camino: Instituto ‘Bernaldo de
Quirós’, 1978], 177); Cabranes (Marı́a Josefa Canellada, El bable de Cabranes, Revista de
Filologı́a Española, Anejo 31 [Madrid: CSIC, 1944], 39); San Ciprián (Fritz Krüger, El dialecto
de San Ciprián de Sanabria, Revista de Filologı́a Española. Anejo 4 [Madrid: lmpr. de los
Sucesores de Hernando, 1923]).
27 Valle de Bielsa (Chabier Lozano Sierra y Ángel Luis Saludas Bernad, Aspectos
morfosintácticos del Belsetán: aragonés del valle de Bielsa [Zaragoza: Gara d’Edizions, 2007],
82); Baixo Peñas (Fernando Blas Gabarda y Fernando Romanos Hernando, El aragonés de
Baixo Peñas [Zaragoza: Gara d’Edizions, 2005], 25!26).
28 Cf. Conde Sáiz, El habla de Sobrescobio, 177; Canellada, ‘El bable de Cabranes’, 39;
Krüger, ‘El dialecto de San Ciprián’.
678 BSS, LXXXIX (2012) PAUL O’NEILL

subjunctive (and at times in the first person singular present indicative)


suggests that the [j] was present prior to the diphthongization of the
mid-open vowels of spoken Latin and was able to raise these vowels
metaphonically and thus exclude them from its effects.

Table 12
verb 3sg. pres. indic. 3sg. pres. subj.

correr cuerre corria


vender viende vendia
goler güele golia
morrer muere morria

Similar paradigms to those above are also attested in Somiedo where,


depending on the locality and the lexeme, the rhizotonic forms of either all
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

the L-pattern cells (case of morder),29 or exclusively the present subjunctive


cells (case of poder and querer)30 lack the diphthongized forms. The inter-
esting point to note about this variety is that, as in Spanish, there are no
vestiges of desinential [j]. The most logical explanation is that the yod was
present and had a metaphonic effect upon the vowels, thus preventing
diphthongization, but subsequently was lost.31 Such a hypothesis also
supports the persistence of yod in all the cells of the L-pattern of those
aforementioned varieties in which it is now only present in the first person
plural and second person plural present subjunctive.
I would also argue that such forms suggest that yod could have persisted in
the desinences of pre-literary Castilian verbs, where they were in the first
instance limited to verbs which displayed root final voiced bilabials but were
also subject to analogical extension. Indeed, it is my opinion that the L-pattern
desinences which contain etymological yod intact could provide an explana-
tion for the lack of the effects of yod on the preceding consonant and for the
attested metaphonic effects on the preceding vowel in Spanish -ir verbs.
Such a development is strange and at the same time rather contradictory,
since if we take the Latin verb MĒTIO(R) which very often is used as an
example of the aforementioned phenomenon, the regular effects of yod (see
Table 1) on the first person singular present indicative form of this lexeme
would have produced the form mezo (cf. Portuguese meço), in which there
ought to have been no raising of the vowel of the preceding syllable in
accordance with the attested lack of metaphonic effects upon the sequence /t/

29 Present Indicative: mordo, muerdes, muerde, mordimos, muerden; Present


Subjunctive: morda, mordas, morda, mordan (cf. Ana Marı́a Cano González, El habla de
Somiedo [Santiago de Compostela: Univ. de Santiago de Compostela, 1981], 156).
30 Cf. Cano González, El habla de Somiedo, 164.
31 The reasons as to why it was lost in this variety but remained in others are unclear.
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 679

"[j] (cf. VITIUM !vezo [old Spanish] ‘custom’). Therefore, it is difficult to


conceive how yod had such influences when, as noted in Table 1, the effects of
metaphony upon the preceding vowel are not extensive and are only really
prevalent in the cases in which the glide is maintained in the following
syllable (cf. RUBEU !rubio ‘blonde’);32 and thus the majority of verbs, whose
root vowels could be susceptible to raising would be those which could have
presented yod in the L-pattern desinences*that is, the verbs whose root
ended in the following consonants: /b m p s /. In what follows I shall offer a
possible explanation for this metaphonic raising.

3 The Metaphonic Effects of Yod on -ir Verbs


In accordance with the regular effects of yod, the situation in Ibero-Romance,
as hypothesized here, would be one in which, in addition to the allomorphic
roots as suggested in Table 9, there also existed allomorphic desinences for
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

the L-pattern cells, as shown by the proposed proto Ibero-Romance forms in


Table 13 of the continuants of the present tense forms of the verbs MORIŌ(R)
‘I die’, SERVIŌ ‘I serve’, and DORMIŌ ‘I sleep’:

Table 13
Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive

1SG [mo jo] [mo ja] [se bjo] [se bja] [do mjo] [do mja]
2SG [m& es] [mo jas] [so bes] [se bjas] [d& mes] [do mjas]
3SG [m& e] [mo ja] [so be] [se bja] [d& me] [do mja]
1PL [mo imos] [mo jamos] [se bimos] [se bjamos] [do mimos] [do mjamos]
2PL [mo ites] [mo jates] [se bites] [se bjates] [do mites] [do mjates]
3PL [m& en] [mo jan] [so ben] [se bjan] [d& men] [do mjan]

Given the lack of such allomorphy in the Castilian documentation, one


might imagine that analogical levelling was responsible for the eradication of
such patterns. With regard to the root allomorph the expected situation would
be for the L-pattern allomorph to be replaced with that of the rest of the
paradigm; thus for MĒTIŌ(R) the root [met' s ] would have been replaced by
[met]. What I am proposing is that, due to the allomorphic desinences for
the L-pattern cells of -ir verbs, there was confusion and perhaps variation,
whereby some verbs which etymologically did not contain a desinential yod
came to acquire one and some verbs in which yod was attested came to lose it.
Thus given the existence of two types of desinences, the reestablishment of the
root [met] for Latin MĒTIŌ ![met' s o] could have produced a proto-Castilian

32 Or by metathesis, when it is adjacent to the vowel. This latter process took place at a
relatively late stage in Old Spanish (cf. Penny, A History of the Spanish Language) subsequent
to diphthongization, and therefore at the pre-literary stage in which I am supposing the yod
would still have been present in the following syllable.
680 BSS, LXXXIX (2012) PAUL O’NEILL

form [meto] or [metjo]; from this latter form the desinential yod can exert its
effects upon the root vowel but the consonant could show resistance to this
process.33
Evidence for this hypothesis comes from the attested forms of the reflexes
of the L-pattern cells of a number of -ERE verbs, whose root allomorph
ended in /d/ and etymologically lacked a yod in these cells (e.g., CADŌ ‘I fall,
RĀDO ‘I scratch’, CREDŌ ‘I believe’, RŌDŌ ‘I gnaw’, VADŌ ‘I wander’).
The subsequent attestations of these verbs in a number of Romance
languages suggest that they acquired a yod relatively early: thus, CADAM !
[kadja] !caya ‘I believe’; CRĒDAM![k edja] !creya ‘believe’; RĀDĀM !
[radja] !raya ‘scrape’; RŌDAM ![rodja] !roya ‘gnaw’; VĀDAM!
[badja] !vaya ‘go’.34 Likewise the reflexes of the Latin verb PETŌ ‘I aim at’
and PONŌ ‘I put’ in Portuguese (peço, poño) both suggest a proto-form *[petjo]
and *[ponjo]; so, I would argue, do the Spanish forms, pido and pongo,
although in these cases the effects of yod are obscured by analogical processes.
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

This desinential yod was subsequently deleted in Castilian, most proba-


bly due to analogy with the great majority of other verbs which did not
display this element in the desinences. However, remnants of it survive in
lexemes in the -er and -ir conjugation in a number of both Asturian and
Aragonese varieties. This desinential yod could, before disappearing, exert a
metaphonic effect upon the preceding vowels but not alter the pronunciation
of the preceding consonant. Thus, I claim that the vacillation over desi-
nential yod persisted long after the consonantal palatalizations were done
with, but continued to produce metaphony.
The fact that yod was eradicated from the earliest attestations of the
Spanish language in a large number of verbs in which it was etymological,
most notably the -er verbs, might lead one to ask: what is the point of proposing
the existence of a now defunct pre-literary Castilian/proto-Ibero-Romance
verbal paradigm which contained the glide? I would argue that it is of great
importance, in order to understand the abstract workings of the morphological
system of the language, to appreciate that such patterns of allomorphy existed
and to comprehend what mechanisms Spanish had for dealing with the
allomorphy caused by yod. Also, by recognizing such pre-literary forms we are
both better able to integrate Castilian in a continuum with adjacent Romance

33 It ought to be noted that whilst it is problematic to suggest that verbs could be


resistant to the effects of regular sound change, in this particular case the effects of yod, such a
phenomenon is attested in -ar verbs whose lexical root ended in a velar consonant and which
did not undergo the palatalization effects of the following '
front vowels in the present
subjunctive e.g., PLICEM ! llegue, not the expected [^ETSE]. Cf. Maiden, ‘Di nuovo sulle
alternanze velari’.
34 Cf. Paul M. Lloyd, From Latin to Spanish. Vol. I: Historical Phonology and
Morphology of the Spanish Language (Philadelphia: American Philosophical Society, 1987),
296!97; Malkiel, ‘New Problems in Romance Interfixation’, 333; Penny, A History of the
Spanish Language, 174!75.
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 681

varieties, most notably Portuguese, and, in addition, to explain differences


between abstract morphological patterns of allomorphy in Spanish and
Portuguese. This will be the subject of the remainder of this article.

4 The Effects of Yod and Autonomous Morphological Structures


(MORPHOMES)
My claim is that, for proto-Ibero Romance, there was a semantically and
phonologically heterogeneous collection of cells, those of the L-pattern which
were the domain of various types of allomorphy of both the lexical root and
the desinences. This, in part, was an effect of yod, but another contributing
factor were the verbs of the Latin 3a conjugation (see above, section 1.1)
whose root ended in a velar occlusive of which DĪCERE ‘to say’ is the classic
example.35 These verbs displayed a different allomorph exclusively in the L-
pattern cells on account of the differential phonological developments of velar
occlusives in the history of Spanish (and Romance in general), depending
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

upon whether the consonant was followed by a back or front vowel. Before
back vowels these consonants remained velar (although /k/was subject to
voicing and /g/ to become an approximant or to be lost),36 whilst before front
vowels there were various developments, depending upon the phonetic
context.37 A selection of such allomorphy is given in Table 14 for the verbs
decir ‘to say’, esparcir ‘to scatter’ and ceñir ‘to gird’:38

Table 14
Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive

1SG digo diga espargo esparga cingo cinga


2SG dizes digas esparzes espargas ciñes cingas
3SG diz(e) diga esparz(e) esparga ciñe cinga
1PL dezimos digamos esparzemos espargamos ceñimos cingamos
2PL dezides digades esparzedes espargades ceñides cingades
3PL dizen digan esparzen espargan ciñen cingan

35 Other verbs with attested etymological velar allomorphs in Spanish are: lucir ‘to
shine’, frañir ‘to break’, esparcir ‘to scatter’, erguir ‘to erect’, tañir/tañer ‘to play’, ceñir ‘to gird’,
teñir ‘to dye’. To these lexemes two other significant groups of verbs ought to be added: (a)
those verbs which etymologically are derivatives of DUCERE: reducir ‘to reduce’ aducir ‘to
adduce’ etc., and (b) the verbs which contained the inchoative suffix -ESC-: conocer ‘to know’,
merecer ‘to be worth’ etc.
36 Cf. NEGĀRE ! negar ‘to deny’, LIGĀRE ! liar ‘to tie’ (cf. Penny, A History of the
Spanish Language, 68).
37 In Castilian, in the cases when the velar consonant was preceded by a vowel or the
'
consonant / /, it was subject to the normal process of affrication to [dz ] (cf. decir ‘to say’, yacer
‘to lie prostrate’, aducir ‘to adduce’); when preceded by a nasal consonant, however, the
outcome was /D/ (cf. frañir ‘to break’, tañir ‘to play’, ceñir ‘to gird’).
38 These were respectively continuants of the Latin verbs DĪCŌ ‘I say’, SPARGŌ ‘I
scatter’ and CINGŌ ‘I surround’.
682 BSS, LXXXIX (2012) PAUL O’NEILL

According to Maiden,39 the cumulative result of such allomorphy, which, as


this author has argued convincingly, defied any explanation in terms of
semantic relatedness and could not be derived from phonological rules, was
that speakers marked this collection of cells as being the locus of a certain
type of identical allomorphy, and the grouping of the cells themselves came
to form part of the linguistic grammar of the speakers.40 In this way the
collection of cells constitutes an autonomous morphological structure which
Maiden has named, following Aronoff, a morphome;41 which is to be un-
derstood as an abstract grouping of cells with tight psychological links
based on purely morphological factors, usually on the basis of some identical
shared form (in the case of the L-pattern, a root allomorph). With particular
reference to the L-pattern cells, Maiden has named the morphome: the
L-pattern; hence my usage of the term.42
It is Maiden’s claim that this collection of cells forms a psychologically
real and distinctive sub-group of paradigmatic cells. With particular refer-
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

ence to the allomorphs given in the examples above, this would mean that,
on account of the fact that these forms and their distribution cannot be
derived by phonological rules or semantic ones, the allomorphs correspond
to lexically memorized roots. It is not the case, however, that speakers
memorize each different inflectional form (e.g. digo, diga, digas . . .); rather,
because the cells which display this identical root allomorphy are the locus
of different types of allomorphy in a number of other lexemes, the purely
morphological distribution (morphome) is learned and it comes to form part
of the grammar of the speakers and to be associated with the root allomorphs
(e.g. dig-). Evidence for this hypothesis comes from diachrony in that the
forms which constitute a morphome manifest two tendencies that Maiden
has termed coherence and convergence.43 The first of these tendencies,
coherence, refers to the attested tendency that the forms of a morphome
have to ‘show persistent resistance to any morphological change liable to
disrupt their peculiar paradigmatic distribution. If an analogical change

39 Martin Maiden, ‘Morphological Autonomy and Diachrony’, Yearbook of Morphology


(2004), 137!75.
40 Cf. Maiden, ‘Irregularity as a Determinant of Morphological Change’; Maiden, ‘Di
nuovo sulle alternanze velari’; Maiden, ‘Morphological Autonomy’; Paul O’Neill, ‘The Origin
and Spread of Velar Allomorphy in the Spanish Verb’ (under consideration).
41 Mark Aronoff, Morphology by Itself (Cambridge, MA: MIT Press, 1994).
42 The term ‘L-pattern’ has no particular semantic significance but rather was adopted
by Maiden on account of the fact that in conventional representations of the verb paradigm on
paper, the cells which comprise the L-pattern resemble a rotated letter ‘L’.
43 Maiden refers to them as principles which prove the psychological reality of the
morpheme (cf. Maiden, ‘Morphological Autonomy’). However, O’Neill has argued that they
ought to be conceived as tendencies and diagnostics of the morphome since highly frequent
irregular verbs sometimes can disrupt the coherence of the morphomic pattern on account of
their forms being so frequently used that they correspond to individual memorized items.
Examples of such forms from Old Spanish are the verbs ir ‘to go’ and haber ‘to have’ which
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 683

affects one ‘‘cell’’ of the paradigm in which the relevant allomorph occurs, it
affects all the others in the same way. The relationship of mutual implication
between ‘‘cells’’ always survives intact’.44
This formal coherence is attested in the history of the velar allomorphs
in Spanish. Old Castilian possessed more verbs which displayed such
allomorphs and whose distribution was limited to the L-pattern cells. This
is either due to an etymological velar consonant which was later lost (see the
examples in Table 15), or to the analogical extension of the velar to another
verb (see Table 16). For brevity, only the first person singular present
indicative has been given. However, it is of crucial importance to note that
the velar was attested in all of the L-pattern cells.

Table 15
Old Spanish: 1sg. Modern Spanish: 1sg.
present indicative present indicative Gloss
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

cuego cuezo I cook


ungo unzo/uño I yoke/I unite
espargo esparzo I scatter
tango taño I play
plango plaño I mourn
frango fraño I break
cingo ciño I gird

Table 16
Old Spanish attested
velars Modern Spanish Form Gloss

suelgo suelo I am accustomed


muelgo muelo I grind
duelga duela I hurt
calga - it is necessary
fuigo huyo I flee
fiergo hiero I harm
remango remano I remain
tuelgo - I take away
distruigo distruyo I destroy
restituigo restituyo I restitute

displayed the velar allomorphs vaiga and haiga in Old Spanish only in the present subjunctive
cells; and forms such as *vaigo or *haigo are not attested in the first person singular present
indicative. Cf. Paul O’Neill, The Ibero-Romance Verb: Allomorphy and the Notion of the
Morphome (Doctoral thesis, University of Oxford, 2011).
44 Cf. Maiden, ‘Morphological Autonomy’. With particular reference to the velar
allomorphy in Italian, Burzio is of the opinion that if the phonological environment is still
there, there is ipso facto phonological conditioning, and in this way the velar allomorphy can
684 BSS, LXXXIX (2012) PAUL O’NEILL

The main point about these examples is that the appearance or loss of a velar
in any one of the L-pattern cells implies its form in all of the other cells.45
The second tendency, convergence, manifests itself because the set of
paradigmatic cells in the morphome ‘tend over time to acquire certain common
phonological characteristics across all verbs in which they occur’.46 And this is
limited to the morphomically defined set of cells. On the basis of the proto-
Castilian forms for the L-pattern, this could also be demonstrated for the velar
verbs in Spanish. Thus, of all the different allomorphy present in the L-
pattern cells (see Tables 9, 14, 15, 16) there have been two noted tendencies:

(a) Standard analogical levelling whereby a one-to-one relationship


between form and meaning is restored to the lexeme of the verb,
usually via the eradication of allomorphy within the L-pattern.47
(b) morphomic levelling (convergence) whereby the irregularity is
maintained but is made more predictable by reducing the types of
allomorphy present in the morphome.
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

Both tendencies are noticeable in old Spanish for the velar verbs: the forms
listed below in A are all examples of analogical levelling as described in
(a) above, whilst those under B constitute cases of morphomic levelling or
convergence in which the varied allomorphic effects of yod were replaced by
velar allomorphy:

A TANGŌ !tango! !taño ‘I play’, FRANGŌ !frango !!fraño


‘I break’, CINGŌ!cingo! !ciño TINGŌ !tengo !!tiño ‘I dye’,
SPARGŌ !espargo ! !esparzo ‘I scatter’, RINGŌ(R) !ringo ! !

be explained via phonological rules (cf. Luigi Burzio, ‘Paradigmatic and Syntagmatic
Relations in Italian Verbal Inflection’, in Contemporary Approaches to Romance Linguistics.
Selected Papers from the 33rd Linguistic Symposium on Romance Languages [LSRL],
Bloomington, Indiana, April 2003, ed. Julie Auger, J. Clancy Clements and Barbara Vance
[Amsterdam/Philadelphia: John Benjamins, 2004], 17!44). Maiden argues against this
methodology on the basis of specific evidence from Italian historical-comparative data (cf.
Martin Maiden, ‘From Pure Phonology to Pure Morphology: The Reshaping of the Romance
Verb’, Recherches Linguistiques de Vincennes, 38 [2009], 45!82).
45 Bybee argued that the subjunctive forms are derived from the form of the first person
singular present indicative, since this is the most frequent and less marked form (cf. Joan L.
Bybee, Morphology: A Study of the Relation between Meaning and Form [Amsterdam/
Philadelphia: John Benjamins, 1985], 68!74). However, O’Neill and Malkiel have proved that
this hypothesis is not supported by the historical data, since there is no evidence to suggest
that the velar element appeared first in the first person singular present indicative forms;
rather, the generalization which confirms the morphome is confirmed: its appearance in one of
the L-pattern cells usually implies its existence in all of these cells. Cf. O’Neill, ‘The Origin
and Spread of Velar Allomorphy’; Malkiel, ‘New Problems in Romance Interfixation’, 307.
46 Cf. Maiden, ‘Morphological Autonomy’.
47 In the case of erguir B ERIGERE the L-pattern allomorph has been extended to all
the paradigm.
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 685

riño ‘I scold’, VINCŌ ![beEko] !!venzo, COQUŌ!cuego ! !


cuezgo !cuezo ‘I cook’ etc.48
B CADŌ !! cayo !caigo ‘I fall’, AUDIŌ !oyo !oigo ‘I hear’,
‘TENEŌ ![teDo] !! tengo ‘I have’, SALIŌ![sa^o]! ! salgo
'
‘I leave’, FACIŌ! [fats o] !!fago !hago ‘I do’ etc.

The origin and analogical extension of the verbs which display velar
allomorphy in Spanish is a much studied and debated topic.49 Here it is
proposed that, in line with other Romance varieties (most notably Italian),
the extension of the velar allomorphs to other lexemes in the early history of
Spanish could be construed as evidence that demonstrates the full effects of
yod upon the verbal paradigm at an early stage in the Spanish language.
Such effects of yod on a number of consonants were effaced from the
historical record. In verbs such as respondo B B[responuo] BRESPONDEŌ,
muerdo B B[mwe uo] BMORDEŌ, huelo BB [gwe^o]BOLEŌ this is due to
analogical levelling, whilst for verbs such as tengo BB [teDo] BTENEŌ,
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

salgo BB[sa^o]BSALIŌ it is the result of morphomic levelling, the latter


process producing the verbs which display velar allomorphy.
In what follows I suggest that the lack of the effects of yod in Spanish
-er verbs is also merely a result of analogical levelling and that this was
a characteristic of the -er conjugational class in Spanish. Interestingly, in
Portuguese the same class of verbs was characterized by morphomic levelling
on the L-pattern cells, and in the historical development of these verbs the
effects of yod are apparent upon the preceding vowels.

5 Spanish and Galician and Portuguese -er Verbs


Spanish and Galician and Portuguese -er verbs originated from a mix of the
Latin 2nd, 3a and 3b conjugations, the first two of which displayed an
etymological desinential yod (see above 1.1). Given my claim that yod should
have had the same effects upon the verbal paradigm as the nominal one, the
L-pattern cells might have been expected to be the domain of various types of
allomorphy. In modern Spanish, however, with the exception of the velar
verbs tener ‘to have’, caer ‘to fall’, traer ‘to bring’, and the verbs saber ‘to
know’ and caber ‘to fit’, this allomorphy is not attested, nor are there in
any other verbs of this class metaphonic effects of yod upon the vowel.
In what follows I compare the development of the Spanish -er verbs with
those of Galician50 and show that whilst Galician clearly shows signs of

48 In these verbs the velar allomorph is either present in Old Spanish and later levelled
or a velar allomorph was predicted via regular sound change, and thus must have been
levelled at a pre-literary stage of the language.
49 Cf. O’Neill, ‘The Origin and Spread of Velar Allomorphy’, for an overview.
50 My reasons for the use of examples from Galician and not Portuguese are outlined in
the following footnote.
686 BSS, LXXXIX (2012) PAUL O’NEILL

morphomic levelling and thus displays a similar type of allomorph across


lexemes exclusive to the L-pattern cells, Spanish displays standard analogi-
cal levelling.
I illustrate this point with a comparative analysis of a selection of verbs
which passed into the -er class in both Galician-Portuguese and Spanish and
whose origin was variably the Latin 2nd, 3a or 3b conjugation. These are:
Latin TIMEŌ ‘I fear’, BIBŌ ‘I drink’, DEPENDEŌ ‘I hang’, INCENDŌ ‘I set
fire to’, UOLUŌ ‘I return’, MOUEŌ ‘I move’, CŌNSUŌ ‘I sew together’
and SORBEŌ ‘I absorb’. These verbs all have reflexes in both languages,51
they all passed into the -er class, and they demonstrate the full range of
etymological vowels /e o o & u/ for the -er class. Moreover, for each type of
vowel I have provided an example of a verb which etymologically contained a
desinential yod and one in which there was no glide present. In Table 17
I have provided the proto-Ibero Romance forms for the present tense of these
verbs.
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

Note that these proto Ibero-Romance forms are in accordance only with
the metaphonic effects of yod on the preceding vowel (see section 3), and are
subsequent to the deletion of yod from the desinences where it variably
must have remained in a number of verbs. Thus for the Latin, DEPENDEŌ
‘I hang’, I do not give the expected form [depent's o] but the form [dependo];
here the glide has had an effect on the vowel only.
These sets of verbs present three different types of paradigm:

(a) paradigms in which there are different types of vocalic allomorphy


in accordance with the L-pattern.
(b) paradigms in which there is an invariable root.
(c) paradigms in which an open mid-vowel in the rhizotonic forms of
the verb (all persons of the singular and the third person plural)
alternates with a close mid-vowel elsewhere.

Given that (c) was, in the first instance, purely a matter of phonology,
since open mid-vowels were merged with their closed counterparts in favour
of the latter forms in unstressed position, then there are only two different
paradigmatic types: L-pattern allomorphy or an invariable root. As was

51 My examples are from Galician because two of the verbs sorber and volver, whilst
present in most Portuguese dictionaries, are both considered antiquated and are no longer
frequently in use in modern standard Portuguese. Also, the present tense reflexes of
INCENDŌ in modern Portuguese do not display the alternation between a mid-open vowel
and mid-closed vowel in accordance with the L-pattern, since in Portuguese this alternation,
unlike in many varieties of Galician, is absent in verbs whose vowel occurs in a syllable having
a nasal consonant in the coda. This is due to a phonological generalization which bans open
mid-vowels in such syllables (see inc[e]ndo vs. inc[e]nde, p[e]ndo vs. p[e]nde from the verbs
incender ‘to set alight’ and pender ‘to hang’). For these reasons I illustrate my point with
examples from Galician. Note also that in Galician there are two irregular verbs querer ‘to
want’ and poder ‘to be able’ which do not display this alternation.
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE
Table 17
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

/e/"[j] /e/ $[j] /o/"[j] /o/$[j]

indicative subjunctive indicative subjunctive indicative subjunctive indicative subjunctive


' '
1SG [timo] [tima] [bebo] [beba] [dependo] [dependa] [entsondo] [entsonda]
' '
2SG [temes] [timas] [bebes] [bebas] [depondes] [dependas] [entsondes] [entsondas]
' '
3SG [teme] [tima] [bebe] [beba] [deponde] [dependa] [entsonde] [entsonda]
' '
1PL [tememos] [timamos] [bebemos] [bebamos] [dependemos] [dependamos] [entsndemos] [entsendamos]
' '
2PL [temetes] [timates] [bebetes] [bebates] [dependetes] [dependates] [entsendetes] [entsendates]
' '
3PL [temen] [timan] [beben] [beban] [deponden] [dependan] [entsonden] [entsondan]
/&/ $[j] /&/"[j] /o/ $[j] /o/"[j]
1SG indicative subjunctive indicative subjunctive indicative subjunctive indicative subjunctive
2SG [b&lbo] [b&lba] [mobo] [moba] [koso] [kosa] [su bo] [su ba]
3SG [b&lbes] [b&lbas] [m&bes] [mobas] [koses] [kosas] [so bes] [su bas]
1PL [b&lbe] [b&lba] [m&bes] [moba] [kose] [kosa] [so be] [su ba]
2PL [bolbemos] [bolbamos] [mobemos] [mobamos] [kosemos] [kosamos] [so bemos] [su bamos]
3PL [bolbetes] [bolbates] [mobetes] [mobates] [kosetes] [kosates] [so betes] [su bates]
1SG [b&lben] [b&lban] [m&ben] [moban] [kosen] [kosan] [so ben] [su ban]

687
688 BSS, LXXXIX (2012) PAUL O’NEILL

the case with the velar verbs (see above, section 4) there are three noted
morphological tendencies in Ibero-Romance to deal with such different
patterns of allomorphy:

(i) morphomic levelling in which the allomorphy is made predictable


by making the different types of allomorphy converge on one type
within the domain of the morphome.
(ii) Classical analogical levelling in which the different allomorphs
are levelled in favour of one form, thus simplifying the
relationship between form and meaning.
(iii) The alignment of the root allomorph with the morphosyntactic
feature of the present subjunctive.

All three options are attested in Ibero-Romance, although the last option
is attested only in a limited number of Asturian varieties, and, within these
varieties, in a limited number of lexemes (San Ciprián and to a certain
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

extent Cabranes).52 The second process clearly underlies the development of


Castilian, as will be noted on examination of the modern Spanish reflexes in
Table 18.
In the modern Spanish paradigms the verbs either display an invariable
root or a diphthongized allomorph in the rhizotonic forms of the verb.
Thus, bearing in mind the proto-paradigm for -er verbs in Table 17, one must
suppose the substitution of the metaphonically raised vowel in the L-pattern
in favour of the other rhizotonic forms of the present indicative: thus,
temio !timo ! !temo, sorbio !surbo ! !sorbo. Note that if the verb pos-
sessed an etymological open vowel, the result at a later stage of the language
would have been a diphthongized allomorph: thus m[&]vio !m[o]vo ! !
m[&]vo!muevo.
Compare the modern Spanish paradigms with those of modern Galician
in Table 19 in which the first process listed above must have taken place: i.e.
morphomic levelling, since the -er verbs have all converged on a close mid-
vowel in the L-pattern and on an open mid-vowel in the second person
singular, third person singular and third person plural present indicative.
Such developments defy explanations in terms of natural paths of analogical
extension, since both this latter collection of cells and the L-pattern cells
constitute semantically and phonologically diverse inflectional forms; and
thus, if any levelling were to take place, it would be expected either to
eradicate the allomorphy, as was the case in Castilian, or align the allo-
morphy with some semantic feature, as is the case in one or two Asturian
varieties. This did not take place; rather, the allomorphy was organized in
accord with the L-pattern morphome.

52 Cf. Krüger, El dialecto de San Ciprián; Canellada, El bable de Cabranes, 39!43.


NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE
Table 18
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

/e/"[j] /e/ $[j] /o/"[j] /o/ $[j]

indicative subjunctive indicative subjunctive indicative subjunctive indicative subjunctive

1SG temo tema bebo beba dependo dependa enciendo encienda


2SG temes temas bebes bebas dependes dependas enciendes enciendas
3SG teme tema bebe beba depende dependa enciende encienda
1PL tememos temamos bebemos bebamos dependemos dependamos encendemos encendamos
2PL teméis temáis bebéis bebáis dependéis dependáis encendéis encendáis
3PL temen teman beben beban dependen dependan encienden enciendan
/&/$[j] /&/"[j] /o/ $[j] /o/ [j]
indicative subjunctive indicative subjunctive indicative subjunctive indicative subjunctive
1SG vuelvo vuelva muevo mueva coso cosa sorbo sorba
2SG vuelves vuelvas mueves muevas coses cosas sorbes sorbas
3SG vuelve vuelva mueve mueva cose cosa sorbe sorba
1PL volvemos volvamos movemos movamos cosemos cosamos sorbemos sorbamos
2PL volvéis volváis movéis mováis coséis cosáis sorbéis sorbáis
3PL vuelven vuelvan mueven muevan cosen cosan sorben sorban

689
690
BSS, LXXXIX (2012)
Table 19
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

/e/ " [j] /e/ $ [j] /o/ " [j] /o/ $ [j]

indicative subjunctive indicative subjunctive indicative subjunctive


1SG t[e]mo t[e]ma b[e]bo b[e]ba dep[e]ndo dep[e]nda enc[e]ndo enc[e]nda
2SG t[o]mes t[e]mas b[o]bes b[e]bas dep[o]ndes dep[e]ndas enc[o]ndes enc[e]ndas
3SG t[o]me t[e]ma b[o]be b[e]ba dep[o]nde dep[e]nda enc[o]nde enc[e]nda
1PL Tememos temamos bebemos bebamos dependemos dep[e]ndamos encendemos enc[e]ndamos
2PL Temedes temades bebedes bebais dependedes dep[e]ndades encendedes enc[e]ndades
3PL t[o]men t[e]mam b[o]bem b[e]bam dep[o]nden dep[e]ndan enc[o]nden enc[e]ndan
/&/ $ [j] /&/ " [j] /o/ $ [j] /o/ $ [j]
Indicative subjunctive indicative subjunctive indicative subjunctive indicative subjunctive
1SG v[o]lvo v[o]lva m[o]vo m[o]va c[o]so c[o]sa sorbo sorba
2SG v[&]lves v[o]lvas m[&]ves m[o]vas c[&]ses c[o]sas s[&]rbes sorbas
3SG v[&]lve v[o]lva m[&]ve m[o]va c[&]se c[o]sa s[&]rbe sorba
1PL volvemos volvamos movemos movamos cosemos cosamos sorbemos sorbamos
2PL volvedes volvades movedes movades cosedes cosades sorbedes sorbades
3PL v[&]lven v[o]lvan m[&]ven m[o]van c[&]sen c[o]san s[&]rben sorban

PAUL O’NEILL
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 691

As stated, this development was not the case in Castilian, which eradi-
cated the majority of the vocalic allomorphy, etymological in the L-pattern,
by levelling it in favour of the non-metaphonized stem in those verbs which
were to become the -er verb class. The implications of this process for
verbs which displayed open mid-vowels in the third person singular present
indicative was that they would undergo diphthongization. Initially this was
a purely phonological process but eventually the diphthongized allomorphs
would become morphologized.
This process of diphthongization had a profound impact upon the
morphology of Spanish, due to the sheer number of lexemes which were
affected both in the -ar class and also the -er verbs (in this case due to the
elimination of the effects of yod as described above). In fact, Maiden and
O’Neill have maintained that the allomorphy present in these rhizotonic
forms of the verb, which was initially the result of the differential phonological
effects on stressed and unstressed vowels, became morphologized; and that
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

subsequently these rhizotonic cells became marked as the domain of identical


allomorphy and came to form another morphome, the N-pattern morphome,
which marked the first person singular, second person singular, third person
singular and third person plural present tense forms as sharing a common
allomorph.53
In Portuguese this morphome was not as predominant in the language;
not merely because of the lack of diphthongization, but due to the effects of
metaphony of yod on !er and !ir verbs and the concomitant convergence
upon a particular vowel exclusively in the L-pattern cells, a process which
eradicated any types of allomorphy according to the N-pattern in these
conjugations and relegated such allomorphy to the -ar verbs, which displayed
different vocalic root vowels initially conditioned purely by the accent. Thus
the type-frequency of these verbs was significantly reduced.
Portuguese has therefore converged upon the close mid-vowels in the
L-pattern, whereas Castilian eradicated the allomorphy by levelling in
favour of the non-metaphonized stem in those verbs which were to become
the -er verb class. This is an alternative explanation to the hypothesis that
yod was merely deleted in such verbs in Castilian.
The question that remains is whether this explanation, based upon
analogical levelling, is a better one than the claim that the yod was simply
deleted in the Latin of the region in which Castilian arose. From a com-
parative Ibero-Romance perspective, the former account seems more feasible
since it is clear that yod did have an effect on -er verbs of surrounding Ibero-
Romance variants. To disregard such comparative evidence in favour of a
narrowly ‘Castilian’ internal perspective as to the effects of yod in verbs
fails to collate Castilian properly among the Ibero-Romance dialects, and to
appreciate the deeper morphological developments taking place. Indeed,

53 Cf. Maiden, ‘Morphological Autonomy’; O’Neill, The Ibero-Romance Verb.


692 BSS, LXXXIX (2012) PAUL O’NEILL

the effects of desinential yod on verbs are directly attested in Castilian for
the verbs saber, caber and for a selection of verbs whose roots ended in /d/ or
/g/ in Latin. Moreover, I have argued that the velar allomorphy in tener, caer,
traer, poner etc. is intimately linked with these verbs displaying consonantal
L-pattern allomorphs as a consequence of yod. Thus I would conclude that,
whilst the claims of the loss of yod in the verbal paradigm of Castilian are
attractive because they obviate a number of thorny problems, the result is
an excessively ad hoc and unattractive explanation of the evolution of the
Spanish verb, one in which the effects of yod are sometimes fully evoked,
other times partially evoked and on occasions thought not to have taken
place. The problem is that the effects of morphology have obscured the effects
of sound change. Rather than to suppose the lack of a sound change, it is
greatly preferable, on the basis of the comparative data, to address and
analyse the morphological effects. In what follows I argue that taking
account of the morphological processes which have eradicated the original
effects of yod, and which I detail below, can greatly illuminate the purely
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

morphological structures around which both the Spanish and Portuguese


verbs are structured. I illustrate this with reference to the development of
-ir verbs in both languages.

6 !
The Development of ir Verbs in Spanish and Portuguese
The evolution of the vocalic developments of Spanish and Portuguese -ir
verbs is a highly debated and complex matter.54 For the present purposes
it is sufficient to reiterate that the evidence of the metaphonic effects upon
the root vowel in a number of verbs and inflectional forms of the Spanish -ir
class clearly suggests the legacy of yod (see above section 2). As with the -er
verbs the simple effects of sound change upon the verbs in the -ir conjugation
cannot, on their own, account for the modern-day paradigms. This is dem-
onstrated by Table 20 which displays the expected phonological outcome55
for the present tense of the Latin verbs RĪDEŌ ‘I laugh’, MĒTIŌ ‘I measure’,
SENTIŌ ‘I feel’, DORMIŌ ‘I sleep’, SUBEŌ ‘I rise’ and SŪMŌ ‘I undertake’56
in accordance with the supposed effects of metaphony on verbs and prior to
the diphthongization of open mid-vowels. These particular verbs have
been chosen because they are representative of all the different vocalic
qualities that stems could display in stressed position, namely /i e o & o u/.
From this point, I refer to the paradigms in Table 20 as the proto-paradigms

54 For an in-depth analysis of such verbs see Paul O’Neill, ‘Alternancias vocálicas en el
presente de los verbos en -ir. Un análisis desde la morfologı́a autónoma’, Revista de Historia de
la Lengua Española (forthcoming).
55 In the reconstruction of these proto-forms I have assumed that Latin initial R was
pronounced as an alveolar trill, that voicing of intervocalic stops had not taken place, and that
the shift had taken place in the accent from the root to the desinence of the first and second
person plural present tense of verbs of the Latin 3a and 3b conjugations.
56 Modern Spanish reı́r, medir, sentir, dormir, subir, sumir.
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 693

for Ibero-Romance, and these ought to be compared with the attested vocalic
alternations for these verbs, given in the Appendix, for the Spanish and
Portuguese reflexes. It should be noted that in Table 20 the cells which share
the same root vowel have identical shading.
With reference to these proto-paradigms three points need to be
highlighted:

(a) They exhibit a large degree of variation with regard to the vocalic
qualities of the root vowels and the types of alternations within
individual lexical items.
(b) With the exception of the verbs whose root vowel is an
etymological reflex of a Latin long high vowel (Ū/Ī), all verbs
display allomorphy within the L-pattern.
(c) A high vowel is etymologically predicted in all verbs for the first
and second person plural present subjunctive.
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

Comparing these forms with those of modern Portuguese given in the


Appendix, Portuguese has generally made this allomorphy more predictable
for this class of verbs by organizing it with reference to the autonomous
morphological structure of the L-pattern; that is, there has been morphomic
convergence upon a high-vowel in the L-pattern whilst the other rhizotonic
forms of the present indicative have converged upon an open mid-vowel:
thus, there has been ‘morphomic levelling’.57
In this respect the modern Spanish -ir verbs, whose congener forms to
the Portuguese verbs above can be found in Table 22 in the Appendix,
comprise a much more varied selection of types than the Portuguese verbs.
As with the -er verbs, there has been a tendency to level the allomor-
phy and spread an invariable high vowel to all forms of the paradigm
(subir B Bsobir SUBĪRE, and also acudir BBacodir BACCUTERE,
sufrir B Bsofrir BSUFFERRE, escupir B Bescopir B*EXCONSPUERE).58

57 Once again, any explanation of these forms on the basis of sound change and analogy
from the Latin etyma is hazardous since, in the case of the L-pattern and on the basis of the
four vowel types /i, o, e, u/ present in the proto-paradigms, a sound change would have to be
invented which would raise mid-vowels in both stressed and unstressed positions and which
variably would not apply to the rhizotonic forms of the second person singular, third person
singular and third person plural present indicative; in which case there would have to be a
different rule which would lower all close mid-vowels and high vowels to open mid-vowels.
Likewise, any recourse to analogy would have to explain the unusual paths taken: that the
strange pattern of allomorphy present in the proto-paradigms is consolidated and extended to
other verbs which lack such allomorphy. A more common or natural course of action would be
either to level the allomorphy completely and thus create a biunique relationship between
sound and meaning, or else to align the allomorphy with some semantic criteria (e.g. confining
the phonological feature just to the present subjunctive).
58 Cf. O’Neill, ‘Alternancias vocálicas’.
694
BSS, LXXXIX (2012)
Table 20
Proto-Paradigm
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

/i/ BĪ /e/BĬ/Ē/OE /o/BĔ/AE

indicative subjunctive indicative subjunctive indicative subjunctive


1 1 1 1 1
1SG [ ri o] [ ri a] [ mito] [ mita] [ sento] [1senta]
2SG [1ries] [1ri as] [1metes] [1mitas] [1sontes] [1sentas]
3SG [1rie] [1ri a] [1mete] [1mita] [1sonte] [1senta]
1PL [ri1emos] [ri1 amos] [me1timos] [mi1tamos] [sen1timos] [sin1tamos]
2PL [ri1etes] [ri1 ates] [me1tites] [mi1tates] [sen1tites] [sin1tates]
3PL [1rien] [1ri an] [1meten] [1mitan] [1sonten] [1sentan]
/&/BŎ /o/ BŌ/Ŭ /u/ BŪ
indicative subjunctive indicative subjunctive indicative subjunctive
1SG [1do mo] [1do ma] [1subo] [1suba] [1sumo] [1suma]
2SG [1d& mes] [1do mas] [1sobes] [1subas] [1sumes] [1sumas]
3SG [1d& me] [1do ma] [1sobe] [1suba] [1sume] [1suma]
1PL [do 1mimos] [du 1mamos] [so1bimos] [su1bamos] [su1mamos] [su1mamos]
2PL [do 1mites] [du 1mates] [so1bites] [su1bates] [su1mates] [su1mates]
3PL [1d& men] [1do man] [1soben] [1suban] [1sumen] [1suman]

PAUL O’NEILL
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 695

The most outstanding difference, however, with regard both to the proto-
paradigm and the Portuguese forms, lies in the different distributions of
allomorphy, which in Spanish is, to a large extent, characterized by the N-
pattern forms sharing the same root allomorphs, which can either be a high
vowel (servir, medir) or a diphthong (sentir, dormir).
In the historical development of both Spanish and Portuguese verbs there
has been a similar tendency for verb forms to display formal coherence across
a collection of cells with respect to the root allomorph, and for the allomorphy
in these cells to display formal convergence across lexemes. The major
difference between Spanish and Portuguese is the set of cells which display
such coherence and convergence; in Portuguese the cells are those of the L-
pattern, while in Spanish they are those of the N-pattern. Thus, for reflexes
of the verb DORMĪRE, the etymological open vowel of the second and third
person singular and third person plural present indicative has spread, in
Spanish, to the other rhizotonic forms of the verb (dormo ! !
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

d[&]rmo !duermo, dorma !!d[&]rma !duerma). In Portuguese, on the


other hand, no analogical extension has taken place and in its stead the
etymological closed mid-vowel of the L-pattern cells has converged upon a
high vowel (*dormo !! durmo, *dorma !! durma). For the verb medir ‘to
measure’, Portuguese has maintained the original allomorphy produced by
yod (MĒTIŌ(R) !meço) exclusively in the L-pattern cells, whilst Spanish
must have levelled this allomorphic root but still allowed the desinential
yod to have a metaphonic effect on the root vowel (MĒTIŌ(R) !mezo! !
metio !mido; see above, section 3). This high vowel, which was only
etymological in the L-pattern cells, has then spread to the other N-pattern
cells, the rhizotonic cells of the present tense.
Such different patterns of analogical extension and formal convergence
in the history of the Spanish and Portuguese verbs are, I would argue,
intimately connected with the different morphological responses to the
effects of desinential yod in the verbal paradigm. As previously explained,
Spanish and Portuguese differed in their treatment of yod in -er verbs, in
that Portuguese displayed morphomic levelling, whereby the morphomic
patterns etched out by sound change were heightened and all verbs of this
conjugation which displayed mid-vowels in the root came to converge upon a
particular allomorph in the L-pattern cells. Castilian, however, did not
simply delete or lose the etymological yod but displayed analogical levelling
in which the allomorphy in the verbal paradigm was levelled in favour of the
most frequent form. Such levelling, when applied to the reflexes of spoken
Latin verbs with root vowels /o/ and /&/ in the rhizotonic forms, resulted in
the situation whereby the open vowels were reinstated in these cells and
could undergo diphthongization (MOVEŌ !m[o]vio !!m[&]vo !muevo).
This created a pattern of alternation in which the rhizotonic forms of the
verb all shared the same allomorph.
696 BSS, LXXXIX (2012) PAUL O’NEILL

Such different patterns of alternation are crucial for the understanding


of the Galician and Portuguese and Spanish verbs, since synchronically
they determine the distribution of allomorphy, and diachronically one must
presuppose the existence of these patterns in order to account for certain
historical developments.

Conclusion
In the foregoing discussion I have provided evidence, from within Spanish
but also largely from comparative Ibero-Romance varieties, to suggest that
the effects of yod on the verbal paradigm could have been similar to
those attested for nouns. However, due to a number of analogies and
different types of levellings, these effects were largely effaced from the first
testimonies of Spanish. Although the obvious option of the philologist faced
only with the Spanish historical data is to propose a more or less wholesale
deletion of desinential yod, the closely related languages testify differently.
Downloaded by [Paul O'Neill] at 15:21 10 July 2012

Moreover, a full appreciation of the effects of yod and of the processes by


which Spanish came to eliminate such effects can help to illuminate the
abstract morphological structures upon which the verbal systems of
Spanish, Portuguese and Galician are based. Such an appreciation can
assist us to understand the spread of the velar allomorphs and to appreciate
this spread viewed within a larger and more coherent Romance context.
Also, as regards the vocalic developments of -ir verbs, by appreciating
what happened to the yod and its effects, we can reach a better under-
standing of the differential patterns of vocalic allomorphy and of the paths of
analogical extensions which took place in the Spanish, Portuguese and
Galician verbs.
NEW PERSPECTIVES ON THE EFFECTS OF YOD IN IBERO-ROMANCE 697

Appendix

!
Table 21
Portuguese ir verbs which are reflexes of those of the
proto-paradigm in Table 20
/i/BĪ /e/ BĬ/Ē/OE /o/BĔ/AE

indicative subjunctive indicative subjunctive indicative subjunctive

1SG rio ria meço meça sinto sinta


2SG ris rias medes meças s[e]ntes sintas
3SG ri ria mede meça s[e]nte sinta
1PL rimos riamos medimos meçamos sentimos sintamos
2PL rides riais medis meçais sentis sintais
3PL riem riam medem meçam s[e]ntem sintam

/&/BŎ /o/ BŌ/Ŭ /u/ BŪ


Downloaded by [Paul O'Neill] at 15:21 10 July 2012

indicative subjunctive indicative subjunctive indicative subjunctive

1SG durmo durma subo suba sumo suma


2SG d[&]rmes durmas s[&]bes subas s[&]mes sumas
3SG d[&]rme durma s[&]be suba s[&]me suma
1PL dormimos durmamos sobimos subamos somimos sumamos
2PL dormis durmais sobis subais somis sumáis
3PL d[&]rmem durmam s[&]bem subam s[&]mem sumam

!
Table 22
Spanish ir verbs which are reflexes of those of the proto-paradigm
in Table 20
/i/BĪ /e/ BĬ/Ē/OE /o/BĔ/AE

indicative subjunctive indicative subjunctive indicative subjunctive

1SG rı́o rı́a mido mida siento sienta


2SG ries rı́as mides mida sientes sientas
3SG rı́e rı́a mide mida siente sienta
1PL reı́mos riamos medimos midamos sentimos sintamos
2PL reı́s riais medı́s midáis sentı́s sintáis
3PL rı́en rı́an miden midan sienten sientan
/&/BŎ /o/ BŌ/Ŭ /u/ BŪ
indicative subjunctive indicative subjunctive indicative subjunctive
1SG duermo duerma subo suba sumo suma
2SG duermes duermas subes subas sumes sumas
3SG duerme duerma sube suba sume suma
1PL dormimos durmamos subimos subamos sumimos sumamos
2PL dormı́s durmáis subı́s subáis sumı́s sumáis
3PL duermen duermen suben suban sumen suman

You might also like