| 44

Revista de Variación y Cambio Lingüístico

ISSN: 3020‐9854

/s/ aspiraon in Andalusian Spanish in word internal posion and across

word boundaries: an experimental study of four cies

La aspiración de /s/ implosiva en el interior de palabra y en la fonética sintáctica:

un estudio experimental de cuatro ciudades andaluzas

Paul O’Neill

Instute for Romance Philology, Ludwig‐Maximilians‐Universität (LMU)

Munich, Germany

Recibido / Submied: 01/11/2024

Aceptado / Accepted: 02/12/2024

Abstract

This study presents acousc data relang to the realisaon of coda /s/ in four cies in

Andalusia. It diﬀers from the majority of recent studies in that it analyses the eﬀects of coda /

s/ on all types of consonants, both word‐internally and across word‐boundaries. In the laer

context, diﬀerent morphological values of /s/ are also analysed. The results suggest that the

phonological system of Andalusian Spanish is being radically altered. In word internal

posion, the historical eﬀects of /s/ aspiraon are producing a series of new incipient

phonemic disncons of which the most robust are elongated and/or aspirated consonants.

As for across word‐boundaries, the ﬁndings are only signiﬁcantly diﬀerent from word‐internal

coda /s/ for the city of Seville. For the speech of this city, it seems that /s/ has been lost / is

being lost when it is a marker of 2sg on verbs and plural on nouns. However, when /s/ is part

of lexeme, i.e. in the word dos ‘two’, the phonec cues are robust and similar to those in

word‐internal posion. I end the arcle with a tentave hypothesis as to how this situaon

could be eﬀecng the morphological system and point to possible future developments.

Keywords: Andalusian Spanish; aspiraon; geminaon; sound change; morphologisaon

Resumen

Este estudio presenta datos acúscos relavos a la realización fonéca de /s/ implosiva en

cuatro ciudades de Andalucía. Se diferencia de la mayoría de los estudios recientes en que

analiza los efectos de /s/ implosiva ante todo po de consonantes, tanto en el interior de

palabra y en la fonéca sintácca. En este úlmo contexto, también se analizan los disntos

valores morfológicos de /s/. Los resultados sugieren que el sistema fonológico del español de

Andalucía se está cambiando notablemente. En el interior de palabra, los efectos históricos

de la aspiración de /s/ están produciendo una serie de nuevas disnciones fonémicas de las

que las más robustas son las consonantes alargadas y/o aspiradas. En cuanto a la fonéca

sintácca, los resultados solo son signiﬁcavamente diferentes del contexto de interior de

palabra para la ciudad de Sevilla. Para el habla de esta ciudad, parece que /s/ se ha perdido o

se está perdiendo cuando es un marcador de 2sg en los verbos y de plural en los sustanvos.

Sin embargo, cuando /s/ forma parte del lexema, es decir, en la palabra dos, las pistas

fonécas son sólidas y similares a las del contexto de interior de palabra. Concluimos el

arculo con una hipótesis provisional sobre cómo esta situación podría estar afectando al

sistema morfológico y señalamos posibles desarrollos futuros.

Palabras clave: Español de Andalucía; aspiración; geminación; cambio fonéco y fonológico;

morfologización

DOI: hps://doi.org/10.30827/3020.9854rvcl.1.2.2024.32243

Variación. Revista de variación y cambio lingüísco, 1(2), diciembre 2024, pp. 44‐69

| 45

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

1. INTRODUCTION

Recently, aspiraon of coda /s/ before voiceless occlusives in Andalusian Spanish has

aracted substanal academic aenon due to its various phonec realisaons. The most

notable of these is the lack of phonec cues of aspiraon aer the vowel and, in its stead, the

most frequent pronunciaon is a voiceless aspirated stop in Western Andalusian Spanish (WAS)

and a voiceless aspirated and elongated stop in Eastern Andalusian Spanish (EAS). There has

been much debate on the phonec movaon and phonological status of such sound changes.

Regarding the laer, Torrerira (2007b, 2007a, 2012) has argued that these post‐aspirated

pronunciaons are not intended by speakers but are the result of purely phonec processes

relang to the synchronisaon and overlap of arculatory gestures. However, others have argued

that at least some, if not the majority, of post‐aspirated pronunciaons are intended by speakers

(O’Neill, 2010, 2009; Parrell, 2012). All scholars agree, however, that there is a sound change in

progress in Andalusian Spanish, resulng in a process whereby the historic eﬀect of /s/ aspiraon

is undergoing a metathesis and producing post‐aspirated stops, which also have a tendency to

also be elongated in EAS ((Gerfen, 2002; Moya Corral, 2007; O’Neill, 2010, 2009; Parrell, 2012;

Ruch, 2012; Ruch and Harrington, 2014; Torrerira, 2007b, 2007a, 2012; Torrerira and Ernestus

2011). Regarding the phonec movaon of the sound changes. Issues have centered around

whether an elongated consonant [paʰtˑa] [pat:a] (due to the overlap between the gloal gesture

for the aspiraon and the occlusive gesture) is a necessary inial step, or not, for the triggering

of post aspiraon via increased air‐pressure during the closure and concomitant early release of

the stop (O’Neill, 2010, 2009; Torrerira, 2007b, 2007a, 2012; Torrerira and Ernestus 2011; Ruch,

2012; Ruch and Harrington, 2014). Alternavely, it has been defended that post‐aspirated stops

could have emerged via the reorganizaon of the gloal spreading gesture for /s/ and the oral

closure gesture for the stop, whereby instead of the gestures ocurring sequenally, they come to

be synchronised with each other (Parrell, 2012). The diﬀerent explanaons make diﬀerent

predicons regarding the trade‐oﬀs between duraons of pre‐aspiraon, post‐aspiraon and the

occlusive closure gesture. However, diﬀerent experiments produce diﬀerent and, at mes,

contrasng results (see Ruch & Harrington (2014) and Ruch & Peters (2016) for comprehensive

overviews).

Common to all the aforemenoned studies, however, is (a) their exclusive focus on /s/ before

voiceless occlusives in word internal contexts (e.g. pasta, caspa) and (b) the variety in the

geographical origin of the parcipants selected to represented both EAS and WAS. Thus, whilst

the parcipants in the study of Ruch and Harrington (2014) all came from either the city of

Seville (WAS) or the city of Granada (EAS), Torreira’s three speakers of WAS were from the city of

Cádiz, and the 20 parcipants in Parrell’s (2012) study came from the large provinces of both

Cádiz and Seville. Torreira’s (2007b) study stands out as having parcipants from geographically

distant locaons spread across seven diﬀerent cies and four diﬀerent provinces of Andalusia

(Seville, Écija (85 km from Seville) , Lepe (121 km from Seville), Camas (8 km from Seville);

Chiclana de Segura (209 km from Granada), Javalquinto (135 km from Granada), La Rábita (56 km

from Granada), Almuñécar (46.5 km from Granada)).

In normal speech, /s/ tends not to have an alveolar fricave pronunciaon before any

consonants in Andalusia but there is a marked lack of research as to (a) what these

pronunciaons are, (b) the eﬀects of external sandhi and (c) whether the morphological status of

<s> inﬂuences its pronunciaon. Moreover, Andalusia is the largest autonomous community in

Spain and one which is parcularly rich in diatopic variaon, thus it is expected that the speech

of people in a village in Cádiz will be markedly diﬀerent from that of Seville capital, despite both

being WAS; the former is usually characterised by ceceo whilst the laer by seseo/disnción.

The present study, therefore, does not primarily intend to explore the contested academic

quesons, summarised above, regarding the phonec origin and phonological status of any new

| 46

Paul O'Neill

Variación 1(2) (2024), 18‐30

For those words in the corpus which it was impossible to solicit via the naming of images, subjects read the words

embedded in a carrier phrase, from a card.

Note that speakers were presented with two images which corresponded to the word in ﬁrst phrase and just one image in

the second phrase to ensure that they produced the singular and plural forms.

types of pronunciaons; rather, the aim is to (a) expand on previous research in analysing more

linguisc contexts of coda /s/, speciﬁcally: word internal posion before all types of consonants

and across word boundaries pre‐vocalically and in all preconsonantal posions, and (b) explore

the diﬀerent types of morphological <s>, speciﬁcally: as a marker of 2SG on verbs, PLURAL on

nouns and in the word dos; (c) dispense with the terms WAS and EAS and focus on the speech of

parcular geographical locaons, speciﬁcally: the cies of Almería, Cádiz, Granada and Seville.

2. METHOD

The data for this experiment came from a corpus of recordings collected by the author in

2005. The recordings were of 4 subjects (2 male and 2 female), in the four cies menoned

above; all parcipants were 18 years old and nave inhabitants of their respecve cies. The

corpus contained words containing the sequences V[owel]S[ibilant]C[onsonant]V[owel] (VSCV) in

both word internal posion and across words (VS#CV) and their minimal or near minimal pairs

(VCV and V#CV sequences respecvely). All the diﬀerent types of consonants were included

([O]cclusives, [N]asals, [A]pproximants, [F]ricaves) but the numbers of each type of consonant

were uneven. The corpus also contained sets of minimal pairs in which coda orthographic <s>

occurred/did not occur in word ﬁnal prevocalic contexts (VS#V vs. V#V). Given that the

pronunciaon of coda orthographic <s> is a sociolinguisc variable and speakers in formal

situaons and when reading can pronounce it as a voiceless sibilant, subjects were presented

with photographs depicng the words in the corpus, along with a number of control words that

did not contain orthographic <s>; parcipants were requested to name what they saw

. In word

internal posion the following carrier phrase was used dame un ____ pa[ra] mí ‘give me a ____

for me’, whilst in the external sandhi contexts there were three diﬀerent types of carrier phrase

depending on what type of <s> was being tested and in what phonological contexts. The ﬁrst

type of <s> tested was the marker of 2SG (S2SG) in external sandhi before all types of consonants

and in prevocalic posion; its pronunciaons were contrasted with phrases with 3SG verb forms

in which there was no <s> (VS2SG#CV vs. V#CV contexts or VS2SG#V vs. V#V contexts). Speakers

were required to name the images they saw in the photographs in the phrase tu siempre comes

____ ‘you always eat ____’ and then again in the phrase él siempre come ____ ‘he always eats

____’. The second type of <s> tested was that in the word dos (SDOS) before all types of

consonants in external sandhi and prevocalically; speakers were required to name what they saw

in the photographs in the phrase digo dos ____ pa[ra] ’ ‘I say two ____ for you’ and then again

in the phrase digo un/una ____ pa[ra] ’ ‘I say a ____ for you’ (VSdos#CV vs. V#CV contexts)

These same carrier phrases were also used to examine the eﬀects of the presence/absence of

the third type of <s>, the nominal plural marker (SPLURAL) and its eﬀects in external sandhi

exclusively on the following /p/ consonant in the phrase digo dos ____ pa[ra]  (VSPLURAL#C/p/V

vs. V#C/p/V contexts). In order to test the eﬀects of external sandhi in this context prevocalically

(VSPLURAL#V vs. V#C/p/V) the following phrases were used: digo dos ____ ahora ‘I say two ____

now’ and digo un(a) ____ ahora ‘I say a(n) ____ now’ (VSdos#V vs. V#V and VSPLURAL#V vs. V#V

contexts). The words for the prevocalic sandhi contexts are those in (2) and the words used in

the word internal and preconsonantal sandhi contexts are given in (1). For clarity I give examples

of the phrases used in accordance with the diﬀerent phonological and morphological contexts in

(3), along with the total number of words analysed. Note that due to space restricons in the

sandhi contexts only data from Seville and Granada will be presented.

| 47

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

(1) The words used in the carrier phrases for word internal contexts and the ﬁrst two types

of sandhi contexts.

(2) The words used in the carrier phrases for tesng the eﬀects of external sandhi in

prevocalic contexts.

abeja(s) ‘bee(s)’, ajo(s) ‘garlic(s)’ árboles ‘tree(s)’, autobuses ‘bus(es)’, escudo(s) ‘shield(s)’,

espada(s) ‘sword(s)’, higo(s) ‘ﬁg(s)’, hormiga(s) ‘ant(s)’, ojo(s) ‘eye(s)’, oreja(s) ‘ear(s)’,

ordenador(es) ‘computer(s)’, uva(s) ‘grape(s)’.

(3) Examples of the carrier phrases in accordance with the diﬀerent phonological and

morphological contexts.

(a) VSCV vs. VCV (11 pairs of words per subject = 704 words overall for the four cies)

digo caspa pa[ra]  vs. digo capa pa[ra] 

I say ‘dandruﬀ’ for you vs. I say ‘cape’ for you

(b) VS2SG#CV vs. V#CV (36 pairs of words per subject = 288 for the city of Seville)

Tú siempre comes peras vs. él siempre come peras

You always eat ‘pears’ vs. he always eats ‘pears’

| 48

Paul O'Neill

Variación 1(2) (2024), 18‐30

(c)

VSdos#CV vs. V#CV and VSPLURAL#C/p/V vs. V#C/p/V (36 pairs of words per subject

per

context = 576 for the city of Seville)

digo dos peras pa[ra]  vs. digo una pera pa[ra] 

I say two ‘pears’ for your vs. I say one ‘pear’ for you

(d) VSdos#V vs. V#V and VSPLURAL#V vs. V#V (12 pairs of words per subject = 96 for the city

of Seville)

digo dos uvas ahora vs. digo una uva ahora

I say two ‘grapes’ now vs. I say one ‘grape’ now

The overall number of words analysed was 1,664, although some had to be discarded due to

either the speakers not saying the appropriate word or the quality of the recording. The data

were analysed acouscally using Praat (Boersma & Weenink, 2005) and note was taken of the

amount and type of cues for aspiraon of <s>, which was termed GLOT and the manner of

arculaon and voicing quality of the following consonant. Addionally, the following

measurements were taken for each word: (i) duraon of the gloal gesture before the consonant

(GLOT); ii) duraon of the consonant (C) and, exclusively with voiceless occlusive consonants, the

duraon of the VOT.

A stascal analysis was conducted using R (R Core Team, 2018) and lme4 (Bates, Maechler &

Bolker, 2015) to perform a linear mixed eﬀects analysis of the relaonship between the

phonological context (presence or lack of /s/) and the length and type of the following

consonant. Diﬀerent linear mixed eﬀects analyses were carried out in accordance with the

diﬀerent contexts in (3a‐c) above for each city and each type of consonants. The laer was

necessary since consonants diﬀer greatly in length and so all occlusive consonants had their own

stascal analysis and all nasal consonants etc. These analyses mostly all shared the same

method: phonological context and gender were coded as ﬁxed eﬀects. Intercepts for subjects

and consonants, as well as by‐subject random slopes for the eﬀect of phonological context, were

added to the model as random eﬀects. Visual inspecon of residual plots did not reveal any

obvious deviaons from homoscedascity or normality. P‐values were obtained by likelihood

rao tests of the full model with the eﬀect in queson against the model without the eﬀect in

queson. Also, categorical variables (presence of GLOT, quality of consonant) were converted

into dummy codes so they could be incorporated into a regression framework. In the cases of

VSCV vs. VSCV minimal pairs which were not occlusives (e.g. pairs such as cisne vs. cine; isla vs.

ila; desde vs. dedo), two‐tailed t‐tests were used to establish if the length of the consonants in

VSCV sequences was signiﬁcantly longer than in VCV contexts.

3. RESULTS

3.1 VSCV vs. VCV contexts: e.g. pasta vs. pata; cisne vs. cine; isla vs. ila; desde vs. dedo

A disncon is made between the sequences in which there is evidence of aspiraon aer the

vowel, termed VSCVGLOT, and those in which there is not, termed VSCVNOGLOT. The former are

only really present in VSOV sequences, outside these sequences aspiraon is only aested in

Seville and Cádiz and it is a very infrequent pronunciaon; it only appeared in 7% of occurrences

in Seville (4/52 tokens) and 9% in Cádiz (4/43 tokens). In both cies, the aspiraon is always

voiced and oen appears as either breathy voice at the end of the vowel (see (4)a) or throughout

the whole vowel (see (4)b).

| 49

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

(4) Spectrographs showing cues of aspiraon before the consonant in VSCV sequences

Within VSOV sequences, however, in all cies the pronunciaons with aspiraon before the

consonant (VSOVGLOT) outnumber those without it, as demonstrated by the ﬁgures in (5). As to

the phonec realisaons of this aspiraon, in the data, as illustrated by the spectrographs in (6),

there was only one token in which there was an idenﬁable segment of voiced aspiraon similar

in length to that of [s] in non‐aspirang variees (6a); all other tokens were characterised by the

presence of post‐aspiraon on the consonant combined with either breathy voice on the vowel

(6c), pre‐aspiraon on the consonant (6b) or a combinaon of the both (6d). In VSOVNOGLOT

contexts all variees are characterised by aspirated voiceless occlusives (6f); in Granada the

occlusive was also elongated (6e).

(5) The rates of pronunciaons of VSOV with cues of aspiraon (or nor) before the

consonant.

(6) The phonec realisaons of aspiraon of VSOV with cues of aspiraon before the

consonant

| 50

Paul O'Neill

Variación 1(2) (2024), 18‐30

In the stascal analysis which was used to esmate the extent to which O and VOT could be

predicted from the phonological context (presence or lack of /s/), as illustrated by the p‐values

below the boxplots in (7), the phonological context was only signiﬁcant in Granada for length of

O but in all cies for length of VOT. Speaker was signiﬁcant in all cies for the length of O, but

only signiﬁcant in Almería and Cádiz for VOT; gender was not signiﬁcant in any city.

(7) Boxplots of the length of O and VOT in the diﬀerent cies depending on the diﬀering

contexts (VSOV vs. VOV) and diﬀerent speakers.

| 51

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

In the theorecal literature much emphasis is placed on the interacons between the lengths

of GLOT, O and POST with respect to the origins of the post‐aspirated pronunciaons and

whether they are phonological or phonec; restricons of space prevent a detailed discussion

here but one theory is that these pronunciaons could be the result of an early occlusive gesture

which masks the aspiraon aer the vowel (O’Neill, 2010, 2009; Torrerira, 2007b, 2007a, 2012;

Torrerira and Ernestus, 2011; Ruch, 2012; Ruch and Harrington, 2014). This gestural overlap

coupled with the increased length of VOT can lead to a phonological reanalysis of these

sequences as consisng of either (a) voiceless aspirated occlusives or (b) elongated voiceless

aspirated occlusives. The fact that speaker was signiﬁcant for length of O in all cies could

suggest that there is sll much inter‐speaker variaon as to whether these sequences have been

phonologized. In this respect, observe the behaviour of speakers in Almería. The two female

| 52

Paul O'Neill

Variación 1(2) (2024), 18‐30

speakers seem to favour pronunciaon (b) whilst speaker M1 favours (a) and M2 displays much

variaon. Observe also the behaviour of speaker F2 in Seville, she stands out from the other

speakers as having longer occlusives in VSOV contexts which could suggest that she is not

intenonally producing short voiceless occlusives all of the me and that the elongated

occlusives are the historical forerunners of these pronunciaons (contrary to Parell (2012)). Note

that this speaker also consistently had more VSOVGLOT pronunciaons and, in general, VSCV

pronunciaons with more indicaons of the presence of an open glos corresponding to

orthographic <s>.

As to the voicing quality of the occlusives these were always voiceless in VSOV contexts and

mainly voiced in VOV contexts. In these laer contexts, as illustrated in (8), the occlusive

consonants could also be realised as approximants; sex was a contribung factor to such voiced

pronunciaons with men voicing and producing more approximants than women (O’Neill, 2010).

(8) Quality of consonant in VOV sequences

Moving away from occlusive consonants, all other consonants in VSCV sequences were

signiﬁcantly longer than in VCV sequences with the excepon of fricave consonants (see (9) in

which the p‐vales correspond to two‐tailed t‐tests); Granada was the city in which the diﬀerence

between the length of the consonants in the diﬀerent sequences was greater and also the city

where the consonants were generally the longest in VSCV sequences. Note that due to space

restricons the data is presented by city and not by speaker. As to the manner of arculaon of

the consonants, there was greater evidence in Seville and Cádiz for the coarculaon of

aspiraon with the following consonant and the concomitant changes to its manner of

arculaon (see (10) for an overview). Thus, as illustrated by the spectrographs in (11), in Seville

and Cádiz nasals could be realised with breathy voice (a), lateral approximants with fricaon (c),

and there was a greater tendency to produce spirant approximants as fricaves (e) which could

also be voiceless (i). Contrasvely in Granada and Almería nasals and laterals were simply

elongated ((b) and (d) respecvely) and the spirant approximants could also be elongated (f) but

they could also be realised as voiced stops (g) and voiced fricaves ((h) & (j)). Note that in all

cies modal voice was only aested for nasals in VNV sequences; in VLV contexts all consonants

were produced as approximants, and in VAV contexts all consonants were either approximants or

were elided (see also O’Neill (2010)).

| 53

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

(9) Length of consonants in VSCV vs. VCV sequences (p=vales below graphs from two‐tailed

t‐tests)

(10) Manner of arculaon of consonant in VSCV and VCV contexts

| 54

Paul O'Neill

Variación 1(2) (2024), 18‐30

(11) Spectrographs of VSCV and VCV sequences

| 55

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

In conclusion to this secon, all VSCV contexts, with the excepon of fricave consonants, are

longer than their VSV minimal pairs (especially in Granada) and, with the excepon of occlusives

consonants, there are no real cues for aspiraon before the consonants; there is a tendency in

the western cies for the following consonants to be produced with some evidence of an open

glos (breathy voice/voiceless fricaon). In VSOV contexts, the post‐aspiraon and voiceless

quality of these consonants is a pan‐Andalusian phenomenon and it is only in Granada where,

for all speakers, the occlusive consonant is signiﬁcantly and consistently longer in VSOV vs. VOV

contexts.

3.2. VS#CV vs. V#CV contexts

Recall that due to space limitaons only pronunciaons from Seville and Granada will be

analysed for external sandhi contexts, in which a disncon will be made between (a) occlusive

consonants (VS#OV contexts) and other consonants and (b) diﬀerent morphological types of <s>:

marker of 2SG in verbs (S2SG), marker of PLURAL in nouns (SPLURAL), part of the lexeme in the

word dos (SDOS).

3.2.1 VS2SG#CV vs. V#CV: tú siempre comes ____ vs. él siempre come ____

When <s> was a marker of 2SG there were only tokens for cues of aspiraon before occlusive

consonants (that is VS2SG#OV sequences). However, the diﬀerences between both cies was

qualitavely diﬀerent. In Seville, there were only two tokens (one for M2 and one for F2) in

which there were cues of aspiraon before the occlusive whereas in Granada cues were present

in 23 tokens, which constuted 64% of all VS2SG#OV contexts. There were also stark diﬀerences

between the two cies regarding the quality, voicing and length (both O and VOT) of the

occlusives in comparison to V#OV contexts. In Seville the occlusives in VS2SG#OV contexts paern

with the occlusives in V#OV contexts: they have the same propensity to be voiced/realised as

approximate consonants, and both O and VOT have similar duraons. In Granada, however, the

occlusives in VS2SG#OV contexts are markedly diﬀerent from the V#OV contexts and paern with

the VSOV contexts: they tended to be more voiceless, elongated and more heavily aspirated.

These diﬀerences are illustrated by the spectrograms in (12) the percentages in (13), and the

box‐plots in (14). Regarding the laer, note that there are clear gender diﬀerences with respect

to the length of O, but not VOT, in Granada.

(12) Spectrograms of VS2SG#CV vs. V#CV contexts

| 56

Paul O'Neill

Variación 1(2) (2024), 18‐30

(13) Percentages relang to the phonec realisaon of the occlusives in VS#OV and V#OV

(14) Boxplot of the length of O and VOT in VS2SG#OV and V#OV contexts

| 57

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

The impressionisc conclusion that in Seville the occlusives in VS2SG#OV contexts paern

with those in V#OV whereas in Granada they are markedly diﬀerent ( VS2SG#OV paerning more

with VSOV contexts) is conﬁrmed by the inferenal stascal analysis which compared VS2SG#OV

and V#OV contexts. The results were the following:

• VS2SG#OV contexts did not reliably predict the presence of aspiraon in Seville (χ2 = 0.237,

p= 0.626) but did in Granada (χ2 = 42.05, p= < 0.000)

• VS2SG#OV contexts did not have any signiﬁcant eﬀect on the type of consonant in Seville (χ2

= 1.478, p= 0.224) but did in Granada (χ2 = 16.865, p= <0.000) where the presence of <s>

increased the number of voiceless pronunciaons

VS2SG#OV contexts did not have any signiﬁcant eﬀect on the length of the following

consonant in Seville (χ2 = 1.546, p= 0.218) but did in Granada (χ2 = 10.003, p= 0.002) where the

presence of <s> increased the length of the consonant by 49.11ms ± 7.3 (standard errors).

As for the other consonants (contexts VS2SG#NV, VS2SG#AV and VS2SG#FV), in Seville when

proceeded by S2SG there were no indicaons of any aspiraon coarculated with the consonants,

in contrast to when these sequences occurred in internal posion; all nasals were pronounced

with modal voice, all laterals as approximants and all spirant approximants as approximants or

elided. The consonants were also not signiﬁcantly longer with respect the S2SG‐less contexts

(V#NV, V#AV and VS#FV). The same was not true, however, in Granada since the consonants

were signiﬁcantly longer. Compare the box‐plots below in (15).

(15) Boxplots of length of C according to speaker for Seville in the context VS2SG#CV vs. V#CV

Again, the impressionisc conclusion that in Seville the consonants in VS2SG#CV contexts

paern with those in V#CV contexts whereas in Granada they are markedly diﬀerent (VS2SG#CV

| 58

Paul O'Neill

Variación 1(2) (2024), 18‐30

paerning more with VSCV contexts) is conﬁrmed by the inferenal stascal analysis R (R Core

Team, 2018) and lme4 (Bates, Maechler & Bolker, 2015) were used to create a mixed linear

model to esmate the extent to which the length of the consonant could be predicted from the

ﬁxed factors phonological context ((VS2SG#CV vs. V#CV), and sex. Intercepts for subjects and

consonants, as well as by‐subject random slopes for the eﬀect of phonological context were

added to the model as random eﬀects. The results were that, in Seville, the length of the

consonant it is signiﬁcantly aﬀected by speaker (χ2(1)= 67.74, p= <0.000), by the type of

following consonant (χ2(13)= 255.51, p= <0.000) but not by the phonological context = preceding

orthographic <s> (χ2(1)= 1.242, p= 0.265) or sex (χ2(1)=0, p=1). In Granada, however, the length

of the consonant is signiﬁcantly aﬀected by the phonological context (χ2(1)= 4.3854, p= 0.036),

in addion to being aﬀected by the speaker (χ2(1)= 14.651, p= <0.000) and the following

consonant (χ2(13)= 57.786, p= <0.000). As with Seville, however, the sex of the speaker was not

signiﬁcant (χ2(0.4084)=0, p=0.522).

In summary, in Seville there are no phonec cues to disnguish between 2SG and 3SG present

indicave forms of the verb in sentences such as tú siempre comes limas vs. él siempre come

limas and therefore the orthographic <s> corresponds to no mental unit. The same, however,

cannot be said for Granada since in this city, in VS2SG#C contexts (e.g comes limas), the

consonants are signiﬁcantly longer when compared with VS3SG#V contexts (e.g. come limas). In

Granada, therefore, there is sll some mental unit at the end of 2SG and 3SG present indicave

forms which leaves phonec cues in sandhi contexts.

3.2.1.2 VS2SG#V vs. V#V: tú siempre comes ajos vs. él siempre come ajos

On the basis of the data and conclusion from the previous secon one would expect that in

VS2SG#V contexts there should be phonec cues for aspiraon in Granada but not in Seville,

since, in this city, S2SG corresponds to no mental element. This hypothesis is borne out by the

results of VS2SG#V vs. V#V contexts in Seville, in which there were no cues of aspiraon in the

former contexts which enrely resembled the laer contexts. Note, however, this was also the

result for these contexts in Granada; in no VS2SG#V context were there any cues for aspiraon

between the vowels. Therefore, in Granada, VS2SG#V contexts contrast with VS2SG#CV contexts,

since in the laer S2SG has robust phonec cues in the following consonants.

3.2.2 VSPLURAL#C/p/V vs. V#C/p/V contexts: dame dos ___s pa mí vs. dame un/una ____ pa mí

When <s> was a marker of PL the carrier phrase used meant that only the eﬀects on the

consonant /p/ could be analysed. In these contexts, as in the previous secon, the diﬀerences

between both cies were qualitavely diﬀerent. Cues for aspiraon before the consonant

appeared only in 3% of tokens from Seville but in 32% of tokens from Granada. Moreover, in <s>

contexts, the quality, voicing and length (both O and VOT) of the occlusives did not diﬀer from

the non‐<s> contexts for all but one speaker in Seville but did notably diﬀer in Granada for all

speakers. As with the VS2SG#OV contexts of the previous secon, VSPLURAL#OV contexts in

Granada paerned with VSOV contexts: they tended to be more voiceless, elongated and more

heavily aspirated. These diﬀerences and the contrasts with Seville are illustrated by the

spectrograms in (16), the percentages in the Table in (17) and the box‐plots in (18). Once again,

the impressionisc conclusion that in Seville the occlusives in VSPLURAL#OV contexts paern with

those in V#OV contexts whereas in Granada they are markedly diﬀerent (VSPLURAL#OV paerning

more with VSOV contexts) is conﬁrmed by the inferenal stascal analysis which compared

VSPLURAL#OV and V#OV contexts. The results were the following:

| 59

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

• In <s> contexts cues for aspiraon before the consonant appeared in 32% of cases in

Granada but only 3% of cases in Seville. Stascal analysis conﬁrmed that the diﬀerent

contexts did not reliably predict the presence of aspiraon in Seville (χ2 = 0.967, p= 0.325)

but did in Granada (χ2 = 7.025, p= 0.008)

• <s> contexts did not have any signiﬁcant eﬀect on the length of the following consonant in

Seville (χ2 = 1.515, p= 0.218) but did so in Granada (χ2 = 10.003, p= 0.002) where the

presence of <s> increased the length of the consonant by 49.11ms ± 7.3 (standard errors)

• <s> contexts did not have any signiﬁcant eﬀect on the type of consonant in Seville (χ2 =

1.478, p= 0.224) but did so in Granada (χ2 = 16.865, p= <0.000) where the presence of <s>

increased the number of voiceless pronunciaons.

(16) Spectrograms of VSPLURAL#CV vs. V#CV contexts. Note only the parts in square brackets

are represented in the spectrograms.

(17) Type of consonant in VSPLURAL#OV and V#OV sequences

| 60

Paul O'Neill

Variación 1(2) (2024), 18‐30

(18)

Boxplots of length of O and VOT according to speaker in the VSPLURAL#C/P/V (pl)

vs. V#C/

(sg.) contexts

The results of this secon align themselves closely, though not exactly, with those of the

previous secon, in that they suggest SPLURAL, as was the case with S2SG, corresponds to no

mental element in the city of Seville. However, upon close inspecon of the boxplots in (18)

above it becomes clear that speaker F2 paerns diﬀerently from other speakers: she consistently

has longer O and VOT in in VS2SG#OV contexts. This impression was also conﬁrmed by the

stascal analysis since the variable speaker in Seville was signiﬁcant, when speaker was

controlled for context. Also, the diﬀerence between the lengths of O and VOT in the diﬀerent

contexts for speaker F2 of Seville was conﬁrmed as being signiﬁcant in a paired t‐test (length of O

(p= 0.014) length of VOT (p= 0.001)). Witness also the diﬀerence between the occlusives in the

spectrograms below in (19) for idencal phrases produced by M1 and F2 of Seville; F2 displays an

| 61

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

aspirated voiceless occlusive when preceded by SPLURAL but M1 produces a closed approximant/

unreleased voiced stop, very similar to that produced in the V#CV context.

(19) Spectrograms of VSPLURAL#CV vs. V#CV contexts for two speakers of Seville

One, however, must not be hasty to conclude that all speakers in Seville, save F2, have lost

the morphological marker which corresponds to orthographic <s>, since close inspecon of the

data in (17) relang to the diﬀerent voicing qualies and manner of arculaon of the consonant

/p/ in the VS2SG#OV contexts vs their near minimal pairs (V#OV) reveal interesng tendencies.

Speciﬁcally, it seems that there is a tendency to produce /p/ as a voiced stop as opposed to an

approximant when proceeded by SPLURAL. These data are extremely interesng since it could

signal the ﬁnal days of a sound change SPLURAL > Ø whereby speakers can sll make a slight

disncon in their producon but it is doubul that this will be perceived by listeners since

intervocalic <p t k> are readily realised as both approximants and voiced stops in Andalusia and

the best cues for the sequence VS or V#S is usually a voiceless stop as in VSOV contexts.

3.2.2.1 VSPLURAL#V vs. V#V contexts: dame dos ___s ahora mí vs. dame un/una ____ ahora

Further evidence to support the living but moribund status of SPLURAL in Seville comes from

prevocalic external sandhi contexts (VSPLURAL#V, fresas ahora, cerezas ahora) as illustrated by the

data in (20). Here, despite there only being cues for aspiraon in just under a third of all phrases,

these percentages contrast with those from the last secon (tú siempre comes ajos vs. él siempre

come ajos) in which for VS2SG#V contexts there were no cues of aspiraon at all and these

contexts resembled V3SG#V contexts enrely. Note also that in the data below in (20) there are

clear individual diﬀerences with M1 displaying no tokens of aspiraon at all and F2 displaying the

most tokens (75%).

(20) Table showing the percentages of aspiraon in VSPLURAL#V contexts

| 62

Paul O'Neill

Variación 1(2) (2024), 18‐30

In Granada, as with S2SG in which there were phonec cues for /s/ on the following

consonants (e.g comes limas) but none in prevocalic posion (e.g. comes ajos), so too with

SPLURAL. For VSPLURAL#V contexts (peras ahora vs. pera ahora) there were no phonec cues for

SPLURAL in any of the tokens analysed, which contrast not only with the data from Seville but also

with the robust phonec cues for Granada when SPL was followed by a consonant (peras

pa[ra] mí).

3.2.3 VSDOS#CV vs. V#CV contexts: dame dos ___ pa mí vs. dame un/una ____ pa mí

The results of this secon contrast with those of the previous two secons. Within VSDOS#OV

sequences there were not only some cues of aspiraon before the consonant in both cies but

the occlusives in these sequences were diﬀerent in the quality, voicing, and length (both O and

POST) when compared with those in the V#OV contexts in both Seville and Granada. These facts

can be appreciated in the spectrograms in (21), the percentages in (22), and the box‐plots in (23).

Unlike the occlusives following S2SG which paerned with those in V#OV contexts, those

following SDOS show a greater tendency to be voiceless.

(21) Spectrogram of the sequence VSDOS#CV and V#CV

(22) Type of consonant in VSDOS#OV and V#OV sequences

| 63

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

(23) Boxplots of length of O and VOT according to speaker

The impressionisc conclusion from the ﬁgures above that in Seville the occlusives in

VSDOS#OV contexts do not paern with those in V#OV contexts but rather with VSOV contexts is

conﬁrmed by the inferenal stascal analysis which compared VSDOS#OV and V#OV contexts

(dame dos peras pa[ra] mí vs. dame una pera pa[ra] mí). The results were the following:

• In <s> contexts cues of aspiraon before the consonant appeared in 50% of <s> contexts in

Seville and 75% of cases in Granada. Stascal analysis conﬁrmed that the diﬀerent

contexts did reliably predict the presence of aspiraon in Seville (χ2 = 10.03, p= 0.002) and

in Granada.

• <s> contexts had a signiﬁcant eﬀect on the length of the following consonant in both Seville

and Granada.

• <s> contexts also did have a signiﬁcant eﬀect on the type of consonant in both Seville and

Granada, increasing the number of voiceless pronunciaons.

It is clear therefore that there is some mental element at the end of the word dos which

causes the following occlusive consonant to be longer and more heavily aspirated.

Regarding the other consonants in VSDOS#CVcontexts , the results are consistent with those of

the occlusives for both cies: they adhere more to paerns in VSCV contexts than those in VCV

contexts. The consonants are consistently longer in VSDOS#CV than in the minimal VS#CV pairs as

illustrated by the boxplots in (24) for Seville. Regarding the quality of the consonants, although

there were no cases of nasal vowels being visibly pronounced with breathy voice on the

spectrogram, there were modiﬁcaons in the approximant pronunciaon of laterals and spirants,

which were grouped together as approximants; there were even two instances of breathy voice

before the vowel which then was coarculated parally with the approximant. The data are

presented in the table in (25) for Seville only. For this city, I also carried out a mixed linear model

for the contexts VSDOS#CV and VS#CV. The results were that the phonological context was

signiﬁcant (χ2(1)= 17.643,p= 0.001) and speaker was signiﬁcant (χ2(1)= 7.8933, p= 0.019),

gender was not signiﬁcant. The manner in which the consonant was realised was not controlled

for since there were too many diﬀerent realisaons.

| 64

Paul O'Neill

Variación 1(2) (2024), 18‐30

(24) Boxplots of length of C according to speaker for Seville in the VSDOS#CV vs. V#CV (una)

contexts

(25) Comparison of the pronunciaons of lateral and spirant aproximants in VSDOS#CV vs.

V#CV contexts in the city of Seville

These data suggest that orthographic <s> sll corresponds phonecally to some type of

opening of the glos. The resulng increased airﬂow can then have an eﬀect on the following

consonant whose gesture is med at the same moment as the glos gesture, leading to none,

or very few cases of pre‐consonantal aspiraon.

| 65

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

3.2.3.2 VSDOS#V vs. V#V contexts: dame dos orejas/ojos vs. dame una oreja/ojo

In pre‐vocalic posion (VSDOS#V) aspiraon is robustly aested for Seville as illustrated in (26);

note, however, that percentages did not reach 100% in any speakers but there were more cases

of resyllabiﬁcaon (pronunciaon of [s] in dos ojos) than of no‐aspiraon.

(26) Table showing the percentages of aspiraon in VSDOS#V contexts

These data contrast with those from Granada in which there were no cues for aspiraon in

none of the phrases analysed. This result is consistent with the other cases of pre‐vocalic /s/ in

external sandhi contexts for Granada and contrasts with the cases of pre‐consonantal /s/ for

which there are robust phonec cues for /s/ on the following consonants.

3.3 Summary of results

In VSCV contexts, word‐internal orthographic <s> does not correspond to any phonec cues in

VSFV contexts. In VSOV contexts it usually realised as some type of voiced aspiraon before the

consonant (breathy voice or pre‐aspiraon on the occlusive), which is always accompanied by a

voiceless aspirated occlusive. This consonant tends to be elongated with respect to VOV

sequences in Granada and more heavily aspirated in Seville. Outside these sequences the

consonantal gesture of the following consonant is, with few excepons, med to occur at the

end of the vowel and the consonants are all elongated with respect to VCV sequences. In

western variees there are indicaons of a more open glos gesture which is coarculated with

the consonant; this coarculaon manifests itself as breathy voice on nasals and fricaon on

approximants. In VS#CV contexts, it is necessary to make a disncon between the diﬀerent

morphological types of <s>: S2SG, SPLURAL and SDOS. The ﬁrst of these (S2SG), on the basis of the

evidence in this arcle, cannot be considered to represent a mental element for speakers of

Seville since it is bere of any phonec cues. In Granada, however, the phonec cues are similar

to those in word‐internal posion (VSCV contexts) indicang that in VS2SG#CV contexts there

remains some mental element at the end of these words. Contrasvely, in pre‐vocalic contexts,

(comes ajos/orejas) there are no phonec cues for this S2SG in Granada. The same is true for

SPLURAL in Granada: VSPLURAL#V contexts (tartas ahora) show no cues for /s/ but VSPLURAL#CV

contexts do. These contexts paern with VSCV contexts and contrast with their minimal pairs in

V#CV (tartas pa[ra] mí vs tarta pa[ra] mí). In Seville, however, SPLURAL seems to be heading in the

same direcon as S2SG. That is, it is undergoing a sound change towards its complete eliminaon.

Indeed, only one speaker (F2) showed consistent and robust phonec cues for SPLURAL. Finally, for

SDOS, with the excepon of pre‐vocalic contexts in Granada (dos ajos/orejas), in both cies there

were considerable phonec cues for this element, especially in prevocalic contexts in Seville. In

preconsonantal contexts, the phonec cues in VSDOS#CV contexts were similar to those in word

| 66

Paul O'Neill

Variación 1(2) (2024), 18‐30

internal posion and diﬀerent from their near minimal pairs in V#CV (dos peras vs. una pera).

4. DISCUSSION

It is clear that the phonological system of Andalusian Spanish is being radically altered due to

the historical eﬀects of /s/ aspiraon and that there is much diatopic and inter‐speaker variaon

(Stevens and Harrington 2014). In word internal posion, this is producing a series of new

incipient phonemic disncons of which the most robust are elongated consonants (cine [sine] /

[θine] vs. cisne [sin:e] / [θin:e]; hila [ila] vs. isla [il:a]) and aspirated consonants (pata [pata] vs.

pasta [patha], caco [kako] vs. casco [kakho]). It is worth remembering that the ﬁeldwork for the

current study dates from 2005 and that, in this study, only 46% and 23% of tokens in VSOV

contexts in the cies of Seville and Cádiz respecvely showed no evidence of aspiraon before

the (VSOVNOGLOT pronunciaons, see (5)). Contrast this with Parrell’s (2012) study focussing on

speakers from these same cies and their wider autonomous communies in which 75%

(936/1239) of the VSOV tokens had no phonec cues of aspiraon before the consonant but

rather ‘aer’ the consonant as post‐aspiraon. This increase in VSOVNOGLOT pronunciaons is,

perhaps, indicave of how the sound change is progressing historically and how such new

pronunciaons are creang a new phonemic disncon.

As for the realisaons of word ﬁnal /s/ in external sandhi, it seems that in Seville only /s/

behaves diﬀerently depending on its morphological status. When /s/ is a marker of 2SG on verbs

there are no phonec cues for its presence in Seville. When /s/ is a marker of PLURAL on nouns,

the cues are extremely scant (with the excepon of one speaker) in Seville and will in the future

possibly disappear (if it has not already done so). However, when /s/ is not a morphological

marker but it forms part of lexeme, as is the case for the word dos, the phonec cues are robust.

The conclusion from these data could be that sound change is sensive to the morphological

environment, which would have interesng theorecal implicaons. I do not think that this is the

case, however, since there are parallels with the Andalusian data presented here and

phonological changes in the history of Italian which suggest that the phonological eﬀects of word

ﬁnal /s/ on the following consonants are becoming opaque and lexicalized.

The diachronic developments in Italian that menoned above are those which took place in

central and southern variees of Italo‐Romance and produced the phenomenon known as

raﬀorzamento, or raddoppiamento (fono)sintaco. This is a process, illustrated by the examples

in (27), whereby a lexicalised set of words produce geminaon or strengthening of the inial

consonants of the following words e.g. Italian. da casa ‘from home’ [dak:asa] vs. la casa ‘the

house’[lakasa] (for an overview see Wheeler and O’Neill (2019)).

(27) words which trigger raﬀorzamento (fono)sintaco in standard Italian.

a ‘to’: a lui ‘to him’ [al:ui]; a Napule ‘to Naples’ [an:apulǝ]

né ‘neither/nor’: né caldo né freddo ‘neither hot nor cold’ [ne k:aldo ne f:ɾeddo]

che: che dite? ‘what are you saying?’ [ke d:ite]

è: è vero ‘it is true’ [ɛ v:ero]

fa’: fammi un favore ‘do me a favour’

tre: tre cani ‘three dogs’ [tre k:ani]

The origin and historical evoluon of this phenomenon is complex (see Loporcaro (1997)) but

the incipient stages are essenally similar to the developments underway in Andalusian Spanish.

In late Lan, word ﬁnal consonants were totally assimilated to the following consonantal onset

| 67

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

Note that maintenance of word ﬁnal /s/ is not ensured and speaker M2 in Seville did not have any cues for S

DOS

in 30% of

tokens in VS

DOS

#V contexts (see (26)).

<sud die> for SUB DIE (Loporcaro, 1997, pp. 42, 121). The consequences of this assimilaon

process can be seen word internally in the history of Italian dialects where the outcome was

elongated consonants (DICTU(M) > deo ‘said’; RUPTU(M) > roo ‘broken’; FRIG(I)DU(M) >

freddo ‘cold’). In word ﬁnal posion, it is hypothesised that this same process was acve in all

pre‐consonantal external sandhi contexts, as is the case with the phonological eﬀects of /s/ in

the word dos in Seville and all cases of word ﬁnal /s/ in Granada. In Italian, the ﬁnal consonants

which triggered changes in the following words eventually came to be deleted. Again, there are

pararells with the Andalusian data since in Seville /s/ can be considered to have deleted when it

is a marker of 2SG in verbs it is currently being deleted when it is a maker of PLURAL in nouns. In

the relevant variees of Italo‐Romance, as illustrated by the data in (28), in certain monosyllabic

words (note that the words in (27) are all monosyllables, such words have a special status

historically, especially if stressed, see Wheeler and O’Neill (2019) for an overview) there are

traces of these lost consonants in the phonological eﬀects that these words have on following

word‐inial onset, making them geminate. Likewise, in the Spanish spoken in Seville, the <s> in

the monosyllabic word dos is triggering changes in following consonants but, unlike the Italian

data, it cannot be claimed that this consonant has been deleted yet.

(28) Historical summary of the development of raﬀorzamento (fono)sintaco in Italian.

Word ﬁnal /s/ is clealy undergoing a process of deleon in the Spanish spoken in Seville but it

is being maintained in the monosyllable dos and, I would hypothesise, in other monosyllables

(más, los, las, sus, mis, mes, les, es, das, vas)

. In the future, /s/ could be deleted in these words

but, like with what happened in Italian, one may be able to appreciate the eﬀects of this

historical consonant in the inial consonants of following words. However, unlike in Italian in

which the lexical items only trigger gemminaon. In Andalusian Spanish, the lexical items could

also make following occlusives post‐aspirated, and turn approximants into fricaves.

A future development of Andalusian Spanish, therefore, could be that the historical eﬀects

of /s/ aspiraon are only apparent (a) word internally via new phonemes (obis[p

]o ‘bishop’) and

(b) due to the eﬀects that a speciﬁc set of monosyllabic words have on consonant‐inial words

(dos [p

]ersonas ‘two people’). All other word‐ﬁnal contexts in which /s/ occurred, as could be

the case in the current Spanish spoken in Seville, would not trigger such changes in following

consonants (e.g. comes [p]an ‘you eat bread’ and importantes [p]ersonas ‘important people’).

Moreover, given that the plural arcles in Spanish are monosyllables, it could be hypothesised

that the morphological disncon between singular and plural is currently being transferred

from moribund noun ﬁnal /s/ to word inial consonants via the phonological eﬀects on the

following nouns of ﬁnal /s/ in the arcles los/las. That is, the marker of plurality might be in a

| 68

Paul O'Neill

Variación 1(2) (2024), 18‐30

process of being metathesised to the start of the word due to the eﬀects of external sandhi of

word ﬁnal /s/ in the arcles. Thus, the historical development of los toros ‘the bulls’ would be

the following [lohtoɾoh] > [lo

toɾo] > [lo

oɾo] > [lot

oɾo], whereby the disncon between the

singular el [t]oro and plural lo[t

]oros is expressed phonecally via a diﬀerent inial consonant,

producing a hypothecal situaon in which one one would have un toro ‘one bull’ [untoɾo] but

cinco toros ‘ﬁve bulls’ [siŋkotʰoɾo].

Anecdotally, around the city of Seville the word todos ‘everyone’ seems to be pronounced

with a more aspirated inial consonant than the corresponding singular form todo ‘everything’.

Also, Pons Rodriguéz (2019) in a discussion of the phenomenon of the pluralizaon of the

parcle que in exclamave and interrogave sentences (qués cosas ‘what things’ instead of qué

cosas), notes how in the tweets in which this form appears, orthographic <s> is used to signal

some type of aspiraon. Another interpretaon of the forms qués cosas, qués ganas, etc., which

could be applied in parallel to the existence of pluralised ques, could be that the orthographic

use of <s> in these speciﬁc contexts is a sign that my future hypothesis for Andalusian Spanish is

already underway. That is, that the marker of PLURAL on nouns has been transferred from the

end to the start of the word and is phonecally realised as a more aspirated/breathy consonant.

Evidence in favour of such an interpretaon comes from the tweet below in (29) in which the

addional orthographic <s> appended to the clic te seems to indicate a following aspirated

consonant. Likewise, the orthographic rendering qués, in the same tweet, is clearly singular and

therefore not a case of a pluralised que. The situaon could be that, just as with nouns in which

the disncon between singular and plural is being morphologized via an inial aspirated

consonant, so with verbs the same disncon is analogically being adopted to express the

disncon between 2SG and 3SG. A disncon which, as with PLURAL on nouns, was originally

expressed via a ﬁnal /s/. Thus, the orthographic forms tu ques queieres and tu tes calla would

represent the forms [kekʰjeɾes] and [t

ekʰaja], respecvely, and would contrast with que quiere

[kekjeɾe] and se calla [sekaja].

(29) Tweet taken from the study by Pons Rodriguéz (2019, p. 23):

@FaliGuerrez jajaja , tu ques quieres pelea ehhhhhhhh que si err negro te vacila tu tes calla

y lo asimila :P [Tuit de @LauritaBravo99, 29/12/2012, Sevilla‐Cádiz].

“@FaliGuerrez hahahah, what do you want? A ﬁght? Ehhhhhhh if the black man teases you,

you should shut up and take it”.

In the foregoing, much experimental evidence from Andalusian Spanish has been presented

showing how the historic eﬀects of /s/ aspiraon are radically changing the phonological system

of this variety of Spanish. It has also been suggested that there could be changes in progress

which are eﬀecng the morphological system. Most recent research on Andalusian /s/ aspiraon

has focussed mainly on its eﬀects on occlusives in word internal posion and on how these

occlusives are becoming post‐aspirated and, in some places, elongated. All types of consonants

in word internal posion and in external sandhi contexts have been analysed in the present

paper. The results from this arcle show just how rich a ﬁeld Andalusian dialectology is and its

potenal for exploring the mechanisms involved in sound change, phonologisaon and

morphologisaon.

| 69

/s/ aspiraon in Andalusian Spanish in word internal posion...

Variación 1(2) (2024), 44‐69

REFERENCES

Gerfen, C. (2002). Andalusian codas. Probus, 14, 247–77.

Loporcaro, M. (1997). L’origine del radoppiamento fonosintaco. Saggio di fonologia diacronica romanza.

Francke: Basel/Tübingen.

Moya Corral, J. A. (2007). Nocia de un sonido emergente: la africada dental procedente del grupo ‐st‐ en

Andalucía. Revista de Filología de la Universidad de La Laguna, 25, 457–65.

O’Neill, P. (2009). The eﬀect of s‐aspiraon on occlusives in Andalusian Spanish. Oxford University

Working Papers in Linguiscs. Philology & Phonecs, 12, 73‐86.

O’Neill, P. (2010). Variación y cambio en las consonantes oclusivas del español de Andalucía. Estudios de

Fonéca Experimental, XIX, 11–41.

Parrell, B. 2012. 'The role of gestural phasing in Western Andalusian Spanish aspiraon. Journal of

Phonecs, 40, 37–45.

Pons Rodriguéz, L. (2019). Nuevos plurales en la gramáca del español meridional: un paradigma en

formación. Revista Internacional De Lingüísca Iberoamericana, XVII 34(2), 15‐32.

Ruch, H. (2012). Aﬀricaon of /st/ in Western Andalusian Spanish: Variaon and change from a

sociophonec point of view. In S Calamai, C Celata and L Ciucci (Eds.), Proceedings of “Sociophonecs,

at the crossroads of speech variaon, processing and communicaon” (pp. 61‐64). Edizioni della

Normale: Pisa.

Ruch, H, and Harrington, J. (2014). Synchronic and diachronic factors in the change from pre‐aspiraon to

post‐aspiraon in Andalusian Spanish. Journal of Phonecs, 45, 12‐25.

Ruch, H. and Peters, S. (2016). On the Origin of Post‐Aspirated Stops: Producon and Percepon of /s/ +

Voiceless Stop Sequences in Andalusian Spanish. Laboratory Phonology: Journal Of The Associaon

For Laboratory Phonology, 7(1), 2. hps://www.journal‐labphon.org/arcle/id/6170/.

Stevens, M. and Harrington, J. (2014). The individual and the actuaon of sound chang. Loquens, 1.

Torrerira, F. (2007a). Coarculaon between aspirated‐s and voiceless stops in Spanish: An interdialectal

comparison. In N Sagarra and A. J. Toribio (Eds.), Selected proceedings of the 9th hispanic linguiscs

symposium (pp. 113‐120). Cascadilla Proceedings: Somerville, MA.

Torrerira, F. (2007b). Pre‐ and postaspirated stops in Andalusian Spanish. In P. Prieto, P Mascaró and M. J.

Solé (Eds.), Segmental and prosodic issues in Romance Phonology (pp. 67‐82) John Benjamins.

Torrerira, F. (2012). Invesgang the nature of aspirated stops in Western Andalusian Spanish. Journal of

the Internaonal Phonec Associaon, 42, 49–63.

Torrerira, F. and Ernestus, M. (2011). Realizaon of voiceless stops and vowels in conversaonal French

and Spanish. Laboratory Phonology, 2, 331–53.

Wheeler, M. and O’Neill, P. (2019). Sandhi phenomena. In M. Maiden and A. Ledgeway (Eds.), The

Cambridge Handbook of Romance Linguiscs (pp. 209‐233). CUP.

ACADEMIC AND PROFESSIONAL PROFILE

Paul O'Neill completed his undergraduate degree in Classics and Spanish at the University of

Oxford (Brasenose College). He then moved to Madrid to carry out the Spanish equivalent of a

MPhil in Lingüísca Teórica y sus Aplicaciones at the Instuto Universitario de Invesgación

Ortega y Gasset. He wrote his MPhil thesis on the phonecs of the Andalusian variety of Spanish.

He then returned to Oxford to complete a DPhil on historical morphology in Ibero‐Romance). He

is currently a Professor of Romance Philology at LMU, Munich. His research focuses on language

variaon and change.