Salvadoran Spanish /-s/ Aspiration and Deletion

in a Bilingual Context1
Michol F. Hoffman

1 Introduction
One of the best known features of Peninsular and Latin American Spanish
is the aspiration and deletion of syllable- and word-final /-s/. This variable
has been documented as early as the sixteenth century (Lipski 1984) and
has been investigated in many varieties of Spanish (Cedergen, Poplack,
Lipski). This variable is of interest for its social as well as linguistic conditioning. It appears in monomorphemes, e.g. menos 'less', as well as an
inflectional marker indicating person for verbs, and serves as the plural
marker, e.g., casas 'houses'. It is thus subject to a variety of phonological,
morphological, and syntactic constraints. The aspiration and deletion of
1-s/ is a feature of the Spanish ofEl Salvador, a lesser-studied variety.
This paper provides a preliminary investigation of this variable in the
speech of Spanish/English bilingual youth from El Salvador residing in
Toronto, Canada. The paper begins with a discussion of the frequency of
aspiration and deletion in three speech styles. It then offers a varbrul
analysis of social factors predicting aspiration and deletion and continues
with a Varbrul analysis of linguistic factors predicting aspiration and deletion.
1.1 The Spanish-speaking Population in Toronto
Numerous linguistic studies have examined United States Spanish as well
as issues of language contact and maintenance within Spanish-speaking
communities. That Canada, and Toronto in particular, are home to large
and vibrant Spanish speaking communities, is less well-known. Toronto's
Spanish speaking population has many countries of origin. The 1996 Census of Canada reports approximately 116,500 residents of Latin American
ethnicity in Toronto, although members of the community say that this is
a low estimate and that the number is closer to 200,000.
Though Spanish speaking Latin American immigration to Canada can
be traced from the mid-twentieth century, it expanded exponentially dur1

Many thanks to Greg Guy for his advice and support in the preparation of this paper.
U. Penn Working Papers in Linguistics, Volume 7.3 (2001)

116

MICHOL HOFFMAN

ing the early 1970s. Thousands of Chileans fleeing the military coup obtained landed immigrant status and are still a great presence within the
Spanish speaking community. Beginning in the 1970's, improved Canadian immigration policies facilitated the arrival of many immigrants from
Colombia and Ecuador. The early 1980's brought large-scale immigration
from Central America, due to the adverse socio-political conditions in El
Salvador, Nicaragua, and Guatemala. The escalation of civil war and its
consequences in El Salvador forced thousands to flee to Canada, mostly
as political and economic refugees. An estimated 34,000 Salvadorans
came to Toronto as immigrants from the eighties through 1991. Since then
the number has continued to grow. The more recent immigrants are motivated more by economic hardship and the resulting adverse social conditions than by political persecution.

1.2 The Sample

The data for this paper come from a judgment sample of eighteen first
language speakers of Salvadoran Spanish, aged fifteen to twenty three,
with varying ages of arrival and durations of stay in Canada. The majority
of the speakers come from San Salvador, the capital of El Salvador. I conducted sociolinguistic interviews with them in Spanish, consisting of three
styles: conversation, reading passage, and word list. I drew and coded approximately 9000 tokens of 1-s/ from the conversation portion of the sociolinguistic interviews for use in this paper.
This study focuses on the Spanish of youth from El Salvador for a
number of reasons: the Salvadorans are one of the largest groups within
Toronto's Spanish speaking community, and furthermore, the wave of
immigration from El Salvador is more recent than for other countries.
Since most of the youth were born in El Salvador rather than Toronto, I
thought they would be more likely to retain their use of Spanish. The issue
of language maintenance among Spanish speaking youth in Toronto is
complicated. It is one I shall attempt to address in future work. Although
the youth I interviewed claim to have strong language loyalty to Spanish,
most admit to speaking English with their siblings. All the youth have
Spanish speaking friends, both from El Salvador and from other Spanish
speaking Latin American countries, primarily Ecuador, Colombia, and
Guatemala. In general they reported that they are more likely to speak
Spanish with their friends than with their siblings, a pattern common in
situations of impending language loss. Speakers indicated that in some
cases, they spoke Spanish in order to assert their identity as Hispanic or

l-SI ASPIRATION AND DELETION IN A BILINGUAL CONTEXT

117

Latino. They also report code switching, or speaking what many of them
call "Spanglish" with friends who speak both Spanish and English.
1.3 Salvadoran Spanish

There have been comparatively few studies of Central American Spanish

compared with Caribbean and South American varieties. Canfield's
(1981) survey of Spanish American phonology offers a limited description
of features characteristic of the Spanish of El Salvador, 1-sl aspiration and
deletion among them. Lipski (1985) conducted a survey of 1-sl aspiration
deletion in the Central American capitals. His Salvadoran sample consisted of 10 adult speakers with a minimum of high school education.
Overall, Lipski's data for these speakers reveal a higher rate of overall aspiration and deletion than my data: 59%. He offered no breakdown according to social factors or individual speakers.

2 1-s/ Aspiration and Deletion According to Speech Style

In this discussion of speech style and social factors, I shall focus my
comments on all 1-sl weakening-aspiration and deletion-in contrast
with 1-sl retention. Of the three styles recorded in this sample, 1-sl weakening is highest in the least formal style, free conversation, lower in the
reading passage and lowest-almost non-existent-in the word list, the
most formal style.
As reported in Table 1, in free conversation, speakers aspirate 1-sl at a
rate of 25%, and delete at a rate of 17%; altogether /-s/ is weakened at a
rate of 42%. In more careful styles, speakers aspirate and delete /-sl at a
rate of 7% in the reading passage and 2% in the word list. This suggests
that speakers have a negative evaluation of 1-sl weakening and is consistent with the attitudes expressed in interviews by some of the speakers.
When we spoke about their opinions of different varieties of Spanish,
some speakers mentioned 1-s/ deletion and aspiration or "comerse las
eses" ["eating one's s's"] as something that other people, "from the country" or who don't speak properly do. They were consistent in the view that
it wasn't a feature of proper Spanish. Although they did identify it, along
with syllable initial l-si aspiration, e.g. nohotros [for nosotros 'we'] as
characteristic of Salvadoran Spanish.

118

All Speakers

MICHOL HOFFMAN

Conversation
42%

Reading Passage Word List

7%
2%

52%
34%

10%
4%

4%
1%

51%
40%
28%

10%

5%

4%
1%

1%

6%
6%
8%

2%
2%
3%

Men
Women
Age of Arrival
11-15
6-10

<5

Duration of StllX
33%
12-18
44%
9-10
48%
3-6 years

Table 1: /-s/ Weakening (Aspiration and Deletion) by Style

3 Social Factors
The social factors I considered for the analysis of /-s/ weakening were sex,
age of arrival in Canada, duration of stay in Canada, and parents' level of
education in El Salvador. As indicated in Table 1, men weaken /-s/ more
than women in all three styles, as do speakers with the later ages of arrival
and shorter durations of stay. This points to the hypothesis that speakers
who have lived in Canada the longest are most likely to retain word and
syllable final/-s/.
For the most part, the Varbrul analysis of these social factors reflects
the raw statistics. For this analysis, I combined the two shortest duration
factors (3-6 years and 9-10 years}, leaving two factors: speakers who have
been in Canada for 3-10 years, and for 12-18 years.
Results in Table 2 reveal that men are more likely to weaken /-s/ than
women, with factor weights of .57 for men and .42 for women. This is
consistent with other studies of 1-s/ deletion and aspiration and suggests
that women are less likely to use the more stigmatized variants. What
could account for this result? This gender difference could be reflective of
the expectations of women's behavior within the Salvadoran community:
speakers mentioned that women should be seen to be "proper" and "modest". This discrepancy could also reflect the women's desire to distance
themselves from the negative stereotype associated with young Salvadoran men. These youth, often associated with gangs, have a bad reputation within the community and in Toronto in general.

1-S/ ASPIRATION AND DELETION IN A BILINGUAL CONTEXT

119

Age of arrival was also a solid predictor of 1-s/ weakening: speakers

who arrived in Canada aged 5 years or younger discourage aspiration and
deletion with a probability of .42. For those who arrived aged 6-10 there's
a weaker effect, in the opposite direction: encouraging weakening at .52,
and those who arrived in Canada the latest-aged 11-15 are most likely to
weaken /-s/ with a factor weight of .56.
For the factor group duration of stay, speakers who have been in Canada the longest-12 years or more-inhibit weakening at .48, while the
rest of the speakers promote weakening with a factor weight of .52.
Overall, speakers who came from El Salvador the earliest, and who
have been in Canada the longest seem to not adopt or to move away from
this feature of Salvadoran speech. The factor weights for complete deletion follow the same pattern as those for weakening.
I had included another social factor group, parents' level of education
in El Salvador, as a measure of class. That factor group was not significant for the combined /-s/ weakening calculations, but it was significant
for deletion, the most stigmatized variant. Speakers whose parents attended some form of post secondary education in El Salvador are least
likely to delete 1-s/, with a factor weight of .48, and those whose parents
did not complete high school are most likely to delete 1-s/, at .52.
Probability

Sg

Men
.58
2267
.42
Women
1635
Age of arrival
11-15
.56
2097
7-10
.52
1221
.42
1-5 years
584
Duration of stay
3-10
.52
3052
12-18
.48
850
Parents' level Qf educatiQn in El Salvador
(significant for deletion only)
Elementary
.52
374
High school
.50
474
Post-secondary .48
708

52
34
51
40
28
46
33

19
17
16

Table 2: Factors predicting /-s/ aspiration and deletion (all weakening)

120

MICHOL HOFFMAN

The Varbrul calculations, along with the raw statistics, point to a

number of different issues: all the speakers have diverse friend groups,
consisting of speakers of different varieties of Spanish. Some of these varieties are non-aspirating and deleting. Exposure to other, non-/-s/ aspirating and deleting varieties combined with a negative evaluation of
Spanish from El Salvador, could cause speakers to accommodate to the
more prestigious norm of /-s/ retention. This could also be evidence of
koineization taking place in the varieties of Spanish spoken in Toronto. Of
course, it is still too early to confirm this speculation, and more research is
needed to investigate this possibility.
The speakers' feelings were mixed on whether their contact with
other varieties of Spanish has influenced the way they speak. Most indicated an awareness of the difference between their Spanish and the Spanish of their friends from other countries. In conversation many identified
Peninsular Spanish as the "best" variety. Another factor contributing to
/-s/ retention could be speakers' exposure to Standard Spanish taught in
high schools. This standard, taught as 'proper Spanish', is non /-s/ deleting.
The question of how the ethnicity of these speakers is manifested linguistically through their Spanish is complex. Generally, speakers do not
evaluate Salvadoran Spanish highly. Some even go as far as to take pride
in their impression that other Hispanics cannot tell they are from El Salvador. This alleged confusion is attributed to their' good' speech and general appearance. Speakers reported an awareness that Central Americans
in general and Salvadorans in particular, have the lowest social status
within the Spanish speaking community. This can be attributed more generally to their more recent arrival to Canada and their comparative lack of
education-many Chileans and Argentineans, for example, came as political refugees and had a high level of education-and more specifically
to stereotypes associating Salvadoran youth with gangs.
Consequently, there is a strong social motivation for Salvadoran
youth to adopt features of Spanish not associated with their country of
origin. This is at odds with the insistence of most speakers that they identify as Salvadoran first and Latino or Hispanic second. This consistent and
purposeful identification as Salvadorans is evident among speakers from
all categories of age of arrival and duration of stay. It will be interesting to
see how this tension, between their desire to assert their identity as Salvadorans-perhaps as a reaction against the negative evaluation by their
community-and their movement away from linguistic features associated
with Salvadoran Spanish, is reconciled.

l-SI ASPIRATION AND DELETION IN A BILINGUAL CONTEXT

121

4 Linguistic Factors
Investigations of 1-sl aspiration and deletion have considered variety of
phonological, morphological, and syntactic constraints. With a view to
providing an exhaustive examination of linguistic factors affecting this
variable, the preliminary analysis coded for approximately ten constraints.
Many were found to be not significant and were discarded. These included
following syllable stress and plural 1-sl type, as well as the number of
syllables in the word and a factor group designed to measure functional
effect of disambiguating information in the discourse.
This paper will discuss four significant linguistic constraints governing word final 1-sl deletion: the phonological factors 'following segment'
and 'current syllable stress', for which I shall present calculation results
both for all 1-sl weakening and for deletion; the factors of morphological
category and position of plural within the noun phrase, for which I shall
discuss results for deletion only. 2
4.1 Phonological Constraints

Previous studies have exhibited differing results with regard to the effect
of following segment on aspiration and deletion. In most varieties of
Spanish, aspiration and deletion are most frequent in pre-consonantal position as compared to following vowels and pause (Mason 1994: 153).
Lipski's (1985) Salvadoran data show high rates of aspiration and deletion
before consonants (68%) and before vowels (58%), but a much lower
weakening rate before pause (15%) (144). He provides no break down of
consonant features.
The following segment factor group initially consisted of individual
consonants. After testing for significance, I combined like-featured factors
into the following groups: obstruents except for It/, sonorants, vowels,
pause, and It/, which was kept separate as it appeared to have a distinct
effect on 1-sl on weakening. Varbrul calculations shown in Table 3 reveal
that for 1-sl weakening or combined aspiration and deletion, obstruents
(both voiced and voiceless) other than It/, and sonorants promote aspiration and deletion. In contrast, following It/ inhibits strongly. The effect of
following It/ on aspiration, promoting 1-sl retention, is expected as It/ and
lsi are homorganic and maximally similar. The aspiration qualities associ2
This paper discusses 1-s/ aspiration and deletion in word- final position only.
Phonological constraints on word internal aspiration and deletion in this corpus
were similar.

122

MICHOL HOFFMAN

ated with /tl are acoustically or phonetically associated with /s/. In articulatory terms, the tongue has to pass through an /sf-like position on its way
to articulating /tl. This is not the case for other consonants. Following
vowels promote aspiration and deletion, albeit very weakly. Following
pause exhibits a strong effect discouraging all weakening.
Probability
Following Segment
.71
obstruents
sonorants
.74
vowels
.51
.28
pause
Cun:ent S~llable Stre~s
.45
stressed
.55
unstressed

1346
724
819
297

63
67
44
23

908
2348

42
51

Table 3: Phonological Factors Predicting Word Final/-s/ Weakening (Aspiration and Deletion Combined)
The calculations for deletion only (Table 4), reveal a definite voicing
effect, with voiced obstruents (.72) and sonorants (.65) encouraging deletion, and all voiceless obstruents discouraging with a factor weight of .33.
Guy (1996) found the same result for following consonant voicing on 1-sl
deletion in Buenos Aires Spanish. This difference in obstruent voicing
between alli-s/ weakening and /-s/ deletion only is due to the general effect of voiceless obstruents on aspiration: the non-coronal voiceless obstruents promote aspiration and this effect is absent when we consider
deletion alone. Following vowels and following pause discourage deletion
as they did for combined aspiration and deletion.
The effect of following vowel for deletion only is consistent with
other studies of deletion in other varieties of Spanish (e.g., Poplack 1980,
Guy 1996), but the results for following pause are generally inconsistent
(cf. Poplack 1980, Terrell 1975). Lipski's Salvadoran data show fewer
instances of weakening before pause, but more weakening before vowels
than in my sample.
With regard to syllable stress, For both aspiration and deletion combined, and for deletion alone, current stressed syllables discourage deletion and unstressed syllables encourage deletion. This is consistent with
the general findings that stressed syllables are more resistant to reduction
than unstressed syllables. Following syllable stress, a significant con-

l-SI ASPIRATION AND DELETION IN A BILINGUAL CON1EXT

123

straint in the study of 1-s/ deletion for other varieties of Spanish, was
found not significant in this sample. 3
Probability
%
N
Following Segment
voiceless obstruents .33
216
13
voiced obstruents
.72
306
42
sonorants
.65
389
36
vowels
.40
319
17
pause
.39
206
16
Current Syllable Stress
stressed
.40
318
15
unstressed
.60
1118
24
Table 4: Phonological Factors predicting Word Final /-s/ Deletion

4.2 Morphological Constraints

Initial coding for morphology included an extremely long and elaborate
list of factors that distinguished between different verb forms, pronouns,
adjectives, determiners, and quantifiers, as well as number of common
lexical items.
After testing for significance, I refined the analysis by combining like
factors into categories exhibiting the same effect on deletion. These
recodes left the following factors: nominals, which includes modifying
adjectives, determiners, and some plural pronouns; predicate adjectives;
quantifiers, including todos, muchos, pocos; the pronoun nosotros;4
mono morphemes; object and indirect object pronouns las and les, 5 and
two frequent lexical exceptions, entonces and pues, combined.
Results in Table 5 show that monomorphemes inhibit deletion with a
strong factor weight of .32. Both verbal and nominal 1-s/, which include
many examples of inflectional I-s/, discourage deletion, nominals with a
factor weight of .32, and verbs with a somewhat weaker effect at .44.
Predicate adjectives have a neutral effect at .50. This effect could due to
their distance from the plural NP with which they agree. They may require

Although not significant, following stressed syllable and unstressed syllable discouraged aspiration and deletion weakly in contrast to other studies where following unstressed syllable favored deletion.
4
Nosotros did not produce an effect that lent itself to combination with other
factors.
5
This factor consisted mainly of the oblique pronoun les.

124

MICHOL HOFFMAN

some distinctive syntactic operation to be marked. Quantifiers stand out as

a group encouraging deletion with a factor weight of .61. This could be
attributed to their position, coming before determiners, possibly not within
the domain of the NP. The combined factor les and las, consisting largely
of the indirect pronoun les, also promotes deletion. This higher rate of
deletion could be attributed to case marking, with the oblique losing plural
marking first. This already occurs in Spanish: the indirect object pronoun
se, when used in conjunction with direct object pronouns, can have a singular or plural meaning. It has also lost a gender distinction, e.g., context
determines whether Se lo di means "I gave it to her/him or I gave it to
them."
The lexical exceptions, entonces and pues have the most powerful effect promoting deletion with a factor weight of .71. This supports the theory that these extremely frequent discourse markers have multiple underlying representations-with and without /-s/ (Guy 1996, 2000).

%
Probability
N
Verbs
.44
333
23
Nominals6
.32
551
18
Predicate adjectives .50
43
31
Las!les as pronouns .61
64
45
.61
99
45
Quantifiers
Nosotros
.50
35
36
Monomorphemes
.32
173
13
Entonces, pues
.71
138
51
Table 5: Morphological Factors predicting Word Final /-s/ Deletion
4.3 Syntactic Constraints
With regard to the effect of position in the NP on deletion, or absence of
plural marker, one can discern a clear pattern: generally, deletion is more
likely farther to the right in the noun phrase. As reported in Table 6, positions 1 and 2 inhibit deletion with factor weights of .39 and .40, and positions 3 and 4 encourage it at .57 and .64. When we consider preceding
plural marker with position, words in second position preceded by a plural
maker are least likely to delete. This is true of words preceded by both
plural marked nominals as well as by numbers. In contrast, words in second position preceded by no plural marker favor deletion. Words in third

Includes nouns, modifying adjectives, determiners, plural marked pronouns

l-SI ASPIRATION AND DELETION IN A BILINGUAL CONTEXT

125

and fourth position 7 preceded by only plural marked forms discourage

deletion, though to a lesser degree than second position, at .44. Third and
fourth position preceded by no plural marked forms encourages deletion
strongly at .76. Third and fourth position preceded by mixed plural
marked and deletion has a neutral effect on deletion. Thus, the presence of
a preceding plural favors retention, and the absence of a preceding plural
encourages deletion. This result is similar to other studies of 1-sl deletion
(Poplack 1980, Guy 1996): redundancy seems to be favored within the
noun phrase and mixed antecedents have no effect. These results could
also be evidence of a micro-stylistic effect in the three-word NPs: if the
speaker begins the NP with with a plural marker, he or she will continue
to use plural marking.
Probability

.39
.40
.57

482
173
41
9

22
18
29
32

482
121
52
12
21
17

22
14
40
20
28
53

Position in NP
1

2
3

4
.64
Position Preceded by Plural Marker

I
2 prec. by pl.
2 prec. by 0
3+4 prec. by pl.
3+4 prec. by mixed
3+4 prec.by 0

.35
.31
.61
.44
.50
.76

Table 6: Linguistic Factors predicting Word Final/-s/ Deletion

5 Discussion
These results raise several questions with regard to linguistic and social
constraints on 1-sl aspiration and deletion in Salvadoran and other varieties of Spanish. Lipski ( 1985) suggests that since the aspiration and deletion of /-sf is not as advanced as in other varieties, Salvadoran Spanish offers an opportunity to examine the variation in earlier stages and look at
the "spread of the weakening process" (148). With regard to phonological
constraints, several findings, such as the universal conservative effect of
current syllable stress in this study, confirm previously documented con-

I combined 3ro and 4'h positions because there were so few tokens for 4th position.

126

MICHOL HOFFMAN

straints on the variable. The import of others, e.g., the effect following
pause, has yet to be established.
Morphological category is another area that begs for further investigation. Guy (2000) provides evidence for multiple underlying representations for lexical exceptions. The data for lexical exceptions in this corpus
are a good fit for that model. There is also the question of functional motivation for the retention of the plural marker. Both inflectional categories,
nominals and verbs, discourage deletion. Nonetheless monomorphemes,
where discourse ambiguity is not at issue, show higher rates of deletion
and similar factor weights to the inflectional categories.
This paper provided evidence of the state of 1-sl aspiration and deletion in the Spanish of young Salvadoran immigrants to Toronto, Canada.
Of course, more research is needed to shore up these preliminary observations. More specifically, it would be revealing to expand this study to include young speakers of other /-s/ weakening varieties, as well as more
Salvadoran youth, more closely associated with groups perceived as socially marginal. The continued effect of duration of stay should also be
monitored. I predict that changes to this variable are largely dependent on
social factors and on the situation and status of Spanish spoken in Toronto. Only time and more research will tell.

References
Canfield, D. Lincoln. (1981). Spanish Pronunciation in the Americas. Chicago:
The University of Chicago
Cedergren, Henrietta. (1973 ). The Interplay of social and linguistic factors in Panama. Unpublished Ph.d dissertation, Cornell University.
Guy, Gregory R. (1996). The general and the particular: Constraints on /s/ deletion
in Argentine Spanish. Paper presented at NW AVE 25, Las Vegas.
Guy, Gregory R. (2000). Modelling Lexical Exceptions to Phonological Processes.
Paper presented at 4"' Bilingual Workshop on Theoretical Linguistics, Toronto.
Labov, William.(l994). Principles of Linguistic Change. Vol. 1: Internal Factors.
Oxford: Blackwell.
Lipski, John M.(l985). "lsi in Central American Spanish". Hispania 68. 143-49.
Mason, Keith W. (1994). Comerse las eses: a selective bibliographic survey of Is/
aspiration and deletion in dialects of Spanish. Unpublished Ph.D. dissertation,
University of Michigan.
Poplack, Shana. ( 1980). "The notation of the plural in Puerto Rican Spanish: Competing restraints on Is/ deletion". in Locating language in time and space.
William Labov, ed 55-67. New York: Academic Press.

l-SI ASPIRATION AND DELETION IN A BILINGUAL CONTEXT

127

Terrell, Tracey. (1975) La interacci6n de Ia aspiraci6n y Ia elisi6n sobre Ia /s/ implosiva y fial en el espaiiol de Puerto Rico. Manuscript. Irvine: University of
California.
Terrell, Tracey. (1980). "Diachronic reconstruction by dialect comparison of variable constraints: s-. aspiration and deletion in Spanish". in Variation Omnibus.
David Sankoff and Henrietta Cedergren, eds. 115-125. Edmonton: Linguistic
Research.

346 Brunswick A venue

Toronto, Ontario
M5R2Y9
Canada
[email protected]