Hualde, J. I., Simonet, M., and Torreira, F. (2008). Postlexical contraction of non-high vowels in Spanish. Lingua, 118:1906–1925.

Available online at www.sciencedirect.com Lingua 118 (2008) 1906–1925 www.elsevier.com/locate/lingua Postlexical contraction of nonhigh vowels in Spanish Jose´ Ignacio Hualde *, Miquel Simonet, Francisco Torreira Department of Spanish, Italian and Portuguese, 4080 FLB, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA Received 17 September 2006; received in revised form 20 June 2007; accepted 2 October 2007 Available online 26 November 2007 Abstract Although at the lexical level (in the citation form of words) Spanish has only high glides, mid vowels may also become nonsyllabic by a postlexical process of syllable contraction. In this paper we report on an experiment designed to study the contraction of sequences of nonhigh vowels in Castilian Spanish. Traditional phonological descriptions of this Spanish dialect state that in sequences of equal height such as /eo/, /oe/, the first of the two vowels undergoes greater shortening, becoming nonsyllabic. In sequences of different height, such as /ae/, /ea/, on the other hand, the higher of the two vowels becomes nonsyllabic. Our results offer only partial support for these hypotheses. In agreement with other recent studies, we find the realization of these sequences to be very variable. In particular, we discuss two seemingly opposite tendencies present in our data and elsewhere for the elimination of mid glides. One is their deletion: /ea/ [ a] [a]. The other one is their raising to high glides: /ea/ [ a] [ja]. # 2007 Elsevier B.V. All rights reserved. Keywords: Spanish; Glides; Contraction; Syllabification 1. Introduction Most commonly, glides (nonsyllabic vocoids) are [+high] segments. Nevertheless, in principle, glides can also be [!high]. The fact that the IPA has special symbols for high glides, [j], [w], but not for mid glides, which require the use of diacritics on the symbols for mid vowels, nonsyllabic, [ ], [ ] or extrashort [ ], [o˘], appears to reflect the relative crosslinguistic rarity of nonhigh glides. In this paper we examine the results of the contraction in a single syllable of sequences of nonhigh vowels in Spanish. By definition, in such tautosyllabic sequences, one vocoid will be the syllabic nucleus and the others will be nonsyllabic vocoids or ‘‘glides’’. * Corresponding author. E-mail addresses: [email protected] (J.I. Hualde), [email protected] (M. Simonet), [email protected] (F. Torreira). 0024-3841/$ – see front matter # 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.lingua.2007.10.004  J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 1907 In Spanish, like in many other languages, heterosyllabic sequences of vowels tend to contract in a single syllable in running speech, especially in unstressed contexts. This process affects all phonemic vowels. High vowels in contact with other vowels become glides: mi amigo m[ja]migo ‘my friend’. Mid vowels in the same context, on the other hand, are reported to show a more variable behavior. This variability is perhaps to be related to the relative typological rarity of the expected mid glides from the contraction of mid vowels. Some questions remain regarding the realization of sequences of nonhigh vowels in Spanish. In section 2, we examine the syllabification of VV sequences in citation form and the distribution of lexical glides in Spanish. In section 3, we describe a process of syllable contraction affecting lexically heterosyllabic sequences, summarizing previous work on their phonetic realization. Prior work on vowel sequences across word boundaries in Spanish employ a qualitative, categorical classification of tokens (Jenkins, 1999; Aguilar, 2003, 2005; Alba, 2005). Arguably, however, vowel contraction is a gradient process. An examination of continuous features such as vowel formant values may thus contribute to a better understanding of the phenomenon. In section 4, we report on an experiment where we use duration and formant values to determine the outcome of the contraction of sequences of nonhigh vocoids. In particular, one question that we attempt to answer is whether one of the two vowels in contact systematically undergoes greater shortening and becomes nonsyllabic when a heterosyllabic sequence is reduced to a single syllable. The paper ends with some general considerations regarding the contraction of sequences of nonhigh vocoids (section 5). 2. Syllabification of VV in citation forms in Spanish In Spanish, in the syllabification of sequences of two vocoids (VV sequences) we find near- complementary distribution between heterosyllabic (or hiatus) and tautosyllabic (or diphthong) sequences in the citation form of words. These sequences are realized in hiatus if (a) neither of the two vocoids is high or (b) one of the two vocoids is high but has word-stress. On the other hand, a VV sequence normally forms a diphthong if one of the vocoids is high and does not receive stress. This distribution is exemplified in (1). Examples are given in conventional Spanish orthography, except that word-stress is marked on all words and a hiatus is indicated by a dot between the two vowels: (1) Syllabification of VV sequences in citation form a. Both Vs [-hi]: hiatus ea te.a´tro ‘theater’ a´re.a ‘area’ ae ca.e´r ‘to fall’ pa.elle´ra ‘type of pan’ oe po.e´ma ‘poem’ po.etı´sa ‘poetess’ eo be.o´do ‘drunk’ petro´le.o ‘petrol’ b. One of the Vs is [+hi] and receives stress: hiatus ´ıa Marı´.a aı´ ca.ı´da ‘fall’ ´ıo desvı´.o ‘detour’ oı´ o.ı´do ‘ear’ c. Otherwise: diphthong (with some exceptions) jo Ma´rio ja´ varia´do ‘varied’ o´j bo´ina ‘beret’ a´j ca´igo ‘I fall’ 1908 J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 Given this distribution, glides can be considered allophones of the high vowels in the context defined in (1c). That is, we can assume the following rule (Hualde, 2005:80): (2) Lexical Gliding Rule V[+hi] [!syll]/if unstressed and adjacent to another V It should be noted that sequences that undergo this rule are obligatorily realized with a diphthong. There are, nevertheless, some lexical contrasts in syllabification in Spanish. This is because the Lexical Gliding Rule has exceptions (‘‘exceptional hiatus words’’). Thus, for instance [pje´] ‘foot’ (regular pattern) contrasts with [pi.e´] ‘I chirped’, which is an exception to the Lexical Gliding Rule. A word like pie´ ‘I chirped’ contains a hiatus in citation form even though it presents the context for the Lexical Gliding Rule. In most words with an exceptional hiatus there are morphological or paradigmatic reasons for this syllabification; for instance, the hiatus of pie´ ‘I chirped’ is explained by the fact that this word is paradigmatically related to pı´o ‘I chirp’, with stress on the high vocoid. Nevertheless, there are also instances of exceptions to the Lexical Gliding Rule without morphological justification. The distribution of exceptional hiatus in the lexicon has been the object of some recent work (see Hualde, 1999; Hualde and Prieto, 2002; Chitoran and Hualde, 2007; Hualde, 2005:81–86, Cabre´ and Prieto, 2004, 2006, etc.). It appears that Peninsular Spanish has more exceptions to Lexical Gliding than many Latin American varieties. It follows from the description above that Spanish has a lexical contrast between heterosyllabic and tautosyllabic VV sequences, even though, arguably, it does not have phonemic glides (Roca, 1997; Harris and Kaisse, 1999; Hualde, 2005). As mentioned, lexical or obligatory glides are allophones of the high vowels in the context of the rule in (2). The contrasts that have been noticed arise because of the existence of lexical exceptions to the rule. 3. Syllable contraction A further complication in the syllabification of sequences of vocoids is introduced by another phenomenon, the existence of a Postlexical Syllable Contraction Rule: Many sequences that are syllabified as hiatus in citation form can be reduced to a single syllable in connected speech (Navarro Toma´s, 1977 [1918]; Hualde, 1994; Aguilar, 1999, among others). (3) Postlexical Syllable Contraction Rule: V.V 3.1. Contraction of sequences containing high vowels The actual result of the contraction process depends on the vowels in contact. To begin with, ‘‘exceptional hiatus’’ words which, in citation form, have a hiatus in the context of the Lexical Gliding Rule, may also be realized with a diphthong. In the case of sequences with unstressed high vocoids, the actual lexical contrast is thus between (a) regular words with an obligatory diphthong in citation form in the context of the Lexical Gliding Rule, and (b) words that have an exceptional hiatus in citation form, but may also be realized with a diphthong by Postlexical Gliding.  J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 1909 (4) VV sequences in the context of the Lexical Gliding Rule a. Regular pattern: obligatory diphthong duelo [dwe´lo], *[du.e´lo] ‘sorrow’ cambiar [kambja´r], *[kambi.a´r] ‘to change’ b. Exceptional hiatus: dueto [du.e´to] " [dwe´to] ‘duet’ enviar [embi.a´r] " [embja´r] ‘to send’ The existence of a process of Postlexical Syllable Contraction seems to explain why there is considerable dialectal variation in the lexical distribution of exceptional hiatus. The Postlexical Syllable Contraction Rule may apply both to word-internal sequences and to sequences across word-boundaries. The variable application of the Postlexical Syllable Contraction Rule across word boundaries results in the following contrasts, where the diphthong realization is ambiguous between two meanings, but the hiatus realization can only correspond to a sequence across word boundaries1: (5) VV Sequences within and across words Pronunciations Meanings [sje´mbra] (a), (b) (a) siembra ‘s/he sows’ [si.e´mbra] only (b) (b) si hembra ‘if female’ [demilja´na] (a), (b) (a) de Emiliana ‘of Emiliana’ [demili.ana] only (b) (b) de mi liana ‘of my liana’ (Exceptional hiatus) [sje´ntes] (a), (b) (a) sientes ‘you feel’ [si.e´ntes] only (b) (b) si entes ‘if entities’ [swe´ko] (a), (b) (a) sueco ‘Swedish’ [su.e´ko] only (b) (b) su eco ‘its echo’ Reduction to diphthong has a more categorical status in the case of unstressed sequences of falling sonority across word boundaries. Thus, for instance, esta´ bailando /esta´#baila´ndo/ ‘s/he is dancing’ and estaba hilando /esta´ba#ila´ndo/ ‘s/he was weaving’ are homophonous unless a pause is produced between the two words. (Word internal exceptions to the Lexical Gliding Rule involving sequences of falling sonority are also very rare). 3.2. Contraction of sequences of nonhigh vowels Sequences containing only [!high] vowels can also be reduced to a single syllable by application of the Postlexical Syllable Contraction Rule. This possibility is taken into account in the Spanish poetic tradition. For instance, a sequence such as tengo alguno ‘I have some’ may count as containing four syllables: ten-goal-gu-no. The phonetic results of the reduction process have been described as being very variable when the sequence contains only nonhigh vowels. For Peninsular Spanish, Navarro Toma´s (1977:68) notices that, for instance, cuatro reales ‘four quarters’ may be realized as [kwa´tro ja´les] or as 1 The existence of hiatus sequences at word boundaries may explain why nonmorphological exceptional hiatus is limited to word-initial sequences (‘‘Initiality condition’’). It would be straightforward to design an experiment to test the claims in (5) regarding interpretation. 1910 J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 [kwa´tro a´les] in popular (‘‘vulgar’’) speech, in addition to the ‘‘standard’’ solutions with a hiatus [e.a] or a sequence with a reduced mid vowel. There appears to be both interdialectal variation and variation within the same dialect and even in the speech of a single speaker. For instance, Jenkins (1999:51) reports the following variable outcomes for three sequences of two words that are very frequent in the corpus of spoken New Mexico Spanish that he examines: (6) Jenkins (1991: 51): variable results of Syllable Contraction Diphthong V1 del. V2 del. Coalesc. la escuela ‘the school’ __ 14 5 41 me acuerdo ‘I remember’ 41 11 13 10 cuerpo espı´n ‘porcupine’ 2 7 6 __ In general, Jenkins notices that many vowel sequences may be realized either as tautosyllabic, in one of the four ways indicated above, or as heterosyllabic. Jenkins’ study is based on his perceptual interpretation of the facts, not on spectrographic analysis. Alba (2005), employing the same New Mexico corpus as Jenkins, finds that frequency factors (in particular, patterns of cooccurrence between words) play a role in the realization of vowel sequences across word boundaries that is almost as important as that of phonological factors like stress and vowel quality. This study is limited to sequences where /a/ is followed by another vowel across a word boundary. For the analysis, tokens are classified as showing either maintenance or reduction of the hiatus (although Alba acknowledges that this is a gradient phonetic dimension). Aguilar (2003) conducted an experiment where five speakers of Peninsular Spanish were instructed to read sentences containing the sequences /a-a/, /a-e/, /a-i/ across word boundaries and varying the position of the stress. On the basis of spectrographic observation, she classifies the results as falling into three categories: hiatus (= two vowels in two syllables), reduction (= two vowels in a single syllable; i.e. a diphthong) and monophthong (= one vowel). Tokens are classified as containing a hiatus if two distinct vowels are clearly observable in the waveform display. For the criteria that determine that a vowel sequence has been reduced to a diphthong, the author refers to her earlier work on word-internal sequences: ‘‘As for diphthongs, two steady-state parts are rarely present in their formant tracks, instead, a continuous transition from one frequency area to another is generally observed’’ (Aguilar, 1999:62). In instances of monophthongization, only one vowel can be observed in the spectrographic display. In addition to giving a breakdown of the data by prosodic context according to these three phonetic categories, Aguilar reports on durational values. She finds that stress is an important factor affecting the duration of vowel sequences. As expected, the highest incidence of non-hiatus realizations and the shortest durations are found in sequences where neither vowel in the sequence is lexically stressed. Comparing the behavior of sequences with and without high vowels, Aguilar concludes that both can be reduced to monosyllabic groups. That is, nonhigh vowels can be phonetically realized as nonsyllabic, even though at the lexical level Spanish has only high glides. Aguilar (2005) examines the effect of prominence and phrasing on syllabification of vowel sequences across word boundaries, also with reading data from five speakers of Peninsular Spanish. Again, the study is based on a categorical classification of acoustic data. Similarly to Aguilar (2003), sequences are categorized as showing heterosyllabification (= hiatus), reduction (= diphthong) or deletion (=a single vowel), based on spectrographic analysis. Hiatus is found in only 14% of the tokens, the majority of these in sequences where the (word-initial) second vowel  J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 1911 is lexically stressed. Aguilar gives prototypical examples of spectrograms corresponding to her three categories, but she does not report formant values. In VV sequences reduced to a diphthong, one of the lexical vowels is assumed to become nonsyllabic. One question concerns the phonetic nature of these nonsyllabic vocoids. Awidespread dialectal phenomenon in Spanish is that mid vowels are realized as high glides in this context (e.g. /ea/ [ja], /oa/ [wa]). Across word-boundaries, an example from Jenkins (1999:18) is vinimos a este [binimosa´jste] (with stress displacement). There is much social awareness of this process, since it produces neutralization between high and mid vowel phonemes, both realized as high glides. This phenomenon is common in several Latin American regions (see Vaquero, 1996:55 for the Antilles, Lope Blanch, 1996:82 for Mexico, Quesada Pacheco, 1996:103 for Central America, Sedano and Bentivoglio, 1996:120 for Venezuela, and Garrido, in preparation, for an experimental study comparing Mexican and Colombian Spanish). In some areas, however, it is subject to stigmatization, perhaps because it may lead to spelling errors; e.g., *pior for peor ‘worse’, *peliar for pelear ‘to fight’, *traira´n for traera´n ‘they will bring’. In Peninsular Spanish, in particular, this pronunciation of word-internal sequences with orthographic <e> and <o> has been heavily stigmatized. Nevertheless, as Aguilar (2003, 2005) notices, mid vowels are still often realized as nonsyllabic in this dialect in VV sequences, but without becoming high glides. Regarding sequences of nonhigh vowels, such as /ae/, Aguilar (2005:32) indicates that the vowel that becomes nonsyllabic is centralized. Aguilar (2003:2114, 2005:32, fn. 2) proposes that such sequences should be considered as having the same syllabic structure as diphthongs containing high vocoids. The observation is thus that, phonetically, Spanish has both high and nonhigh nonsyllabic vocoids. All the studies just reviewed classify VV token sequences as showing one of two or more distinct results. Nevertheless, since, as Alba (2005) points out, what we actually have is a continuum of realizations (see also Baltazani, 2006 for Greek), arguably the use of continuous variables for the analysis may offer us a better picture of the facts. This concerns both the type of contraction and the nature of nonsyllabic segments, when these are found. In the next section we develop four specific hypotheses regarding the realization of mid vowels in contexts where they are likely to undergo syllable contraction and describe an experiment we designed in order to test these hypotheses. 4. Experiment 4.1. Research question In this experiment our focus is on sequences of two nonhigh vowels across word boundaries, where neither of the two vowels carries stress. According to Navarro Toma´s (1977 [1918]:149) both of these features, lack of stress and the presence of an intervening word boundary, favor contraction in a single syllable. The role of stress was confirmed in Alba’s (2005) and Aguilar’s (2003, 2005) studies. We examine Peninsular Spanish. The four sequences that are tested are those in (7): (7) VV sequences analyzed examples /a-e/ espa´da enemı´ga ‘enemy sword’ /e-a/ ge´nte ama´ble ‘nice people’ /o-e/ mo´no ena´no ‘dwarf monkey’ /e-o/ tı´nte oscu´ro ‘dark shade’ 1912 J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 Following Navarro Toma´s (1977:67–69), in VV sequences of different height, the higher vowel should reduce more in duration, whereas in sequences of equal height, the first one of the two is the one that is reduced (see also Hualde, 2005:89–91, among others).2 This is supported by dialectal developments in word-internal contexts, where a mid vowel has become a ‘‘true’’ high glide in the contexts of reduction indicated by Navarro Toma´s. For instance, teatral ‘theatrical’ may become [tja.tQa´l] and empeorar ‘to worsen’ may become [empjoQa´r], but usually not *[empewQa´r]. (8) Dialectal evolutions: raising of mid vowels to high glides /ae/ [aj] traera´n [tQajQa´n] ‘they will bring’ /ea/ [ja] paseara´n [pasjaQa´n] ‘they will pass’ /eo/ [jo] empeora´r [empjoQa´r] ‘to worsen’, not *[empewQa´r] /oe/ [we] poesı´a [pwesı´a] ‘poetry’, not *[pojsı´a] As noted above in section 3.2, the developments in (8), with high glides for etymological mid vowels have been stigmatized in Spain and are not common in the speech of educated speakers of this variety. However, it is possible that a tendency to raise unstressed mid vowels in this context may still be found at the subphonemic level, especially in the contraction of sequences across word boundaries. The same sociolinguistic considerations would not apply as long as phonemic /e/, /o/ are kept distinct from /i/, /u/. As mentioned in section 3.2, Navarro Toma´s mentions two ‘‘vulgar’’ developments for sequences with mid vowels. In addition to the raising phenomenon just considered, a prevocalic mid vowel may be deleted. It thus appears that we must contemplate two possible paths of development for shortened mid vowels: raising to high glide (e.g. me alegro m[ a]legro m[ja]legro ‘I am happy’) and deletion (e.g. me alegro m[ a]legro m[a]legro), both with a continuum of intermediate realizations. 4.2. Hypotheses If contraction takes place, our hypothesis, following Navarro Toma´s, is that one of the two vowels should reduce more than the other, depending on both relative phonological height and order. This should be apparent from formant trajectories. In particular we want to test the following hypotheses: (9) Hypotheses I) Contraction is manifested by shortening. If contraction applies, the duration of VV should be much shorter than the sum of the duration of the two vowels in a VC context. II) In sequences of different heights (mid and low vowels, /ea/, /ae/), the mid vowel will show greater shortening. Formant values should approach those of /a/ at the middle point of the sequence. III) In sequences of same height (two mid vowels, /eo/, /oe/), on the other hand, the first one will reduce more in duration. Formant values should approach those of the second vowel already as early as the middle point of the sequence. 2 For Mexican and Mexican-American Spanish, however, slightly different claims have been made in the phonological literature (see Harris, 1970; Hutchinson, 1974; Martı´nez-Gil, 2006).  J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 1913 IV) If reduction in duration is accompanied by raising (as in the dialectal developments in (8)), the raising of mid vowels, e.g. /ea/ [ja], would be manifested in more extreme values at the beginning (or at the end, in the case of /ae/) than those of the corresponding vowel in the interconsonantal context. 4.3. Method Sixty sentences were constructed containing six examples of each of the four sequences tested plus six examples of each of the three vowels in interconsonantal position (both word-initial and word-final). The same lexical items were used for sequences and for single vowels; e.g. mon[o e]nano ‘dwarf monkey’ vs. mon[o] gigante ‘giant monkey’ vs. gorrio´n [e]nano ‘dwarf sparrow’. The analyzed vowels are always found in a noun+adjective construction towards the beginning of the sentence. This context was chosen because it is an unlikely environment for a prosodic break to be inserted. A couple of sample sentences are given in (10) (the complete list can be found in Appendix A): (10) Sample sentences una espada enemiga acabo´ con su vida. ‘an enemy sword ended his life’ el mono enano es el ma´s pequen˜o de todos los monos. ‘the dwarf monkey is the smallest monkey of all.’ The 60 sentences were presented to all participants in the same random order. Participants were asked to read the list three times (=180 analyzed tokens per subject). The speech was recorded directly onto a computer using a unidirectional microphone. For the analysis of the data, the program PRAAT (Boersma and Weenink, 1997–2006) was used. In each sentence, boundaries were manually added at the beginning and end of the test V or VV based on both waveforms and spectrograms. Stop release bursts were used to segment vowels from stops. Changes in amplitude and onset and offset of F2 were used for the placement of initial and final boundaries. Formant values were extracted using a window length of 25 ms with time steps of 25 ms/0.25 = 6.25 ms. F1 and F2 values were automatically extracted at three points: at 25%, 50% and 75% of the segmented V or VV sequence. Duration values were also automatically obtained. Four speakers of Peninsular Spanish participated in this experiment, three female and one male. The speakers were in their 20s and 30s and all had university education. 4.4. Results 4.4.1. Duration (Hypothesis I) The average duration in milliseconds of each of the vowels and VV sequences measured (with standard deviation in parentheses) is shown in Table 1 for all four speakers. The duration measurements in Table 1 demonstrate the tendency for VV sequences to contract. In every case and for all speakers, the duration of a given VV sequence is considerably shorter than the sum of its two constituent vowels in the interconsonantal context. In some cases, in fact, durations of sequences are reduced to approximately the duration of a single vowel. An ANOVA followed by pairwise comparisons (Fisher’s PLSD) shows that, although, generally, sequences are still significantly longer than single vowels, in the following cases there is no significant difference in duration: 1914 J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 Table 1 Durations for all speakers in ms., mean (S.D.) Na Spkr1 Spkr 2 Spkr 3 Spkr 4 a 36 60.9 (13.4) 66 (17.1) 81.9 (14.2) 71.4 (15.9) e 36 58 (19.8) 55.7 (16.6) 70 (17.2) 61.9 (23.9) o 36 59.6 (18.2) 61.3 (15.7) 60.7 (20) 69.8 (18.1) ae 18 82 (10.6) 91 (17.3) 95.3 (25.3) 87 (20.2) ea 18 79.9 (17.3) 81.2 (13.2) 80.6 (18) 76.6 (16.2) eo 18 85.7 (25) 85.5 (25.2) 85.5 (21) 81.9 (26.6) oe 18 91.5 (26.9) 89 (23.8) 95.1 (19.8) 96.5 (36.1) a For Spkr 2 one token of /ae/ and another one of /o/ were discarded. For Spkr 3 one token of /o/ was discarded. Table 2 Mean F1/F2 values in Hz at mid point: /a/, /ae/, /e/, /ea/ Spkr 1 Spkr 2 Spkr 3 Spkr 4 /a/ 796/1852 886/1872 889/1707 617/1545 /ae/ 639/2243 762/2183 758/1882 545/1695 /e/ 554/2308 585/2265 615/2102 445/1890 /ea/ 653/2288 754/2222 843/1746 595/1658 Fig. 1. (Four panels) Plots of formant values at mid point of V or VV for each of the four speakers. The ellipses enclose one standard deviation on each direction from the mean for each of the two formants.  J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 1915 Table 3 Euclidean distance at mid point Spkr 1 Spkr 2 Spkr 3 Spkr 4 /ae/ - /a/ 421 335 219 166 /ae/ - /e/ 107 195 262 219 /ea/ - /a/ 459 374 24 115 /ea/ - /e/ 101 174 423 276 /ae/ - /ea/ 47 40 160 62 (11) Sequences not significantly different in duration from a single vowel Spkr 3: /a/ = /ea/; /a/ = /eo/, /e/ = /ea/ Spkr 4: /a/ = /ea/, /o/ = /eo/, /o/ = /ea/ The next question is then how that duration reduction is accomplished. For that we turn to the analysis of formant values. 4.4.2. Formants. Sequences of different height /ea/, /ae/ (Hypothesis II) Regarding sequences of unequal height (/ea/, /ae/), the hypothesis is that the higher of the two vowels will show greater reduction. This reduction, however, may be accomplished in different ways. Hypothesis II concerns primarily formant values at the 50% point. For both /ea/ and /ae/, values at mid point should be closer to those of /a/ than to /e/, since the higher vowel in a sequence is predicted to have shorter duration. Means at mid point are given in Table 2 (both means and standard deviations at all three measured points are given in Appendix B). The values in Table 2 and Table 5 below are plotted in Fig. 1. The Euclidean distance between each sequence and each of the single vowels /a/, /e/ was calculated on the basis of the average values given above in Table 2.3 The results of this calculation are given in Table 3. These data do not uniformly support the hypothesis. Both sequences /ae/ and /ea/ tend to have very similar formant values at 50%, as shown by the small Euclidean distances between these two sequences for all speakers. For Spkrs 1, 2 and 4, this distance is much smaller than any other in Table 3. These values, however, are clearly not always closer to those of /a/ than to those of /e/. Instead, our four speakers appear to show two different patterns. We see one pattern for Spkrs 1 and 2 and another pattern for Spkrs 3 and 4. For Spkr 1 and Spkr 2, /ae/ is less distant from /e/ than from /a/ at mid point. The same is true regarding the sequence /ea/. For Spkr 3 and Spkr 4, on the other hand, both sequences /ae/ and, especially, /ea/ are closer to /a/ than to /e/ at mid point. ANOVAs followed by post-hoc Fisher tests were performed for the value of each of the first two formants at each of the three measured points for all tokens of all vowels and sequences, separately for each speaker. The alpha value was set at 1%. As expected, all the ANOVAs returned highly significant values ( p < 0.0001). All pairwise comparisons, however, did not result in significant differences. For the vowels and sequences relevant for this hypothesis (/a/, /ae/, /e/, /ea/), the pairs in Table 4 are not significantly different. The formant for which a significant difference was not found is indicated in the table. qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3 The formula that was used is: ED ¼ ððF1V1 ! F1V2Þ2 þ ðF2V1 ! F2V2Þ2 Þ (see Baltazani, 2006). 1916 J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 Table 4 Non-significant differences in formant values at 50% point Spkr 1 Spkr 2 Spkr 3 Spkr 4 /ae/ - /a/ F2 /ae/ - /e/ F2 F2 F2 /ea/ - /a/ F1, F2 F1 /ea/ - /e/ F2 F2 /ae/ - /ea/ F1, F2 F1, F2 F2 F2 The pairs of values that turned out not to be significantly different at mid point are identical for Spkrs 1 and 2. For F2, both /ae/ and /ea/ are similar to /e/. Again, the data for these two speakers are not consistent with the hypothesis of greater reduction of the higher vowel in the sequence. The clearest result for Spkrs 3 and 4 concerns the sequence /ea/, which shows convergence with /a/. In fact, as is apparent from Fig. 2 (below), already at the first measured point, 25% into the V or VV, values for both formants for /ea/ are very close to /a/ for these two speakers. That is, from the beginning /ea/ approaches /a/. This suggests a deletion/assimilation strategy: /ea/ [a]. Fig. 2. (Four panels) Plots of formant values at 25% point of V or VV for each of the four speakers. The ellipses enclose one standard deviation on each direction from the mean for each of the two formants.  J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 1917 Table 5 Mean F1/F2 values in Hz at mid point: /e/, /eo/, /o/, /oe/ Spkr 1 Spkr 2 Spkr 3 Spkr 4 /e/ 554/2308 585/2265 615/2102 445/1890 /eo/ 539/1680 568/1798 604/1521 495/1417 /o/ 586/1322 602/1434 662/1487 483/1247 /oe/ 518/1518 605/1732 631/1854 452/1505 From the duration results reported above, it appears that for these two speakers the duration of /ea/ is also not significantly different from that of /a/, again suggesting a strategy of V1 deletion. This deletion is consistent with Casali’s (1997:497) crosslinguistic finding that ‘‘at the boundary between two lexical words, elision is always of V1’’. These two speakers’ data thus offer some support for Hypothesis II, at least regarding /ea/. 4.4.3. Formants. Sequences of same height /eo/, /oe/ (Hypothesis III) Hypothesis III states that in sequences of equal phonological height the first of the two vowels will show greater reduction. Therefore, formant values at mid point should be closer to those of the second vowel in the sequence. Means at the 50% point are given in Table 5 (the values for /e/, which were also given in Table 2, are repeated here for convenience). Euclidean distances between sequences and component vowels in isolation based on the data in Table 5 are shown in Table 6. Regarding the sequence /eo/, for three speakers (Spkrs 2, 3 and 4) it is closer to /o/ than to /e/ at mid point, showing more reduction of the first vowel in the sequence, which is consistent with Hypothesis III. For the remaining speaker, Spkr 1, the values are almost identical for both comparisons. As for the sequence /oe/, for three speakers (Spkrs 1, 2 and 4), distances are smaller with respect to /o/ than with respect to /e/ (whereas the opposite is true for Spkr 3). For the vowels and sequences relevant for Hypothesis III, the pairwise comparisons in Table 7 do not show a significant difference. From Table 7, it is clear that F1 values tend to be similar for all vowels and sequences considered. Spkr 3’s data are fully consistent with the hypothesis, since for both /eo/ and /oe/, both formant values are statistically nondistinct from those of the second vowel at mid point. 4.4.4. Hypothesis IV Hypothesis IV concerns the possibility of finding evidence for dissimilation, where a vowel has a more extreme starting point as first element in a sequence than when produced as a single Table 6 Euclidean distance at mid point Spkr 1 Spkr 2 Spkr 3 Spkr 4 /eo/ - /e/ 628 467 581 476 /eo/- /o/ 630 365 67 170 /oe/ - /e/ 791 533 248 385 /oe/ - /o/ 207 298 368 260 /eo/- /oe/ 162 76 334 98 1918 J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 Table 7 Non-significant differences in formant values at 50% point Spkr 1 Spkr 2 Spkr 3 Spkr 4 /eo/ - /e/ F1 F1 F1 /eo/- /o/ F1 F1 F1, F2 F1 /oe/ - /e/ F1 F1 F1, F2 F1 /oe/ - /o/ F2 F1 F1 F1 /eo/- /oe/ F1, F2 F1, F2 F1 F2 vowel in interconsonantal position (or a more extreme final point in the case of sequences of falling sonority). Average formant values at 25% into the V or VV are plotted in Fig. 2. In Table 8, F1 values at the 25% point are shown for mid vowels and sequences starting with a mid vowel. An asterisk indicates those cases where the value of a sequence is significantly lower than that of the corresponding initial mid vowel (as determined by the ANOVA and post hoc tests). For Spkr 1 and Spkr 2, thus, /eo/ has a significantly higher starting point than /e/, suggesting a dissimilation process, /eo/ [jo]. For Spkr 1, the same phenomenon is also observable for /oe/ [we] (see Fig. 2). A more extreme gesture for /eV/ can mean not only an increase in height but also an increased fronting. In Fig. 2 we can observe that for Spkr 2 /ea/ has a more fronted realization than /e/ at 25% (F2 values are statistically different, p = .0001). On the other hand, no evidence was found for raising of mid vowels as the second member of sequence. Average values at the 75% point are plotted in Fig. 3 4.4.5. Summary of results As predicted, our experimental data show contraction in sequences of vowels across word boundaries (Hypothesis I). This is manifested in considerable shortening, so that VV sequences are often only slightly longer than a single vowel in the same context. Regarding our hypotheses of asymmetrical shortening of one of the vowels in contact (Hypotheses II and III), evidence for both of them was found in the data from some of the speakers, but not in a general manner. Finally, two speakers provided some evidence of raising of mid vowels in prevocalic position, an option that is at the root of certain dialectal developments (Hypothesis IV). Our data confirm the variation in reduction strategies that has been noticed in other studies on the basis of more subjective categorization. Table 8 F1 at 25% Spkr 1 Spkr 2 Spkr 3 Spkr 4 /e/ 539 574 620 437 /ea/ 563 644 731 550 /eo/ 466* 502* 560 469 /o/ 566 582 673 481 /oe/ 490* 549 625 465  J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 1919 Fig. 3. (Four panels) Plots of formant values at 75% point of V or VV for each of the four speakers. The ellipses enclose one standard deviation on each direction from the mean for each of the two formants. 5. Discussion Sequences of nonhigh vowels in Spanish show considerable variation in their realization, even when, as in our experiment, dialect, word stress and phrasal context are kept constant. Our data confirm the variation in reduction strategies that has been noticed in other studies. In this section, we want to consider some aspects of this variation. Consistently with other descriptions, in addition to temporal compression, our spectrographic analysis reveals two opposite tendencies. One strategy is for the difference between the vowel gestures in the sequence to be reduced (vowel assimilation). On the other hand, we also find that mid vowels may become more peripheral in contact with other nonhigh vowels, producing dissimilation within the sequence. Here we would like to remark on what seem to be two ‘‘opposite’’ hiatus resolution strategies. (12) ‘‘opposite’’ hiatus resolution strategies: Assimilation: ea > a Dissimilation (gliding): ea > ja 1920 J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 Assimilation and gliding seem to be universally available and competing strategies for the resolution of hiatus sequences involving mid vowels. As mentioned above in section 3.2, Navarro Toma´s (1977:68) points out the existence of these two solutions for word-internal hiatus sequences. Interestingly, in neighboring Basque dialects we find that one or the other strategy for the resolution of VV sequences with mid vowels has been phonologized (Hualde, 1990; Erdozia Mauleo´n 2001). In an area of Navarre, intervocalic / / and /g/ [ ] have been deleted, creating sequences of vowels in hiatus. In Uharte, where this deletion process is still optional, sequences of the type [ea], [eo] are found as a consequence of intervocalic consonant deletion. Both in Etxarri and Arbizu, the deletion of / / and /g/ has become obligatory and the VV sequences thus produced have been altered, but in the two ‘‘opposite’’ ways mentioned in (11). In Etxarri hiatus sequences have been solved by gliding of the first mid vowel: [ea] > [ja], [eo] > [jo]. In Arbizu, on the other hand, a mid vowel assimilated to the following vowel with which it came in contact: [ea] > [aa], [eo] > [oo], as shown in (13): (13) Neighboring Navarrese Basque varieties (Erdozia, 2001:406, see also Hualde, 1990) Uharte Etxarri Arbizu eran "ean jan aan ‘drink’ egon "eon jon oon ‘stay’ egosi "eosi josi oosi ‘boil’ We may surmise that after the deletion of intervocalic / / and /g/, both hiatus resolution strategies arose and that sociolinguistic factors determined that Etxarri and Arbizu speakers ultimately opted for different resolutions. Whereas assimilatory changes of the type [ea] > [a(:)] are easy to understand as a consequence of temporal compression, the transformation of mid vowels into high glides in the context of another nonhigh vowel is more difficult to conceptualize. As was mentioned in section 3, in word-internal position, the reduction of sequences with mid vowels to diphthongs is a widespread phenomenon in Latin American Spanish: /pelea´r/ [pelja´r] ‘to fight’, /peo´r/ [pjo´r] ‘worse’, /koe´te/ [kwe´te] ‘fire cracker’. In Spain such forms with ‘‘true diphthongs’’ are nowadays generally considered ‘‘rural’’ or ‘‘vulgar’’ (Navarro Toma´s, 1977:68) and may have receded from perhaps an earlier more general use. The same evolution; e.g. /ea/ > /ja/ is also well documented in the history of the Romance languages, e.g. VINEA > /vinja/ > /vi a/ ‘vineyard’. Since [j], [w] have more peripheral formant values than [e], [o], indicating a more extreme gesture, this change does not seem to follow directly from temporal reduction and overlap of the sequence. That is, changes of the type /ea/ [ja], /oe/ [we] increase the difference in height within the sequence and thus appear to be dissimilatory (the two vowels become more different from each other). The question is then how vowel contraction can produce dissimilation. A possible explanation involves both speaker and hearer, using Ohala’s view of sound change (e.g. Ohala, 1974, 1993). Along these lines, Gussenhoven (in press) suggests that, given the intrinsic relationship between vowel duration and height, shorter vowels may be perceived by the listener as higher than they really are, and in turn realized as such. In our case, very short mid vowels may be perceived as high glides, in Spanish dialects showing evolutions of the type /pelear/ > [pelja´r]. A. Nevins (p.c.) points out that more extreme vowel gestures may also allow for easier recoverability. Given the short duration of the sequence, increasing the dispersion within it will make it more clear that it is a sequence and not a single vowel. We found some evidence for a higher initial element in /eo/, /oe/ and for a more fronted initial articulation in /ea/ in the speech of some speakers. This suggests that perhaps dissimilation involves different strategies in sequences of equal and unequal phonological height. We may  J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 1921 have a higher target for /e/ in /eo/, but a more fronted target for /e/ in /ea/. Eventually, both a raised nonsyllabic /e/ and a more palatal nonsyllabic /e/ may be phonologized as a high glide. Further research would be needed to test this hypothesis. Appendix A Test sentences [o-e] 1. el mono enano es el ma´s pequen˜o de todos los monos. 2. un cuarto elegante estaba a disposicio´n de los congresistas. 3. el canto eterno de las aves inundaba el bosque de sonido. 4. un lago enorme se abrı´a ante nuestros ojos. 5. un saco entero siempre pesara´ ma´s que medio saco. 6. con un gesto hero´ico se despidio´ de sus soldados. [o-C] 7. un mono pequen˜o saltaba entre las ramas. 8. un cuarto privado estaba reservado para los congresistas. 9. el canto triste de los colibrı´es nos llenaba de angustia. 10. un lago seco aparecio´ en el horizonte. 11. un saco mediano es ma´s pequen˜o que un saco grande. 12. con un gesto poe´tico dijo adio´s al pu´blico. [C-e] 13. el gorrio´n enano canta al atardecer. 14. el andar elegante de las garzas sorprendio´ a los presentes. 15. un amor eterno no tiene fin. 16. un mar enorme era el que tenı´an que cruzar. 17. un camio´n entero trajo todo el mobiliario. 18. con un adema´n hero´ico dijo adio´s y se fue. [e-o] 19. la gente honrada no miente. 20. el enlace obligatorio habı´a que hacerlo en Lisboa. 21. el tinte oscuro de este cuadro le da un matiz enigma´tico. 22. la parte optativa aparece al final del capı´tulo. 23. un parque olvidado servı´a de reposo al pobre hombre. 24. era un bosque horrible, sin apenas vida animal. [e-C] 25. era gente feliz la de aquel paı´s. 26. un enlace lento dificultaba el transporte de mercancı´as. 27. su tinte particular hacı´a que fuera un cuadro inconfundible. 28. una parte menor de los fondos se empleaba para gastos diversos. 29. un parque tranquilo adornaba el centro de la ciudad. 30. era un bosque tenebroso, oscuro y sin vida. [C-o] 31. una mujer honrada nos devolvio´ el dinero. 32. la leccio´n obligatoria tenı´a que ver con los requisitos mı´nimos. 1922 J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 33. el color oscuro de la hierba indicaba que habı´a llovido. 34. la leccio´n optativa es la u´ltima del libro. 35. era un lugar olvidado de la mano de Dios. 36. era una cancio´n horrible, llena de notas discordantes. [a-e] 37. una casa hermosa se escondı´a entre los cipreses. 38. una espada enemiga acabo´ con su vida. 39. la carta elogiosa del anciano profesor disipo´ nuestras dudas. 40. su mirada enigma´tica me dejo´ desconcertado. 41. era una moneda etrusca de mucho valor. 42. los gatos de raza etı´ope son muy apreciados para cazar ratones. [e-a] 43. el elefante africano es de mayor taman˜o que el asia´tico. 44. era gente amable la de aquel lugar. 45. un paisaje agradable se extendı´a a ambos lados. 46. la corriente alterna cambia repetidamente de polaridad. 47. el enlace apropiado entre las dos ciudades serı´a el tren. 48. el toque artı´stico me extran˜o´ en un primer vistazo. [a-C] 49. la casa ma´s fea de todas era la del alcalde. 50. una espada francesa del siglo XV adornaba la pared. 51. una carta perdida debı´a de haber anunciado su llegada. 52. su mirada tranquila despejo´ mi temor. 53. la moneda peruana es el sol. 54. una raza primitiva de caballos se encuentra en Mongolia. [C-a] 55. el caima´n africano es ma´s peligroso que el americano. 56. un sen˜or amable nos explico´ como llegar. 57. era un lugar agradable para pasar el mes de agosto. 58. una direccio´n alterna de correo se encuentra en el sobre. 59. el papel apropiado para Juan serı´a el de cocinero. 60. la inspiracio´n artı´stica no llega siempre que queremos. Appendix B Means (and standard deviations) of formant values in Hz at 25%, 50% and 75% of Vor VV for each speaker (Table 2) 25% 50% 75% Spkr 1 a F1 735.55 (52.9) 796.63 (40.6) 753.77 (98.9) F2 1807.86 (250.7) 1852.25 (200.2) 1789.86 (264.5) ae F1 631.44 (55.3) 639.11 (53.3) 596.61 (82.1) F2 2203.5 (205.0) 2243.83 (120.8) 2318.33 (199.1) e F1 539.88 (79.8) 554.36 (83.5) 528.52 (88.1) F2 2315.11 (255.7) 2308.69 (287.2) 2238.72 (317.0)  J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 1923 Appendix B (Continued ) 25% 50% 75% ea F1 563.22 (82.85) 653.72 (73.0) 648.77 (110.2) F2 2443.16 (190.1) 2288.94 (227.8) 2027.27 (354.7) eo F1 465.66 (50.93) 539.61 (85.1) 516.72 (110.5) F2 2245.88 (273.2) 1680.77 (480.7) 1468.0 (389.86) o F1 566.30 (97.03) 586.52 (106.9) 561.55 (118.7) F2 1432.72 (299.3) 1322.44 (266.7) 1272.02 (311.9) oe F1 490.66 (63.64) 518.72 (73.0) 541.61 (74.4) F2 1366.22 (314.3) 1518.61 (501.0) 2119.72 (399.1) Spkr 2 a F1 829.30 (54.9) 886.66 (46.1) 861.80 (60.3) F2 1868.80 (110.9) 1872.22 (74.9) 1832.83 (111.9) ae F1 771.11 (61.0) 762.05 (44.8) 682.17 (45.9) F2 2061.88 (140.1) 2183.00 (137.2) 2318.8 (178.1) e F1 574.02 (55.8) 585.83 (65.1) 542.63 (80.7) F2 2230.5 (148.6) 2265.77 (146.3) 2242.63 (168.4) ea F1 643.83 (77.2) 754.83 (67.5) 783.72 (57.8) F2 2448.61 (140.7) 2222.44 (106.6) 2008.27 (82.4) eo F1 501.66 (69.6) 568.50 (56.6) 563.66 (50.6) F2 2322.16 (287.1) 1798.50 (220.0) 1564.38 (137.7) o F1 581.94 (72.5) 602.02 (72.8) 568.42 (80.1) F2 1475.88 (255.5) 1434.48 (192.7) 1386.54 (150.4) oe F1 549.16 (66.7) 605.11 (44.3) 576.72 (67.2) F2 1511.27 (239.6) 1732.88 (154.4) 1993.05 (247.6) Spkr 3 a F1 824.50 (154.7) 889.88 (126.4) 854.88 (80.1) F2 1670.47 (393.3) 1707.33 (346.8) 1694.66 (248.1) ae F1 757.61 (75.3) 758.61 (96.4) 687.44 (96.9) F2 1887.72 (462.3) 1882.61 (498.8) 1895.61 (475.8) e F1 619.83 (84.6) 615.16 (77.4) 575.66 (80.7) F2 2216.44 (386.1) 2102.05 (496.8) 1970.38 (494.6) ea F1 731.5 (110.9) 843.27 (54.0) 783.88 (74.0) F2 1790.83 (516.8) 1746.05 (319.5) 1695.16 (333.4) eo F1 560.05 (72.0) 604.44 (80.4) 604.50 (110.3) F2 1911.66 (299.4) 1521.61 (326.6) 1631.00 (195.0) o F1 673.00 (84.0) 662.02 (74.8) 596.8 (69.8) F2 1504.88 (294.1) 1487.97 (213.1) 1467.82 (227.6) oe F1 625.72 (55.4) 631.05 (56.1) 580.9 (78.8) F2 1672.16 (302.0) 1854.05 (217.6) 2011.38 (295.9) Spkr 4 a F1 577.91 (44.5) 617.16 (41.0) 611.91 (43.7) F2 1547.27 (74.6) 1545.27 (72.0) 1507.50 (81.6) ae F1 540.05 (25.1) 545.11 (28.7) 509.44 (46.5) F2 1642.44 (67.8) 1695.66 (83.6) 1739.11 (117.5) 1924 J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 Appendix B (Continued ) 25% 50% 75% e F1 437.52 (43.6) 445.19 (43.4) 428.05 (39.5) F2 1875.66 (135.5) 1890.33 (116.8) 1857.77 (121.5) ea F1 550.22 (66.8) 595.61 (51.1) 586.66 (46.8) F2 1638.22 (261.8) 1658.33 (143.1) 1584.00 (95.3) eo F1 468.94 (34.5) 495.33 (37.7) 480.00 (38.8) F2 1576.72 (211.3) 1417.22 (155.3) 1289.88 (101.1) o F1 481.55 (55.5) 483.61 (54.7) 456.25 (56.6) F2 1319.41 (164.3) 1247.83 (151.8) 1224.05 (164.6) oe F1 464.61 (45.5) 452.94 (37.0) 420.72 (42.2) F2 1402.44 (223.1) 1505.16 (154.2) 1644.44 (161.5) References Aguilar, L., 1999. Hiatus and diphthong: acoustic cues and speech situation differences. Speech Communication 28, 57–74. Aguilar, L., 2003. Effects of prosodic and segmental variables on vowel sequences pronunciation in Spanish. In: Sole´, et al., pp. 2111–2114. ? Aguilar, L., 2005. Los enlaces voca´licos: cuestio´n de dominios proso´dicos?. Revista Internacional de Lingu¨´ıstica Iberoamericana 6, 29–48. Alba, M., 2005. Hiatus resolution between words in New Mexico Spanish: a usage-based account. Doctoral dissertation, University of New Mexico, Albuquerque. Baltazani, M., 2006. Focusing, prosodic phrasing, and hiatus resolution in Greek. In: Goldstein, et al. (Eds.), pp. 473–494. Cabre´, T., Prieto, P., 2004. Prosodic and analogical effects in lexical glide formation in Catalan. Probus 16, 113–150. Cabre´, T., Prieto, P., 2006. In: Martı´nez-Gil, F., Colina, S. (Eds.), Exceptional hiatuses in Spanish. pp. 205–238. Casali, R., 1997. Vowel elision in hiatus contexts: which vowel goes? Language 73, 493–533. Chitoran, I., Hualde, J.I., 2007. From hiatus to diphthong: the evolution of vowel sequences in Romance. Phonology 24, 37–75. Erdozia Mauleo´n, J.L., 2001. Sakana erdialdeko euskara. Nafarroako Gobernua, Pamplona/Irun˜a. Garrido, M., in preparation. Diphthongization of nonhigh vowels in Latin American Spanish. Doctoral dissertation, University of Illinois at Urbana-Champaign. Gussenhoven, C., in press. A vowel height split explained: compensatory listening and speaker control. In: Cole, J., Hualde, J.I. (Eds.). Harris, J., 1970. Vowel sequences in Spanish. Linguistic Inquiry 1, 129–134. Harris, J., Kaisse, E.M., 1999. Palatal vowels, glides and obstruents in Argentinian Spanish. Phonology 16, 117–190. Hualde, J.I., 1990. Vowel Lengthening in Basque. Folia Linguistica 24, 269–288. Hualde, J.I., 1994. La contraccio´n sila´bica en espan˜ol. In: Demonte, V. (Ed.), pp. 629–647. Hualde, J.I., 1999. In: Gutie´rrez-Rexach, J., Martı´nez-Gil, F. (Eds.), Patterns in the lexicon: Hiatus with unstressed high vowels in Spanish. pp. 182–197. Hualde, J.I., 2005. The sounds of Spanish. Cambridge Univ. Press, Cambridge. Hualde, J.I., Prieto, M., 2002. On the diphthong/hiatus contrast in Spanish: some experimental results. Linguistics 40, 217–234. Hutchinson, S.P., 1974. Spanish vowel sandhi. In: Bruck, A., et al. (Eds.), pp. 184–192. Jenkins, D., 1999. Hiatus resolution in Spanish: Phonetic aspects and phonological implications from northern New Mexican data. PhD Dissertation, University of New Mexico, Albuquerque. Lope Blanch, J., 1996. Me´xico. In: M. Alvar (Ed.), pp. 81–89. Martı´nez-Gil, F., 2006. Dialectal variation and phonological change as reranking of constraints: Hiatus resolution by syllable merger in three dialects of contemporary Spanish. Ohio State University, ms. Navarro Toma´s, T., 1977 [1918]. Manual de pronunciacio´n espan˜ola, 19th ed. CSIC, Madrid. Ohala, J., 1974. Experimental historical phonology. In: Anderson, J.M., Jones, C. (Eds.), pp. 353–389. Ohala, J., 1993. The phonetics of sound change. In: Jones, C. (Ed.), pp. 237–278. Quesada Pacheco, M.A., 1996. El espan˜ol de Ame´rica Central. In: Alvar, M. (Ed.), pp. 101–115.  J.I. Hualde et al. / Lingua 118 (2008) 1906–1925 1925 Roca, I., 1997. There are no ‘glides’, at least in Spanish: an optimality account. Probus 9, 233–265. Sedano, M., Bentivoglio, P. 1996. Venezuela. In: Alvar, M. (Ed.), pp. 116–133. Vaquero, M., 1996. Antillas. In: Alvar, M. (Ed.), pp. 51–67. Further reading Alvar, M. (Ed.), 1996. Manual de dialectologı´a hispa´nica, vol. 2, El espan˜ol de Ame´rica. Ariel, Barcelona. Anderson, J.M., Jones, C. (Eds.), 1974. Historical Linguistics, vol. 2. North-Holland, Amsterdam. Boersma, P., Weenink, D., 1996–2006. Praat: doing phonetics by computer (Version 4.4.13) www.praat.org. Bruck, A., Fox, R.A., La Galy, M. (Eds.), 1974. Papers from the Parasession on Natural Phonology. Chicago Linguistic Society, Chicago. Cole, J., Hualde, J.I. (Eds.), in press. Laboratory Phonology, vol. 9. Mouton de Gruyter, Berlin. Demonte, V. (Ed.), 1994. Grama´tica del espan˜ol. Publicaciones de la Nueva Revista de Filologı´a Hispa´nica VI, Mexico, D.F. Frota, S., Viga´rio, M., Freitas, M.J. (Eds.), 2005. Prosodies. Mouton de Gruyter, Berlin. Goldstein, L., Whalen, D., Best, C. (Eds.), 2006. Laboratory Phonology, vol. 8. Mouton de Gruyter, Berlin. Gutie´rrez-Rexach, J., Martı´nez-Gil, F. (Eds.), 1999. Advances in Hispanic Linguistics: Papers from the 2nd Hispanic Linguistics Symposium, vol. 1. Cascadilla Press, Somerville. Jones, C. (Ed.), 1993. Historical Linguistics: Problems and Perspectives. Longman, London, pp. 237–278. Martı´nez-Gil, F., Colina, S. (Eds.), 2006. Optimality-Theoretic Studies in Spanish Phonology. John Benjamins, Amsterdam. Simonet, M., 2005. Prosody and syllabification intuitions of [CiV] sequences in Spanish and Catalan’’. In: Frota, S., et al. (Eds.), pp. 247–267. Sole, M.J., Recasens, D., Romero, J. (Eds.), 2005. Proceedings of the 15th International Congress of Phonetic Sciences, Barcelona. Causal Productions, CD-Rom, Adelaide, Australia.