4.1.1. Vowel doubling in EAS: arguments in favour

The next study to examine EAS vowel doubling after ^{Navarro Tomás (1938}, ¹⁹³⁹) was ^{Rodríguez-Castellano and Palacio (1948a}). This was the first article to study the vowels of a specific area within Eastern Andalusia (Cabra, in the province of Cordoba) in detail.

In their impressionistic study, ^{Rodríguez-Castellano and Palacio (1948a}) stated that differences between open and closed vowels were more noticeable in Cabra than in Castilian Spanish; this was especially obvious for /e/ and /o/. Decades later, ^{Peñalver Castillo (2006}) re-examined the speech of Cabra and found vowel harmony and lengthening of vowels preceding /s/ deletion, concluding that vowel opening affects all vowels, especially /a/, /e/, and /o/. Likewise, ^{Martínez Melgar (1986}) also reported that all EAS vowels open when they precede a deleted /s/. According to ^{Martínez Melgar (1986)}, this opening is more obvious in /e/ and /o/ and less in /i/ and /u/, although these views are challenged by other authors who reject opening of /i/ and /u/ preceding /s/ deletion (e.g. ^{Sanders, 1998}).

Interestingly, ^{Rodríguez-Castellano and Palacio (1948a}) register the same pronunciation for /as/, /aƟ/, and /að/ once the final consonants are deleted, while ^{Salvador (1957}) claims that /e/ preceding deleted /s/ is different from /e/ preceding deleted /θ/; this is proved acoustically in ^{Herrero de Haro (2016b}). However, ^{Salvador (1957)} did not believe this was the case for /i/.

^{Alonso et al. (1950}) also considered the effect of the deletion of different consonants on the preceding vowel, being the first authors to describe different types of /a/. According to ^{Alonso et al. (1950)}, /a/ is more open in palos 'sticks' than in barcos 'boats'; there is aspiration in más 'more' and the vowel is more nasal and palatal than in mar 'sea' and in mal 'bad'; /a/ in más is pronounced with the tongue moving towards the lower teeth, and in mar and mal it moves towards the upper ones. In mal the tongue is raised as soon as the vowel stops being pronounced and in mar this happens later. Despite all this, ^{Alonso et al. (1950)} believe that, generally, vowels express the difference between singular and plural, although their analysis did not include perceptual analyses. For ^{Zubizarreta (1979}), however, mal, mar and más sound the same once /l/, /r/ and /s/ have been deleted. Likewise, ^{Alarcos Llorach (1983}) claims that all vowels are pronounced the same before deleted /r/, l/, or /x/, and ^{Cerdà Massó (1992}) claims that es and el are pronounced ENT#091;e̞ENT#093; in EAS (the symbol ̞ indicates lowered vowels).

The debate regarding vowel quality in EAS was quite intense around the 1950s, with an extensive body of impressionistic studies. ^{Rodríguez-Castellano and Palacio (1948a}) described different types of phonetic variants of /e/ depending on the following consonant; however, they also reported a totally open /e/ with phonemic value in Cabra, which was more obvious in stressed syllables (e.g. pelos ENT#091;ˈpe̞lo̞ENT#093; 'hairs'). On the other hand, ^{Alonso et al. (1950}) did not find any obvious differences between the /e/ of tres 'three', miel 'honey', and pies 'feet'. ^{Rodríguez-Castellano and Palacio (1948a)} claimed that, in Cabra, /e/ in singular words was much more closed than in Castilian Spanish, although they thought it was due to the effect of comparing open and closed /e/. Likewise, Rodríguez-Castellano and Palacio (1948a) and Alonso et al. (1950) found that /o/ tended to be more closed in singular. ^{Alonso et al. (1950)} also believed that singular vowels were more closed, especially /e/, and for ^{Alvar (1955a}), ^{Salvador (1957}), ^{Alarcos Llorach (1958}) and, years later, for ^{Peñalver Castillo (2006}), singular vowels were more closed in Andalusian than in Castilian Spanish. This closing of vowels in EAS seems to be a way of making vowel opening more noticeable. ^{Lloret and Jiménez (2009}) explained how something similar happens in the Italian dialect of Ascrea, where high unstressed final vowels (which are grammatically meaningful) close medial stressed vowels to reinforce that morphological element. ^{Alonso et al. (1950)} and ^{Salvador (1957)} also believed that vowels were more closed when they appeared with vowels of the same quality in a word (e.g. eje 'axle'). Furthermore, ^{Alonso et al. (1950)} and ^{Salvador (1957)} equated the openness of /e/ in plural to that of /a/ in singular.

Four decades later, ^{Martínez Melgar (1994}) carried out an acoustic study in the provinces of Jaen, Granada and Almeria, and concluded that vowels were more open when preceding /s/ deletion. ^{Martínez Melgar (1994)} claimed that /a/, /e/, and /i/ were opened and backed preceding /s/ deletion, and that /o/ and /u/ were opened and fronted, with /e/ and /o/ presenting the biggest differences in F1. ^{Martínez Melgar (1994)} concluded that differences between singular and plural vowels in EAS were always meaningful (e.g. consistent). ^{Corbin (2006}) additionally noted that "the laxing of a vowel is characterised by centralisation"; however, she found that in the centralisation of vowels, fronting and backing was a less robust change than lowering.

The debate regarding EAS vowels is somewhat more complex when it comes to /i/ and /u/. ^{Rodríguez-Castellano and Palacio (1948a}) found subtle differences between different types of /i/ and /u/, but these were less obvious than those for /e/ and /o/; they agreed with ^{Navarro Tomás (1938}, ¹⁹³⁹) on the fact that /i/ and /u/ do not double. ^{Navarro Tomás (1938}, ¹⁹³⁹), ^{Rodríguez-Castellano and Palacio (1948a)}, ^{Salvador (1977}), ^{Alarcos Llorach (1983}), ^{Martínez Melgar (1994}), ^{Sanders (1998}), ^{Jiménez and Lloret (2007}), and ^{Lloret and Jiménez (2009)} considered vowel opening a compensation strategy, as it only appears when the consonant is deleted, and they believed that the aspiration of a sound caused vowels to open. ^{Mondéjar Cumpián (1979}) went a bit further and insisted that the aspiration of /x/, /s/, /θ/, /r/, and /l/ opened the previous vowel, being the first author to posit that all consonants can be aspirated in coda in Andalusian Spanish. This aspiration marks the semantic differences, but once this disappears, openness becomes the distinctive feature, considering vowel opening as the historic end of the process of /s/ elimination in coda. In other words, ^{Mondéjar Cumpián (1979)} saw aspiration and opening, respectively, as a step in the process of disintegration of /s/ in coda, and the final stage (at least for the time being) of that very same process.

However, ^{Contreras Jurado (1975}) thought that in EAS, singular/plural distinction was maintained thanks to the intensification of a characteristic already present prior to /s/ loss, which ^{Cerdà Massó (1992}) considered the most original approach; ^{Contreras Jurado (1975)} posited that /s/ had already caused the preceding vowel to open before /s/ was deleted, and that this deletion did not increase the openness of the vowel. EAS intensified this feature extending it to the stressed vowel and, in several other cases, to the rest of the vowels in a word. ^{Contreras Jurado (1975)} was the first researcher to posit this, which would also be posited with slight modifications in ^{Hualde and Sanders (1995}). In their theoretical analysis, ^{Hualde and Sanders (1995)} posited that vowels preceding /s/ were raised in EAS before /s/ was deleted, which made it possible to omit /s/ and still mark the plural. According to ^{Hualde and Sanders (1995)}, this would explain why EAS is the only variety of Spanish which deletes final /s/ and presents vowel doubling, meaning that a previous distinction existing before /s/ deletion was phonemicised once that that deletion had taken place. ^{Hualde and Sanders} believe that:

the raising of unstressed mid vowels in absolute final position which, in our hypothesis, was the original trigger of a sg/pl distinction based on vowel quality, appears to be on its way out, given its negative connotations; whereas, lowering or laxing in plural, a secondary development which does not carry the same social stigma, is establishing itself as the significant feature in the opposition. (¹⁹⁹⁵, p. 432)

^{García Mouton (1992}) explained how some women maintained /s/ and still opened the preceding vowel, which supports ^{Contreras Jurado's (1975}) and ^{Hualde and Sanders' (1995}) position; however, ^{García Mouton (1992)} also noticed that women in Jaen, Granada, and Almeria alternated between open vowel and ENT#091;ahENT#093;, which presents vowel opening as an effect of ENT#091;hENT#093; deletion.

^{Corbin (2006}) analyses this issue examining constraint rankings and claims that "if lenition were ordered before laxing, laxing would not occur, therefore laxing must occur first" (p. 56), which supports ^{Hualde and Sanders (1995}). Furthermore, ^{Corbin (2006)} believes that "the relationship between laxing and lenition appears to be one of counter-bleeding, since an ordering of lenition before laxing would eliminate the environment for laxing" (Kiparsky, 1973, cited by ^{Corbin, 2006}, p. 32).

^{Rodríguez-Castellano and Palacio (1948a}) pointed out how young people in the Cabra area no longer needed a brief aspiration of /s/ to express plurality; they could do it just with vowel opening, which also presented the phenomenon as an emerging trend, an opinion also defended by ^{García Mouton (1992}). Furthermore, ^{Rodríguez-Castellano and Palacio (1948a)} argue that vowels before deleted /l/, /r/, /s/, or /θ/ in words that are not plurals are not as open as in plural words, and they do not give those vowels a phonemic value. For ^{Alonso et al. (1950}), vowels of singular words open when they precede deleted /r/, /l/, /s/, or any other aspirated consonant, and the only consonant that closes vowels in these contexts is /n/ (e.g. ratón 'mouse'). Rodríguez-Castellano and Palacio (1948a) argue that in those cases, the last vowel is the only one that opens. I disagree with this, as a preliminary analysis of my data for a future article indicates that both vowels open in words like crater 'crater' following /r/ deletion.

^{Alonso et al. (1950}) decided to study high socio-cultural classes to prove that vowel doubling was present in all socio-cultural strata of the city of Granada. ^{Salvador (1957}) and ^{Zamora Vicente (1960}, p. 290) also considered it present across all levels of society and, more recently, ^{Peñalver Castillo (2006}) found it across all socio-cultural levels and age groups in Cabra. However, ^{Alonso et al. (1950)} and ^{García Mouton (1992}) indicated that women showed a higher tendency to maintain /s/.

According to ^{Salvador (1957}), vowel opening started in the nineteenth century and this system is the only one present in young people and for men of all ages; even more, the very few older ladies who maintain /s/ understand (and even use at times), understand the system based on vowel doubling. ^{Alonso et al. (1950}) maintained that vowel doubling appeared in the speech of Granada and Almeria, that it could also be found in people from Cordoba, and that vowel opening reaches the stressed vowel, not only the final one. ^{Alonso et al. (1950)} conclude that these vowels are constant in all speakers, but that consonants seem to vary between speakers and even within the same person.

Some authors (e.g. ^{Navarro Tomás, 1938}; ^{Rodríguez-Castellano and Palacio, 1948a}; ^{Salvador, 1957}: and ^{García Marcos, 1987}) believe that the process of /s/ loss in coda has not finished yet in EAS. The fact that Andalusian Spanish is a variety in evolution, as ^{García Marcos (1987)} and ^{O'Neill (2010}) suggest, could explain why the use of /s/ and of vowel lengthening are so variable.

^{Alvar (1955a}) explained that, despite the limitation of the data gathered thus far, he could point out that the loss of /s/ in plurals had caused various processes: a) vowel opening with a very noticeable phonological value; b) vowel opening with no phonological value, which was the beginning of distinction; c) no distinction between singular and plural vowels. Interestingly, ^{Salvador (1977}) also commented on this, arguing that, in some places in western Andalusia, speakers pronounce vowels similar to those in Eastern Andalusia, although they cannot distinguish between them phonologically. This might be, I believe, because they have recently lost the aspiration of /s/ in those areas and they have not had enough time to phonologise those open vowels. ^{Poplack (1980}) also mentions that in Puerto Rican Spanish, vowels preceding /s/ deletion are phonetically different, but that there is no phonological difference. ^{Zubizarreta (1979}) presents an idea which seems to be in the middle, as she believes that to understand Andalusian vowel harmony, we need to differentiate between phonological and phonetic laxing.

Even those who support vowel doubling do not agree on which vowels double. ^{Navarro (1938}, ¹⁹³⁹), ^{Rodríguez-Castellano (1948a}), ^{Alonso et al. (1950}), ^{Zamora Vicente (1960}), ^{Ueda (1993}), and ^{Martínez Melgar (1994}) support vowel doubling only in /a/, /e/, and /o/. On the other hand, ^{Alvar (1955a}, ^1955b) was the first author to support vowel doubling in the five Castilian vowels. ^{Martínez Melgar (1986)} claims that vowels tend to be open in plural, and although she noticed certain opening in /i/ and /u/, she did not consider it relevant. ^{Jiménez and Lloret (2007}) also studied word-final vowel laxing following /s/ deletion and concluded that "mid vowels open at the end of words, low vowels further front, and final high vowels open to a lesser degree."

^{Salvador (1957}) also supported the idea of vowel doubling of the five Castilian vowels, although he said that open /i/ is usually found in word-final position in verbs and that opening of /u/ is not as regular as in other vowels. ^{Salvador (1977)}, on the other hand, would reject /u/ doubling. However, other authors would reject ^{Salvador's (1957}, ¹⁹⁷⁷) claims: ^{Zubizarreta (1979}) distinguished between phonetic and phonological laxing and she believed that /i/ and /u/ undergo phonetic but not phonological laxing; for ^{Martínez Melgar (1994}), /i/ and /u/ do not have vowel harmony; and for ^{Jiménez and Lloret (2007}) and ^{Lloret and Jiménez (2009)}, /i/ and /u/ do not present vowel harmony in stressed position. Likewise, ^{Jiménez and Lloret (2007)} and ^{Lloret and Jiménez (2009)} believe that /i/ and /u/ are the only vowels that do not lax in post-tonic position.

^{Salvador (1977}) believes that openness enables speakers to differentiate between singular and plural vowels in EAS, as this openness is always there and it was the first feature that researchers noticed. However, he analysed EAS vowels using Trubetzkoy's binary categories and concluded that the difference between vowels and their open counterparts is one of articulatory tension.

^{Alvar (1973}, map 1696) studies vowel doubling before deleted word-final /s/, with western Almeria and the southern half of Granada being the areas where the phenomenon has been studied the most and where more evidence of vowel doubling has been found. ^{Alvar (1973)}, ^{García Marcos (1987}) and ^{García Mouton (1992}) also notice differences regarding vowel doubling: different degrees of openness in verbs and in plurals, such as in the towns that ^{Alvar (1973)} identifies as Ma301 (Ronda), Ma402 (Salares), and J101 (Santa Elena); differences between men and women, as in locations J204 (Villacarrillo), Al202 (Contador), Al203 (Oria), and Al200 (Topares); differences depending on age and social background, such as in Al508 (Almeria) and Gr513 (Lujar); and towns where vowel splitting is a recent and imported phenomenon (e.g. Gr200 (Puebla de Don Fabrique)). ^{Alvar (1973}, map 1696) shows that the most common way of distinguishing between vowel and vowel plus deleted /s/ in EAS is through vowel opening, although Alvar (1973, map 1696) also shows vowel lengthening, aspiration, and metaphony as relevant features in some towns. However, ^{Alvar (1973)} also noticed that in some towns, vowel splitting is being lost and vowel differences based on vowel quality are being neutralised. ^{García Marcos (1987)}, for example, noted an increase in the tendency to insert aspiration to mark deleted /s/ on the coast of the province of Granada, which contrasts with ^{Peñalver Castillo's (2006}) findings. ^{Peñalver Castillo (2006)} found some aspirations as well, but he believed that the most common feature was to delete /s/, especially in younger speakers. ^{Peñalver Castillo (2006)} found aspiration mainly in rural areas around the town of Cabra, although ^{García Marcos (1987)} believed that ENT#091;hENT#093; preceded by opened vowel had prestige on the coast of the province of Granada, which is why its frequency increases as we go up the social scale. ^{García Marcos (1987)} indicated that the only places which showed a different tendency were those with less access to media or schooling, or those places where social prestige was given to solutions that were more common outside of that region (either Castilian ENT#091;sENT#093; or Sevilian ENT#091;hENT#093;).

^{Alvar (1973}, maps 1613, 1620, and 1625) studies the words zagal 'kid', mar 'sea', and más 'more'. In most towns, /al/, /ar/, and /as/ are either reduced to the same vowel following consonant deletion, or mar and más are distinguished thanks to aspiration, vowel lengthening, or vowel quality. However, in some exceptional cases, the three vowels have different quality following consonant deletion, as in Al507 (Berja). Alvar (1973, maps 1626 and 1629) studies the words tos 'cough' and voz 'voice'; /os/ and /oӨ/ are neutralised into the same vowel once the final consonant is deleted, although we do not know whether that is due either to seseo, ceceo, or other causes. However, /os/ and /oθ/ are distinguished in some towns even when the consonant is deleted (e.g. Gr303 (Algarinejo), Gr503 (Escuzar), or Al501 (Alboloduy)).

^{Sanders (1998}) studied whether EAS laxing was truly systematic and to what degree, as some factors, such as sociolinguistic variation, can influence the quality and consistency of a phonetic sampling. ^{Sanders (1998)} found that the alternation in vowel quality between singular and plural forms in EAS was consistent and phonemic, and that those alternations involve mid and low vowels. However, ^{Sanders (1998)} argued that there is limited evidence that native speakers of EAS consistently use vowel laxing to make distinctive contrasts in normal speech. ^{Sanders (1998)} asserted that the variation between plural and singular /e/ and /o/ was significant in all contexts, but that in /a/ it was statistically significant either in F1 or F2, depending on the context. ^{Sanders (1998)} supports, at least in part, a modified version of EAS vowel doubling.

^{Jiménez and Lloret (2007}) noticed another degree of opening already identified in ^{Navarro Tomás (1938}), positing that /l/ and /r/ do not systematically trigger opening of vowels, and that if they do, the vowels are not as open as when they precede /s/ deletion. However, to my knowledge, no further study has been carried out to analyse whether this opening has phonemic value.

In ^{Lloret and Jiménez (2009}), the same authors also consider the implications of /x/ deletion and said that obstruents, other than /s/ and /x/, are omitted word-finally as in many other varieties of Spanish without causing vowel opening in the preceding vowel. However, ^{Herrero de Haro (2016b}) posits that vowel opening also happens in vowels preceding other deleted consonants, such as /r/ and /θ/, which he analyses acoustically.

Regarding vowel doubling, ^{Lloret and Jiménez (2009}) believe that even though fronting and -ATR (advanced tongue root) are transferred from the omitted /s/ to the preceding /a/, only -ATR is transferred to other vowels in the harmonic process because -ATR is a new category in the system and it can be transferred to other vowels without altering the existing system. Meanwhile, fronting already existed in the vowel system and if it were transferred to other vowels, it would alter distinctive information that already existed.

To date, ^{Alvar (1973}) is the work with more information on EAS and it shows some interesting instances of further vowel doubling. The author presents several cases in which some vowels seem to acquire a phonological value after certain consonants are deleted. ^{Alvar (1973}, maps 1532 and 1533) shows the pronunciation of the words mies 'grain' and miel 'honey'. The maps show that /es/ and /el/ are reduced to the same vowel after /s/ and /l/ are deleted, although the vowel in question varies from area to area. However, in some towns, both vowels are still distinguished once the consonant is deleted; in J101 (Santa Elena) and in Gr301 (Colomera), for example, the vowels have different quality, thus, showing further vowel splitting. In other towns, however, those vowels are distinguished thanks to aspiration.

^{Alvar (1973}, maps 1543 and 1544) studies the words farol 'lantern' and flor 'flower'. In AL509 (Balerma) and J103 (Bafios de la Encina), the /o/ pronounced after the consonant is deleted has a different quality, which shows a distinction between /o/ preceding deleted /l/ and /o/ preceding deleted /r/. Alvar (1973, maps 1552 and 1553) studies the words horno 'oven' and hoz 'sickle', and the vowel /o/ before deleted /r/ and before deleted /θ/ have different quality in some towns as well (e.g. Al509 (Balerma) or Al401 (Tahal)). ^{Herrero de Haro (2016b}) shows further vowel splitting, with different vowels resulting before /-s/, /-θ/, and /-r/ deletion. Further vowel doubling like this, however, has not been studied to date.

If, as the authors above have claimed, vowel opening has been phonologised, then the question of why that happened is also relevant. ^{Hernández-Campoy and Trudgill (2002}) review several studies in which this is explained as a process of functional compensation in which semantically relevant information is preserved following /s/ deletion; however, their position is different. ^{Hernández-Campoy and Trudgill (2002)} believe that, unlike other varieties of Spanish, functional explanations are not valid for Southern Peninsular Spanish /s/ deletion. They accept an eight-vowel system in EAS (and in Murcian Spanish), due to the phonologisation of originally allophonic differences but they do not believe this was motivated to compensate for /s/ loss. They support their position by the fact that vowels have become open in cases where plurals can be differentiated from singulars even after /s/ has been lost (árbol 'tree' vs. árboles 'trees'). Furthermore, the vowel change happens word-medially and word-finally, and the change also occurs following the deletion of other consonants (e.g. calor 'heat'). Finally, forms which were distinct before consonant deletion word-finally now are homophonous (e.g. mar 'sea', más 'more', and mal 'bad'), and there are no open equivalents for /i/ or /u/. Hernández-Campoy and Trudgill (2002) believe that if vowel opening were due to functional compensation, then open vowels of /i/ and /u/ would have arisen.

4.1.1.1. EAS vowel system models

The debate regarding EAS vowels has not been limited to the phonological value of vowel doubling, but also to how to represent this vowel system. The different systems proposed for EAS vowel phonemes are explained below.

The first graphical illustration of a vowel system for EAS appeared in ^{Alarcos Llorach (1949}), although this was based on the data presented in ^{Navarro Tomás (1938}, 1939) (see Figure 1).

Figure 1:  EAS Vowel system posited by ^{Alarcos Llorach (1949}) for the data presented in ^{Navarro Tomás (1938}, ¹⁹³⁹). 

^{Alvar (1955a}) posited that the triangular vowel system of Spanish had developed into a quadrangular one where vowels have two types of timbre, and he considered it necessary to add EAS to those systems of five degrees of openness considered "extremely rare", as seen in Figure 2.

Figure 2: EAS vowel system posited by ^{Alvar (1955a}). 

^{Salvador (1957}) also proposed a vowel system for EAS. His system had six degrees of openness (Figure 3), and at the time of the article the only language known to have six degrees of openness was Gweabo, a language from Liberia which ^{Trubetzkoy (1949}, p. 119) mentioned as a rare example:

Figure 3: EAS vowel system posited by ^{Salvador (1957}) 

^{Alarcos Llorach (1958}) approached the study of EAS vowels from a totally different angle. At that point, linguists had only talked about vowel doubling, but ^{Alarcos Llorach (1958)} proposed vowel system doubling. He argued that EAS singular and plural vowels present a similar case to stressed and unstressed vowels; they are both the same phoneme. Alarcos posited that x (a co-vowel) makes vowels longer before pause, geminates the following consonant and opens the vowel. ^{Alarcos Llorach (1958)} concluded that x would be voiceless as it devoices the following consonant; the closest sound to this would be ENT#091;hENT#093;. ^{Alarcos Llorach (1958}, ¹⁹⁸³) mentioned 10 different vowels, but they belonged to two different systems (see Figure 4).

Figure 4: EAS vowel system posited by ^{Alarcos Llorach (1983}). 

Thus, ^{Alarcos Llorach (1958}) posited a phoneme /h/ in EAS with various pronunciations:

1. ENT#091;hENT#093; voiceless (and sometimes voiced) at the beginning of a word;

2. ENT#091;hENT#093; voiced between vowels;

3. ENT#091;hENT#093; or gemination before stop, nasal, or lateral consonant;

4. Spirant voiceless or semi-voiced consonant before fricatives; and

5. It lengthens final vowels before pause.

^{Alarcos Llorach (1983}) supported the same theory, although he admitted that vowels preceding deleted /s/ were not always longer. However, he continued defending his position that the limitation of contrast between word-final vowels and vowels preceding word-final /s/ deletion to certain positions meant that both vowels belonged to different systems.

^{Mondéjar Cumpián (1979}) and ^{Cerdà Massó (1992}) supported the latter system. However, ^{Gómez Asensio (1977}) rejected it, as he considered it too schematic and claimed that it did not operate within phonetic reality. ^{Gómez Asensio (1977)} argued that Andalusian vowels were not always long in plurals, that openness was the relevant distinctive feature, and that vowel quality was a redundant feature. Years later, ^{Alarcos Llorach (1983}) still defended the theory posited in Alarcos Llorach (1958) and rejected vowel doubling in EAS, positing that an extra prosodeme distinguished singular and plural EAS vowels.

^{Contreras Jurado (1975}) supported the idea that the EAS vowel system is the same one as in Castilian Spanish, but argued that there was a prosodeme which did not exist in Castilian Spanish (see Figure 5). This prosodeme is used as a numeric marker and it is redundant when there is a hangover of a deleted /s/. When the consonant /s/ disappears totally, the opposition -prosodeme/+prosodeme as numeric marker can be observed in EAS, and vowel closing seems to indicate a tendency to make this numeric prosodeme more effective. The singular/plural distinction is much more effective thanks to this.

Figure 5:  ^{Contreras Jurado's (1975}) singular vs. plural scheme for EAS. 

According to ^{Contreras Jurado's (1975}) hypothesis,vowels in singular EAS words would become less closed in the future, while plural EAS vowels would remain unchanged.

^{Salvador (1977}) revised the model presented in ^{Salvador (1957)} and concluded that the vowels in tu 'your -before sing. noun-' and tus 'your -before plu. noun-' are not distinguished, so he removed open /u/ from his 1957 model. However, he asserted that there are nine vowel phonemes in EAS. ^{Salvador (1977)} understood that a system with 6 degrees of openness can be very complex, but he considered that there might be other features, such as lengthening, which support such weak differences between vowels. Likewise, ^{Salvador (1977)} acknowledged that the lengthening which we associate with open vowels in final position could be a general term we use to describe other phenomena which we cannot comprehend (Figure 6).̞

Figure 6: EAS vowel system posited by ^{Salvador (1977}) 

^{Gómez Asensio (1977}), on the other hand, did not believe that it was possible to explain all the differences found within EAS vowels with the four features which authors had been using until then, so he used generative phonology to analyse these vowels (see Figure 7). ^{Martínez Melgar (1986}) had a similar opinion, as she thought that phonemic models cannot explain the vowel system of EAS. ^{Gómez Asensio (1977)} supports ^{Salvador's (1957}) system, but he decided to add openness as a feature to be able to distinguish between all the vowels suggested by ^{Salvador (1957)}. ^{Gómez Asensio (1977)} eliminated redundant information and changed from a binary system +/- to one with a numerical coefficient (0 meaning no feature, blank meaning redundant feature, and a number depending on the intensity of a feature). As a results, ^{Gómez Asensio (1977)} was the first author to scale the features of EAS vowels.

Figure 7:  ^{Gómez Asensio's (1977}) representation of EAS vowel 

These different models highlight the complexity of EAS vowels and illustrate how the debate is not limited just to whether or not vowel doubling exists. Choosing a vowel system which represents EAS vowel doubling accurately proves difficult, but what remains clear is that the triangular vowel system of Castilian Spanish cannot be used to represent EAS vowels.

4.1.2. Vowel doubling in EAS: arguments against

As explained in ^{Navarro Tomás (1938}, ¹⁹³⁹) and in ^{Rodríguez-Castellano and Palacio (1948a}, 1949b), vowel doubling appears in syllable final position preceding /s/ deletion. Some authors, the first being ^{Alarcos Llorach (1949}), have focused on this limitation of context to refute vowel doubling. ^{Alarcos Llorach (1949)} argued for the importance of considering not only the phonemic inventory of a language, but also the context where each phoneme can appear; however, this would not be until ^{Alarcos Llorach (1958)} when he would openly oppose vowel doubling in EAS. ^{Mondéjar Cumpián (1979}) supported ^{Alarcos Llorach's (1958)} position that five or six levels of openness are too many for a vowel system. These authors claimed that, as vowel doubling is only relevant in certain positions and /a/ opposes open /a/ but not open /e/, then we have two vowel systems, not one. Furthermore, ^{Mondéjar Cumpián (1979)} argued that there are two vowel systems as well due to the fact that only open vowels can be long.

^{Alarcos Llorach (1981}, p. 122) explained that the evolution of a system means that some distinctions are lost, but the system comes up with other ways of solving ambiguity, so he did not reject vowel doubling because he did not think that EAS did not mark deleted /s/, but because he thought that EAS marked deleted /s/ differently. ^{Alarcos Lorach (1981}, p. 123) explained how systems follow the principle of economy and tend to differentiate the highest possible number of phonemes with the minimum number of distinctive features, avoiding isolated oppositions; thus, he deemed a vowel system with 5 degrees of openness as too complex. Furthermore, ^{Alarcos Llorach (1981}, p.127) explained that phonological mutation rarely happens in isolation, and that it normally affects other parts of the system as well, so he had these principles in mind when working on ^{Alarcos Llorach's (1958)} theory. However, I believe that if EAS vowel opening is a relatively new phenomenon, as posited in ^{Rodríguez-Castellano and Palacio (1948a}), then maybe there has not been enough time for EAS to simplify the system and reduce its phoneme inventory.

The dephonologisation of vowel quality in Latin created a system with five degrees of height and a very small security margin between them (^{Alarcos Llorach, 1981}, p. 217). Differences in quantity turned into differences of quality, but it was difficult to maintain that system and the vowel system of Western vulgar Latin was reduced to four degrees of openness and seven vowels. This might be what made Alarcos think that EAS vowel doubling was too complicated to be a durable system; however, I believe that it is worth noting that EAS is an innovative geolect and its vowel system might not have time to readjust yet.

^{Alarcos Llorach (1958}) and ^{Mondéjar Cumpián (1979}) posited that what differentiates rosca 'doughnut' and roca 'rock' is not the open or closed vowel, but the alteration of the following consonant or ENT#091;hENT#093;, thus, ENT#091;'rohka/ 'rokkaENT#093; vs. ENT#091;'rokaENT#093;. ^{Alarcos Llorach (1958)} asserted that in ENT#091;a+xENT#093;, ENT#091;e+xENT#093;, etc. x can be quantity, openness (quality), tone, etc. therefore, he does not talk about 10 different vowels in EAS, but about 2 subsystems: system /a/, /e/, /i/, /o/, /u/ and system ENT#091;A*ENT#093;, ENT#091;E*ENT#093;, ENT#091;I*ENT#093;, ENT#091;O*ENT#093;, ENT#091;U*ENT#093; (short and narrow vs. long and wide vowels). Thus, ^{Alarcos Llorach (1958)} talked about system doubling, not vowel doubling. Furthermore, ^{Contreras Jurado (1975}) considers the vowel system of EAS the same as Castilian Spanish, and ^{Cerdà Massó (1992}) also defended the possibility of having 10 vowels split into two or more phonological vowel systems in EAS.

^{Contreras Jurado (1975}) also observed that in words like masa ENT#091;'masaENT#093; 'mass' and masas ENT#091;'ma̞sa̞ENT#093; 'masses', more than one sound is substituted. Thus, ^{Contreras Jurado (1975)} claimed that EAS speakers need the context to distinguish between singular and plural, and that they cannot do this just by opening one vowel. All vowels in a word, or at least the stressed one and the final one, need to be substituted. ^{Herrero de Haro (2016a}), on the other hand, claimed that EAS speakers can identify underlying /s/ without the help of contextual elements or vowel harmony, which he supported with the results of two perception experiments.

^{Contreras Jurado (1975}) defended his position adding that, if two sounds of the same language appear exactly in the same phonic context and can be substituted for each other without changing the meaning of the word, these two sounds are allophones of one phoneme; thus, pluralised vowel opening must not be understood as a vowel feature. According to ^{Contreras Jurado (1975)}, vowel opening in EAS can change the meaning of words without changing the nature of the phonemes, which is why vowel opening belongs to the same group as stress and tone, as its presence or absence does not alter the phonological entity of a vowel. ^{Contreras Jurado (1975)} also posited that vowel opening is a prosodic feature and affects the word globally, a position also defended by ^{Cerdà Massó (1992}). ^{Contreras Jurado (1975)} argued that the singular/plural distinction does not lie, in the current state of EAS, in the non-open vowel phoneme/open vowel phoneme opposition, but in the opposition word not affected by the prosodeme of openness/affected word. For him, the opposition distinguishing singular and plural in EAS was: - openness prosodeme/ + openness prosodeme.

^{Mondéjar Cumpián (1979}) and ^{Alarcos Llorach (1983}) considered that not all EAS vowels belong to the same vowel system, as they do not have a distinguishing value in all the positions where vowels can appear in Spanish. However, ^{Mondéjar Cumpián (1979)} thought that ENT#091;hENT#093;, which is the last trace of the consonant ENT#091;-sENT#093;, has distinctive value in all positions, explaining how ENT#091;hENT#093; is more tense in onset and not as weak as in coda. Despite not supporting vowel doubling, ^{Cerdà Massó (1992}) rejected the limitation of the phonological value of plural EAS vowels to final position only, as defended by ^{Mondéjar Cumpián (1979)} and ^{Alarcos Llorach (1983)}. ^{Cerdà Massó (1992)} believed that in a word like horrorosos 'horrible plu.', EAS speakers do not need to hear the last syllable to know it is plural. He believed that, due to vowel harmony, the first /o/ shows what is going to happen to the other vowels, and that the semantic load of /s/ has been moved to the first syllable, as it happened in French in cases like des maîtres 'teachers'. This was also supported by ^{Poplack (1980}) for Puerto Rican Spanish.

For ^{López Morales (1984}), vowel opening can only have phonological value if it appears as the only plural marker and happens only when we have /s/ deletion; however, he also attested vowel opening before deleted /r/ and /l/. Furthermore, ^{López Morales (1984)} studied vowel doubling in the southeast of the province of Granada and concluded that vowel opening in EAS is a phonetic feature, not a phonological one, as there is almost always another element that carries the semantic load of the deleted /s/. ^{López Morales (1984)} found numerals, articles, and subject pronouns in 97% of the analysed cases, meaning that vowel opening is a redundant marker of plurality or the subject tú. He also found that /s/ is deleted more often when there are other ways of marking deleted /s/; ^{Terrell (1975}) reached similar conclusions for Cuban speakers. ^{Herrero de Haro (2016a}), however, rejects ^{López Morales' (1984)} theory and argues that EAS speakers can identify underlying /s/ without contextual elements.

In an acoustic study, ^{Martínez Melgar (1986}) found that F1 and F2 movements from singular to plural are simply a tendency. Some movement happens in most cases, but not in all of them, which is why she considered EAS vowels as phonetic but not phonological variants. ^{Martínez Melgar (1986)} did not find vowel splitting in EAS vowels, not even a system of allophones; however, given the existence of tendencies in the movements of vowel formants, she could not be sure that vowel doubling had not existed before. ^{Martínez Melgar (1986)} wondered whether EAS vowels show an evolutionary process that tends towards lack of differentiation of plurals, as it happens, apparently, in western Andalusia. Many linguists, such as ^{Ueda (1993}) and ^{Torreira (2007}), divide Andalusia into an eastern part (where /s/ loss is marked by vowel opening), and a western part (where /s/ loss is marked by ENT#091;hENT#093;). According to ^{Morris (2000}), different outputs for /s/ debuccalisation in different geolects can be explained by suppression or redistribution of different markedness and faithfulness constraints.

For ^{Alarcos Llorach (1983}), EAS vowels are either projected or non-projected. Projected vowels are more advanced in the mouth and pronounced with the lips more retracted and with special tension of the tongue. ^{Alarcos Llorach (1983)} believed that that is what differentiates singular and plural EAS vowels, that vowel opening is redundant, and that plural EAS vowels are not phonemes. He also thought that this projected feature of the vowels is what made ^{Navarro Tomás (1938}) consider plural /a/ as velar.

^{Corbin (2006}) did not reject vowel doubling explicitly; however, ^{Corbin (2006}, p. 27) argued that "there was no significant difference between the formants of vowels in syllables whose codas had been deleted and those in syllables whose codas had not been deleted". Likewise, ^{Carlson (2012}) found no consistent qualitative vowel alteration following word-final /s/ deletion in speakers of Andalusian Spanish, although ^{Corbin (2006}, p. 27) attested vowel lengthening: "It appears that the crucial factor in determining whether a vowel laxes is whether or not the coda position of the syllable in the underlying structure is filled."

4.1.3. EAS Vowel harmony.

^{Alonso et al. (1950}), the same as ^{Navarro Tomás (1938}), identified vowel harmony as a feature of EAS, although they suggested it worked differently. For ^{Alonso et al. (1950)}, the closing or opening of a vowel only extends across a word to its vowel counterparts (e.g. the first /a/ would open in patas 'legs' but not in patos 'ducks'). ^{Alonso et al. (1950)} also believed that initial unstressed /a/, /e/, and /o/ follow the same changes as stressed vowels and they also argued that stressed vowels are longer in EAS than in Castilian Spanish; ^{Alvar (1955b}) also identified vowel harmony, which also happens in Germanic languages. On the other hand, ^{Alarcos Llorach (1958}) argued that some vowel opening is due to assimilation, rather than vowel harmony, and considered word-internal vowel opening a non-pertinent variation. Similarly, ^{Contreras Jurado (1975}) explained that in plural words, final and stressed vowels (and in many cases all vowels), are more open than in Castilian Spanish. ^{Contreras Jurado (1975)} supported the argument that vowel opening in plurals is used by EAS as a plural marker and that this opening is slightly bigger than the one produced by coarticulation. Lastly, ^{Gómez Asensio (1977}) and ^{Mondéjar Cumpián (1979}) also believed that vowel opening is the distinctive feature for plurals; as ^{Gómez Asensio (1977)} pointed out, vowel harmony transfers openness to other vowels, not other features like palatalisation.

^{Zubizarreta (1979}) studied vowel harmony as a set of rules ranked by order of appearance, following a similar process to that used by ^{Gómez Asensio (1977}). In her study, ^{Zubizarreta (1979)} analysed vowel harmony in Andalusian vowels, describing which vowels undergo phonetic and phonemic laxing, which vowels open as part of vowel harmony, and which vowels let vowel harmony influence other vowels in a word. ^{Jiménez and Lloret (2007}) and ^{Lloret and Jiménez (2009)} also focused on vowel harmony and they considered it a result of languages attempting to minimise the resetting of articulators (a type of assimilation). However, this would only explain the cases where all vowels harmonise, not those where not all vowels lax. ^{Kaplan (2012}) approached this issue as well, and concluded that multiple-ranking analysis explains EAS vowel harmony better than other theories, as it is capable of generating the all-or-nothing patterns of post-tonic EAS vowels, something that other theories fail to do. However, unlike ^{Kaplan (2012)}, ^{Lloret and Jiménez (2009)} believed that there is a similar harmonisation process in words ending in /x/ (ENT#091;hENT#093; in Andalusian Spanish).

Optimality Theory has proved useful to tackle the analysis of EAS vowel harmony, especially in the last decade. ^{Jiménez and Lloret (2007}) and ^{Lloret and Jiménez (2009)} explained vowel harmony through the concept of "positional (perceptual) markedness," according to which the harmonising feature is attracted to strong positions in the word, because it becomes more perceptible that way. However, it was not until ^{Lloret and Jiménez (2009)} when they defined the harmonising feature as "-ATR." ^{Kaplan (2012}) presented a similar explanation for vowel harmony in EAS, although he preferred the term "Generalized Licensing" (from ^{Walker, 2011}); he adds that the perception of the feature is also increased by repeating the feature as many times as possible. Furthermore, ^{Jiménez and Lloret (2007)} believed that the vowels before the stress can either open or not, which I personally believe is governed by a series of sociolinguistic rules.

4.1.4. Vowel lengthening.

As with other features related to EAS vowels, quantity has also been widely debated. ^{Rodríguez-Castellano and Palacio (1948a}) and ^{Alonso et al. (1950}) considered /a/ longer in plural than in singular. Likewise, ^{Alonso et al. (1950)} also reported that stressed vowels are always lengthened and longer than their Castilian counterparts. For ^{Salvador (1957}), vowels are longer in plurals and stressed syllables, and /a/, /e/ and /o/ preceding /s/ deletion are more open in stressed than in unstressed syllables. ^{Alarcos Llorach (1958}) claimed that a vowel preceding an omitted /s/ is not only more open, but also longer, and he explained this by saying that the gap created by the omission of /s/ is filled by lengthening the vowel or by geminating the following consonant.

On the other hand, ^{Salvador (1977}) pointed out that lengthening is neither regular nor exclusive to open vowels; however, ^{Martínez Melgar (1994}) claimed that all vowels are shorter before deleted /s/, which contradict what ^{Alarcos Llorach (1958}) posited; for ^{Sanders (1998}), however, only high vowels are shorter in plural than in singular.

^{Gerfen (2002}) studied vowel and consonant lengthening, and vowel aspiration, and he concluded that there is a relationship between vowel and consonant lengthening: if the ratio of consonant lengthening grows following /s/ aspiration, the degree of vowel lengthening decreases, and vice versa. For ^{Gerfen (2002)}, gemination of the following onset consonant is a result of word-internal /s/ aspiration, not a consequence of positioning in the word; nevertheless, I believe that gemination also happens as a result of deleting other consonants, as in apto ENT#091;'ahttoENT#093; 'capable'. Furthermore, ^{Gerfen (2002)} claimed that the duration of aspiration tends to become shorter as the voiced portion of the vowel becomes longer. ^{Peñalver Castillo (2006}) reported something similar in Cabra, claiming that vowels are shorter when /s/ is aspirated than when it is not.

In his study, ^{Gerfen (2002}, p. 265) analysed whether consonant and vowel lengthening are correlated, concluding that "the variability in vowel and consonant lengthening is constrained within a less variable vowel + stop temporal target." Furthermore, ^{Gerfen (2002)} posited that vowels are significantly longer when /s/ is aspirated only if we consider the period of aspiration after the vowel to be part of the vowel gesture. ^Gerfen concluded that consonant lengthening is a more robust cue than vowel lengthening in marking a missing /s/ in coda, being the first author to posit that:

the degree of variation between the phenomena of consonant and vowel lengthening is tightly constrained. Specifically, although there is variation in the degree of gemination and lengthening under s-aspiration both within and across speakers, the results reveal a trade-off relation between the two phenomena. (²⁰⁰², p. 275)

^{Campos-Astorkiza (2003}), however, found consonant gemination in EAS rather than vowel lengthening.

In her analysis of phonemicisation of vowel duration word-finally, ^{Carlson (2012}) found an average increase of 24.2% in the duration of vowels preceding deleted /s/. Her sample, however, included 4 EAS and 2 WAS native speakers, so it is not possible to have a specific percentage for EAS alone. According to ^{Carlson (2012)}, this vowel lengthening allowed participants to distinguish pairs such as buque 'ship' and busque 'I/he/she looks for -subj-' with a success rate of 79%. These results are similar to those reported in ^{Hammond (1978}) for Miami-Cuban Spanish and in ^{Figueroa (2000}) for Puerto Rican Spanish, but they contradict ^{Alemán (1977}), who does not report any compensatory lengthening in vowels preceding deleted /s/. Likewise, ^{Carlson's (2012)} results oppose those presented by ^{Gerfen and Hall (2001}), ^{Gerfen (2002)}, ^{Bishop (2007}), and ^{O'Neill (2010}), who believe that the cue to recognise word-medial /s/ deletion lies in the gemination of the consonants.

It is interesting to note ^{Hammond's (1978}) findings, which claimed that in word-internal vowels, length, not vowel opening, provides Miami-Cuban Spanish speakers with the cues to distinguish between minimal pairs once /s/ has been deleted, an opinion backed by ^{Carlson (2012}). Following the Distinctiveness Condition proposed by ^{Kiparsky (1982}), ^{Carlson (2012)} argues that semantically relevant information tends to be retained, so there must be a mechanism to maintain the information lost in EAS when /s/ is deleted.

4.1.5. Velarisation or palatalisation of /a/

Another widely debated feature of EAS is the quality of /a/. ^{Rodríguez-Castellano and Palacio (1948a}) and ^{Navarro Tomás (1938}, ¹⁹³⁹) asserted that when /s/, /θ/, or /ð/ are deleted, /a/ becomes slightly velarised. Likewise, ^{Salvador (1957}) claimed that /a/ is velarised in Cullar-Baza (now Cullar) when it precedes aspirated /s/ plus a velar consonant. He also claims that /a/ can have a slight velar timbre when deleted /s/ is followed by another consonant.

Regarding /a/ preceding /s/ deletion, ^{Navarro Tomás (1938}, ¹⁹³⁹), ^{Rodríguez-Castellano and Palacio (1948a}), ^{Martínez Melgar (1986}), ^{Ueda (1993}), and ^{Sanders (1998}) have maintained that it is velar, as opposed to the descriptions presented in ^{Wulff (1889}), ^{Alonso et al. (1950}), ^{Alvar (1955b}), ^{Alonso (1956)}, ^{Salvador (1957}, ¹⁹⁷⁷), ^{Alarcos Llorach (1958}), ^{Zamora Vicente (1960}), ^{Llorente (1962)}, ^{Jiménez and Lloret (2007}), and ^{Lloret and Jiménez (2009)}, who considered it a palatal vowel. However, ^{Alonso et al. (1950)} accepted that open /a/ can have a velar aspect before a velar consonant. Alonso et al. (1950) posited that this /a/ is fronted, but still more back than /e/. Interestingly, ^{Jiménez and Lloret (2007)} and ^{Lloret and Jiménez (2009)} believed that /a/ fronting preceding deleted /s/ acts as a cue to preserve the place of articulation of EAS coronal /s/ but that this feature only transfers to /a/, as /e/ and /i/ are already front vowels and /o/ and /u/ do not adopt this feature. However, unlike ^{Jiménez and Lloret (2007)} and ^{Lloret and Jiménez (2009)}, ^{Herrero de Haro (2016b}) believes that all consonants leave a trace of place of articulation on any preceding vowel.

^{Alvar (1973}, map 1697) studied the palatalised solutions for unstressed /a/ preceding deleted word-final /s/ (e.g. casas 'houses'). According to this map, /a/ can be palatalised to different degrees (e.g. casas ENT#091;ˈka̞sa̞jENT#093;/ ENT#091;ˈka̞se̞ENT#093; 'houses') and it can be slightly velarised as well (e.g. casas ENT#091;ˈka̞sa̞ɣENT#093;), but the map shows that a palatal solution is more common than a velar one. ^{Martínez Melgar (1994}) reported that /a/ tended to be velar in the province of Granada but alveolar in Jaen. She even described some alternation between velar and palatal /a/ within the same speaker. Considering all this, the different velar and alveolar pronunciations of /a/ reported by various linguists might have differed due to the location studied in each case. ^{Gómez Asensio (1977}), however, went a bit further and believed that EAS /a/ can be neuter, velar, or palatal.

In a more global sense, ^{Salvador (1977}) considered all Andalusian vowels more fronted than their Castilian counterparts; vowels are more fronted when they precede /s/ deletion but plural /e/ will always be more back than /i/.

4.1.6. EAS Vowels: other phenomena

^{Alonso (1956}) was the first researcher to describe an interesting phenomenon regarding /a/ . He found that stressed and unstressed word-final /a/ preceding deleted /l/, /r/, /s/, or /Ɵ/ was pronounced as an open /e/ and that this opening was more common in women and children, an opinion also supported by ^{García Mouton (1992}). Thus, ^{Alonso (1956)} found an opposition vaya ENT#091;ˈbaʝaENT#093; 'he/she goes 'subj.-' vs. vayas ENT#091;ˈbaʝe̞ENT#093; 'you sing. go 'subj.-' (also reported vayáis ENT#091;ba'ʝeiENT#093; 'you plu. go 'subj.-'). He located the phenomenon in a triangle between Genil (Cordoba), Estepa (Sevilla), and Alameda (Malaga) and he was surprised to see that it was not present in some nearby towns. According to ^{Alonso (1956)}, that phenomenon was in decline and it had covered a much larger area in the past; this was later confirmed by ^{García Marcos (1987}), who found traces of it in older speakers and children on the coast of the province of Granada (approximately 150km outside of the area identified by ^{Alonso, 1956}).

^{Rodríguez-Castellano and Palacio (1948a}) and ^{Salvador (1957}) describe how /je/ can be reduced to /e/ or /i/ in Eastern Andalusia. ^{Rodríguez-Castellano and Palacio (1948a)} and ^{Salvador (1957)} also found that other diphthongs which had been broken in Castilian Spanish were still present in Cabra and Cullar-Baza, and that pre-tonic /a/ closes into ENT#091;eENT#093; and /e/ closes into ENT#091;iENT#093;; these two articles also show examples where pre-tonic /o/ closes into ENT#091;uENT#093;, where /u/ opens into ENT#091;oENT#093;, and where /i/ opens into ENT#091;eENT#093;. ^{Rodríguez-Castellano and Palacio (1948a)} and ^{Salvador (1957)} claimed that two similar vowels are merged into one (as in Granada ENT#091;graˈnaENT#093;), although ^{Rodríguez-Castellano and Palacio (1948a)} argue that that merged vowel in a stressed position is longer than the final /a/ in mama 'mum'.

On the other hand, ^{Rodríguez-Castellano and Palacio (1948a}) and ^{Salvador (1957}) also argued that two adjacent vowels within a word tend to keep their value if one of them is stressed (e.g. cadera ENT#091;kaˈeɾaENT#093; 'hip'). ^{Salvador (1957)} had a comprehensive list of solutions for adjacent vowels in word boundary positions and he also described a tendency to turn diphthongs into hiatuses in his town.

According to ^{Rodríguez-Castellano and Palacio (1948a}), there are different options when the vowels belong to different words, such as diphthongs (ni una ENT#091;ˈnjunaENT#093; 'not even one') and hiatuses (cada uno ENT#091;kaˈunENT#093; 'each one'). They might also merge into one vowel, especially if the first vowel is unstressed (te ha engañado ENT#091;taŋgaˈɲaoENT#093; 'he/she has deceived you sing.'; ese imbécil ENT#091;esimˈbesilENT#093; 'that stupid person'). ^{Rodríguez-Castellano and Palacio (1948b)} also found metathesis (e.g. nadie ENT#091;ˈnaiðeENT#093; instead of ENT#091;'naðjeENT#093; 'nobody').

Although less extensively, nasality has also been considered in the study of EAS vowels. ^{Rodríguez-Castellano and Palacio (1948a}) briefly mentioned nasality in EAS, although it was ^{Alonso et al. (1950}) who first reported phonologisation of vowel nasality in Spanish; they attested word-final /n/ deletion in Albaicin (a neighbourhood in Granada) and posited that the nasality of vowels preceding word-final /n/ deletion distinguished viene 'he/she comes' and vienen 'they come'. According to ^{Lipski (1986}), word-final /n/ is never pronounced ENT#091;nENT#093; in Granada, it is pronounced ENT#091;ŋENT#093; in 77% of cases and as a nasalised vowel in the remaining 23% of cases. ^{Lipski (1986)} claims that in word-final position followed by a word beginning with a vowel, /n/ is pronounced ENT#091;nENT#093; 48% of the time in Granada, ENT#091;ŋENT#093; in 35% of cases, and as a nasalised vowel in 17% of cases; this phenomenon is also reported in ^{Zamora Vicente (1960}, p. 324). This feature demonstrates further similarities between EAS and French, with authors such as ^{Llorente (1962)} claiming that French and Andalusian Spanish are the most phonetically developed Romance varieties. According to ^{Alonso et al. (1950)}, the speech of Granada is also more nasal than Castilian Spanish and even /s/ aspiration in Granada has a nasal timbre. For ^{Zamora Vicente (1960}, p. 323), however, all aspirations in Granada (and some in Cordoba, Jaen, and Malaga) have a clear nasal timbre.

To conclude this section on vowels, it warrants pointing out to a different approach to analysing vowel contrast. ^{Alarcos Llorach (1949}) proposed basing the description of phonemes not on what they are, but on what they are not (e.g. /p/ is phonemically described against /b/ as not voiced. ^{Alarcos Llorach}:

The concept of distinction or difference assumes the concept of contrast, opposition. We cannot distinguish between two things if there is no difference between them ... A phonic characteristic will have distinctive function when it is opposed to another one, this is, when both form a phonic opposition. ... Two distinctive features can be opposed to another two, meaning that one of the oppositions is redundant. However, if in some circumstances the distinctive opposition disappears, then the other one becomes relevant (¹⁹⁸¹, p. 39) (my translation).

This idea would later be followed by ^{Martínez Melgar (1986}), who described EAS vowels as open vs. non-open, as opposed to open vs. closed. Regarding vowels, ^{Alarcos Llorach (1949}) explained the opposition of Spanish vowels as: 1) gradual and isolated opposition (a/e, a/o, a/i and a/u); 2) proportional and equipollent oppositions (e/o and i/u); and 3) proportional and gradual opposition (e/u, o/i, e/i, o/u). Regardless of which position one supports in the debate about EAS, what has become clear is that much work is still needed to fully understand this variety of Spanish.

4.2.1. Aspirations in EAS

Von ^{Wartburg (1952}, as cited in ^{Alvar 1955b}), "the most characteristic and important phonetic change and the one which had more consequences amongst all the ones that happened in the Roman Empire was the one related to final /s/ ENT#091;deletionENT#093;" (p. 287) (my translation) . ^{Alvar (1955b)} also argued that the conservation or loss of that /-s/ is what caused the linguistic difference between the Eastern and Western Roman Empires, so it does not seem exaggerated to attribute so many vowel changes to /s/ deletion.

For ^{Rodríguez-Castellano and Palacio (1948a}), aspiration is one of the main characteristics of EAS. This is used instead of initial Latin f or of /s/, /θ/, and /x/ in coda and in intervocalic position; ENT#091;hENT#093; can also be the realisation of other consonants in coda and various consonants (e.g. /f/) can be aspirated word-initially.

However, ^{Peñalver Castillo (2006}) studied the speech of Cabra and claimed that aspiration of initial Latin f is almost lost. ^{Jiménez Fernández (1999}, p. 45) supported Menéndez Pidal's theory that aspiration of initial Latin f followed the sequence /f/( /h/ ( zero, and that it started in the Basque Country and Gascon due to a Basque-Iberian substrate, as Basque did not have /f/. ^{Zamora Vicente (1960}, p. 55) supported this evolution sequence with a Castilian document from 863 which shows that initial Latin f was already aspirated at that time. ^{Rodríguez-Castellano (1952}) asserted that words like hacha 'axe' and hambre 'hunger' maintain the feminine article in Cullar-Baza, as the initial h is still pronounced and we do not have the avoided ENT#091;a ˈaENT#093; sequence.

According to ^{Mondéjar Cumpián (1979}), the maintenance of word-initial /h/ is the only case where Andalusian is less phonetically evolutioned than Castilian Spanish and it is the only archaic phonetic feature of Andalusian Spanish. ^{Jiménez Fernández (1999}, p. 46) explained how after ENT#091;hENT#093; was common in the North, it became established in the South, where it remained despite being lost in northern Spain. ^{Jiménez Fernández (1999}, p. 47) then described how it appeared in areas of Latin America (it was taken there in the sixteenth century), and posits that the lack of ENT#091;hENT#093; for initial Latin f in Eastern Andalusia could be due either to the fact that the area was repopulated by people from areas where ENT#091;hENT#093; had been lost or because the area is further away from Seville, and from its norm. As ^{Jiménez Fernández (1999}, p. 48) explained, ENT#091;hENT#093; is used at the beginning of words which come from Latin, such as farina 'flour'; however the /f/ was maintained in words introduced later (e.g. fuente 'fountain' or fuerte 'strong'), and it is normally followed by /we/ or /je/.

Regarding /s/ aspiration, it is worth noting that ^{Alvar (1955b}) has explained how the Andalusian tendency to drop /s/ in coda was already a feature in Latin in Malaga and Seville in the second century. In the seventh century, this was documented and considered something vulgar, which suggests that syncope/apocope of /s/ could be much older than initially thought.

^{Rodríguez-Castellano and Palacio (1948a}) described EAS aspiration as a laryngeal or laryngo-pharyngeal sound; nevertheless, they accepted the terms bilabial, dental or alveolar aspiration to show that the aspiration is assimilated to the bilabial, dental, or alveolar character of the following consonant. They believe, however, that the small differences perceived are due to the intensity of the friction, not to the place of articulation. ^{Salvador (1957}), ^{Tejada Giráldez (2012}), ^{Rodríguez-Castellano and Palacio (1948a}, ^1948b), ^{Zamora Vicente (1960}, p. 319) and ^{Zubizarreta (1979}) also supported that /s/ aspiration assimilates the place of articulation of the following consonant.

This type of assimilation is not rare. Word-medial /s/ assimilates to the next consonant in Old Latin and Pali, while in Ancient Greek, Medieval French and Norman French /s/ was assimilated to the following consonant and then it was lost and the previous vowel lengthened (^{Alvar, 1955b}). This sequence of events seems to be in operation in EAS as well.

^{Rodríguez-Castellano and Palacio (1948a}) believed that the aspiration, at least at some point, was voiceless, as there is devoicing of consonants preceded by aspirated /s/ (e.g. dos dientes ENT#091;do̞ƟˈƟje̞nte̞ENT#093; 'two teeth'). The aspiration can also become nasalised, and it is stronger the lower the level of instruction the speaker has and in more informal registers (^{Salvador, 1957}). ^{Alvar (1955b}) believed that the pronunciation of /s+bilabial/, as /f/ in Valais (Switzerland), was another example of the same phenomenon. Rodríguez-Castellano and Palacio (1948b) also analysed the different results of /s/ aspiration before consonants (e.g. desván ENT#091;deɸ'ɸanENT#093; 'attic'; obispo ENT#091;oˈβihppoENT#093; 'bishop'; usted ENT#091;uhˈteENT#093; 'you sing. formal'; muslo ENT#091;mulˈloENT#093; 'thigh'; desgranar ENT#091;(d)ehxgraˈnaENT#093; 'to dekernel'; and escoba /ehˈkoβa/ 'broom'), pointing out that something similar happened in French when /s/ was lost before /t/ in the twelfth century, as in tête 'head'. Alvar (1955b) asserted that /s/ is maintained in Provenzal, although some varieties drop it in final position. Those different results of /s/ aspiration fall within the different solutions that ^{Mondéjar Cumpián (1979}) identified for /s+Consonant/: ENT#091;hCENT#093;, ENT#091;hCCENT#093;, ENT#091;CCENT#093;, and ENT#091;CENT#093;.

^{Salvador (1957}) described some assimilations in detail: in /s+voiceless stop/, the aspiration assimilates part of the place of articulation of the voiceless stop; /sg/ becomes ENT#091;xENT#093; (e.g. es gordo ENT#091;eˈx:orðoENT#093; 'he is fat') ; /sb/ becomes ENT#091;fENT#093; (e.ɡ. las botas ENT#091;la̞ˈfːo̞ta̞ENT#093; 'the boots'); and /sd/ becomes ENT#091;θENT#093; (e.g. los dedos ENT#091;lo̞ˈθːe̞ðo̞ENT#093; 'the fingers'), although other intermediate solutions are also possible. Furthermore, ^{Alvar (1955b}) explained how in /s+voiceless fricative/, /s/ assimilates the place of articulation of the fricative; in /s+nasal/, /s/ is nasalised and assimilated to the place of articulation of the nasal; and, finally, /s+l/ tends to result in ENT#091;llENT#093; (ENT#091;l:ENT#093;). ^{Alarcos Llorach (1958}) did not provide specific examples of these alterations, but he represented them in a scheme that explained that /s/ deletion can open and lengthen the preceding vowel, and it can also geminate and modify the following consonant.

More recently, ^{Torreira (2007}) also studied /s+voiceless stop/ and found out that in Andalusian Spanish, voiceless stops are consistently post-aspirated when they are preceded by aspirated /s/. He argued that "postaspiration should be considered as a potential perceptual cue of /s/ aspiration," (p. 70) and believed that postaspiration could possibly lead to a new category of voiceless stops in Andalusian, as it occurred in Pali. Torreira (2007) reported post-aspiration in Eastern and western Andalusian Spanish, although it is stronger in the latter. Finally, although he could only demonstrate this for /st/, he found evidence of a similar situation for /sk/ and /sp/. ^{Ruch and Harrington (2014}) also studied post-aspiration and concluded that it serves as a cue to identify underlying /st/. In EAS, younger speakers have shorter pre-aspirations and longer post-aspirations than older speakers, but these differences are not as noticeable as the ones found in western Andalusia. Furthermore, ^{Ruch and Harrington (2014)} reported a trading relationship between oral closure duration and post-aspiration, which they believe indicates a sound change.

For ^{Rodríguez-Castellano and Palacio (1948b}), coronal /s/ was always pronounced except in sporadic cases, such as in word-internal codas preceding a consonant and in word-final position, as in those cases they noted an aspiration, usually voiceless.

In word boundaries, ^{Salvador (1957}) found the following solutions: aspirated /s/ does not appear before words containing /x/; /s+voiceless stop/ tends to become an aspiration brought towards the place of articulation of the stop, and the aspiration can be lost and /s/ marked by vowel opening and tensing of the following consonant, as it happens in /s+fricative/ or in /s+affricate/; /sb/, /sd/ and /sg/ behave as in word-internal position; /s+nasal/ results in total loss of aspiration before /m/ and a very nasalised short aspiration before /n/; and /sl/ results in ENT#091;llENT#093; (ENT#091;l:ENT#093;). ^{Salvador (1957)} believed that /s/ was aspirated in word-internal coda and either aspirated or deleted in word-final position. Some years later, ^{Lipski (1986}) studied the pronunciation of /s/ in various contexts in Granada and found that /s/ is hardly ever pronounced ENT#091;sENT#093; in coda, but is often pronounced ENT#091;hENT#093; or deleted. ^{Alvar (1955a}) had also noticed how the aspiration of /s/ caused a whole range of possibilities in /s+consonant/, many of them similar to the ones found around Europe, such as the pronunciation of /st/ as ENT#091;htENT#093; in Fribourg and Vaud.

Likewise, ^{Alvar (1955b}) describes how in Latin, final /s/ used to be pronounced if the next word started with a vowel, as it happens in French now. In Bergamasco, /s/ tends to be aspirated when it becomes a final consonant due to the loss of a vowel, and in Brazilian Portuguese, /s/ is also lost in the final syllable of plural words. Moreover, aspiration of /s/ in articles also happens in Armenian. In some areas, Provenzal /s/ is substituted by aspiration before any consonant, as in Bergamasco ( e.g. vespa ENT#091;ˈvehpaENT#093; 'motor-scooter').

According to ^{Rodríguez-Castellano and Palacio (1948b}), the highest degree of assimilation is reached when /s/ precedes a labial, dental, or velar consonant, and the process seems to be more advanced if the following consonant is voiced. ^{Alvar (1955b}) added that within voiced consonants, the process must have affected labial and velar consonants first, then dentals. ^{Rodríguez-Castellano and Palacio (1948b)} believed that this was also the case in French, where /s/ was lost first before voiced than before voiceless consonants and where the vowel preceding deleted /s/ also opens and is longer. ^{Alvar (1955b)} also explained how in Latin, the omission of /s/ started to appear first before voiced consonants. For ^{Rodríguez-Castellano and Palacio (1948b)}, the most advanced version of the phenomenon can be heard when /s/ is followed by a voiceless velar (e.g. casco /ˈkakko/ 'helmet'). ^{Rodríguez-Castellano (1948b)} believed that with this tendency to omit consonants in coda, in a few years Andalusian Spanish would have lost all consonants in that position, although this has not been the case, yet. ^{Alvar (1955b)} also put this Andalusian phenomenon in connection with other languages: he said that aspiration of /s/ in word-final position is another feature of Sanskrit; /s/ is also omitted before a pause or vowel in Armenian; and in old Slavic /s/ is neutralised in polysyllables. Furthermore, insular Celtic lost that /s/ at the latest in the sixth century and Gallic was losing it when the Roman invasion took place.

The realisation of /s/ varies depending on sociolinguistic factors. ^{Salvador (1957}) found that /s/ deletion and vowel opening was only present in women under 25; in women between 25 and 30 years of age, /s/ appeared at times but the loss of /s/ was more common; in women between the ages of 30 and 40, loss and conservation of /s/ alternated and they appeared with equal frequency; /s/ was more common in women over 40; and loss of /s/ was sporadic in women over 60 years of age. These pronunciations back ^{Salvador's (1957)} theory which posits that vowel doubling was an emerging feature of speech in the area. ^{Alonso (1956}) argued that in the area later studied by ^{Salvador (1957)}, there is conflict between two systems, a new invading one, spoken by men and of an Andalusian type, and one being lost, of a Castilian type, preserved amongst women. This view of complete /s/ deletion as a spreading phenomenon was also backed by ^{Peñalver Castillo (2006}). ^{Salvador (1957)} found three older ladies who did not have yeísmo and maintained /s/ and ^{Peñalver Castillo (2006)} found traces of a maintained /s/ in Cabra in 2% of speakers. ^{Peñalver Castillo (2006)} claimed that speakers with a higher socio-cultural level preferred deleting /s/, while speakers from lower socio-cultural levels preferred aspiration. ^{Tejada Giráldez (2012}) also analysed linguistic factors in the pronunciation of /s/ in coda by speakers with a high socio-cultural background from the city of Granada. ^{Tejada Giráldez (2012)} reported that the most common realisation word-finally was the /s/ deletion preceded by vowel opening and she believed that coda /s/ pronunciation was determined by socio-linguistic forces.

^{Tejada Giráldez (2012}) concluded that speakers of high socio-cultural levels in the city of Granada present the following percentages for each realisation of word-final /s/: 1.4% for ENT#091;sENT#093;; 24.5% for ENT#091;hENT#093;; and 70.5% for /s/ deletion. ^{Tejada Giráldez (2012)} reported a tendency towards aspiration of /s/ word-internally and towards elision of /s/ word-finally. Furthermore, ^{Tejada Giráldez (2012)} argued that speakers of higher socio-cultural levels geminate consonants following /s/ deletion in 1.9% of cases and pronounce ENT#091;hENT#093; plus gemination 1.7% of the time. For ^{Tejada Giráldez (2012)}, social variables (e.g. age), have an effect on the preferred realisations in each case; maintenance of /s/ and aspiration increase with the age of the speaker; however, gemination and elision are lower the older the speaker is.

^{García Marcos (1987}) believed that the loss of /s/ was a change originated in low sociolects. ^{García Marcos (1987)} found that lengthened vowel followed by ENT#091;hENT#093; was more common in high sociolects, likewise, lower sociolects preferred an open and lengthened vowel. ^{García Marcos (1987)} and ^{García Mouton (1992}) also pointed out a tendency in men for more recent solutions: ENT#091;hENT#093;, ENT#091;sENT#093;, or ENT#091;hENT#093; following an open and lengthened vowel.

Another unsettled debate is the one regarding when EAS /s/ aspiration first emerged. ^{Menéndez Pidal (1940}) presented an example in which Christopher Columbus's son (raised in Cordoba, where EAS is spoken) had written Sophonisba as Sofonifa. For ^{Menéndez Pidal (1940)}, this is a graphic representation of the EAS pronunciation /sb/ ( ENT#091;fENT#093; and he considered this the oldest documentation of /s/ deletion in Spanish, dating EAS /s/ deletion to the beginning of the sixteenth century; ^{Alvar (1955b}) and ^{Zamora Vicente (1960}, p. 321) also supported this. However, ^{Mondéjar Cumpián (1979}) believed that /x/ aspiration started after the seventeenth century and that /s/ and /θ/ aspiration started in Andalusia after the eighteenth century, with the aspiration then spreading to the groups /sg/ and /θg/. For ^{Jiménez Fernández (1999}, p. 52), the aspiration of offset /s/ started in the eighteenth century, although he believed that the aspiration of Castilian /x/ started in the sixteenth century; ^{Jiménez Fernández (1999}, p. 53) pointed out how the pronunciation of /x/ as ENT#091;hENT#093; has extended more socially and geographically than initial Latin f. ^{Peñalver Castillo (2006}), for example, explained how /x/ tends to be an aspirated pharyngeal consonant in Cabra. ^{Jiménez Fernández (1999}, p. 52) explained the aspiration of Castilian /x/ in Andalusia as a result of merging /ʃ/ with the sound used for initial Latin f.

Finally, another phenomenon related to aspiration is heheo (pronouncing word-initial or intervocalic /s/ as ENT#091;hENT#093;). For ^{Rodríguez-Castellano and Palacio (1948b}), heheo only appears in intervocalic position, creating sí señor ENT#091;si heˈɲoENT#093; 'yes, sir' and el señor ENT#091;el seˈɲoENT#093; 'the gentleman'. This could have been the case then, but the situation is different now; I have heard multiple EAS speaker friends and relatives (and even myself on certain occasions), using heheo word-initially (e.g. sí ENT#091;hiENT#093; 'yes'). Interestingly, a similar phenomenon is reported in the Nice area, in Ancient Greek, and in Irish (^{Alvar, 1955b}).

4.2.2. Seseo and ceceo

^{Alonso et al. (1950}) reported seseo (pronouncing /Ɵ/ as ENT#091;sENT#093;), in Granada even in speakers from high socio-cultural backgrounds, and in Albaicin, men use ceceo (pronouncing /s/ as ENT#091;ƟENT#093;), and women seseo. ^{Peñalver Castillo (2006}) claimed that seseo is the norm in Cabra, although speakers of higher socio-cultural levels, especially under 50 years of age, tend to distinguish and distinción is more common in men. ^{Zamora Vicente (1960}, p. 304) explained how Arias Montano (b.1527) described how confusion of s, z and ç appeared in Seville between 1546 and 1566, although Zamora Vicente (1960, p. 305) believed that Montano was talking about a small minority and that the change had happened earlier in the lower classes. Zamora Vicente (1960, p. 305) posited that the lack of distinción first affected Seville and then the coastal south of Andalusia. ^{Zamora Vicente (1960}, p. 305) proved that in 1651 distinción was the old custom, and ceceo was the norm for western Andalusia. ^{Jiménez Fernández (1999}, p. 24) believed that seseo and ceceo appeared at the end of the sixteenth century.

There has also been some sociolinguistic variation reported regarding seseo and ceceo. ^{Garía Mouton (1992)} identified ceceo in men from low socio-economic backgrounds in Malaga, but seseo in women and men from high socio-cultural backgrounds. ^{García Mouton (1992}) also noticed further differences in an area of Jaen, where people have ceceo in the countryside and seseo in the city but women and older men present distinción. However, to my knowledge, a study on the spread of seseo and ceceo in Eastern Andalusian has not been carried out since ^{Alvar (1973}).

4.2.3. Confusion of /r/ and /l/

For ^{Rodríguez-Castellano and Palacio (1948a}), word-initial EAS /l/ and /r/ are similar to their Castilian counterparts but with less muscular tension, and in word-final coda they are very relaxed and weak; they hardly touch the palate and are more like fricatives.

^{Alonso et al. (1950}) claimed that alternation of /l/ and /r/ varied significantly, although it was more constant in word-final position. For ^{Salvador (1957}), /r/ and /l/ in word-final position is pronounced as a relaxed /l/, as in mujer ENT#091;muˈxe̞lENT#093; 'woman'), and la mujer es ENT#091;la muˈxe̞l e̞:ENT#093; 'the woman is', but mujeres ENT#091;mu̞ˈxe̞re̞:ENT#093; 'women'. However, for ^{Rodríguez-Castellano and Palacio (1948b}), /l/ is pronounced ENT#091;rENT#093; in word-final posision is pronounced. ^{Salvador (1957)} also noticed that /l/ and /r/ before /t/ or /d/ tend to be ENT#091;ɹENT#093;, but they can also be ENT#091;lɹENT#093; and ENT#091;lENT#093;. He believed that ENT#091;ɹENT#093; is more common before /θ/, and that ENT#091;lENT#093; is preferred before /s/, but that /rn/ is never substituted by ENT#091;lnENT#093;. ^{Rodríguez-Castellano (1952)} also encountered this phenomenon in Cabra and claims that in el 'the sing. masc.', al 'to the sing. masc.', and del 'from the sing. masc.' /l/ is pronounced ENT#091;rENT#093;. For ^{Zamora Vicente (1960}, p. 314), neutralisation of /r/ and /l/ happens in coda but not in onset; /l/ is not altered when it is followed by a vowel, farol ENT#091;faˈɾorENT#093; 'lantern', but faroles ENT#091;faˈɾolesENT#093; 'lanterns'. ^{Zamora Vicente (1960}, p. 316) and ^{Jiménez Fernández (1999}, p. 63) presented examples from this confusion registered from the twelfth century, such as the surname Álvarez written Árbarez in Toledo in 1161.

^{Navarro (1938}, ¹⁹³⁹) identified /l/ and /r/ deletion in word-final position; nevertheless, ^{Salvador (1957}) defended the notion that these were still not deleted in Cullar-Baza, but pronounced very weakly. ^{Alvar (1973}, map 1699) studied the pronunciation of stressed /a/ preceding deleted word-final /l/ or /r/, and the map showed that the only areas in Andalusia where those two consonants were not deleted were the border of Granada with Almeria, Northeastern Jaen, the border of Almeria and Granada with Murcia, and in a few towns in northern Cordoba; in other various locations, those two consonants are not deleted very frequently.

^{Rodríguez-Castellano (1952}) explained how final /r/ in infinitive is assimilated to the initial /l/ of enclitic pronouns. There are three variations of this: /r+l/ ( ENT#091;ʝENT#093; (e.g. comerlo ENT#091;koˈmeʝoENT#093; 'to eat it'), found amongst older people in Doña Mencía; /r+l/ ( ENT#091;hllENT#093; (e.g. comprarlo ENT#091;komˈprahlloENT#093; 'to buy it'); and /r+l/( ENT#091;lENT#093; (e.g. hacerlo ENT#091;haˈseloENT#093; 'to do it'), pronunciation also recorded in ^{Jiménez Fernández (1999}, p. 62). ^{Rodríguez-Castellano (1952)} also found complete /r/ deletion in cases like matarme ENT#091;maˈtameENT#093; 'to kill me' and decirte ENT#091;deˈsiteENT#093; 'to say to you'. On the other hand, for ^{Jiménez Fernández (1999}, p. 62), /rl/ can also be pronounced as ENT#091;llENT#093; (e.g. decirlo ENT#091;deˈθilloENT#093; 'to say it'). I believe, however, that /r/ neutralisation in these instances without modifying the preceding vowel or consonant is only found in lower sociolects.

^{Salvador (1957}) reported /rn/ pronounced as ENT#091;ɹnENT#093; and as an intermediate sound ENT#091;ɹnENT#093;, which he believed is the first step before assimilation. The assimilation process is even more developed in /rl/, with the solutions ENT#091;lɹENT#093;, ENT#091;lrlENT#093; and ENT#091;llENT#093;. However, ^{Wulff (1889}) already documented ENT#091;ˈkanneENT#093; for carne 'meat', and ENT#091;kanniseˈɾiaENT#093; for carnicería 'butcher's'.

4.2.4. Yeísmo

^{Salvador (1957}) reported yeísmo (pronouncing /ʎ/ as /ʝ/), in Cullar-Baza (Granada), as a recent phenomenon there. ^{Salvador (1957)} found three older ladies who did not have yeísmo and who maintained /s/; nevertheless, he did not find /ʎ/ in any male speakers. Likewise, ^{Zamora Vicente (1960}, p. 312) reported that in some places, the distinction /ʎ/ vs. /ʝ/ was maintained by women. This supports the claim that women's speech is more normative than men's (e.g. ^{García Mouton, 1992}; ^{Cepeda, 1990}; and ^{Calero, 1990}). ^{García Mouton (1992)} theorised that in some places, women's speech is more archaic because that is what the norm says, but in cities it is more modern because the prestigious norm of the city is innovative, theory also supported by ^{Cortés Rodríguez (1990}).

A few years later, ^{Peñalver Castillo (2006}) would report yeísmo as the norm in Cabra, which contrasted with ^{Alvar's (1955a}) and ^{Llorente's (1962)} findings of lack of yeísmo in the northeast of the province of Granada. ^{Zamora Vicente (1960}, p. 311) found lack of yeísmo in parts of the provinces of Huelva, Granada, Seville, and Jaen, and in smaller areas between Cadiz and Malaga. ^{Llorente (1962)} found that /ʎ/ was maintained mainly in the surroundings of the Guadalquivir river and in areas closer to Portugal.

^{Zamora Vicente (1960}, p. 309) argued that Andalusian yeísmo is older than the one found in Madrid, or at least it was considered Andalusian before it was known in Madrid. It is worth mentioning that some intermediate sounds between /ʝ/ and /ʎ/ have been reported (^{Rodríguez-Castellano and Palacio, 1948a}). A detailed account of the spread of yeísmo across Andalusia as well as information on how /ʎ/ and /ʝ/ are pronounced across the region can be found in ^{Alvar (1973}, map 1703).

4.2.5. Gemination

^{Schuchardt (1881}) was the first author to report gemination in EAS (e.g. este ENT#091;ˈetteENT#093; 'this sing. masc.', casco ENT#091;ˈkakkoENT#093; 'helmet'), and this would not be reported again until ^{Rodríguez-Castellano and Palacio (1948a}). Later on, ^{Alvar (1955b}) would also identify this feature, which he explained as a phenomenon also found in Greek, where /s/ was aspirated in the sequence /sm/ and the final outcome became ENT#091;mmENT#093;.

^{Gerfen and Hall (2001}) were the first ones to analyse other aspiration phenomena apart from word-final /s/. ^{Gerfen and Hall (2001)} thought that as aspiration can occur with any obstruent in coda in EAS, not just in /-s/, we could have new homonyms in EAS, such as casta /ˈkasta/ 'cast' and capta /ˈkapta/ 'he/she captures' being pronounced ENT#091;ˈkaht.taENT#093;. However, ^{Gerfen and Hall (2001)} proved that that contrast is not totally neutralised, as they found that there is a "subtle yet consistent production pattern distinguishing forms derived from an underlying /s/ and those derived from an underlying voiceless stop (/p/ or /k/)" (p.2), which ^{Bishop (2007}) would also support later on. ^{Gerfen and Hall (2001)} found that "aspirating an /s/ results in a longer duration of aspiration, while aspirating a /p/ or /k/ results in more medial consonant gemination" (p. 5), which coincides with ^{Bishop's (2007)} findings. However, ^{Gerfen and Hall (2001)} also found variation between the duration of vowels preceding deleted /s/, /p/, or /k/, so they did not find any significant difference between the total duration of words with aspirated /s/, /p/, or /k/. ^{Gerfen and Hall (2001)} concluded that aspiration occurs with all coda obstruents and that it causes aspiration of the vowel and gemination of the following consonant. However, a preliminary study of data I have gathered in western Almeria suggests that deletion does not affect /l/, /r/, and /n/ in word-medial coda, at least not in all contexts.

^{Campos-Astorkiza (2003}) studied EAS lengthening using Optimality Theory and she saw lengthening as a compensatory feature. However, she found consonant gemination rather than vowel lengthening. ^{Campos-Astorkiza (2003)} considered gemination a way of keeping the same number of elements as the input and believed that this "repairs a banned structure deleting the features of the targeted consonant while keeping the same number of input segments" (p. 2) . This concept was later also studied by ^{Carlson (2012}). Furthermore, quoting ^{Wilson (2001}), ^{Campos-Astorkiza (2003)} said that deletion processes in biconsonantal clusters consistently delete the first consonant across different languages. ^{Campos-Astorkiza (2003)} concludes that "compensatory lengthening arises as a device to maintain the same number of segments as in the input" (p. 10).

^{Bishop (2007}) also studied gemination in EAS and he concluded that the length of a stop consonant following word-medial /s/ aspiration, but not the aspiration itself, acts as "a strong, disambiguating cue to listeners in making phonemic decisions as to an underlying coda" (p. 1765). This is in line with the findings reported by ^{Gerfen and Hall (2001}), who have also supported phonemicisation of gemination. I believe that this word-medial phonemicisation can be extended to other contexts, as we would have the same case in word boundaries (e.g. las casas 'the houses'). For ^{Bishop (2007)}, this incomplete neutralisation is characterised by durational differences of less than 15 or 20 milliseconds. ^{Bishop (2007)} admitted that this difference is much smaller than anything assumed to represent linguistic contrast; however, he explained how these differences are consistent in each of the analysed underlying contrasts and that there are cases in German and in Dutch where linguists have found that such small durational differences can be used by native listeners to distinguish between neutralised segments.

As ^{Bishop (2007}) explained, other works, such as ^{Gerfen and Hall (2001}) and ^{Gerfen (2002)}, had studied how duration differences in /s/ aspiration represent a case of incomplete neutralisation; nevertheless, ^{Bishop (2007)} provided the first analysis of the perceptual consequences of such durational differences. ^{Bishop (2007)} asserted that increasing the closure duration is a cue to underlying /p/, as it made more people perceive the word hasta 'until' as apta 'capable'; however, increasing the length of aspiration had no effect on the judgment of the listeners. The results presented in ^{Gerfen and Hall (2001)}, ^{Gerfen (2002)}, ^{Bishop (2007)}, and ^{O'Neill (2010}) showed that the cue to recognise word-medial /s/ deletion lies in the gemination of the consonants, giving gemination a more important role than the secondary one given to it by earlier researchers (e.g. ^{Mondéjar Cumpian, 1979}).

^{O'Neill (2010}) also analysed gemination related to /s/, /p/, and /k/; nevertheless, his analysis had a different approach and focus to the ones in ^{Gerfen and Hall (2001}) and ^{Bishop (2007}). ^{O'Neill (2010)} studied the phonetic cues that speakers of Andalusian Spanish use to distinguish between pairs such as // /V/ +/s/+ /p, t, k/ + /V/ // (VsOV), as in costa 'coast', and // /V/ + /p, t, k/ + /V/ // (VOV), as in cota 'level'. ^{O'Neill (2010)} concluded that the difference between VsOV and VOV is never neutralised, and that the difference between them does not lie, at least mainly, in the presence of aspiration before the stop but in the pronunciation of the stops, given that in one third of the cases there was no aspiration before stops. In EAS, stops in VsOV are always voiceless, more aspirated and longer, and in VOV, stops are not aspirated and they are voiceless only in 13% of cases, voiced in 69% of cases, and pronounced as approximants in 19% of cases. As ^{O'Neill (2010)} explained, this last pronunciation of stops as approximants means that the contrast /p, t, k/ vs. ENT#091;ß, ð, ɣENT#093; can be neutralised in 12.5% of cases. ^{O'Neill (2010)} also claimed that ENT#091;ßENT#093;, ENT#091;ðENT#093;, and ENT#091;ɣENT#093; are pronounced as approximants in 66% of cases and deleted 34% of the time, with men showing a stronger tendency to delete approximants and voice voiceless stops. ^{O'Neill (2010)} concluded that there can be a great degree of variation in the pronunciation of consonants in VsOV and VOV in Andalusian Spanish; however, the phonetic cues that distinguish those two pairs in EAS seem to be aspiration, voicing, and the length of the stop.

Regarding gemination, ^{Mondéjar Cumpián (1979}) presented an interesting view: he believes that vowel opening carries the semantic load of deleted /s/, although he argued that in pairs like pisto 'ratatouille' and pito 'whistle', the semantic load is carried by gemination, which he considered to be biphonemic.

4.2.6. EAS consonants: other phenomena

In this section, I will deal with EAS consonant phenomena not covered in previous sections. ^{Rodríguez-Castellano and Palacio (1948b}) claimed that in word-final position before a consonant, /l/ is pronounced ENT#091;rENT#093;; nevertheless, when the following consonant is /n/, /r/ is aspirated and nasalised, with the final product being something between a nasal aspiration and a geminated ENT#091;nENT#093;. ^{Rodríguez-Castellano and Palacio (1948a)} also believed that /n/ in the Cabra region has no alveolar character, but a weak velar one; ^{Zamora Vicente (1960}, p. 324) also said that /n/ can be velarised in Andalusia.

^{Rodríguez-Castellano and Palacio (1948b}) expanded the work presented in ^{Rodríguez-Castellano and Palacio (1948a)} and they focused on the consonants of Cabra. The authors described how initial /f/ is only altered before /w/ (e.g. fuente ENT#091;ˈhwenteENT#093; 'fountain'), and they also found that the neighbouring town palatalised /n/ into ENT#091;ɲENT#093; as in Leonés. ^{Salvador (1957}) also mentioned ENT#091;ŋENT#093; in coda in Cabra and the area of Cullar-Baza and found /nj/ merged in ENT#091;ɲENT#093;, as in poniéndose ENT#091;puˈɲendoseENT#093; 'putting something on'.

In word-initial position, /d/ is only omitted in donde 'where', and ^{Rodríguez-Casellano (1952)} found that intervocalic /d/ deletion causes changes in the conjugation of verbs (e.g. ayudo ENT#091;aˈʝuoENT#093; 'I help'; suda ENT#091;ˈsuaENT#093; 'he/she sweats'). According to ^{Jiménez Fernández (1999}, p. 70), the loss of intervocalic consonants is registered from the sixteenth and seventeenth century as in ENT#091;keˈdaoENT#093; for ENT#091;keˈdaðoENT#093; (quedado 'stayed') in the Cancionero of Pedro del Pozo (1547). According to ^{Jiménez Fernández (1999}, p. 70), ENT#091;ðENT#093; is the most frequently deleted consonant in intervocalic position, followed by ENT#091;ßENT#093;, ENT#091;ɾENT#093; and ENT#091;nENT#093;.

^{Rodríguez-Castellano and Palacio (1948b}) accepted that /n/ is very relaxed, but they argued that it is the only consonant which is not ommitted. Likewise, /d/, /n/, and /l/ are the initial voiced consonants that are less influenced by aspiration. ^{Rodríguez-Castellano and Palacio (1948b)} report that, in los dientes ENT#091;lo(ð(je̞nte̞ENT#093; 'the teeth', ENT#091;ðENT#093; is fronted and slightly devoiced and that in las nueras ENT#091;lah ˈnwerahENT#093; 'the daughters-in-law', /n/ is not modified and aspirated /s/ is maintainted and heavily nasalised and semi-voiceless. Equally, /l/ is not altered by a previous aspirated /s/ and this is maintained usually as a simple aspiration, more or less assimilated in some cases. ^{Rodríguez-Castellano and Palacio (1948b)} were also the first authors to report that /s/ is not aspirated when a word already has another aspirated sound, as in los ojos ENT#091;los ˈohoENT#093; 'the eyes', which is due to dissimilation.

Assimilation is also very important to understand EAS. ^{Alarcos Llorach (1981}, p. 241) explained how common word-internal consonant clusters were reduced in Latin through assimilation, as in ipse ( isse. ^{Alarcos Llorach (1981}, p. 243) also explained that some intervocalic voiced stops turned into fricatives in Latin, voiceless consonants were voiced, and geminated consonants were simplified, in this exact order. I believe that these changes are what EAS is exhibiting.

^{Alonso et al. (1950}) reported frequent examples of ENT#091;vENT#093; in Granada (e.g. fiebre ENT#091;ˈfjevreENT#093; 'fever'), but these do not appear regularly, not even within a single speaker, and Granada speakers seem to lose it after being away from the city for some time. For ^{Salvador (1957}), ENT#091;vENT#093; is a result of aspirated /s+b/ in Cullar-Baza, but not in other positions.

^{Salvador (1957}) considered word-final /d/ deletion older than /s/ deletion word-finally, although the vowel preceding deleted /d/ is not altered. Despite this, he claimed that speakers from high socio-cultural levels pronounce verdad 'truth' as ENT#091;be̞rˈða̞jENT#093;. ^{Salvador (1957)} also argued that the distinction between /s/ and /θ/ is neutralised word-finally in EAS, although this was challenged in ^{Herrero de Haro (2016b}).

In the present paper, we have seen multiple examples of how phonologisations are usually accompanied by a desphonologisation that causes it and /n/ deletion could be considered another example of this. ^{Alarcos Llorach (1981}, p. 135) explained how phonological fusion changed Old French vowel + /n/ into nasalised vowels, which has also been identified in EAS. ^{Zamora Vicente (1960}, p. 324) and ^{Jiménez Fernández (1999}, p. 72) described nasalisation of Andalusian vowels preceding /n/ deletion in coda (a phenomenon registered in the neighbourhood of Zaidin, Granada). ^{Alvar (1955b}) described how nasalisation of vowels also happened in Latin after /m/ deletion and it happens with any nasal in Sanskrit. However, Alvar (1955b) also claimed that in some parts of Andalusia and in the Canaries, this nasalisation has been lost, which represents a more advanced stage in this evolution.

The fricativisation of /tʃ/ is another phenomenon that EAS has in common with French. ^{Zamora Vicente (1960}, p. 313) claimed that /tʃ/ is pronounced ENT#091;ʃENT#093; in Granada and the surrounding areas, in the south of Seville, in Cadiz province, in western Malaga and on the coast of Almeria close to the capital, with /tʃ/ being maintained in the rest of Andalusia. However, ^{Jiménez Fernández (1999}, p. 69) and ^{Melguizo Moreno (2007}) thought that ENT#091;ʃENT#093; was being lost to ENT#091;tʃENT#093; in Granada. In her sociolinguistic study, ^{Melguizo Moreno (2007)} reported that, although very common in the 1950s, fricativisation of /tʃ/ is now stigmatised in Granada and she asserted that ENT#091;ʃENT#093; is found mainly word-initially and almost exclusively in men from low sociolects. According to this study, ENT#091;ʃENT#093; is more common in older men, although it can also appear in men between 25 and 54 years of age; in the female population, it can also be found amongst older women with medium level instruction. Other socio-linguistic studies (e.g. ^{Villena Ponsoda et al., 2003}; ^{Hernández-Campoy and Villena Ponsoda, 2009}) have analysed the sociolinguistic forces at play in EAS and in WAS; however, these will not be discussed in the present paper as they do not focus on phonetic-phonological characteristics of EAS.

The final consonant phenomena to be discussed in this paper will be variation in EAS stops. ^{O'Neill (2010}) argued that voicing is a continuum in EAS, as he found different degrees of voicing in different consonants. Likewise, he found different degrees of occlusivity, having encountered some stopless stops. Following on from that, ^{O'Neill (2010)} claimed that, in Andalusia, the most frequent pronunciation for stops in the sequence VOV is a voiced stop, and the second one is an approximant (except in Granada). ^{O'Neill (2010)} also noticed a tendency amongst men from Granada city to voice the stops more and to pronounce them as approximants; however, women maintain voiceless stops to a greater degree. This tendency, however, is the opposite in Almeria. Furthermore, ^{O'Neill (2010)} believed that stops that are mainly voiceless are longer than those that are mainly voiced, and the latter ones are longer than the ones pronounced as approximants. For ^{O'Neill (2010)}, the most common pronunciation for intervocalic /p/, /t/ and /k/ in Andalusia is a voiced stop, but when /s/ precedes them the occlusion is always voiceless and aspirated. ^{O'Neill (2010)} concluded that there is a great degree of variation in the pronunciation of voiceless stops in Andalusian and he suggested that this variation could mean that there is an ongoing change where the phonemic system of Andalusian stops is being adjusted. ^{O'Neill (2010)} explained how this already happened in Spanish, as geminated voiceless stops in Latin became simple consonants in Spanish (e.g. gutta - gota 'drop'), voiceless stops became voiced (e.g. catena - cadena 'chain'), and voiced stops disappeared (e.g. sedere- seer- ser 'to be').