Comparing the AzBio Sentence-in-Noise Test in English and Spanish in Bilingual Adults

Abstract

Background:

The AzBio Sentence-in-Noise Test was developed in 2011 and was successful in minimizing speech-recognition ceiling effects, giving clinicians and researchers a more accurate representation of a listener’s speech-in-noise recognition. Recently, the Spanish version of the AzBio corpus was developed as a sentence-recognition test that could similarly be used to reduce speech-recognition ceiling effects in Spanish-speaking patients. The developers of the AzBio in Spanish included researchers and clinicians from the United States and Colombia.

Purpose:

The aim of this study was to determine whether the AzBio test batteries in English and Spanish are comparable in difficulty to proficient Spanish–English bilingual adults residing in the United States.

Research Design:

The study was designed as a standard group comparison.

Study Sample:

Participants included 20 Spanish–English bilinguals between the ages of 18 and 30 years with hearing thresholds no greater than 25 dB HL in both ears.

Data Collection and Analysis:

Participants listened to three lists of 20 sentences from the AzBio Sentence-in-Noise Tests in English and in Spanish over two test sessions. Sentences were presented at a +5 dB signal-to-noise ratio in 10-talker babble. Sentence-recognition scores were calculated from total words repeated correctly out of total words presented for all three lists (60 sentences in total) in each language condition. A language experience survey was used to quantify and explore language experience in different dialects of Spanish.

Results:

Our results indicate that bilingual listeners scored similarly on the English and Spanish test corpora on the group level. On an individual level, participants who spoke a Colombian Spanish dialect were among the highest-performing listeners for the Spanish test corpus and among the lowest-performing listeners for the English corpus.

Conclusions:

The AzBio in Spanish is a highly valuable clinical tool for evaluating speech recognition in Spanish-speaking patients. Our results suggest that listeners who spoke a Colombian Spanish dialect, consistent with the location where the AzBio in Spanish test was developed, tend to perform better on the Spanish version of the test compared to the English version of the test. Thus, dialectical factors may affect sentence-recognition scores on the AzBio in Spanish corpus. Clinicians in the United States must consider dialect when administering this test corpus because the most common dialect in the United States is Mexican Spanish. Future research should evaluate the education level of listeners to determine the impact of language-specific vocabulary on sentence-recognition performance on both AzBio language corpora.

Spanish is the fourth-most-used language in the world, with around 600 million speakers. Unlike other commonly spoken languages such as Mandarin Chinese—whose speakers are largely concentrated in Asia—Spanish spans six continents and is the official language of more than 20 countries (Instituto Cervantes, 2023). Large geographical and cultural differences in Spanish speakers have resulted in vast phonological, semantic, and syntactical differences between Spanish dialects, with some dialects being more intelligible than others depending on the language experience of the listener (Fernández and Ueda, 2018).

The rise in racial and ethnic diversity within the United States will soon be accompanied by an increase in linguistic diversity. In 2019, 20 percent of Americans reported proficiency in both English and a second language, making the United States the second-largest Spanish-speaking country in the world with more than 41 million Spanish speakers (Grosjean, 2018; Instituto Cervantes, 2023). By 2050, it is projected that the United States will become the largest Spanish-speaking country in the world, with an estimated 138 million Spanish speakers, and one third of U.S. citizens are projected to report Spanish as their primary language (Diez, 2015).

Despite the apparent rise in linguistic diversity in the United States, the demographics of audiologists and hearing scientists in the United States do not adequately reflect the general population. The American Speech-Language-Hearing Association has more than 14,000 members who are audiologists and/or hearing scientists, but only 384 are bilingual English–Spanish service providers—that is, only 2.7 percent of all hearing professionals within the national organization reported providing care in Spanish (American Speech-Language-Hearing Association, 2023). As previously mentioned, there are approximately 41 million Spanish-speaking individuals in the United States. Whereas hearing loss generally affects about 10 percent of the U.S. population (Lin et al, 2011), hearing loss in the U.S. Hispanic community is at a slightly higher rate: 15 percent (Cruickshanks et al, 2015). This means that for every one audiologist who is a Spanish-speaking service provider there are more than 16,000 Spanish-speaking individuals with hearing loss.

The alarming discrepancy between the racial and ethnic population of the United States and the demographics of the professionals in hearing health care is also apparent in hearing-related research studies. Research in audiology and hearing science, as in many other fields, suffers from sampling bias within populations of individuals who are “WEIRD”—White, Educated, Industrialized, Rich, and from Democratic Societies (Henrich et al, 2010; Shin and Doraiswamy, 2016; Pittman et al, 2021). There are negative consequences of limiting our research questions to these populations, including reduced generalizability of research findings to the greater population, experimenter bias, and threats to ecological validity.

In 2020, Rivas and colleagues addressed one of these limitations by developing a Spanish version of the AzBio corpus that is designed as the Spanish-language equivalent to the English version of the AzBio sentence-recognition corpus (Rivas et al, 2021). The original AzBio Sentence-in-Noise Test in English is the established “gold standard metric” for measuring sentence-recognition performance before and after cochlear implantation (Spahr et al, 2012). The English AzBio was originally designed to avoid ceiling effects commonly associated with traditional speech-recognition batteries, such as the Hearing in Noise Test (HINT) (Gifford et al, 2008). The AzBio in Spanish was designed to similarly provide an alternative to existing Spanish sentence-recognition tests, including the Spanish HINT, that demonstrated substantial ceiling effects for individuals with hearing loss (Nilsson et al, 1994; Soli et al, 2002; Barón de Otero et al, 2008; Huarte, 2008). An external analysis of the Spanish HINT and AzBio in Spanish compared the linguistic content between the two test batteries (Turnbull et al, 2023). Results from this analysis suggested that the AzBio in Spanish is linguistically more complex than the Spanish HINT in terms of sentence length, complexity, and grammatical structure. The AzBio in Spanish corpus fills a critical gap in the clinical tools that are available to non-English speakers.

However, it is not yet clear whether the AzBio in Spanish materials represent a test corpus that is functionally equivalent to the gold-standard AzBio in English corpus. There have been no systematic comparisons between the English- and Spanish-language tests. Additionally, there is limited information regarding the ethnolinguistic background of the speakers of the recorded Spanish sentences. This is important to note because dialectical differences in Spanish can substantially affect the outcomes of a listener’s score; features in some Spanish dialects are not mutually intelligible in others (Fernández and Ueda, 2018).

The primary focus of this study was to determine whether the English AzBio and the Spanish AzBio are comparable in difficulty level in adult speakers of both English and Spanish. If these lists are comparable in difficulty level, we can establish that the administration of the Spanish version of the test would not likely add an additional barrier for cochlear implant candidacy for Spanish-speaking individuals. A significant difference in difficulty level across language corpora could greatly influence a listener’s sentence-recognition scores depending on what language the test is administered in—thus possibly creating an additional barrier for cochlear implant candidacy for one of these groups. In other words, if the two corpora are comparable in difficulty, clinicians who routinely evaluate patients for cochlear implant candidacy using the AzBio sentences in English could apply the same clinical decision-making when evaluating patients using the AzBio sentences in Spanish in the same listening condition.

The secondary focus of our study was to determine the impact of Spanish-language dialect to participants’ sentence-recognition scores. There is limited information regarding the recorded speakers’ ethnolinguistic background of the AzBio in Spanish materials beyond reporting that the test was developed in Colombia. It is possible that a difference in speech-recognition scores may be influenced by the listeners’ ethnolinguistic background, which could ultimately affect which individuals are determined to meet candidacy criteria for cochlear implantation.

MATERIALS AND METHODS

Participants

Participants were recruited using flyers posted to the Purdue University campus and surrounding areas, as well as through community outreach to local and university Latino/Hispanic cultural centers. To be included in the study, individuals were required to be between 18 and 30 years old, have hearing thresholds no greater than 25 dB HL at octave frequencies between 250 and 8,000 Hz in both ears, and self-report proficiency in both English and Spanish. Individuals were excluded from the study if they reported clinically diagnosed hearing loss or self-rated their expressive language proficiency in either English or Spanish lower than “7” or “good” on the Language Experience and Proficiency Questionnaire (LEAP-Q). Twenty English–Spanish bilingual adults—13 males and 7 females—aged 19 to 30 years (mean = 23.95 ± 3.4 years) participated in the study. Of the 20 subjects recruited for the study, 19 were identified as having “good” expressive language proficiency in both languages as defined by the LEAP-Q. Data from these 19 participants are included in the results and subsequent analyses. Procedures were approved by the Purdue University Institutional Review Board (protocol no. IRB-2022-687) and all participants were compensated for their time.

Language Experience Questionnaire

When assessing one’s language status and history, it is important to consider the distinction between fluency, proficiency, and dominance. Although these terms may be used interchangeably, there are distinct differences. Fluency describes the use of language with minimal interruptions or self-corrections (Chambers, 1997), whereas proficiency describes the mastery of a language (Bedore et al, 2012). Language dominance describes the relative proficiency between two or more languages and can be affected by a multitude of factors, including age of acquisition and the amount of input in the language (Gathercole and Thomas, 2009). It is important to note that a bilingual or multilingual individual can have a dominant language and be proficient in one or more languages—proficiency and language dominance are not always inclusive.

Proficiency in both languages was required to participate in the study. We administered the LEAP-Q through a Qualtrics survey to investigate a participant’s language background. This instrument is a self-assessment tool designed to gain information about the respondent’s language experience, dialectal exposure, and cultural background (Marian et al, 2007). An 11-point Likert scale is used for all questions with each point anchored to a descriptive label (i.e., 0 = “none,” 5 = “adequate,” 10 = “perfect”). The LEAP-Q was modified slightly to remove any questions regarding immigration or citizenship status; questions were added to assess the participant’s ethnolinguistic background in both languages (e.g., specific dialect[s] of English or Spanish spoken). Participants were encouraged to be as specific as possible regarding their language experiences. Completion of the LEAP-Q was verified prior to the first session of the experiment. However, results of the questionnaire were not examined until both language conditions of the experiment were completed to avoid potential scoring biases due to dialectical differences. The participants’ ethnolinguistic information can be found in Table 1. Although the LEAP-Q is a helpful tool to determine a person’s level of proficiency in a language, this measure may introduce response bias because it is based on the participant’s self-report.

Table 1.

Language Experience of Recruited Participants in English and Spanish

ID	Age (years)	Sex	Dialect of English	Age Acquired English (years)	Dialect of Spanish	Age Acquired Spanish (years)
1	30	M	Spain	12	Spain	Birth
2	30	M	Colombia	15	Colombia	Birth
3	24	M	Colombia	10	Colombia	Birth
4	30	M	Colombia	16	Colombia	Birth
5	24	F	Venezuela	5	Venezuela	Birth
6	27	F	Puerto Rico	5	Puerto Rico	Birth
7	23	M	Mexico	6	Mexico	Birth
8	26	M	United States	7	Puerto Rico	Birth
9	19	M	United States	8	Mexico	Birth
10	20	F	Colombia	11	Colombia	Birth
11	20	M	Mexico	5	Mexico	Birth
12	19	F	Puerto Rico	Birth	Puerto Rico	Birth
13	25	F	United States	6	Mexico	Birth
14	21	F	Puerto Rico	5	Puerto Rico	Birth
15	26	M	United States	Birth	Puerto Rico	Birth
16	22	F	United States	Birth	United States	Birth
17	23	M	United States	Birth	United States	4
18	24	M	United States	1	United States	Birth
19	25	M	United States	Birth	United States	Birth
20	21	M	United States	Birth	United States	13

Mean age = 23.95 ± 3.4 years. Participant 20 reported “less than adequate” expressive language skills in Spanish and was excluded from further analysis. F = female; M = male.

AzBio Sentence-in-Noise Test

The English and Spanish AzBio test materials comprise 15 lists of 20 individual sentences. The sentences are spoken by four speakers, two male speakers and two female speakers. The lengths of the sentences range from 3 to 12 words in the English version and 3 to 14 in the Spanish version (mean in English = 7 words; mean in Spanish = 7.3 words). The total word count of each list ranges from 133 to 154 words in the English version and 139 to 154 in the Spanish version (mean in English = 140.87; mean in Spanish = 144.47). Other than the language of the test materials, test sessions were identical in nature regarding the test protocol, presentation level, and signal-to-noise ratio (SNR).

PROCEDURES

Experiment Paradigm

Before the first session of the experiment, participants had their hearing screened at 25 dB HL at octave frequencies from 250 to 8,000 Hz using a Grason–Stadler (GSI-18) portable audiometer with TDH-49 supra-aural headphones in a sound-treated booth. Otoscopy was performed to confirm there were no contraindications for testing (e.g., excessive cerumen, suspected fluid behind the eardrum).

Participants were seated 1 meter away at 0° azimuth to an Audioengine loudspeaker in a sound-treated booth. Sentences were presented over the loudspeaker at 50 dB SPL at +5 dB SNR in multitalker babble (colocated) using a custom MATLAB script. A sound-level meter was used to ensure the presentation level of the stimuli prior to testing. The test procedures were consistent with the standard clinical protocols for cochlear implant candidacy evaluation in the Purdue University Speech and Hearing Clinic. Each test session was recorded for reliable scoring purposes.

Participants were instructed to repeat each sentence to the best of their ability exactly as they heard it. Three 20-sentence lists within a single language condition (either English or Spanish) were presented during a single session, for a total of 60 sentences tested in each language condition. Testing was performed over two sessions with the target language being counterbalanced to avoid order effects. Sessions were typically scheduled within 1 week of each other to eliminate overfamiliarity with the task and to minimize fatigue.

Scoring

Participants’ performance on the AzBio sentences were scored independently by two English–Spanish bilingual scorers who were graduate student clinicians familiar with administering these test batteries. Scores for each sentence list were calculated by the sum of words correctly repeated divided by the total number of words in the list. Scores were averaged across the three lists within each language condition. Final percent-correct scores were compared at group-level, within-subject, and by dialectical proximity across both language conditions.

It is important to note that the AzBio in English and in Spanish do not provide scoring guidance for clinicians or researchers for either test. The clinician can either score a word as “correct” or “incorrect.” However, Spanish dialects can vary phonologically, semantically, and syntactically due to historical language interference with indigenous languages spoken in its geographic proximity. This may cause scoring discrepancies across individual providers. For consistency of scoring for the Spanish-language condition, both scorers agreed on scoring parameters using the assumed dialect of the test: Colombian Spanish. Thus, dialectical differences of the participants were not considered when scoring in English or Spanish. Strict scoring criteria were used in which words were only scored as “correct” if the participant repeated the entire word (e.g., correct repetition of the root of the key word, but with the final /s/ omitted was scored as incorrect). Rigorous scoring criteria is typically implemented in Spanish-language speech-perception testing, because changing a single phoneme can change the meaning of a word.

Interrater reliability scores reflected a 97.54 percent agreement between the two bilingual scorers. Discrepancies between interrater scores were discussed and resolved after the recorded responses were reviewed.

RESULTS

Language Experience

Following the participants’ completion of the experimental task, their responses to the LEAP-Q were inspected to identify their ethnolinguistic background. Nineteen of the 20 participants self-identified as proficient in both English and Spanish with more than adequate expressive language skills in both languages. Eighteen of the participants identified as Hispanic and Latino; one participant was from Spain, and thus this participant identified as Hispanic but not Latino. Eighteen of the 19 participants completed some college-level education. Specifically, six participants were enrolled in an undergraduate degree program, three had completed a bachelor’s degree, and nine were completing a graduate degree or postdoctoral fellowship.

As shown in Figure 1, the geographic location of English- and Spanish-language acquisition varied within our group of participants. About half of them (n = 8) acquired English while residing in the United States, and the second-largest group (n = 5) acquired English while residing in Colombia/Venezuela. The remaining participants acquired English while residing in Puerto Rico, Mexico, or Spain. For Spanish-language acquisition, five participants acquired Spanish while residing in Colombia/Venezuela, five participants while residing in Puerto Rico, four participants while residing in Mexico, four participants while residing in the United States, and only one while residing in Spain.

Figure 1.

Ethnolinguistic background of the 19 participants in our study. Circles and triangles represent English and Spanish, respectively. The color within the shape represents the country where the participant acquired either English or Spanish.

Did Scores Across Languages Differ?

Our primary experimental question asked whether sentence-recognition scores for the English and Spanish tests differed significantly in a group of English–Spanish bilingual listeners. Percent correct sentence-recognition scores for each language condition are plotted in Figure 2. Mean performance scores for the group of participants revealed that there were no significant differences in recognition scores between the English and Spanish AzBio Sentence-in-Noise Tests. Mean level of performance ranged from 56.8–92.6 percent correct in English (mean in English = 80.9 percent, standard deviation = 11.0 percent) and 58.1–97.9 percent correct in Spanish (mean in Spanish = 88.8 percent, standard deviation = 18.0 percent). A paired, two-tailed t-test revealed no significant differences in sentence-recognition scores across two language conditions [t(18) = −1.89417, p = 0.074387]. In summary, for this group of bilingual adults there was no group-level difference in sentence-recognition ability between the English and Spanish versions of the AzBio test.

Figure 2.

Box-and-whisker plot of mean sentence-recognition scores for English and Spanish AzBio with a +5 dB SNR. Scores from a single participant are connected with a solid-colored line and the color represents where the participant acquired Spanish.

Are There Potential Differences in Sentence Recognition Related to the Listeners’ Dialect?

Figure 3 highlights the observed dialect-dependent trends in sentence-recognition scores across the two language conditions. The participants falling into the highest quartile of the distribution of scores on the Spanish AzBio (ranging from 95.5–97.9 percent) largely identified as having Colombian or Venezuelan heritage; these countries having a much closer geographical and dialectical proximity to the dialect of the recorded speakers used in the Spanish test battery. Yet, these high-performing Colombian and Venezuelan speakers tend to fall into the lowest quartile of recognition performance on the English test battery (ranging from 58.1–70.3 percent correct). On average, participants who identified as speaking a Colombian Spanish dialect demonstrated a 28.9 percentage point difference in their scores between language conditions.

Figure 3.

Dot plot of mean sentence-recognition scores for the AzBio in English and Spanish with a +5 dB SNR. Shapes represent scores in either English or Spanish and are connected with a solid-colored line that represents where the participant acquired Spanish. The plot is arranged by greatest mean difference in Spanish to greatest mean difference in English.

It is also important to note that the highest-performing participants in the Spanish-language condition all received formal education in Spanish-speaking schools and acquired English as an adolescent in their home country. Likewise, the lowest-performing participants in the Spanish-language condition reported acquiring English and/or Spanish while residing in the United States and received no formal education in Spanish. Participants who had the lowest scores in the Spanish-language condition demonstrated average or high-average performance on the English-language materials, with scores ranging from 85.8–89.5 percent correct, falling within the highest two quartiles.

DISCUSSION

The AzBio Sentence-in-Noise Test was developed primarily to be used by clinicians to measure sentence-recognition ability of patients being evaluated for cochlear implant candidacy. Although the developers of the AzBio in Spanish carefully compared the variability and validity within and across sentence lists, there is no systematic comparison between the English and the Spanish AzBio corpora. The aim of this study was to determine whether the English and Spanish AzBio Sentence-in-Noise Tests were comparable in difficulty level for bilingual adults. The secondary aim of this study was to determine whether Spanish-language dialects—representing a vast range of phonemic and semantic differences—influenced sentence-in-noise recognition performance.

There are major disparities in the existing literature on the efficacy of using existing audiometric speech test batteries for bilingual speakers and speakers of languages other than English. It is unclear whether there are significant differences in sentence-recognition scores for native speakers compared to proficient nonnative speakers in the same language condition. This gap in knowledge is also reflected in the clinic; this is one of the barriers in providing comprehensive audiological services for non-English-speaking individuals.

Previous research in this area has explored using clinical speech test materials in English and Spanish for bilingual adults by evaluating the ethnolinguistic background of their participants (Von Hapsburg and Peña, 2002; Shi and Sánchez, 2010). Age of acquisition, length of immersion, language use, listening proficiency, and language dominance were considered as potential predictors of speech-recognition scores in both languages. Von Hapsburg and Peña (2002) reported that language status, language usage, and language competency are among the three most important factors in determining how successful a listener will be in a speech-perception task in the target language. Shi and Sánchez (2010) echo the previous finding, reporting that age of acquisition and language dominance appear to have the highest predictive power for the language in which an individual will have the best speech-perception score. That being said, both studies agree that most studies looking at speech perception in bilinguals do not report these findings and may prevent accurate replication of research studies. Although both studies investigated bilinguals’ performance on speech-perception tasks, these studies did not consider the impact of the dialect of the speech test material used to obtain speech-recognition scores compared to the native dialect of their participants.

A focused review by Cowan et al (2022) reviewed 28 articles describing bilingualism in speech-recognition and audiometry research. Out of the 28 articles that were analyzed for content, only 26 of these articles reported language competency data (e.g., interviews, questionnaires, standardized testing) to report on their participants’ bilingualism profile. All 26 of these studies identified their participants’ “bilingual status” and most of them inquired about linguistic factors such as language competency, language use, and language history. Although most of these studies were quite comprehensive, only 3 of 26 collected data related to “accent” or dialectical experience.

Cowan et al (2022) recommends that not only should research intentionally collect and disclose data related to listeners’ traditional bilingual profile (i.e., bilingual status, language competency, language use, and language history), but it should also report the participants’ education level and socioeconomic status, because these factors could account for differences in performance in speech-recognition tasks across listeners with traditional bilingual profiles. Although our research was able to retrieve educational experiences for our participants, the LEAP-Q administered before the first session did not inquire about socioeconomic status.

Our findings provide insight into the scores earned by participants and how these differences in scores across languages were potentially affected by listeners’ ethnolinguistic background. Spanish dialects are not mutually intelligible. Whereas communication breakdowns can occur in other languages due to dialectical differences, it is more prominent in Spanish dialects because of the geographical span of the language, the language interference with indigenous languages, and the sociocultural factors that can influence each dialect differently. The result of the aforementioned influences on Spanish dialects have created vast phonological, semantic, and syntactic differences across Spanish dialects. Thus, it is important to discuss dialectical experience to determine whether a particular test is appropriate to assess speech perception, whether in a research or clinical setting.

Implications

There is a lack of well-normed Spanish-language audiometric speech materials, as well as a lack of research in this area. Results from the current study highlight the need for Spanish-language clinical evaluation tools, particularly those that have been tested with a heterogenous group of Spanish speakers with a wide range of dialects. Our findings suggest that, on the group level, sentence recognition in noise is comparable between the English and Spanish versions of the AzBio test corpus in our group of English–Spanish bilinguals. Within-subject differences in recognition scores across language conditions were observed, which were related to the listener’s individual dialect of Spanish. Listeners who acquired Spanish in Colombia or Venezuela, and thus spoke with a dialect most similar to the AzBio in Spanish material itself, achieved the highest sentence-recognition scores in Spanish.

The AzBio in Spanish test has filled a substantial clinical need by adapting test materials to accommodate the growing language demographic in the United States. The use of Colombian Spanish speakers—a nondominant dialect of Spanish in the United States—is a potential important consideration for clinicians and researchers who interact with Spanish-speaking individuals. Our results suggest that Colombian-Spanish speakers will typically perform better on the current version of the AzBio in Spanish compared to speakers of other Spanish-language dialects, including Mexican Spanish, which is the most dominate Spanish dialect spoken in the United States (Lipski, 2015). There are multiple clinical implications of this finding. First, for nonnative speakers of English who are being evaluated for cochlear implant candidacy in the United States, their sentence-recognition scores in English may significantly underestimate their speech-recognition ability. In other words, their scores in English may not be driven primarily by their perceptual hearing deficits, but rather by their language experience. Second, evaluating nonnative speakers’ sentence-recognition scores in their primary language does not completely overcome this limitation. The dialect of the speakers of the test materials and the dialect of the listener affects sentence-recognition scores. It appears that language experience and dialectical differences, particularly for Spanish-language materials, influence recognition scores even when listening to materials spoken in the listener’s native language. Third, the results of this study further emphasize the need for better representation of Spanish-speaking audiologists and speech-language pathologists. There is already a high demand and low supply of Spanish-speaking audiologists who can readily administer, instruct, and score Spanish-language materials without the use of an interpreter.

LIMITATIONS

There are two main limitations of this study: a relatively homogenous sample of participants and the use of only three lists to evaluate each language condition. Despite our recruitment efforts at local, non-university-affiliated community centers for Latinos and Hispanics, only one participant did not attend an institution of higher education. More than half of participants identified as graduate students or as receiving postdoctoral training at the university. It is logical to assume that our sample of participants possess a larger vocabulary, and thus their scores could be slightly higher than the national average of English- and Spanish-speaking individuals in the United States. In addition to education level, this study did not explore age-related effects or the effects of cognitive abilities on speech-perception performance.

The first three lists of the English and Spanish AzBio were used to determine whether the test batteries were equivalent in difficulty level. The number of lists were chosen to minimize the effects of listener fatigue and to replicate a clinical scenario more closely. Although the lists contained within both the English and Spanish AzBio corpora are equivalent within their own test battery, we cannot fully deduce that the results from our study are entirely representative of the entire English and Spanish test corpora.

CONCLUSIONS

The goal of the current study was to determine whether the English and Spanish versions of the AzBio tests are comparable in difficulty when administered to the same individuals in the same listening condition. Results suggest that these tests are comparable in difficulty level across language conditions at the group level, but that individual differences in participants’ ethnolinguistic background and dialectical experience influenced sentence-recognition scores.

The clinical implications of this finding are that the choice of language material, the dialect of the recorded materials, and the dialect of the listener may significantly affect the sentence-recognition score that is ultimately achieved by that listener. Clinicians should acknowledge these factors in their clinical decision-making because different speech test batteries will come with different trade-offs—largely being a higher ceiling effect versus a score being influenced by dialect effects. In addition, all clinicians must understand their responsibilities in treating culturally and linguistically diverse populations, even if they themselves are from a different background. This includes using interpreters when appropriate, continuing education for providing care to culturally and linguistically diverse populations, expanding their clinical education to include language development opportunities, and promoting inclusion of culturally and linguistically diverse individuals in our profession starting at the institutional level.

In addition, results from this study should encourage researchers to be inclusive in their participant recruitment and in their research team to ensure that speech test batteries can be applicable to a larger population. Future research should include participants with a diverse range of socioeconomic backgrounds (including diverse education levels), who speak other dialects of Spanish that are not represented in our sample, and who use a hearing aid or cochlear implant to further improve the generalizability of our results and provide direct applications to clinical populations. Large-scale multisite studies would be helpful in obtaining more data from Spanish-speaking populations across many different dialects, as well as exploring the effects of education level, vocabulary size, and cognition on this sentence-recognition task.

Abbreviations:

HINT

Hearing in Noise Test

LEAP-Q

language experience and proficiency questionnaire

SNR

signal-to-noise ratio