Charting the cognitive sphere: tactile-kinesthetic performance within the structure of intelligence.
The majority of studies investigating the structure of human cognitive abilities utilize tasks that are administered to the individual through the visual medium. In the remaining paradigms aurally presented stimuli are employed. As a result of these methodological practices, several sensory modalities remain under-represented and, therefore, poorly understood in individual differences research. This state of affairs would appear to reflect an implicit assumption in many extant models of intelligence. According to these models, the investigation of "higher mental processes" is thought to depend (critically) on complex cognitive mechanisms (compared to which sensory processes are relatively simple). Hence, it is argued that nothing of importance may be gained by employing tests that are based on different modalities. This thesis entails that each sensory modality is an equally good candidate for accessing the "highest level" of thinking which, by extension, is the defining feature of intelligence. However, this theoretical position tends to ignore the opinions of early psychometric researchers who saw intelligence as the "integrative function of the mind" (Burt, 1949) or as the sum-total of all cognitive processes (cf. Thurstone & Thurstone, 1941). This viewpoint also tends to disregard contemporary theories about the structure of human abilities including the "full-blown" theory of fluid and crystallized intelligence (see Horn, 1985), Humphreys' (1979) conceptualization of general cognitive ability, and Carroll's (1993) three-stratum framework. In each of these models, although the term "cognitive abilities" may be preferred over "intelligence," the concepts employed are clearly used in a generic sense to refer to any given cognitive act.
The importance of studying human cognition in its entirety should not be underestimated. In a recent article, Carroll (1995) has attempted to establish the frames of reference for methodology in the study of individual differences. Carroll argues that the goal of our sub-discipline is the exploration of "the diversity of intellect in the people of this planet--the many forms of cognitive processes and operations, mental performances, and creations of knowledge and art" (p. 429). To achieve this end it is clear that there are untapped domains of mental activity that must be discovered and encapsulated within a comprehensive taxonomic model. Furthermore, the demonstration of the meaning of these concepts and their relevance to real-world activities and problems is critical to progress in the science of individual differences.
In extending these arguments to present concerns, there would appear to be a number of complex processes (ranging from psychophysiological to cognitive) that are unique to a given sensory modality. The importance of these processes in, for example, creations of knowledge and art is self-evident. However, as noted above, many sensory phenomena have not been investigated with those tests of intelligence currently available and typically employed. Elsewhere, the development of auditory tests has pointed to the existence of distinct abilities that could not otherwise be assessed. The specification of broad and narrow factors underlying the domain of auditory reception (e.g., tonal memory, speech perception under distraction/distortion; see in particular Carroll, 1993, Chapter 9; Stankov & Horn, 1980) has, in turn, provided for a more complete model of human cognitive abilities. The development of a still more comprehensive model of intelligence demands attention to the investigation of further sensory factors.
Apart from a need to study cognition in its entirety, there would appear to exist both pragmatic and theoretically driven rationales for examining each of the sensory modalities from the perspective of an individual differences framework. For instance, the findings obtained with auditory factors alerted researchers to a need for the incorporation of a temporal dimension into psychometric assessment. In due course this allowed for successive presentation of both visual and auditory items in computer-based testing. Note also that the importance of both sensory and perceptual processes to various individual differences phenomena (apart from intelligence) is now fully recognized (Stankov, Boyle & Cattell, 1995). The realization that such processes are influential in the patterns of age-related changes observed in human cognition has provided a particularly compelling impetus (e.g., Anstey, Stankov & Lord, 1993; Lindenberger & Baltes, 1994). Equally, neuropsychological research indicates that it may be profitable to examine loss of function in modalities other than vision in order to gain an understanding of the effects of brain damage. In neurological testing the damage detected in any of the remaining sensory systems is often viewed as indicative of broader cognitive dysfunction.
The main focus of the present paper is on a variety of measures of tactile (i.e., haptic) and kinesthetic abilities. The aim of this research is to gain an insight into the place of these abilities within the overall organization of human cognition. It is assumed that summaries of factor analytic work distilled within structural theories of intelligence provide a useful framework for the current investigation. To this end, this research employed a battery of tactile and kinesthetic tests adopted and adapted from the experimental and clinical literature.(1) The tests were selected because they proved useful in the past for some practical purpose and/or as a means of elucidating a particular theoretical issue. A battery of visual tests and measures of mental speed (whose factorial structure is known) were also selected. A major issue involves establishing whether tactile tests cluster together to form a separate factor or whether they are defined by existing factors derived from visually presented measures. If a separate tactile factor were to appear its correlation with the well-known factors of fluid and crystallized intelligence would also need to be established. Further details pertaining to this multivariate design are provided in subsequent passages of the present paper.
PSYCHOMETRIC, EXPERIMENTAL, AND CLINICAL STUDIES OF TACTILE AND KINESTHETIC ABILITIES
In exploring a relatively uncharted domain of human cognitive ability it would appear necessary that consideration extend beyond previous research undertaken within the field of individual differences to incorporate experimental studies of perceptual and cognitive processes. Because clinical applications of sensory paradigms are commonplace, this literature should also be consulted. In the passages that follow, empirical studies of tactile-kinesthetic constructs, conducted across diverse fields of psychological inquiry, are surveyed.
In his meta-analysis of major factor analytic studies published this century, Carroll (1993, Chapter 14) points to a noticeable lack of rigorous investigation into a number of areas of human cognitive ability. The role played by tactile sensory input in contributing to individual differences in intelligent behavior is one such area that has yet to be systematically explored. The paucity of research conducted within the tactile-kinesthetic domain is reflected in the fact that there are only four studies (out of some 477 data sets) that Carroll (1993, p. 546ff.) could re-examine in his extensive treatise. Because each of these isolated studies provide information that would seem relevant to the concerns of the present investigation, the results are summarized below:
1. There would appear to be evidence for a tactile localization ability that is assessed by tasks requiring the identification of where upon the skin one has been touched (Adevai, Silverman, & McGough, 1968; Royce, Yeudall, & Bock, 1976). The relationship of this factor to well-known cognitive ability constructs is uncertain.
2. There may exist a tactual performance ability that is tapped by tests such as those found in the tactile section of both the Halstead-Reitan Neuropsychological Test Battery (Jarvis & Barth, 1984; Royce et al. 1976) and the Luria-Nebraska Neuropsychological Battery (Golden, Hammeke & Purisch, 1981). As noted with a tactile localization factor, the relationship of this construct to broad cognitive factors is equivocal. However, from clinical studies to be reported shortly, it would appear evident that this factor shares moderate correlation with measures of fluid intelligence, and low to zero correlation with indices of crystallized intelligence.
3. Tactile ability measures appear to be related to certain visualization tasks (cf. Adevai et al., 1968) and to writing speed and reaction time (Moursy, 1952). Largely because of deficiencies in design features of these isolated studies, it is unclear whether the linkage between visualization and haptic indices occurs at the first-, second- or third-stratum of human cognitive abilities.
4. Carroll's reanalysis of Moursy's (1952) data indicate that writing speed, reaction time, and touch discrimination parameters all share salient loading (of approximately 0.31) on a third-order factor interpreted as general intelligence. This may be taken to suggest that aspects of touch (and/or kinesthesia) are related to constructs of mental speed--an issue worth considering given the extent to which chronometric indices are envisaged to provide a better understanding of intelligence (e.g., Eysenck, 1995; Jensen, 1987).
Gardner's (1983) theory points to the existence of bodily-kinesthetic intelligence. The emphasis in this model is on kinesthetic rather than on tactile processes and on gross and complex motor movements that are characteristic of dance and various sporting activities. These emphases notwithstanding, Gardner's (1983) analyses point to an important area of human endeavor that has largely been neglected in psychometric studies of intelligence.
Experimental Cognitive Studies
Experimental psychologists have provided important information about the nature of tactile and kinesthetic perceptual processes (cf. Loomis & Lederman, 1986). Studies within the area of cognition have argued for the independence of constructs tied to various sensory systems. Within the tactile domain, an important distinction is related to Gibson's (1962, 1968) seminal research on active and passive touch mechanisms, research that has subsequently been regarded as a turning point in the psychology of perception (Harre, 1981).(2) In addition the relationship between different modalities, particularly vision and tactile sensitivity, has attracted considerable attention. For example, Heller (1989) demonstrated that visual imagery and visual experience establish a proper frame of reference that may be helpful in passive tactile retention and pattern recognition (see also Schultz & Petersik, 1994). Kazen-Saad (1986) compared short-term retention of tactile-kinesthetic and verbal information and showed that the latter can be more easily transformed into a visual code. Other researchers (e.g., Kiphart, Hughes, Simmons & Cross, 1992) have provided evidence for a short-term haptic memory function. It has even been suggested that this represents a system that is independent from both iconic and echoic memory, although whether this claim is entirely justified would appear contentious.
Studies of cognitive processes amongst the visually impaired have, as would be expected, placed a heavy reliance on tactile measures. Thus, Duncan, Weidl, Prickett, Vernon & Hollingsworth-Hodges (1989) used tactile and kinesthetic: stimuli to develop a new IQ test for the blind, whilst Ungar, Blades, Spencer & Morsley (1994) studied the use of tactile maps as means of finding directions. However, it would appear that some of this literature is either poorly referenced or simply misunderstood by those working within the domain. This may be demonstrated by recourse to the following account. A test that is commonly known as the Tactile Progressive Matrices Test reportedly represents a standardized instrument that has been used to evaluate the cognitive ability of blind individuals. In attempting to locate this psychometric test, Taylor and Ward (1990) found that it has never actually been employed in either clinical or research practice. (Its use has been restricted solely to doctoral studies). These authors go on to note that there is frequent (mis)citation regarding its common usage, theoretical interpretation, validity and established norms!
Clinical Uses of Tactile-Kinesthetic Tasks
Tactile and kinesthetic tasks have been employed in studies of cerebral lateralization (e.g., Maxwell & Niemann, 1984; Reitan, Wolfson, & Hom, 1992), aging (e.g., Anstey et al., 1993; Verriflo, 1993) and in the assessment of the effects of medical (e.g., Copeland, Dowell Jr., Fletcher, Sullivan, Jaffe, Cangir, Frankel & Judd, 1988) and psychopharmacological intervention (e.g., Hennessy, Kirkby & Montgomery, 1991). Many of these findings have interesting implications for theories of individual differences. For instance, Woodward (1992) reports that age-related decrement in tactile sensitivity cannot be attributed to changes in mechanical properties of the skin but rather to "changes in the nervous system affecting the speed, quantity, or quality of information processing" (p. 63). An even greater importance to tactile and kinesthetic abilities is given in the child developmental literature. Both infant attachment and learning (e.g., Ainsworth, 1969; Harlow, 1974; McCarron & Horn, 1979; Piaget, 1952) have been closely linked to this sensory modality.
Neuropsychological investigators have produced test batteries that provide additional clues as to the nature of tactile ability. One such instrument, the Halstead-Reitan Neuropsychological Test Battery (HRNTB; see e.g., Halstead, 1947; Reitan & Wolfson, 1985) has been subjected to extensive analysis, and normative tables (e.g., Thompson & Heaton, 1991), correlations with the Wechsler scales (Wechsler, 1981) and factor analytic studies are available in the clinical literature (e.g., Moehle, Rasmussen & Fitzhugh-Bell, 1990; Yeudall, Reddon, Gill & Stefanyk, 1987). The results presented in this research program are highly suggestive of a relationship between tactile-kinesthetic sensitivity and conventional psychometric indices. For instance, the Tactual Performance Test from the HRNTB (which is used in the current study) correlates around 0.26 with WAIS Performance IQ, with these correlations being slightly higher (in the mid 0.30s) for a male sample investigated by Yeudall et al. (1987). In addition, Boyd and Hooper (1993) report similar results (i.e., moderate correlation with Performance IQ) when examining the rather different tactile tests of the Luria-Nebraska Neuropsychological Battery (LNNB). The magnitude of these correlations suggests that it is likely that tactile-kinesthetic abilities are related to fluid (or perhaps broad visualization) capabilities rather than the broad crystallized intelligence factor.
The Speed-Level Dichotomy: Implications for Experimental Design
Carroll (1993) has suggested that: "If any broad taxonomic classification of cognitive ability factors were to be formulated, in fact. it might be one based on the distinction between level and speed" (p. 644).(3) In recent studies such distinctions appear to have been supported empirically (e.g., Roberts & Stankov, 1997). Consequently, in attempting to systematically examine new constructs of human cognitive ability, it would seem judicious to include measures both of "level" (i.e., accuracy) and "speed." Accordingly, there are three sub-divisions outlining the concepts investigated in the present multivariate design: cognitive ability (level) measures, mental speed constructs and the subset of experimental paradigms assessing various tactile-kinesthetic processes.
Cognitive Ability Tests: Measures of Level. The aim of this aspect of the investigation was to define from tests of primary mental ability second-order factors corresponding to those traditionally identified by [G.sub.f]/[G.sub.c] theory. In all there were twelve tests: seven administered under a computerized format, five given as paper and pencil tests. (All variables used in this study are described in the Method section. The number given to each measure corresponds to that used throughout the various Tables reported in this paper.) There were three marker tests of fluid intelligence ([G.sub.f], Variables 1-3). three marker tests of crystallized intelligence ([G.sub.c], Variables 4-6), three marker tests of short-term acquisition and retrieval (SAP, Variables 7-9), and three marker tests of broad visualization ([G.sub.v], Variables 10- 12).(4) While justification for including the first two factors is self-evident, some comment is in order concerning the decision to include SAR and [G.sub.v]. The former was included because of the importance attached to memory constructs in many studies of tactile-kinesthetic abilities (e.g., Kiphart et al., 1992), while the latter was included because of the postulated complex interactions between the visual and haptic mediums (e.g., Kazen-Saad, 1986).
Mental Speed Measures. Because recent research has indicated that mental speed measures may constitute a separate taxonomy from level abilities and some studies have used speed measures to define tactile-kinesthetic constructs (e.g., Moursy, 1952), fourteen mental speed measures were included in the present investigation. Furthermore, since most mental speed measures (Movement Time in particular) involve tactile and kinesthetic processes (e.g., Fagot, Lacreuse & Vauclair, 1994; Shaffer, 1991) it would seem important to establish the amount of communality shared between the tactile and speed domains. Indeed, the possibility should not be ruled out that the correlation between mental speed and intelligence (which has recently preoccupied so much of the individual differences literature [see Stankov & Roberts, 1997]) is a function of shared variance between say [G.sub.f] and tactile-kinesthesis and reaction time and tactile-kinesthesis. The measures employed (with a brief discussion of the conceptual status of each mental speed factor) is given below:
1. Test-Taking Speed. Computerized testing allowed speed of test-taking to be collected from each psychometric task administered in this fashion (i.e., Variables 14-20). There would now appear sufficient evidence to indicate that this ability, which involves the speed at which individuals solve items of non-trivial difficulty, has a well-replicated factor structure (e.g., Carroll, 1993; Horn & Hofer, 1992; Stankov, Roberts & Spilsbury, 1994).
2. Clerical-Perceptual Speed ([G.sub.s]). This factor represents individual differences in the speed with which visual stimuli (of relatively trivial item difficulty) are compared or otherwise codified. Two measures of [G.sub.s] were employed in the present design: the number of items responded to in Digit Symbol (Variable 22), and mean speed of response to items of Number Comparison (Variable 21). Note that in the past, this ability was assessed by paper and pencil tests given within strict time limits. More recently this ability has been assessed by computerized testing, with the ensuing factor well defined by speed of response (Roberts & Stankov, 1994; Stankov, 1988).
3. Choice Reaction Time (CRT). This construct actually represents a composite of two mental speed factors that have been isolated in the literature: a) Decision Time (DT--the time required to determine and initiate an appropriate response to a stimulus [or stimuli]) and b) Movement Time (MT--the time associated with sensory and motor control of movement) (see Carroll, 1993, p. 478; Roberts, 1997a). The degree of choice is manipulated in these tasks by varying the number of response alternatives. In general, the number of choice alternatives ranges from I through to 8, and are re-scaled into binary digits (i.e., bits). In the present study three tests assessing CRT were employed (Variables 23-25).
4. Movement Time. There were two pure measures of MT: Variables 26 and 27. Interestingly, some recent accounts emphasize the importance of this construct to human intelligence (e.g., Buckhalt, Reeve & Dornier. 1990). However, an alternative perspective, views this as a factor that though well replicated, is somewhat peripheral to human cognitive abilities per se (e.g., Carroll, 1993, Chapter 13; Roberts, 1997b).
Experimental Tasks: Assessment of Tactile and Kinesthetic Performance. In addition to these psychometric and mental speed tests, participants completed seven tasks assessing various aspects of tactile-kinesthesis. Details concerning the relationship that these tasks share with either level or speed measures are sparse. Thus, predictions concerning structural relations must be guarded. Two tasks have analogues in the traditional psychometric domain. Because of this feature it might be expected that Finger Counting (Variable 28), like Letter Counting (Variable 3), will load on [G.sub.f] and that Tactile Bead Memory (Variable 30), like Visual Bead Memory (Variable 9). will load on SAR. Two tests have been taken directly from the HRNTB. One of these tests (Tactual Performance) produces both level and speed scores (Variables 31-32). The second test, Finger Writing (Variable 34), gives only a level score. As mentioned previously, the available evidence indicates that scores from the HRNTB are con-elated with measures of general intelligence. Little is known about the likely correlates of the remaining two tasks. Intuitively. both appear less cognitively demanding than any of the variables previously discussed. Because Tactile Texture (Variable 29) clearly involves discrimination at a more sensory level it might plausibly share low correlations with most other variables of the present battery. Gibson's Active/Passive Touch Test (Variable 35), on the other hand, is perhaps more perceptual than sensory. As a consequence, it should have higher correlations with the other variables relative to Tactile Texture (but not as high a magnitude of correlation as these other variables will share among themselves).
Overall, the present battery of seven tactile-kinesthetic tasks (yielding as it does eight measures) is somewhat biased in favor of relatively complex processes. As a result, it is predicted that these tasks will be related to non-verbal measures of intelligence--[G.sub.f], SAR, and perhaps certain aspects of the previously elucidated mental speed constructs. The main issues considered are: a) whether these tasks will define a factor that is separate from all other factors involving level and speed measures and b) what relationships exist between this empirical construct (assuming it is found) and other factors of intelligence assessed in the study.
The participants in this study were 87 First Year Psychology students from the University of Sydney and 108 participants recruited via newspaper and billboard advertisements. The former group completed the experiment as part of course requirements while the later were paid $30 for their participation. The last two tests of this battery (34. Finger Writing and 35. Gibson's Active/Passive Touch Test) were given only to the people recruited from outside the University. Most statistical analyses of this paper will exclude these two tests. However, where separate analyses involving these two tests are reported, only the reduced sample was used. Details of the total sample are presented in Table 1. This Table also includes: a) intercorrelations between age, level of education, gender, handedness and first language and b) the coding of extension variables for each respective group.
Table 1. Statistical Breakdown of Sample by Demographic Variables and the Correlations That These Extension Variables Share Among Each Other. Descriptive (or Frequency) Data Reported Involves the Full Sample, The University Sample and Outside Group.
Total University Outside Extension Sample Sample Sample Variable (n = 195) (n = 87) (n = 108) Age Mean 23.59 20.55 26.04 S.D. 6.66 3.90 7.39 Years of Education Mean 14.40 13.84 14.81 S.D. 2.17 1.68 2.42 Gender Females (1) 96 44 52 Males (0) 99 43 56 Handedness Right-Handed (1) 174 73 101 Left-Handed (0) 21 14 7 First Language English (1) 174 74 100 Other (0) 21 13 8 Extension Correlation Correlation Variable with with Age Education Age Mean S.D. Years of Education .331 - Mean S.D. Gender .121 .149 Females (1) Males (0) Handedness .121 .061 Right-Handed (1) Left-Handed (0) First Language .043 -.008 English (1) Other (0) Extension Correlation Correlation Variable with with Gend r Hand Age Mean S.D. Years of Education Mean S.D. Gender - Females (1) Males (0) Handedness .011 - Right-Handed (1) Left-Handed (0) First Language .210 -.014 English (1) Other (0)
It should be noted from this Table that the mean age of the entire sample was higher than is commonly reported for University populations. This is due largely to the presence of the second group recruited from the general population, which, on average, is five and a half years older. Nevertheless, this second group was well educated (the majority held University degrees). This is reflected in relative correspondence between means of the two groups on the level of education variable. Hence, interpretative problems that might have been encountered in using widely diverse samples were not expected to eventuate in the present investigation. It should also be noted from Table 1 that there were an almost equal number of males and females. This is desirable from the perspective of examining possible gender differences in tactile performance. In elaboration, recall that the correlation between tactile measures and Performance IQ is slightly higher for males (Yeudall et al., 1987). As there is equal representation on the gender variable, the present study should provide valuable data on the relationship between tactile-kinesthetic performance and gender. Table 1 also shows the proportion of left-handed participants in the sample. In investigating tactile-kinesthetic measures, past researchers (particularly in the clinical domain) have pointed to the influence of cerebral lateralization in, for example, finger writing tasks (e.g., Maxwell & Niemenn, 1984). The percentage of left-handed individuals found in the current study matches that observed in the general population.
Psychometric, Mental Speed and Tactile-Kinesthetic Tasks
Measures of Level (i.e., number correct scores). The tests measuring these constructs were those used routinely in studies of intelligence carried out at the University of Sydney and elsewhere. Indeed, within the context of a higher-stratum design these tests are often selected because of the clarity they bring to the structural model provided by [G.sub.f]/[G.sub.c] theory (see Anstey et al., 1993; Horn, 1988; Roberts, Pallier & Stankov, 1996). These psychometric tests (of level) were:
* [G.sub.]: 1. Raven's Standard Progressive Matrices Test; 2. Letter Series Test; 3. Letter Counting.
* [G.sub.c]: 4. Synonyms Vocabulary Test; 5. General Knowledge Test; 6. Esoteric Analogies Test.
* SAR: 7. Digit Span Forward; 8. Digit Span Backward; 9. Visual Bead Memory. This last test was taken directly from the Stanford-Binet Intelligence Scale: Fourth Edition (Thorndike, Hagen & Sattler, 1985).
* [G.sub.v]: 10. Card Rotations Test; 11. Hidden Figures Test; 12. Hidden Words Test.
* Carefulness: 13. Number Comparisons Test-number correct score. Under paper and pencil conditions this measure would serve as a marker of Perceptual Speed. However, when administered within a computerized format this measure has been found to define a Carefulness factor (French, 1957; Roberts & Stankov, 1997). Even so, since this is the only measure of this construct and it is more highly correlated with [G.sub.s] than any other broad cognitive factor, Variable 13 is likely to share salient loadings on [G.sub.s]. The more commonly used measure in this test is, of course, speed (see Variable 21).
Mental Speed Measures. The constructs, along with tests defining various types, of mental speed examined in this study are fisted below. Tasks cited relatively infrequently in the individual differences literature are described in detail.
Measures of Test-Taking Speed (TTSp). The following seven tests from the above list of thirteen were computerized which, in turn, allowed for separate time scores (measured in seconds) to be obtained: 14. Letter Series Time; 15. Letter Counting Time; 16. General Knowledge Time; 17. Esoteric Analogies Time; 18. Digit Span Forward Time; 19. Digit Span Backward Time; 20. Hidden Words Time.
Clerical-Perceptual Speed ([G.sub.s]). To demarcate this factor, two tests that have been used extensively in studies examining [G.sub.s] were employed: 21. Number Comparison Time; 22. Digit Symbol. (Note that Test 22 comes from the Performance IQ scale of the WAIS-R [Wechsler, 1981]).
Choice Reaction Time (CRT). To assess CRT, three tasks were used:
23. Math-Classification Test. Participants were required to give the arithmetic operator that would solve a simple numeric problem (e.g., 2 ? 3 = 5, Answer "+"). To maintain consistency in the presentation of the mean results obtained with mental speed measures (and allow some comparisons to be made across tasks), the number of such problems solved within 60 sec was obtained and transformed into the average speed of solving each item (in seconds).
24. Stimulus-Response Compatibility Test. This involved a series of three tasks in which participants were required to match letters or numbers to the direction of arrows. In the most compatible condition the participant was required to write "U" (for up) to an arrow pointing at 12 o'clock and similarly "D" (down), "R" (right) or "L" (left) for arrows pointing in these directions. In the least compatible condition, participants were required to write "Z," "P," "T" or "M" for the same stimuli. The third task required the substitution of "1," "2," "3" or "4" (which was presented in a clockwise orientation) for the direction of the arrow. The number of such problems solved within 60 sec was obtained for each task, then averaged across the three conditions and subsequently transformed into the mean speed of solving each item (in seconds).
25. Card-Sorting Test. This task, modeled after Crossman (1953; see also Roberts, Beh & Stankov, 1988), utilized the informational properties of a simple deck of playing cards. In it participants were required to sort the deck of playing cards into suits (2 bits). The number of cards correctly sorted within 60 sec was obtained and then transformed into the average speed of sorting each card (in seconds).
Movement Time (MT). Two tests were used for the measurement of MT:
26. Fitts' Movement Task. Participants had to tap a small metal-tipped stylus between two targets as quickly and as accurately as possible. Task difficulty was increased by requiring participants to complete this task five times with successively smaller target areas. The number of cycles made in 60 sec was recorded, averaged across the five conditions and then transformed into the average movement speed across the levels of task difficulty (in seconds).
27. Card-Dealing Test. Participants dealt decks of playing cards face up into two piles (and within designated target areas) as quickly and accurately as possible. The number of cards correctly sorted within 60 sec was obtained and then transformed into the average movement speed per card (in seconds).
Tactile and Kinesthetic Tasks. In each of these tests the dependent variable was number correct although for one of these tasks (Variable 31) solution time (measured in seconds) under three different experimental manipulations of handedness was also recorded. During the performance of all tactile-kinesthetic tasks, participants were either blindfolded or had their hand placed inside a box to obscure objects from direct visual inspection.
28. Finger Counting Test This task, first reported by Kainthola and Singh (1992), bears a number of similarities to Letter Counting (Variable 3). In Test 28, fingers of the participant's dominant hand were tapped one or more times with a small rubber stylus. Item administration followed a pre-arranged series of sequences, with a one second delay between presentation of each element of the sequence. Participants were required to recall bow many times each finger was tapped. The test consisted of 24 such items. In an attempt to manipulate the difficulty level of this task, the first twelve items involved three fingers (little finger omitted), while the remaining items involved all four fingers.
29. Tactile Texture Test. This task is an adaptation of a task used by Hollins, Faldowski, Rao and Young (1993; see also Kudoh, 1988). The stimuli were various gradations of sandpaper. Blindfolded participants were presented with a board containing a test item and five samples, one of which matched the test item. The participants dominant hand was placed on the test sandpaper while the other was left free to inspect each of the choice alternatives. Participants were required to determine (within 30 sec) the alternative matching the test sandpaper in perceived texture (with the response simply the participant pointing to their solution). The test contained 24 such items.
30. Tactile Bead Memory Test. This task involved a modification of Test 9--the (Visual) Bead Memory Test of the Stanford-Binet Intelligence Scale: Fourth Edition (Thorndike et al., 1986). Blindfolded participants were given 20 seconds to feel (with both hands) predetermined configurations of the various beads comprising this sub-test. Item difficulty was manipulated by increasing the number of beads to be correctly recalled in each trial. The participants task was to reconstruct the configuration haptically (i.e, whilst still blindfolded) after the 20 seconds had lapsed. Testing was discontinued after participants made two consecutive incorrect responses at a given item level.
31. Halstead-Reitan Tactual Performance (Time) Test This test was taken directly from the HRNTB and administered in the usual fashion (see Reitan & Wolfson, 1985). An index of tactile form discrimination, this task makes use of a modified Seguin-Goddard form board. Blindfolded participants were required to place ten geometrically shaped wooden blocks into this form board on three separate occasions: firstly with their preferred hand, then their non-preferred hand and finally with both hands. The time spent on each occasion was recorded.
32. Halstead-Reitan Tactual Performance (Level) Test. Upon completion of the third trial of Test 3 1, the form board was hidden from view. At this point the participant was permitted to remove their blindfold, and then asked to draw each shape in its correct location. Participants were not informed of this requirement until after the tactile part of the test had been completed. (The number correct score may therefore be regarded as a measure of incidental learning). Two performance measures were obtained: the number of shapes correctly recalled (referred to in the HRNTB as the "Memory" score) and the number of shapes recalled in, their correct position (i.e., "Location" score). For present purposes a composite of the "Memory" and "Location" scores was employed.
33. Tactile Shapes Test. In this task, adapted from Kazen-Saad (1986), blindfolded participants were presented with cut-out patterns of vertical and horizontal line segments varying in length. The participants were required to move the index of their preferred hand twice along these gratings (which took, on average, 20 seconds). Participants were then instructed to decide (upon visual inspection of analogues) from amongst five printed patterns the correct choice alternative. Item complexity was manipulated by varying line length and the number of horizontal and vertical line segments (24 items).
34. Finger Writing Test The HRNTB contains a number-writing task in which the digits "3," "4," "5" and "6" are written once on each finger of both of the participants hands in a predetermined random order (Reitan & Wolfson, 1985). A number of papers examining the microstructure of this task report left-hand advantages in participants who are normally right-handed (e.g., Maxwell, Wise, Pepping, Townes, Peel & Preston, 1984). In addition, researchers employing this task report moderate correlation with the Performance (but not Verbal) scale of the WAIS-R (Maxwell & Niemann, 1984). Envisaging that the number writing task may be too simple for normal participants, Test 34 employed letters as stimuli. Invoking the principles of information theory, the current authors note that this should extend the stimulus uncertainty from 2 bits in the normal number writing test to 5 bits for Test 34 (see Lehrl & Fischer, 1988; Roberts et al., 1996 for the underlying rationale relevant to this modification). Unbeknownst to the participants, only capital letters that could be written in one stroke (e.g., "Z") were employed as stimuli. The actual task involved participants placing both hands into a cut-out cardboard box, wherein the experimenter would write a letter on the index finger of the left-hand and move on to the next finger of that hand with a new stimulus. Having finished the left-hand, the experimenter proceeded to the right-hand tittle finger then through to the index finger then back again before proceeding to the left-hand little finger and so forth. The participant's task was simply to identify (vocally) the letter drawn on respective fingertips. No letter was repeated more than twice (24 items).
35. Gibson's Active/Passive Touch Test. In this task, participants were again instructed to place their hands inside a card-board box, which for this task had numbered figural representations of the stimuli to be presented (5 shapes) attached at its apex. There were two experimental conditions. In the first (passive condition), one of five cookie cutters was pressed into the palm of the participant's preferred hand (see Gibson, 1962). In the second (active condition) the shape was pressed and rotated 180 degrees and back again. In both conditions the participant was required to give the number corresponding to the shape that was perceived to have been presented (12 items per condition).
Total testing time for the whole study was about four hours although, since the computer administered tests were all self-paced, the time taken varied from participant to participant. Testing was generally broken up into two sessions lasting two hours. Rest pauses of ten minutes were given to all participants at the end of one hour of work.
In the first test session, single participants were allotted a trained research assistant who administered all "face-to face" tests (i.e., each of the tactile-kinesthetic paradigms, MT measures and Card-Sorting). In the second session. participants completed the paper and pencil and computerized tasks. This later format allowed group testing with, on average, 4 participants per group. Macintosh[TM] computers, available from the Psychology Department of the University of Sydney, were used during this test session. Because some of the participants may have had little prior experience with these computers. time was also provided during this phase for familiarization with the equipment. Data were subsequently either stored or entered onto disk for later statistical analyses. Two statistical packages were used throughout the analyses reported in this paper: SPSS (Norusis, 1990) and EQS (Bentler & Wu, 1995).
The means and standard deviations calculated for each test are presented in Table 2. Inspection of Table 2 shows that most tests were subject neither to floor nor ceiling effects--rendering interpretation of these measures more easy than would be expected had this not been the case. Furthermore, as expected, Digit Span Backward is more difficult than Digit Span Forward (compare means for Variables 7 and 8, and means for Variables 18 and 19). In terms of providing an understanding of the microstructure of the tactile-kinesthetic tasks, it should be noted that Tactile Bead Memory was more difficult than its Visual counterpart (compare Variables 9 and 30). However, there would appear to be little difference between Letter Counting and its tactile analogue when the number of items comprising these tasks is taken into consideration (compare Variables 3 and 28).
[TABULAR DATA 2 NOT REPRODUCIBLE IN ASCII]
Each of the tactile-kinesthetic measures was analyzed from a variety of experimental perspectives. These findings are reported in a subsequent paper that focuses both on task microstructure and additional personality correlates obtained in this study (Stankov, Pallier & Roberts, in preparation). However, some indication of certain features of these paradigms is in order. Relevant findings include each of the following:
1. The reliabilities of each of the tactile-kinesthetic tests were not only found to be acceptable (i.e., Cronbach-alpha's exceeding 0.75) but within the range provided by standardized neurological tests of tactile function (e.g., Golden et al., 1981).
2. The complexity levels of several of the tasks in which difficulty was manipulated (i.e., Tests 28, 29, 33) exhibit simplex pattern (see Guttman, 1955). This finding is encouraging from the perspective of establishing the validity of tactile-kinesthetic measures used throughout this study (cf. Roberts et al., 1988).
3. If one partitions the two different samples of the study the means for all variables are relatively close (see final column of Table 2). The only measures that show statistically significant differences are Variables 4, 13, 21, 26 and 33. Indeed, close equivalence in mean performance is also observed in a third sample investigating microstructural aspects of these tactile-kinesthetic measures (Pettersen, 1996). Moreover, exploratory factor analysis of common tests used in these two studies bear close correspondence, testifying to factor replicabitity.
4. Performance in many of the tactile-kinesthetic tasks of the current investigation that were established tests reflected very similar findings to those reported in earlier studies. By way of illustration, Thompson and Heaton (1991) conducted a large scale study (n = 489) of the HR Tactual Performance Test. The means reported for the Memory and Location scores were 7.59 (S.D. = 1.58) and 4.43 (S.D. = 2.45) respectively. These statistics are highly similar to the ones presently obtained (Mean = 7.34, S.D. = 1.51 for Memory score; Mean = 5.75, S.D. = 2.25 for Location score). Perhaps more strikingly, Thompson and Heaton (199 1) note that some 19.6% of their participants obtain what they refer to as reversals in their HR Tactual Performance Time scores--longer latencies on non-dominant hand trials relative to the (initial) preferred hand trial. In the present study, 38 participants (19.5%) revealed a similar pattern of performance.
Correlations Between Variables
Pearson correlations between all level, speed, experimental (i.e., tactile-kinesthetic) and extension variables are presented in Table 3. In order to interpret these correlations more easily, all speed variables have been reflected. These results are critical to two major concerns: a) preliminary evidence for structural relationships between the established and experimental measures and b) determination of the degree of association among ability measures and variables such as age, sex and handedness. With respect to the former, this is important from the perspective of establishing the construct validity of the measures that were employed in this study. For example, there is a body of evidence that suggests the type of relationships that [G.sub.f] and [G.sub.c] abilities should share with chronological age (e.g., Stankov et al., 1995). Demonstration of consistency with the published literature renders findings from the present study less problematic.
[TABULAR DATA 3 NOT REPRODUCIBLE IN ASCII]
Correlations Between Psychometric Variables. Inspection of Table 3 indicates the existence of positive manifold. This is worth noting in the context of the speed-level dichotomy. Note, therefore, not only that the correlations within level and speed domains are positive, but that this also occurs between level and speed measures. An interesting exception is that of Number Comparison (Variable 13) which shares consistently negative correlations with all Test-Taking Speed Measures (Variables 14-20). This may be taken to reflect the fact that rapid response in tests of non-trivial difficulty reflects a slight predisposition towards carelessness.
The correlations obtained between each of the elementary cognitive tasks (i.e., Variables 23-27) and traditional psychometric indices also require a brief comment. Elsewhere, Roberts and Stankov (1997) have demonstrated that various RT parameters (but not MT measures) extracted from these tasks share differential relationships with broad cognitive ability constructs. In particular, RT shares substantial to moderate correlation with [G.sub.f] markers, low to moderate correlations with [G.sub.v] and SAR marker tests and near zero correlation with [G.sub.c] abilities. These findings would appear to be replicated in the present data set.
However, it is with respect to the tactile-kinesthetic measures that this investigation is devoted. Inspection of the magnitude of correlations that Variables 28-35 share in Table 3 is most informative. There are a number of relevant outcomes requiring comment. These findings include each of the following:
1. The average raw correlation between tactile-kinesthetic measures is quite substantial (r = 0.239). However, this value is reduced somewhat by the low correlation which Tactile Texture shares with all other haptic indices. Thus, when Variable 29 is removed from the intercorrelation matrix, the average correlation between tactile measures is notably higher (r = 0.276)
2. In general, the tactile-kinesthetic measures share high correlation with [G.sub.f] measures. For example, Tactile Shapes correlates quite substantially with the following tests: Ravens Progressive Matrices (r = 0.472), Letter Series (Level) (r = 0.388), and Letter Counting (Level) (r = 0.446). Tactile-kinesthetic measures also share consistently high correlations with [G.sub.v] marker tests and mental speed measures (especially those speed measures that include a DT component). These same measures do not appear to share correlation with [G.sub.c], Psychometric Test-Taking Speed or MT variables.
3. The correlations observed between tactile-kinesthetic constructs and SAR marker tests require special mention. The correlation obtained between the Backward and Forward versions of the Digit Span Tests is modest (r = 0.240). In turn, correlations between each of the experimental measures and Digit Span Forward are low. Correlations between the experimental measures and Visual Bead Memory and Digit Span Backwards Tests are, on the other hand, substantial. Indeed, for every tactile-kinesthetic measure these two memory tasks share higher correlation than that found for Digit Span Forward. These correlation coefficients preface the findings with CFA and EFA--apparently performance in tactile-kinesthetic tests shares more in common with working memory than with short-term memory.
Correlations Between Psychometric and Extension Variables. With respect to the age variable and the correlation that it shares with the psychometric tests, most generally, the observed outcomes are as predicted from the available literature. For example, tests that serve as markers for [G.sub.c] (especially Esoteric Analogies and Vocabulary) share substantial positive correlation with chronological age. In a similar vein, [G.sub.c] measures also share substantial correlation with level of education. While correlations between [G.sub.f] marker tests and age are low, it is worth noting that the direction of this relationship is as predicted from previous findings, in that all are negative in sign (see Stankov, 1988). Indeed, with respect to additional factor marker tests, it is generally the case that poorer performance is associated with advanced age. Elsewhere, Horn and Hofer (1992) posit an important theoretical distinction between "vulnerable" and "maintained" abilities. Vulnerable abilities are susceptible to the influence of aging and neurological dysfunction while maintained abilities (such as [G.sub.c] are not. While the present sample is very much restricted with respect to the age variable, the direction of correlation exhibited by measures of tactile-kinesthesis and age indicates that this construct may constitute a vulnerable ability.
On the basis of the available literature it would also appear necessary to consider the influences of gender on the tactile tests and some of the cognitive ability and mental speed measures. For example, there is a literature that suggests female advantages in [G.sub.c] abilities and male advantages in [G.sub.v] abilities (see Brody, 1992 for a review). While the former outcome is currently observed, there would appear a slight female advantage in the present sample also on some [G.sub.v] measures. Possibly because of this feature of the sample, a potential male advantage in the tactile-kinesthetic tasks was not observed.
Finally, some brief comment on the correlations that the psychometric tests share with language and handedness is in order. As expected, the first language spoken at home is quite highly correlated with [G.sub.c] Conversely, handedness shares nonsignificant correlations with all psychometric tests of this study. In as much as the lowest correlations (of approximately zero-order magnitude) are shared between tactile-kinesthetic variables and first language, the measures assessing these constructs might be construed to be rather minimally influenced by direct cultural influences.
Charting the Cognitive Sphere: Introduction to Factor Analyses
The design of studies that aim to chart the structure of human abilities has, of necessity, to be both exploratory and confirmatory. The exploratory aspect of the current study is with respect to new variables (i.e., tactile and kinesthetic measures) that are of particular scientific interest. Although some clues as to possible relationships between variables are available to the researcher, the evidence is presently too imprecise to allow strong, definitive statements. The confirmatory aspect involves establishing what is already known (i.e., the structure of the level and, to some extent, speed abilities of the present battery). Even within these two domains a "perfect" replication of previous results is not anticipated. While this may be due partly to peculiarities in the sample of participants employed in a single study, it is even more a function of the limited number of variables that may (practically) be chosen to represent a factor adequately for any new domain under investigation. Thus, even though two or three variables are selected to demarcate a given hypothetical construct, it is expected that some factors may not appear in a particular study. For example, measures of [G.sub.f] sometimes do not separate from measures of SAR (or [G.sub.v]), and measures of broad verbal fluency may remain attached to [G.sub.c]. It is therefore expedient to combine the exploratory and confirmatory approaches to arrive at a meaningful solution for any individual differences construct subject to empirical investigation (cf. Carroll, 1995).
In this paper exploratory factor analysis (EFA) served three purposes. First, the maximum likelihood analysis of the 33 x 33 correlational matrix of Table 3 was conducted in order to gain some impression of the appropriate number of factors. This analysis also provided the preliminary factor pattern for subsequent confirmatory factor analyses (CFA). Finally, a smaller sample of 23 variables was analyzed. This EFA was implemented in order to resolve certain issues that remained intractable from the preceding confirmatory analysis.
Confirmatory Factor Analyses
Procedures used in Structural Equation Modeling. Exploratory factor analyses were carried out with the 33 by 33 correlational matrix using maximum likelihood procedures outlined above. While root-one criteria and Cattell's (1966) scree test indicated that up to ten factors could be extracted, rotated factors for the ten, nine and eight factor solutions resulted in either singlets or doublets. A subsequent solution with seven factors was better interpretable. This solution also gave an acceptable Chi-square value of 377.41 with d.f. = 318 and p = 0.0122. Considerations related to the initial design features of this study and the pattern suggested by the seven-factor EFA solution were used to arrive at a factor pattern matrix that was entered into the EQS for Macintosh program (Bender & Wu, 1995). This package was used to conduct CFA. Note, as suggested by Joreskog and Sorbom (1989), all CFAs reported in this study were based on the covariance (rather than correlation) matrix.
A two step procedure was adopted. The first model to be fitted was the first-order oblique seven-factor solution. The best fitting version of that model produced a Chi-square value of 587.384 with d.f. = 460 and p [is less than] 0.001. The comparative fit index was satisfactory: CFI = 0.929. In addition, several nonsignificant factor intercorrelations (i.e., coefficients equal to zero in Table 5) were fixed at this value. Table 4 contains the standardized factor loadings of this solution. All non-standardized coefficients corresponding to those presented in Table 4 have test statistics that are significant at the 0.05 level.
[TABULAR DATA 4 NOT REPRODUCIBLE IN ASCII]
In the next step, the second-order structure was explored. The coefficients (involving a covariance metric rather than the standardized solution of Table 4) were all fixed, and two second-order models were fitted. The first model assumed the existence of a single second-order factor and it produced a Chi-square value of 625.152 with d.f. = 518 and p [is less than] 0.001. The comparative fit index, CFI = 0.941, was better than the fist-order solution but somewhat inferior to a second model that was attempted. This last model contained three correlated second-order factors. It produced a Chi-square value of 606.913 with d.f. = 514 and p = 0.003 and an acceptable comparative fit index: CFI = 0.949. These two (standardized) second-order solutions are reproduced in Figure 1. In both diagrams of this Figure, the lowest-order (i.e., the tests themselves) are not presented since the tests' loadings are already given in Table 4. Interpretation of the CFA results presented in the sections that follow is based on the data reproduced in Tables 4 and 5 and also in Figure 1.
[Figure 1 ILLUSTRATION OMITTED]
Table 5. Intercorrelation Between the Factors of Table 4 Factors Gc Gf TK/Gv TTSp ECSp DsF DsB Gc - Gf - TK/Gv .807 - TTSp .205 .394 - ECSp .649 .665 .248 DsF .354 .223 - - DsB .171 .212 .321 .434 -
Note: All elements of the Factor intercorrelation matrix that were blank were fixed at zero in the model.
Because of their relevance to certain theoretical issues, both second-order solutions are considered in this paper. Although these two solutions point to slightly different aspects of the structure of cognitive abilities, the substantive aspects of these models are in general agreement with current conceptualizations of human intelligence. Furthermore, the difference in statistical indices of fit between the two solutions is relatively small.
Interpretation of Factors at the First-Order
The data presented in Table 4 contain seven first-order factors. At the top of Table 4, these factors are further classified with respect to a second-order solution presented in Figure 1a.
Factor 1: Crystallized Intelligence ([G.sub.c]). This factor is well-defined by the three postulated markers of [G.sub.c] (Vocabulary, Esoteric Analogies [Level] and General Knowledge [both Level and Speed]). nose measures that have low loadings (i.e., Letter Series, Hidden Words, Digit Symbol and Number Comparison) do not detract from the overall interpretation of this factor, more especially because each task involves the manipulation of either verbal or numerical material. In as much as an element of acculturation is critical to successful performance in tasks having such content, each additional factor loading is interpretable.
Factor 2: Fluid Intelligence ([G.sub.f]). This factor is reasonably well-defined in the present data set. Thus, two markers of [G.sub.f] (Raven's Standard Progressive Matrices and Letter Series [Level]) have the highest loading on this factor. In addition, the low salient loading of Esoteric Analogies (Level) on this factor is not surprising. In previous work with undergraduate students at the University of Sydney, this test has been found to be factorially complex (e.g., Roberts, 1997b). The only well-known marker test of [G.sub.f] that failed to exhibit loading on this factor was Letter Counting--an explanation for which is taken up shortly. Notwithstanding, Tactile Texture has very low communality in this battery. Thus, whatever it shares with other variables of the battery is captured by salient loading on Factor 2. Even so, it would seem plausible that the extent to which participants were able to find an adequate code for the surfaces they were touching might, at least in part, be reflected by incidental learning experiences that typically define [G.sub.f] (see Kazen-Saad, 1986).
Factor 3: Tactile-Kinesthetic and Broad Visualization Ability (TK/[G.sub.v]). Although seven out of eight tactile-kinesthetic tests load on this factor, it appears unlikely that this represents a "pure" tactile-kinesthetic ability. Thus, an equally high set of loadings is exhibited by three measures of broad visualization ([G.sub.v]; i.e., Card Rotations, Hidden Figures [Level], and Hidden Words [Level]). Apparently, as suggested by several studies in the literature (e.g., Heller, 1989; Schultz & Petersik, 1994), tactile processing and visual processing share much in common. As a result, these two groups of abilities cannot be separated in the present analyses. The `dual' nature of this factor is not compromised by the presence of two additional tests (Visual Bead Memory and Letter Counting [Level]) sharing relatively high loadings on this factor. As explicated earlier, these tests have tactile counterparts (i.e., Tactile Bead Memory and Finger Counting). This alone may account for their communality. Furthermore, both of these non-tactile tests contain a critical visual component. This assertion is obvious with respect to Visual Bead Memory. In this task, the participant is shown a picture of beads stacked upon a rod. After removing the picture from the participant's vision, the person is asked to manually arrange a set of `real' beads in the way shown on the picture. While Letter Counting would appear to contain only a weak visual component, the most frequently employed strategy involves the use of visual imagery (Monty, 1973). In this particular version of the task, the participants are asked to keep track of the number of times each of the three letters is repeated. Many participants imagine three bins corresponding to these letters and hence mentally place incoming letters into ordered bins. In sum, interpretation of this factor as TK/[G.sub.v] would appear justified.
It should be noted that Letter Counting is a well-known marker of the Temporal Tracking primary mental ability factor and of [G.sub.f] at the second-order (e.g., Stankov & Raykov, 1995). Its strong loading on this factor suggests that this TK/[G.sub.v] factor may be closer to the higher-order fluid intelligence construct than it is to any of the lower-order sensory and/or perceptual abilities of [G.sub.f]/[G.sub.c] theory. This proposition finds additional support in consideration of the cognitive demands imposed by most of the tactile tests, and the absence of a salient loading from the more sensory orientated Tactile Texture Test.
To the best of present knowledge, the Gibson's Active/Passive Touch Task has not been employed at any level in individual differences research. The fact that this task (together with the Finger Writing Test) loads on this factor suggests that this ability is more TK than [G.sub.v] in nature. Indeed, this statement would seem to be further justified in separate analyses involving the two versions of Gibson's Touch Task. Thus, the passive version of Gibson's Task (which is essentially perceptual) shares nearly as high a loading on this factor as does the active analogue. Note, however, that this claim remains guarded, since being able to translate the stimuli presented tactually into a visual code clearly remains crucial even to this less demanding pair of tasks.
Factor 4: Test-taking Speed (TTSp). This factor clearly represents the speed of working through cognitive tests of non-trivial difficulty. It has now been replicated several times both in our laboratory (e.g., Roberts, 1995; Stankov et al., 1994) and elsewhere (Horn & Hofer, 1992).(5) Two tests of the Clerical-Perceptual Speed factor did not define this factor in previous research (see Roberts & Stankov, 1994, 1997) nor do they appear to do so in the present study. Accordingly, two scores obtained from Number Comparison have relatively low loadings on the TTSp factor, while the Digit Symbol Test loads on Factor 5. This may be explained by virtue of the fact that there were insufficient markers of [G.sub.s] employed in the present investigation. As a consequence, tests that might have defined this factor at the first-order split their variance across the two mental speed constructs that were more clearly defined. (It might also be recalled that Number Comparison acted somewhat differently between the two samples-possibly accounting for this outcome; see Table 2).
Factor 5: Elementary Cognitive Speed (ECSp). This factor is defined by six measures of the speed of mental operations. Each of these tests is a rather non-demanding chronometric paradigm (where speed is not the by-product of traditionally complex assessment of level abilities). These six measures are known to define at least two distinct factors when embedded within a larger battery of mental speed paradigms (Roberts, 1995; Roberts & Stankov, 1994, 1997). Respectively, these are the Decision Time factor (Math-Classification, Stimulus-Response Compatibility, and Card-Sorting) and Movement Time factor (Fitts' Movement and Card-Dealing). In the present study these two factors failed to separate, plausibly because only two marker tests of the Movement Time factor were employed. Moreover, there was no attempt to assess DT and MT independently in the former subset of these tasks. While it is not clear why Tactile Bead Memory shares loading on this factor, it seems likely that fast hand movement speed would be advantageous to successful performance in this paradigm. Nevertheless, this explanation is not entirely satisfactory given the (negative) loadings of the two scores from the Halstead-Reitan Tactual Performance Test, both of which invoke a similar relationship between hand speed and memory. The direction of these loadings makes this factor bipolar. However, there would appear to be several good reasons to de-emphasize this factor's bipolarity at this stage. This issue will be taken up shortly in more detail.
Factor 6. Digit Span Forward (DsF?). While this factor shares loadings from both level and speed measures of Digit Span Forward, it is notably broader than that constituting a seemingly analogous Factor 7. Indeed, to a large extent, its nomenclature is justified because of its obvious similarities with Factor 7. It is predicted, therefore, that if memory span is an important aspect of this factor, Factors 6 and 7 should define a common factor at the second-order.
It should not go unnoticed that, on the basis of past findings, the level scores from the two Digit Span Tests should have loaded on a single SAR factor. The present outcome may be accounted for by the fact that speed measures from these two tests are not routinely collected. Nevertheless, on the assumption that what happens with all other psychometric tests would also happen with Digit Span, it was postulated that these speed measures would share salient loading on the Test-Taking Speed factor (i.e., Factor 4 of this study). Since this did not happen, the present result suggests that speed measures from the Digit Span Tests should be examined further in future studies. They are clearly different from the other test-taking speed measures and, importantly, they break an otherwise strong SAR factor into two task-related factors.
The additional (small) positive loadings on this factor that result from Esoteric Analogies (Time) and General Knowledge (Time) might reflect the verbal (i.e., acculturated) nature of digits. However, given that negative loadings (particularly from Digit Span Backward [Time]) are unexpected, this proposition would seem premature. Indeed, bipolarity on any factor should not presently be emphasized. Further issues related to the nature of the memory abilities of this study will be taken up in relationship to the exploratory analyses conducted with a reduced set of variables.
Factor 7. Digit Span Backward (DsB). This factor is defined by two measures derived from the same task-speed and accuracy score from Digit Span Backwards. Notably, two tests having low salient loadings on this factor (Tactile Bead Memory and Finger Counting) likewise contain a strong working memory component.
Negative Loadings of the Halstead-Reitan Tactual Performance Test
It is somewhat surprising that two measures derived from the HRNTB Tactual Performance Test have significant negative loadings on Factors 5 and 6. The raw correlation between these measures is 0.425 indicating that the "speed" and "level" measures from this test are indeed sufficiently different and may thus be treated as distinct variables. When these four negative loadings were fixed to zero, the goodness-of-fit statistics for the complete model became less favorable although still acceptable. It was decided to leave these loadings free in the model since they appear to point to an interesting outcome that may well be replicated sometime in the future. However, as mentioned above, these negative loadings should not be taken too seriously at present. There are at least two reasons for this viewpoint. Firstly, the exploratory solution indicates that the two scores from the Tactual Performance Test should load exclusively on the TK/[G.sub.v] factor. Secondly, the loadings of these two variables on the TK/[G.sub.v] (Factor 3) are very high. Thus, it would appear that their negative loadings on each of the additional factors is needed to compensate for the effects of these substantially high loadings. As it turns out, the latter account probably explains why Digit Span Backward (Time) also shares negative loadings on Factor 6 (i.e., DsF).
Correlations among the previously elucidated factors are presented in Table 5. Three features are readily apparent:
1. The [G.sub.c] factor has zero (low) correlations with all other factors.
2. The correlation between [G.sub.f] and TX/[G.sub.v] is high (r = 0.807). Moreover, both [G.sub.f] and TK/[G.sub.v] have relatively high con-elation with ECSp. While the later is not entirely unexpected, the former may indicate conceptual similarity between [G.sub.f] and TK at a second-stratum.
3. The two memory factors (DsB and DsF) show a pattern of correlations that indicate that they share much in common. This may suggest that the inclusion of response time alone has contributed to differences with previous results involving memory processes. Note, however, Carroll (1993, Chapter 7) points to problems interpreting memory constructs such that EFA solutions involving these measures might be seen as informative.
Interpretation of Factors at the Second-Order
As mentioned earlier, the second-order solution was obtained by fixing the factor loadings of the non-standardized first-order solution and fitting second-order matrices instead of factor intercorrelations. The resulting second-order structure(s) are displayed as path coefficients in Figure 1.
One feature is common to both second-order solutions of Figure 1: the [G.sub.c] factor from the first-order solution does not have a significant loading on the second-order factor(s). Clearly, the nature of the "general" factor is influenced by the selection of variables from the battery of cognitive tasks. This point has been repeatedly stressed by some researchers working within the area of intelligence (cf. Carroll, 1993; Horn, 1988) and virtually ignored by others (cf. Jensen & Weng, 1994; Ree, Earles & Teachout, 1994) with possibly serious theoretical and practical implications. In this case, measures of Vocabulary, General Knowledge, and Esoteric Analogies are not loading on the higher-order factor(s). However, it would be absurd to claim that these tests are not measuring intelligence. The nature of the factors at the second-order of analysis is determined by the large number of non-crystallized intelligence (or vulnerable ability) marker tests included in the study.
The solution presented in Figure 1 a shows that the six first-order factors conveniently divide into three groups--"level" factors (TK/[G.sub.v] and [G.sub.f]), "speed" factors (TTSp and ECSp), and "mixture" factors (DsB & DsF). Of the two "level" factors, the TK/[G.sub.v] is the "defining" ability since its loading on the second-order factor is 1.000. Notice, however, that loading of [G.sub.f] on this factor is also rather high (0.846). Another feature of this particular solution is an extremely high correlation between the "level" and "speed" second-order factors (0.985). This correlation is so high that a single factor underlying both "level" and speed" related cognitive factors appears to provide the most efficacious model.
Figure 1b represents the solution with only one factor at the second-order. The highest loadings on this factor come from TK/[G.sub.v] and [G.sub.f] and in particular, the former cognitive factor. The "speed" constructs (TTSp and ECSp) are next in order of magnitude, with DsB and DsF factors having the lowest loadings. It is tempting to conclude that this second-order factor is just the TK/[G.sub.v] factor in the sense in which the general factor (psychometric g) is sometimes said to be no broader than [G.sub.f] (cf. Gustafsson, 1984). In fact, this second-order factor is arguably akin to the fluid intelligence factor, [G.sub.f] obtained in the present investigation (i.e. it is a [G.sub.f] factor with a very strong tactile-kinesthetic component).
Exploratory Factor Analysis with a Reduced Set of Variables
In this section, attention is directed to an EFA (oblimin-rotated principal axis factor solution) examining a subset of twenty-three variables selected from the complete set of thirty-three.
This analysis had three aims. First, there would appear a need to eliminate all variables that were experimentally dependent. For example, since solution time and accuracy scores from the same psychometric test are experimentally dependent, time scores demarcating a Psychometric Test-Taking Speed factor were discarded. Thus, nine measures (Variables 14 to 21 and 31) were removed. In a similar context, one variable (Card-Sorting) was formed by combining scores from Variables 25 and 27 since the former is a function of the later. This reduced battery contains no experimental redundancy that might have affected the factorial structure obtained previously through CFA. Second, EFA rather than CFA was employed in order to gain information about the across-method generality of the preceding results. Elsewhere, Carroll (1995) has alerted theoreticians to certain interpretative problems where EFA and CFA solutions do not converge. Third, correlations rather than covariances were employed as input. This procedure provides information about the across-metric stability of the ensuing factor analytic solution.
The removal of the time-based psychometric scores was expected to result in the TTSp factor no longer being present in the pattern matrix. In addition, because both the DsF and DsB factors are a conglomerate of "level" and "speed" factors, it is possible that these too might be affected. Extending this logic (while remaining faithful to the previous CFA solution) six factors were extracted for this EFA solution.
Table 6 presents the factor pattern matrix for the six factor solution. Comparison with the results of Table 4 shows that two factors [G.sub.c] and ECSp) remain essentially the same as before. A "skeleton" speed factor, resembling the TTSp factor is also visible. This is defined largely by a single variable (21. Number Comparison Time). While this might more readily be identified as a [G.sub.s] factor (see discussion of Factor 4 in the confirmatory factor analysis section of this paper) the nomenclature TTSp is retained in order to draw similarities between the CFA and EFA solutions.
Table 6. Exploratory Factor Analytic Solution (Principal Axis Factoring With Oblimin Rotation) of Psychometric, Mental Speed and Tactile-kinesthetic Measures Selected to Ensure Experimental Independence
[G.sub.c] [G.sub.f]/TK/ TTSp [G.sub.v] Cognitive Ability Measures Level/Accuracy Scores 1. Ravens Progressive Matrices .122 .509 -.022 2. Letter Series .275 .467 -.179 3. Letter Counting -.097 .333 .137 4. Vocabulary .833 .064 -.110 5. General Information .732 -.008 .119 6. Esoteric Analogies .657 .190 -.075 7. Digit Span Forward .039 -.081 -.016 8. Digit Span Backward .148 -.020 -.019 9. Visual Bead Memory .058 .203 -.018 10. Card Rotations .027 .315 .218 11. Hidden Figures .165 .590 .151 12. Hidden Words .147 .192 .098 Mental Speed Measures 21. Number Comparison Time .061 -.088 .716 22. Digit Symbol -.119 .096 .231 23. Math Classification .048 .145 -.038 24. Compatibility (Mean) .144 -.076 .074 25. Card-Sorting DT + 27 -.018 .050 -.075 26. Fitts' MT -.034 -.084 -.084 Tactual Performance Measures 28. Finger Counting -.035 .201 -.046 29. Tactile Texture -.006 .005 -.036 30. Tactile Bead Memory .001 .260 .107 32. HR Tactual Perform Level -.015 .438 .052 33. Tactile Shapes -.109 .442 -.024 ECSp SAR MW [h.sup.2] Cognitive Ability Measures Level/Accuracy Scores 1. Ravens Progressive Matrices .211 .011 -.241 .600 2. Letter Series .157 .004 .116 .477 3. Letter Counting .155 .309 .135 .424 4. Vocabulary -.092 .077 -.045 .710 5. General Information -.007 -.072 -.029 .530 6. Esoteric Analogies .052 .048 .026 .554 7. Digit Span Forward .066 .669 -.085 .436 8. Digit Span Backward -.128 .386 .302 .326 9. Visual Bead Memory .085 .083 .386 .342 10. Card Rotations .090 .182 .076 .313 11. Hidden Figures .149 .151 .201 .513 12. Hidden Words .005 .084 .075 .141 Mental Speed Measures 21. Number Comparison Time -.036 -.042 .056 .519 22. Digit Symbol .648 .020 -.114 .516 23. Math Classification .591 .001 .098 .460 24. Compatibility (Mean) .732 -.040 .262 .715 25. Card-Sorting DT + 27 .223 .212 .421 .309 26. Fitts' MT .311 .034 .005 .112 Tactual Performance Measures 28. Finger Counting .048 .285 .351 .386 29. Tactile Texture -.021 -.045 .001 .061 30. Tactile Bead Memory .062 .063 .612 .455 32. HR Tactual Perform Level -.146 -.023 .157 .261 33. Tactile Shapes .023 .239 .282 .503
Note: Salient loadings (above 0.200) are underlined
Important differences are apparent with respect to the remaining three cognitive ability factors. Thus, the two "level" abilities from the CFA solution [G.sub.f] and TK/[G.sub.v]) have now merged to form a single, broad [G.sub.f]/TK/[G.sub.v] factor. This was not totally unexpected given the high correlation obtained between the [G.sub.f] and TK/[G.sub.v] factors (see Table 5) and the high correlation reported between the F1 and F2 factors of Figure 1. Clearly, the processes underlying fluid intelligence, tactile/kinesthetic performance and broad visualization are rather close in this data set. While the presence of timed measures in the battery may keep the [G.sub.f] apart from the other two psychological processes, the link among these three different domains is indisputable.
The exploratory solution produces somewhat different results regarding the two Digit Span factors (DsF and DsB), leading to slight discrepancies in the interpretation of factors relative to the confirmatory solution reported in Table 4. The exploratory solution of Table 6 includes a well- defined short-term acquisition and retrieval function (SAR) representative of that obtained from the theory of fluid and crystallized intelligence. As expected, the two Digit Span tests load on this factor while the other tests having low but salient loadings each appear to involve an aspect of the SAR function. The last factor in the exploratory solution has loadings from the two Bead Memory Tests (Variables 9 and 30) and the Digit Span Backward Test (Variable 8). These three variables imply a specific memory process. Indeed, it may be argued that this represents a type of Working Memory (WM) factor since Tactile Finger Counting (Variable 28), Tactile Shapes (Variable 33), and Raven's Progressive Matrices (Variable 1) all have high loadings on the current Factor 6. A Working Memory interpretation is supported by salient loading from a composite card-sorting measure (Variables 25 and 27) since performance on this task would appear to benefit from the engagement of WM processes as one sorts cards according to increased task demands (see Roberts et al., 1988). As can be seen in Table 7, this putative WM function correlates relatively highly (r = 0.460) with the [G.sub.f]/TK/[G.sub.v] factor. It should also be noted that the two memory factors show a considerable amount of overlap and that tactile processing is involved in Working Memory.
Table 7. Factor Intercorrelation Matrix of the Solution Given in Table 6 Factors Gc Gf TTSp ECSp SAR WM Gc -- Gf .241 -- TTSp .151 .133 -- ECSp .011 .208 .128 -- SAR .245 .252 .071 .129 -- WM .151 .460 .124 .281 .224 --
Discrepancies between the memory factors reported in the EFA and CFA solutions are rather considerable. This finding calls for further examination of the relationship between memory span, tactile processing, and measures of timed performance. At this stage it would appear injudicious to favor either the DsF and DsB interpretation of CFA, or the SAR and WM interpretation of EFA. Clearly, memory factors do exist in these data but their exact interpretation is unclear. As elaborated in the next section of this paper, these data suggest that the role played by mental speed in short-term memory processes deserves close examination.
Towards a More Complete Taxonomic Model of Human Cognitive Abilities
The main outcome of this study is the presence of a first-order factor that is defined largely by the tactile-kinesthetic performance measures. It is necessary to underscore three important characteristics of this factor.
1. This is not a pure tactile-kinesthetic ability since visual perceptual processes also exhibit salient loadings on this factor. The two types of processes--visual and tactile--cannot be separated easily in these data. As a result it is not known whether tactile and broad visualization abilities can presently be separated empirically. Perhaps a study with a larger number of spatial tasks, or a different selection of tactile tasks, will produce a division between these two domains. Nevertheless, the tactile-kinesthetic aspect of performance assessed with this battery is somewhat more salient than are certain visual features of this factor.
2. The available evidence points to the relative complexity of tactile and kinesthetic abilities captured by this factor. This is apparent from the analyses of the psychological processes that may be involved in various tactile tasks. Thus, several tests have a strong working memory component (e.g., HRNTB; Tactual Performance Test, Tactile Shapes, Bead Memory Tests, Finger and Letter Counting, Gibson's Active and Passive Touch) and this kind of memory is known to be an important characteristic of many fluid intelligence measures. As mentioned in the introductory section of this paper, Carroll (1993) follows the example of others and uses the term -Tactual Performance" ability to refer to a factor defined mainly by the Halstead-Reitan Tact" Performance Test. The present factor is clearly broader than the tests of the Halstead-Reitan battery and it appears that the label "Factile-Kinesthetic Working Memory" captures its non-visual component particularly well. This interpretation is supported not only in the various CFA solutions reported in the present paper, but in the EFA factor intercorrelation matrix of Table 7 that clearly links WM to tactile-kinesthetic processes.(6)
3. This tactile-visualization first-order factor is closely related to the first-order fluid intelligence factor. As a result this tends to define the broad second-order factor of this study. In fact, it is only slightly easier to separate [G.sub.f] processes from a TK/[G.sub.v] factor than it is to separate [G.sub.v] processes from the latter factor--all three types of processes are closely related. This, too, agrees with an interpretation that stresses the relative complexity of the tactile processes captured by this psychological factor. The remaining five first-order constructs of this study are less strongly related to the TK/[G.sub.v] factor than is fluid intelligence.
The other broad factors of this study, by and large, indicate the hypothetical structure that was anticipated. Slight peculiarities of the obtained factorial solution that do appear in this data set reflect on several well-established cognitive abilities rather than on the nature of the tactile-kinesthetic construct per se. One of these anomalies is the appearance of two Digit Span factors. Thus, Forward and Backward Digit Span Tests defined two distinct factors rather than (as hypothesized) a single construct (SAR). The reason for this outcome seems to derive from the peculiar nature of the test-taking speed measures obtained from the Digit Span Tests. Hence, when these speed measures are taken out, a rather typical SAR factor appears (see EFA solution of Table 6). It should also be noted that for the majority of psychometric tests measures of test-taking speed tend to share common variance, defining as they do a broad factor. Inspection of the correlation coefficients obtained in this study provides a possible account for the split between measures of the SAR factor. While for all other tests of this battery the speed measures correlate negatively with the number correct score from the same test (a finding well-repticated in the literature [see Carroll, 1993, Chapter I I]), this correlation is positive for each version of the Digit Span Test. Thus, people who provide their answers to the Digit Span Tests slowly also tend to get larger memory span scores. This is consistent with the proposition that participants who take more time to rehearse the information provided by these test items (or who invest a longer period of time developing an efficient strategy) suffer less memory decay. It appears that it may be profitable to collect information on solution times from the Digit Span Tests routinely and examine this "rehearsal hypothesis" and its impact on the measures of intercorrelations more closely.
The factors of mental speed identified in this study are rather similar to those that were predicted. The Test-Taking Speed factor is exactly as was suggested in the literature review. The second mental speed factor (Elementary Cognitive Speed) is a composite of at least two mental speed factors that exist in the literature. These are the broad Decision Time ([DT.sub.G]) and Movement Time ([MT.sub.G]) factors (see Carroll, 1993, Chapter 15; Roberts, 1995; Roberts & Stankov, 1997). The small number of marker tests of Movement Time coupled with failure to extract hand movement speed measures from several of the paper and pencil chronometric tasks plausibly accounts for the failure to separate these two speed factors in the present study. Note also two interesting features of these speed constructs as they relate to the second-order structure of human cognitive abilities. In Figure 1a, these two factors define a single construct that is elsewhere acknowledged to be a broader speed factor than any so far encapsulated in psychometric models (see Roberts & Stankov, 1997). In Figure 1b, the speed obtained from the elementary cognitive tasks has a quite substantial loading on a general factor (cf. Jensen, 1987; Vernon & Weese, 1993).(7)
Finally, a well-defined crystallized intelligence factor of this study exhibits a very weak relationship to the remaining six factors. As a result it does not load on any broad second-order factors (see Figure 1a). This does not mean, of course, that crystallized abilities lay outside the overall structure of human intelligence--the results of a single study cannot be used to deny the outcomes of a body of research that suggests the obverse. However, this may suggest that more careful consideration needs to be given to [G.sub.c] constructs than has been afforded in the past (cf. Ackerman, 1996).
Tactile Tasks and the Structure of Intelligence: Implications for Psychometrics and Cognitive Psychology
In this section the status of tactile-kinesthetic constructs within the overall structure of human cognitive abilities is discussed. Owing to the abovementioned anomalies, the present account remains speculative at this stage. Initial comments revolve around what might be considered the proper position of the MG, factor within the hierarchy of human intelligence. Basically, two options are worth evaluating. These depend on whether the TK/[G.sub.v] factor is viewed as being (potentially at least) a broad factor like [G.sub.f], [G.sub.v], SAR, and [G.sub.s], or a primary factor like, say, the Test-Taking Speed factor of this study. Under the first option, the TK part of the TK/[G.sub.v] factor may be seen as an indication of a yet-to-be discovered broad tactile function (i.e., [G.sub.h]).(8) This notion depends on the reasonably strong assumption that it will be possible to eliminate (or unconfound) the [G.sub.v] (and/or [G.sub.f]) component of the TK/[G.sub.v] factor in future studies. Figure 2a illustrates this possibility.
Alternatively, as depicted in Figure 2b, the TK/[G.sub.v] factor may be akin to first-order primary factors. In this scenario, both [G.sub.f] and [G.sub.v] processes are implicated and, assuming again that the [G.sub.v] portion can be partialled-out, the TK portion is best interpreted as another primary mental ability underlying fluid intelligence. This interpretation is somewhat pessimistic as to the existence of a broad tactile function ([G.sub.h]). However. the interpretation appears to be more fully supported by the present data set. The existence of a tactile-kinesthetic primary ability that is a measure of fluid intelligence at the second-order does -not diminish the importance of the tactile modality. This primary mental ability would differ from the other primary abilities of fluid intelligence (such as the Inductive Reasoning ability that is captured by Letter Series or the Cognition of Figural Relations ability that is captured by Raven's Progressive Matrices). Thus, rather than distributing variance of tactile tests across different, well-replicated factors, the present study has demonstrated that tactile tests contain much common variance that is not captured by other [G.sub.f] tasks.
A number of suggestions for the design of future studies of tactile-kinesthetic abilities derive from the present findings. In particular, any subsequent work with such tasks should contain a broader battery of [G.sub.f] and [G.sub.v] marker tests. This would allow for an improved understanding of the role of [G.sub.v] and [G.sub.f] in tactile processing. Mental speed measures and the Digit Span Tests appear to be important to our understanding of tactile abilities and it will be necessary to continue exploring the relationship among these three processes. In addition, it is essential to include more simple perceptual and sensory tasks (like Tactile Texture) in future studies. Although "tactile localization" was identified in previous work (see Carroll, 1993), measures of this ability were not included in the present investigation. The existence of measures of such lower-order abilities is important if one is interested in separating tactile from other modalities and in discovering a broad tactile ability factor. Finally, it may be critical to separate kinesthetic and tactile processes during performance within the same cognitive act, in much the same way as it is important to derive independent measures of Decision Time and Movement Time during a sample of choice behavior.
The results presented in this paper are also suggestive of at least three further issues that have been taken up in detail in the individual differences literature.(9) The first is whether or not it is possible to construct a measure of intelligence that is less influenced by the processes of acculturation. The present results open up the possibility of developing a different (and possibly better) culture-fair test since tactile processing (as opposed to the visual processing of geometric shapes currently employed in intelligence tests) is not formally taught within the educational system. Secondly, these tactile tests might offer interesting possibilities to those working with biological correlates of intelligence. For example, brain imaging studies of cognitive performance may use tactile-kinesthetic paradigms to supplement findings with visual and auditory tasks. Finally, contemporary theories of intelligence emphasize the role of limited capacity systems like working memory (e.g., Kyllonen & Christal, 1990; Necka. 1992). A theoretically valid means of studying these systems is the so-called competing (or dual) task paradigm (Stankov, 1983). The existence of tactile tests like those used in the present study opens the possibility of constructing competing tasks by combining visual and auditory tests with those from the tactile domain. It is also possible to construct competing tasks exclusively within the tactile-kinesthetic domain itself. The construction of such tasks and their relationship to intelligence is obviously noteworthy, particularly if "intelligence" is to rid itself of its problematic operationalized basis.
Horn and Noll (1994) have suggested that human cognitive capabilities (and thus conceptions of intelligent behavior) are changing as a function of cultural evolution and technological change. While there can be no denying this statement, there would also appear to be a number of "lower-level" abilities (like tactile-kinesthesia) that are (and will continue to be) pivotal to human cognition. Exploration of such constructs is clearly an important undertaking that has, in recent times, been overshadowed in individual differences research by a desire to establish the cognitive correlates of intelligence, with mental speed providing a generic focus (Stankov & Roberts, 1997). It is our conviction that the latter research program is valid. However, it must progress in parallel with a more complete charting of the structural domain if it is to provide an adequate understanding (i.e., explanation) of intetlligence.
Acknowledgments: This paper was supported in part by a University of Sydney Research Grant to the first author and a small ARC grant to the second author. We would like to thank the following people for their efforts in test construction, data collection and/or scoring of protocols: Anne Antonios, Robyn Bartlett, Elle Bednall, Fleur Buffier, Evelyn Kim Boon, Michaela Davies, Emad Hanna, Charissa, Mak, Greg Petterson, Eugene Phang, Isa Stankov and Rhonda Ting. We would also like to thank Agnes Petocz, Gerry Fogarty, Rod McDonald, Philip L. Ackerman, Lloyd Humphreys and John B. Carroll for their thoughtful comments on earlier drafts of this manuscript. Portions of this manuscript were written while the first author held an NRC Fellowship at Armstrong Laboratory, Brooks AFB, TX to which institutions due acknowledgments are given.
(1.) The term "tactile" is used in preference to "tactual" throughout this paper, although in informal usage these terms may be used interchangeably. The only exception to this rule will be in the referencing of the titles of the cognitive tests employed in the present study that traditionally make use of the latter term.
(2.) This distinction has spawned several studies into tactile constructs (e.g., Heller, 1984: Richardson. Wuillemin & MacKintosh, 1981) that clearly might be worthy of consideration sometime in the future with respect to individual differences.
(3.) Carroll (1993) has, in making a distinction between number correct and time measures, referred to the former as a measure of "level", the latter as a measure of "speed". This distinction, borrowed from Tbomdike et al. (1926), is employed throughout the present paper.
(4.) It may be noted here (and in subsequent passages) that the marker tests of these various factors have been designated by "Variable" numbers. This is because of the presence of computerized tests that provide both measures of "level" and "speed".
(5.) Horn and Hofer (1992) label this factor Correct Decision Speed in order to emphasize that only correct solution times should be analyzed when examining the speed with which subjects attempt complex cognitive tasks. However, using a variety of different techniques, Roberts (1995) has demonstrated that, whether correct or incorrect solution times (or composites) are used in factor analysis, timed performances on psychometric tests tend to cluster together. A similar point has been made by Carroll (1993).
(6.) Baddeley's (1986) model of WM postulates two mechanisms: the visual-spatial scratchpad and the articulatory loop. It is worth noting that the present remit might be taken to suggest that the former concept requires some degree of reformulation.
(7.) With regard to the status of the mental speed constructs, it is likely that what originally was conceptualized as a broad second-order construct --[G.sub.s]--is in actual fact a lower order factor of a much broader speed factor ([G.sub.t]) that minimally encapsulates this earlier construct, Test-Taking Speed, DT and MT, and also possibly constructs such as natural tempo, inspection time, coincidence timing and the like (Roberts & Stankov, 1997; Stankov & Roberts, 1997). Its place within a three-stratum model is unclear given that it would appear to somewhere between stratum I and stratum II.
(8.) The letter "h" (representing haptic) will be reserved for this construct so as to differentiate the notation of this from that used recently to demarcate the broad cognitive speed construct, [G.sub.t] (Roberts & Stankov, 1997).
(9.) The literature review and subsequent analyses reported in this paper also raise a point worthy of discussion regarding clinical applications of psychometric tests. As there exists a relationship between cognitive abilities and two tests contained in the HRNTB, it would appear that "non-tactile" [G.sub.v] marker tests may serve to provide equivalent information regarding brain damage in subjects unable to exhibit tactile abilities (e.g., stroke victims, Parkinsonism).
Ackerman, P. L. (1996). A theory of adult intellectual development: Process. personality, interests, and knowledge. Intelligence, 22, 227-257.
Adevai, G., Silverman, A. J., & McGough, W. E. (1968). Perceptual correlates of the Rod-and-frame test Perceptual and Motor Skills, 26, 1055-1064.
Ainsworth, M. D. S. (1969). Object relations, dependency and attachment: A theoretical view of the mother-infant relationship. Child Development, 40, 969-1025.
Anstey, K., Stankov, L., & Lord, S. (1993). Primary aging, secondary aging and intelligence. Psychology and Aging, 8, 562-570.
Baddeley, A. D. (1986). Working Memory. Oxford: Clarendon Press.
Bentler, P. M., & Wu, E. J. C. (1995). EQS for Macintosh: User's Guide. Encino, CA: Multivariate Software, Inc.
Boyd, T. A., & Hooper, S. R. (1993). WISC-R IQ estimates from the Luria-Nebraska neuropsychological battery. Perceptual & Motor Skills, 77, 683-688.
Brody, N. (1992). Intelligence (Second Edition). New York: Academic Press.
Buckhalt, J. A., Reeve, T. G., & Domier, L. A. (1990). Correlations of movement time and intelligence: Effects of simplifying response requirements. Intelligence, 14, 481-491.
Bum C. (1949). The structure of the mind: A review of the results of factor analysis. British Journal of Educational Psychology, 19, 100-111,176-199.
Carroll, J. B. (1993). Human Cognitive Abilities: A Survey of Factor-Analytic Studies. New York: Cambridge University Press.
Carroll, J. B. (1995). On methodology in the study of cognitive abilities. Multivariate Behavioral Research, 30, 429-452.
Cattell, R. B. (1966). The scree test for the number of factors. Multivariate Behavioral Research, 1, 245-276.
Copeland, D. R., Dowell Jr., R. E., Fletcher, J. M., Sullivan, M. P., Jaffe, N., Cangir, A., Frankel, L. S., & Judd, B. W. (1988). Neuropsychological test performance of pediatric cancer patients at diagnosis and one year later. Journal of Pediatric Psychology, 13, (2). 183-196.
Crossman, E. R. F. W. (1953). Entropy and choice reaction time: The effect of frequency imbalance on choice response. The Quarterly Journal of Experimental Psychology. 5, 2, 41-51.
Duncan, E., Weidl, J., Prickett, H., Vernon, M., & Hotlingsworth-Hodges, T. (1989). The Tactile TONI: A possible new performance IQ test for blind adults? Journal of Visual Impairment & Blindness, 83 (10), 510-511.
Eysenck, H. J. (1995). Can we study intelligence using the experimental method? Intelligence, 20, 3, 217-228.
Fagot, J., Lacreuse, A., & Vauclair, J. (1994). Hand-movement profiles in a tactual-tactual matching task: Effects of spatial factors and laterality. Perception & Psychophysics. 56 (3). 347-355.
French, J. W. (1957). The factorial invariance of pure-factor tests. Journal of Educational Psychology, 48,93-109.
Gardner, H. (1983). Frames of Mind: The Theory of Multiple Intelligences. New York: Basic Books.
Gibson, J. J. (1962). Observations on active touch. Psychological Review, 69,(6),477-491.
Gibson, J. J. (1968). The Senses Considered as Perceptual Systems. London: George Allen & Unwin.
Golden, C. J., Hammeke, T. A., & Purisch. A. D. (1981). The Luna-Nebraska Neuropsychological Battery. Los Angeles, CA: Western Psychological Services.
Gustafsson, J-E. (1984). A unifying model for the structure of intellectual abilities. Intelligence, 8,179-203.
Gunman, L. (1955). A generalized simplex for factor analysis and a faceted definition of intelligence Psychometrika, 20,173-192.
Halstead, W. C. (1947). Brain and Intelligence: A Quantitative Study of the Frontal Lobes. Chicago: University, of Chicago Press.
Harlow, H. F. (1974). Learning to Love. New York: Aronson.
Harre, R. (1981). Great Scientific Experiments. Oxford: Phaidon.
Heller, M. A. (1984). Active and passive touch: The influence of exploration time on form recognition. The Journal of General Psychology, 110, 243-249.
Heller, M. A. (1989). Tactile memory in sighted and blind observers: The influence of orientation and rate of presentation. Perception, 18, 121-133.
Hennessy, M. J., Kirkby, K. C., & Montgomery, I. M. (1991). Comparison of amnesiac effects of midazolam and diazepam. Psychopharmacology, 103, 545-550.
Hollins, M., Faldowski, R., Rao. S., & Young, F. (1993). Perceptual dimensions of tactile surface texture: A multidimensional scaling analysis. Perception and Psychophysics. 54, (6). 697-705.
Horn, J. L. (1985). Remodeling old models of intelligence. In B. B. Wolman (Ed.). Handbook of Intelligence: Theories, Measurements and Applications. New York: Wiley.
Horn, J. L. (1988). Thinking about human abilities. In J. R. Nesselroade & R. B. Cattell (Eds.). Handbook of Multivariate Experimental Psychology. New York: Plenum.
Horn, J. L., & Hofer, S. M. (1992). Major abilities and development in the adult period. In R. J. Sternberg & C. Berg (Eds.), Intellectual Development. New York: Cambridge University Press.
Horn, J. L., & Noll, J. (1994). System for understanding cognitive capabilities: A theory and the evidence on which it is based. In D. K. Detterman. (Ed.), Current Topics in Human Intelligence: Volume IV. New York: Springer.
Humphreys, L. G. (1979). The construct of general intelligence. Intelligence, 3. 105-120.
Jarvis, P. E., & Barth, J. T. (1984). Halstead-Reitan Test Battery: An Interpretive Guide. Odessa, FL: Psychological Assessment Resources.
Jensen, A. R. (1987). Individual differences in the Hick paradigm. In P. A. Vernon (Ed.), Speed of Information-processing and Intelligence. Norwood. New Jersey: Ablex.
Jensen, A. R.. & Weng, L-J. (1994). What is a good g? Intelligence, 18. 231-258.
Joreskog, K. & Sorbom. D. (1989). LISREL 7-A Guide to Program and Applications. Chicago: SPSS Publications.
Kainthola, S. D., & Singh, T. B. (1992). A test of tactile concentration and short term memory. Journal of Visual Impairment and Blindness. May. 219-221.
Kazen-Saad, M. (1986). Visual recognition of equivalent tactile-kinesthetic and verbal information in short-term memory. Perceptual & Motor Skills 63, 547-552.
Kiphart, M. J., Hughes, J. L., Simmons, H. J. P., & Cross. H. A. (1992). Short-term haptic memory for complex objects. Bulletin for the Psychonomic Society. 30 (3). 212-214.
Kudoh, N. (1988). Tactile perception of textured surfaces: Effects of temporal frequency on perceived roughness by passive touch. Tohoku Psychologica Folia. 47, (14). 21-28.
Kyllonen, P. C., & Christal, R. E. (1990). Reasoning ability is (little more dun) working memory capacity) Intelligence, 14,389-433.
Lehrl, S., & Fischer, B. (1988). The basic parameters of human information processing: Their role in the determination of intelligence. Personality and Individual Differences. 9, 5, 883-896.
Lindenberger, U., & Baltes. P. B. (1994). Sensory functioning and intelligence in old age: A strong connection. Psychology and Aging, 9, 339-335.
Loomis, J. M., & Lederman, S. J. (1986). Tactual perception. In K. R. Boff, L. Kaufman, & J. P. Thomas (Eds.), Handbook of Perception and Human Performance, Vol. 2: Cognitive Processes and Performance. New York: John Wiley & Sons.
McCarron, L., & Horn, P. W. (1979). Haptic visual discrimination and intelligence. Journal of Clinical Psychology, 35 (1), 117-121.
Maxwell, J. K., & Niemann, H. (1984). The finger-tip number writing test: Practice effects versus lateral asymmetry. Perceptual & Motor Skills, 59, 343-351.
Maxwell, J. K., Wise, F., Pepping, M., Townes, B. D., Peel. J., & Preston, M. (1994). Fingertip number-writing errors by psychiatric patients. Perceptual & Motor Skills, 59, 933-934.
Moehle, K. A., Rasmussen, J. L., & Fitzhugh-Bell, K. B. (1990). Factor analysis of neuropsychological tests in an adult sample. International Journal of Clinical Psychology, 12, (3-4), 107-115.
Monty, R. A. (1973). Keeping track of sequential events. Ergonomics, 4, 443-454.
Moursy, E. M. (1952). The hierarchical organization of cognitive levels. The British Journal of Psychology, 5 (III), 151-180.
Necka, E. (1992). Cognitive analysis of intelligence: The significance of working memory processes. Personality and Individual Differences, 13, 1031-1046.
Norusis, M. J. (1990). SPSS Reference Guide. Chicago, Illinois: SPSS Inc.
Pettersen, G. (1996). Tactile abilities within the structure of human cognitive abilities. Unpublished B.A. (Hons.) Thesis: University of Sydney.
Piaget, J. (1952). Autobiography. In E. G. Boring (Ed.), A History of Psychology in Autobiography, Vol. 4. Worcester: Clark University Press.
Ree, M. J., Earles, J. A., & Teachout, M. S. (1994). Predicting job performance: Not much more than g. Journal of Applied Psychology, 79, 4, 518-524.
Reitan, R. M., & Wolfson, D. (1985). The Halstead-Reitan Neuropsychological Test Battery: Theory and Clinical Interpretation. Tucson, AZ: Neuropsychology Press.
Reitan, R. M., Wolfson, D., & Horn, J. (1992). Left cerebral dominance for bilateral simultaneous sensory stimulation? Journal of Clinical Psychology, 48(6), 760-766.
Richardson, B. L., Wuillemin, D. B., & MacKintosh, G. J. (1981). Can passive touch be better than active touch? A comparison of active and tactile maze learning. British Journal of Psychology, 72, 353-362.
Roberts, R. D. (1995). Speed of processing within the structure of human cognitive abilities. Unpublished Ph.D. Thesis: University of Sydney.
Roberts, R. D. (1997a). Intraindividual parameters of elementary cognitive tasks (ECTs): Troublesome trends. Submitted for publication in Intelligence.
Roberts, R. D. (1997b). Fitts' law, movement time and intelligence. Personality and Individual Differences, 23, 2,227-246.
Roberts, R. D., Beh. H. C., & Stankov. L. (1988). Hick's law, competing tasks, and intelligence. Intelligence, 12, 2, 111-131.
Roberts, R. D., Pallier, G., & Stankov. L. (1996). The basic information processing (BIP) unit, mental speed and human cognitive abilities: Should the BIP R.I.P? Intelligence, 23, 133-155.
Roberts, R. D., & Stankov, L. (1994). Speed of processing within the structure of human cognitive abilities. Paper presented at the Symposium on Australian Studies of Human Intelligence, 21st Annual Australian Experimental Psychology Conference. University of Sydney, Sydney, April. 1994.
Roberts, R. D., & Stankov, L. (1997). Mental speed and intelligence: A multivariate investigation. Submitted for publication in Psychological Review.
Royce, J. R., Yeudall, L. T., & Bock, C. (1976). Factor analytic studies of human brain damage. I. First and second-order factors and their brain correlates. Multivariate Behavioral Research, 11, 381-418.
Schultz, L. M., & J. T. Petersik (1994). Visual-haptic relations in a two-dimensional size-matching task. Perceptual & Motor Skills, 78, 395-402.
Shaffer, L. H. (1991). Cognition and motor programming. Tutorials in Motor Neuroscience. Kluwer Academic Publishers: Netherlands.
Stankov, L. (1983). The role of competition in human abilities revealed through auditory tests. Multivariate Behavioral Research Monographs. No. 83, pp. 66 & VII.
Stankov, L. (1988). Aging, attention and intelligence. Psychology and Aging. 3, 2, 59-74.
Stankov, L., Boyle. G. J., & Cattell, R. B. (1995). Models and paradigms in intelligence research. In D. Saklofske & M. Zeidner (Eds.), International Handbook of Personality & Intelligence. New York: Plenum Publishing Corporation.
Stankov, L., & Horn, J. L. (1980). Human abilities revealed through auditory tests. Journal of Educational Psychology, 72, 1, 19-42.
Stankov, L., Pallier, G., & Roberts, R. D. (in preparation). Cognitive and personality correlates of tactile-kinesthetic performance.
Stankov, L., & Raykov, T. (1995). Modeling complexity and difficulty in measures of fluid intelligence. Structural Equation Modeling, 2, 335-366.
Stankov, L., & Roberts, R. D. (1997). Mental speed is not the `basic' process of intelligence. Personality and Individual Differences, 22, 1, 69-85.
Stankov, L., Roberts, R. D., & Spilsbury, G. (1994). Attention and speed of test-taking in intelligence and aging. Personality and Individual Differences, 17, 2, 273-284.
Taylor, R. E., & Ward, K. M. (1990). The tale of our search for the Tactile Progressive Matrices. American Psychologist, 45 (1), 69.
Thompson, L. L., & Heaton, R. K. (1991). Pattern performance on the tactual performance test. The Clinical Neuropsychologist, 5 (4), 322-328.
Thorndike, R. L., Hagen, E. P., & Sattler, J. M. (1985). Stanford-Binet Intelligence Scale, 4th Edition. Chicago, IL: Riverside Publishing.
Thurstone, L. L., & Thurstone, T. G. (1941). Factorial Studies of Intelligence. Chicago: The University of Chicago Press.
Ungar, S., Blades, M., Spencer, C., & Morsley, K. (1994). Can visually impaired children use tactile maps to estimate directions? Journal of Visual Impairment & Blindness, May-June, 221-234.
Vernon, P. A., & Weese, S. E. (1993). Predicting intelligence with multiple speed of information-processing tests. Personality and Individual Differences, 14, 3. 413-419.
Verrillo, R. T. (1993). The effects of aging on the sense of touch. In R. T. Verrillo (Ed.), Sensory Research: Multimodal Perspectives. Hillsdale, NJ: Lawrence Erlbaum.
Wechsler, D. (1981). Manual for the Wechsler Adult Intelligence Scale-Revised. New York: The Psychological Corporation.
Woodward, K. L. (1992). The relationship between skin compliance, age, gender, and tactile discriminative thresholds in humans. Somatosensory & Motor Research, 10, (1), 63-67.
Yeudall, L. T., Reddon, J. R., Gill, D. M., & Stefanyk, W. O. (1987). Normative data for the Halstead-Reitan Neuropsychological tests stratified by age and sex. Journal of Clinical Psychology, 43, 346-367.
|Printer friendly Cite/link Email Feedback|
|Author:||Roberts, Richard D.; Stankov, Lazar; Pallier, Gerry; Dolph, Bradley|
|Date:||Sep 1, 1997|
|Previous Article:||Self-confidence and performance on tests of cognitive abilities.|
|Next Article:||Intelligence and regional brain volumes in normal controls.|