Determinants of Adult Age Differences on Synthetic Work Performance

Journal of Experimental Psychology: Applied Copyright 1996 by the American Psychological Association, Inc. 1996, Vol. 2, No. 4, 305-329 1076-898X/96/$3.00 Determinants of Adult Age Differences on Synthetic Work Performance Timothy A. Salthouse, David Z. Hambrick, Kristen E. Lukas, and T. C. Dell Georgia Institute of Technology Synthetic work research is designed to simulate complex work activities by requiring participants to perform several concurrent tasks. The current project consisted of 2 studies in which adults of different ages performed 4 tasks during 25 sessions in a synthetic work situation in 5-min periods across 3 days. Large age differences were evident in the total score in both studies, and they were maintained across all stages of practice. Detailed analyses revealed that with increased age adults in this time management activity were less likely to perform self-paced tasks and to attempt difficult auditory discrimination judgments. Very little independent age-related influences were evident after the initial few sessions on the task. More than 70% of the age-related variance after nearly 2 hr of practice was shared with measures of processing speed obtained before performing the tasks. These results suggest that age-related differences in basic processing efficiency may be responsible for a large proportion of the age-related influences on the performance of moderately complex activities presumed to be similar to those likely to be encountered in a variety of work situations. A potential implication of the results of these studies is that increased age is likely to be a disadvantage in at least the initial phases of performance in many jobs.' As the average age of the population increases there has been growing interest in the abilities and capacities of older workers. Unfortunately, relatively little is known about the actual work performance of adults of different ages because it is often difficult to obtain detailed assessments while individuals are performing real jobs (however see Salthouse & Maurer, 1996, for a recent review). Not only is the process of assessment potentially intrusive and disruptive to normal operations, and thus may not be well re- Timothy A. Salthouse, David Z. Hambrick, Kristen E. Lukas, and T. C. Dell, School of Psychology, Georgia Institute of Technology. This research was supported by National Institute on Aging Grant AG R376826. We thank Richard Sit for collecting auditory discrimination data in the retest session. Correspondence concerning this article should be addressed to Timothy A. Salthouse, School of Psychology, Georgia Institute of Technology, Atlanta, Georgia 30332-0170. Electronic mail may be sent via Internet to tim. salthouse @ psych.gatech.edu. ceived by management, but many workers could be reluctant to participate if they cannot be convinced that the project is not intended to evaluate their individual level of productivity. Attempts should still be made to examine the relations between age and performance in actual job situations, but other approaches to predicting the functioning of people in the world outside the laboratory should also be explored. Three quite different approaches have been used to relate research in cognitive psychology to functioning outside of the laboratory. By far the most common approach has been to rely on psychometric ability measures to link cognitive constructs to performance in natural situations. That is, within this approach the researcher concentrates on measures designed to assess primary cognitive abilities, and then reference is made to the large literature in industrial and personnel psychology in which scores on psychometric tests have been found to be related to criteria such as job performance and occupational level (e.g., J. E. Hunter & R. E Hunter, 1984; Schmidt & J. E. Hunter, 1992). 305

306 SALTHOUSE, HAMBRICK, LUKAS, AND DELL The fundamental assumption underlying the psychometric approach is that a small number of abilities are sufficient to account for a large proportion of the variance in performance across a wide range of activities because those abilities are presumably required for successful performance in many and possibly all moderately complex tasks. The psychometric approach has the advantage of providing very efficient assessment, and there is a large literature concerned with the validity of psychometric measures. It also provides a parsimonious set of predictors to a wide range of occupational and academic activities. A possible disadvantage of this approach is the concern that prediction may not be optimal because the ability measures are usually quite abstract and often have little face validity for many activities. A second approach to linking functioning in the laboratory to that in life outside the laboratory consists of detailed analyses of specific jobs or activities. That is, one particular job or activity is selected, and then an attempt is made to analyze performance on it in terms of cognitive constructs. The target activity could be as broad and complex as automobile driving, but is often more limited, such as learning to use selected functions of a word processing package. The task analysis approach offers a potentially rich set of data that could be informative about the strategies, processes, and mechanisms used in a specific activity. However, generalizability may be limited to very similar types of situations. For example, a task analysis of word processing skill would probably not be very useful in understanding the on-the-job functioning of an airline pilot. Furthermore, it is often very time consuming to perform a task analysis and to obtain valid and reliable measures of each of the hypothesized components. A third approach that can be used to relate laboratory research to functioning in real-world activities involves the use of simulations of the characteristics postulated to be common to many complex activities. In this approach, the goal is to abstract what are assumed to be the critical aspects of a wide range of activities and then to measure them in a controlled setting. However, it is important to note that the purpose is not just to assess the component abilities in isolation but also to evaluate the ability to time share among several concurrent activities. That is, the synthetic work perspective emphasizes the coordination of two or more dynamically changing components with varying levels of difficulty and importance, not simply the abilities required for performance of the individual components (A1- luisi, 1967; Morgan & Alluisi, 1972). The usefulness of the synthetic work approach depends on the extent to which relevant aspects of actual jobs have been incorporated in the simulation. If the simulation is successful, it may incorporate the major advantages of both the psychometric approach and the task analysis approach. A disadvantage is that although synthetic work situations have been designed to incorporate components of a wide variety of work activities, little validity information is typically available to indicate that the simulations are truly accurate or realistic. Each approach has strengths and weaknesses, and all three are probably needed to obtain a complete picture of how cognition relates to functioning outside the laboratory. In this project, we relied primarily on the third approach by using Elsmore's SYNWORK1 (1994) computer program. The program is designed to incorporate the dynamic aspects of complex work activities, and consists of four tasks that are all quite simple when performed in isolation. They include remembering items on demand, performing a self-paced task requiring concentration, and monitoring and reacting to both visual and auditory information. In addition to the abilities required in the individual tasks, the synthetic work situation requires time management skills to deal with multiple concurrent demands. Although there are apparently no studies in which performance in synthetic work situations has been related to measures of work performance, synthetic work activities were developed to assemble what were assumed to be important components of many different work situations together in a manner that would allow sensitive measurement of all relevant aspects (Alluisi, 1967; Elsmore, 1994). Even if subsequent research were to find only weak relations to actual measures of job performance, synthetic work activities should be of interest because they are clearly more complex than most laboratory tasks. Therefore, the motiva-

SYNTHETIC WORK 307 tion for the current project was that understanding the sources of individual differences in moderately complex synthetic work activities would likely be informative about differences that might occur in a variety of actual job situations. At the end of Study 2, we asked participants whether the synthetic work activity resembled anything from their own experience. Only a few individuals did not provide a response, and thus most apparently perceived some similarity to real-world activities. In addition to various specific work situations (e.g., working as a receptionist in a busy office, preparing and baking pizzas), the activities mentioned by 2 or more respondents included driving in traffic or on an interstate highway, preparing a multicourse meal, and monitoring and interacting with small children, including a response by a woman who said that it reminded her of having to babysit nine children. Study 1 The primary goal in Study 1 was to determine if there are adult age differences in the initial level of performance and in the rate of improvement on this synthetic or artificial work situation. If there is, we want to specify how the level of performance and type of improvement differ across age groups. Moreover, we seek to identify the factors that predict the level of proficiency reached after several hours of practice on the task. We are also interested in determining whether independent age-related influences are no longer evident after a certain amount of practice on the task. If this is true, then most of the age differences were attributable to factors operating in the earliest periods of experience with the situation. Method Participants. There were 39 college students and 33 older adults recruited from newspaper advertisements who participated in the study (for their descriptive characteristics, see Table 1). There were no age differences in self-ratings of health, but older adults reported more healthrelated limitations, more frequent use of antihypertension medications, and more incidents of cardiovascular surgery. Older adults had higher vocabulary scores but were slower in the perceptual comparison and reaction time tasks. The older adults were also slower and less accurate on the mouse control task. This pattern of differences is typical of that reported in many studies comparing adults of different ages in measures of cognitive functioning. Procedure. All participants came to the laboratory on 3 separate days within a 10-day period. Most completed the project within the same week that they started. Testing was conducted in groups of 1 to 5 individuals but participants each had their own microcomputer workstation. On Day 1, a background questionnaire was administered followed by two paper-and-pencil perceptual comparison speed tasks, 10-item multiple choice synonym and antonym vocabulary tests, and two computer-controlled reaction time tasks. These tests were identical to those described in earlier articles (e.g., Salthouse, 1995; Salthouse, Fristoe, Lineweaver, & Coon, 1995); hence they are described only briefly here. The two perceptual comparison tests (Letter Comparison and Pattern Comparison) each consisted of a page of instructions with examples, followed by two test pages. The test pages contained pairs of 3 to 12 letters (Letter Comparison) or line segments (Pattern Comparison). Participants had to reply by writing an S for same or a D for different on the line between the members of the pair according to whether the members of the pair were the same or different. We allowed 30 s for each test page, and the score was the average of the number of correct responses minus the number of incorrect responses across both pages. The two reaction time tasks involved either physical identity decisions for a pair of digits (Digit Digit) or decisions whether a digit and a symbol were equivalent according to a code table displayed at the top of the screen (Digit Symbol). Same decisions were communicated by pressing the "slash" key, and different decisions were communicated by pressing the z key on the keyboard. In both tasks, a practice block of 18 trials was administered followed by an experimental block of 90 trials. Because average accuracy was greater than 95% in both tasks, the median reaction time was used as the primary measure of performance. A mouse control training exercise was then administered in which the participant used the mouse to move a cursor through a maze shaped

308 SALTHOUSE, HAMBRICK, LUKAS, AND DELL Table 1 Characteristics of Participants in Study I Younger (n = 39) a Older (n = 33) b Variable M SD M SD t(70) Age (years) 19.7 Education (years) 13.5 Health rating 1 2.1 2 2.3 Health satisfaction 2.1 Health-related limitations 1.4 Cardiovascular surgery 0.0 Hypertension medications 0.0 Head injury 0.05 Neurological treatment 0.05 Comparison Letter 11.9 Pattern 20.5 Vocabulary Synonym 4.7 Antonym 4.8 Reaction time (ms) Digit digit 551 Digit symbol 1,056 Mouse time (s) First 3 trials 7.54 Last 3 trials 4.52 Errors First 3 trials 0.9 Last 3 trials 0.6 72 176 1.5 68.7 1.3 15.3 4.9 1.7 --5.02* 0.9 2.3 1.1 --0.63 0.8 2.3 0.9 --0.13 0.8 2.4 0.7 -- 1.52 0.7 1.9 1.0 --2.94* 0.2 0.4 --2.60* 0.3 0.5 --3.77* 0.2 0.2 0.4 -- 1.77 0.2 0.06 0.2 --0.17 2.4 8.6 2.8 5.40* 3.6 14.3 3.4 7.48* 2.1 7.8 2.3 6.6 2.5 --5.99* 3.0 --2.88* 719 230 --4.33* 1,595 516 --6.13" 3.0 27.51 19.1 -- 6.44* 1.3 18.92 14.6 --6.14" 0.7 5.1 4.7 -- 5.40" 0.6 2.4 2.5 --4.32* Note. Health rating, health satisfaction, and health-related limitations were determined by a 5-point scale where lower numbers indicate better health, and responses to the cardiovascular surgery, hypertension medications, head injury, and neurological treatment variables were yes or no questions; thus the means correspond to the proportion of individuals reporting a positive response. ~66.7% were women, b54.5% were women. *p <.01. like the letter W. Eight trims were presented on this task. The participant was able to see the time in seconds and the number of errors displayed after each attempt. The participants were then introduced to the synthetic work program through written instruction and a short practice session on each of the component tasks. The written instructions explained the task, and an examiner was present to answer questions about procedures. Each task was then presented in isolation for 1 min and followed by the four tasks presented together for 1 rain. At this point, the participants were reminded that the goal was to obtain the highest possible number of points in each session. The remainder of Day 1 was devoted to the four tasks together for the 5 sessions, which were 5 min each. On Days 2 and 3, participants performed 10 sessions of 5 min each with all four tasks together. Figure 1 illustrates the SYNWORK1 display screen when all four tasks are operative. Displays of the tasks occupied a full screen on a color display monitor. However, because the viewing distance was not constrained, it was impossible to specify the exact size of the stimuli in terms of visual angle. All responses within the synthetic work program were executed with a mouse. In

SYNTHETIC WORK 309 I BCWRA I-;I I YES I NO I RE.SET J 829 563 0000 / / \ \ + + + + F ] I HIGH SOUND REPORT I I IIIIIIIIIIIIIIIIIIIIIIIIII Figure 1. program. Illustration of computer screen with all four tasks in the SYNWORK1 the upper left of the screen was a memory task consisting of a set of five letters. This set was then removed and followed by periodic displays of a probe letter, which was to be classified as yes (for a member of the set) or no (not a member of the set). At any point during the session, the participant could retrieve the set of memory letters by moving the cursor to the box that previously contained the letters and pressing a button on the mouse. The upper right of the display contained an arithmetic task, where two 3-digit numbers were to be added by adjusting plus and minus signs to produce the correct sum in the row below the addends. This task was completely selfpaced. The lower left quadrant of the display contained a visual monitoring task, where the participant monitored the position of a pointer moving continuously along a horizontal scale and attempted to reset it before it reached the end of the scale. The lower right quadrant of the display contained an auditory monitoring task. High and low tones were presented periodically throughout the session, and the task was to respond whenever a high tone occurred. The SYNWORK1 program (Elsmore, 1994) allows considerable flexibility in terms of the presentation of the tasks and manipulation of relevant parameters. Values of the parameters used in this study were as follows: In the memory task, there were five letters in the memory set, and they remained constant throughout each 5-rain session but were varied across sessions. Probe stimuli occurred every 10 s; 10 points were awarded for every correct response, and 10 points were subtracted for every incorrect response and for every retrieval of the memory list after the initial display at the beginning of the session. In this task, the highest possible score was 300 because 10 points could be acquired with a correct response to each of the 30 probe stimuli

310 SALTHOUSE, HAMBRICK, LUKAS, AND DELL (i.e., one probe every 10 s for 5 min). Parameters in the arithmetic task specified that two 3-digit numbers were to be added. Again, 10 points were awarded for every correct response, and 10 points were subtracted for every incorrect response. Because this task was self-paced, the maximum score depended on how quickly and accurately the individual could perform the arithmetic operations. In the visual monitoring task, the pointer line required 15 s to move from the middle to the end of the scale. We awarded 1 point for every 10 pixels in which the line was away from the middle of the scale and deducted 10 points for every second in which the line was on the end of the scale. Therefore, the maximum possible score in this task was 200, if the line was reset each time at the very end of the scale (i.e., 10 points 20 opportunities.!f in 5 min if the reset I interval was 15 s). Finally, parameters within the auditory monitoring task specified that the events occurred every 5 s, had a.2 probability of targets, with a target frequency of 1319 Hz and a nontarget frequency of 1046 Hz. We awarded 10 points for every hit (correct detection) but subtracted 10 points for every miss (false alarm). If exactly 12 targets (i.e.,.2 60 events with an event every 5 s for 5 min) were presented, then the maximum score would be 120. An index of overall performance, corresponding to the sum of the points accumulated in the four tasks, was continuously updated and was visible in the middle of the display. Results Figure 2 displays the mean total points (sum of points across the four tasks) by session for the Day I Day 2 Day 3 300 100 aa~ 0 /.L w e -100-I i I I I i I I I I I I I I I I I I i I I I I i i 2 3 4 5 6 7 8 9 10111213141516171819202122232425 Young Old Session Figure 2. Means and standard errors of the total number of points by session for the young and old adults in Study 1.

SYNTHETIC WORK 311 Memory Arithmetic Day1 Day2 Day3 Day1 Day2 Day3 3O0 O 200 '~" EL,.. 100 D.Q E :3 Z o -loo 13,,,,,, J,, J,, 5 7 9 11 13 15 17 19 21 23 25 Visual Monitoring Day1 Day2 Day3 2OO loo.f -100 _ ~-~.r~ 43- tg 43) 'ID" B "Q" (g* B "O" E]'B "ID" 0 - I vou_ I no 300 ;a,(3. ErJa,B. td.b "Q- S,421 I i,,,. i,,,,,,,,, 1 3 5 7 9 11 13 15 17 19 21 23 25 Auditory Monitoring Day1 Day2 Day3 I1. E o Z -11111,' ~I 13 ~Be e m.g,i.[3 200 100 0-100..-v'-".DHE}.B,ID" B 3" D. id,4~.e). B.(3. El" [:] ~ 1~ m J 1 3 a i i * i * a a i i i 5 7 9 11 13 15 17 19 21 23 25 Session 1 3 I i l I l I l l I I I I I I I 5 7 9 11 13 15 17 19 21 23 25 Session Figure 3. Mean points in each task by session for the young and old adults in Study 1. two age groups. 1 Notice that there were large age differences but that similar patterns of improvement were evident in both age groups. An Age x Practice analysis of variance (ANOVA) was conducted on these data after first grouping the sessions into 5-session blocks in order to increase reliability. The ANOVA revealed significant effects of age and practice (both Fs > 113), but no Age x Practice interaction, that is, F(4, 280) = 2.62. Figure 3 illustrates the mean points on each of the four tasks for the two age groups. All main effects of age and block were significant (Fs > 14), and all Age x Block interactions were significant (Fs > 4.5) except for that on the memory task, that is, F(4, 280) = 2.69. The lack of an interaction in the memory task likely occurred because both groups were very close to the maximum score (i.e., 300). Figures 2 and 3 indicate that there were large age differences in total score and in the scores for every component. The Age X Block interaction was not significant for total score because the greater improvement for the younger adults in the arithmetic task was apparently offset by greater improvement for the older adults in the visual monitoring and auditory monitoring tasks. Predictors of performance. We conducted a series of regression analyses to identify the determinants of performance at each stage of 1 Because of the large number of statistical comparisons, an alpha level of.01 was used for all significance tests.

312 SALTHOUSE, HAMBRICK, LUKAS, AND DELL practice. First, composite measures were formed for use as predictors in the regression equations. We created a composite speed index by subtracting the average of the z scores for the Digit Digit reaction time and Digit Symbol reaction time measures (r =.74) from the average of the z scores for the Letter Comparison and Pattern Comparison (r =.68) measures. Because higher reaction times represent slower performance and higher comparison scores represent faster performance, the effect of this subtraction was such that it created a composite speed index in which higher scores indicated faster speed. A composite vocabulary score was created from the average of the synonym and antonym vocabulary scores (correlation =.66), with higher scores representing higher vocabulary levels. A mouse control measure was derived from the average time (in seconds) needed to complete the last three trials on the mouse maze. Higher scores on this measure reflect poorer or slower performance. We then used simultaneous regression equations to predict the number of points in each 5-session block from the age, vocabulary, speed, and mouse control measures (see Table 2). The increase in the intercept across successive blocks for many of the criterion measures indicates that factors other than the predictors in the regression equations are responsible for some practice-related improvement. That is, if only the relation of the predictors to the scores were to change, then the intercept would remain constant and the regression coefficients would vary. However, there were relatively few systematic changes across practice in the strength of the predictors. Mouse control ability appeared somewhat less important in later sessions (primarily in the visual monitoring task), but there were few other shifts in the predictive relations. Because significant independent effects of age were evident only on arithmetic and auditory monitoring, most of the age effects on the memory and visual monitoring measures can be inferred to have been mediated through other variables. Independent age-related influences. The next series of analyses (see Table 3) examined where in the sequence of successive blocks independent age-related influences occurred. The standardized regression coefficients are in the first two rows followed by the total proportion of variance accounted for by these two variables, the total proportion of variance accounted for by age, and then the proportion of variance uniquely accounted for by age (i.e., the increment in R 2 associated with age in a hierarchical regression analysis after control of the prior variable in the sequence). Finally, the percentage of the agerelated variance that was unique (i.e., the value in row 5 divided by the value in row 4 100) is in the last row. Statistically significant unique age effects were evident on several measures after the first in the sequence (row 5), but in all cases it was a rather small percentage of the total age-related variance (row 6). Only in Block 1 of Day 2 for the memory measure was the value greater than 15%, and even then it was only 23.5%. Therefore, it can be concluded that most age-related influences in these synthetic work tasks are present within the first 30 min of performing the tasks. Detailed analyses of component task performance. Measures of each task were next examined to determine the source of the age and practice effects in more detail. That is, which aspects of the tasks contribute to improvement with practice, and to the existence of age differences? Results of Age Practice (five blocks with five sessions each) ANOVAs of specific measures from each component task, are summarized in Table 4. In the memory task, significant age differences were evident in the accuracy measure and in the number of list retrievals. There was a significant increase in accuracy with practice, but it was no greater for young adults than for older adults. The older adults were also much more variable than the young adults in the later blocks, but this may simply reflect the fact that the young adults were near the measurement ceiling on the accuracy measure and near the measurement floor with the number of list retrievals measure. There was a large increase with practice in the number of arithmetic problems attempted for the younger adults but much less of an increase for the older adults. Some older adults did not attempt the arithmetic task, but for those who did, accuracy was significantly lower than for the younger adults. Efficiency also improved with practice and was reflected in the shift toward fewer operations (i.e., sum of plus and minus operations) per problem in later sessions. The absolute shift in the number of operations per

SYNTHETIC WORK 313 Table 2 Regression Analyses of Total and Component Scores Across All Blocks in Study I Variable 1 2A 2B 3A 3B Total Inmrcept 632 793 873 900 880 Age -6.7* -7.4* -6.2* -7.4* -6.4* Vocabulary 7.0 10.0" 5.1 7.0* 5.9 Speed 55.5* 32.3* 31.7" 34,0* 38.1" Mouse -4.0-5.7* -3.9* -2.2 -.02 R 2.67*.85*.84*.88*.81" Memory Intercept 245 266 282 277 279 Age -0.5-0.5-0.2-0.4-0.3 Vocabulary 1.9 2.6* 1.3 1.9 1.3 Speed 9.0 5.7 4.1 2.0 2.5 Mouse 0.2-0.5-0. I -0.4-0.1 R 2.18".33*.18".21".15 Arithmetic Intercept 156 234 294 325 325 Age -2.1" -3.5* -3.7* -4.6* -4.3* Vocabulary 0.9 3.2 1.7 3.9 3.4 Speed 11.8" 10.3 9.9 11.4 11.2 Mouse -0.2-0.3-0.1-0.6-0.1 R 2.74*.80*.73*.79*.73* Visual monitoring Intercept 161 155 164 170 144 Age - 1.3-0.1 0.5-0.1 0.3 Vocabulary 1.1 1.8-0.2-0.4 0.4 Speed 23.8 11.9 10.3 9.5 11.3 Mouse -4.4-4.8* - 3.0* - 1.2-1.3 R 2.35*.53*.47*.34*.26* Auditory monitoring Intercept 69 137 133 129 132 Age -2.8* -3.3* -2.8* -2.4* -2.1" Vocabulary 3.1 2.5 2.4 1.6 0.7 Speed 10.9 4.5 7.5 11.1 13.1 Mouse 0.4-0.1-0.0-0.0 0.1 R 2.72*.75*.64*.68*.60* Note. All blocks contained five sessions. *p <.ol.

314 SALTHOUSE, HAMBRICK, LUKAS, AND DELL problem was greater for the older adults. However, it is important to note that the average for the older adults in the last block was nearly the same as the average in Block 1 for the younger adults. Younger adults appeared to have increased their use of minus signs relative to plus signs with practice, but this did not occur for the older adults, and the overall practice effect on the measure of the ratio of plus-to-minus operations was not statistically significant. In the visual monitoring task, there were more lapses in the early blocks for the older adults. However, there were no significant age or practice differences in the average reset distance. Older adults were less accurate than young adults across all blocks in the auditory monitoring task, although the absolute amount of practice-related improvement was greater for the older adults. There was a large age difference in hit rate and a smaller but still significant difference in the false alarm rate. Prediction of final level of performance. In the last analysis, we examined how much of the age-related variance in the final level of performance (i.e., total number of points) was independent of the index of speed assessed from the reaction time and perceptual speed tasks administered at the beginning of the study. The speed variable was selected because it was a significant predictor across all blocks (cf. Table 2), and past research has indicated that this variable shares a large percentage of the age-related variance with many cognitive measures (e.g., Salthouse, 1993, 1994, 1995; 1996; Salthouse et al., 1995). The R 2 associated with age in the regression equation for Block 3B overall score with age as the only predictor was.739, but the increment in R 2 associated with age after control of the speed measure was.206. Therefore, it can be inferred that only 27.9% (i.e.,.206 -.739) of the total age-related variance in the last block of trials was independent of the speed with which very simple cognitive operations could be performed before beginning the synthetic work activity. Discussion The results of Study 1 indicate that there are sizable initial age differences in synthetic work performance and that they are largely maintained across approximately 2 hr of practice. Although the age differences in overall score remained fairly constant across practice, the nature of the practice-related improvement differed across the two groups, with the younger adults improving by increasing the number of arithmetic problems attempted and solved correctly and the older adults improving by reducing the number of lapses in visual monitoring and by increasing detection of targets in auditory monitoring. What was responsible for the age differences in the initial session and for the differences in the nature and magnitude of improvement across sessions? The large speed relations suggest that how quickly many processing operations can be executed contributes to the age differences at all stages of practice. In fact, the discovery that over 70% of the age-related variance in the total points measure in the last block was shared with measures of perceptual comparison and reaction time speed indicates that simple processing efficiency may be a major determinant of the age-related differences in complex activities. This has also been found to be true in simpler activities (e.g., Salthouse, 1993, 1994, 1995, 1996; Salthouse et al., 1995). However, other factors could also have contributed to the observed age differences. For example, mouse experience, auditory or visual sensitivity, or general experience with computers and related activities may all have been important factors affecting performance. There could also be age-related declines in time-sharing ability because several older participants reported that they simply ignored some of the tasks, usually the arithmetic and auditory monitoring tasks, because they felt that they could not cope with them simultaneously. This interpretation is consistent with earlier reports of age differences in divided attention situations (e.g., Kramer, Larish, & Strayer, 1995; Salthouse et al., 1995). Study 2 Two approaches can be used when a potential productivity problem is identified in a work situation. One is to introduce training to attempt to improve the proficiency of the poorer performing individuals. The second is to try to redesign

SYNTHETIC WORK 315 Table 3 Standardized Regression Coefficients and Estimates of Proportions of Variance Across All Blocks in Study I Variable 1 2A 2B 3A 3B Total Age -.76* -.41" -.26* -.25*.11 Priorvariab~.59*.71".74*.98* R 2 Total.58*.89*.89*.94*.96* Age.58*.74*.76*.81".74* Unique age.58*.07*.02*.02*.00 Percentage ofage R2unique 100.0 9.5 2.6 2.5 0.0 Memory Age -.31 * -.22 -.02 -.07 -.02 Prior variable.61".77*.83*.83* R 2 Total.10".46*.58*.72*.70* Age.10".17".09.10".08 Unique age.10".04.00.00.00 Percentage of age R 2 unique 100.0 23.5 0.0 0.0 0.0 Arithmetic Age -.84* -.24* -.09 -.24* -.09 Prior variable.75*.86*.75* 1.05" R 2 Total.70*.93*.89*.92*.96* Age.70*.76*.71".76*.69* Unique age.70*.02*.00.02*.00 Percentage ofage R2unique 100.0 2.6 0.0 2.6 0.0 Visual monitoring Age -.50* -.15 Prior variable.65* R 2 Total.25*.54* Age.25*.23* Unique age.25*.02 Percentage ofager 2 unique 100.0 8.7 -.02 -.20*.14.80*.73*.92*.62*.68*.74*.13".21".08.00.03*.01 0.0 14.3 12.5 Auditory monitoring Age -.82" -.38" Prior variable.58* R 2 Total.67*.84* Age.67*.73* Unique age.67*.05* Percentage of age R 2 unique 100.0 6.8 Note. All blocks contained five sessions. *p <.01..10 --.16".10 1.03".82* 1.05".89*.91".95*.61".65*.56*.00.01".00 0.0 1.5 0.0

316 SALTHOUSE, HAMBRICK, LUKAS, AND DELL Table 4 Component Measures and Other Statistical Data Across All Blocks for Each Age Group in Study I 1 2A 2B 3A 3B Variable M SD M SD M SD M SD M SD Memory Correct (%) Age (A): F(1, 70) = 6.75 Practice (P): F(4, 280) = 25.18 Younger 90.3 12.1 94.2 6.1 97.1 3.1 Older 81.9 10.9 84.7 11.3 91.5 9.9 t(70) 3.05" 4.51" 3.31" Number of list retrievals Age (A): F(1, 70) = 8.44* Practice (P): F(4, 280) = 2.79 Younger 0.14 0.23 0.13 0.23 0.04 0.09 Older 1.48 3.01 1.28 2.96 0.78 1.94 t(70) -2.79* -2.41-2.38 Arithmetic Number of problems attempted Age (A): F(1, 51) = 161.49" Practice (P): F(4, 204) = 159.70" Younger 18.65 6.36 24.81 7.30 30.35 9.29 Older 3.28 5.00 4.95 6.00 6.88 7.57 t(70) 11.26" 12.46" 11.61" Correct (O~)a Age (A): F(I, 51) = 57.83* Practice (P): F(4, 204) = 15.50" Younger 86.5 6.0 91.5 4.5 91.3 6.8 Older 58. I 25. I 65.4 25.0 73.0 21.0 t(df) 6.69* (53) 6.35* (60) 5.02* (59) Number of operations per problem a Age (A): F(I, 56) = 7.24* Practice (P): F(4, 244) = 18.17" Younger 11.36 2.29 10.79 3.10 10.42 1.91 Older 15.49 9.10 12.51 4.78 11.38 3.86 t(djo -2.71" (60) -1.74 (61) -1.31 (59) Ratio of plus to negative operations a Age (A): F(1, 50) = 0.12 Practice (P): F(4, 200) = 2.72 Younger 7.41 5.72 7.23 9.40 6.61 6.29 Older 4.99 5.01 11.12 24.68 16.48 37.91 t(djo 1.47 (53) -0.87 (57) - 1.59 (55) Reset distance Age (A): F(1, 70) = 0.04 Younger 69.9 Older 70.6 t(70) -0.14 Visual monitoring Practice (P): F(4, 280) = 2.32 21.3 66.9 20.1 66.6 17.2 69.0 17.0 68.7-0.46-0.46 19.7 18.0 A P: F(4, 280) = 2.23 95.9 4.6 94.8 5.4 89.2 9.5 89.9 10.1 3.91" 2.60 A x P: F(4, 280) = 188 0.06 0.12 0.12 0.20 0.85 1.77 0.76 1.63-2.76* -2.45 A x P: F(4, 204) = 47.89 32.30 8.46 33.06 9.62 7.08 7.22 7.49 7.85 13.47" 12.21" A P: F(4, 204) = 3.40 93.5 4.1 92.3 4.9 77.0 13.8 80.3 12.0 6.97* (58) 5.50* (58) A P: F(4, 244) = 4.20* 9.46 1.61 9.88 1.74 10.65 4.15 11.24 3.28-1.59 (59) -2.10 (58) A x P: F(4, 200) = 1.51 4.30 5.12 3.76 4.29 10.84 27.28 12.09 24.39-1.46 (54) -2.08 (55) A P: F(4, 280) = 0.57 67.5 19.5 67.6 20.5 67.5 18.9 66.9 21.3-0.01 0.13

Table 4 (continued) SYNTHETIC WORK 317 1 2A 2B 3A 3B Variable M SD M SD M SD M SD M SD Visual monitoring (continued) Number of lapses Age (A): F(1, 70) = 6.38 Practice (P): F(4, 280) = 6.41" A P: F(4, 280) = 18.71 Younger 0.18 0.21 0.23 0.36 0.29 0.60 0.28 0.49 0.42 0.67 Older 0.99 0.83 0.64 0.58 0.41 0.56 0.39 0.53 0.33 0.44 t(70) -5.89* -3.66* -0.83-0.92 0.64 Auditory monitoring Hit rate Age (A): F(1, 70) = 122.59" Practice (P): F(4, 280) = 32.70* A P: F(4, 280) = 3.42* Younger 77.9 16.4 93.2 8.2 94.8 6.2 95.2 5.4 96.7 4.1 Older 24.7 23.7 29.4 30.1 39.4 38.0 42.5 35.0 46.4 37.9 t(70) 11.22" 12.72" 8.97* 9.29* 8.23* False alarm rate Age (A): F(1, 70) = 12.84" Practice (P): F(4, 280) = 14.34" A P: F(4, 280) = 2.55 Younger 2.2 1.8 1.1 1.5 0.3 0.5 0.6 0.7 0.5 0.7 Older 5.7 7.6 3.4 5.8 2.4 3.9 2.1 2.7 1.2 2.0 t(70) - 2.74* - 2.41-3.37 * - 3.35" - 2.03 athe df differs because some older adults did not complete any arithmetic problems in this session; hence, the df appears in parentheses to the right of the respective t value. *p <.01. the job situation to minimize sources of difficulty. A version of the second approach was adopted in this study in part because there was little evidence of a shift in the size of the age differences in overall score across over 2 hr of task performance time in Study 1. Suitable training may eventually eliminate the age differences, but differential training may not be needed if the differences can be eliminated with job redesign. Furthermore, if the source of the age differences could be identified, then eventually it might be possible to design interventions that eliminate the initial differences and again make training unnecessary. Job redesigns typically begin by identifying aspects of the situation creating problems followed by attempts to modify those aspects to reduce the problems. In the previous study, many older adults performed at low levels on the arithmetic and auditory monitoring tasks. Therefore, we attempted to redesign parameters of the synthetic work situation to make these tasks more compatible with the capabilities of older adults. This was accomplished by deemphasizing the arithmetic task and increasing the distinctiveness of the tones in the auditory monitoring task. Major changes from Study 1 were as follows: First, the participants were recruited to form a continuous age distribution, but to the best of our knowledge, none was currently enrolled as a full-time student. Study 1 age comparisons are open to question because all of the younger adults in Study 1 were students at a technological university, and most had considerable previous experience with computers and mouse control devices. Second, we transferred participants to different priorities at the middle of Day 3 (Block 3B) to examine adaptability to new conditions. Although there was little independent age-related variance from Block 3A to Block 3B in Study 1, unique age-related variance may have emerged if there was a change in the conditions. To assess awareness of the altered task conditions, we did not inform the participants of this change, but instead we asked them whether they had noticed it when they had completed the final session. Third, different initial task conditions (payoffs

318 SALTHOUSE, HAMBRICK, LUKAS, AND DELL and event frequencies) were used to examine generalizability of the practice-related improvements. Fourth, the auditory discrimination was made easier by increasing the difference between the signal and nonsignal tones. Fifth, a test of near-vision acuity was administered at the end of Day 2. Finally a postexperimental questionnaire was administered at the end of Day 3. Items in the questionnaire referred to the amount of prior experience with computers, a computer mouse, and video games, as well as self-appraisals of level of motivation and of satisfaction with the amount of improvement in performance. We also requested information about how performance improved and any factors that might have limited further improvement. Method Participants. Descriptive characteristics of the participants are summarized in Table 5. The young adults in this study were slower than the college-age adults in Study 1 in the perceptual comparison, reaction time, and mouse control tasks, but they had higher scores in the vocabulary tests. The older adults in the two studies were fairly similar. It is particularly noteworthy that there were strong negative age relations on the visual acuity and experience measures. This is consistent with the interpretation that factors related to sensory ability and to relevant experience contribute to the age differences in synthetic work performance. Procedure. Most aspects of the procedure were identical to those in Study 1, and thus only the differences between the two studies are described below. The vision test was conducted with a card 2 containing columns of 2-digit numbers and Landolt C stimuli at 10 different font sizes. Participants read numbers or stated the orientation of the gap in the C while holding the card at a distance of approximately 30.0 cm and wearing necessary corrective lenses if needed. The fonts at this distance correspond to Snellen visual acuity ratios of 0.1 to 1.0 in steps of 0.1. Participants began with the largest font and continued to the smaller fonts until they were unsuccessful on two or more items in a set of a given size. The smallest font size at which they made fewer than two errors served as the measure of visual acuity. We made separate assessments with the number and Landolt C material with the left eye while the fight eye was covered and with the right eye while the left eye was covered. The questionnaire administered at the end of Day 3 contained nine items. The first three were ratings of the amount of experience on a scale of 1 = very much to 5 = none with computers, use of a computer mouse and playing video games. The level of interest or motivation in the study and the degree of satisfaction with the amount of improvement were also rated on a scale with 1 = high and 5 = low. Finally, there were open-ended questions that asked how individuals felt they had improved across practice, what factors limited his or her performance, whether any difference was noticed in the conditions in the last day, and whether the synthetic work activity reminded him or her of any particular situation. The parameters in the synthetic work program were selected to emphasize the memory task more than the arithmetic task. This was accomplished by reducing the costs and payoffs in the arithmetic task to 5 points for a correct response and -5 points for an incorrect response, increasing the frequency of the memory probes to one every 5 s, and changing the payoffs in the memory task to 20 points for correct responses and -20 points for an error or a miss, with the cost for a list retrieval remaining at -10 points. Tones in the auditory monitoring task were changed to 523 Hz for the low (nontarget) tone and 2092 Hz for the high (target) tone, from the values of 1046 and 1319 Hz, respectively, in Study 1. For the last five sessions on Day 3, the following changes were introduced: the memory probes occurred every 10 s instead of every 5 s; the payoffs in the memory task were changed from 20 and -20 to 10 and -10 for correct and incorrect responses, respectively; and the payoffs for correct and incorrect responses in the arithmetic task were changed from 5 and - 5 to 10 and - 10, respectively. Therefore, the maximum score in the memory task for Blocks 1 to 3A was 1,200 (i.e., 20 points for each of 60 probes), and for Block 3B it was 300 (i.e., 10 points for each of 30 probes). The initial instructions to the participants were also modified slightly to enhance comprehension, and additional practice was provided on the 2 The vision chart was Scalae Typographicae Birkhausen (Birkhauser Verlag; Basel, Switzerland).

Table 5 Characteristics of Participants in Study 2 Age group 18-39 40-59 60-80 (n = 21) a (n = 30) b (n = 26) c Variable M SD M SD M SD F(2, 76) Age 28.9 Education (years) 15.9 Health rating 1 1.9 2 2.2 Health satisfaction 2.0 Health-related limitations 1.3 Cardiovascular surgery 0.0 Hypertension medications 0.0 Head injury 0.0 Neurological treatment 0.1 Comparison Letter Pattern Vocabulary Synonym Antonym Reaction time (ms) Digit Digit Digit Symbol Mouse time (s) First 3 trials Last 3 trials Errors First 3 trials Last 3 trials Vision (Snellen ratio) d Left Eye Numbers Landolt C Right Eye Numbers Landolt C Experience level Computers Mouse Video games Level of interest Satisfaction with improvement Noticed Block 3B change 10.7 18.3 Demographicd~a 6.3 49.4 5.3 69.2 5.2 2.0 16.1 2.9 14.8 2.1 2.00 -.18 Health 0.7 2.1 1.0 2.3 0.9 0.95.17 0.9 2.4 0.9 2.6 0.9 1.06.19 0.7 2.4 0.8 2.4 0.9 1.50.17 0.6 1.8 0.9 1.8 0.8 3.10.24 0.1 0.3 0.1 0.3 1.05.16 0.1 0.3 0.6 0.5 19.94".53* 0.1 0.3 0.04 0.2 1.32.01 0.4 0.1 0.3 0.04 0.2 0.91 -.10 Other comparison measures 2.2 9.8 2.4 8.8 2.3 3.68 -.36* 2.8 16.6 3.4 14.1 2.7 11.74" -.54* 6.4 2.5 7.2 2.6 7.7 2.1 1.66.22 6.1 3.4 6.4 3.2 6.8 2.7 0.35.07 681 239 730 162 804 204 2.28.35* 1,311 281 1,467 246 1,844 439 16.56".61" 13.65 9.16 16.03 13.21 33.82 26.17 9.51".49* 8.07 4.76 9.95 8.39 18.99 12.60 9.74*.47* 2.3 1.8 2.1 2.3 6.4 7.0 7.78*.43* 0.8 0.9 1.5 2.0 2.1 3.3 1.87.29 0.65 0.23 0.44 0.23 0.37 0.18 10.39" -.50* 0.84 0.21 0.44 0.20 0.39 0.18 36.59* -.64* 0.60 0.22 0.42 0.18 0.36 0.18 9.53* -.47* 0.78 0.22 0.45 0.23 0.40 0.18 22.44* -.55" 2.3 1.3 2.4 1.2 3.8 1.1 11.36".45* 2.5 1.5 3.0 1.3 4.2 1.2 9.90*.49* 3.1 1.3 3.9 1.2 4.5 1.2 7.08*.45* 1.7 1.1 1.5 0.8 1.4 0.8 0.80 -.19 2.4 1.1 2.6 1.2 2.7 1.4 0.29.08 0.91 0.30 0.86 0.35 0.54 0.51 6.28* -.32* Note. Health rating, health satisfaction, and health-related limitations were determined by a 5-point scale where lower numbers indicate better health, and responses to the cardiovascular surgery, hypertension medications, head injury, neurological treatment, and noticed Block 3B change were yes or no; thus the means correspond to the proportion of individuals reporting a positive response. a52.4% were women, b60.0% were women, c46.2% were women, dhigher scores indicate better vision. *p <.01.

320 SALTHOUSE, HAMBRICK, LUKAS, AND DELL component tasks in isolation (i.e., two blocks of 1 min each instead of one single block). The probability of the target signals in the initial instruction sessions on the auditory monitoring task was also increased to.5 from.2 in order to maximize the opportunity to experience both tone types. Results and Discussion Figure 4 displays the mean total points in the three age groups across the 25 sessions. The age and practice (five-session blocks) main effects (Fs > 21.67) and the Age X Practice interaction, F(8, 296) = 5.12, were all significant in an ANOVA on the total points. The mean points in each component task as a function of practice are illustrated in Figure 5. Age X Practice (5-session blocks) ANOVAs revealed that all main effects (Fs > 7.00) and interactions (Fs > 3.29) were significant for all of the component scores. Age X Practice ANOVAs were also conducted 18-39...,...0.--- 40-59 60-80 1200,'g,~" ~2" e. u 0 n m I" 9OO 6OO 300 i I I : -i I i i i/-"..." "" i o..o 0 Day I Day 2 Day 3 12345 6 7 S 9 10111213141516171819202122232425 Session Figure 4. Means and standard errors of the total number of points by session for the three age groups in Study 2.

SYNTHETIC WORK 321 with only Blocks 3A and 3B, before and after the switch in the payoffs in the memory and arithmetic tasks and in the frequency of the memory probes. Interactions of age and block were significant on the arithmetic; F(2, 74) = 15.45, and visual monitoring, F(2, 74) = 9.71, scores. Young adults increased their scores more than older adults after the shift in the arithmetic task, but older adults had more of an increase than young adults in the scores on the visual monitoring task. This latter effect was surprising and may have occurred because older adults had more time to attend to visual monitoring when the frequency of memory events was reduced. Young and middle-aged adults did not improve in visual monitoring because they were already performing near the maximum level. Responses to the questionnaire item on whether any differences were noticed on Day 3 were coded as yes or no (see Table 5). Table 5 shows that although 91% of the younger adults and 86% of the middle-aged adults reported noticing a change in the conditions, only 54% of the older adults reported a change. Therefore, the smaller increase in the number of attempted arithmetic problems on the part of the older adults may have been partially attributable to many of them not having noticed the change in payoffs. Predictors of performance. Composite measures of vocabulary, speed, and mouse control ability were formed as in Study 1. We also created composite measures of vision because the vision measures were all moderately correlated (r =.52 to.80) with one another. The experience measures were also moderately correlated with one another (r =.32 to.68). Thus an experience composite was created by averaging the z scores on the computer experience, mouse experience, and video game experience items. Table 6 shows results from the regression!day1 Memory Day2 Day3 200 Day I Arithmetic Day 2 Day 3 150 ~ 900 '.: ~oo!...".o 30O E :3 7 0-3OO 100 50 7" 4D. B~B'43.~...~.~ k.,~.,ok...,.~. ~ kl 2OO ~ a i i I i * * i * i J = 1 3 5 7 9 11 13 15 17 19 21 23 25 0 ~d' Visual Monitoring ~...,'~'*'*' "~ "5-200. ~.".." i, i i i m, 1 3 5 7 9 11 13 15 17 19 21 23 25 Auditory Monitoring Day1 Day 2 100 50 0 f '~' ~" B'P" m'~' 'u"~ t~ Day3 Z." ~,00 -/ ; Day2 ~3-100 i I! I I I * * * * = j * * * * * * * * * -1000 1 3 5 7 9 11 13 15 17 19 21 23 25 Session -150 1 3 5 7,~.~.- '~.'.~..~...y" #.' ",k,, i,ll,, = i i t i t a, 9 11 13 15 17 19 21 23 25 Session Figure 5. Mean points in each task by session for the three age groups in Study 2.

322 SALTHOUSE, HAMBRICK, LUKAS, AND DELL Table 6 Regression Analyses of Total and Component Scores Across All Blocks in Study 2 Variable 1 2A 2B 3A 3B Total Intercept 1,493 1,459 1,183 1,273 681 Age -9.6-6.1-0.5-1.2-3.9 Vocabulary 3.4 3.0 3.2 3.4 7.5 Speed 75.5 100.6" 118.4" 90.9* 43.8* Mouse -9.5-11.4-11.1-11.4" -7.5* Vision - 1.6 243.3 330.5 290.2 66.5 Experience -99.1-51.5-21.5-23.9-6.8 R 2.46*.47*.48*.52*.68* Memory Intercept 1,087 1,063 902 969 251 Age -3.6-0.9 3.1 1.4-0.0 Vocabulary 2.7 0.9-2.7 0.4 0.6 Speed 28.1 37.6 64.0* 36.3* 5.2 Mouse -2.0-3.9-3.3-2.9-0.3 Vision 10.2 111.5 207.1 162.8 20.2 Experience -22.0-3.2-0.2-0.1 5.4 R 2.27*.26*.29*.25*.06 Arithmetic Intercept 9 30 46 49 165 Age -0.3-0.5-0.7-0.7-2.1" Vocabulary 0.7 1.2" 1.5" 1.5 4.2* Speed 1.6 3.5 3.4 5.4 11.9 Mouse -0.1-0.3-0.3-0.3-0.7 Vision 16.5 9.5 7.5 3.7-3.5 Experience -0.4-2.1-2.5-2.9-9.1 R 2.41".47*.45*.46*.53* Visual monitoring Intercept 381 241 53 95 105 Age -4.7-2.3-0.3 0.8 0.8 Vocabulary -5.9-4.1-0.1-3.2-1.1 Speed 36.9 57.2 49.9* 46.0* 21.0" Mouse -6.5-4.8-5.9-5.3* -3.5* Vision - 32.0 123.0 142.8 130.5 56.8 Experience - 56.2-30.7 3.1-11.5 1.7 R 2.44*.39*.38*.41" ~40" Auditory monitoring Intercept 17 125 182 160 161 Age - 1.1-2.4* -2.6* -2.7* -2.5* Vocabulary 5.9" 5.0* 4.6* 4.5* 4.0* Speed 8.8 2.2 1.2 3.3 5.7

Table 6 (continued) SYNTHETIC WORK 323 Variable 1 2A 2B 3A 3B Auditory monitoring (continued) Mouse -0.9-2.3-1.6-2.9* -3.0* Vision 3.7-2.8-26.9-6.7-6.9 Experience -20.6-15.5-22.0-9.5-4.9 R E.41".49*.51".54*.62* Note. All blocks contained five sessions. *p <.01. analyses with the set of six predictors. It is noteworthy that there were no significant independent effects of the composite experience or vision measures on any session. As in Study 1, there were significant speed and mouse effects on later sessions of the memory, visual monitoring, and auditory monitoring tasks. The significant relation of vocabulary on auditory monitoring performance was surprising, but it may reflect an influence of general intellectual ability on timesharing capabilities. There were no significant independent age effects except in the posttransfer block of arithmetic and in all blocks except Block 1 of auditory monitoring. There was a decrease from Blocks 1 to 3A in the intercepts for some of the measures, but there was relatively little systematic increase in the regression coefficients for any of the predictors. One exception was in the transfer (Block 3A vs. Block 3B) comparison, where speed seems to have been less important in the posttransfer block, particularly in the memory task where performance was at ceiling for all groups (cf. Figure 5). Independent age-related influences. Table 7 contains the results of the regression analyses designed to identify independent age-related influences on the total scores and the scores in each component task. As in Study 1, most of the unique age-related effects were evident early in practice, with the highest value prior to the transfer block of only 15.2%. Significant independent age-related effects occurred in the transfer block, particularly for the arithmetic and auditory monitoring tasks. We expected an increase in emphasis on the arithmetic task when payoffs were shifted, but it is noteworthy that the shift was smaller with increased age. The reason for the emergence of independent age-related vari- ance in later blocks of the auditory monitoring task was not obvious. Detailed analyses of component task performance. Table 8 shows that significant practice effects occurred on memory percentage correct, arithmetic number of problems attempted, percentage correct, and number of operations per problem, visual monitoring number of lapses, and auditory monitoring hit rate and false alarm rate. Significant age effects were evident in memory percentage correct, arithmetic number of problems attempted, and hit rate in the auditory monitoring task. Finally, significant interactions of age and practice were evident only on the number of problems attempted in the arithmetic task, and this occurred because the increases were smaller with greater age. The only interaction of Age x Practice in ANOVAs contrasting performance in Block 3A (before transfer) and Block 3B (after transfer) was on the number of arithmetic problems attempted, F(2, 74) = 5.50. Prediction of final level of performance. The total points on Block 3A (pretransfer) and Block 3B (posttransfer) were predicted from age and the composite speed index. The R 2 for age in the prediction of total points on Block 3A was.292, and after control of the speed measure it was.054, indicating that only 18.5% (.054 +.292) of the age-related variance in the final pretransfer performance was independent of speed. The R 2 for age in the prediction of total points in Block 3B was.430, and after control of speed it was.117, indicating that 27.2% (.117 +.430) of the age-related variance in the posttransfer performance was independent of speed. Although there is clearly much more than speed involved in the age differences in these tasks, it is nevertheless noteworthy that speed accounts for such a large

324 SALTHOUSE, HAMBRICK, LUKAS, AND DELL Table 7 Standardized Regression Coefficients and Estimates of Proportions of Variance Across All Blocks in Study 2 Variable 1 2A 2B 3A 3B Tot~ Age -.60* -.15.06 -.10 -.29* Prior variable R 2.71".98*.87*.68* Total.35*.66*.89*.85*.76* Age.35*.33*.26*.29*.43* Unique age.35*.02.00.01.06* Percentage ofage R2unique 100.0 6.1 0.0 3.4 14.0 Memory Age -.46* Prior variable R 2 Total.21" Age.21" Unique age.21" Percentage ofager 2 unique 100.0 -.01.08 -.07.10.83*.86*.82*.75*.70*.69*.71".53*.16".07.08.01.00.01.01.01 0.0 14.3 12.5 100.0 Arithmetic Age -.53" -. 11 Prior variable R 2.85" Total.28*.83* Age.28*.31" Unique age.28*.01 Percentage ofager 2 unique 100.0 3.2 -.07 -.05 -. 12".90*.90*.89*.88*.87*.92*.33*.32*.38*.00.00.01" 0.0 0.0 2.6 Visu~monitofing Age -.58* -.19 Prior variable.58* R 2 Tot~.34*.50* Age.34*.28* Unique age.34*.03 Percentage ofager 2 unique 100.0 10.7.07 -.08.03.97*.83*.86*.88*.76*.71".20*.20*.13".00.00.00 0.0 0.0 0.0 Auditory monitofing Age -.43* -.23* Prior variable.79* R 2 Tot~.18".84* Age.18".33* Unique age.18".05* Percentage ofage R2unique 100.0 15.2 Note. All blocks contained five sessions. *p <.01. -.09 -.08 -.18".88*.90*.78*.88*.90*.8 I*.36*.38*.43*.01.00.02* 2.8 0.0 4.7 proportion of the age-related variance after nearly 2 hr of practice. For purposes of comparison, we carried out similar analyses with the experience and vision composites as predictors. The percentages of unique age-related variance after control of the experience composite variable were 42.8% in Block 3A and 47.9% in Block 3B. The values