One of the key challenges within research on cognitive abilities has been to identify simple cognitive tasks that can explain individual variation in more complex cognitive- performance criteria such as composite ability (IQ) mea- sures, fluid intelligence (Gf),1 and academic achievement (Deary, 2001 Engle, Tuholski, Laughlin, & Conway, 1999 Jensen, 1998 McGrew, 2005 Schweizer, 2005). Most of the simple cognitive tasks that have been identified within this research tradition were originally developed as experimen- tal measures of underlying ���information-processing��� com- ponents that were derived from various models of human cognition (see Floyd, 2005, for a review). Two information- processing components that have received considerable in- terest and support as correlates of more complex cognitive- performance measures are processing speed and working memory (Conway, Cowan, Bunting, Therriault, & Minkoff, 2002 Engle et al., 1999 Fry & Hale, 1996 Luo, Thomp- son, & Detterman, 2006 McGrew, 2005). Processing speed (Gs) is measured using tasks that re- quire rapid performance of simple cognitive operations. Gs tasks are so uniformly simple that people generally differ only in their speed of responding rather than in their accuracy (Carroll, 1993 Horn & Noll, 1997). In contrast, working memory (WM) measures are primarily measures of accuracy, requiring the maintenance of ���memory rep- resentations in the face of concurrent processing, dis- traction, and/or attention shifts��� (Conway et al., 2002, p. 164). What Gs and WM tasks share are their relatively homogeneous measurement units: Gs is measured using the number of uniformly simple items completed within a given time limit WM is measured using the number of uniformly simple representations maintained. The precise relationship between Gs measures and more complex ability measures has been investigated and debated for decades (Burns & Nettelbeck, 2003 Danthiir, Roberts, Schulze, & Wilhelm, 2005 Jensen, 1982, 1998 Nyborg, 2003 Stankov & Roberts, 1997). Similarly, a voluminous amount of research and commentary has fo- cused on the relationship between WM and more complex ability measures (Ackerman, Beier, & Boyle, 2005 Gath- ercole & Pickering, 2000 Just & Carpenter, 1992 Kane, Hambrick, & Conway, 2005 Kyllonen & Christal, 1990 Oberauer, Schulze, Wilhelm, & S����, 2005 Wickelgren, 1997). More recently, researchers have begun to investi- gate the complex interrelationships among Gs, WM, Gf, and broader composite ability measures (Conway et al., 2002 Fry & Hale, 1996, 2000 McGrew, 2005 Schmid & Leiman, 1957 S����, Oberauer, Wittmann, Wilhelm, & Schulze, 2002). The relatively consistent findings of sub- stantial relationships among these variables has led some researchers to propose that Gs and WM tasks are viable assessment tools in educational and clinical settings (Len- genfelder et al., 2006 Luo et al., 2006). Luo, Thompson, and Detterman (2006) argued that Gs and WM tasks require little formal instruction and do not seem to require the higher order processes thought to underlie complex reasoning. Additionally, the relatively homogeneous nature of items in Gs and WM tasks means that item banks can be generated easily and that they are amenable to experimental manipulation. This ease of ma- nipulation also extends to the content and modality fea- tures of the stimuli used, so that Gs and WM tasks can be constructed using verbal, numerical, or spatial stimuli and can be presented visually or aurally in many cases. 969 Copyright 2008 Psychonomic Society, Inc. Assessing the validity of computer-game-like tests of processing speed and working memory Jason McPherson and nicholas r. Burns University of Adelaide, Adelaide, South Australia Processing speed (Gs) and working memory (WM) tasks have received considerable interest as correlates of more complex cognitive performance measures. Gs and WM tasks are often repetitive and are often rigidly pre- sented, however. The effects of Gs and WM may, therefore, be confounded with those of motivation and anxiety. In an effort to address this problem, we assessed the concurrent and predictive validity of computer-game-like tests of Gs (Space Code) and WM (Space Matrix) across two experiments. In Experiment 1, within a university sample (N 5 70), Space Matrix exhibited concurrent validity as a WM measure, whereas Space Code appeared to be a mixed-ability measure. In Experiment 2, Space Matrix exhibited concurrent validity as well as predictive validity (as a predictor of school grades) within a school-aged sample (N 5 94), but the results for Space Code were less encouraging. Relationships between computer-game-like tests and gender, handedness, and computer- game experience are also discussed. Behavior Research Methods 2008, 40 (4), 969-981 doi: 10.3758/BRM.40.4.969 J. McPherson, jlmcpher@psychology.adelaide.edu.au

970 McPherson and Burns Space Code may be a viable measure of Gs, albeit one requiring further research, but there was also some evi- dence suggesting that the task might have a more complex psychometric profile. The development of Space Code was partially inspired by the observation that many computer games inherently require rapid response of a nature similar to that of many Gs tasks. Our development of Space Code was also made easier by the fact that Gs tasks can be readily adapted to different stimulus classes, including game-like graphics in this instance. Similarly, it could also be argued that many modern games require the maintenance of memory repre- sentations in the face of other distracting tasks���much as WM tasks do���such as recalling where objects, doorways, or other rewards are located. Additionally, as with Space Code, there appears to be no reason why game-like graph- ics cannot be used as stimuli within a WM task. The present article describes two related experiments. The overall aim of both experiments was to further assess the validity of Space Code as a measure of Gs, as well as to evaluate a new computer-game-like WM task, Space Matrix. Space Matrix was developed by using the core task of Space Code and adding a simultaneous task based on the dot matrix task developed by Miyake, Friedman, Rettinger, Shah, and Hegarty (2001). In Experiment 1, the concurrent validity of both tests was assessed in relation to established Gs, WM, and Gf measures, using a university student sample. In Experiment 2, concurrent validity was assessed as it was in Experiment 1, and predictive valid- ity was assessed using school subject grades, both within a sample of school-aged children. Additionally, each ex- periment assessed relationships with gender, handedness, and computer-game experience measures. ExPEriMEnt 1 Method Participants There were 70 participants. All participants were Level I psychol- ogy students from the University of Adelaide who received course credit for their participation. The mean age for the whole sample was 19.6 years (SD 5 4.00) 40 were female (mean 5 19.20 years, SD 5 2.37), and 30 were male (mean 5 20.13 years, SD 5 5.48). Materials nonability measures. Participants were asked to indicate their age (in years), their gender, and whether they considered themselves predominantly right-handed, somewhat ambidextrous, or predomi- nantly left-handed (scored 1, 2, and 3, respectively). Additionally, three questions were aimed at measuring computer game experience. The questions were (1) ���On average how many hours do you play computer games each week?��� (2) ���On average how many hours do you play computer games using a mouse each week?��� and (3) ���At any stage in the past what was the maximum number of hours you played computer games in a week?��� Participants were able to freely estimate the number of hours for each question. tests of processing speed (Gs). These tests included the following. Digit Symbol. This test, from the Wechsler Adult Intelligence Scale, required the filling in of blank cells according to a key that was provided at the top of each test sheet. The standard 2-min time limit was used. Visual Matching. This test, taken from the Woodcock���Johnson III Tests of Cognitive Abilities (WJ���III COG Woodcock, McGrew, & Although there is some evidence that domain-specific processes contribute to Gs and WM task performance, research has also suggested that the same underlying pro- cesses are measured at a broader level (McGrew, 2005 Miyake, 2001). Despite the aforementioned advantages and the finding that Gs and WM measures were predic- tors of scholastic achievement, however, Luo et al. also acknowledged that these tasks tend to be rigidly presented and often appear ���impersonal.��� Consequently, they sug- gested that measurement of Gs and WM might be con- founded to a greater extent with psychological effects such as anxiety and weakened motivation (p. 81). We and other researchers have suggested that computer- game-like tasks may provide a less anxiety-provoking and more motivating environment for those being tested (McPherson & Burns, 2005, 2007 Porter, 1995 Wash- burn, 2003). This may be especially true for children, who are increasingly likely to be familiar with computers as a result of their experience with computer games (Gentile & Walsh, 2002 Yelland & Lloyd, 2001). Computer games incorporate a range of unique structural features that act to engage and motivate players, such as real-time positive and negative feedback (via graphics, sounds, and scoring), real- istic graphics and sounds, and player advancement through levels (Wood, Griffiths, Chappell, & Davies, 2004). With the incorporation of such features, computer-game-like Gs and WM tasks could potentially maintain many of the ad- vantages inherent in such measures, while also addressing the concerns raised by Luo et al. (2006). Alternatively, introducing game-like elements might distract those being tested to the point that the construct of interest is no longer being measured reliably. Indeed, since Gs and WM measures are also thought to involve strong attentional demands (Conway et al., 2002 Horn & Blank- son, 2005 Schweizer & Moosbrugger, 2004)���and since such tests may also measure the ability to concentrate on somewhat monotonous tasks���altering motivationally sensitive aspects of the tasks could affect their psychomet- ric properties. We believe, however, that the development of game-like tests may help researchers investigate the separate contributions of speed and attention in Gs tasks once their psychometric properties have been established. In our previous research, we have attempted to address this problem, as well as the additional ergonomic issue of using a computer mouse response method, by developing a computer-game-like test of Gs in distinct stages. In a preliminary study, we developed a simple speeded coding task, using a mouse response method and non- game- like stimuli, and found that the task loaded on a Gs factor (McPherson & Burns, 2005). In two subsequent studies, we investigated the impact of introducing game- like features (McPherson & Burns, 2007). In the first of these studies, we introduced complex game-like graphics as stimuli and found that the task shared variance with both Gs and visuospatial (Gv) tests. In the final study, we added further dynamic game-like features (e.g., scoring, feedback, and additional graphics and sound features) and assessed the task within a larger range of ability tests. The task, Space Code, was found to load strongly on a Gs fac- tor and only minimally on a Gv factor. We concluded that