At a closer inspection of the findings however, it should be noted that two dimensions of the items had been changed from those of the VVIQ. The item type (real world objects/scenes vs. line drawings of shapes) and the nature of the item (recalled from long term memory vs. perceived and imagined). Thus it is not clear from the original study which of these manipulations was central to the predictive ability of the questionnaire.
The current experiment aimed put these unsettled issues at rest. An extended version of the imagery questionnaire used in the original study (Dean & Morris, 2003) was employed in order to contrast all possible dimensions of the four variables. Additionally the VVIQ (Marks, 1979) was included to test the predictive ability of the new questionnaire with that of the previously failed measures. In line with Dean & Morris (2003) the current experiment aimed to investigate the relationship between self-reports of imagery based on the newly developed questionnaire and performance on five spatial tests. Specifically it is the role of the current paper to report on the correlation between the new ratings across different item types with performance on the Comprehensive Ability Battery Spatial Test (CAB-S; Hakstian & Cattell, 1976, cited in Dean & Morris, 2003).
Method
Design
The current experiment employed a correlation design. The independent variables were the individual items and the ratings for the new questionnaire and the VVIQ. There were five dependant variables reflecting a measure of spatial ability; Comprehensive ability battery spatial test (CABS), the Vandenberg test of mental rotation (VAND), visualisation of rotations (VR), visualisation of views (VV) and space relations (SR).
Participants
An opportunity sample of 226 undergraduate students volunteered to take part in the experiment as part of their course.
Materials
The five spatial tests used in the current experiment were the (CABS) test of mental rotation, the Vandenberg test of mental rotation, the Visualisation of views, the Visualisation of rotation and the Space relations test. The imagery questionnaires selected were the VVIQ (Marks, 1973) and an extended version of the questionnaire used in the study by Dean & Morris (2003). The questionnaire consisted of 14 questions and 8 items. Each item was systematically variated in order to contrast different dimensions of the variables. Items were manipulated by virtue of a change in either the source of the item (perceived or recalled) or the type of image (everyday image or line drawing).
Procedure
The measures were completed by participants in two sessions with a 1 week interval in between sessions. In the first session participants completed the VVIQ and the new imagery questionnaire, in the second session participants completed the five spatial tests. The spatial tests were completed in the following order: CABS, Vandenberg, Visualisation of rotations, visualisation of views and space relations.
Participants were allowed as long as they wanted to complete the VVIQ and the new imagery questionnaire. Time constraints of 2 minutes were placed on the CABS and the Vandenberg tests, and 12 minutes each for the remaining spatial tests.
Results
Correlations of spatial tests and VVIQ
By looking at the correlations of the spatial tests with the scores on the VVIQ
(Table. 1) it is possible to note that the VVIQ does not correlate with any of the spatial tests however; all five spatial tests do correlate significantly with one another.
Table 1. Intercorrelation of scores on the VVIQ and the five Spatial tests (N = 226).
** = P < 0.01 ; * = P < 0.05
Correlations of questionnaire ratings for the 2D shape (retrieved) with the CABS test and VVIQ scores.
By referring to table 2 it is clear that a large proportion of the ratings for the 2D shape (retrieved) correlate significantly with the CABS test and the VVIQ.
For the CABS test the only two ratings that do not correlate significantly are size and vividness. For the VVIQ the ratings that correlate significantly are: form, detail, clarity, maintain, proportion, vividness, rotation, proportion during rotation and vividness during rotation.
Table 2. Correlation of ratings for the 2D shape (retrieved) with the CAB-S test and scores on the VVIQ (N = 226).
** = P < 0.01 ; * = P < 0.05
Intercorrelation of questionnaire ratings for 2D shape retrieved
Table 3. Presents the inter-correlation data for item S3. Inspection of the pattern of correlations between ratings for shape 3 suggests that 12 of the 14 ratings are significant. Due to these strong inter-correlations between the ratings it is necessary to determine whether this is because the ratings on the questionnaire are all measuring one underlying factor or whether there is a unique variance related to a certain group of ratings?
Regression of ratings for the 2D shape against the CABS test
In answer to the previous question, a stepwise multiple regression analysis was performed (see. Appendix. 1). The mean scores on the CABS test were treated as the dependant variable; the mean rating scores for the 2D shape (retrieved) were entered as the independent variables. Two variables were entered into the regression resulting in a multiple R of .28 (P < .001). They were; ease of rotation (ROT) Beta = .234 (P < .001) and change in clarity once rotated (RCLARC) Beta = -.163 (P = .02).
Correlation of questionnaire ratings for ease of rotation across all items and the CABS test.
Table 4. Correlation of Rot variable across all items of the questionnaire and with scores on the CABS test (N = 226).
** = P < 0.01 ; * = P < 0.05
In reference to the correlations between the rating (ROT) for each item type (Table 4) it can be noted that ease of rotation correlates significantly across the different item types apart from item B which was the item retrieved from long-term memory and imagined i.e. Flake.
The correlations of the rotation ratings for each item type with the CABS test
(Table 4) highlight two items that do not correlate with spatial performance; item B as referred to previously and item D which was the line drawing of a stool i.e. real world object also recalled from long term memory.
Discussion
The underlying theory behind the current experiment was that imagery had been documented as having a central role to play in the solution of spatial problems and performance on spatial tests (Dean & Morris, 2003). Yet apart from Dean & Morris (2003) no previous self-report measures had found correlations between the subjective experience of imagery and spatial ability. The question that arose from the results concerned the extent to which aspect of the manipulation of the items was most predictive of spatial performance on the previous questionnaire.
Interpretations of the current findings highlight several issues that need to be addressed. The first step of the analysis was to make sure that spatial ability could be accurately predicted by the use of the current spatial tests. In the original experiment only two spatial tests were used (see Dean & Morris, 2003) whereas the current experiment employed five spatial tests, thus larger variations in performance within the tests could be accounted for by variations between the tests. Table.1. Clearly identifies that all spatial tests seem to be measuring one underlying ability. Thus we can refer to any of the current measures used as accurate predictors of spatial ability.
Table 1. Also provides evidence that the VVIQ, is unable to predict performance on the spatial tests. In line with Dean & Morris (2003) this is largely thought to be due to item type used suggesting that other processes may be involved in the generation of items from long-term memory that are not captured by the VVIQ.
Table 2. Highlights a significant correlation between ratings on the new imagery questionnaire for the 2D shape retrieved, performance on the CAB-S test and with ratings of the VVIQ. The fact that the new questionnaire correlates with both the VVIQ and the spatial tests is sufficient evidence to suggest that it must be capturing more of the imagery processes involved in spatial ability than the VVIQ.
The next step of the analysis was to check the internal reliability of the questionnaire ratings themselves (see Table 3.). The results showed strong intercorrelations between most of the questionnaire ratings. This suggests that the questionnaire is an internally reliable measure of imagery i.e. all the ratings are relevant measures of the imagery system. Because most of the questionnaire ratings all correlate and also correlate with the spatial tests, the question arose as to whether there was unique variance associated with the individual ratings that also has predictive validity against the spatial tests.
Results from the stepwise multiple regression suggest that ease of rotation and clarity once rotated account for a larger portion of the variance than any of the other ratings. Such interpretations also suggests they accounted for a unique variance that has a predictive ability on spatial performance. This suggests that at least one aspect of the imagery system identified by Kosslyn (1980, 1994) seems to be predictive of spatial performance on the CAB-S test. As Dean & Morris (2003) note, however interpretations of results of regression analysis may not always be what they seem.
As stepwise regressions just enter the variables with the highest variation it is possible that they were capturing variance shared with other ratings that were not included into the regression analysis.
The final step in the analysis is to determine what aspect of the change in the items was responsible for the predictive ability of the previous questionnaire. The current findings reveal that ease of rotation of the 2D shape (recalled) seems to capture a unique variance in spatial ability, and is thus independently predictive of performance on the CAB-S test. However this does not necessarily mean that it is predictive of performance for other shapes as Dean & Morris (2003) observed. The correlations between the ratings for ease of rotation across items found that all apart from items B and D correlated significantly with performance on the CAB-S test, which identifies that the ratings are clearly measuring an underlying ability or process. The fact that items B and D do not correlate however supports the notion that different stimuli require the involvement of different processes in the imagery system (Dean & Morris, 2003).
At this point of the discussion it should be noted that both items B and D were objects recalled from long-term memory. This must be compared with similarities of items on the VVIQ, and thus provide further evidence as to why previous measures have found no relationship with spatial ability. However item D was also a line drawing of an object that was perceived and then recalled whereas item B relied entirely on the constructs of long term memory. In line Dean & Morris (2003, p.268) this would suggest that even for perceived stimuli visual processing takes place and is reflected in the reinforcement or maintenance of the image in the visual buffer”
It is thus interesting to examine why these items showed no relationship with spatial performance for ease of rotation. There are several possibilities, the most obvious of answers would be because they are recalled from long-term memory in contrast items S3 and S4 were also retrieved from long-term memory and showed significant correlations. Further Kosslyn (1994, p.7) believes that images we experience are “actively generated from information in the long term store” thus implies that differences in item nature should not effect the surface images we experience. Such interpretations are in conflict to the ones found in the current experiment however.
A second possibility is that item type effects the way in which the processes of the imagery system are reflected in the surface properties of the image experienced.
Such suggestions would be more in line with those of the working memory model (Baddeley, 1990, cited in Dean & Morris, 2003) and further contradict those of Kosslyn (1980, 1994). A major problem relating such findings to models of the imagery system notably Kosslyn’s (1980, 1994) is that because they are structural models they do not allow for the individual prediction of separate processes and how variations or slight manipulations will effect the way these processes work or interact. The working memory model on the other hand suggests “the visuo-spatial sketch pad can be decomposed into separable components with separate rehearsal systems for each” (Logie, 1995, cited in Dean & Morris, 2003).
A third and final possibility could simply be result of information content or information load of the imagined objects. Kosslyn et al., (1999) provide evidence that suggests mental imagery is more difficult when subjects form representations of high resolute images. Their findings of increasing difficulty with high-resolution discrimination implies additional effort to represent or retrieve visual images with high resolution. If so interpretations of the current findings would suggest that participants images of items B and D do not correlate with spatial performance because the resolution of the images cause a degraded representation of the surface properties of the image experienced. As yet however no such explicit models of the imagery system that predict spatial performance exist.
In conclusion, the findings of the current experiment support those found by Dean & Morris (2003). It was found that changing the ratings on the previous questionnaire lead to introspective ratings that predicted spatial performance. The new ratings seem to capture more of the imagery processes than vividness alone. The largest change was due to item type imagined; real world items recalled or constructed from long term memory were unable to predict spatial performance on the CAB-S test. Such findings raise important questions that need to be addressed by current models of the imagery system. It can be suggested that the phenomenological experience of an image is not in its purest form strictly a result of the processes of the visual buffer in operation but a result of other visual and spatial process that have not yet been explained (Dean & Morris, 2003).
References
Dean, G. M., & Morris, P. E. (2003). The relationship between self reports of imagery and spatial ability. British Journal of Psychology, 94, 245-273.
Kosslyn, S. M. (1994). Image and mind. Cambridge, MA: MIT Press.
Kosslyn, S. M., Sukel, E. L., & Bly, B. M. (1999). Squinting with the minds eye: Effects of stimulus resolution on imaginal and perceptual comparisons. Memory & Cognition, 13, 276-287.
Kyllonen, P. C. (1996). Is working memory capacity Spearman’s g? In I. Dennis & P. Tapsfield (Eds), Human abilities. Hillsdale, NJ: Erlbaum.
Nisbitt, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84, 231-259.
Marks, F. D. (1973). Visual imagery in the recall of pictures. British Journal of Psychology, 64, 17-24.
Paivio, A. (1970). On the functional significance of imagery. Psychological Bulletin. 73, 385-392.
Richardson, A. (1977a). The meaning and measurement of memory imagery. British Journal of Psychology, 68, 29-43.
Sheehan, P. W. & Neisser, U. (1969). Some variables affecting the vividness of imagery in recall. British Journal of Psychology, 60, 71-80.
Shephard, R. N., & Metzler, J. (1971). Mental Rotation of three-dimensional objects. Science, 171, 701-703.