O n f e r e n c e

The design-based research of the GALA system 
A design-based approach to the instructional design of the GALA system is to justify the 
theory-driven underpinnings that facilitate the acquisition of grammar. The prototype 
program of the learning tasks has been designed (Liu & Chao, 2012), including (A) 
structured input for processing instruction (VanPatten, 1990; VanPatten & Cadierno 
1996), (B) feedback, (C) 4 grammar areas, and (C) 4 levels of difficulty. 
(A) Structured input 
I like swimming better than to dive. 
I like swimming better than diving. 

---

-84- 
2014 CALL Conference 
LINGUAPOLIS
www.antwerpcall.be 
(B) Feedback design 
Example: I like 
swimming
better than 
diving
. 
(C) Grammar areas (GA) 

GA1: Verb-tenses consistency (VT) 

GA2: Subject-Verb agreement (SV) 

GA3: Consistency of pronounce (CP) 

GA4: Consistency in a series of forms (SF) 
(D) Levels of difficulty 

Level 1: one of the grammar topics in a task (4 options). 

Level 2: two combined grammar topics in a task (6 options) 

Level 3: three combined grammar topics in a task (4 options) 

Level 4: The acquisition level that involves all the four grammar topics 
The GALA program was first designed with adaptive learning for users to choose the level 
they want to try and learn and the first phase of DBR was conducted via the GALA 
contest held in the English department in 2012. A total of 26 students participated and 
completed the questionnaire of System Usability Scale (SUS). Results show that those 
went through progressive levels of learning outperformed those who tried the targeted 
Level 4 most. The second phase of quasi-experiment was conducted in two classes of 
Freshman Composition to compare the two learning paths of advancing levels of 
difficulty. Class P (progressive learning, N=23) started at Level 1 and moved to the next 
Level 2 after passing all of Level 1, while Class A (adaptive learning, N=21) could choose 
whatever level to start but the level locked and forced them do the lower level if they 
failed to pass the chosen level. Results of pre-post paired samples t-tests of Level 4 
reveal Class P outperformed Class A.