Research Article
Dialectal Effects on a
Clinical Spanish Word Recognition Test
Lu-Feng Shia and Luz Adriana Canizalesa
Purpose: American Spanish dialects have substantial phonetic
and lexical differences. This study investigated how dialectal
differences affect Spanish/English bilingual individuals’
performance on a clinical Spanish word recognition test.
Method: Forty Spanish/English bilinguals participated in the
study—20 dominant in Spanish and 20 in English. Within each
group, 10 listeners spoke the Highland dialect, and 10 spoke
the Caribbean/Coastal dialect. Participants were maximally
matched between the 2 dialectal groups regarding their
demographic and linguistic background. Listeners were
randomly presented 4 lists of Auditec Spanish bisyllabic
words at 40 dB SL re: pure-tone average. Each list was
randomly assigned with a signal-to-noise ratio (SNR) of quiet,
+6, +3, and 0 dB, in the presence of speech-spectrum noise.
Listeners responded orally and in writing.
I
n the past decade, clinicians who work in urban areas
with a large number of Spanish-speaking residents have
seen the pressing need for research focused on this
fast-growing population. Data regarding proper hearing
evaluation and aural rehabilitation for these individuals have
been identified to be highly critical for evidence-based
practice (American Speech-Language-Hearing Association
[ASHA], 2009).
For research in the field of Spanish speech audiometry,
the issue of dialectal difference has thus far not been the
center of clinical investigation. Spanish is widely used in
America, and many dialects exist. Phonetically, American
Spanish demonstrates many diverse characteristics (Canfield,
1981; Dalbor, 1997), and these dialectal differences may
contribute to perception errors. Although a detailed com-
parison of these characteristics is beyond the scope of the
current report, key phonetic differences across American
Spanish, as summarized in Caballero, Moreno, and
Nogueiras (2009), are re-tabulated here in Table 1.
Traditionally, American Spanish can be categorized into the
a
Long Island University–Brooklyn Campus, NY
Correspondence to Lu-Feng Shi: lu.shi@liu.edu
Editor: Larry Humes
Received June 30, 2012
Accepted August 23, 2012
DOI: 10.1044/1059-0889(2012/12-0036)
74 American Journal of Audiology N Vol. 22 N 74–83 N June 2013 N ß American Speech-Language-Hearing Association
Results: Dialect and language dominance both significantly
affected listener performance on the word recognition test.
Higher performance levels were obtained with Highland than
Caribbean/Coastal listeners and with Spanish-dominant than
English-dominant listeners. The dialectal difference was
particularly evident in favorable listening conditions (i.e., quiet
and +6 dB SNR) and could not be explained by listeners’
familiarity with the test words.
Conclusion: Dialects significantly affect the clinical
assessment of Spanish-speaking clients’ word recognition.
Clinicians are advised to consider the phonetic features of the
dialect when scoring a client’s performance.
Key Words: word recognition, Spanish/English bilingual,
dialect, linguistic background, word familiarity
Highland versus Coastal group due to early Castilian
settlement in the former area, followed by settlement of
people from southern regions of Spain in the latter area.
Of the two groups, the Coastal dialect is less faithful to
‘‘standard’’ Castilian Spanish.
Over the centuries, region-specific dialects have
evolved from these two groups, resulting in a liberal
phonetic-orthographical system. Dalbor (1997) suggested
five main dialectal groups: Mexican, Central American,
Caribbean (including Panama as well as coastal areas of
Colombia and Venezuela), Highland (mainly Colombia,
Ecuador, Peru, Bolivia, and Chile), and Argentinean.
Phonetic variances across these dialects are abundant. For
example (perhaps an oversimplified one), the Castilian
phoneme /ʎ/ for the orthographic ‘‘ll’’ may be somewhat
maintained in American Highland Spanish but has under-
gone ‘‘z(h)eı́smo’’ (/ʒ/) in Rioplatense Spanish (spoken in
Argentina, Paraguay, and Uruguay) or ‘‘yeı́smo’’ (/ʝ/) in most
of the other American Spanish dialects (e.g., the word calle is
pronounced as /kaʒe/ or /kaʝe/ in most American Spanish
dialects instead of /kaʎe/, as in Castilian Spanish). Another
example may be ‘‘seseo/ceceo,’’ where only Castilian Spanish
retains the dental /h/ (‘‘ceceo’’), but all American Spanish
dialects have undergone ‘‘seseo’’ where /h/ is replaced by /s/.
Compared to other American Spanish dialects,
Caribbean Spanish has developed many unique features
that are not typically seen in other variations. First, /s/ in
Caribbean Spanish tends to be aspirated (e.g., the word más
Table 1. Main phonological characteristics of the Highland versus
Caribbean/Coastal American Spanish dialects (adapted from
Caballero, Moreno, & Nogueiras, 2009).
Characteristic Highland Caribbean/Coastal
Seseo + +
Yeı́smo +a +
Aspiration of /s/ – +
Velarization of /n/ – +
Interchange of /l-r/ – +
Elision of /d/ (colloquially) – +
Interchange of s-z – +
(orthographically)
Note. + means that a feature is present in the dialect; – means that
a feature is not present in the dialect.
a
Caballero et al. (2009) considered yeı́smo to be a feature of the
Highland dialect. However, according to other sources (Canfield,
1981; Dalbor, 1997), some Highland regions may retain lleı́smo.
may be pronounced as /ma/ instead of /mas/). Second, there
is a velarization process of the nasal /n/, resulting in /ŋ/ (e.g.,
the word manta may be pronounced as /maŋta/ instead of
/manta/). Third, syllable final /l/ is accepted in lieu of /r/ (e.g.,
the word carta may be pronounced as /kalta/ instead of
/karta/; orthographically, the word may be written as
‘‘calta’’). Fourth, in colloquial speech, elision of /d/ is
common (e.g., the word edad may be pronounced as /eda/
instead of /edad/). Last, due to the effect of leveling, no
distinction between orthographic ‘‘s’’ and ‘‘z’’ is typically
made (e.g., the word mesa may be written as ‘‘meza’’).
These considerations may present serious challenges
for clinicians who are not proficient in Spanish or are not
familiar with its dialects. Indeed, Caballero et al. (2009)
found that a voice recognition system using Castilian
Spanish as the standard dialect failed at a high error rate
when trying to recognize the same words spoken in various
American Spanish dialects. In addition, this monodialectal
model erred more in recognizing speakers of a Mexican and
Highland Colombian dialect than those of an Argentinean
and Caribbean dialect. Multidialectal systems that include
phonetic variations as shown in Table 1 improved their
overall rate of correct recognition.
American Spanish dialects can also differ in their
lexicon. Some words may have different meanings or be used
in different contexts in different regions. For example, pipa
means ‘‘pipe’’ in Spanish, but Puerto Ricans may associate
the word with ‘‘belly fat.’’ The word grifo means ‘‘faucet’’ in
Spanish, but in some countries (e.g., Peru), the word is often
used to refer to the nozzle of a gas pump; to refer to the
faucet, llave (the handle of the faucet) is used instead. On the
other hand, one construct can be represented by different
words or phrases. The construct of ‘‘bus,’’ for instance, may
be expressed as camión (Mexico), busito (Central American
and some areas in Mexico), guagua (Caribbean nations such
as Cuba, Dominican Republic, and Puerto Rico), colectivo
(Argentina), micro (Chile), and many other words depending
on the nation and region.
Furthermore, communities of different American
Spanish origins differ in their culture and degree of exposure
to other cultures and languages. In the New York City area,
the largest Spanish American population has traditionally
been of the Caribbean origin, including Puerto Rican,
Dominican, and others (Canfield, 1981). In recent decades,
communities of South American origins, such as Colombian
and Peruvian, are booming. Residents of Caribbean
American communities oftentimes have been exposed to
English since childhood. American English, one of many
characteristics of the American culture, is prevalent in
Caribbean nations due to their geographic and economic ties
with the United States. Consequently, individuals growing up
in these communities could have developed bicultural identity
(Arnett, 2002) as well as advanced English language skills.
By contrast, American culture has relatively less influence
over South American nations, although the influence of pop
culture from the United States may have increased in recent
years. Individuals who are open to foreign cultures or who
view English learning as an integrative part of globalization
may develop English language skills more eagerly than those
who regard American culture as a threat to the native culture.
Given these phonetic, lexical, and cultural differences,
it seemed imperative to examine current Spanish word
recognition tests in light of dialectal variations. Findings
from studies focused on English have provided unequivocal
evidence that dialect must be considered when listener
performance is analyzed (e.g., Wilson & Moodley, 2000).
However, few studies have considered potential dialectal
effects while interpreting and/or analyzing their data
(Schneider, 1992; Shi & Sánchez, 2010, 2011; Weisleder &
Hodgson, 1989). For example, Weisleder and Hodgson
(1989) conducted a study of 16 native Spanish-speaking
participants from different countries including Mexico,
Panama, Venezuela, Honduras, Colombia, and Spain. The
participants were presented with four lists of 50 Auditec
bisyllabic Spanish words spoken by a male Mexican talker.
The results of the test scores indicated that the listeners of
Mexican origin obtained higher percentage correct scores
than the listeners originating from other countries.
Like Weisleder and Hodgson (1989), most previous
studies were conducted based on Mexican Spanish (Berruecos
& Rodrı́guez, 1967) or mainly recruited bilingual college
students to be the listeners, most of whom were reported to
be of Mexican origin (Mexican/total participants = 5/10,
Cokely & Yager, 1993; 15/20, Flores & Aoyama, 2008; 9/16,
Weisleder & Hodgson, 1989), therefore reducing the
generalizability of the findings to bilinguals of other
nationalities (Dalbor, 1997). Mexican immigrants comprise
the largest Spanish-speaking population in the United States
(30 million), but there are other nationalities with >1 million
immigrants in this country, including Cuba, El Salvador, the
Dominican Republic, Guatemala, and Puerto Rico (U.S.
Census Bureau, 2010). Spanish used in many of these nations
has its nation- or region-unique phonetic and lexical features
(Caballero et al., 2009; Canfield, 1981; Dalbor, 1997).
The current study was designed to explore the
possible effects of dialectal variations on recognition of
Shi & Canizales: Spanish Dialect and Word Recognition 75
Spanish words. In particular, two groups of native
Spanish-speaking participants who differed in their dialect
(Highland, mainly Colombian, vs. Caribbean/Coastal,
mainly Dominican and Puerto Rican) were included.
These two dialects have significant phonetic and lexical
differences (Caballero et al., 2009; Canfield, 1981; Dalbor,
1997); thus, it was surmised that dialectal effects on a word
recognition test, if any, would most likely be captured by
comparing the performance of individuals speaking these
two dialects.
Because most Spanish-speaking individuals in the
United States are bilingual (Kayser, 1995), we also examined
listener performance according to their dominance in
language (Spanish vs. English). This aspect was important,
especially given the fact that many Caribbean/Coastal
individuals have substantial exposure in English and consider
themselves English dominant. Comparison of their perfor-
mance on a Spanish word recognition test with individuals of
a Highland background but Spanish dominance could
overestimate the magnitude of the dialectal effect (due to the
confound of language dominance). As such, results pertain-
ing to differences between the two dialects, Highland versus
Caribbean/Coastal areas, were examined with regard to
bilinguals’ dominant language.
Additionally, potential difference in word recognition
could be mediated by dialectal-specific phonetic and/or
lexical characteristics. To tease out possible lexical contri-
bution, we obtained listeners’ familiarity rating of the test
words. We surmised that listeners of Caribbean/Coastal
versus Highland dialect might prescribe different ratings to
test words and that higher familiarity might lead to improved
performance.
Method
Participants
Forty normal-hearing, adult, native Spanish listeners
participated in this study. Normal hearing was defined as
thresholds ≤20 dB HL at octave frequencies 250–8000 Hz
(American National Standards Institute [ANSI], 2004). All
participants were currently living in the greater New York
City area. They belonged to two dialectal groups (Canfield,
1981; Dalbor, 1997). Half of them came from areas where the
Highland dialect is spoken, including the Andean regions
of Colombia, Ecuador, Bolivia, and Peru. All had had
significant experience with English. Ten of the 20 Highland
participants regarded themselves as dominant in Spanish,
and the other 10 regarded themselves as dominant in English.
The second group of 20 participants came from areas where
the Coastal/Caribbean dialect is spoken, including the
Dominican Republic, Puerto Rico, and the Caribbean
coastal region of Colombia. Similar to the Highland group,
half of the 20 Caribbean/Coastal participants regarded
themselves as dominant in Spanish, and the other half
regarded themselves as dominant in English. Thus, four
groups of participants who differed in dialectal region
(Highland vs. Caribbean/Coastal) and language dominance
(Spanish vs. English) were included in this study.
Information regarding the participants’ demographics
and language background as obtained through the Language
Experience and Proficiency Questionnaire (LEAP-Q;
Marian, Blumenfeld, & Kaushanskaya, 2007) is summarized
for each group in Table 2. To facilitate visualization, the two
dialectal groups are shown in adjoining columns for the
same dominant language. All four groups of listeners had

Table 2. Study participants’ demographic and linguistic background information.

Spanish dominant English dominant

Variable Caribbean/ Coastal Highland Caribbean/ Coastal Highland

Age (years) 28.60 ¡ 7.32 31.30 ¡ 3.27 26.70 ¡ 3.06 27.90 ¡ 7.42
Gender (female/male) 8/2 4/6 4/6 6/4
Originality (number of participants) Coastal Colombia (2) Andean Colombia (9) Dominican Republic (6) Andean Colombia (7)
Dominican Republic (8) Ecuador (1) Puerto Rican (4) Bolivia (1)
Ecuador (1)
Peru (1)
Current daily exposure in Spanish (%) 53.50 ¡ 18.86 46.30 ¡ 16.57 39.00 ¡ 22.21 32.50 ¡ 18.45
Length of residence in a 15.46 ¡ 5.87 15.50 ¡ 6.46 2.08 ¡ 4.30 3.73 ¡ 3.71
Spanish-speaking country (years)
Length of residence in a 24.34 ¡ 8.04 30.18 ¡ 4.01 26.48 ¡ 3.12 25.70 ¡ 9.59
Spanish-speaking family (years)
Length of overall education (years) 16.00 ¡ 0.67 14.90 ¡ 2.60 16.20 ¡ 0.42 14.90 ¡ 2.73
Length of education in Spanish (years) 14.38 ¡ 4.20 14.50 ¡ 7.03 6.20 ¡ 6.21 7.60 ¡ 7.04
Self-rated proficiency in Spanish (0–10)
Listening 9.30 ¡ 0.48 9.50 ¡ 0.53 8.50 ¡ 0.97 8.50 ¡ 0.97
Speaking 8.90 ¡ 0.74 9.50 ¡ 0.53 6.90 ¡ 1.60 7.50 ¡ 1.72
Reading 8.80 ¡ 1.03 9.10 ¡ 0.99 6.80 ¡ 1.48 6.70 ¡ 2.16

76 American Journal of Audiology N Vol. 22 N 74–83 N June 2013

comparable age, F(3, 36) = 1.180, p = 0.331, and gender
composition (Fisher’s exact test, p = 0.170). All groups were
currently exposed to Spanish to a comparable degree on a
daily basis, F(3, 36) = 2.252, p = 0.099.
In terms of language background, age of Spanish
acquisition is not listed because all participants reported
Spanish as their native language and started learning the
language at birth. Their exposure in Spanish was measured in
terms of their length of residence in a Spanish-speaking
country as well as in a Spanish-speaking family. Because the
absolute values of these variables might not fairly represent
the extent of exposure (e.g., for a 20-year-old, 20 years of
living in a Spanish-speaking family might mean ‘‘pure’’
exposure in that language, whereas for a 40-year-old, the
same length of exposure suggests that half of the individual’s
activities at home involve a language other than Spanish),
these values were divided by the biological age for each
participant (Weiss & Dempsey, 2008). Nonetheless, data
treated in either the raw form or the ratio yielded similar
results: When the two dialect groups (Caribbean/Coastal or
Highland) were compared within one dominant language
(English or Spanish), no statistical differences (p > 0.05) were
found for any of the exposure parameters. As expected,
however, English-dominant participants in either dialect
group had resided in a Spanish-speaking country for a
significantly shorter time period than Spanish-dominant
participants (one-way ANOVA followed by two-sample
t tests, p < 0.001). English-dominant participants in the
Caribbean/Coastal group had a somewhat greater exposure
to Spanish in the family than their Spanish-dominant
counterparts (p = 0.035); however, this difference did not
achieve statistical significance with Bonferroni correction for
multiple-pair comparisons (p = 0.05/4 = 0.0125).
Participants were compared in their education back-
ground. Although the Caribbean/Coastal group appeared to
have more education than their Highland counterparts, the
two dialectal groups did not differ for their overall length of
education F(3, 36) = 1.312, p = 0.286. Participants were also
compared in their education in Spanish. In this respect,
significant differences were found between English- and
Spanish-dominant participants within the Caribbean/Coastal
group (p = 0.002) but not within the Highland group with
Bonferroni correction (p = 0.042). No difference was
found between the Caribbean/Coastal and Highland groups
(p > 0.05).
Self-ratings were obtained for listening, speaking, and
reading proficiency in Spanish (Marian et al., 2007). These
ratings were based on an 11-point scale (0 = not proficient
at all to 10 = perfectly proficient). Significant differences
(with Bonferroni correction for each language domain,
p = 0.05/4 = 0.0125) were found between English- and
Spanish-dominant participants for the Caribbean/Coastal
group (speaking: p = 0.002; reading: p = 0.002; except for
listening: p = 0.032) and Highland group (listening: p =
0.010; speaking: p = 0.002; reading: p = 0.005), but not
between Caribbean and Highland participants within
either the English- or Spanish-dominant group (p > 0.05
in all cases).
In summary, participants were demographically
similar but did differ in their language experience and
proficiency. However, the differences existed exclusively
between those dominant in Spanish versus in English, but
not between those speaking a Highland dialect versus a
Caribbean/Coastal dialect.
Stimuli
The Auditec recording of the bisyllabic Spanish
word recognition test was employed as the test stimuli. This
test has four lists of 50 words (200 in total) that are spoken
by a male talker with a Mexican origin (Weisleder &
Hodgson, 1989). This test is among the most widely used
Spanish speech materials in hearing clinics (Flores &
Aoyama, 2008), and its psychometric characteristics, such as
interlist equivalency (Weisleder & Hodgson, 1989), perfor-
mance-intensity function (Flores & Aoyama, 2008; Shi &
Sánchez, 2010; Weisleder & Hodgson, 1989), and word
familiarity (Shi & Sánchez, 2011), have been relatively well
investigated.
Procedure
Listeners were tested diotically in quiet and in noise.
For noise, three signal-to-noise ratios (SNRs, +6, +3, and
0 dB) were selected, largely based on previous findings (Shi
& Sánchez, 2010, 2011). These four listening conditions were
presented to each listener in a random order. For each
condition, a word list was randomly assigned and was
presented at 40 dB SL re: pure-tone average (PTA) of 500,
1000, and 2000 Hz through the GSI-61 audiometer
(Grason-Stadler) and a pair of TDH-50 headphones
(Telephonics). For noise conditions, speech-weighted noise
was generated by the audiometer and was presented at the
specified SNR.
Listeners orally repeated every word they heard and
also wrote them down. Credit was given if the response was
correct in either form. For both forms, all of the dialect-
sensitive characteristics as summarized earlier were observed,
especially within the Caribbean/Coastal group. In these
cases, credit was given if the response was consistent with the
dialectal feature. For example, if a Dominican listener
pronounced or wrote ‘‘calta’’ in lieu of carta, this item was
considered as correctly recognized. The same response,
however, would be considered incorrect for a Peruvian
listener. Responses that clearly related to spelling issues were
never counted as wrong (e.g., missing or substituting a letter
with another letter that shares the same phoneme, ‘‘ilo’’
for hilo, ‘‘hardin’’ for jardin, or ‘‘varca’’ for barca).
The second author, who is fluent in both Spanish and
English, conducted the experiment. She originally came from
Highland Colombia but also had significant exposure and
training in the Caribbean/Coastal dialect. Verbal instruction
was provided in Spanish and English, following the specifics
in Shi and Sánchez (2010, 2011). The first author was an
elective learner of Spanish, and his knowledge of Spanish
was not extensive beyond the phonetics/phonology of the
language; however, he had considerable experience working
Shi & Canizales: Spanish Dialect and Word Recognition 77
with Spanish-speaking individuals with normal hearing or
hearing loss and was highly familiar with the test in question.
The first author was present in 50% of the experiments, and
in these cases, he participated in scoring. Interexperimenter
agreement was high (>95%). Scoring discrepancies typically
occurred in cases of spelling issues and were reconciled
through discussions between the two authors or consultation
with a simultaneous Spanish/English bilingual phonologist/
speech-language pathologist (SLP) who had decades of
experience working with Spanish-speaking individuals of
different backgrounds.
Once the listening part of the experiment concluded,
the listeners were asked to review a list of all test words and
rate their familiarity with them on a 1–7 scale (1 = do not
know the word at all to 4 = have heard of the word, but do not
know how to use it to 7 = know the word very well), developed
by Nusbaum, Pisoni, and Davis (1984) that had previously
been used by Shi and Sánchez (2011).
Results
Percentage correct scores were transformed to ratio-
nalized arcsine units (RAUs; Studebaker, 1985) due mainly
to the near-perfect performance in the quiet condition across
all groups. A three-way repeated measures analysis of
variance (ANOVA) was employed to analyze the data, with
language dominance (Spanish and English) and dialect
(Caribbean/Coastal and Highland) as the two between-
subjects factors and SNR (quiet, +6 dB SNR, +3 dB SNR,
and 0 dB SNR) as the within-subjects factor. Results
indicated that all three main effects were significant
(Table 3): language dominance, F(1, 36) = 26.440, p < 0.001,
ŋp2 = 0.423; dialect, F(1, 36) = 6.380, p = 0.016, ŋp2 = 0.151;
and SNR, F(3, 108) = 560.347, p < 0.001, ŋp2 = 0.940. None
of the two- or three-term interactions was significant. That is,
relating to the main purpose of the study, Highland listeners
achieved a higher score on the test than Caribbean/Coastal
listeners, regardless of their dominant language and/or test
conditions.
In clinical practice, it is important to know how much
difference is to be expected between the two dialectal groups
at each SNR. To this end, Figure 1 plots the difference
between Caribbean/Coastal and Highland listeners as a
function of SNR for the English-dominant (left panel) versus
Spanish-dominant (right panel) groups. The dialectal groups
performed at a comparable level for the two low SNRs;
however, for the two favorable conditions (quiet and +6 dB
SNR), Highland listeners’ performance was higher than
Caribbean/Coastal listeners’. Indeed, although the two
dialectal groups yielded comparable performance levels for
other SNRs, pairwise t tests indicated a higher average score
for Highland than Caribbean/Coastal listeners in quiet
(114.51 vs. 100.84 RAUs) for the English-dominant groups
(p = 0.002, significant at Bonferroni corrected p = 0.05/4 =
0.0125). In the same vein, Highland listeners significantly
outperformed their Caribbean/Coastal counterparts at +6 dB
SNR (87.48 vs. 77.20 RAUs) for the Spanish-dominant
groups (p < 0.001). These results suggest that dialectal
difference existed when listening conditions were not severely
compromised and that the presence of substantial noise was
sufficient to ‘‘level’’ out any dialectal effects on the test.
To explore whether language background might
possibly confound this conclusion, the variables listed in
Table 2, with the exception of age and gender, were entered,
individually or in aggregate, into the three-way repeated
measures ANOVA model as covariates. None of these
covariates reached statistical significance, suggesting that
language variables (Table 2) did not account for variances in
performance that were not due to the three main effects. In
addition, their presence in the model reduced the significance
of the language dominance factor to p > 0.05, whereas the
dialect factor remained statistically significant. Evidently,
language background provided information redundant to
dominance (Table 2) but was not a confounder for findings
that Highland and Caribbean/Coastal listeners performed
differently on the test.
Finally, to estimate lexical contribution to differences
in performance, we analyzed listeners’ familiarity with the
test words. Due to a clerical error, one English-dominant
Highland listener’s familiarity rating was not properly
recorded. The average word familiarity rating for the
remaining 39 listeners across groups is summarized in
Table 4. In general, the listeners appeared to be familiar with
the test words, as the average rating for each group was >6.6
out of 7; however, variability was much larger with the
English- dominant than Spanish-dominant listeners, sug-
gesting that the Spanish vocabulary was not homogeneously
large in the former group. Pairwise t tests indicated that the
English-dominant listeners overall had lower ratings than

Table 3. Main effects (language dominance, dialect, and signal-to-noise ratio, SNR) of the analysis of
variance. Shown is the average performance for each level within the effect as well as the level of statistical
significance.

Main effect Condition (performance in RAUs) p value

Dominance English (64.37) < Spanish (73.99) <0.001

Dialect Caribbean/Coastal (66.82) < Highland (71.54) 0.016
SNR 0 dB SNR (33.20) < +3 dB SNR (54.21) <0.001 for all pairwise
< +6 dB SNR (77.09) < Quiet (112.22) comparisons

Note. RAUs = rationalized arcsine units.

78 American Journal of Audiology N Vol. 22 N 74–83 N June 2013

 Figure 1. Word recognition in rationalized arcsine units in English-dominant (left panel) and Spanish-
dominant (right panel) listeners with Caribbean/Coastal dialect (CD) and Highland dialect (HD) in four listening
conditions. Error bars represent one standard error of the mean.
the Spanish-dominant listeners, which is consistent with the
findings reported in Shi and Sánchez (2011). Significant
difference, however, was not seen between the dialects,
suggesting that word familiarity may explain the perfor-
mance difference between English- and Spanish-dominant
listeners, but it was not a contributor to that between
Caribbean/Coastal and Highland listeners. This notion was
further confirmed by including familiarity rating as a
covariate in the ANOVA. Similar to language background
variables, familiarity did not significantly account for
variance in RAU, F(1, 34) = 0.020, p = 0.888. Hence, it is
deduced that differences in performance between the
Caribbean/Coastal and Highland groups were most likely
the result of dialectal differences in phonetics.
Discussion
It is well known that American Spanish not only differs
substantially from Castilian Spanish, but it also has phonetic,
lexicosemantic, and morphosyntactic differences among its
dialectal denominations (Canfield, 1981; Dalbor, 1997).
Clearly, audiologists and other hearing health care profes-
sionals must be aware of these differences when evaluating
and treating clients who are American Spanish speakers. The
Table 4. Intergroup comparison in familiarity rating (M ¡ SD) of the
Audiect bisyllabic Spanish words.
Caribbean/Coastal Highland p value
Spanish dominant 6.87 ¡ 0.08 6.93 ¡ 0.06 0.072
English dominant 6.65 ¡ 0.21 6.69 ¡ 0.18 0.642
p value 0.005 <0.001
current study provides data to guide clinical practice
concerning clients who are speakers of one of two American
Spanish dialects: South American Highland versus
Caribbean or Coastal South America. It was found that, on a
widely used clinical Spanish word recognition test, indivi-
duals of the Highland background performed significantly
better than those of the Caribbean/Coastal background,
whether they were dominant in Spanish or English.
Listener Performance and Previous Studies
Three studies have obtained the performance-intensity
function of the Auditec Spanish bisyllabic word recognition
test (Flores & Aoyama, 2008; Shi & Sánchez, 2010; Weisleder
& Hodgson, 1989), but none has specifically examined
dialectal differences. The dialectal profile of these three
studies is summarized in Table 5. Note that participants
were bilingual in all studies, and no normative data are
readily available for monolingual Spanish speakers. This
characteristic makes it appropriate to compare the current
study to the above studies. Both Weisleder and Hodgson
(1989) and Flores and Aoyama (2008) included mainly
Mexican listeners plus individuals from a variety of Spanish-
speaking countries. According to Dalbor’s dialectal system
(1997), results from these two studies may better represent
the Mexican dialect than other dialects, but results may have
been confounded by data from listeners of other dialects. In
particular, Colombia and Venezuela, which were included
in both studies, belong to two different dialectal regions
(Coastal vs. Highland). Without specification of individual
participants’ origin, it is impossible to know which dialects
these Colombian and Venezuelan participants were speaking
in those studies. Shi and Sánchez (2010) included listeners
from 12 Caribbean, Central American, and South American
Shi & Canizales: Spanish Dialect and Word Recognition 79
Table 5. Listener dialectal profile in three studies that employed the Auditec Spanish word recognition test.
Weisleder & Hodgson (1989)
Main purpose Interlist equivalency;
Performance-intensity function in quiet
Listener number 16
Listener group 1 2
Listener dialect Mexican (9)
Central American (1)
Caribbean/Coastal American (at least 2)
N Two Venezuelan and one Colombian
included, dialectal region unspecified
N One Castilian Spanish listener also
included
nations to obtain basic psychometric properties of the
Auditec test. No single nation was overrepresented, and
fairly balanced representation was achieved across four
American Spanish dialects (excluding Argentinean).
Current Spanish-dominant participants’ performance
in quiet can be compared to Weisleder and Hodgson (1989)
and Flores and Aoyama (2008), both of which obtained
performance-intensity functions for the Auditec test in quiet.
The 40-dB SL re: PTA presentation level in the current
study can only be compared to the highest level employed in
those two studies (32 dB HL in Weisleder & Hodgson, 1989;
40 dB SPL in Flores & Aoyama, 2008). Weisleder and
Hodgson’s listeners, all of whom were immigrants from
foreign nations, averaged 94.5% correct across the four test
lists, whereas Flores and Aoyama’s second language listeners
(bilinguals who were more experienced with Spanish than
English) obtained 90% on List A, which was the only list
used in that study. In the current study, Spanish-dominant
listeners, whose language profile was comparable to the
listeners in Weisleder and Hodgson and Flores and Aoyama,
yielded an average performance of 98.6% across the
Caribbean/Coastal and Highland groups (97.8 and 99.4%,
respectively). This performance was close to Shi and Sánchez
(2010), which reported an average of 115 RAUs (roughly
99%) in their Spanish-dominant listeners, obtained at 45 dB
HL in quiet.
The higher performance reported in Shi and Sánchez
(2010) and the current study as compared to the other two
studies could be due to methodological differences across
the studies. In Weisleder and Hodgson (1989) and Flores
and Aoyama (2008), stimuli were presented via one ear
(randomly selected in Weisleder & Hodgson, 1989; preferred
in Flores & Aoyama, 2008) but were presented binaurally
in the Shi and Sánchez (2010) and the current study.
80 American Journal of Audiology N Vol. 22 N 74–83 N June 2013
Flores & Aoyama (2008) Shi & Sánchez (2010)
Intertest performance-intensity function Comparison between English/
in quiet; Spanish bilingual listeners
Comparison between Spanish listeners who were dominant in English
who obtained English as second versus in Spanish;
versus first language Prediction of performance in quiet
and in noise based on listener
language background variables
20 30
2
English as a second language – Foreign- ED – English-dominant;
born, Spanish as first language; SD – Spanish-dominant
Balanced bilingual – Spanish and English
both first languages
Mexican (15) Mexican (6)
N Four Venezuelan and one Colombian Central American (4)
included, dialectal region unspecified Caribbean/Coastal American (10)
Highland American (10)
Presentation level was also different across the studies—
higher in Shi and Sánchez and the current study than in
Weisleder and Hodgson and Flores and Aoyama. For
English monosyllabic words, performance in quiet asymp-
totes at 32 dB SL re: PTA (Beattie, Edgerton, & Svihorec,
1977). Spanish test words should theoretically asymptote
at a lower level due to their bisyllabic nature; that is, the
higher presentation level used in the current study should not
have resulted in better performance than previous studies.
Hence, it is not clear whether, or how much, presentation
level could account for the interstudy difference.
Because Weisleder and Hodgson (1989) and Flores and
Aoyama (2008) both relied predominantly on Mexican
listeners, whereas the recording was made with a Mexican
talker, it may appear counterintuitive for listeners in those
studies to yield performance lower than that reported by Shi
and Sánchez (2010) and the current study, which included
either a small number of Mexican listeners or none at all.
Two explanations could be proposed here. First, it is possible
that the interstudy differences in participant composition
were less important than the differences in methodology as
discussed in the last paragraph (e.g., test ear, presentation
level). Second, it is possible that the Venezuelan and
Colombian participants in Weisleder and Hodgson and
Flores and Aoyama came from the coastal region of their
country, as individuals of the Coastal dialect could yield a
reduced performance level on the Auditec test (based on
current findings). These listeners might have brought down
the average performance of the entire group.
Performance of the English-dominant listeners in the
current study could be compared to Flores and Aoyama
(2008) and Shi and Sánchez (2010). The balanced bilingual
listeners in Flores and Aoyama obtained 75.0% at the highest
presentation level, whereas the English-dominant listeners in
Shi and Sánchez achieved 110 RAUs (roughly 98.0%–99.0%).
In the current study, the English-dominant listeners in the
Caribbean/Coastal group achieved 94.0% in quiet, and those
in the Highland group obtained 98.2%. Thus, the English-
dominant Caribbean/Coastal listeners performed more
poorly than the average reported by Shi and Sánchez. The
large difference between Flores and Aoyama and the other
two studies could be related to the differences in listeners’
language background. Perhaps the simultaneous bilingual
listeners in that study were limited in their Spanish skills.
Regarding performance in noise, the current findings
could be compared to Shi and Sánchez (2010) at two SNRs.
At +6 dB SNR, their Spanish-dominant listeners scored in
the higher 70s in RAUs, whereas their English-dominant
listeners achieved ,70 RAUs. This result was very similar
to the current English-dominant listeners (71.10 and 72.60
RAUs for the Caribbean/Coastal and Highland group,
respectively) or Spanish-dominant Caribbean/Coastal
listeners (77.20 RAUs). However, the Spanish-dominant
Highland listeners achieved 87.48 RAUs at that SNR, which
is higher than previously reported. Perhaps the mixed
dialectal profile of their participants might have led to the
relatively lower performance in Shi and Sánchez. At 0 dB
SNR, Shi and Sánchez’s Spanish- and English-dominant
listeners scored in the 30s and 20s, respectively, in RAUs. In
the current study, the average was 30.50 and 28.73 RAUs for
the Caribbean/Coastal and Highland group, respectively, for
English-dominant listeners; and the average was 34.52 and
39.04 RAUs for the Caribbean/Coastal and Highland group,
respectively, for Spanish-dominant listeners. These values
were largely comparable, again suggesting that difficult
listening conditions may level out dialectal differences.
Language Profile and Word Familiarity
One strength of the current study is to have grouped
Spanish-speaking participants according to their language
dominance. The reason to have included English-dominant
listeners is that nowadays, many Spanish-speaking indivi-
duals, especially in North America, are using English on a
daily basis. Previous research has shown that these listeners’
performance on an English word recognition test may not
compare to the performance of their English monolingual
counterparts, particularly in degraded conditions (Rogers,
Lister, Febo, Besing, & Abrams, 2006); hence, some
clinicians may avoid evaluating English-dominant bilinguals
in English. Consequently, refusing to evaluate these indivi-
duals in Spanish would further limit opportunities for
hearing professionals to reach them. If the charge of
audiologists is to promote everyday communication, then a
client should be evaluated in all of the languages that he or
she uses on a daily basis.
Another strength of the study is to have matched the
Highland and Caribbean/Coastal groups in their language
profile for listeners who were dominant in Spanish as well as
for those who were dominant in English. Comparisons were
only made within each dominant group. Data in Table 2
indicate that all significant differences in language profile
existed between the English- and Spanish-dominant listeners,
but not between the Highland and Caribbean/Coastal
listeners, suggesting that matching was successful and that
any potential differences in performance between the
Highland and Caribbean/Coastal listeners was likely the
result of dialects. Inclusion of language variables in an
ANOVA did not improve model fit; nor did such a strategy
reduce the significance of the main effect of dialect. All of
this evidence reinforces the notion that performance on the
Spanish word test employed here is dialect dependent.
The lexical aspect of the word recognition test was
assessed through word familiarity rating and was found to
differ only between the English- and Spanish-dominant
listeners, but not between the Caribbean/Coastal and
Highland listeners. Lower familiarity in English- than
Spanish-dominant listeners was reported previously in Shi
and Sánchez (2011) and was not the focus of this study.
Thus, so long as one reports Spanish to be his or her ‘‘better’’
language, the test words are orthographically equally
familiar to him or her regardless of his or her dialect.
Although a handful of test words, such as galgo, tretas, and
grifos, previously reported in Shi and Sánchez, were found to
be of low familiarity in the current study as well, these words
were similarly difficult across dialectal groups. Overall, the
Auditec test words seemed familiar enough for clinic use.
There were other possibilities why word familiarity did
not turn out to be a key factor in the current study. Had
more complex stimuli such as sentences been employed,
lexicosemantic and morphosyntactic differences might have
played a critical role. Familiarity rating is subjective and was
not verified (e.g., by asking the participants to describe each
word or replace it with a synonym) in this study. Participants
might have made honest misjudgments of words that were or
were not familiar to them, especially those whose reading
skills in Spanish were not advanced.
Phonetic Differences Across Dialects
Based on the discussion above, it is tenable to conclude
that phonetic differences comprised the main reason for
differences between the dialectal groups. Of the two included
dialects, the Highland dialect resulted in higher performance
than the Caribbean/Coastal dialect. This difference existed
not only in English-dominant participants, but also in
Spanish-dominant participants. This result is likely due to
the fact that the Highland dialect is the more linguistically
conservative of the two. That is, not only has it retained
many of the Castilian Spanish characteristics, but it is also
comparable over a vast region across the Andean area. In
contrast, the Caribbean/Coastal dialect experienced sub-
stantial changes in linguistic features over the years as a
result of trade and communication with non Spanish–
speaking areas in the Americas (Canfield, 1981; Dalbor,
1997). Given these historical, geographic, and sociolinguistic
differences, it seems understandable for the Highland and
Caribbean/Coastal dialects to have developed linguistic
features unique to each own.
The phonetic differences across American Spanish are
well known, but the current study provides evidence that
such differences are substantial enough to affect clinical
Shi & Canizales: Spanish Dialect and Word Recognition 81
assessment. It should be noted that credit was already given
to Caribbean/Coastal listeners even though their verbal or
orthographical responses may be considered unacceptable
according to other American Spanish dialects. Indeed,
almost all of the characteristics specific to the Caribbean/
Coastal dialect (Table 1) were seen in this study. The
interchange of /l-r/ was the most prevalent (especially the
replacement of /r/ with /l/). The elision of /d/ and interchange
of s-z in writing were also common. The performance of
Caribbean/Coastal listeners could be further unfavorably
viewed if clinicians used ‘‘standard’’ Spanish when scoring
their responses. Due to the design of the current study,
prevalence data were not available for each of the unique
phonetic variations in listener responses. Future investiga-
tions should gain insights in this direction, as findings may
better prepare clinicians to predict pronunciation or spelling
‘‘errors’’ made by clients of a Caribbean/Coastal origin.
The test material was recorded with a male Mexican
talker (Weisleder & Hodgson, 1989); therefore, the words
may have been pronounced with phonetic features that
are characteristic of the Mexican dialect. However, most
participants in the current study, who were naı̈ve to this type
of experiment, reported that they could not confidently
recognize the talker’s dialect. It is possible that the talker
might have been instructed by the developer(s) of the test to
speak ‘‘neutrally,’’ as a word recognition test maximally free
of dialectal influence may be desirable for clinical practice.
Indeed, most test words seemed to be equally familiar to the
participants across dialects, suggesting that they might have
been chosen from a large corpus to represent the vocabulary
that is most general for American Spanish speakers. These
said, the current report suggests that dialectal influence was
still present in the test, more so for the Caribbean/Coastal
listener group than for the Highland group. Phonetically, the
Mexican dialect may be considered somewhat more similar
to the Highland dialect than to the Caribbean/Coastal dialect
(Caballero et al., 2009). Phonetic closeness to the Mexican
dialect could be accountable for the better performance in
the Highland than the Caribbean/Coastal group.
Previously, Cokely and Yager (1993) showed that
clinicians were able to reliably score performances on a
Spanish word recognition test, regardless of their knowledge
of Spanish. However, some knowledge of Spanish dialects,
especially the phonetic and orthographical variations (of the
Caribbean dialect, in particular), as demonstrated in the
current study, is probably needed to serve Spanish-speaking
clients across dialects. Even for experienced clinicians, fair
scoring of listeners’ performance could still be difficult due to
dialectal differences. In Rogers et al. (2006), two Spanish/
English bilingual clinicians (of Argentinean and Dominican
descent, respectively) failed to consistently rate 12 bilingual
listeners’ accent due to the variety of dialects these bilinguals
presented. Hence, clinicians are advised to be familiar with
the culture of the local Spanish community and to be
sensitive to individual differences. If audiologists familiar
with different American Spanish dialects are not readily
available, perhaps a phonologist or SLP who is experienced
in this area should be consulted for difficult cases.
82 American Journal of Audiology N Vol. 22 N 74–83 N June 2013
The dialectal difference in performance was consistent
across noise levels but was most clinically significant for the
two easier listening conditions, +6 dB SNR and in quiet. The
difference between the Caribbean/Coastal and Highland
listeners for quiet was ,15 RAUs for English-dominant
listeners. This difference at such a high performance level was
equivalent to an error rate of 4%–6% (i.e., 2–3 words per
50-word list). Some clinicians may consider this error
noncritical. However, dialectal difference for +6 dB SNR
was >10 RAUs for the Spanish-dominant listeners, which
translates to ,10% (i.e., ,5 words per list). There is no
denying that the difference in the latter case is significant
enough to result in misdiagnosis. According to Thornton
and Raffin’s binomial model (1978, revised by Carney &
Schlauch, 2007), a difference of five words on a 50-item word
recognition test can be considered a critical difference. Thus,
when scoring a Spanish-speaking client’s performance, the
audiologist must exercise caution regarding the dialect the
client speaks.
Conclusion
The current study compared four groups of Spanish/
English bilingual individuals’ recognition of Spanish words
in quiet and in noise (+6, +3, and 0 dB SNR). The bilingual
individuals differed in their dialect (Caribbean/Coastal vs.
Highland) and dominance (Spanish vs. English) but were
matched in other aspects of their language profile. Results
showed a significant effect for dialect, dominance, and SNR.
The dialectal effect was independent of dominance and SNR
effects and could not be accounted for by listeners’ language
background and familiarity rating with the test words. Thus,
the dialectal effect on the Auditec Spanish word recognition
test is ‘‘real.’’ In conclusion, when evaluating clients who
speak different American Spanish dialects, clinicians should
be sensitive to the phonetic characteristics of each dialect.
Acknowledgment
The authors would like to thank all of the volunteers who
participated in this study. The authors would also like to thank
Sylvia Yúdice Walters for her insight on American Spanish
dialects. Portions of this work were presented at the 2012
American Speech-Language-Hearing Association Convention in
Atlanta, GA.
References
American National Standards Institute. (2004). Specifications for
audiometers (ANSI-S3.6-2004). New York, NY: Author.
American Speech-Language-Hearing Association. (2009).
Membership profile: Highlights and trends. Retrieved from
http://www.asha.org/NR/rdonlyres/71229952-6205-4965-9970-
37A72CA13292/0/2008MemberCounts.pdf.
Arnett, J. J. (2002). The psychology of globalization. American
Psychologist, 57, 774–783.
Beattie, R. C., Edgerton, B. J., & Svihorec, D. J. (1977). A
comparison of the Auditec of St. Louis cassette recordings of
NU-6 and CID W-22 on a normal-hearing population. Journal
of Speech and Hearing Disorders, 42, 60–64.
Berruecos, P. T., & Rodrı́guez, J. L. (1967). Determination of the
phonetic percent in the Spanish language spoken in Mexico City,
and formation of P. B. lists of trochaic words. International
Audiology, 6, 211–216.
Caballero, M., Moreno, A., & Nogueiras, A. (2009). Multidialectal
Spanish acoustic modeling for speech recognition. Speech
Communication, 51, 217–229.
Canfield, D. L. (1981). Origins of American Spanish. In Spanish
pronunciation in the Americas (pp. 1–18). Chicago, IL: University
of Chicago Press.
Carney, E., & Schlauch, R. S. (2007). Critical difference table for
word recognition testing derived using computer simulation.
Journal of Speech, Language, and Hearing Research, 50,
1203–1209.
Cokely, J. A., & Yager, C. R. (1993). Scoring Spanish word-
recognition measures. Ear and Hearing, 14, 395–400.
Dalbor, J. B. (1997). Language and dialects. In Spanish
pronunciation: Theory and practice (pp. 12–25). Boston, MA:
Thomson-Henle.
Flores, L., & Aoyama, K. (2008). A comparison of psychometric
performance on four modified Spanish word recognition tests.
Texas Journal of Audiology and Speech-Language Pathology, 31,
64–70.
Kayser, H. (1995). Bilingual speech-language pathology: An
Hispanic focus. San Diego, CA: Plural.
Marian, V., Blumenfeld, H. K., & Kaushanskaya, M. (2007). The
Language Experience and Proficiency Questionnaire (LEAP-Q):
Assessing language profiles in bilinguals and multilinguals.
Journal of Speech, Language, and Hearing Research, 50, 940–967.
Nusbaum, H. C., Pisoni, D. B., & Davis, C. K. (1984). Sizing up the
Hoosier mental lexicon: Measuring the familiarity of 20,000 words
(Research on Speech Perception Progress Report No. 10).
Bloomington, IN: Speech Research Laboratory, Psychology
Department, Indiana University.
Rogers, C., Lister, J., Febo, D., Besing, J., & Abrams, H. (2006).
Effects of bilingualism, noise, and reverberation on speech
perception by listeners with normal hearing. Applied
Psycholinguistics, 27, 465–485.
Schneider, B. S. (1992). Effect of dialect on the determination of
speech-reception thresholds in Spanish-speaking children.
Language, Speech, and Hearing Services in Schools, 23, 159–162.
Shi, L.-F., & Sánchez, D. (2010). Spanish/English bilingual listeners
on clinical word recognition tests: What to expect and how to
predict. Journal of Speech, Language, and Hearing Research, 53,
1096–1110.
Shi, L.-F., & Sánchez, D. (2011). The role of word familiarity in
Spanish/English bilingual word recognition. International
Journal of Audiology, 50, 66–76.
Studebaker, G. A. (1985). A ‘‘rationalized’’ arcsine transform.
Journal of Speech and Hearing Research, 28, 455–462.
Thornton, A. R., & Raffin, M. J. M. (1978). Speech-discrimination
scores modeled as a binomial variable. Journal of Speech and
Hearing Research, 21, 507–518.
U.S. Census Bureau. (2010). The Hispanic population 2010 [Press
Release C2010BR-04]. Washington, DC: Author.
Weisleder, P., & Hodgson, W. R. (1989). Evaluation of four Spanish
word-recognition-ability lists. Ear and Hearing, 10, 388–392.
Weiss, D., & Dempsey, J. J. (2008). Performance of bilingual
speakers on the English and Spanish versions of the Hearing
in Noise Test (HINT). Journal of the American Academy of
Audiology, 19, 5–17.
Wilson, W. J., & Moodley, S. (2000). Use of the CID W-22 as a
South African English speech discrimination test. South African
Journal of Communication Disorders, 47, 57–62.
Shi & Canizales: Spanish Dialect and Word Recognition 83
Copyright of American Journal of Audiology is the property of American Speech-Language-
Hearing Association and its content may not be copied or emailed to multiple sites or posted
to a listserv without the copyright holder's express written permission. However, users may
print, download, or email articles for individual use.