Background 

To build on the newly adopted intelligibility-based approach to pronunciation in our Centre, John, Venus, and Yin wished to explore the notion of the intelligibility threshold, addressing these three questions: 1) What is the intelligibility threshold; 2) How do you measure it; and 3) Are those existing intelligibility thresholds applicable to our context? A systematic analysis of five studies apparently could shed some light on the three questions. Hopefully, the results of this study will help teachers understand descriptors related to pronunciation, and ultimately improve inter-rater intelligibility. 

The premise underlying the notion of the intelligibility threshold (IT)

The concept of intelligibility (IT) or the intelligibility threshold could be challenging to grasp as listeners or teachers may determine the intelligibility of a speaker while assessing comprehensibility and/or acceptability as well. In fact, the analytic scale proposed by Issacs et al (2018) addresses comprehensibility from the listener’s perspective, whether the listener has to put in little effort, some effort, enormous effort to understand the speech or not.  Therefore, in order to explore the notion of the intelligibility threshold and to ensure that all teachers are on the same page, we could first of all examine three related terms – acceptability, comprehensibility, and intelligibility, and how they are commonly defined in the field or context of pronunciation assessment.  

• Acceptability is defined as “The degree of annoyance and irritability experienced by listeners” (Kozlowska, 2014, as cited in Thomson, 2017, p. 25).
• Comprehensibility (Gass and Varonis, 1984, as cited in Munro and Derwing, 1995, p.77) is defined as “ease of interpretation”, or viewed as “a measure of the processability of speech” (Thomson, p.24).
• Intelligibility refers to “the extent to which a speaker’s message is actually understood by a listener” (Munro and Derwing 1995, p.76) or “actual level of understanding” (1995, p.92) 

What is the intelligibility threshold (IT)?

Depending on the context and the purpose of communication, the index or percentage varies. According to a study conducted by Okim et al (2020), the intelligibility threshold can be expressed by an index, or a percentage of the total number of content words understood by a listener, that is, a ratio of the total number of content words understood by a listener and the total number of content words spoken by a speaker. The researchers proposed an index as high as 96% (in other words, a divergence rate of 4%) as they were looking for speakers who would play role as speakers reading aloud passages for a listening comprehension test, that is, the TOEFL listening paper.

Measuring the IT focusing on content words has limitations, one of which is that the importance of prepositions in phrasal verbs (verb + preposition) and other function words (conjunctions, pronouns, determiners) that may carry special meanings or enhance understanding is ignored. Other researchers may also investigate various aspects of the individual listener such as accent familiarity, topic familiarity, listeners’ voluntary and deliberate efforts to decipher speech when determining one’s intelligibility (Hardman, 2010; Munro & Derwing, 1995).

Another study carried out by Kosheleva and Kreinovich (2013) attempted to quantify the intelligibility threshold. In their investigation into dialects and languages, they argued that if 70% or more of words spoken by a person from one region are understood by other speakers from another region, then it is usually considered a dialect; however, if less than 70% of words are mutually intelligible, they are usually considered as different languages. The researchers mathematically demonstrated that this equation p > p2 def = √ 0.5 or 70.71% could suggest the intelligibility threshold.

Other methods of collecting speech data for measuring the intelligibility threshold

There are at least three ways to collect speech data. 

1. Subjects read aloud a script usually consisting of various grammatical and syntactic structures such as relative clauses, complement clauses, prepositional phrases, and so on.
2. Subjects write their scripts and read them aloud. (e.g. SIPE and oral presentations; usually scripted, some improvisation may be required if a speaker forgets some part of the speech) 
3. Subjects are engaged in a spontaneous conversation. (e.g. mock job interview, improvisation is usually required when answering unprepared questions) 

Are those existing intelligibility thresholds applicable to our context? 

Arguably, our IT should not be as high as the one identified by Okim et al’s. Their purpose was to select speakers for reading aloud passages for listening comprehension purposes in the context of developing test papers. Similarly, the IT for a platform manager working on a train station will not be as high as the IT for an air traffic controller working at the airport. Further research into alternative methods may be needed for determining whether our descriptor “pronounces quite clearly and accurately; and uses rhythm and intonation quite appropriately in general” or whether other qualitative statements can serve as our intelligibility threshold for our Center.

References 

Hardman, J. B. (2010). The intelligibility of Chinese-accented English to international and American students at a U.S. university (Order No. 3428659). Available from ProQuest Dissertations & Theses A&I; ProQuest Dissertations & Theses Global A&I: The Humanities and Social Sciences Collection. (758452801).

Issacs, T., Trolimovich, P., & Foote, J.A.  (2018). Developing a user-oriented second language comprehensibility scale for English-medium universities. Language Testing, 35(2). https://journals.sagepub.com/doi/full/10.1177/0265532217703433 

Kang, O., Thomson, R. I., & Moran, M. (2018). Which features of accent affect understanding? Exploring the intelligibility threshold of diverse accent varieties. Applied Linguistics, 41(4), 453–480. https://doi.org/10.1093/applin/amy053

Kosheleva, O., & Kreinovich, V. (2013). Dialect or a new language: a possible explanation of the 70% mutual intelligibility threshold. International Mathematical Forum, 9(4), 189-192.

Munro, M. J. & Derwing, T. M. (1995). Foreign Accent, Comprehensibility, and Intelligibility in the Speech of Second Language Learners. Language learning, 45(1), 73-97.

Munro, M. J., & Derwing, T. M. (2015). Intelligibility in research and practice: Teaching priorities. In M. Reed & J.M. Levis (Eds). The handbook of English pronunciation. Wiley Blackwell: Malaysia. 

Thomson, R. (2017). Measurement of accentedness, intelligibility, and comprehensibility. In O. Kang & A. Ginther (Eds). Assessment in second language pronunciation. Routledge: NY. 

Short Descriptions

Aim Building Business English Students’ Confidence in Using English is a small corpus-based project aiming at, as the title suggests, helping business students become even more confident and effective communicators. Rationale Using a linguistic approach, students work on various subjects, including syntax, semantics, and sounds, with authentic materials taken from their coursebooks. In this way, students will kill two birds with one stone, or rather, engage in a win-win situation because they can learn not only language skills, but also revise their subject knowledge. Even though Business students in general are more able students when it comes to communication skills, they may still benefit from working on topics such as the end-weight principle and parallelism under syntax; hyponyms and hypernyms under semantics; English stress rules such as rules governing stress falling on the penultimate syllable or ante-penultimate syllable under sounds or phonology.

Possible Benefits

Further Research The materials will be tried and tested. If the project, in particular, the materials, receives positive feedback, other subjects such as Finance, Marketing, and Information Systems could be examined using this approach.

Deliverables

Deliverables Student helpers/interns were hired to assist in developing materials for this project. With some basic training in corpus linguistics and using AntConc, a tool for analyzing text, student helpers/interns built a corpus focusing on Macroeconomics, Microeconomics, and Accounting with a size of over 740,000 words, and created learning materials addressing syntax, semantics, and sounds and the first draft of materials have been uploaded onto Lumi Syntax: https://app.lumi.education/run/u6z-39 Semantics: https://app.Lumi.education/run/xXRrDm Sounds: https://app.Lumi.education/run/iGLZ1U A document about FAQ has been created by a student helper/intern as well FAQ.docx.