**4. Implications and future research directions**

There are some important implications that researchers and educators can draw from the findings of the existing research investigated in this desktop review. First, researchers should be informed of the fact that NSs' judgments of NNSs' speech performance involve a varying degree of listener bias. Any research that involves subjective judgments of NS listeners should control such confounding variables and curtail their intervening effects. Second, intelligibility is as much a function of listeners' effort as that of speakers, meaning that the intelligibility of speech is not solely determined by speakers but by listeners. In measuring the intelligibility of an L2 speech variety, listeners play an active role. A speech sample might be highly intelligible for a particular group of listeners but less intelligible for another group. Communication is a two-way street. Sharing responsibility between both speakers and listeners would be the first step to take for successful communication. Finally, acknowledging the active role of listeners in determining L2 speech intelligibility challenges the research instruments prevalently used in research for measuring speech intelligibility. The stimuli used in intelligibility-based research are recorded speeches that do not provide an opportunity for listeners to employ effective communication strategies and to negotiate meaning with speakers [28]. Thus, intelligibility remains to be determined still solely by speakers. Future research should take the initiative to utilize interactive stimuli in which intelligibility is more realistically measured as a joint effort between a listener and a speaker as it is what is expected to happen in real-life interactions.

Given the importance attached to listeners as active contributors to speech intelligibility and communication success between NSs and NNSs, L2 pronunciation educators would require to dedicate a substantial amount of their instruction to introducing and practicing successful communication strategies such as how to paraphrase, do circumlocution, and use nonlinguistic signals. Additionally, in lieu of setting NS pronunciation norms and encouraging L2 learners to approximate their pronunciations to those norms, educators should emphasize the pronunciation features which contribute more to speech intelligibility and prioritize them over those less critical features in their instruction [28]. As of yet, research on the efficacy of this type of instruction is scarce. Future research will be needed to determine to what extent this type of instruction promotes the communication between NS and NNS.

It is also important to note that those L2 learners who experience varying degrees of bias in their encounter with NSs tend to report having lower self-efficacy expectations, which would, as a consequence, impair their self-estimate of their language abilities. Self-efficacy expectations would predict one's performance beyond their true abilities. Those L2 learners with lowered self-efficacy expectations would be less willing to initiate and maintain interaction with NSs both in classroom and

real-life settings. This reluctance on the part of L2 learners would exacerbate the existing bias toward them by NS. To mitigate the bias, if not eliminate it, educators can devise interventions which would foster L2 learners' self-efficacy through strategy instruction and other methods such as providing them with supportive mentors. By fortifying L2 learners' self-efficacy, they would be more encouraged to participate in interactions involving NSs. As a result of more interaction, the intensity of bias held by NSs toward that particular group of NNSs decreases.

With respect to listener background variables, researchers should be cognizant of the fact that some listener-related variables such as listeners' first language status and the amount of their exposure to accented varieties of English are more influential to listeners' perceptions of NNSs' speech than other variables (e.g., listeners' linguistic knowledge and their prior teaching experience). Thus, those variables warrant more attention in selecting listeners for rating NNSs' speech, as they might exert more potent influence and compromise the reliability of the assessment. On the other hand, as pointed out by Kang et al. [3], the effects of the latter variables are more contextually determined and should be considered regarding the type of assessment being administered. For example, if the purpose of speech assessment is evaluating the nuances of pronunciation, listeners' linguistic knowledge and their educational background should be considered for selecting listeners as raters. Additionally, the way that the prior studies have operationalized listener variables needs to be rectified. For example, Kang et al. [3] operationally defined listeners' linguistic sophistication as a function of three factors: (a) the number of foreign languages they spoke, (b) the number of years they studied foreign languages, and (c) the number of linguistic courses they had taken. However, this definition needs to be improved given that with the spread of globalization and emergence of English as an international language (EIL), the concept of foreign language is becoming ambiguous. Thus, future research should take initiatives to address these limitations marked in the prior studies.
