Distraction and Speed 1 Running head: DISTRACTION AND SPEED Distraction as a Determinant of Processing Speed

Processing speed is often described as a fundamental resource determining individual (e.g., I.Q.) and group (e.g., developmental) differences in cognition. However, most tests that measure speed present many items on a single page. Because many groups with slowed responding are also distractible, we compared younger and older adults on high(i.e., standard) versus low-distraction versions of two classic speed tasks. Reducing distraction improved the performance of older adults while having little or no effect on younger adults, suggesting that the ability to limit attentional access to task-relevant information can affect performance on tests designed to measure processing speed. Distraction and Speed 3 Distraction and Processing Speed The idea that “faster is better” is powerful in cars, computing, and cognitive psychology. Group differences, especially age differences, are often ascribed to the better-performing group’s faster processing. We report two studies that assessed the contribution of an attentional-perceptual variable, visual distraction, in determining age differences in classic speed tasks. Many standard speed tests use items that are individually simple, but fit many such items onto a single page, resulting in a cluttered, potentially distracting display. Many groups thought to have deficits in processing speed also have difficulties regulating attention, and thus might be especially vulnerable to distraction. These groups include children, older adults, poor readers, and young adults who score less well on intelligence tests (e.g., Casey, Giedd & Thomas, 2000; Dempster & Corkill, 1999, Engle, Tuholski, Laughlin, & Conway; 1999; Fry & Hale, 1996; Gernsbacher & Faust, 1991; Hasher & Zacks, 1988; Kail,1993; Salthouse 1996a, 1996b). Our interest in distraction’s potential role on tests of processing speed stems from a longstanding theoretical framework emphasizing inhibitory control mechanisms that, together with goals, determine what information enters the focus of attention (e.g., Hasher & Zacks, 1988; Hasher, Zacks, & May, 1999) . Weakened inhibitory control allows distraction to impede the speeded performance of older adults in other settings, including well-practiced skills such as reading (e.g., Carlson, Hasher, Connelly, & Zacks, 1995; Duchek, Balota, & Thessing, 1998; Dywan & Murphy, 1996; Madden, 1983; Rabbitt, 1965). Distraction and Speed 4 To assess distraction’s potential role on tests of processing speed, we computerized two standard speed tasks and administered them to younger and older adults in one of two formats. The "high-distraction" format resembled the standard, paper-and-pencil versions of these tasks, with many items presented at the same time. The "low-distraction" format reduced the opportunity for distraction by presenting items individually, so that only the currently relevant item was present on the screen. Our hypothesis was simple: If vulnerability to distraction contributes to group differences in processing speed, then older adults should be faster on the “lowdistraction” versions of processing tests than on the “high-distraction” versions that resemble the standard, but distraction should make little difference to young adults.. Experiment 1 The paper and pencil versions of the Letter Comparison and Pattern Comparison tasks (Salthouse & Babcock, 1991) are widely used as measures of processing speed (e.g., Hambrick & Oswald, 2005; Salthouse, 1993). Both meet our critieria for “highdistraction”, with many items presented on a single page. We computerized the Letter Comparison task and presented it in either a highor low-distraction format. Correlations between the computerized Letter Comparison tasks and the standard paper-and-pencil version of the Pattern Comparison task were examined to ensure that the high-distraction computerized task was representative of performance on standard measures of processing speed. Method Participants. Strict exclusionary criteria helped ensure that any differences were the result of age and our distraction manipulation rather than extraneous problems with Distraction and Speed 5 vision, health, or motor functioning. Participants’ data were discarded if they (a) had health problems or were taking medications that might affect vision or motor functioning (e.g., attention deficit disorder, dyslexia, macular degeneration), b) made incorrect responses on more than one-third of the trials, or c) failed to meet criterion on either our vocabulary measure (a score of at least 13 out of 48 possible on the Extended Range Vocabulary Test (ERVT), Version 3, Educational Testing Service (ETS), 1976) or our dementia screen (a score less than 6 on the Short Blessed Test; Katzman et al., 1983). In both experiments, data from several other participants were discarded due to experimenter or computer error, or because the participant had completed one of the speed tasks in a previous session. In both studies, participants in each age group were randomly assigned to the lowor high-distraction condition (see Table 1 for demographics). As is common, older adults had more education and higher vocabulary scores than did young adults. Participants within an age group but in different distraction conditions did not differ in age, education, or vocabulary. Three hundred and twelve young adults and 239 older adults participated. After discarding data from participants who did not meet one or more of our exclusion criteria, the final sample had 146 young adults and 92 older adults in the low-, and 138 young adults and 99 older adults in the high-distraction condition. Materials and Procedure. In the paper-and-pencil version of the Letter Comparison Test, each of two pages has 21 pairs of letter strings with 3, 6, or 9 letters (e.g., RXL_____RXL) presented in a random order in a single column. Participants indicate whether the two strings are the same or different. Our computerized version Distraction and Speed 6 consisted of 48 pairs of letter strings, with three, six, or nine letters per string. In the high-distraction condition, pairs were presented in two columns of 12 pairs each. Columns were separated by .6cm vertical space. A cursor directly under the first character in the string marked the participant’s progress down the screen. The screen refreshed after the first 24 pairs. For participants in the low-distraction condition, each stimulus item was presented individually in the center of the screen. All items were presented in black text on a white background. Participants adjusted the chair to sit at a distance most comfortable to them. The font (Turbo C graphics SMALL_FONT) had characters that up to .4cm wide and .5cm high, with .15 horizontal space between characters and a 1.2cm line separating the letter strings that made up a pair. Items in the high-distraction condition were separated by .6cm vertical space. Participants were to press one key (the “z” key, covered by a red sticker) if the strings were identical and a different key (the “/” key, covered by a blue sticker) if the strings were different. Reaction time (RT) was measured as the time between the cursor’s movement to the stimulus (high-distraction) or the appearance of the stimulus (low-distraction) and the participant’s keypress. Participants first completed a health questionnaire and a practice test (30 singledigit pairs, e.g., 5_8) to familiarize them with the display and response mapping. Following practice, participants completed five additional trials using letter strings before beginning the 48 trials that constituted the Letter Comparison Test. Participants were instructed to respond as quickly as possible, but not so fast that they made mistakes. Distraction and Speed 7 Participants also completed a paper-and-pencil version of the Pattern Comparison Test (Salthouse & Babcock, 1991). Due to experimenter error, four young adults and one older adult were not given this task. Results Error and RT data from the Letter Comparison test were analyzed using 2 (Age: young, old) X 2 (Distraction: high, low) X 3 (Length: 3, 6, or 9 letters in a string) ANOVAs, followed by planned contrasts within each age group comparing the highand low-distraction conditions. Within-subjects analyses used the Huyn-Feldt sphericity correction implemented in SPSS, resulting in non-integer degrees of freedom. Error rates gradually increased as a function of string length, F(1.97, 924.56) = 243.68, p < .0001, but did not differ by group (all ps > .19; see Table 2). Within each string length, we computed the mean RT across correct trials for each participant first deleting outlying RTs that were more than 2.5 SD faster or slower than the participant’s mean. Outlying trials made up 1.5% of the total, and all patterns in the data remain the same if these outlying RTs are included. Mean RTs are shown in Figure 1. The three-way interaction was not significant, F(1.30, 610.03) = 1.91, p = .16, although Length interacted significantly with Age, F(1.30, 610.03) = 86.93, p < .001, and Distraction, F(1.30, 610.03) = 12.77, p < .001, and had an obvious main effect, F(1.30, 610.03) = 2813.61, p < .0001. Of primary importance, the Age X Distraction interaction was reliable, F(1, 471) = 8.58, p = .004. Young adults were equally fast across conditions (F(1, 471) = 1.13, p = .29, d = .15), but older adults were significantly faster in the low-distraction condition than in the high-distraction condition, F(1, 471) = 23.79, p = .0001, d = .58. Similar Distraction and Speed 8 patterns were found in separate analyses done at each string length, with the exception that at length 6, young adults showed a small benefit of reduced distraction F(1, 471), = 4.15, p = .04, d = .28. Performance on the paper-and-pencil Pattern Comparison speed task (Table 3) replicated standard findings of better performance by young adults (e.g., Salthouse, 1993; Salthouse, 1996a), F(1,466) = 463.34, p < .0001, and did not interact with group assignment (highor low-distraction) for the Letter Comparison test, F(1,466) = 2.55, p = .11. Thus the Age X Distraction interaction found for the computerized Letter Comparison task is not an artifact of subject selection problems across the groups. Correlations between Pattern Comparison and the different computerized versions of the Letter Comparison task helped to validate our manipulation. If distraction critically influences the speed of older adults, performance on the paper-and-pencil test should correlate more highly with the highthan the low-distraction version of the computerized test. A different speed test, Pattern Comparison, was chosen as the criterion task rather than a paper-and-pencil version of Letter Comparison to increase the probability that any correlations would reflect relations among speed tasks in general, rather than being idiosyncratic to Letter Comparison. For older adults, the highdistraction version of the Letter Comparison task tended to be a better predictor of performance on the paper-and-pencil task than was the low-distraction version (Table 3), although perhaps due to low power (.50) this medium-sized difference did not reach statistical significance (q = 30, Fisher’s Z = 1.76). For young adults, the lowand highdistraction versions of the computerized Letter Comparison task were equally good predictors of performance on the paper-and-pencil Pattern Comparison test. Distraction and Speed 9 The relations between the computerized tasks and the paper-and-pencil test across age groups were compared in a post-hoc analysis using structural equation models implemented in LISREL. The first model served as a conceptual null hypothesis, constraining correlations between tasks to be equal for all groups, regardless of age or distraction condition. This model did not fit the data well, Chi-square (df = 3) = 7.30, p = .06. The second model constrained the between task correlations to be equal only for the two young adult groups and low distraction older adults. It fit the data well, Chi-square (df = 2) of 0.30. A Chi-Square difference test comparing these two models yielded a significant result (df = 1, Chi-Square difference = 7.00, p < .01). The contrast between these models supports the suggestion that correlations with the paper-and-pencil test were different (higher) for older adults tested with the high-distraction version of Letter Comparison than they were for young adults or older adults tested in low distraction. Nearly identical results were found for a second paper-and-pencil task added later in data collection (Identical Pictures Test, ETS, 1976; completed by 63% of participants). (See Table 2.) Further, correlations between the two paper and pencil tasks (r = .77 for older adults, r = .48 for young adults) were in the same range as those between Letter Comparison and Pattern Comparison for older adults tested in high distraction, and for young adults overall. In other words, the variance shared between the high-distraction computerized task and the paper and pencil-and-pencil tasks was similar to that shared between the two paper-and-pencil tasks themselves. Letter Comparison string length did not systematically influence correlations with the paper-and-pencil tests. Variability (standard deviation of RT) also showed an Age X Distraction interaction, F(1,471) = 4.29, p < .05; Figure 2. Results were generally similar to those on Distraction and Speed 10 mean RT, with the following exceptions: Young adults also showed a significant effect of Distraction, F(1,471) = 21.06, p < .0001, and Length did not interact with Distraction, F(1.69, 797.52) = 2.23, p = .12. Variability did not show strong correlations with paperand-pencil test performance for any group, all r < .30. To examine whether effects on RT per se were greater than those on variability, each trial RT was transformed into a z-score based on the participant’s mean RT and standard deviation of RTs for all correct trials across string lengths (Faust, Balota, Spieler, & Ferraro, 1999). Using the same ANOVA design as for the raw RTs, the two main effects of Length and Distraction were significant, as was their interaction, Length X Distraction (F(1.74, 820.77) = 30.23, p < .0001). The main effect of Age was not significant and did not enter interactions. At length 6, means were higher in high(.14 young; .12 old) than low-distraction (.08 young, .05 old), but the reverse was true at length 9 (high distraction: .87 young, .90 old; low distraction: 1.02 young, 1.05 old). These patterns are generally consistent with our hypothesis that distraction can lead to slowed and more variable performance, and that its effects are greater on older adults. In the following experiment, we ask whether the results found for this simple two-choice task would generalize to a more complex test.