Off-line Questionnaire
5.3.2 Off-line Questionnaire
Individual Differences
A One-way ANOVA for materials analysis, with a repeated design, was performed to obtain a measure of individual differences. (This analysis is identical to the one reported in the previous study: Individual differences in attachment preference, p. 105.) Without these differences, there is no variability in attachment preference to explain using reading span. The ANOVA was unorthodox because the main factor was not a measure but participant identity. That is, each level in the factor Participant corresponded to an individual participant. The dependant variable was the binary response for each of the experimental items. If the factor Participant contributes significantly to the variability in attachment preference then we can say that the sample contains individual differences. In other words, differences in questionnaire performance may be attributed in large part to differences among individual participants. Two such ANOVAs were carried out A One-way ANOVA for materials analysis, with a repeated design, was performed to obtain a measure of individual differences. (This analysis is identical to the one reported in the previous study: Individual differences in attachment preference, p. 105.) Without these differences, there is no variability in attachment preference to explain using reading span. The ANOVA was unorthodox because the main factor was not a measure but participant identity. That is, each level in the factor Participant corresponded to an individual participant. The dependant variable was the binary response for each of the experimental items. If the factor Participant contributes significantly to the variability in attachment preference then we can say that the sample contains individual differences. In other words, differences in questionnaire performance may be attributed in large part to differences among individual participants. Two such ANOVAs were carried out
2.8, p < 0.001. Therefore, in both cases, there are significant differences in attachment preference between individuals.
Table 10 shows the correlations in NP1 bias between Phases I and II. None of the measures is significant. If NP2 animacy has had a systematic effect, we would expect a positive correlation for overall NP1 bias, NP1 bias for animate items and bias for inanimate bias because these were manipulated similarly on both occasions. If the effect is non-systematic, then it suggests that uncontrolled variables have influenced the dependent measure. In this experiment the crucial manipulation is the plausibility of the Saxon genitive alternative, and the reaction of higher-span readers to it.
It may be argued the failure to find a correlation between Phases I and II indicates a flaw in the preparation of the materials that renders any subsequent analysis redundant. However, if the plausibility of the Saxon genitive was not correctly manipulated – i.e. one questionnaire contained a greater number of plausible items than another – then this balance may be addressed if we (i) determine plausibility empirically (using an independent sample) and (ii) use this new measure to test the experimental hypothesis. Thus, while there may be no correlation between Phases I and II, we can address the likely factor that underlies this null result and continue to investigate the relationship between reading span and Saxon genitive plausibility.
Table 10: Attachment preference correlations between Phases I and II
Correlation between Phases I and II (Pearson, one-tailed) N = 34
Overall NP1 Bias
0.15 (NS)
NP1 Bias for Animate Items
0.14 (NS)
NP1 Bias for Inanimate Items
0.12 (NS)
Saxon Genitive Manipulation
For the sample as a whole, there was an overall effect of animacy of NP2: T 48 = 2.28, p
< 0.05. The means for this effect are shown in Figure 13. Sentences with an inanimate
NP2 were resolved in a Late Closure fashion (significantly different from chance T 48 =
2.96, p < 0.01), whereas animate NP2 sentences showed a random bias.
Anim ate NP2 Inanimate NP2
Figure 13: Mean percentage NP1 bias for sentences with an animate NP2 and an inanimate NP2 (with standard error).