First, change were categorized regarding height: symbol, term, phrase, and you may sentence/clause. Zero transform were made at section top since pupils had been coached so you’re able to edit as opposed to change L2 texts. Up coming, the new mistakes you to definitely took place new revised texts was basically than the those in this new L2 texts and you can were analyzed considering if or not the fresh new errors have been corrected (confident change), remained incorrect (basic transform), or the newest errors was indeed placed into the changed text (bad alter). Finally, the final mistakes have been classified towards 12 mistake models.
Through the data, per judges very carefully tested this new L2 trials, classified brand new errors, cross-looked at for each and every other people’s analysis, talked about inaccuracies between the two, drew opinion, and you may went on to make use of the brand new decided-through to measures to analyze this amazing trials. Using proceeded dialogues while in the investigation, the newest boffins been able to identify and get rid of the majority out of inaccuracies. The latest inter-coder reliability of the finally study shortly after several rounds of cross-examining hit 93%.
All the errors found in the L2 and you can modified messages was in fact recorded regarding error versions, and you can multiple steps out-of quantitative analyses have been conducted. Basic, detailed analytics was in fact utilized to dictate this new wavelengths and you may technique of for each and every mistake kind of. Next, t-assessment was in fact held to verify whether people significant distinctions were seen for every single error sorts of plus the final amount away from problems regarding the 
completely new L2 and you may revised texts. Second, the young had been classified towards around three communities in accordance with the regularity regarding errors on brand spanking new versions. Communities step 1 (Letter = 19), 2 (N = 23), and you can step three (N = 16) made up children just who introduced anywhere between step 1 and you will 10, between 11 and you will 20, as well as 20 problems for every single 100 terms, respectively. ANOVA was useful to contrast the outcome, for instance the amount of problems, changes, modifications, and you will the latest problems, as listed on the students’ messages among grammar proficiency teams. Finally, regression try presented to decide if the level of transform made by the people swayed the final number of mistakes on modified texts.
cuatro. Performance
The common term number of students’ modern L2 messages is , in addition to mistakes you to definitely taken place inside particularly L2 texts have been analyzed in terms of the a dozen error products. At least appear to going on mistake sorts of is actually Fragments, followed closely by Word Purchase and you will Incomprehensibility. An average level of complete errors occurring on the L2 messages try each 100 created conditions.
Results revealed that the word count increased (M = ) in the revised texts, whereas the number of errors per 100 words was reduced in all areas of the revised texts. The results also show that the error-free T-units of the revised drafts significantly increased (p < 0.001) in comparison to the original L2 texts. Table 3 reveals that the students made 7.02 changes per 100 words on average in the revised texts. Regarding the level of change, students made changes most frequently at the word level, followed by the changes at the phrase, clause, and symbol levels.
Table 2 gifts the outcome with regards to the regularity out of mistakes, where Insertions/Deletions is actually one particular apparently occurring error variety of, with Noun Plurals and you can Substitutions
Since shown when you look at the Desk cuatro, the students produced self-confident alter (mistake correction) usually, followed closely by neutral change (no change) and you can negative transform (introduction of brand new mistakes). The bottom line is, the outcome showed that pupils enhanced its modified texts because of the dos.67 mistakes for each and every one hundred words. Taking a closer look, the results provide research you to MT oftentimes assisted students to help you create corrections in regards to the errors inside the Stuff, Prepositions, Noun Plurals, and you can Substitutions. The fresh imply of your own full volume out-of errors on modified messages are for each one hundred conditions when compared to on L2 messages. The outcomes of t-decide to try affirmed that the differences when considering the 2 designs was in fact tall for everybody portion apart from Fragments and you will Incomprehensibility (Dining table dos).