Content uploaded by Ander Beristain
Author content
All content in this area was uploaded by Ander Beristain on Jun 17, 2024
Content may be subject to copyright.
Citation: Beristain, Ander. 2024. The
Role of (Re)Syllabification on
Coarticulatory Nasalization:
Aerodynamic Evidence from Spanish.
Languages 9: 219. https://doi.org/
10.3390/languages9060219
Academic Editor: Rebeka
Campos-Astorkiza
Received: 21 February 2024
Revised: 29 May 2024
Accepted: 13 June 2024
Published: 17 June 2024
Copyright: © 2024 by the author.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
languages
Article
The Role of (Re)Syllabification on Coarticulatory Nasalization:
Aerodynamic Evidence from Spanish
Ander Beristain
Department of Languages, Literatures and Cultures, Saint Louis University, St. Louis, MO 63108, USA;
ander.beristain@slu.edu
Abstract: Tautosyllabic segment sequences exhibit greater gestural overlap than heterosyllabic
ones. In Spanish, it is presumed that word-final consonants followed by a word-initial vowel
undergo resyllabification, and generative phonology assumes that canonical CV.CV# and derived
CV.C#V onsets are structurally identical. However, recent studies have not found evidence of this
structural similarity in the acoustics. The current goal is to investigate anticipatory and carryover
vowel nasalization patterns in tautosyllabic, heterosyllabic, and resyllabified segment sequences
in Spanish. Nine native speakers of Peninsular Spanish participated in a read-aloud task. Nasal
airflow data were extracted using pressure transducers connected to a vented mask. Each participant
produced forty target tokens with CV.CV# (control), CVN# (tautosyllabic), CV.NV# (heterosyllabic),
and CV.N#V (resyllabification) structures. Forty timepoints were obtained from each vowel to observe
airflow dynamics, resulting in a total of 25,200 datapoints analyzed. Regarding anticipatory vowel
nasalization, the CVN# sequence shows an earlier onset of nasalization, while CV.NV# and CV.N#V
sequences illustrate parallel patterns among them. Carryover vowel nasalization exhibited greater
nasal spreading than anticipatory nasalization, and vowels in CV.NV# and CV.N#Vstructures showed
symmetrical nasalization patterns. These results imply that syllable structure affects nasal gestural
overlap and that aerodynamic characteristics of vowels are unaffected across word boundaries.
Keywords: coarticulatory nasalization; resyllabification; aerodynamics; Spanish
1. Introduction
Traditionally, generative phonology assumes that, in connected speech in Spanish,
word-final consonants undergo resyllabification before a word that starts with a vowel,
e.g., un amigo [u.n#a.mi.
G
o] ‘a friend’ (Harris and Kaisse 1999) (henceforth, ‘C’: consonant,
‘V’: vowel, ‘N’: nasal consonant, ‘.’: syllable boundary, ‘#’: word boundary). The process
is understood as complete and a part of the phonology of Spanish. Nevertheless, recent
laboratory studies have questioned the status of ‘obligatory’ or ‘complete’ resyllabification
of C#V sequences in Spanish by providing evidence for acoustic and phonetic differences
between ‘canonical’ onsets, e.g., CV.CV#, and ‘derived’ (resyllabified) onsets, e.g., CV.C#V
(Bradley et al. 2022;Hualde and Prieto 2014;Jiménez-Bravo and Lahoz-Bengoechea 2023;
Strycharczuk and Kohlberger 2016).
To date, most of the studies investigating resyllabification have focused on the conso-
nantal segments, which were, in most cases, oral fricative sounds, disregarding how neigh-
boring vowels or other types of segments such as nasal structures might be affected by
(re)syllabification rules. Furthermore, research has shown that syllabic structure plays a role in
nasalization degree (Byrd et al. 2009;Cohn 1993); thus, tautosyllabic VN segments will show
greater gestural overlap than heterosyllabic V.N segments (see Krakow 1989,1993,1999).
Remarkably, the effect of (re)syllabification of nasal consonants on neighboring vowels
in Spanish has not been thoroughly investigated, especially via articulatory methods.
The present study seeks to understand the aerodynamics (Beristain 2022,2023a,2023b;
Cohn 1993;Huffman and Krakow 1993;Shosted 2009;Shosted et al. 2012;Solé1992;
Languages 2024,9, 219. https://doi.org/10.3390/languages9060219 https://www.mdpi.com/journal/languages
Languages 2024,9, 219 2 of 23
Stoakes et al. 2020) of the spread of anticipatory and carryover coarticulatory nasalization
across various syllabic contexts within word boundaries and across word junctures. The
study determines that tautosyllabic segments show greater articulatory overlap and that
carryover nasalization is unaffected by resyllabification, which provides evidence for
complete resyllabification.
The structure in the current article is as follows: first, the relevant literature on syllable
structure, resyllabification, and nasalization is provided; second, research questions and
hypotheses are presented; third, the methods are included; fourth, the results pertaining
to anticipatory and carryover nasalization are summarized, followed by the discussion
and conclusions.
2. Literature Review
2.1. Resyllabification Revisited
Resyllabification is a process by which a word-final coda delinks from its original
syllable structure and attaches to the following vowel-initial word as an onset. It should
be noted that this process is not universal across languages and/or dialects. For instance,
English has ambisyllabicity (Kahn 1976) or glottal stop insertion (Bissiri et al. 2011), German
has glottal stop insertion (Kohler 1994), and eastern Abruzzese dialects opt for voiced glottal
fricative [
H
] or velar fricative [
G
] as an empty onset repair mechanism (Passino et al. 2022,
p. 93). On the other hand, resyllabification is found in Spanish (Harris 1983), French
(Durand et al. 2011), and dialects of Occitan (Sauzet 2012), among others. Harris and
Kaisse (1999) explain that for a segment to be resyllabified, three stages need to happen:
(i) initial syllabification (no change); (ii) delinking (change initiates); and (iii) attach onset
(resyllabification occurs, change is complete). Below is Figure 1, which illustrates the
derivation of resyllabification for the word sequence un año ‘a year’ in Spanish (adapted
from Harris and Kaisse 1999, p. 137).
Languages 2024, 9, x FOR PEER REVIEW 2 of 24
Remarkably, the effect of (re)syllabification of nasal consonants on neighboring vow-
els in Spanish has not been thoroughly investigated, especially via articulatory methods.
The present study seeks to understand the aerodynamics (Beristain 2022, 2023a, 2023b;
Cohn 1993; Huffman and Krakow 1993; Shosted 2009; Shosted et al. 2012; Solé 1992;
Stoakes et al. 2020) of the spread of anticipatory and carryover coarticulatory nasalization
across various syllabic contexts within word boundaries and across word junctures. The
study determines that tautosyllabic segments show greater articulatory overlap and that
carryover nasalization is unaffected by resyllabification, which provides evidence for
complete resyllabification.
The structure in the current article is as follows: first, the relevant literature on sylla-
ble structure, resyllabification, and nasalization is provided; second, research questions
and hypotheses are presented; third, the methods are included; fourth, the results pertain-
ing to anticipatory and carryover nasalization are summarized, followed by the discussion
and conclusions.
2. Literature Review
2.1. Resyllabification Revisited
Resyllabification is a process by which a word-final coda delinks from its original
syllable structure and aaches to the following vowel-initial word as an onset. It should
be noted that this process is not universal across languages and/or dialects. For instance,
English has ambisyllabicity (Kahn 1976) or gloal stop insertion (Bissiri et al. 2011), Ger-
man has gloal stop insertion (Kohler 1994), and eastern Abruzzese dialects opt for voiced
gloal fricative [ɦ] or velar fricative [ɣ] as an empty onset repair mechanism (Passino et al.
2022, p. 93). On the other hand, resyllabification is found in Spanish (Harris 1983), French
(Durand et al. 2011), and dialects of Occitan (Sauzet 2012), among others. Harris and
Kaisse (1999) explain that for a segment to be resyllabified, three stages need to happen:
(i) initial syllabification (no change); (ii) delinking (change initiates); and (iii) aach onset
(resyllabification occurs, change is complete). Below is Figure 1, which illustrates the der-
ivation of resyllabification for the word sequence un año ‘a year’ in Spanish (adapted from
Harris and Kaisse 1999, p. 137).
Figure 1. Derivation showing (re)syllabification in word sequence un año ‘a year’ in Spanish [where
(i) initial syllabification; (ii) delinking; (iii) aach onset (resyllabification)].
In this view, resyllabified codas should behave like onsets. Remarkably, weakening
phenomena that affect word-internal coda consonants may also affect word-final conso-
nants, even if they are resyllabified as onsets preceding a vowel in the following word.
These cases have been modeled using either rule ordering (Harris 1983) or constraint rank-
ing within optimality theory (Colina 1995, 1997, 2009) approaches. In Spanish varieties
where /s/ reduces to [h] in the coda (but not in word-internal position), word-final /s/ is
aspirated before resyllabification occurs in Vs#V structures, suggesting that a sequence
such as los amigos /los amigos/ ‘the friends’ would undergo [loh. a.mi.ɣoh → lo.ha.mi.ɣoh]
versus VsV structures, e.g., losa /losa/ ‘slab’ → *[lo.ha] (Colina 2002; Harris 1983; Kaisse
1999). Nevertheless, as pointed out and illustrated by a reviewer, in more conservative
Figure 1. Derivation showing (re)syllabification in word sequence un año ‘a year’ in Spanish [where
(i) initial syllabification; (ii) delinking; (iii) attach onset (resyllabification)].
In this view, resyllabified codas should behave like onsets. Remarkably, weakening
phenomena that affect word-internal coda consonants may also affect word-final conso-
nants, even if they are resyllabified as onsets preceding a vowel in the following word.
These cases have been modeled using either rule ordering (Harris 1983) or constraint rank-
ing within optimality theory (Colina 1995,1997,2009) approaches. In Spanish varieties
where /s/ reduces to [h] in the coda (but not in word-internal position), word-final /s/
is aspirated before resyllabification occurs in Vs#V structures, suggesting that a sequence
such as los amigos /los amigos/ ‘the friends’ would undergo [loh. a.mi.
G
oh
→
lo.ha.mi.
G
oh]
versus VsV structures, e.g., losa /losa/ ‘slab’
→
*[lo.ha] (Colina 2002;Harris 1983;Kaisse
1999). Nevertheless, as pointed out and illustrated by a reviewer, in more conservative
varieties, the sibilant [s] is retained. Brown (2008) states that lower levels of word-final
[s] retention correlate with high-frequency words, unless word-strings are taken into con-
sideration and the following word starts with a stressed vowel, in which case the relation
is inversed. For instance, Brown (2008, pp. 200–1) indicates that, because of the higher
frequency of the word combination dos años ‘two years’, this word-string is stored as a
single unit in memory, i.e., /dosa
ñ
os/. This results in word-final /s/ being reanalyzed as
Languages 2024,9, 219 3 of 23
word-medial, and being less likely to undergo reduction. On the other hand, the author
points out that a low-frequency word-string such as dos asnos ‘two donkeys’ is stored as
two separate words, i.e., /dos asnos/. As such, the word-final /s/ is being processed as
word-final, and it is more likely to be reduced (Brown 2008, pp. 201–2). As far as nasal
segments are concerned, in velarizing dialects, word-final /n/ is velarized to [
N
] in Vn#V
structures, e.g., van a casa /ban a kasa/ ‘they go home’ we find [ba
N
. a. ka.sa
→
ba.
N
a.ka.sa]
versus in VnV structures, e.g., vana /bana/ ‘vain’
→
*[ba.
N
a] (Harris 1983;Robinson 2012).
The generative literature assumes that the syllabic structure of the resyllabified C#V
segments (also termed ‘derived onsets’) and ‘canonical’ CV onsets are identical. From the
perspective of connected speech, such an assumption could be well received. However,
the postulate of structure similarity has been questioned, as will be shown in the next
paragraph. It should be noted that the number of studies that have explored word-internal
and across-word syllabic structure implementing articulatory approaches is scarce (see
Byrd et al. 2009;Krakow 1999, and references therein), even though those approaches
provide a more reliable and less ambiguous description of linguistic patterns and speech
gestures than acoustic studies.
In recent years, there has been an effort by phoneticians to identify the acoustic
correlates of resyllabified onsets. The language under study in this paper is Spanish,
which shows phonotactic restrictions as to which consonants can appear in the coda
position, those being /n, l,
R
, d, s,
θ
/ (Hualde 2014, p. 62). Most recent studies on the
acoustic correlates of resyllabified consonantal segments have focused on the sibilant
fricative /s/. As will be explained, word-final /s/ is shorter and more voiced in the
coda position; as such, it is expected that resyllabified onsets will have longer durations.
Hualde and Prieto (2014) analyzed spontaneous acoustic data of /s/ in Madrid Spanish
and investigated uninterrupted voicing, voicing frames, and duration of that segment
in word-initial, medial, and final positions. Their results showed that intervocalic word-
final position (i.e., resyllabification) led to higher rates of fully voiced /s/ and shorter
durations, indicating differences from ‘canonical’ conditions (see Torreira and Ernestus
2012 for similar results). Furthermore, Strycharczuk and Kohlberger (2016) analyzed the
fricative /s/ in northern and central Peninsular Spanish and compared the effects of the
position within the syllable and word on the acoustic properties of the segment. They found
that word-final, derived onset consonants were shorter than canonical ones. Strycharczuk
and Kohlberger (2016) is among the few studies that also provided an acoustic description
of neighboring vowels. Unlike for the duration of /s/, the researchers found evidence for
complete resyllabification in terms of vowel durations. Strycharczuk and Kohlberger (2016,
p. 11) encountered that the duration of vowels was not significantly different between word-
initial onsets and derived onsets (the latter were 0.09 ms longer) but it was significantly
greater compared to word-medial codas (the latter were 17.28 ms shorter) and word-medial
onsets (the latter were 11.43 ms shorter). Moreover, Jiménez-Bravo and Lahoz-Bengoechea
(2023) conducted an acoustic study where they compared canonical /s, n, l/ onsets, e.g.,
vende naves ‘(s/he) sells ships’ and derived ones venden aves ‘(they) sell birds’. Similar to
previous studies, their results illustrate that derived (resyllabified) onsets were shorter than
canonical ones in duration, respectively; /s/: 97.6 vs. 90.6 ms, /n/: 53.1 vs. 50.2 ms, and
/l/: 63.5 vs. 58.7 ms, which provides additional evidence for incomplete resyllabification
patterns in Spanish.
In the case of word-final /n/ in Spanish, an allophonic variation in certain varieties
is its velarization, i.e., [
N
] (see Bongiovanni 2021). The velarization of word final /n/
has been attested in the north-west and south of Spain, the Canary Islands, Caribbean
varieties, and certain Spanish varieties in South America (Hualde 2014, p. 174). Robinson
(2012) provided impressionistic data from Ecuadorian Spanish, specifically from Quito and
Cuenca, where participants were asked to resyllabify words that contained a final /n/ in a
velarizing context across word junctures, and all participants exhibited a significant pause
between words. This would indicate that resyllabification rules might not be applicable
Languages 2024,9, 219 4 of 23
universally in Spanish, as the ‘expected’ syllabic production of the word was misaligned in
connected speech.
2.2. Syllable Structure and Vocalic Nasalization
Coarticulatory vowel nasalization is the gestural coordination between a vowel (V)
next to a nasal consonant (N). The effects of coarticulation can spread from left to right
(carryover), e.g., no /no/ [nõ] ‘no’ in Spanish, or right to left (anticipatory), e.g., en [ẽn]
‘in’ in Spanish. The outcome in both conditions is gestural overlap with partial phonetic
nasalization of the vowel. Cross-linguistic differences have been found with regard to
the degree of gestural overlap across carryover and anticipatory, and within each type of
process (see Beristain 2023a,2023b;Clumeck 1976;Goodin-Mayeda 2016;Martínez 2021).
Previous studies have analyzed the effect of syllable structure on the nasalization
of the vowel (Beristain 2022;Byrd et al. 2009), and the general resolution that they have
found is that nasalization develops earlier in vowels in tautosyllabic VN sequences than in
heterosyllabic V.N ones because of gestural timing differences (Krakow 1989,1999). For
instance, Diakoumakou (2005) analyzed coarticulatory vowel nasalization in Modern Greek
and compared it to Spanish, Chinese, Japanese, Thai, Ikalanga, French, Italian, American
English, Hindi, and Bengali. As far as Greek is concerned, she provides acoustic evidence
to point out that the differences found in the nasalization patterns of vowels in tautosyllabic
or heterosyllabic environments is due to different articulatory patterns found in syllable-
initial and syllable-final contexts, with the latter having a lower velum position than the
former. Diakoumakou (2009) shows that only the final 17% of the vowel was nasalized
in heterosyllabic V.N sequences, whereas 33% of the vowel was nasalized in tautosyllabic
VN ones. Similarly, Cohn (1990) finds that the onset of nasalization in tautosyllabic VN
sequences is earlier than in those of heterosyllabic V.N ones in French. Moreover, Krakow
(1999, p. 27) analyzed the effect of syllable structure on nasalization patterns in English,
comparing word sets such as see more vs. seam ore and pa made vs. palm aid (i.e., CV#NV vs.
CVN#V). She used the Velotrace (Horiguchi and Bell-Berti 1987), which collected velum
raising and lowering movements with an LED attached to the lower lip with the Selspot
System to capture labial aperture and closure during the bilabial nasal /m/ and contiguous
vowels. Similarly to Cohn (1990), Krakow found that, in carefully read speech, tautosyllabic
VN sequences show an earlier onset of nasalization than heterosyllabic V.N ones. Under
similar contexts, Lahoz-Bengoechea and Jiménez-Bravo (forthcoming) compared the degree
of nasalization in VN#V and V#NV sequences using Nasalization from Acoustic Features
(NAF) measurements (see Carignan 2021 for a methodological overview). Their acoustic
data came from 19 individuals from Spain whose first language was Spanish. The authors
found no significant differences in the degree of nasal coarticulation between the two
contexts: VN#V: 0.714; V#NV: 0.713. These results provide evidence in support of complete
resyllabification in connected speech.
We may surmise that an earlier onset of nasalization in tautosyllabic segments is
correlated with the development of nasal vowels in Romance languages. As Sampson (1999,
p. 35) points out, “the process of vowel nasalization has often been significantly affected by
whether the conditioning nasal is tautosyllabic or heterosyllabic, reflecting the importance
of the syllable as a structural unit in the diachronic and synchronic phonological patterning
of Romance.” Notice how vowels in VN tautosyllabic sequences developed and eventually
became nasal vowels in some languages, e.g., BONU ‘good, masc.’ > [bon] > [bõn] > [bõ] (in
Modern French), whereas vowels in heterosyllabic sequences showed more coarticulatory
resistance, in some cases even leading to denasalization, e.g., BO NA ‘good, fem.’ > [bona] in
Northern Italian (Hajek 1997, p. 10).
Diakoumakou (2005) points out that languages with a tendency for open syllables show
greater carryover nasalization, while languages with a preference for close syllables have
more extensive anticipatory nasalization. She conducted an acoustic study investigating
the temporal extent of vowel nasalization in Modern Greek. She obtained data from six
native speakers and found that the temporal extent of nasalization was 27 ms long in the
Languages 2024,9, 219 5 of 23
heterosyllabic anticipatory condition, 48 ms long for the tautosyllabic anticipatory condition,
and 70 ms for the carryover condition. Diakoumakou discusses these results, comparing them
to what has been found in other languages, and explains that the tendency for open syllables,
such as in Greek, Spanish, or Italian, seems to be conducive for greater carryover nasalization.
On the other hand, she mentions that languages that show a preference for closed syllables
show greater anticipatory nasalization (see results for English in Krakow (1999)).
When comparing anticipatory versus carryover nasalization patterns, Cohn (1990) pro-
vides airflow contours for various contexts that include tautosyllabic VN and heterosyllabic
V.NV and V#NV sequences within a word and across word junctures in French. Her results
indicate that there is greater carryover nasalization (e.g., nez ‘nose’, p. 123) than anticipatory
nasalization (e.g., bonne ‘good’, p. 97). Similar aerodynamic results were found by Delvaux
et al. (2009). In Cohn (1990), airflow contours in derived onsets, e.g., bonne ode ‘good ode’
(p. 123) vs. canonical onsets, e.g., beau nez ‘beautiful nose’ (p. 101) do not exhibit significant
differences in French. Both conditions show a cline-like rise that appears late in the vocalic
segment and then a rapid drop after the offset of the nasal consonant, which is followed
by a constant and smooth transition throughout the nasalized segment. This indicates
that articulatory correlates are similar across heterosyllabic V#NV and resyllabification
V.N#V structures.
3. Research Questions
A vast amount of research about resyllabification has focused on the properties of the
consonant that changes its syllabic linking. The features associated with such consonants
were usually [+oral][-nasal]. Evidence has been provided for sub-phonemic differences
between canonical and derived onsets (see Bradley et al. 2022). The role that resyllabification
may have on the vowels surrounding the resyllabified consonant has been less investigated,
especially in the context of coarticulatory nasalization. Studies on gestural timing and
nasalization point out that the syllable structure of contiguous segments plays a significant
role, as tautosyllabic segments will show greater gestural overlap than heterosyllabic ones
(Krakow 1989;Byrd et al. 2009). Moreover, Diakoumakou (2009) and Delvaux et al. (2009)
state that languages that show a tendency towards open syllables usually exhibit greater
carryover coarticulatory nasalization than anticipatory nasalization.
Based on the previous findings, the specific research questions (RQs) and hypotheses
that the present study will investigate are as follows:
•
RQ 1: Are there differences between nasal airflow contours in the degree of phonetic
implementation of carryover and anticipatory coarticulatory nasalization in Spanish?
Hypothesis 1: Yes. Considering that Spanish shows a tendency for open syllables (Hualde 2014,
p. 59), according to Cohn (1990), Diakoumakou (2009), and Delvaux et al. (2009), it is expected that
carryover vocalic nasalization should show greater nasal spreading than anticipatory nasalization.
•
RQ 2: Does syllable structure play a role in the degree of anticipatory coarticulatory
nasality in Spanish?
Hypothesis 2: Yes. Based on Krakow (1989,1999) and Cohn (1990), it is expected that vowels
in tautosyllabic CVN# segment sequences will show an earlier onset of nasalization than vowels in
canonical CV.NV# and derived CV.N#V heterosyllabic segment sequences. Furthermore, vowels in
heterosyllabic (canonical and derived) contexts should exhibit similar anticipatory nasalization patterns.
•
RQ 3: Do vowels across word junctures (resyllabification) show parallel carryover
nasalization patterns as those within word boundaries?
Hypothesis 3: Yes. According to derivational rules of Spanish phonology and studies such as
Cohn (1990) and Lahoz-Bengoechea and Jiménez-Bravo (forthcoming), it is expected that vocalic
nasalization patterns in resyllabification CV.N#V contexts should present symmetrical patterns to
those in heterosyllabic CV.NV# sequences.
Languages 2024,9, 219 6 of 23
4. Methodology
4.1. Participants
Nine native speakers of Northern Peninsular Spanish (7F, 2M) participated in this
experiment. Their average age was 26 (age range = 23–29). All the participants were
graduate students at a US university when the experiment took place. They were originally
from the Basque Country, spoke Spanish as a native language, and had different levels of
proficiency in Basque. There is no reason to believe that the internal structure of Basque
could interfere with Spanish nasalization patterns, because, firstly, both Basque and Spanish
show resyllabification (Hualde and Ortiz de Urbina 2003;Hualde 2014, respectively);
secondly, there are no phonologically nasal vowels in the Basque varieties spoken by the
participants (Central and Western), nor is there direct contact with the French language
(Zuazo 2014); and, thirdly, none of the participants were proficient in any language that
included phonologically nasal vowels. On the other hand, strictly speaking, the evidence
presented here describes the Spanish variety of the Basque Country. In principle, there
could be dialectal differences in this respect.
4.2. Stimuli
The target tokens were adapted from Beristain (2022, p. 50) and fell under the following
four different conditions: (1) CV.CV, tautosyllabic oral C and oral V sequences (as oral
control tokens); (2) CVN#, tautosyllabic nasalized V and contiguous coda N; (3) CV.NV#,
heterosyllabic nasalized V and canonical onset N, as well as a proceeding nasalized V; and
(4) CV.N#V, derived onset N, with preceding and proceeding nasalized Vs (resyllabification
context). The first vowel of each word was each of the vowels in Spanish, /i, e, a, o, u/,
and the second vowel in the nasal structures was always /a/. The wordlist can be found in
Table 1.
Table 1. Target tokens.
CV.CV
(Oral Control)
CVN#
(Tautosyllabic)
CV.NV#
(Heterosyllabic)
CVN#V
(Resyllabification)
/i/ tita
‘aunt’
patín
‘rollerblade’
patina
‘s/he rollerblades’
patín atado
‘tied rollerblade’
/e/ cateto
‘ignorant’
ten
‘you have, IM P.’
tena
‘timber’
ten atado
‘have (it) tied’
/a/ tato
‘little brother’
tan
‘so’
gitana
‘gypsy’
tan atado
‘so tied’
/o/ pitote
‘fuss’
botón
‘button’
botona
‘s/he buttons’
botón atado
‘tied button’
/u/ batuta
‘baton’
atún
‘tuna’
gatuna
‘cat-like’
atún atado
‘tied tuna’
As can be noted, the stress patterns across words were uniform: the first vowel (V1)
was always stressed, and the second vowel (V2), in the CV.NV# and CV.N#Vcontexts, was
an unstressed /a/. This was undertaken to normalize the data because research has shown
that differences in stress may lead to varying degrees of nasalization (Byrd et al. 2009;Cohn
1990;Krakow 1989,1999).
The list was presented to participants before the experiment started, and the meaning
of the words was explained in case they were not familiar with it, which did not happen
in the current experiment. Target words were included in the carrier phrase Digo TARGET
ligeramente ‘I say TARGET softly’.
4.3. Equipment
The aerodynamic data were collected via a vented Scicon OM-2 oral mask (Scicon R&D,
Inc., Beverly Hills, CA, USA) that was connected to two TSD160A pressure transducers
(operational pressure 72.5 cm H
2
O; Biopac Systems, Goleta, CA, USA). Simultaneous
Languages 2024,9, 219 7 of 23
acoustic data were obtained to facilitate the segmentation process. For that, the signal
was preamplified using a Grace M101 microphone preamplifier (Grace Designs, Boulder,
CO, USA) and digitized at 2 kHz, the highest allowed in the Biopac System (BIOPAC
2020). Participants wore an AKG C-520 head-mounted microphone (Harman International,
Stamford, CT, USA) that was located approximately 3 cm (1 inch) away from their mouths.
For a more detailed report about the equipment, see Beristain (2022,2023a).
4.4. Data Collection and Procedure
The author of the current study calibrated the equipment manually before every
session by utilizing an AFTA6A Calibration syringe of 600 mL (Beristain 2023a,2023b) and
annotating the correction value of the signal for every word uttered and the surface area
of the air volume expelled by the syringe. Those numbers were inserted in a calibration
formula that was applied to the raw data of each participant. The conversion equation used
is presented in (1) (Beristain 2023a,2023b;Shosted 2009), as follows:
s′=filtfilt (s×v
Rs+c)(1)
As cited in Beristain (2023b, p. 9): “s
′
is the new, resulting signal; filtfilt is the MATLAB
function; sis the original, unaltered signal;
R
means integration in order to calculate the
area under the signal curve; vis the total volume of the syringe (= 600 mL), and cis the
correction number (specific to every word)”.
Data were collected in a sound-attenuating booth inside a phonetics laboratory of
a university in the USA. In order to avoid as much linguistic co-activation as possible,
the author of the current study (who is a native speaker of Northern Peninsular Spanish)
communicated with the participants in Spanish at all times. Participants signed a written
consent before starting the experiment.1
Participants were then informed about the experiment and shown how to operate the
equipment. Before every experimental procedure, a trial session took place to ensure no air
leakage was present. None of the current participants showed anomalous results in their
Spanish trial productions. Once the experiment began, participants held the mask holding
the handle attached to it. Recordings were stopped after every minute, in order to alleviate
any possible discomfort.
The software that was utilized to obtain the data was AcqKnowledge (version 3.9.1).
Three different and simultaneous signals were collected: (1) nasal airflow, (2) oral airflow,
and (3) audio (acoustics). The current study will only present nasal airflow data. Future
studies will provide a multidimensional analysis presenting a combination of several signal
types. MATLAB (version 2020a; MATLAB 2020) was used to convert the AcqKnowledge
files into .wav files that were readable in Praat (Boersma and Weenink 2020).
4.5. Data Annotation
In Praat, data were segmented and annotated by inspecting the nasal airflow, oral
airflow, and audio signals. The onset of the first vowel (V1) was located after the visible
burst of air of /t/ decreased and a more periodic signal was present both in the oral airflow
and audio signals, indicating vowel periodicity (see Beristain 2023a,2023b;Delvaux et al.
2009). The amplitude of the spectrogram was also a clear cue for distinguishing the vowel
from contiguous consonants, both oral and nasal. The cues used to segment oral second
vowels (V2) were: the oral air burst of /t/, the more periodic signal apparent in oral and
audio channels, and the greater amplitude. The cues for nasalized V2s were as follows: the
onset of oral airflow following the decrease in nasal airflow (indicating the N-V transition)
and the greater amplitude present in the vowel. The offset of nasalized V2s was located
when the amplitude decreased significantly from the V-/C/ transition. Figure 2presents
two examples with segmentations that include a fully oral CVCV token, batuta ‘baton’, and
a CV.N#V (resyllabification) one, atún atado ‘tied tuna’. Notice how the signal in Channel 1
Languages 2024,9, 219 8 of 23
(nasal airflow) is virtually flat, showing essentially zero nasal airflow in batuta (N: nasal; O:
oral; A: audio/acoustic).
Languages 2024, 9, x FOR PEER REVIEW 8 of 24
In Praat, data were segmented and annotated by inspecting the nasal airflow, oral
airflow, and audio signals. The onset of the first vowel (V1) was located after the visible
burst of air of /t/ decreased and a more periodic signal was present both in the oral airflow
and audio signals, indicating vowel periodicity (see Beristain 2023a, 2023b; Delvaux et al.
2009). The amplitude of the spectrogram was also a clear cue for distinguishing the vowel
from contiguous consonants, both oral and nasal. The cues used to segment oral second
vowels (V2) were: the oral air burst of /t/, the more periodic signal apparent in oral and
audio channels, and the greater amplitude. The cues for nasalized V2s were as follows:
the onset of oral airflow following the decrease in nasal airflow (indicating the N-V tran-
sition) and the greater amplitude present in the vowel. The offset of nasalized V2s was
located when the amplitude decreased significantly from the V-/t/ transition. Figure 2 pre-
sents two examples with segmentations that include a fully oral CVCV token, batuta ‘ba-
ton’, and a CV.N#V (resyllabification) one, atún atado ‘tied tuna’. Notice how the signal in
Channel 1 (nasal airflow) is virtually flat, showing essentially zero nasal airflow in batuta
(N: nasal; O: oral; A: audio/acoustic).
(a)
(b)
Figure 2. Sample segmentation of CVCV and CV.N#V sequences: (a) batuta ‘baton’; (b) atún atado
‘tied tuna’; (Speaker #21, Rep. #1).
4.6. Data Analysis
Taking into consideration the goal of this experiment, the time dynamics of nasal air-
flow in V1 and V2 were analyzed. Each segment was further divided into 40 equidistant
timepoints. The total number of datapoints gathered and submied for the statistical anal-
ysis is presented in Table 2.
Figure 2. Sample segmentation of CVCV and CV.N#V sequences: (a)batuta ‘baton’; (b)atún atado
‘tied tuna’; (Speaker #21, Rep. #1).
4.6. Data Analysis
Taking into consideration the goal of this experiment, the time dynamics of nasal
airflow in V1 and V2 were analyzed. Each segment was further divided into 40 equidistant
timepoints. The total number of datapoints gathered and submitted for the statistical
analysis is presented in Table 2.
Table 2. Number of tokens analyzed.
Number
Anticipatory nasalization
(V1)
9 speakers ×5 vowel conditions [i, e, a, o, u] ×40 time-points ×4 syllable conditions
[CVCV; CVN#; CV.NV#; CV.N#V] ×2 repetitions ×= 14,400
Carryover nasalization
(V2)
9 speakers ×5 vowel conditions [i, e, a, o, u] ×40 time-points ×3 syllable conditions
[CVCV; CV.NV#; CV.N#V] ×2 repetitions = 10,800
Total: 25,200 datapoints
Languages 2024,9, 219 9 of 23
To statistically analyze the time dynamics of nasal airflow, generalized additive mixed
models (GAMMs) were run in R (R Core Team 2020) and RStudio (R Studio Team 2020)
using the mgcv, v. 1.8.31 package (Wood 2011). Data visualization was conducted using the
itsadug package, v. 2.3 (Van Rij et al. 2017). The optimal statistical models for V1 and V2
were chosen by inspecting Akaike Information Criterion (AIC) scores (Akaike 1974). Such
models are listed below (see Appendix Afor further details).
•
Optimal model for V1:
Context * Vowel + Sex + s(NormTime, by = interaction
(Context, Vowel)) + s(NormTime, by = Sex) + s(NormTime, Speaker, by =
interaction(Context, Vowel), bs = "fs", m = 1) + s(Word, bs = "re")
•
Optimal model for V2:
Context + Sex + s(NormTime, by = Context) + s(NormTime,
by = Sex) + s(NormTime, Speaker, by = Context, bs ="fs", m = 1) + s(Word,
bs = "re")
As shown, the statistical model to analyze anticipatory vowel nasalization (V1) in-
cluded an interaction between CONT EX T (CVCV; CVN#; CV.NV#; CV.N#V) and VOWEL
(i; e; a; o; u) as well as the fixed factor of SEX (male; female). A smooth function was
included through NO RM TIM E (from 0 to 1 with 0.025 increments). The statistical model to
analyze carryover nasalization (V2) was similar to that of V1 without VOWEL in it. This is
because V2 was always /a/ in CV.NV# and CV.N#Vtokens.
5. Results
This section is divided into two parts: (1) V1 (anticipatory vowel nasalization) and
(2) V2 (carryover vowel nasalization). As is customary with GAMM reports, figures will
present contours and difference curves. The x-axis of each figure indicates normalized time
(NormTime), which analyzes nasal airflow (in L/s) across 40 equidistant timepoints from
the onset to the offset of the vocalic segment. The darker line within the contour indicates
the mean values, while the lighter color indicates confidence intervals, i.e., variability in
the data. The nasalized target and oral control contexts are presented. Note that even
the oral control context shows a certain degree of nasal airflow, as air will inevitably be
exhaled through both oral and nasal apertures when producing speech. Most importantly,
the quantity and shape of nasal airflow in the oral control context is significantly lesser and
stable, as opposed to rising contours in sequences that contain nasal structures.
The way in which statistical significances are provided and onset and offset of nasal-
ization are calculated is by means of statistical pair-wise comparisons, retrieved via the
plot_diff() function under the itsadug package
2
. Oral control segments and nasalized target
segments are compared, and the difference curves of their nasal airflow are analyzed, thus
pointing out and locating the onset and offset of the nasal gesture in the time window of the
whole segment (Beristain 2023a). The area in which the two airflow curves are significantly
different is the “time region of significance”, and that will be delimited by red dotted lines
in the difference curve figures. Those areas are what will be reported in the results section.
The full statistical model output for V1 can be found in Appendix B, while that for V2 can
be found in Appendix C.
5.1. Anticipatory Vowel Nasalization
The results for V1 nasal airflow curves are illustrated in Figure 3.
Languages 2024,9, 219 10 of 23
Languages 2024, 9, x FOR PEER REVIEW 10 of 24
Figure 3. Nasal airflow GAMM curves of V1 in CVCV (oral control), CVN# (tautosyllabic), CV.NV#
(heterosyllabic), and CV.N#V (resyllabification)3.
As can be seen, the nasal airflow contour in the oral control (CVCV) context is stable
and virtually non-existent. The tautosyllabic (CVN#) sequence shows the greatest amount
of nasal airflow for V1, followed by resyllabification (CV.N#V), and then the heterosyl-
labic (CV.NV#) one. The pair-wise comparisons between each nasalized context and the
oral control exhibit differences. Such differences are shown and described in Figures 4a,b,
5a,b and 6a,b.4
Figure 4a compares anticipatory nasalization contours in the oral control CVCV
structure and the tautosyllabic CVN# one and illustrates a minimal and stable develop-
ment of nasal airflow contour for CVCV, while a rise is seen towards to end of the segment
for CVN#. This is to be expected, as the segment to follow V1 in CVN# is a nasal conso-
nant, and as such, due to anticipatory coarticulatory nasalization, the onset of the nasal
gesture that applies to the nasal consonant will begin before the offset of the previous
vowel, i.e., V1. As shown in Figure 4b, the time region of significance starts at 0.86. Aero-
dynamic data for Spanish has shown delayed onsets of nasalization, as opposed to other
languages like English (Beristain 2023a).
(a)
Figure 3. Nasal airflow GAMM curves of V1 in CVCV (oral control), CVN# (tautosyllabic), CV.NV#
(heterosyllabic), and CV.N#V (resyllabification)3.
As can be seen, the nasal airflow contour in the oral control (CVCV) context is stable
and virtually non-existent. The tautosyllabic (CVN#) sequence shows the greatest amount
of nasal airflow for V1, followed by resyllabification (CV.N#V), and then the heterosyllabic
(CV.NV#) one. The pair-wise comparisons between each nasalized context and the oral
control exhibit differences. Such differences are shown and described in Figures 4a,b, 5a,b
and 6a,b.4
Figure 4a compares anticipatory nasalization contours in the oral control CVCV
structure and the tautosyllabic CVN# one and illustrates a minimal and stable development
of nasal airflow contour for CVCV, while a rise is seen towards to end of the segment for
CVN#. This is to be expected, as the segment to follow V1 in CVN# is a nasal consonant,
and as such, due to anticipatory coarticulatory nasalization, the onset of the nasal gesture
that applies to the nasal consonant will begin before the offset of the previous vowel, i.e.,
V1. As shown in Figure 4b, the time region of significance starts at 0.86. Aerodynamic data
for Spanish has shown delayed onsets of nasalization, as opposed to other languages like
English (Beristain 2023a).
Languages 2024, 9, x FOR PEER REVIEW 10 of 24
Figure 3. Nasal airflow GAMM curves of V1 in CVCV (oral control), CVN# (tautosyllabic), CV.NV#
(heterosyllabic), and CV.N#V (resyllabification)3.
As can be seen, the nasal airflow contour in the oral control (CVCV) context is stable
and virtually non-existent. The tautosyllabic (CVN#) sequence shows the greatest amount
of nasal airflow for V1, followed by resyllabification (CV.N#V), and then the heterosyl-
labic (CV.NV#) one. The pair-wise comparisons between each nasalized context and the
oral control exhibit differences. Such differences are shown and described in Figures 4a,b,
5a,b and 6a,b.4
Figure 4a compares anticipatory nasalization contours in the oral control CVCV
structure and the tautosyllabic CVN# one and illustrates a minimal and stable develop-
ment of nasal airflow contour for CVCV, while a rise is seen towards to end of the segment
for CVN#. This is to be expected, as the segment to follow V1 in CVN# is a nasal conso-
nant, and as such, due to anticipatory coarticulatory nasalization, the onset of the nasal
gesture that applies to the nasal consonant will begin before the offset of the previous
vowel, i.e., V1. As shown in Figure 4b, the time region of significance starts at 0.86. Aero-
dynamic data for Spanish has shown delayed onsets of nasalization, as opposed to other
languages like English (Beristain 2023a).
(a)
Figure 4. Cont.
Languages 2024,9, 219 11 of 23
Languages 2024, 9, x FOR PEER REVIEW 11 of 24
(b)
Figure 4. (a) Nasal airflow GAMM curves of V1 in CVCV and CVN# syllable structures; (b) differ-
ence curve of CVN# and CVCV (in V1).
Figure 5a compares anticipatory nasalization contours in CVCV and the heterosyl-
labic CV.NV# sequences, and it shows similar paerns to those in Figure 4a. The main
difference is that the nasal airflow contour for CV.NV# exhibits lower amplitude, and con-
fidence intervals are wider. Figure 5b reveals that the time region of significance starts at
0.92, which is later than for CVN#.
(a)
Figure 4. (a) Nasal airflow GAMM curves of V1 in CVCV and CVN# syllable structures; (b) difference
curve of CVN# and CVCV (in V1).
Languages 2024, 9, x FOR PEER REVIEW 11 of 24
(b)
Figure 4. (a) Nasal airflow GAMM curves of V1 in CVCV and CVN# syllable structures; (b) differ-
ence curve of CVN# and CVCV (in V1).
Figure 5a compares anticipatory nasalization contours in CVCV and the heterosyl-
labic CV.NV# sequences, and it shows similar paerns to those in Figure 4a. The main
difference is that the nasal airflow contour for CV.NV# exhibits lower amplitude, and con-
fidence intervals are wider. Figure 5b reveals that the time region of significance starts at
0.92, which is later than for CVN#.
(a)
Languages 2024, 9, x FOR PEER REVIEW 12 of 24
(b)
Figure 5. (a) Nasal airflow GAMM curves of V1 in CVCV and CV.NV# syllable structures; (b) dif-
ference curve of CV.NV# and CVCV (in V1).
In Figure 6a, the CVCV sequence is compared to the CV.N#V resyllabification con-
text. The paerns found are more similar to those in CVN#. The first time region of signif-
icance in Figure 6b (0.05 to 0.20) should not be considered an indicator of onset of nasali-
zation because the downward directionality of the nasal airflow could pertain to air inha-
lation and not velum lowering. The second time region of significance starts at 0.89.
(a)
Figure 5. (a) Nasal airflow GAMM curves of V1 in CVCV and CV.NV# syllable structures; (b) differ-
ence curve of CV.NV# and CVCV (in V1).
Languages 2024,9, 219 12 of 23
Languages 2024, 9, x FOR PEER REVIEW 12 of 24
(b)
Figure 5. (a) Nasal airflow GAMM curves of V1 in CVCV and CV.NV# syllable structures; (b) dif-
ference curve of CV.NV# and CVCV (in V1).
In Figure 6a, the CVCV sequence is compared to the CV.N#V resyllabification con-
text. The paerns found are more similar to those in CVN#. The first time region of signif-
icance in Figure 6b (0.05 to 0.20) should not be considered an indicator of onset of nasali-
zation because the downward directionality of the nasal airflow could pertain to air inha-
lation and not velum lowering. The second time region of significance starts at 0.89.
(a)
Languages 2024, 9, x FOR PEER REVIEW 13 of 24
(b)
Figure 6. (a) Nasal airflow GAMM curves of V1 in CVCV and CV.N#V syllable structures; (b) dif-
ference curve of CV.N#V and CVCV (in V1).
Below is Table 3, which summarizes the onset of significant differences between na-
salized and oral V1s, which we consider a cue for onset of nasalization.
Table 3. Onset of nasalization (0–1) of V1.
Onset of Nasalization
CVN# (tautosyllabic) 0.86
CV.NV# (heterosyllabic) 0.92
CV.N#V (resyllabification) 0.89
Regarding inferential statistics, the only significant difference between the nasal air-
flow contours in nasalized vowels (CVN#, CV.NV#, and CV.N#V) was retrieved between
CVN# and CV.NV#, as illustrated in Figure 7a,b. The tautosyllabic CVN# sequence exhib-
ited a greater amount of nasal airflow than the resyllabification CV.N#V context. The time
region of significance started at 0.91.
(a)
Figure 6. (a) Nasal airflow GAMM curves of V1 in CVCV and CV.N#V syllable structures; (b) differ-
ence curve of CV.N#V and CVCV (in V1).
Figure 5a compares anticipatory nasalization contours in CVCV and the heterosyllabic
CV.NV# sequences, and it shows similar patterns to those in Figure 4a. The main difference
is that the nasal airflow contour for CV.NV# exhibits lower amplitude, and confidence
intervals are wider. Figure 5b reveals that the time region of significance starts at 0.92,
which is later than for CVN#.
In Figure 6a, the CVCV sequence is compared to the CV.N#V resyllabification context.
The patterns found are more similar to those in CVN#. The first time region of significance
in Figure 6b (0.05 to 0.20) should not be considered an indicator of onset of nasalization
because the downward directionality of the nasal airflow could pertain to air inhalation
and not velum lowering. The second time region of significance starts at 0.89.
Below is Table 3, which summarizes the onset of significant differences between
nasalized and oral V1s, which we consider a cue for onset of nasalization.
Languages 2024,9, 219 13 of 23
Table 3. Onset of nasalization (0–1) of V1.
Onset of Nasalization
CVN# (tautosyllabic) 0.86
CV.NV# (heterosyllabic) 0.92
CV.N#V (resyllabification) 0.89
Regarding inferential statistics, the only significant difference between the nasal airflow
contours in nasalized vowels (CVN#, CV.NV#, and CV.N#V) was retrieved between CVN#
and CV.NV#, as illustrated in Figure 7a,b. The tautosyllabic CVN# sequence exhibited a
greater amount of nasal airflow than the resyllabification CV.N#V context. The time region
of significance started at 0.91.
Languages 2024, 9, x FOR PEER REVIEW 13 of 24
(b)
Figure 6. (a) Nasal airflow GAMM curves of V1 in CVCV and CV.N#V syllable structures; (b) dif-
ference curve of CV.N#V and CVCV (in V1).
Below is Table 3, which summarizes the onset of significant differences between na-
salized and oral V1s, which we consider a cue for onset of nasalization.
Table 3. Onset of nasalization (0–1) of V1.
Onset of Nasalization
CVN# (tautosyllabic) 0.86
CV.NV# (heterosyllabic) 0.92
CV.N#V (resyllabification) 0.89
Regarding inferential statistics, the only significant difference between the nasal air-
flow contours in nasalized vowels (CVN#, CV.NV#, and CV.N#V) was retrieved between
CVN# and CV.NV#, as illustrated in Figure 7a,b. The tautosyllabic CVN# sequence exhib-
ited a greater amount of nasal airflow than the resyllabification CV.N#V context. The time
region of significance started at 0.91.
(a)
Languages 2024, 9, x FOR PEER REVIEW 14 of 24
(b)
Figure 7. (a) Nasal airflow GAMM curves of V1 in CVN# and CV.NV# syllable structures; (b) dif-
ference curve of CVN# and CV.NV# (in V1).
Regarding the contour differences between CVN# and CV.N#V, the former showed
a higher degree of nasal airflow starting relatively close to the midpoint of the vowel (see
Figure 8a). However, no significant differences were found between the two contexts, as
shown in Figure 8b.
(a)
(b)
Figure 8. (a) Nasal airflow GAMM curves of V1 in CVN# and CV.N#V syllable structures; (b) dif-
ference curve of CVN# and CV.N#V (in V1).
Figure 7. (a) Nasal airflow GAMM curves of V1 in CVN# and CV.NV# syllable structures; (b) differ-
ence curve of CVN# and CV.NV# (in V1).
Regarding the contour differences between CVN# and CV.N#V, the former showed a
higher degree of nasal airflow starting relatively close to the midpoint of the vowel (see
Figure 8a). However, no significant differences were found between the two contexts, as
shown in Figure 8b.
Languages 2024,9, 219 14 of 23
Languages 2024, 9, x FOR PEER REVIEW 14 of 24
(b)
Figure 7. (a) Nasal airflow GAMM curves of V1 in CVN# and CV.NV# syllable structures; (b) dif-
ference curve of CVN# and CV.NV# (in V1).
Regarding the contour differences between CVN# and CV.N#V, the former showed
a higher degree of nasal airflow starting relatively close to the midpoint of the vowel (see
Figure 8a). However, no significant differences were found between the two contexts, as
shown in Figure 8b.
(a)
(b)
Figure 8. (a) Nasal airflow GAMM curves of V1 in CVN# and CV.N#V syllable structures; (b) dif-
ference curve of CVN# and CV.N#V (in V1).
Figure 8. (a) Nasal airflow GAMM curves of V1 in CVN# and CV.N#V syllable structures; (b) differ-
ence curve of CVN# and CV.N#V (in V1).
5.2. Carryover Vowel Nasalization
The results in this section pertain to nasal airflow in the second vowel (V2), i.e., CVCV
(oral control), CV.NV# (heterosyllabic), and CV.N#V(resyllabification). Figure styles are
similar to those in Section 5.1. The x-axis represents normalized time (NormTime) of
40 equidistant points, and y-axis indicates the amount of nasal airflow produced (in L/s).
The left side of figures are the onset of the vowel, while the right side are its offset. Notice
that the CVN# (tautosyllabic) syllable structure will not be included in this section because
there is no V2. The offset of carryover nasalization was calculated by contrasting the V2
nasal airflow contours in CVCV(oral control) versus CV.NV# and CV.N#V(nasalized).
Figure 9illustrates nasal airflow in CVCV, CV.NV#, and CV.N#V. As can be seen,
CVCVshows a stable and virtually non-existent nasal airflow, which is to be expected
as no nasal segment is present in that sequence. With regard to the nasalized contexts
and carryover nasalization patterns, CV.NV# and CV.N#Vshow parallel patterns: a sharp
decline is present at the beginning of V2, induced by the presence of the preceding nasal
segment /n/. After this, nasal airflow stabilizes throughout the remainder of the vowel,
always showing higher nasal airflow values than the oral control context CVCV.
Languages 2024,9, 219 15 of 23
Languages 2024, 9, x FOR PEER REVIEW 15 of 24
5.2. Carryover Vowel Nasalization
The results in this section pertain to nasal airflow in the second vowel (V2), i.e., CVCV
(oral control), CV.NV# (heterosyllabic), and CV.N#V (resyllabification). Figure styles are
similar to those in Section 5.1. The x-axis represents normalized time (NormTime) of 40
equidistant points, and y-axis indicates the amount of nasal airflow produced (in L/s). The
left side of figures are the onset of the vowel, while the right side are its offset. Notice that
the CVN# (tautosyllabic) syllable structure will not be included in this section because
there is no V2. The offset of carryover nasalization was calculated by contrasting the V2
nasal airflow contours in CVCV (oral control) versus CV.NV# and CV.N#V (nasalized).
Figure 9 illustrates nasal airflow in CVCV, CV.NV#, and CV.N#V. As can be seen,
CVCV shows a stable and virtually non-existent nasal airflow, which is to be expected as
no nasal segment is present in that sequence. With regard to the nasalized contexts and
carryover nasalization paerns, CV.NV# and CV.N#V show parallel paerns: a sharp de-
cline is present at the beginning of V2, induced by the presence of the preceding nasal
segment /n/. After this, nasal airflow stabilizes throughout the remainder of the vowel,
always showing higher nasal airflow values than the oral control context CVCV.
Figure 9. Nasal airflow GAMM curves of V2 in CVCV, CV.NV#, and CV.N#V.
Pair-wise comparisons are presented in Figures 10a,b, 11a,b, and 12a,b.
(a)
Figure 9. Nasal airflow GAMM curves of V2 in CVCV, CV.NV#, and CV.N#V.
Pair-wise comparisons are presented in Figures 10a,b, 11a,b, and 12a,b.
Languages 2024, 9, x FOR PEER REVIEW 15 of 24
5.2. Carryover Vowel Nasalization
The results in this section pertain to nasal airflow in the second vowel (V2), i.e., CVCV
(oral control), CV.NV# (heterosyllabic), and CV.N#V (resyllabification). Figure styles are
similar to those in Section 5.1. The x-axis represents normalized time (NormTime) of 40
equidistant points, and y-axis indicates the amount of nasal airflow produced (in L/s). The
left side of figures are the onset of the vowel, while the right side are its offset. Notice that
the CVN# (tautosyllabic) syllable structure will not be included in this section because
there is no V2. The offset of carryover nasalization was calculated by contrasting the V2
nasal airflow contours in CVCV (oral control) versus CV.NV# and CV.N#V (nasalized).
Figure 9 illustrates nasal airflow in CVCV, CV.NV#, and CV.N#V. As can be seen,
CVCV shows a stable and virtually non-existent nasal airflow, which is to be expected as
no nasal segment is present in that sequence. With regard to the nasalized contexts and
carryover nasalization paerns, CV.NV# and CV.N#V show parallel paerns: a sharp de-
cline is present at the beginning of V2, induced by the presence of the preceding nasal
segment /n/. After this, nasal airflow stabilizes throughout the remainder of the vowel,
always showing higher nasal airflow values than the oral control context CVCV.
Figure 9. Nasal airflow GAMM curves of V2 in CVCV, CV.NV#, and CV.N#V.
Pair-wise comparisons are presented in Figures 10a,b, 11a,b, and 12a,b.
(a)
Languages 2024, 9, x FOR PEER REVIEW 16 of 24
(b)
Figure 10. (a) Nasal airflow GAMM curves of V2 in CV.NV# and CV.N#V syllable structures; (b)
difference curve of CV.NV# and CV.N#V (in V2).
As shown in Figure 10a, the V2 nasal airflow contours in nasalized contexts show
similar paerns between CV.NV# and CV.N#V. As explained in Figure 9, a rapid decline
during the initial portion of the vowel is followed by a more stable nasal airflow contour
that does not decrease to zero. When contrasting both nasal airflow contours, no signifi-
cant differences are found (see Figure 10b).
Figure 11a compares the nasal airflow contours of CVCV and CV.NV#. The oral con-
text shows a stable and low degree of nasalization, to be expected from its structure, and
the heterosyllabic CV.NV# sequence shows a rapid decline, with some overlap between
the 0.2–1 time region. The time regions of significance between contours arise in 0–0.35,
0.43–0.51, and 0.97–1 (Figure 11b). One could argue the offset may be when the first time
region of significance culminates, but we should be cautious considering the various time
regions of significance and the physiological nature of the data. If we observe the data, we
will see that the reason for the lack of statistical difference between the two time regions
of significance is due to the downward directionality of the nasal airflow in CV.NV#. Since
the nasal airflow remains relatively stable in that portion of the vowel, there is no reason
to believe that the offset of carryover vocalic nasalization is that first instance, and we
should thus set it at the second one at 0.51. The third time region of significance is not an
indicator or cue of onset of nasalization, as the following consonant is a /t/.
(a)
Figure 10. (a) Nasal airflow GAMM curves of V2 in CV.NV# and CV.N#Vsyllable structures;
(b) difference curve of CV.NV# and CV.N#V(in V2).
Languages 2024,9, 219 16 of 23
Languages 2024, 9, x FOR PEER REVIEW 16 of 24
(b)
Figure 10. (a) Nasal airflow GAMM curves of V2 in CV.NV# and CV.N#V syllable structures; (b)
difference curve of CV.NV# and CV.N#V (in V2).
As shown in Figure 10a, the V2 nasal airflow contours in nasalized contexts show
similar paerns between CV.NV# and CV.N#V. As explained in Figure 9, a rapid decline
during the initial portion of the vowel is followed by a more stable nasal airflow contour
that does not decrease to zero. When contrasting both nasal airflow contours, no signifi-
cant differences are found (see Figure 10b).
Figure 11a compares the nasal airflow contours of CVCV and CV.NV#. The oral con-
text shows a stable and low degree of nasalization, to be expected from its structure, and
the heterosyllabic CV.NV# sequence shows a rapid decline, with some overlap between
the 0.2–1 time region. The time regions of significance between contours arise in 0–0.35,
0.43–0.51, and 0.97–1 (Figure 11b). One could argue the offset may be when the first time
region of significance culminates, but we should be cautious considering the various time
regions of significance and the physiological nature of the data. If we observe the data, we
will see that the reason for the lack of statistical difference between the two time regions
of significance is due to the downward directionality of the nasal airflow in CV.NV#. Since
the nasal airflow remains relatively stable in that portion of the vowel, there is no reason
to believe that the offset of carryover vocalic nasalization is that first instance, and we
should thus set it at the second one at 0.51. The third time region of significance is not an
indicator or cue of onset of nasalization, as the following consonant is a /t/.
(a)
Languages 2024, 9, x FOR PEER REVIEW 17 of 24
(b)
Figure 11. (a) Nasal airflow GAMM curves of V2 in CV.NV# and CVCV syllable structures; (b) dif-
ference curve of CV.NV# and CVCV (in V2).
In Figure 12a, the CV.N#V nasal airflow contour is compared to that of CVCV. As
can be seen, the distance between the two curves is greater and the overlap is more de-
layed than in Figure 11a, which applies to CV.NV#. Approximately, the overlap in Figure
12a begins close to 0.6, and extends until the end of V2. By contrasting both contours, time
regions of significance arise at 0–0.75 and 0.94–1 (see Figure 12b). Notice that the second
significance window frame arises in the nasal airflow due to the rise of the CV.N#V curve
in the final portion of V2.
(a)
Figure 11. (a) Nasal airflow GAMM curves of V2 in CV.NV# and CVCVsyllable structures; (b) differ-
ence curve of CV.NV# and CVCV(in V2).
As shown in Figure 10a, the V2 nasal airflow contours in nasalized contexts show
similar patterns between CV.NV# and CV.N#V. As explained in Figure 9, a rapid decline
during the initial portion of the vowel is followed by a more stable nasal airflow contour
that does not decrease to zero. When contrasting both nasal airflow contours, no significant
differences are found (see Figure 10b).
Figure 11a compares the nasal airflow contours of CVCVand CV.NV#. The oral
context shows a stable and low degree of nasalization, to be expected from its structure, and
the heterosyllabic CV.NV# sequence shows a rapid decline, with some overlap between
the 0.2–1 time region. The time regions of significance between contours arise in 0–0.35,
0.43–0.51, and 0.97–1 (Figure 11b). One could argue the offset may be when the first time
region of significance culminates, but we should be cautious considering the various time
regions of significance and the physiological nature of the data. If we observe the data, we
will see that the reason for the lack of statistical difference between the two time regions of
significance is due to the downward directionality of the nasal airflow in CV.NV#. Since
the nasal airflow remains relatively stable in that portion of the vowel, there is no reason to
believe that the offset of carryover vocalic nasalization is that first instance, and we should
thus set it at the second one at 0.51. The third time region of significance is not an indicator
or cue of onset of nasalization, as the following consonant is a /l/.
Languages 2024,9, 219 17 of 23
Languages 2024, 9, x FOR PEER REVIEW 17 of 24
(b)
Figure 11. (a) Nasal airflow GAMM curves of V2 in CV.NV# and CVCV syllable structures; (b) dif-
ference curve of CV.NV# and CVCV (in V2).
In Figure 12a, the CV.N#V nasal airflow contour is compared to that of CVCV. As
can be seen, the distance between the two curves is greater and the overlap is more de-
layed than in Figure 11a, which applies to CV.NV#. Approximately, the overlap in Figure
12a begins close to 0.6, and extends until the end of V2. By contrasting both contours, time
regions of significance arise at 0–0.75 and 0.94–1 (see Figure 12b). Notice that the second
significance window frame arises in the nasal airflow due to the rise of the CV.N#V curve
in the final portion of V2.
(a)
Languages 2024, 9, x FOR PEER REVIEW 18 of 24
(b)
Figure 12. (a) Nasal airflow GAMM curves of V2 in CV.N#V and CVCV syllable structures; (b) dif-
ference curve of CV.N#V and CVCV (in V2).
Table 4 provides a summary of the offsets of nasalization for V2.
Table 4. Offset of nasalization (0–1) of V2.
Offset of Nasalization
CV.NV# (heterosyllabic) 0.51
CV.N#V (resyllabification) 0.75
6. Discussion
Let us revisit the goals and hypotheses of the current study: first, to document the
differences between anticipatory and carryover coarticulatory nasalization in Spanish.
Based on previous studies such as Diakoumakou (2009) and Delvaux et al. (2009), we hy-
pothesized that carryover nasalization would present to a greater extent than anticipatory
nasalization. Second, to elucidate whether the extent of coarticulatory anticipatory and
carryover nasalization in Spanish is affected by tautosyllabicity and heterosillabicity of
VN sequences. We hypothesized that segments within the same syllable would show
greater gestural overlap (Byrd et al. 2009) and an earlier onset of nasalization than those
in canonical and derived (resyllabification) heterosyllabic sequences. Third, to test the role
of resyllabification on anticipatory and carryover coarticulatory nasalization. Previous lit-
erature (Bradley et al. 2022; Hualde and Prieto 2014; Jiménez-Bravo and Lahoz-Ben-
goechea 2023; Strycharczuk and Kohlberger 2016) has shown differences between canon-
ical and derived onset consonants. However, articulatory data show that the characteris-
tics of neighboring vowels in those contexts are similar (Cohn 1990; Krakow 1989, 1999).
Moreover, Lahoz-Bengoechea and Jiménez-Bravo (forthcoming) find the degree of antici-
patory nasalization in V#NV and VN#V sequences to be identical.
This study showed that the degree of coarticulatory carryover nasalization is greater
than that of anticipatory nasalization in Spanish (RQ1, Hypothesis 1). Anticipatory nasal-
ization was implemented in the final 15% of the segment, while carryover nasalization
extended from the beginning to 51–75% of the vocalic segment. Previous literature had
diverging proposals as to which one of the two shows greater nasal spreading. An argu-
ment that is generally used in favor of this view is how current phonologically nasal vow-
els in Romance languages are a development of contiguous VN segments and not NV
(Sampson 1999). Henderson (1984), by means of data obtained with a fiberoptic endo-
scope, argues that it is anticipatory nasalization which shows greater extent, and provides
Figure 12. (a) Nasal airflow GAMM curves of V2 in CV.N#Vand CVCVsyllable structures; (b) differ-
ence curve of CV.N#Vand CVCV(in V2).
In Figure 12a, the CV.N#Vnasal airflow contour is compared to that of CVCV. As can
be seen, the distance between the two curves is greater and the overlap is more delayed
than in Figure 11a, which applies to CV.NV#. Approximately, the overlap in Figure 12a
begins close to 0.6, and extends until the end of V2. By contrasting both contours, time
regions of significance arise at 0–0.75 and 0.94–1 (see Figure 12b). Notice that the second
significance window frame arises in the nasal airflow due to the rise of the CV.N#Vcurve
in the final portion of V2.
Table 4provides a summary of the offsets of nasalization for V2.
Table 4. Offset of nasalization (0–1) of V2.
Offset of Nasalization
CV.NV# (heterosyllabic) 0.51
CV.N#V(resyllabification) 0.75
Languages 2024,9, 219 18 of 23
6. Discussion
Let us revisit the goals and hypotheses of the current study: first, to document the
differences between anticipatory and carryover coarticulatory nasalization in Spanish.
Based on previous studies such as Diakoumakou (2009) and Delvaux et al. (2009), we
hypothesized that carryover nasalization would present to a greater extent than anticipa-
tory nasalization. Second, to elucidate whether the extent of coarticulatory anticipatory
and carryover nasalization in Spanish is affected by tautosyllabicity and heterosillabicity
of VN sequences. We hypothesized that segments within the same syllable would show
greater gestural overlap (Byrd et al. 2009) and an earlier onset of nasalization than those in
canonical and derived (resyllabification) heterosyllabic sequences. Third, to test the role of
resyllabification on anticipatory and carryover coarticulatory nasalization. Previous litera-
ture (Bradley et al. 2022;Hualde and Prieto 2014;Jiménez-Bravo and Lahoz-Bengoechea
2023;Strycharczuk and Kohlberger 2016) has shown differences between canonical and
derived onset consonants. However, articulatory data show that the characteristics of
neighboring vowels in those contexts are similar (Cohn 1990;Krakow 1989,1999). More-
over, Lahoz-Bengoechea and Jiménez-Bravo (forthcoming) find the degree of anticipatory
nasalization in V#NV and VN#V sequences to be identical.
This study showed that the degree of coarticulatory carryover nasalization is greater
than that of anticipatory nasalization in Spanish (RQ1, Hypothesis 1). Anticipatory nasal-
ization was implemented in the final 15% of the segment, while carryover nasalization
extended from the beginning to 51–75% of the vocalic segment. Previous literature had di-
verging proposals as to which one of the two shows greater nasal spreading. An argument
that is generally used in favor of this view is how current phonologically nasal vowels in
Romance languages are a development of contiguous VN segments and not NV (Sampson
1999). Henderson (1984), by means of data obtained with a fiberoptic endoscope, argues
that it is anticipatory nasalization which shows greater extent, and provides evidence for
the velum reaching lower positions in CVN# sequences than in NVC#. However, it should
be noted that, in Henderson’s study, the coda consonant in the NVC# sequence is a /t/,
which could induce an earlier rise of the velum in the gestural timing and transition. Other
studies such as Diakoumakou (2009) have pointed out that the extent of coarticulatory
nasalization is language-specific. Her investigation focuses on Modern Greek, but also
considers other languages, and proposes that languages that favor open syllables show
a tendency for greater carryover nasalization, while those languages that favor closed
syllables show greater extent of anticipatory vowel nasalization. Spanish is an example of a
language that favors the CVCV type of open syllables, and greater carryover nasalization is
indeed the pattern that is encountered. In recent studies comparing coarticulatory nasal-
ization in CVN sequences in English and Spanish, Beristain (2023a,2023b) found that the
onset of nasalization was significantly earlier in English, and that the ratio of nasal airflow
proportion (to total airflow produced) was significantly higher in English, too. Whether this
can simply be explained by the preference of English for closed syllables or linguistically
internal reasons goes beyond the scope of the current study.
Furthermore, evidence for an earlier onset of vocalic nasalization in tautosyllabic
CVN# sequences as opposed to heterosyllabic canonical CV.NV# and derived CV.N#V
sequences (RQ2, Hypothesis 2) was found. This finding is in alignment with previous
literature that considers (a) articulatory gestures to be an important part of the physiological
part of the syllable (Byrd et al. 2009;Krakow 1999), and (b) that the timing among gestures
of segments within the same syllable possess greater articulatory overlap and coordination
than those in different syllables. In the CVN# sequence, the velum will have started
to reach a spatially lower position during the vocalic segment by the time we have the
onset of N. On the other hand, for a vowel that is contiguous to a nasal consonant in a
heterosyllabic condition, the gestural coordination will be different. As a word-medial
onset, the articulatory effort is not equal to a word-initial onset. Krakow (1999) points out
that syllable-initial consonants are ‘stronger’ because “it is generally associated with tighter
articulatory constrictions and with greater stability” (p. 25). The significant difference
Languages 2024,9, 219 19 of 23
found in anticipatory nasalization between CVN# and CV.NV# could be induced because
of syllable structure. However, the lack of significant differences between CVN# and
CV.N#V complicates that argument. As far as structural similarities are concerned in
the generative literature, one would have expected differences to arise between CVN# vs.
CV.NV#, and CVN# vs. CV.N#V, not solely in the former
5
.Lahoz-Bengoechea and Jiménez-
Bravo (forthcoming) found similar results when comparing VN sequences in VN# and V#N
structures. The authors point out that as coarticulatory nasalization is a purely phonetic and
a mechanical byproduct in Spanish, “[its effect arises in close contact to the nasal consonant,
independently from its phonological structure]” (p. 3, translated by the author of this
manuscript). Previous literature on the resyllabification of non-nasal segments has deduced
that derived onsets exhibit distinct sub-phonemic characteristics from canonical onsets
(Hualde and Prieto 2014;Strycharczuk and Kohlberger 2016). Although those studies have
focused on oral fricatives mostly, it could be hypothesized that some additional cues might
be affected in nasal consonants in resyllabification contexts as well. A future study will
provide a multidimensional view of aerodynamic and acoustic characteristics of the nasal
segments that were examined in the current study.
Third, in alignment with results presented in Cohn (1990), the current data demon-
strated that carryover nasalization patterns were parallel between: (i) within word, het-
erosyllabic CV.NV#, and (ii) across word boundaries, resyllabification CV.N#Vsequences
(RQ3, Hypothesis 3). Most of the previous literature had focused on the way in which
resyllabification affected the consonant, and those studies found differences in terms of
the acoustic cues such as duration and voicing frames of oral fricatives. Little had been
mentioned about how coarticulatory nasalization could spread across word junctures and
across syllabic reorganization processes. This study showed that no significant differences
were found regarding anticipatory or carryover nasalization among CV.NV# and CV.N#V
sequences, which correlates with what Strycharczuk and Kohlberger (2016) found; that is,
that vowels are less malleable in their phonetic production as opposed to the consonant
for which the syllabic linking changes. As pointed out in the previous paragraph, one of
the reasons for this may be that coarticulatory nasalization is a mechanical byproduct in
Spanish, and the generative literature considers CV.NV# and CV.N#Vsequences struc-
turally identical. An important difference between previous studies and the current one
is the nature of the data. While most of the previous studies investigating the effect of
resyllabification explore acoustic correlates, the current one delves into the physiological
characteristics of speech; in other words, it analyzes the way in which air is exhaled during
speech production. This avenue had not been explored in the context of syllable structure
realignment. As far as resyllabification is concerned, generative phonology considers its
structure identical to any other heterosyllabic structure, yet previously mentioned labo-
ratory studies provided evidence for acoustic differences between canonical and derived
onsets. The results obtained in this study may elucidate a novel perspective on connected
speech and its articulatory features. Here, the extent of phonetic characteristics such as
coarticulation can be studied more accurately and without the need to conclude differences
from fine-grained details in the acoustic signal.
Lastly, the limitations of the current study cannot go unnoticed. A sample of nine par-
ticipants and a database of 630 vocalic tokens were used to draw conclusions. Furthermore,
Spanish data were collected in an English-speaking environment. As such, we need to
consider that participants’ English could have influenced their Spanish productions. The re-
searcher, a native Spanish speaker, tried to induce participants into a “monolingual”-mode
to the best of his abilities where only Spanish was used during the experimental setting.
Moreover, while none of the target tokens were uncommon words, their lexical frequency
was not meticulously controlled (cf. Brown 2008). Doing so could have contributed to a
more thorough understanding and analysis of the topic under study. Considering that
aerodynamic experiments have reported small(er) sample sizes (Kochetov 2020), the current
study is a significant contribution to the literature on syllable structure, Spanish phonology,
and articulatory phonetics.
Languages 2024,9, 219 20 of 23
7. Conclusions
The two aims of the current study were to investigate aerodynamic differences between
anticipatory and carryover coarticulatory nasalization in Spanish and to observe whether
nasalization can be used as a cue to resyllabification.
The contours of aerodynamic nasal airflow data illustrate that carryover nasalization
exhibited greater nasal spreading than the anticipatory context. This was in alignment with
previous literature, which points out that such a pattern can be found in languages that
favor open syllables. In that regard, the degree of anticipatory nasalization was greater in
CVN# tautosyllabic structures compared to in heterosyllabic CV.NV# and resyllabification
CV.N#V ones. Regarding carryover nasalization, vowels in the heterosyllabic CV.NV# and
resyllabification CV.N#Vcontexts showed parallel patterns.
These findings indicate that the position and structure of the syllable are crucial
elements to take into consideration and that they affect inter-articulatory gestural overlap.
Furthermore, the current study demonstrates that vowels in heterosyllabic CV.NV# and
resyllabification CV.N#Vstructures are similar in terms of their aerodynamic properties,
thus providing evidence in favor of derivational rules in Spanish phonology as an exemplar
of complete resyllabification in connected speech.
Funding: This research received no external funding.
Institutional Review Board Statement: The study was conducted in accordance with the Declaration
of Helsinki, and approved by the Institutional Review Board of the University of Illinois Urbana-
Champaign (protocol code #20071 and 16 August 2019) for studies involving humans.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: Not applicable due to the IRB protocol does not allow the author to
share the data openly for privacy and confidentiality restrictions.
Acknowledgments: I would like to thank the audience at the 51st Linguistic Symposium on Romance
Languages, the two anonymous reviewers and editor for their comments, JoséIgnacio Hualde for his
input on a previous version of this manuscript, and Jennifer Zhang for stylistic feedback.
Conflicts of Interest: The authors declare no conflicts of interest.
Appendix A
Model selection and AIC scores
Model AIC
Score
Deviance
Explained
Coarticulatory anticipatory nasalization (V1):
Context * Vowel + Sex + s(NormTime, by =
interaction(Context, Vowel)) + s(NormTime, by = Sex) +
s(NormTime, Speaker, by = interaction(Context, Vowel),
bs = “fs”, m = 1) + s(Word, bs = “re”)
−127,117.7 72.9%
Context * Vowel + s(NormTime, by = interaction(Context,
Vowel)) + s(NormTime, Speaker, by = interaction(Context,
Vowel), bs = “fs”, m = 1) + s(Word, bs = “re”)
−114,413.8 67%
Coarticulatory carryover nasalization (V2):
Context + Sex + s(NormTime, by = Context) + s(NormTime,
by = Sex) + s(NormTime, Speaker, by = Context, bs =
“fs”, m = 1) + s(Word, bs = “re”)
−86,151.6 80.7%
Context + s(NormTime, by = Context) + s(NormTime,
Speaker, by = Context, bs = “fs”, m = 1) + s(Word, bs =
“re”)
−77,586.6 78.5%
Languages 2024,9, 219 21 of 23
Appendix B
Model output for anticipatory vowel nasalization (V1)
Formula:
nasal ~ Context * Vowel + Sex + s(NormTime, by = interaction(Context,
Vowel)) + s(NormTime, by = Sex) + s(NormTime, Speaker, by = interaction
(Context, Vowel), bs = “fs”, m = 1) + s(Word, bs = “re”)
Parametric Coefficients: Estimate SE t-Value p-Value
(Intercept) 0.0009 0.002 0.502 0.61580
Context CVN# 0.002 0.003 0.735 0.46240
Context CV.N#V 0.0004 0.002 0.174 0.86202
Context CVCV 0.0007 0.002 0.378 0.70522
Vowel e 0.00006 0.002 0.027 0.97817
Vowel i 0.0008 0.002 0.321 0.74846
Vowel o 0.00000 0.003 0.000 0.99962
Vowel u 0.002 0.003 0.494 0.62124
Sex Male −0.002 0.001 −3.128 <0.01
Context CVN# x Vowel e −0.002 0.003 −0.667 0.50486
Context CV.N#V x Vowel e 0.0031 0.004 0.843 0.39919
Context CVCV x Vowel e 0.0003 0.002 0.121 0.90404
Context CVN# x Vowel i 0.0003 0.004 0.073 0.94213
Context CV.N#V x Vowel i 0.0001 0.003 0.034 0.97307
Context CVCV x Vowel i −0.00008 0.002 −0.036 0.97096
Context CVN# x Vowel o −0.0024 0.004 −0.564 0.57271
Context CV.N#V x Vowel o −0.0003 0.004 −0.075 0.93988
Context CVCV x Vowel o 0.0007 0.003 0.227 0.82066
Context CVN# x Vowel u −0.0009 0.005 −0.183 0.85497
Context CV.N#V x Vowel u 0.0008 0.004 0.191 0.84872
Context CVCV x Vowel u −0.0005 0.003 −0.170 0.86534
Deviance explained = 72.9%.
Appendix C
Model output for carryover vowel nasalization (V2)
Formula:
nasal ~ Context + Sex + s(NormTime, by = Context) + s(NormTime,
by = Sex) + s(NormTime, Speaker, by = Context, bs = “fs”, m = 1) + s(Word,
bs = “re”)
Parametric Coefficients: Estimate SE t-Value p-Value
(Intercept) 0.017 0.005 3.148 <0.01
Context CVCV −0.0149 0.005 −2.781 <0.01
Context CVN#V 0.0018 0.007 0.239 0.81
Sex Male −0.0008 0.001 −0.563 0.57
Deviance explained = 80.7%.
Notes
1IRB Protocol Number: 20071, University of Illinois Urbana-Champaign.
2It is worth mentioning that plot_diff() curve differences can only conduct pair-wise comparisons.
3The colors in this palette are color vision deficiency-friendly in the original version in color.
4
A remark should be made about the visible “negative” airflow results. As Beristain (2023a) pointed out, the equipment’s built-in
electric voltage and DC offset could explain such results, or there might be some phonetic cue involving negative airflow and
vocalic nasalization in Spanish, as such results only appeared in Spanish but not in other languages that were originally analyzed
by Beristain (2022). Considering that the negative values are generally a part of the confidence intervals, it should not pose any
significant alterations in the current results.
5Using vowels with different height and a relatively small corpus may have contributed to a somewhat large standard error.
Languages 2024,9, 219 22 of 23
References
Akaike, Hirotsugu. 1974. A new look at the statistical model identification. IEEE Transactions on Automatic Control 19: 716–23. [CrossRef]
Beristain, Ander. 2022. The Acquisition of Acoustic and Aerodynamic Patterns of Coarticulation in Second and Heritage Languages.
Ph.D. thesis, University of Illinois Urbana-Champaign, Champaign, IL, USA.
Beristain, Ander. 2023a. Gestural timing patterns of nasality in highly proficient Spanish learners of English: Aerodynamic evidence.
Language and Speech. [CrossRef] [PubMed]
Beristain, Ander. 2023b. An airflow analysis of Spanish and English anticipatory vowel nasalization among heritage bilinguals.
Languages 8: 205. [CrossRef]
BIOPAC Hardware Systems, Inc. 2020. AcqKnowledge Data Acquisition and Analysis Software. version 3.9.1. Goleta: BIOPAC Hardware
Systems, Inc.
Bissiri, Maria P., Maria-Luisa García-Lecumberri, Martin Cooke, and Jan Volín. 2011. The role of word-initial glottal stops in recognizing
English words. Paper presented at Interspeech 2011, Florence, Italy, August 27–31.
Boersma, Paul, and David Weenink. 2020. Praat: Doing Phonetics by Computer [Computer Program], Version 6.1.10. Available online:
http://www.praat.org/ (accessed on 25 March 2020).
Bongiovanni, Silvina. 2021. Acoustic investigation of anticipatory vowel nasalization in a Caribbean and a non-Caribbean dialect of
Spanish. Linguistics Vanguard 7: 1–11. [CrossRef]
Bradley, Travis G., Kimberly Morris, and Lillian Jones. 2022. Incomplete resyllabification and ambisyllabic gestural coupling in Spanish.
Loquens 9: e094. [CrossRef]
Brown, Earl K. 2008. A Usage-Based Account of Syllable- and Word-Final /s/ Reduction in Four Dialects of Spanish. Ph.D. thesis, The
University of New Mexico, Albuquerque, NM, USA.
Byrd, Dani, Stephen Tobin, Erik Bresch, and Shrikanth Narayanan. 2009. Timing effects of syllable structure and stress on nasals: A
real-time MRI examination. Journal of Phonetics 37: 97–110. [CrossRef]
Carignan, Christopher. 2021. A practical method of estimating the time-varying degree of vowel nasalization from acoustic features.
Journal of the Acoustical Society of America 149: 911–22. [CrossRef] [PubMed]
Clumeck, Harold. 1976. Patterns of soft palate movements in six languages. Journal of Phonetics 4: 337–51. [CrossRef]
Cohn, Abigail C. 1990. Phonetic and phonological rules of nasalization. UCLA Working Papers 76. Available online: https://escholarship.
org/uc/item/1xq3d5hr (accessed on 20 December 2023).
Cohn, Abigail C. 1993. Nasalisation in English: Phonology or phonetics. Phonology 10: 43–81. [CrossRef]
Colina, Sonia. 1995. A Constraint-Based Analysis of Syllabification in Spanish, Catalan, and Galician. Ph.D. dissertation, University of
Illinois at Urbana-Champaign, Champaign, IL, USA.
Colina, Sonia. 1997. Identity constraints and Spanish resyllabification. Lingua 103: 1–23. [CrossRef]
Colina, Sonia. 2002. Interdialectal variation in Spanish /s/ aspiration: The role of prosodic structure and output-to-output constraints.
In Structure, Meaning, and Acquisition in Spanish: Papers from the 4th Hispanic Linguistics Symposium. Edited by James F. Lee,
Kimberly L. Geeslin and J. Clancy Clements. Somervilla: Cascadilla Press, pp. 230–43.
Colina, Sonia. 2009. Spanish Phonology: A Syllabic Perspective. Washington, DC: Georgetown University Press.
Delvaux, Véronique, Didier Demolin, Bernard Harmegnies, and Alain Soquet. 2009. The aerodynamics of nasalization in French.
Journal of Phonetics 36: 578–606. [CrossRef]
Diakoumakou, Evanthia. 2005. Coarticulatory nasalization in modern Greek: Evidence for a link between coarticulation and syllable
structure. Journal of the Acoustical Society of America 117: 2459. [CrossRef]
Diakoumakou, Evanthia. 2009. Coarticulatory Vowel Nasalization: Evidence from Modern Greek. Riga: VDM Verlag.
Durand, Jacques, Bernard Laks, Basilio Calderone, and Atanas Tchobanov. 2011. Que savons-nous de la liaison aujourd’hui? [What do
we know today about liaison]? Langue Française 169: 103–35. [CrossRef]
Goodin-Mayeda, C. Elizabeth. 2016. Nasals and Nasalization in Spanish and Portuguese: Perception, Phonetics, and Phonology. Amsterdam:
John Benjamins. [CrossRef]
Hajek, John. 1997. Universals of Sound Change in Nasalization. Hoboken: Wiley-Blackwell.
Harris, James. 1983. Syllable Structure and Stress in Spanish: A Nonlinear Analysis. Cambridge, MA: MIT Press.
Harris, James, and Ellen M. Kaisse. 1999. Palatal vowels, glides and obstruents in Argentinian Spanish. Phonology 19: 117–90. [CrossRef]
Henderson, Janette B. 1984. Velopharyngeal Function in Oral and Nasal Vowels. Ph.D. thesis, University of Connecticut, Storrs, CT,
USA.
Horiguchi, Satoshi, and Fredericka Bell-Berti. 1987. The Velotrace: A device for monitoring velar position. Cleft Palate Journal 24: 104–11.
[PubMed]
Hualde, JoséI. 2014. Los Sonidos del Español [The Sounds of Spanish]. Cambridge: Cambridge University Press. [CrossRef]
Hualde, JoséI., and Jon Ortiz de Urbina. 2003. A Grammar of Basque. Berlin: De Gruyter Mouton. [CrossRef]
Hualde, JoséI., and Pilar Prieto. 2014. Lenition of sibilants in Catalan and Spanish. Phonetica 71: 109–27. [CrossRef] [PubMed]
Huffman, Marie K., and Rena A. Krakow, eds. 1993. Nasals, Nasalization, and the Velum (Phonetics and Phonology V). San Diego: Academic
Press. [CrossRef]
Jiménez-Bravo, Miguel, and JoséMaría Lahoz-Bengoechea. 2023. Durational cues to resyllabification in Spanish. Loquens 10: e099.
[CrossRef]
Languages 2024,9, 219 23 of 23
Kahn, Daniel. 1976. Syllable-Based Generalizations in English Phonology. Ph.D. thesis, Massachusetts Institute of Technology,
Cambridge, MA, USA.
Kaisse, Ellen M. 1999. Resyllabification precedes all segmental rules. In Formal Perspectives on Romance Linguistics: Selected Papers
from the 28th Linguistic Symposium on Romance Languages. Edited by Jean-Marc Authier, Barbara E. Bullock and Lisa A. Reed.
Amsterdam: John Benjamins, pp. 197–210.
Kochetov, Alexei. 2020. Research methods in articulatory phonetics II: Studying other gestures and recent trends. Language and
Linguistics Compass 14: e12371. [CrossRef]
Kohler, Klaus J. 1994. Glottal stops and glottalization in German. Phonetica 51: 38–51. [CrossRef]
Krakow, Rena A. 1989. The Articulatory Organization of Syllables: A Kinematic Analysis of Labial and Velar Gestures. Ph.D. thesis,
Yale University, New Haven, CT, USA.
Krakow, Rena A. 1993. Nonsegmental influences on velum movement patterns: Syllables, sentences, stress, and speaking rate. In Nasals,
Nasalization, and the Velum. Edited by Marie K. Huffman and Rena A. Krakow. San Diego: Academic Press, pp. 87–116.
Krakow, Rena A. 1999. Articulatory organization of syllables: A review. Journal of Phonetics 27: 23–54. [CrossRef]
Lahoz-Bengoechea, JoséMaría, and Miguel Jiménez-Bravo. Forthcoming. Mismo grado de nasalización en VN# y V#N en español
[Same degree of nasalization in VN# and V#N in Spanish]. In Proceedings from the 9th International Conference on Experimental
Phonetics (CIFE). Vigo: Universidad de Vigo, in press.
Martínez, Ruth M. 2021. The Bidirectional Acquisition of Oral and Nasal Gestural Timing in L2 Brazilian Portuguese and Rioplatense
Spanish Nasal Structures. Ph.D. thesis, University of Toronto, Toronto, ON, USA.
MATLAB. 2020. 9.8.0.1323502. R2020a. Natick: The MathWorks Inc.
Passino, Diana, Joaquim Brandão de Carvalho, and Tobias Scheer. 2022. Syllable structure and (re)syllabification. In Manual of Romance
Phonetics and Phonology. Edited by Christoph Gabriel, Randall Gess and Trudel Meisenburg. Berlin and Boston: De Gruyter,
pp. 89–126. [CrossRef]
R Core Team. 2020. R: A Language for Statistical Computing. Available online: https://www.R-project.org/ (accessed on 15 January 2024).
R Studio Team. 2020. RStudio: Integrated Development for R. Available online: http://www.rstudio.com/ (accessed on 15 January 2024).
Robinson, Kimball. 2012. The dialectology of syllabification: A review of variation in the Ecuadorian Highlands. Romance Philology 66:
115–45. [CrossRef]
Sampson, Rodney. 1999. Nasal Vowel Evolution in Romance. Oxford: Oxford University Press.
Sauzet, Patrick. 2012. Occitan plurals. A case for a morpheme-based morphology. In Inflection and Word Formation in Romance Languages.
Edited by Sascha Gaglia and Marc-Olivier Hinzelin. Amsterdam and Philadelphia: Benjamins, pp. 179–200.
Shosted, Ryan K. 2009. The Aeroacoustics of Nasalized Fricatives: An Instrumental Study in Phonetic Typology. Saarbrucken: VDM Verlag.
Shosted, Ryan K., Christopher Carignan, and Panying Rong. 2012. Managing the distinctiveness of phonemic nasal vowels: Articulatory
evidence from Hindi. The Journal of Acoustical Society of America 131: 455–65. [CrossRef] [PubMed]
Solé, Maria-Josep. 1992. Phonetic and phonological processes: The case of nasalization. Language and Speech 35: 29–43. [CrossRef]
Stoakes, Hywel M., Janet M. Fletcher, and Andrew R. Butcher. 2020. Nasal coarticulation in Bininj Kunwok: An aerodynamic analysis.
Journal of the International Phonetic Association 50: 305–32. [CrossRef]
Strycharczuk, Patricia, and Martin Kohlberger. 2016. Resyllabification reconsidered: On the durational properties of word-final /s/ in
Spanish. Journal of the Association for Laboratory Phonology 7: 3. [CrossRef]
Torreira, Francisco, and Mirjam Ernestus. 2012. Weakening of intervocalic /s/ in the Nijmegen Corpus of Casual Spanish. Phonetica 69:
124–48. [CrossRef]
Van Rij, Jacolien, Martijn Wieling, R. Harald Baayen, and Hedderik Van Rij. 2017. Itsadug: Interpreting Time Series and Autocorrelated
Data Using GAMMs (R Package Version 2.3). Available online: https://cran.r-project.org/web/packages/itsadug/index.html
(accessed on 15 January 2024).
Wood, Simon N. 2011. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized
linear models. Journal of the Royal Statistical Society (B) 73: 3–36. [CrossRef]
Zuazo, Koldo. 2014. Euskalkiak [The Basque Dialects]. Donostia: Elkar.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.
