This past weekend I attended the Second Language Research Forum conference in Pittsburgh, Pa. It was full of so much information that I think will really move me further along my path to teaching in Korea!

One presentation was titled “Large-Scale Corpus Analysis of Korean Learners of English Segmental Errors” and it was–you guessed it–a large-scale corpus analysis of Korean learners of English segmental errors. I couldn’t believe my luck! I have an interview this Thursday night with an English-teaching program in Korea and I get (what I feel like) is this insider-information that will allow me to speak in a very knowledgeable and specialized way about difficulties learners may face.

So anyway, here is the rundown:

They performed a contrastive analysis for the initial set of errors using American English and Korean Phonemes. Here are some of the results:

Korean doesn’t have a voicing distinction (this explains why it’s either Pusan or Busan; Cheju-do or Jeju-do, depending on who you’re talking to), but it does have a manner distinction (lax, reinforced or aspirated, for example).

Additionally, English has many more fricatives than Korean. This means that when teaching English pronunciation, Korean learners will most likely substitute sounds that English requires (like θ or ð) with sounds they have in Korean (like s). Additionally, there is no tense/lax vowel distinction in Korean, so words like /hɪt/ may surface as /hit/ when pronounced by a Korean learner of English.

Because Korean does not allow consonant clusters (ex: firST; SPeak), learners may insert a vowel (also called vowel epenthesis) between consonants. The analysis also showed a tendency to delete the semivowels /j/ and /w/ so that words like “wood” are pronounced /ud/.

Korean learners of English might also affricatize their z sounds or palatize their s sounds; change English dipthongs to monopthongs or simplify consonant clusters. Additionally, there are a number of error tendencies regarding unstressed vowels.

Surprisingly, the world-famous r-l issue accounted for only 4 percent of errors uncovered by this corpus.

These errors are interesting because they are backed by data. When performing contrastive analysis, it is easy to get lost in how much “intuitive” sense it makes to study the gaps between the L1 and the TL (first language and target language) and forget that these predicted errors are often not supported by data. This is why the error analysis performed in this study has proved to be a more effective way to understand learner errors.