The following figure and table give the frequency distribution and other descriptive statistics (mean, standard deviation, etc.) for the VZ-2 single-task scores (VZS) using the correction formula (c-i/5):

Figure 6. Frequency Distribution of VZ-2 Single-Task Scores (VZS) for Experiment 1.

Table 4

These scores may then be compared to the following dual-task scores. The figure and table below give the frequency distribution and other descriptive statistics (mean, standard deviation, etc.) for the VZ-2 dual-task scores (VZD) using the correction formula (c-i/5):

Figure 7. Frequency Distribution of VZ-2 Dual-Task Scores (VZD) for Experiment 1.

Table 5

The following figure and table give the frequency distribution and other descriptive statistics (mean, standard deviation, etc.) for the Listening Comprehension Single-Task scores (LCS).

Figure 8. Frequency Distribution of Listening Comprehension Single- Task Scores (LCS) for Experiment 1.

Table 6

These scores may then be compared to the following dual-task scores. The figure and table below give the frequency distribution and other descriptive statistics (mean, standard deviation, etc.) for the Listening Comprehension dual-task scores (LCD), number correct out of 20.

Figure 9. Frequency Distribution of Listening Comprehension Dual- Task Scores (LCD) for Experiment 1.

Table 7

Difference Scores

Difference scores, which were used only to collapse or summarize the data, were computed by subtracting the dual-task scores from the single-task scores for each task -- VZ-2 and Listening Comprehension (LC). Such that:

VZÆ = VZD - VZS

and

LCÆ = LCD - LCS.

A positive difference would therefore indicate an improvement from single- to dual- task performance, whereas a negative difference would indicate a decrement in performance.

It was hypothesized that the difference scores between the VZS and the VZD would be less for high SVA individuals than for the low SVA individuals. However, that result was not found. As it turned out a regression analysis showed that an individual's VZ-2 score and difference score were related (F(1,58) = 4.93, p<.05) but that it was the low SVA performers who showed less of a decrement in performance, not the high SVA individuals. Figure 10 shows the relationship between VZ-2 Scores and Difference scores (r=.28, r2 = .08).

Figure 10. Scattergram of VZ Differences (VZÆ) based on VZ-2 Scores for Experiment 1.

An additional hypothesis was that there would either be no significant difference in the difference scores between LCS and the LCD, or the difference would be less for the high SVA individuals than for the low SVA individuals. This hypothesis was supported as a regression test showed there to be no significant functional relationship between VZ-2 score and LC difference scores (F(1,58) = .09, p>.05, r=.04, r2 Å .00).

Since a marked improvement was shown for the low SVA individuals, a paired t-test was carried out to look at the general overall effect. Results indicate that there was not a general decrement in performance from VZS to VZD (t(59) = .45, p>.05 ). This is not surprising since the high SVA individuals tended to show a decrease in performance while the low SVA individuals tended to show an increase. However, there was a relationship between the scores in the single- and dual- tasks (F(1,58) = 119.13, p<.01, r=.82, r2=.67). The following scattergram shows the relationship.

Figure 11. Scattergram of Relationship between VZ-2 Single-Task Scores (VZS) and VZ-2 Dual Task Scores (VZD) for Experiment 1.

A paired t-test was then carried out to determine whether the general effect occurred for the Listening Comprehension (LC) tasks. In fact, the effect was found (t(59) = -3.44, p<.01). A comparison of the means gave the following values, MLCS = 10.68, MLCD = 9.03, and the scattergram below shows the relationship between LCS and LCD (F(1,58) = 7.65, p<.05, r=.34, r2 = .12).

Figure 12. Scattergram of relationship between Listening Comprehension Single Task Scores (LCS) and Listening Comprehension Dual-Task Scores (LCD) for Experiment 1.

Additionally, difference scores were calculated based on LC scores. For the VZ-2 difference scores, as was expected, there was no significant functional relationship, (F(1,58) = .09, p> .05, r = .04, r2 Å .00), but for the LC difference scores there was a relationship (F(1,58) = 40.57, p < .01, r = .64, r2 = .41) as shown in the following figure.

Figure 13. Scattergram of Listening Comprehension differences based on Listening Comprehension (LC) Scores for Experiment 1.

A means comparison between the two groups, as determined by a median split (Low LC = 0-11, High LC = 12-20), for difference scores gave the following: MLow LC=3.84, MHigh LC=.09 indicating a rather large decrement in performance for the high LC individuals and a very small decrement in performance for the low LC individuals.

Correlations between VZ-2 and LC Scores

At this point it would be useful to look at the relationships in terms of performance among the various tests. Even though some of these correlations have already been studied, it would be beneficial to build a correlation matrix to see how all of the relationships compare.

The correlation matrix created by these comparisons gives the following results:

As can be seen from this matrix, the only strong relationships were between VZS and VZD (r = .82), the scattergram of which is shown previously in Figure 11, and also between LCS and LCÆ as discussed in the previous section.

From these correlations and the previous findings the following table was constructed:

Table 9

This table summarizes some of relationships between the tasks in the single- and dual-task conditions. In the single-task columns this table describes the correlations between Spatial Visualization Ability (SVA) level and each task (VZS and LCS), as well as Listening Comprehension (LC) level and each task. In the dual-task columns, this table describes the changes in performance -- overall decrements and increases in performance. The results of this study show, not surprisingly, there were positive correlations between SVA Level and scores on the visualization tasks (VZS and VZD), as well as between LC level and comprehension tasks (LCS and LCD). In addition, there was no significant correlation found for any of the relationships between SVA level and the comprehension tasks or between LC level and the visualization tasks. Of most interest, though, were the relationships between single- and dual- task performance. As predicted, there were performance decrements in the visualization tasks (from VZS to VZD) for the high SVA individuals, and in the comprehension task (from LCS to LCD) for the high LC individuals. What is surprising, though, is that the low LC individuals and the low SVA individuals actually show less of a decrement (LC) and even an improvement (SVA) in their performance levels from the single- to the dual- task condition.

Reliability/Validity and Internal Consistency Scores

The results of this experiment are only meaningful if the tests used for measurement are themselves internally reliable and valid. Reliability refers to an index of a measurement's accuracy -- agreement between successive applications of a measurement procedure. If a dependent variable has low reliability it may lead to a type II error -- overlooking a potentially significant effect which does, in fact, exist (Kirk, 1982). One way to measure the internal reliability of a dependent variable is to use a split-half reliability test -- comparing the number of odd items answered correctly with the number of even items answered correctly for each participant. For the VZS test a moderately strong correlation was found between the two groups which consisted of ten items each -- odd scores vs. even scores: (F(1,58) = 72.45, p<.01, r = .75, , r2 = .56.) However, to correct for the decreased number of items in each half of the split, the Spearman-Brown formula was then applied, such that rtt=.86.

Figure 14. Reliability for VZ-2 Single-Task test using split half comparison.

For the VZD test a high correlation was found between the two groups (F(1,59) = 245.41, p<.01, r = .90, r2 = .81, and the Spearman-Brown rtt =.95.) It is important to note, however, that in the dual-task condition some subjects completed more than the initial twenty questions as they had in the single-task condition. During the dual-task condition, subjects answered questions for the entire six minute session while they listened to the comprehension exercise.

Figure 15. Reliability for VZ-2 Dual-Task (VZD) test using split half comparison (Note, the number of items may total more than 20 per subject).

As for the Listening Comprehension (LC) task, a moderate correlation was found between the two groups, each with ten questions (F(1,59)=39.21, p<.01, r=.64, r2=.40, and the Spearman-Brown rtt =.78.)

Figure 16. Reliability for Listening Comprehension Single-Task (LCS) test using split half comparison.

For the LCD test a lower correlation was found between the two groups of ten questions each (F(1,59) = 8.09, p<.01, r = .35, r2 = .12, and the Spearman-Brown rtt =.52.)

Figure 17. Reliability for Listening Comprehension Dual-Task test (LCD) using split half comparison.

Individual Scores (Condition Plots)

Since the original hypothesis was not validated by the main test of differences, it was necessary to look at individual scores to gain more information as to the possible effects of learning over the different conditions. A condition plot of each subject's performance over the four tasks was made which expressed each score as a function of the task. Scores were sorted from highest to lowest in terms of (1) VZS, (2) LCS, (3) VZD, and finally, (4) LCD. For the sake of reference, and in order to determine two relatively equal groups, a median split divided the individuals into the following groups: Low SVA = 0-12.8, NLow SVA = 31, High SVA = 12.8-20, NHigh SVA=29. Results of the condition plots showed that those individuals who scored highest overall, exhibited a fairly similar performance curve such that:

1) VZS scores were generally higher than or the same as VZD scores and

2) LCS scores were generally higher than or the same as LCD scores.

(These curves followed the hypothesized behavior.)

A typical plot of a high SVA individual is shown in Figure 18.

Figure 18. Condition Plot for a High SVA individual.

(Note: for VZS and VZD, Score = # Correct using the adjustment formula, and for LCS and LCD, Score = # Correct).

However, for low SVA individuals, performance was much more erratic. While some low SVA individuals showed a similar performance curve to the high SVA individuals, other individuals displayed a variety of trends. More often than for high SVA individuals, the VZD score for the low SVA individuals showed an improvement over the VZS score, but with a decrement to the LCD score, as in this case:

Figure 19. Condition Plot for a Low SVA individual.

(Note: for VZS and VZD, Score = # Correct using the adjustment formula, and for LCS and LCD, Score = # Correct).

But some of the low SVA individuals showed improvements for both the VZ-2 and the Listening Comprehension (LC) exercises, such as in this example:

Figure 20. Another Condition Plot for a Different Low SVA Individual. (Note: for VZS and VZD, Score = # Correct using the adjustment formula, and for LCS and LCD, Score = # Correct)

Ways of Thinking: Subjects were asked to complete a "Ways of Thinking" Questionnaire (see Appendix E) which was actually the VVIQ test from the Richardson battery of imaging tests (Richardson, 1994). A Ways of Thinking score (WOT Score) was calculated by taking the 15 preselected questions as indicated by Richardson and from those, subtracting the number of questions answered in the verbal direction from those answered in the spatial direction. These scores were correlated with each subject's scores on the VZ-2 single-task test (VZS). However, no significant functional relationship was found between the WOT scores and the VZS score (F(1,59) = .003, p>.05, r=.01, r2 Å .00).

Additional Spatial-Verbal Preference Questions: Subjects were also asked to answer some additional questions regarding their spatial and verbal preferences (see Appendix D). As in the case of the WOT score, a Spatial-Verbal (S-V) score was calculated by subtracting the number of questions answered in the verbal direction from those answered in the spatial direction. These scores were correlated with each subject's scores on the VZS test. However, no significant functional relationship was found between the six spatial-verbal preferences survey questions and the VZS score (F(1,58) = .39, p>.05,

r= .08, r2 = .01).

Predictive Scores/ Ordering / Frequency Counts

After completing the experiment, subjects were asked to predict their percent correct on each of the four tasks (VZS, LCS, VZD, and LCD). In addition, they were asked to compare each pair of tasks (e.g. VZS vs. VZD, VZS vs. LCS, etc. see Appendix F for the entire list). As for the accuracy of predictions, scores were compared, such that for each test, the actual proportion correct out of total attempted was compared to the predicted proportion correct (this differs for the VZ-2 from the "adjusted" score which has been used for most of this experiment). For the VZS a small positive correlation was found (F(1,58) = 6.29, p<.05, r=.31, r2= .10)

Figure 21. Scattergram for Actual vs. Predicted scores on the VZ-2 Single-Task (VZS) test for Experiment 1.

A calculation of the Goodman-Kruskal gamma correlation (G), which is a measure of association (Nelson, 1984) gave the following result, G=-.50. The calibration curve plotted below shows the means of the actual scores achieved for each of the predicted scores:

Figure 22. Calibration Curve for Actual vs. Predicted scores on the

VZ-2 Single Task (VZS) test for Experiment 1. The Y-axis (VZS Actual) gives the mean score for each predicted score.

Also, for the VZD a small positive correlation was found (F(1,58) = 7.51, p<.01, r=.34, r2= .11. ) A calculation of the Goodman-Kruskal gamma correlation (G), gave the following result, G=-.30. The following calibration curve shows the means of the actual scores achieved for each of the predicted values:

Figure 23. Scattergram for Actual vs. Predicted scores on the VZ-2 Dual-Task (VZD) tests for Experiment 1.

Figure 24. Calibration Curve for Actual vs. Predicted scores on the

VZ-2 Dual-Task (VZD) test for Experiment 1. The Y-axis (VZD Actual) gives the mean score for each predicted score.

For the LCS scores the predictions were slightly more accurate (F(1,58) = 10.41, p<.01, r=.39, r2=.15):

Figure 25. Scattergram for Actual vs. Predicted scores on the Listening Comprehension Single-Task (LCS) test for Experiment 1.

A calculation of the Goodman-Kruskal gamma correlation (G), gave the following result, G=-.40. The calibration curve plotted below shows the means of the actual scores achieved for each of the predicted values:

Figure 26. Calibration Curve for Actual vs. Predicted scores for the Listening Comprehension Single-Task (LCS) test for Experiment 1. The Y-axis (LCS Actual) gives the mean score for each predicted score.

But scores were predicted less accurately for LCD (F(1,59) = .04, p>.05, r=.03, r2Å0.)

Finally, subjects were asked to compare each pair of tasks (e.g. VZS vs. VZD, VZS vs. LCS, see Appendix F for the list of all comparisons). In general, certain orderings appeared more frequently than others. The following table shows the most frequent rankings of preferences:

Table 10

This table shows only those rankings which were considered valid. Some pairings, however, did not lead to valid rankings. For instance, a subject might say that: LCD was harder than LCS, and LCS was harder than VZD, but also, VZD is harder than LCD. This led to a circular ranking and occurred for 27 out of the 60 participants. These intransitivities were not included. From the remaining 33 valid rankings, however, the most frequent was quite clearly:

LCD > LCS > VZD > VZS.

The following matrix gives a frequency rating for all of the subjects. This table shows the number (and percentage) of subjects who thought that the test listed in the vertical column was more difficult than the test listed in the horizontal column:

Table 11

From these results it is now possible to address each of the specific hypotheses.

The main hypothesis addresses the way in which high and low SVA individuals react under single- and dual- task conditions. Specifically: (a) High SVA individuals will show less of a decrement in performance from the single-task to the dual-task condition in terms of the visual/spatial task than low SVA individuals and (b) either, there will be no performance differences from the single- to the dual- task condition in terms of the auditory/verbal task or the high SVA individuals will show less of a decrement in performance than the low SVA individuals.

Results from Experiment 1 suggest that it is not the high SVA individuals, but rather the low SVA individuals who show less of a decrement in performance in the visual/spatial task, with many of the low SVA individuals actually showing an improvement from the single-task to the dual-task condition. While this finding did not support the first hypothesis (H1), the second hypothesis (H2) was supported by this experiment. That is, there was no substantial difference in terms of performance decrements between the high and low SVA individuals for the auditory/verbal task. At first glance these findings might suggest that while the low SVA individuals tend to process the two dual tasks with very little interference, the high SVA individuals show much more interference between the two tasks which occurs primarily as a performance deficit in the visual/spatial task.

However, there are problems with this conclusion. First, since the VZ-2 score was used as a measure of Spatial Visualization Ability (SVA) as well as for single- to dual- task comparison purposes, this may have had a confounding effect. Obviously low SVA individuals did worse than the high SVA individuals in the single-task condition and possibly that low level of performance did, as mentioned above, leave much more room for improvement. In addition, there may have been problems with the design of this experiment. These problems will be addressed in the final section of this paper.

In addition to observing the difference scores, it was of interest to look at the relationship, if one existed, between SVA and Listening Comprehension (LC) as given by these particular results. Specifically, this involves the correlation of the VZ-2 scores (VZ-2 single-task only) and the LCS (listening comprehension single task only) scores. The hypothesis originally stated was that there would be either no correlation or a negative correlation between these two measures. The assumption here is that individuals generally appear to be stronger in one area or the other. That is, people often claim to be "more spatially oriented" or "more verbal", rarely do people state that they are equally strong (or weak, for that matter) in both areas. This assumption would suggest a negative correlation between SVA and listening comprehension ability. However, that which people claim as their strengths and that which are their actual abilities do not always agree. Furthermore, people generally make these statements individually, not comparatively. That is, they may feel that they, themselves are stronger in one area than another, but they may not be aware of how they rate against other individuals. It would seem that there would be great variability in this measure. Some individuals may indeed be strong in both, or weak in both, while others may show a great preference for one ability over the other. The assumption here is that there is not enough consistency to lead to a strong positive or negative correlation, which would therefore suggest that there is no correlation between these two abilities. The results from this experiment support the second assumption which suggested no significant correlation between the VZS and the LCS scores.

Next Page Back