Chapter 3: The Pilot Study
3.7 Results
3.7.2 Pretest and posttest
A paired-samples T test was conducted with the pre and posttest scores of the TGJT. The column comprising the p values on table 5 shows that there was a significant increase in mean score of the in vs. on rule from pre to posttest for both the metalinguistic instruction group, and for the processing instruction group. However, there was no significant increase for the group receiving textual enhancement. The column displaying the p values on table 6 also indicates that none of the three groups
had a significant increase in mean score of the by vs. in/on rule from pre to posttest on the TGJT.
Table 5
Results from the TGJT Pre to Posttest of the rule (in vs. on) for each of the Conditions
Pretest Posttest 95% CI Condition n M (SD) n M (SD) p LL UL Textual Enhancement 5 5.00 (1.58) 5 5.20 (2.48) 0.861 -3.164 2.764 Metalinguistic Instruction 5 3.60 (2.07) 5 7.60 (2.40) 0.009 -6.322 -1.677 Processing Instruction 5 3.80 (1.78) 5 7.00 (2.34) 0.045 -6.291 -0.108 Note. CI = confidence interval; LL = lower limit; UL = upper limit.
Table 6
Results from the TGJT Pre to Posttest of the rule (by vs. in/on) for each Condition Pretest Posttest 95% CI Condition n M (SD) n M (SD) p LL UL Textual Enhancement 5 6.00 (1.87) 5 8.00 (1.22) 0.154 -5.165 1.165 Metalinguistic Instruction 5 6.80 (2.04) 5 5.6 (0.89) 0.109 -0.418 2.818 Processing Instruction 5 7.00 (2.64) 5 7.60 (2.07) 0.683 -4.386 3.186 Note. CI = confidence interval; LL = lower limit; UL = upper limit.
A paired samples T-test was also conducted with the pre and posttest scores of the FIB test. The column displaying the p value on table 7 indicates that the textual enhancement group and the metalinguistic instruction group did not learn the rule in vs.
on from pre to posttest. Only the processing instruction group learned this rule. However, this group also had the lowest score in the pretest which partially explains why this group had a significant learning effect from pre to posttest in contrast to the other two groups which scored higher on the pretest and did not have a learning effect. The column containing the p value on table 8 shows that none of the groups learned the rule by vs. in/on from pre to posttest.
Table 7
Results from the FIB test Pre to Posttest of the rule (in vs. on) for Each Condition
Pretest Posttest 95% CI Condition n M (SD) n M (SD) p LL UL Textual Enhancement 5 5.2 (2.28) 5 4.6 (1.81) 0.5 -1.655 2.855 Metalinguistic Instruction 5 5.0 (2.82) 5 6.8 (4.43) 0.52 -9.061 5.461 Processing Instruction 5 4.0 (1.58) 5 9.0 (1.41 < .001 -5.877 -4.122 Note. CI = confidence interval; LL = lower limit; UL = upper limit.
Table 8
Results from the FIB test Pre to Posttest of the rule (by vs. in/on) for Each Condition
Pretest Posttest 95% CI Condition n M (SD) n M (SD) p LL UL Textual Enhancement 5 8.2 (2.68) 5 8.8 (1.64) 0.68 -4.386 3.186 Metalinguistic Instruction 5 5.40 (3.20) 5 8.00 (2.34) 0.19 -4.483 -0.716 Processing Instruction 5 9.0 (1.00) 5 92. (1.09) 0.704 -1.56 1.1601 Note. CI = confidence interval; LL = lower limit; UL = upper limit.
A paired-samples T test was also conducted with the pre and posttest scores of the MKT. As shown on the column displaying the p value on table 9, there was an increase in mean scores of the in vs. on rule in the MKT test from pre to posttest for the processing instruction group, but no increase for the metalinguistic instruction group or for the textual enhancement group which scored zero points in both the pretest and the posttest. The three groups’ showed evidence of no metalinguistic knowledge of the rule in vs. on from the beginning of the experiment. This floor effect on the pretest simply shows that learners were not able to verbalize the target rules in spite of having some knowledge on the TGJT and FIB pretest. The two groups that learned this rule due to the treatment sessions scored above chance on the immediate posttest suggesting that the increase in metalinguistic knowledge was not due to a low score at the beginning of the experiment, but because of the explicit treatments.
The column with the p values on table 10 also indicates that the same pattern emerged with the mean scores of the by vs. in/on rule where there was an increase from pre to posttest for the processing instruction group, but no increase for the metalinguistic instruction group nor for the textual enhancement group, which again scored zero points in the pretest and the posttest. This floor effect on the pretest shows that although participants had scored above chance on the TGT and FIB pretest, they were not able to explain this target rule prior to the treatment sessions.
Table 9
Results from the MKT Pre to Posttest of the rule (in vs. on) for Each Condition
Pretest Posttest 95% CI Condition n M (SD) n M (SD) p LL UL Textual Enhancement 5 0 5 0 Metalinguistic Instruction 5 0 5 5.80 (5.31) 0.071 -12.39 0.793 Processing Instruction 5 .60 (0.89) 5 10.0 (0) < .001 -10.51 -8.289 Note. CI = confidence interval; LL = lower limit; UL = upper limit.
Table 10
Results from the MKT Pre to Posttest of the rule (by vs. in/on) for Each Condition
Pretest Posttest 95% CI Condition n M (SD) n M (SD) p LL UL Textual Enhancement 5 0 5 0 Metalinguistic Instruction 5 0 5 5.80 (5.31) 0.071 -12.39 0.793 Processing Instruction 5 .80 (1.78) 5 10 (0) < .001 -11.42 6.978 Note. CI = confidence interval; LL = lower limit; UL = upper limit.
To analyse if there was a significant increase in mean scores between groups over time for the TGJT, a Mixed Factor ANOVA was conducted with the pre and posttest scores of in vs. on items as a within subject factor and with the group treatment as a between subject factor. No significant interaction between treatment conditions and improvement over time was found, Wilk’s λ = .606, F(2, 12) 3.909, p = .050, η2= .394, 1-β = .588. The same procedure was followed with the items of by vs. in/on,
and there was no interaction between treatment conditions and improvement over time, Wilk’s λ = .731, F(2, 12) 2.206, p = .153, η = .269, 1-β = .363.
In order to test whether there had been a significant increase in mean scores between groups over time for the FIB results, I conducted a Mixed Factor ANOVA with the FIB pre and posttest scores of in vs. on items as a within subject factor and with group treatment as a between subjects factor. No significant interaction was found for any of the groups’ pre to posttest scores over time, Wilk’s λ = .658, F(2, 12) = 3.116, p = .081, η2 = .342, 1-β = .490. For the items by vs. in/on, the same procedure was followed, and there was also no interaction for any of the groups’ pre to posttest score over time, Wilk’s λ = .756, F(2, 12) 1.937, p = .187, η2 = .244, 1-β = .323.
A Mixed Factor ANOVA was also conducted with the MKT pre and posttest scores of in vs. on items as a within group factor and with group treatment as a between group factor. A significant interaction between treatment conditions and improvement over time, Wilk’s λ = .340, F(2, 12) 11.634, p = .002, η2 = .660, 1-β = .973, was found. A Bonferroni Post Hoc test indicated a significant gain for the metalinguistic instruction group (p = .036) and for the processing instruction group (p < .001).
For the MKT pre and posttest scores of by vs. in/on a Mixed Factor ANOVA revealed that there was a significant interaction between treatment conditions and improvement over time, Wilk’s λ = .367, F(2, 12) 10.338, p = .002, η2 = .633, 1-β = .955. A Bonferroni Post Hoc test indicated that there was a significant increase from pre to immediate posttest for the metalinguistic instruction group (p = .045) and also a significant increase from pre to immediate posttest for the processing instruction group (p = .001).