H6 An example RESULTS section for advanced students

Here is an advanced reporting of the mnemonic study described in the book. Your reports may never become as complex as this. However, you will probably be expected to include at least some of the additional output reported below at some stage. Do not be daunted by this version of the Results! Simply use the relevant parts of this as an illustrative example of how you might report such statistics, as and when you need to. (The table and figure referred to can be found in the example RESULTS in Section 4.6.10 of the book.)

Results

The data met the assumptions of normality, homogeneity of variance and, for the analysis involving the related measures variables, sphericity. There were no missing data. Given the number of separate tests of significance employed, Bonferroni adjustments were made to the significance level to limit the experimentwise type I error rate to an alpha of .05. Individual tests of significance were thus carried out with alpha of .01.

Analysis of the mean recall scores on the pretest using one-way analysis of variance (ANOVA) for unrelated measures with instruction (mnemonic or no mnemonic) as the independent variable indicated that performance was equivalent in the two conditions, (mnemonic group, M = 10.95 words recalled, SD = 1.05; no-mnemonic group, M = 11.15 words recalled, SD = 1.14), F (1, 38) = 0.33, p = .99. Ninety five percent confidence intervals for the two groups were 10.62-11.68 (mnemonic) and 10.46-11.44 (no mnemonic), with partial η² = .01. The narrow confidence intervals and extremely small effect size suggest that it is reasonable to assume that the groups had equivalent recall abilities for lists of words prior to the experimental group being taught how to use the mnemonic.

The manipulation check on the imageability of the words used in the experiment was also satisfactory. Analysis of participants' ratings of the imageability of these words used one-way ANOVA for related measures with imageability (easily imaged or hard to image) as the independent variable. This revealed significantly higher ratings for the easily-imaged words ( M = 5.21, SD = .77) than for the hard-to-image words ( M = 3.39, SD = .95), F (1, 38) = 151.72, p < .001.

Given these satisfactory findings, I proceeded to analyse the principal dependent measure, the mean numbers of words of each type correctly recalled by those in the two conditions, excluding misspellings (Table 2). Analysis employed 2 x 2 ANOVA for mixed designs with imageability (easily imaged or hard to image) as a related measures variable and instruction (mnemonic or no mnemonic) as an unrelated measures variable.

Analysis of the data in Table 2 revealed a statistically significant main effect of instruction, F (1, 38) = 7.20, p = .01. Those in the mnemonic group recalled more items overall than did those in the no-mnemonic group ( M = 15.65, SD = 3.97; M = 12.40, SD = 3.74 respectively). Partial η² = .16, indicating that 16% of the overall variance was attributable to this manipulation. There was also a statistically significant main effect of imageability, F (1, 38) = 145.22, p < .001, with more items from the easily imaged list being recalled than from the hard to image list ( M = 15.98, SD = 4.12; M = 12.08, SD = 4.48 respectively). Partial η² = .79, indicating that 79% of the overall variance was attributable to this manipulation. However, the Instruction X Imageability interaction was also statistically significant, F (1, 38) = 11.55, p = .002. Figure 1 displays this interaction. Partial η² = .23, indicating that 23% of the overall variance was attributable to the interaction between the variables.

To analyse the significant two-way interaction, I ran tests of the simple effects of instruction at the levels of imageability. These followed the procedures recommended by Bowel (2002). These tests revealed that the mnemonic group recalled significantly more of the easily imaged words than did the no-mnemonic group, F (1, 43) = 11.88, p < .01. However, the mnemonic group did not recall significantly more of the hard to image words than did the no-mnemonic group, F (1, 43) = 2.90, ns . The findings are thus consistent with the experimental hypothesis.

The majority of participants (18) in the no-mnemonic condition claimed to have attempted to remember the items by rote repetition, with the remainder employing simple attempts at organizing the material into semantically related clusters. All those in the mnemonic condition reported using the method of loci , and most of these expressed their surprise at the impact this appeared to have had on their capacities to recall.