Speech perception testing is an integral component of the patient’s cochlear implant (CI) journey from an initial evaluation to establish candidacy to post-implantation assessment and counseling to optimize benefit. In this article, we begin by considering speech perception as it refers primarily to clinical assessments commonly used to evaluate spoken word or sentence recognition, and later argue for expanding this approach to better reflect real-world communication. The history of CIs as the first highly successful electronic prosthesis for sensorineural hearing loss (SNHL) cannot be told without persistent references to the results of various speech perception tests used to demonstrate CI benefits. Nevertheless, since the early days of CIs more than half a century ago until now, assessment of speech perception in adult CI users has continued to evolve. Although changes in clinical testing have been rather slow to implement, they have been driven by a variety of separate but interrelated factors. These include the advancements in CI technology and improved patient outcomes, changes in clinical practices and candidacy criteria, as well as expanding theoretical perspectives on speech communication in people with hearing loss. A better understanding of how these factors have influenced speech perception assessment in CI users can assist clinicians and researchers in selecting and interpreting results of specific tests for better alignment with treatment goals of individual patients and a continued improvement in CI outcomes. After briefly reviewing the changes in speech perception testing of CI users over time, we provide specific considerations and potential criteria that can assist CI clinicians and researchers in selecting or developing appropriate tests for outcome measures in clinical and research contexts.

In the 1960s and 1970s, when restoration of functional human hearing by means of electrical stimulation of the auditory nerve was a novel and controversial approach to the treatment of profound hearing loss in adults, the benchmarks for CI success were quite modest by today’s standards (House, 2011). A mere auditory awareness of speech, along with other environmental sounds, was sufficient to demonstrate the promise of the new approach for individuals with profound SNHL using single-channel devices. A noted speech perception benefit demonstrated by many single-channel CI users was an improvement in lipreading scores following implantation. However, without visual cues, open set monosyllabic word scores typically remained near chance (Bilger and Black, 1977; Owens et al., 1985; Wilson and Dorman, 2008).

As CI technology continued to develop, particularly with the introduction of multi-channel CIs, patients began achieving better speech perception outcomes, creating a need for a more comprehensive assessment of speech perception in patients with varying ability levels, with an added emphasis on testing in the auditory-only condition. Development of new tests was especially important because available monosyllabic word lists alone, which were based on materials developed during World War II to improve electronic telecommunications (Egan, 1948; Hirsh, 1947; Lehiste and Peterson, 1959; Bilger, 1984; Pisoni, 2021), did not have contextual cues and were still too challenging for many CI users. Open-set speech recognition tests also did not provide useful information for optimizing programing strategy, monitoring performance over time or guiding rehabilitation. Additional tests were needed to expand the range of speech perception skills assessed in CI users. Subsequently, the Minimum Auditory Capability (MAC) battery was developed beginning in 1976 and revised during the following decade (Owens et al., 1985) to standardize performance evaluation across CI centers and provide a more comprehensive set of assessment instruments. The MAC battery consisted of 13 auditory-only and one lip-reading tests, with all but one (familiar environmental sounds) of these tests focused on various aspects of speech perception. The tests varied in difficulty from simple discrimination of speech vs. noise to word intelligibility with different amounts of context. The battery also included prosodic tasks such as discriminating questions from statements and identifying word stress. Although the MAC battery was not often administered in full due to the length of time required (2–3 h), some tests from the MAC battery became widely adopted as benchmarks of CI performance and remain so to this day. For instance, the Consonant-Nucleus-Consonant (CNC) word recognition test, first published by Lehiste and Peterson (1959), which drew on the materials from Egan (1948), is currently still the most widely cited test used to characterized speech performance in adults with CIs (Figure 1). Importantly, the MAC tests further highlighted some inconsistencies between CI recipients’ pure tone thresholds and their ability to perceive speech. With multi-channel implants, CI-aided pure tone thresholds could often be obtained in the normal or mild hearing loss range, while most patients still struggled to understand speech presented without visual cues, particularly in noisy backgrounds.

As discussed above, clinical assessment of outcomes in adult CI users continue to rely on isolated word and sentence recognition tests conducted in quiet or fixed noise levels, using highly controlled materials. It should be noted that these measures exhibit several useful features – they are easy to score, are sensitive to degradation (e.g., noise, hearing loss), and also demonstrate changes following implantation (e.g., Baese-Berk et al., 2023). Indeed, many postlingually deafened adult CI users achieve reasonably good open-set speech recognition on conventional clinical speech recognition tests and show improvements up until around 3 months post-CI on average (Ma et al., 2023). Nonetheless, these tests offer only a limited view of the broader real-world experiences and challenges faced by adult CI users. Conventional testing materials typically contain carefully articulated, linguistically simple materials produced by a single talker or a small number of talkers. Additionally, these measures are focused solely on the listener’s ability to recognize isolated words or words in sentences. These test attributes differ from real-world speech communications, which involve dynamic understanding of utterance meaning within specific situational context, often within short time windows of an ongoing back-and-forth conversation and response planning.

There is a need for more comprehensive, ecologically valid assessments that not only measure speech recognition in controlled settings but also capture the cognitive and communicative demands of real-world listening environments. While many tests provide important clinical benchmarks, they do not necessarily predict how CI users will perform in more dynamic, real-world settings. For example, speech recognition in quiet does not necessarily predict success in complex, multi-talker environments where CI users must rapidly adapt to new voices, extract meaning from acoustically variable speech, and manage cognitive load (e.g., Tamati et al., 2023). CI users often rely on visual cues to supplement auditory input. Therefore, assessments that integrate both modalities may provide a more ecologically valid measure of real-world communication ability. The importance of visual cues in speech perception became particularly apparent during the recent COVID-19 pandemic, when masks covering speakers’ mouths made speech perception especially challenging for all people with hearing loss (e.g., Homans and Vroegop, 2021; Sönnichsen et al., 2022). Further, standard clinical tests fail to account for the substantial individual differences in how CI users adapt to real-world challenges (e.g., Moberly et al., 2025; Tamati et al., 2023). For instance, the PRESTO sentence test was specifically developed to reflect the talker and linguistic variability present in everyday listening environments. It also captures meaningful individual differences in the ability to rapidly adapt to such variability (Gilbert et al., 2013; Tamati et al., 2013). Thus, assessments that incorporate real-world complexity, such as audio-visual input, speech originating from multiple talkers, or speech in the presence of competing talkers, would provide a more ecologically valid assessment of CI users’ communicative abilities.

Conventional speech recognition tests provide endpoint “product” measures of performance, offering limited insight into the underlying cognitive processes that support successful communication (Moberly et al., 2018). In addition to the limitations of convention speech recognition measures described above, these measures were not developed to explain or identify the factors underlying the variability in outcomes or address individual differences (Pisoni et al., 2008). When a patient does well after receiving the implant he and his care team are satisfied with the outcome. However, when a patient’s performance does not improve, they are not satisfied, and it is often difficult to explain the reasons or find ways to improve performance. It is thus critical to consider factors that are responsible for individual differences, along with the associated behavioral and neurocognitive domains. To that end, “process” measures assess the underlying cognitive mechanisms that contribute to speech perception. These include information processing speed, working memory, and inhibitory control. Assessing these domains using non-auditory cognitive tasks allows researchers to disentangle cognitive contributions from audibility. For example, our recent work demonstrated that speech recognition scores in adult CI users are associated with working memory, nonverbal reasoning, and lexical access speed (Mattingly et al., 2018; Moberly et al., 2016, 2021; Tamati et al., 2020, 2023). These findings demonstrate how cognitive-linguistic abilities contribute to individual differences in speech perception performance.

Assessments should also consider the cognitive demands of listening in complex environments, as well as the compensatory strategies CI users use to manage listening effort in their daily lives. For example, previous research has shown that even CI users who achieve high accuracy on conventional speech recognition tests may still expend significant cognitive effort to understand speech, particularly in challenging environments (Patel et al., 2023; Winn and Teece, 2021, 2022). Measures such as pupillometry and reaction time tasks are used to assess cognitive demand, providing a more comprehensive picture of the listening experience beyond speech recognition accuracy. For example, a recent study by our group (Patel et al., 2023) demonstrated that CI users experience increased listening effort when comprehending sentences produced by specific talkers, despite showing similar levels of comprehension accuracy across talkers. Further, talker-specific effects were potentially related to the talker’s idiosyncratic speaking style or rate. These findings suggest that talker differences may impose additional cognitive demands, even when speech comprehension appears stable. Together, this growing body of literature highlights a limitation of conventional speech recognition tests: they may obscure underlying effort or variability in information processing, leading to an incomplete picture of CI outcomes. Understanding patterns of listening effort may help guide personalized interventions aimed at improving real-world listening experiences.

Speech perception testing has been instrumental in demonstrating the effectiveness of CIs and has been firmly embedded in clinical guidelines and candidacy criteria. As such, it has enabled hundreds of thousands of patients to regain functional hearing and improve aural communication. However, the vast new knowledge about speech perception mechanisms in normal, impaired and electric hearing accumulated since the early days of CIs has had limited penetration into clinical practice. Although potentially discouraging, this is hardly a new predicament. As early as 1984, when CIs were just beginning to be recognized and accepted as a viable treatment option for adults with SNHL, Walden (1984) in his introduction to an edited volume on speech recognition by the hearing impaired remarked on the slow adaptation of novel tests. Specifically, he wrote “In recent years, several new test materials, methodologies, and analyses have been suggested. Yet, due to a variety of limitations, none have received widespread acceptance. As a result, despite our substantially increased knowledge of speech recognition by the hearing impaired, old testing approaches largely persist. Although this may be disturbing, it is probably quite predictable. In the absence of a universally acceptable alternative, and faced with the necessity of doing something, clinicians and researchers are likely to resort to old, familiar approaches despite their obvious limitations.” More than 40 years later, Walden’s words continue to ring true, being as applicable to the present moment.

Our theoretical understanding of the intricacies of speech perception has considerably broadened over the last half century, resulting in new comprehensive information-processing models that address interactions of sensory bottom-up and cognitive top-down factors (Pichora-Fuller et al., 2016). In parallel to that, breakthroughs in electronic speech recognition and synthesis have made them commonplace technologies in everyday communications. These advancements have in turn been accompanied by the development of numerous novel tests and training materials that can be used to assess and modify more specific aspects of speech perception such as accents and dialectal variation, effort, context and listener experience. To date, few of these newly developed tests are routinely utilized in clinical settings, and as illustrated in Figure 1, the original monosyllabic word test, CNC, has remained most common in assessing performance of CI users. But there is also progress -- CNC has been closely followed by AzBio sentence test, which is the second most common, and has been developed in later years to include a variety of talkers and difficulty levels through the use of variable SNRs of multitalker babble.

Over the last decade, both tests, CNC and AzBio, have become the most widely used in clinical settings. Although many factors play a role in the choice of a clinical speech perception test, the successful adaptation of AzBio and the continuing reliance on CNC in CI clinics in North America, may reflect their complementarity and practical utility. Monosyllabic words in quiet with no additional context can indicate the integrity of bottom-up sensory integration and low-level phonemic-lexical encoding, while complex AzBio sentences, especially when delivered in noise, can provide an estimate of speech perception in more challenging real-world conditions. Both tests have a considerable range of difficulty levels that encompass the performance levels of most patients, and can thus be effectively used for establishing candidacy and outcome monitoring. On the other hand, the two tests are less useful in revealing specific challenges that individual CI users encounter in everyday communication environments and informing programming adjustments or rehabilitation targets.

As our understanding of speech perception mechanisms and CI outcomes evolves, we must keep asking: Do our speech perception tests capture the complexity of real-world speech communication, and meet the diverse priorities and needs of adult CI users? It is highly unlikely that any single test can satisfy all demands. However, some tests may better align with patient-specific goals. Currently, no common integrated theoretical framework exists that guides selection of specific tests and their interpretation (Pisoni et al., 2008). Being aware of the limitations of specific tests can ensure that optimal tests are chosen, while the choice of a speech perception test should be based on assessment goals and informed by patient-specific communication needs. For example, updates to MSTB battery can include brief tests that also tap into broader linguistic and cognitive aspects of speech perception, as well as additional PROMs that reflect self-reported benefits. In addition, comprehensive remote batteries, such as BASE battery or similar, could be used for augmenting MSTB protocols with self-administered assessments that profile multiple aspects of auditory and cognitive functioning that are not captured by MSTB. These additional tests could guide selection of individualized rehabilitation targets.

Future research should prioritize assessments that bridge the gap between clinical and laboratory-based measures and everyday communication experiences. Priority should be given to the development and validation of new measures that also satisfy the following characteristics: (1) assess patient-specific communication needs, (2) are quick and easy to administer with no or minimal assistance from a clinician, and (3) provide actionable recommendations for improving patient-specific outcomes. Although the demands of clinical settings constrain the number and length of testing that can be completed, new online technologies enable patient testing in their unique communication environments, potentially providing more accurate and individually-tailored results. However, to be widely adopted, new tools must undergo rigorous validation, including demonstration of reliability, sensitivity to change, and clinical utility. Future CI rehabilitation may also be further enhanced by speech perception training that can be tailored to specific patients’ communication needs. For instance, the ability to quickly synthesize speech materials based on a given speaker’s vocal transfer function can result in fast and inexpensive development of a patient-specific sets of training utterances targeting the voices of conversational partners the patient is most interested in communicating with and their specific dialectal preferences (Kadam et al., 2021; Azzuni and El Saddik, 2025). Indeed, past research indicates that speech perception training using the utterances produced by talkers with whom patients communicate most frequently leads to greater perceived benefit (Tye-Murray et al., 2016). Furthermore, interactive electronic conversational agents, already ubiquitous in daily life (e.g., smartphones, Alexa, Siri), may be leveraged as rehabilitation platforms for dynamic conversational activities, thus combining speech perception and production training customizable to patient-specific needs. By integrating behavioral speech perception tasks, cognitive-linguistic assessments, PROMs, and qualitative methods, research can gain a holistic view of CI users’ everyday communication experiences, abilities, and needs. These advancements will not only enhance the assessment of real-world CI outcomes but also inform individualized intervention strategies that optimize communication success and quality of life for adult CI users.