Visual information processing and intelligence.
It is well established that people have enduring differences in their mental abilities, as tested by psychometric intelligence tests (Pushkar, Gold, Andres, Etezadi, Arbuckle, Schwartzman & Chaikelson, 1995). Moreover, the hierarchical structure of mental abilities--in terms of Spearman's g, group factors and specific abilities--attracts broad agreement (Carroll, 1993). The aggregated facts about psychometric intelligence, though impressive, provide a description of, rather than an explanation for, human mental abilities (Sternberg, 1990). There is a growing recognition of the importance of reductionism in the investigation of human intelligence; reviews of the field increasingly highlight the accumulating corpus of research that attempts to understand intelligence differences by construing them in terms of more basic psychological, psychophysical, psychophysiological and biological phenomena (Matarazzo, 1992; Neisser et al., 1996). It is perhaps useful to make it clear that most of what is referred to here is psychometric intelligence, that is, individual differences in scores on psychometric tests of the IQ-type. Though they have much validity, it is not assumed that IQ-type test scores are synonymous with the broader concept of intelligence.
Reductionism in intelligence may crudely be broken into two forms, (1) cognitive and (2) biological--though there are overlaps (Deary, 1996). Biological reductionism includes studies of intelligence that involve genetics (traditional and molecular), brain size, nerve conduction velocity, electroencephalography and brain potentials and biochemistry (Deary, 1989; Vernon, 1993). Cognitive reductionism includes research on the associations between psychometric intelligence and parameters from various reaction time paradigms and psychophysical procedures, and these tend to be captured by the term `information processing' approaches (Vernon, 1987). The present report aims to contribute to an understanding of these latter, information processing bases of intelligence differences.
One solid finding of the cognitive reductionist approach, which we intend to build upon here, is that psychometric intelligence is significantly associated with individual differences in inspection times. Reviews of this research concluded that individual differences in inspection time (IT) correlate reliably, at about .4 to .5, with IQ-type test scores (Nettelbeck, 1987; Kranzler & Jensen, 1989). Single studies with adequate sample sizes are in accord with these associations (Deary, 1993). The IT procedure is a psychophysical task which assesses the stimulus duration required by a subject in order to make an accurate discrimination. A typical IT stimulus has two parallel, vertical lines, one of which is markedly longer than the other. There are two forms of the stimulus, one with the long line on the right and the other with the long line on the left. A subject in an IT trial is asked to state the position of the long line (i.e., "right" or "left"). The unspeeded examination of the IT stimulus reveals a very easy discrimination, but errors may be induced by exposing the stimulus for brief durations and by following stimulus offset by a backward mask. People who score higher on IQ-type tests, especially Performance IQ measures, can make accurate discriminations on the basis of shorter exposure times than can people who score less. Lower IQ individuals, therefore, must have the stimulus presented to them for a longer period, on average, in order to make an accurate discrimination.
After establishing the IT-IQ correlation, researchers have, appropriately, given much thought to the nature of IT differences (Vickers & Smith, 1986; Levy, 1992; White, 1993, 1996), and to the meaning of the IT-IQ correlation (Deary & Stough, 1996). The reason for this is obvious: if a part of the variance in human intelligence is to be explained in terms of inspection time differences, then we must understand the sensory and/or cognitive processes which contribute to individual differences in IT. The original theory of IT assumed that it was a measure of the efficiency or speed of early visual processing (Vickers, Nettelbeck, & Willson, 1972) or an index of general mental speed (Brand & Deary, 1982). However, other accounts of the essence of IT have been mooted. In particular, there is a view that IT performance draws upon high level cognitive strategies (Pellegrino, 1987; MacKenzie & Bingham, 1985). A difficulty in the straightforward interpretation of inspection time performance, and a spur to the suggestion of higher level factors in IT performance, has been the need for a backward mask to interrupt the processing of the stimulus. Stimulus-mask interactions offer detectable movement artefacts and, though the detection of these artefacts by some participants does not appear to account for the IT-IQ correlation (Egan, 1994; Deary & Stough, 1996), their presence calls into question the simple interpretation of IT in terms of speed of information processing. Different research teams have devised different backward masks to reduce the apparent movement effects in IT testing (Knibb, 1992; Evans & Nettelbeck 1993; Sadler & Deary, 1996). These studies have tended to support the finding that the IT-IQ correlation is best demonstrated in situations where people are not using strategies such as apparent movement.
Vickers, the inventor of the IT task and its theory, saw the stimulus-mask interactions in IT assessment as so problematic that he has developed another task to measure processing speed (the frequency accrual speed task, or FAST; Vickers, 1995; Vickers & McDowall, 1996). However, after detailed analyses of participants' responses to individual FAST items, Vickers, Pietsch, and Hemingway (1995) have acknowledged that FAST is largely a memory task, and Deary and Caryl (1997) have argued that the parameters used for the FAST procedure have precluded the measurement of visual processing speed.
It is not our intention to give a review of all research on IT at present; this has been done to an extent by Deary and Stough (1996). They argue that the IT-IQ correlation is well established, and they adduce enough evidence still to hypothesise that IT does represent a measure of low-level processing speed/efficiency. In particular, the research on the event related potentials collected during the performance of IT tasks (visual or auditory; Zhang, Caryl, & Deary, 1989; Caryl, 1994; Caryl, Golding, & Hall, 1995)--showing that the gradient from the N100 trough to the P200 peak correlates with both IT performance and IQ scores--suggests such an interpretation.
Therefore, this is a time of reassessment in IT research. With the IT-IQ correlation firmly established, research groups have attempted to move on. As shown above, some have concentrated: on the psychophysical/theoretical basis of IT (Deary, Caryl, & Gibson, 1993; White, 1996); on strategies in IT performance (Egan, 1994; Egan & Deary, 1993); on auditory versions of IT (Deary, 1994, 1995); on devising new and supposedly better tasks to get at the elements of human information processing (Vickers, 1995); on the electrophysiological bases of IT (Caryl, 1994); and on the psychopharmacological bases of IT (Petrie & Deary, 1989; Stough, Mangan, Bates, Frank, Kerkin, & Pellet, 1995; McCrimmon, Deary, Huntly, MacLeod, & Frier, 1996). Therefore, the theme of research strategies is: how to proceed? IT affords a small, misty window on human intelligence, and it needs widening and clarifying.
VISUAL CHANGE DETECTION
We shall attempt to widen the window by employing different visual processing tasks in parallel with IT in a study of human intelligence differences. The present study combines empirical-black box and theoretical approaches to understanding the IT-IQ association. With regard to the black box approach, it is important to find out which aspects of the IT task account for its association with IQ-type test scores (i.e., discriminant and convergent validity studies should be performed). This approach treats IT variance as a black box; it recognises that IT differences might arise from a variety of sources of variance but hypothesises that, among the overall variance, there is a significant contribution from a basic information processing parameter, and that it is principally this source of variance that is shared with variance in IQ-type test scores. A rational advance may be made by devising a new task which has in common with IT principally that source of variance which is thought to be responsible for the IT-IQ correlation. In selecting a new task, therefore, it is important to identify the aspects of IT that have proved problematic, for example the stimulus-mask interaction, and eliminate them. In addition, it is a bonus if the new task is theoretically tractable. IT theory is in a state of retrenchment (Levy, 1992; White, 1996), and a successful new task may open up a fresh avenue of theorizing about the information processing bases of human intelligence differences. This study describes a visual processing task that dispenses with some problematic aspects of IT--the backward mask and the narrow attentional locus that can lead to strategy formation--but that retains the essential independent variable of exposure duration.
In seeking out a new visual processing task we have appealed to the psychophysical and psychophysiological studies of On and Off transients in vision that were conducted by Phillips and Singer (1974) and Singer and Phillips (1974). These authors studied subjects' ability to detect appearances and disappearances from a random dot pattern after interruptions of the pattern. The basic structure of each trial in their psychophysical experiments is shown in Figure 1. A dark period is followed by a random dot pattern (t1; containing 50 dots), a variable ISI follows, and the pattern reappears either the same as before (t2), or with one dot added or one missing. The subject's task is to state whether the pattern is the same or has changed. With t1 and t2 at 500 ins, subjects performed almost perfectly at short ISIs, such as 20 ins and 40 ms. With longer ISIs responses fell almost to chance level. With an ISI set at 40 ins, t1 at 60 ins and t2 at 500 ms, the detection of appearance was near-perfect. However, with t1 set at 2 ins, and other parameters set as before, the probability of a correct response fell to about 20%. Phillips and Singer (1974, p. 498) noted that "as t1 decreased the detection of appearance worsened but the detection of disappearance improved." In a further experiment in the same paper, ISIs ranged from 20 ms to 90 ins, t1 was set at 2 ins and t2 was set at 500 ms. The detection of appearance rose from less than 10% at ISI = 20 ins to about 90% at ISI = 90 ms, whereas the detection of disappearance was about 90% for all ISIs. Reductions in t2 duration led to poorer detection of disappearance, but the detection of appearance was less affected. In summary, the longer the duration for which the first pattern may be inspected, either in terms of a long t1 or a long, but not too long, t1+ISI duration, the better was the subject's ability to detect appearance of a new pattern element.
[Figure 1 ILLUSTRATION OMITTED]
Phillips and Singer (1974) concluded that there must be some neural activity that "stands out" in order to be able to detect these small differences in patterns. They focussed on a possible correlation between transient activity in the visual system and performance on their tasks. Their account of the neural mechanisms of the detection of appearance is detailed and only a summary is attempted here (see also Royer & Gilmore, 1985). Their ideas focus on the activity of on-centre cells in the lateral geniculate nucleus. When a dot pattern appears for the first time t1 on-centre cells are depolarised and fall back to resting levels over a few hundred milliseconds. When the stimulus is turned off, the on-centre cells are inhibited by off-centre cells in the vicinity. If the same stimulus re-appears during this period of inhibition, the on-centre response will be weak. However, if a new element in the dot pattern appears in t2, the relevant on-centre cells will not be inhibited, and the new element in the pattern will evoke a strong on-centre cell response. Therefore, the difference in the evoked on-centre response at t2 between the repeated and new elements of the pattern will allow the subject to detect the new element. In a series of experiments on cats, Singer and Phillips (1974) found evidence from intra- and extra-cellular recordings in the lateral geniculate nucleus for the neural mechanisms sketched above. Cellular discharges to reappearing lights were tenfold weaker than discharges to newly appearing lights in a pattern, and the authors noted that "the difference between on-centre cell responses to appearance and interruption decreases as t1 decreases; performance for appearance decreases" (p. 516). Although there has been some disagreement about the necessity of invoking inhibition of off-centre transients--some have suggested that only on-center cell activity is required to explain the psychophysical results of appearance detection (Bourassa, Stelmach, & Di Lollo, 1985; Wilson & Phillips, 1987; Bourassa, Di Lollo, & Stelmach, 1987)--the paradigm devised by Phillips and Singer (1974) and its possible neural bases (Singer & Phillips, 1974) have not been refuted. Moreover, Royer and Gilmore (1985) replicated the key psychophysical results, and explored age-related changes in on/off responses in visual processes.
Given the association between stimulus duration and the probability of the correct detection of appearance in the task devised by Phillips and Singer (1974) we suggest that it provides a psychophysical procedure whose psychometric properties might bear comparison with IT. Royer and Gilmore (1985) also urged integration between visual masking tasks and Phillips and Singer's (1974) procedures. The detection of appearance contains an essential similarity with IT: t1 stimulus duration affects performance. Moreover, in common with IT, no speeded response is required. In addition, the task lacks some features of the IT task. The appearance task requires no mask, and the change to be detected may occur anywhere in a complex pattern across a wide visual field. Thus, there is little possibility of a subject's developing a strategy by focussing attention in a particular part of the display to detect small movement artefacts. Arguably, then, the IT and detection tasks share principally the visual processing variance attributable to the stimulus duration limitations. As such, it will be of interest to investigate the IT-appearance detection association, if any, and their correlations with cognitive ability test scores.
THE PRESENT STUDIES
In this study we describe the development of a task to estimate individual differences in subjects' abilities to detect appearance at brief durations. The task, named Visual Change Detection (VCD) is based on the work of Phillips and Singer (1974) and is very similar to the VCD task used by Wilson, Wiedmann, Phillips and Brooks (1988). In addition, the task used here has similarities with that developed by Royer and Gilmore (1985; Experiment 2) in which they held ISI and t2 constant and used an adaptive staircase psychophysical procedure to find the duration of t1 at which appearance could be detected. Pilot studies on the VCD task are described briefly in Appendix 1 and the main study examines the associations of individual differences in VCD with IQ-type test scores and IT estimates. We shall test two main hypotheses: first, given a battery of visual processing tasks, a latent trait may be derived from those tasks which emphasise stimulus duration and that this latent variable will correlate significantly with cognitive ability test scores; and second, that visual processing tasks which, though potentially difficult, do not emphasise stimulus duration will have near-to-zero correlations with tests of visual processing which are time-limited and with IQ-type test scores.
All participants were health-care workers who responded to advertisements in a hospital for volunteers to help with vision research. Sixty-five participants (37 women, 28 men) took part, all with visual acuity of 6/6 or better (as measured under standard conditions with a Snellen chart). Ages ranged from 20 to 46 years (mean 28.7, SD = 6.3). None of the participants was taking regular medication other than the contraceptive pill, nor had they any significant past medical history (e.g., ocular or neurological disease).
Cognitive Ability Tests. The National Adult Reading Test (NART; Nelson & Willison, 1991) was used as a measure of verbal ability. The test involves reading aloud 50 irregularly-pronounced words. Among normal adults the NART has very high correlations with Full Scale WAIS-R IQ scores. In Crawford's (1992) re-analysis of the NART standardisation sample, the NART correlations with Wechsler Adult Intelligence Scale (WAIS) IQs were, respectively, Full Scale = .74, Verbal = .77 and Performance = .57. In his own representative sample of the Scottish population the respective correlations with WAIS-R were .81, .85 and .57. The Alice Heim IV Test Parts I and II (AH4; Heim, 1970) were administered also. Part I of the test is verbal and numerical in content, whereas PART II is diagrammatic.
Visual Processing Tests. The inspection time (IT) task used in this experiment was run on the same equipment that was used by Deary, Caryl and Gibson (1993). Stimuli and mask were composed of light emitting diodes (LEDs). The cue was an inverted U-shape, 16 mm across by 14 nun high, and it shared the same horizontal cross-bar as the IT stimulus. For the IT stimulus the long line was 29 mm and the short line was 14 mm. The vertical stimulus lines were aligned at the top. The vertical lines of the backward mask were 40 mm. long by 10 mm wide--i.e., they were longer and wider than the stimulus lines and completely covered them. The IT presentation unit was controlled by a BBC computer which also collated the response data.
The cue to begin each trial lasted 300 ms. After the cue offset there was 1000 ms gap before stimulus onset. Stimulus duration could vary from 1 to 400 ms. The duration of the mask was adjusted so that, in every trial, the stimulus+mask duration was 600 ms. After mask offset subjects responded, using a two-choice response box, indicating which line was longer. Subjects were instructed to make responses at leisure, to ensure accuracy. Any responses made before mask offset were not recognised, and a further response was required. Each response initiated the cue for the next trial. No feedback was given about the correctness of responses. Each subject's IT threshold was estimated using the parameter estimation by sequential testing algorithm (PEST; Taylor & Creelman, 1967). This sought the duration at which subjects achieved 85% accuracy. The first stimulus duration was 200 ms, the first step was 75 ms and there was a minimum of five trails per step. The step size halved on reversal and the stopping step size was 1 ms. Most subjects needed about 90-130 trials to reach threshold.
Visual Change Detection (VCD). The design of this task was similar to that described in Pilot Studies 1-3 (Appendix 1), and was based on the ideas of Phillips and Singer (1974) and the VCD task of Wilson, Wiedmann, Phillips and Brooks (1988). (See McCrimmon, Deary, Huntly, MacLeod & Frier (1996) for a full description of the task used here.) The test was designed by one of the authors (IJD) and the control programmes were written in the Department of Psychology, University of Edinburgh. All VCD stimuli were presented on a P22 class, 14", colour, touch-sensitive monitor with a refresh rate of 70 Hz. A virtual 10 x 10 matrix was defined. The stimulus display for this test consisted of an array of 49 small rectangles, positioned randomly in the matrix, to which, after a variable interval, a single (target) rectangle was added. The subject's task in each trial was to identify the newly-appeared rectangle. The overall array size was 185 mm by 105 mm, each rectangle measured 5 mm vertically and 3 mm horizontally, and were non-contiguous. The visual angle of the display was 5 [degrees]. The cue preceded the array by 100 ms and was a circle located in the centre of the screen. The stimulus levels employed--i.e., the intervals between the onset of the 49-rectangle array and the appearance of the target rectangle--were in exact multiples of the screen refresh time (i.e., 14.3, 28.6, 42.9, 57.1, 71.4 and 85.7 ms). Responses to stimuli were made by touching the putative target rectangle on the screen. Feedback was given after each trial. Responses were made at leisure and subjects were encouraged to take their time to ensure accuracy. The onset of the next trial was initiated by the subject. Each of the 6 durations was presented 10 times, making 60 trials in all. VCD accuracy was gauged by taking the total number of correct target identifications.
Visual Movement Detection (VMD). The VMD task resembles the VCD in all aspects except one: the target rectangle, rather than appearing after the rest of the array, appears at the same time as the rest of the array but, after a variable interval, moves to the right or left by a distance identical to its width (i.e., 3 mm). As with the VCD, the stimulus display was generated by lighting 50 rectangles in a potential 10 x 10 matrix. The VMD was run on the same equipment as the VCD and used the same stimulus parameters and response, test and feedback formats.
Contrast Sensitivity. Static contrast sensitivity was measured using the Cambridge Low Contrast Gratings (Clement Clarke International Ltd, Essex, England). In each item of this test the subject views two adjacent pages of a booklet positioned at a distance of 6 m. One of the pages contains horizontal lines (a grating). Each "line" in the grating is made up of small black dots on a white background, separated from each other by equal distances. The opposite page of the booklet has the same number of dots evenly dispersed (i.e., not as lines). Viewed from a distance, the subjective impression is of grey lines with white spaces between them on one page of the booklet, and of a blank page on the other. The subject's task is to identify the page that contains the grating. By varying the number of dots, and the distance between them, a series of gratings are produced with different levels of contrast. The gratings have 11 levels of difficulty, all with the same spatial frequency of 4 cycles/degree. In this study the subject was presented with a block of 50 presentations--10 trials each of the five most difficult gratings, with contrast percentages of 0.37, 0.27, 0.19, 0.14 and [is less than] 0.14--in randomised order. The total number of correctly identified gratings was used as the score.
All subjects were tested on the cognitive ability tests and visual processing tests in the same laboratory. Each subject was tested individually in a single session. The room was blacked-out and the only lighting was artificial (tungsten filament). For the whole testing session, the lighting in the room was adjusted so that the luminance of the non-grating plate in the demonstration pair of the Cambridge Low Contrast Gratings (contrast = 13%) was 100 cd/[m.sup.2]. Subjects undertook the tasks in the same order.
Mean number of correct items (out of 50) on the NART was 36.4 (SD = 8.8, Table 1), which corresponds to a mean (SD) WAIS-R Full Scale IQ of 113.8 (SD = 10.9). Therefore, the present sample of subjects were slightly less than one standard deviation above the mean on IQ level and had about two thirds of the normal population distribution. Means and standard deviations for other measures are shown in Table 1. NART and IT scores were positively skewed (NART raw scores are the numbers of errors made, though note that the number in Table 1 is the number correct; and ITs on the data base were expressed in milliseconds needed to achieve 85% accuracy). Therefore, a logarithmic transformation of these scores was used. Visual Change Detection, Visual Movement Detection and Alice Heim scores were negatively skewed. VCD and VMD scores were cubed and AH4 I and II scores were squared to achieve more normal distributions. All scores for the above variables were then transformed to normalised scores ([micro] = 0; [Sigma] = 1). These are the scores that were used for correlations and multivariate analyses, though the untransformed descriptive statistics are shown as the rightmost column in Table 1. Psychometric functions for the VCD, VMD and contrast sensitivity scores are shown in Figures 2 and 3.
[Figure 2 AND 3 ILLUSTRATION OMITTED]
Table 1. Correlations Among Cognitive Test Scores and Visual Processing Test indices (N = 65). (All correlations are based on scores in which a higher number represents a better performance.)
NART AH4 I A H4 II IT NART - AH4 I .71(***) - AH4 II .58(***) .67(***) - IT (ms) .28(*) .36(**) .45(***) - VCD .22 .31(*) .52(***) .40(***) VMD .18 .36(**) .49(***) .47(***) CS .11 .22 .04 .01 VCD VMD Mean(a)(SD) NART 36.4(8.8) AH4 I 45.1(11.4) AH4 II 50.5(10.5) IT (ms) 56.4(18.5) VCD - 34.8(6.8) VMD .64(***) - 51.8(6.8) CS .05 .10 41.2 (5.6)
Note: NART = National Adult Reading Test;
AH4 = Alice Heim 4 Test;
VCD = Visual Change Detection;
VMD = Visual Movement Detection;
IT = Inspection Time;
CS = Contract sensitivity.
a. The means and SDs refer to the number of correct items in the tests except for IT where the numbers refer to milliseconds. (*) = p <.05; (**) = p < .01; (***) = p [is less than or equal to] .001.
The cognitive ability tests were very highly correlated (Table 1). There was an especially high correlation between the NART and the AH4 I. The AH4 I involves verbal/ numerical reasoning; therefore, one may argue that NART and AH4 I are assessing crystallised/verbal-oriented ability. AH4 II, involving diagrammatic reasoning has an abstract content and is arguably assessing more fluid/performance-oriented ability.
Inspection time correlated significantly with visual change detection and visual movement detection, and VCD and VMD correlated at .64 (Table 1). Correlations between these three visual processing tests and contrast sensitivity were all close to zero. Similarly, none of the correlations between cognitive ability test scores and contrast sensitivity were significant. Correlations between Alice Heim 4 II test scores and IT, VCD and VMD ranged between .45 and .52 (all p [is less than] .001). Alice Heim 4 I scores correlated with IT, VCD and VMD at levels between .31 and .36 (all p [is less than] .01). Only IT, out of the three visual processing tests, correlated significantly with NART scores.
In an exploratory analysis, to discover how much cognitive test ability variance could be accounted for when the visual processing tests were combined, two stepwise regression analyses were carried out, using AH4 I and II scores as the dependent variables. (NART was not examined because it had a significant correlation with only IT.) For AH4 I scores only VMD was entered, so the multivariate R is the same as the univariate r. This indicates that the variance shared among VCD, VMD and IT is related to the variance shared by these tests and IQ-type tests scores. For AH4 II scores both VCD and IT were entered into the equation and together produced a multiple R of .58, indicating that these two visual processing tests accounted for 31.5% (adjusted [R.sup.2]) of the reliable variance in AH4 II scores. However, it should be noted that, again, much of the visual processing test variance that is shared with AH4 II is common: IT added only 5.9% (adjusted [R.sup.2]) of the variance to AH4 II after VCD had been entered first. The next analyses were conducted to model the association between common visual processing variance and measures of psychometric intelligence.
Transformed data from the three cognitive ability tests (NART, AH4 I and AH4 II) and the three visual processing tests (IT, VCD and VMD) were subjected to a confirmatory latent trait analysis using structural equation modelling. The EQS Structural Equations Program was used to perform the modelling (Bentler, 1995; Bender & Wu, 1995). The model tested the following straightforward hypothesis: that a latent trait from the three mental ability tests would have a significant correlation with a latent trait from the three visual processing tests. The size of this correlation was a free parameter in the model (though it was assumed to be significantly greater than zero), as were the loadings of the mental ability tests and the visual processing tests on their respective latent traits (these, also were all hypothesised to be significantly greater than zero). After testing this model using the maximum likelihood method, the Lagrange Multiplier test, which is used to indicate any paths which might be included to improve the fit of the model significantly, indicated a single alteration. In addition to the simple model described above, the pathway between the visual processing latent trait and the AH4 II scores should be included to achieve best fit.
Therefore, the slightly altered, best fit model is shown in Figure 4. The fit of the model will be described first, and then its meaning. The model was tested by the method of maximum likelihood using EQS. The average of the absolute standardised residuals was .029, indicating that most of the variables' covariance was accounted for by the model. The [chi square] of the model was 6.28, d.f. = 7, p = .51, indicating that the residual covariance was not greater than zero in this sample and, therefore, a good fitting model. The fit statistics were as follows: Bentler-Bonett Normed Fit Index =.963; Bentler-Bonett Non-normed Fit Index = 1.01; and Comparative Fit Index = 1.00. These indices may take values from 0 to 1 (though the Bentler-Bonett Non-normed Fit Index achieves values greater than 1 with some very well-fitting models), with values greater than 0.9 indicating a well-fitting model. Therefore, by all these indices, the model in Figure 4 fits the data very well. The parameter estimates in the model, shown adjacent to the arrows in Figure 4, were all highly significant when divided by their standard errors; therefore, each of the paths indicated a significant association between two (latent or manifest) variables. The Wald test indicated there none of these paths could be dropped from the model without significantly impairing the model's fit to the data. Beyond including the F1-AH4 I link in the model, the Lagrange Multiplier test did not indicate any alterations that might be made to improve the fit of the model. The percentage of variance shared by any two adjacent variables may be estimated by squaring the values shown in Figure 4. Therefore, the latent trait from the visual processing test, which correlates at .462 with the latent trait from the cognitive ability tests, shares 21.3% of the variance with general mental ability common to the three ability tests. In addition, the visual processing latent trait contributes 15.8% of the variance directly to AH4 II scores. Using a path-analytic strategy to add the direct and indirect contributions from F1 to AH4 II, 22.5% of the variance in AH4 II scores is shared with the latent visual processing trait.
[Figure 4 ILLUSTRATION OMITTED]
The pilot studies in Appendix 1 were recorded briefly to document the devising of a novel visual processing test. The relationship between t1 (die delay before a target appears in an array after array onset) and the probability of a correct response was such that longer delays afforded better target detection, as was suggested by the results of Phillips and Singer (1974) and Royer and Gilmore (1985). This apparent dependence on speed of processing in the VCD and VMD tests was designed to be shared with inspection time. However, other aspects of VCD and VMD were designed to be different from IT. VCD involves a broad field of attention as the subject awaits the appearance of a target in one of 51 possible positions on a computer screen. Thus, the narrow focus of attention that is required for IT performance, where subjects know exactly the spatial position of forthcoming target information, is not possible. Therefore, attention must be distributed and parallel processing is emphasised in the VCD (and VMD) tasks. In addition, unlike IT, the VCD and VMD stimuli are not masked, and no subjects to date have reported any artefacts that might permit strategies akin to the apparent movement strategy that can arise because of stimulus-mask interactions on the IT task. In summary, VCD and VMD were intended to retain the speed-of-processing element of IT but to differ from IT on aspects that have proved problematic to IT's interpretation and understanding. Contrast sensitivity was included specifically because it was unspeeded and because it made some psychological demands that were similar to IT; it involved a forced, two-choice, visual discrimination with graded levels of difficulty. Therefore, in the main study we were in the position to ask whether, in addition to IT, any of the other visual tasks would correlate with cognitive ability test scores and with IT itself. The upshot was that VCD and VMD--the speeded discrimination tasks--had the same pattern of correlations with cognitive ability tests that IT did, and they were moderately strongly correlated with IT itself. Contrast sensitivity did not correlate significantly with cognitive ability test scores or with IT, VCD or VMD. These convergent and discriminant validity correlations suggest that it might be stimulus duration-related aspects of individual differences in visual processing that lead to the associations with cognitive ability test scores.
In this study, the principal hypothesis was that a general latent trait from the cognitive ability tests would be associated with the general latent trait from the visual processing tests. In fact, the correlation of .46 between these two traits is very similar to some recent estimates of the IT-IQ correlation in near-representative samples of the young adult population (Deary, 1993; Deary & Stough, 1996). Moreover, there was an additional correlation of .40 between the visual processing trait and AH4 II scores. This supports the especially high correlations between IT and performance/non-verbal IQ over more verbal IQ-type tests (Deary, 1993; Kranzler & Jensen, 1989). The fact that it is the latent trait from IT, VCD and VMD that correlates with general mental ability and with AH4 II specifically rebuts explanations of the IT-IQ correlation that focus on peculiarities of the IT setup that afford specific strategies. Instead, we must consider what these tasks share, despite their differences. The obvious feature that is shared is the control of the stimulus duration, and it is arguable that they are all able to detect intersubject variance in the efficiency with which brief visual stimuli are processed. However, is it possible to argue that the latent trait from IT, VCD and VMD represents something less theoretically interesting, such as "visual psychophysical test sophistication?" The results from the contrast sensitivity test argue against this. Individual differences in contrast sensitivity--a two-choice visual discrimination task of graded difficulty, but in which stimulus duration was unlimited--were not related to IT, VCD, VMD, NART, AH4 I or AH4 II.
We note the recent attempt to dissect out those aspects of IT performance that account for its association with cognitive ability test scores. Such aspects range from visual processing speed and general mental speed to high level strategy formation and non-cognitive factors such as motivation, test anxiety, freedom from boredom and personality (Egan, 1994; Deary & Stough, 1996; Stough, Brebner, Nettelbeck, Cooper, Bates, & Mangan, 1996). Out of this attempt to understand the processing elements involved in IT has come a call for IT to be better integrated with other psychophysical research (Levy, 1992; White, 1993, 1996). Such was the approach used here, where a psychophysical task was adapted and developed which shared the processing speed element of IT, but differed in other important ways. There are two benefits, at least, to this approach. First, on the cognitive-psychophysical level, it allows us to manipulate task requirements and to note which aspects are necessary for the retention of the visual processing task-IQ relations. Therefore, whereas we might speculate for ever about which aspects of a single task lead to the IT-IQ correlation, when we have other tasks which share some but not other aspects of the original tasks we can begin to separate out the correlation-critical task requirements (cf. Deary's, 1994, studies in the auditory modality). From the present study one might hypothesise that available stimulus time for discrimination was a key factor, but that the difficulty of the discrimination itself was not.
Second, the introduction of a new task or set of tasks alongside IT opens up a new line of theory for understanding the IT-IQ correlation and for conducting more reductionist research. Although the psychophysics of IT is by no means moribund, and the evoked potential correlates of IT have proved highly replicable, there are clearly many years of research ahead on the investigation of IT, whose initial theoretical bases have been all but cleared away (Levy, 1992; Deary, Caryl & Gibson, 1993; White, 1993, 1996; Vickers et al., 1995). Visual change detection has a body of validating research that is based on both psychophysics and electrophysiology. Also relevant to the concerns of the present study is the finding that older people require significantly longer t1 durations accurately to detect appearance, with ISI and t2 fixed, than younger people matched for educational experience (Royer & Gilmore, 1985). These authors further suggested that the mechanisms studied by Phillips and Singer might be integrated with those of visual masking by appealing to the same physiological mechanisms.
Therefore, VCD offers a parallel set of information on the possible biological bases of the association between visual processing and intelligence as measured by psychometric mental tests. If VCD has parallel electrical activity in the lateral geniculate nucleus (Singer & Phillips, 1974), that is a starting point. However, we know neither whether this activity is associated with successful VCD decisions, nor whether this activity will be the source of individual differences in VCD success. Even if it were, it would still be possible for these individual differences to be a reflection of a general limitation in processing speed, perhaps shared not just by various visual processing tasks, but by tasks in other sensory processing modalities also. In addition it might also be that apparent "processing speed" at the psychological level might be underpinned by a non-speed related mechanism at the physiological level, such as fidelity of stimulus representation (Raz, Willerman, & Yama, 1987). It is commonplace on knowing the time for the completion of a task to infer, by taking the reciprocal of time, a faster speed; if we know that two people arrive at point B at different times having left point A at the same time, we say that one got there faster than the other. At an intuitive level, therefore, it is understandable to speak of "processing speed" when referring to IT, VCD and VMD differences. As Deary (1994) and Rabbitt and Maylor (1991) have argued, however, apparent speed may be subserved by a variety of possible biological sources of variance, some related to speed, some not. Take the electrophysiological correlates of VCD as an example. Singer and Phillips (1974) found activity of On-centre transients in the lateral geniculate nucleus that could register difference at brief stimulus durations. In order maximally to preserve the evanescent difference information that might be contained in a very brief VCD stimulus one can equally well posit that a fast temporal processor (Poppel, 1994), a high-fidelity stimulus representation device (Raz, Willerman, & Yama, 1987) or certain types of inter-cellular interactions (Royer & Gilmore, 1985) could do the job. Such physiological speculations on the bases of behavioural speed measures clearly require specific experiments that can distinguish between them.
It is intended that the principal experiment conducted here win shift the focus of attention within intelligence and information processing research away from a myopic obsession with IT and toward a more integrated approach that shows wider awareness of relevant research in psychophysics and experimental psychology (cf. White, 1996). The processing parameter that IT was intended to measure has been the focus of other psychophysical procedures and theories, some of which have more supporting experimental evidence and theory than IT. VCD is just one example of such a procedure, which deserves further development; another is the identically-named but procedurally distinct "inspection time" used by Bergen and Julesz (1983). Performance on most, if not all, information processing tasks will involve sources of variance other than the one that is of principal interest. However, if two or more procedures are believed to be tapping the same target information processing parameter it is a strong test of this hypothesis to model only the variance that is shared by the tasks.
Bentler, P.M. (1995). EQS structural equations program manual. Encino, CA: Multivariate Software, Inc.
Bentler, P.M., & Wu, E.J.C. (1995). EQS for Macintosh user's guide. Encino, CA: Multivariate Software, Inc.
Bergen, J.R., & Julesz, B. (1983). Rapid discrimination of visual patterns. IEEE transactions on systems, man, and cybernetics, SMC-13, 857-863.
Bourassa, C.M., Di Lollo V., & Stelmach, L.B. (1987). Comment on reply of Wilson and Phillips: General agreement. Vision Research, 27, 1397-1398.
Bourassa, C.M., Stelmach, L.B., & Di Lollo, V. (1985). Equating visibility of brief decrements: Unconfounding duration and luminance. Vision Research, 25, 1241-1244.
Brand, C.R., & Deary, I.J. (1982). Intelligence and `inspection time.' In H.J. Eysenck (Ed.), A model for intelligence. New York: Springer.
Carroll, J.B. (1993). Human cognitive abilities: A survey of factor-analytic studies. Cambridge: Cambridge University Press.
Caryl, P.G. (1994). Event-related potentials correlate with inspection time and intelligence. Intelligence, 18, 15-46.
Caryl, P.G., Golding, S.J.J., & Hall, B.J.D. (1995). Interrelationships among auditory and visual cognitive tasks: an event-related potential (ERP) study. Intelligence, 21, 297-326.
Crawford, J.R. (1992). Current and premorbid intelligence measures in neuropsychological assessment. In J.R. Crawford, D.M. Parker & W.W. MacKinlay (Eds.), A handbook of neuropsychological assessment. Hove, UK: Erlbaum.
Deary, I.J. (1989). Prospects for the biology of human intelligence. Human Evolution, 3, 503-513.
Deary, I.J. (1993). Inspection time and WAIS-R IQ subtypes: A confirmatory factor analysis. Intelligence, 17, 223-236.
Deary, I.J. (1994). Intelligence and auditory discrimination: separating processing speed and fidelity of stimulus representation. Intelligence, 18, 189-213.
Deary, I.J. (1995). Auditory inspection time and intelligence: What is the direction of causation? Developmental Psychology, 31, 237-250.
Deary, I.J. (1996). Reductionism and intelligence: The case of inspection time. Journal of Biosocial Science, 28, 405-423.
Deary, I.J., & Caryl, P. (1997). No so F.A.S.T., Dr. Vickers! Intelligence, 24, 393-400.
Deary, I.J., Caryl, P.G., Egan, V., & Wight, D. (1989). Visual and auditory inspection time: Their relationship and correlations with IQ in high ability subjects. Personality and Individual Differences, 10, 525-533.
Deary, I.J., Caryl, P.G., & Gibson, G.J. (1993). Nonstationarity and the measurement of psychophysical response in a visual inspection time task. Perception, 22, 1245-1256.
Deary, I.J., & Stough, C. (1996). Intelligence and inspection time: Achievements, prospects and problems. American Psychologist, 51, 599-608.
Egan, V. (1994). Intelligence, inspection time and cognitive strategies. British Journal of Psychology, 85, 305-316.
Egan, V., & Deary, I.J. (1993). Does perceptual intake speed reflect intelligent use of feedback in an inspection-time task? The effect of restricted feedback. The Journal of General Psychology, 120, 123-137.
Evans, G., & Nettelbeck, T. (1993). Inspection time: a flash mask to reduce apparent movement effects. Personality and Individual Differences, 15, 91-94.
Heim, A.W. (1970). AH4 group test of general intelligence: Manual. Windsor. NFER-Nelson.
Knibb, K. (1992) A dynamic mask for inspection time. Personality and Individual Differences, 13, 237- 248.
Kranzler, J.H., & Jensen, A.R. (1989). Inspection time and intelligence: a meta-analysis. Intelligence, 13, 329-347.
Levy, P. (1992). Inspection time and its relation to intelligence: Issues of measurement and meaning. Personality and Individual Differences, 13, 987-1002.
Mackenzie, B., & Bingham, E. (1985). IQ, inspection time and response strategies in a university sample. Australian Journal of Psychology, 37, 257-268.
Matarazzo, J.D. (1992). Psychological testing and assessment in the 21st century. American Psychologist, 47, 1007-1018.
McCrimmon, R.J., Deary, I.J., Huntly, B.J.P., MacLeod, K.J., & Frier, B.M. (1996). Visual information processing during controlled hypoglycaemia in humans. Brain, 119, 1277-1287.
Neisser, U., Boodoo, G., Bouchard, T.J., Boykin, A.W., Brody, N., Ceci, S.J., Halpern, D.F., Loehlin, J.C., Perloff, R., Sternberg, R.J., & Urbina, S. (1996). Intelligence: knowns and unknowns.
Nelson, H.E., & Willison, J. (1991). National adult reading test (NART) test manual, 2nd ed. Windsor: NFER-Nelson.
Nettelbeck, T. (1987). Inspection time and intelligence. In P.A. Vernon (Ed.), Speed of information processing and intelligence. Norwood, NJ: Ablex.
Pellegrino, J. (1987). Measuring versus understanding individual differences in cognitive abilities. In Modgil, S. & Modgil, C. (Eds.), Arthur Jensen: Consensus and controversy. New York: Falmer.
Petrie, R.X.A., & Deary, I.J. (1989). Smoking and human information processing. Psychopharmacology, 99, 393-396.
Phillips, W.A., & Singer, W. (1974). Function and interaction of On and Off transients in vision, 1: Psychophysics. Experimental Brain Research, 19, 493-506.
Poppel, E. (1994). Temporal mechanisms in perception. International Review of Neurobiology, 37, 185- 202.
Pushkar, D., Gold, D.P., Andres, D., Etezadi, J., Arbuckle, T., Schwartzman, A., & Chaikelson, J. (1995). Structural equation model of intellectual change and continuity and predictors of intelligence in older men. Psychology and Aging, 10, 294-303.
Rabbitt, P.M.A., & Maylor, E.A. (1991). Investigating models of human performance. British Journal of Psychology, 82, 259-290.
Raven, J.C., Court, J.H., & Raven, J. (1977). Manual for Raven's progressive matrices and vocabulary scales. London: Lewis.
Raz, N., Willerman, L., & Yama, M. (1987). On sense and senses: intelligence and auditory information processing. Personality and Individual Differences, 8, 201-210.
Royer, F.L., & Gilmore, G.C. (1985). Age and functions of the transient component of ON and OFF responses in visual processes. Quarterly Journal of Experimental Psychology, 37A, 147-170.
Sadler, A.S., & Deary, I.J. (1996). Cerebral asymmetries in inspection time? Neuropsychologia, 34, 283-295.
Singer, W., & Phillips, W.A. (1974). Function and interaction of On and Off transients in vision, II: Neurophysiology. Experimental Brain Research, 19, 507-521.
Sternberg, R.J. (1990). Metaphors of mind. Cambridge: Cambridge University Press.
Stough, C., Brebner, J., Nettelbeck, T., Cooper, C.J., Bates, T., & Mangan, G.L. (1996). The relationship between intelligence, personality and inspection time. British Journal of Psychology, 87, 255-268.
Stough, C., Mangan, G., Bates, T., Frank, N., Kerkin, B. & Pellett, O. (1995). Effects of nicotine on perceptual speed. Psychopharmacology, 119, 305-310.
Taylor, M.M., & Creelman, C.D. (1967). PEST: efficient estimates in probability functions. Journal of the Acoustical Society of America, 41, 782-787.
Vernon, P.A. (1987). Speed of information processing and intelligence. Norwood, NJ: Ablex.
Vernon, P.A. (1993). Biological approaches to the study of human intelligence. Norwood, NJ: Ablex.
Vickers, D. (1995). The frequency accrual speed test(FAST): A new measure of "mental speed"? Personality and Individual Differences, 19, 863-879.
Vickers, D., & McDowell, A. (1996). Accuracy in the frequency accrual speed test (FAST), inspection time and psychometric intelligence in a sample of primary school children. Personality and Individual Differences, 20, 463-469.
Vickers, D., Nettelbeck, T., & Willson, R.J. (1972). Perceptual indices of performance: the measurement of "inspection time" and "noise" in the visual system. Perception, 1, 263-295.
Vickers, D., Pietsch, A., & Hemingway, T. (1995). Intelligence and visual and auditory discrimination: Evidence that the relationship is not due to the rate at which sensory information is sampled. Intelligence, 21, 197-224.
Vickers, D., & Smith, P. (1986). The rationale for the inspection time index. Personality and Individual Differences, 7, 609-624.
White, M. (1993). The inspection time rationale fails to demonstrate that inspection time is a measure of the speed of post- sensory processing. Personality and Individual Differences, 15, 185-198.
White, M. (1996). Interpreting inspection time as a measure of the speed of sensory processing. Personality and Individual Differences, 20, 351-363.
Wilson, J.T.L., & Phillips, W.A. (1987). Confounding effects of dark intervals: A reply to Bourassa, Stelmach and Di Lollo. Vision Research, 27, 1393-1395.
Wilson, J.T.L., Wiedmann, K.D., Phillips, W.A., & Brooks, D.N. (1988). Visual event perception in alcoholics. Journal of Clinical and Experimental Neuropsychology, 10, 222-234.
Zhang, Y., Caryl, P.G., & Deary, I.J. (1989). Evoked potentials, inspection time and intelligence. Personality and Individual Differences, 10, 1079-1094.
Pilot Study 1
Fifty eight applicants seeking to join the Royal Air Force (RAF) as officers and aircrew (pilots and navigators) acted as subjects. They were tested at the officer and aircrew selection centre (OASC), RAF Cranwell. All subjects were in their late teens or early twenties and were of high academic ability. A prototype Visual Change Detection (VCD) test was used, very similar to that used in the main study above. The test was designed by one of the authors (IJD), adapted by Science 3b(Air) of the Ministry of Defence, and programmed by RAF technicians to run on the OASC Archimedes computer-based test administration system. The delay between the appearance of the 49 rectangles and the single target rectangle ranged from 0 ins to 116.7 ins, in steps of 16.7 ms. Hereafter, these durations will be called level 0 to level 6. Each stimulus level had 15 repeats, making a total of 105 trials overall. The probability of a correct response increases with increasing stimulus duration (Table 2). There were no significant differences in the probabilities of target detection in the different screen areas (Figure 5).
Table 2. Accuracy, Response Times and Error Distances for each Stimulus Level of the Visual Change Detection Task in Pilot Study 1 (N = 58). Level 0 1 2 3 4 5 6 Number correct 1.03 3.09 9.26 11.69 11.55 12.12 12.19 (out of 15) RT (sec.) 3.69 3.75 4.31 3.78 4.03 3.97 3.69 Deviation X 3.10 2.25 0.77 0.29 0.23 0.14 0.14 Deviation Y 3.35 2.34 0.79 0.32 0.34 0.23 0.21
[Figure 5 ILLUSTRATION OMITTED]
Pilot Study 2
Seventy further candidates for officer training at RAF Cranwell were tested on the VCD task. Thirty one of the subjects completed Raven's Advanced Progressive Matrices (Raven, Court, & Raven, 1977). The mean number of correct responses for stimulus levels 0 to 6 was 1.06, 2.84, 10.27, 12.13, 12.34, 12.56 and 12.51. For the subset of subjects who took the Raven's Advanced Progressive Matrices (APM), their mean score was 23.4, with a standard deviation of 5.8. These are similar to high-level UK university students' data reported by Deary, Caryl, Egan, and Wight (1989). The correlation between the total number of correct responses on the VCD task and scores on the APM was 0.20 (ns).
Pilot Study 3
Three hundred and thirty nine applicants (272 men, 67 women) were tested at OASC, RAF Cranwell. VCD was administered as described above. Only stimulus levels 1 to 4, inclusive, were used, with 20 trials at each level. In addition, subjects undertook the RAF's Critical Reasoning Battery (CRB). This is an RAF, in-house produced series of reasoning questions and has a correlation of .50 with Raven's Advanced Progressive Matrices. The mean number of correct responses (out of a possible 20) for the four stimulus levels was 3.77, 13.57, 16.09 and 16.49. The correlation between the total score on the VCD and the CRB was 0. 17 (p [is less than] .025). The correlation between the odd and even items in the VCD test (split-half reliability) was 0.66 (p [is less than] .001). The correlation between the first and second halves of the test was 0.51 (p [is less than] .001). There was a small but significant practice effect; there was a mean of 23.5 (out of 40; SD = 4.4) correct responses in the first half of the test versus a mean of 26.5 (SD = 4.5) for the second half (p [is less than] .001). As in Study 1, the performance in different areas of the stimulus screen were examined (Figure 5). There was a significant main effect of screen area (F = 36.2, df. = 4, 1316, p [is less than] .001), and 90% of the variance of this effect was explained by the contrast of centre and bottom left versus top right. This effect was found to be caused by a programming error which resulted in a disproportionate number of the early trials being presented in the top right quadrant. Subsequent testing shows no significant differences across screen areas.
DISCUSSION OF PILOT STUDIES
The visual change detection task has been piloted on three separate samples of young people with very high cognitive ability. The relationships between stimulus level and response accuracy show the expected psychometric function. There is a slight, but significant, association between VCD performance and reasoning scores, but the samples thus far have had a very attenuated range of ability and, where the sample was large, the reasoning test was not widely validated. Overall, there is no evidence for any significant effect of screen area on accuracy of change detection. The reliability of even a short test was satisfactory, and the practice within such a test, though significant in a large sample, was not large in effect size.
Acknowledgements: This pilot work has been carried out with the support of the Directorate of Recruiting and Selection at RAF Cranwell.
Direct all correspondence to: Ian J. Deary, Department of Psychology, University of Edinburgh, 7 George Square, Edinburgh EH8 9JZ, Scotland, UK <I.Deary@ed.ac.uk>.
|Printer friendly Cite/link Email Feedback|
|Author:||Deary, Ian J.; McCrimmon, Rory J.; Bradshaw, Jonathan|
|Date:||May 1, 1997|
|Previous Article:||The role of attention in the relationship between Inspection Time and IQ in children.|
|Next Article:||Is the acquisition of a theory of mind linked to a specific competence beyond three years of age?|