TANDEM LANGUAGE LEARNING THROUGH A CROSS-CULTURAL KEYPAL PROJECT

Kaoro Kabata, Univeristy of Alberta, Canada

Yasuyo Edasawa, Doshisa Woman’s College, Japan

TANDEM LANGUAGE LEARNING THROUGH A

CROSS-CULTURAL KEYPAL PROJECT

Penpal ? Regular Project between two universities Opportunities for cultural and language learning Ideal opportunity for incidental learning Message exchanges on a discussion board

Why “Keypal Project”?

Semi-classroom based research

Quantitative research

Types of research

Canadian students (UA) were required to report learning notes 4 times about WHAT they learned and HOW. see

Their reports will be graded in their learning course.

Japanese students (DWC) were just asked but not required to report what they learned, noticed and corrected their partners.

370 entries were collected from UA students and 67 entries from DWCs

Data Collection

Data founded

As reported by UA students Linguistic items learned are

Vocabulary(55%) Grammar (22%) Expression (17%) Kanji (6%)

Preferred learning styles are Others (57%) With explicit error correction (31%) Learn through Q&A (6%) Without explicit error correction (5%)

Results

Notice: DWC students were not assigned to report all of what they found so the data would not match with from UA students.

Explicit error correction is provided 29 times with 18 times of recognition

(the recognition information is correspond with 13 entries reported by UAs)

Implicit error correction is provided 13 times with 0 times of recognition

Other errors were found 20 times but no correction provided

Results

UA data indicated that non-explicit correction may not lead to learning as often as explicit correction.

DWC data showed that UA students often failed to recognize their errors without explicit correction.

However, UA data also yielded that implicit correction and correction through negotiation could lead to better understanding of their errors when they noticed. see

Data analysis

Wrong information

Comments

Limited information Weak argument

Possible problematic variable Data Collection

Limitations