© 2013 American Psychological Association 0882-7974/13/$ 12.00 DOI: 10.1037/a0a34937
Psychology and Aging 2013, Vol. 28. No. 4, 1024-1031
Age Differences in Voluntary Task Switching Karin M. Butler and Christina Weywadt University of New Mexico Task choice processes in older (60-1- years) and younger (18-30 years) adults were compared using a voluntary task switching procedure (Arrington & Logan, 2004). To assess age-related differences in task representation maintenance, preparation times were varied across a large range of responseto-stimulus intervals (100, 500, 1,000, and 5,000 ms) and the environmental influence on task selection was varied by repeating or changing stimuli from trial to trial. Older adults switched less frequently than younger adults and this effect was the same at each RSI. Younger adults were more likely to switch tasks when the stimulus changed than when it repeated suggesting that they used a different process to determine task choices, either endogenous task selection or environmentally supported task repetitions. Older adults' task selection was unaffected by stimulus repetitions indicating that they were less flexible with the processing they used to guide task selection. These findings are consistent with previous observations of age-related increases in goal shielding, but not with age-related deficits in task goal maintenance. Robust age differences in switch costs were observed across RSIs suggesting that task reconfiguration processes are different following endogenous than exogenous task selection. Keywords: aging, task choice, multitasking, task representation, cognitive control
Flexibly moving between tasks in the absence of external cues is important for day-to-day life. As we age, optimal task switching may decline as a result of age-related changes in the ability to inhibit prepotent responses to environmental stimuli (e.g., Butler, Zacks, & Henderson, 1999; Hasher, Zacks, & May, 1999), difficulty maintaining task goals (Braver et al., 2001; Mayr, 2001), or the need for additional time to support task reconfiguration (Salthouse, 1996). In this study we evaluated age differences in task choice in an endogenous task switching environment known as the voluntary task switching (VTS) procedure while manipulating both the time available to prepare a task selection and the presence of an environmental context that supports repetition of the current task rather than task switching. Efficiently moving between tasks is impaired by deficits in maintaining contextually appropriate goals according to the goal maintenance theory of age-related differences in goal-directed behavior (Braver et al, 2001; Braver & West, 2008). Older adults display difficulties maintaining task goals across intervals as short as 5 seconds. For example, in the AX-continuous performance task (AX-CPT), individuals are shown one of four pairs of letters with a delay of 4,900 ms between them (AX, AY, BX, or BY). They are instructed to respond positively only when an "X" probe stimulus
follows an "A" stimulus. On 80% of trials that begin with an "A" stimulus, an "X" probe stimulus follows, biasing the individual to prepare to respond to the probe stimulus. When the "A" is followed by a "Y" probe, false alarms are made often. Older adults made fewer of these errors than younger adults did, though. The result suggests that older adults' goal representations (e.g., prepare to respond to the "X" probe) degrade more quickly than the representations of the younger adults do across interval as short as 4,900 ms. Age-related task maintenance deficits have been demonstrated in a prospective memory task as well. In the context of an ongoing task, if a new goal is introduced that must be maintained across a short 5-second interval (i.e., a prospective memory task), older adults will more often forget to execute that intention than younger adults will (McDaniel, Einstein, Stout, & Morgan, 2003). Age-related deficits in goal representations have also been observed in task switching environments that cued participants to perform a certain task on each trial. The costs associated with being in a task switching environment compared to a single task environment were larger for older than younger adults, particularly when both the relevant stimulus attributes and the response options for each task were the same (Mayr, 2001). Mayr concluded that from one trial to the next older adults do not maintain the previously executed or currently required task as robustly as younger adults do. When a task repetifion is required, this results in more time needed to reactivate the just executed task, as well as greater interference (slowing) when a different task is required but the stimulus or response is the same. The idea that task representations are weaker in older individuals was challenged by Lien, Ruthruff, and Kuhns (2008). They found that older individuals were as affected as younger individuals when task expectancies were violated. Across three experiments, participants used temporal and location cues to predict what
Karin M. Butler and Christina Weywadt, Psychology Department, University of New Mexico. Portions of this work were presented at the 50th Annual Psychonomic Society Conference, Boston, Massachusetts, in 2009. This research was supported by a Ouad-L Eoundation grant to Christina Weywadt. Our thanks to Melakeh Kurdi for her assistance with data collection. Correspondence concerning this article should be addressed to Karin M. Butler, MSC 03 2220; Logan Hall, Department of Psychology, 1 University of New Mexico, Albuquerque, NM 87131-0001. E-mail: kmbutler@ unm.edu 1024
AGE DIFFERENCES IN VOLUNTARY TASK SWITCHING
task to perform on each trial. Occasionally, the stimulus that was presented would not afford the expected task, though. Older adults were slowed by the violation of this expectancy more so than younger adults, suggesting that older adults maintained task information as robustly as younger adults, if not more so. The researchers suggested that the older adults used the predictable environment to establish robust task representations in the procedure that they used. In fact, older adults will rely on the environment to maintain robust task representations even when it is not necessary (Mayr & Liebscher, 2001; Spieler, Mayr, & LaGrone, 2006). After switching tasks in blocks of trials that required following the direction of task cues, older individuals were much more likely than younger adults to look at task-related, but unnecessary cues during single task performance (Spieler et al., 2006). Persisting in processing the cues slowed older adults' responses compared to responses in a single task block performed prior to the cued task switching trials. Older adults who did not have the cues available during single task performance reached the speed of baseline reaction times (RTs) within 60 trials of the single task blocks. Spieler and colleagues (2006) suggested that older individuals are less flexible in their reliance on internal and extemal sources of control. When older adults can rely on a single mode of control (internal or extemal), age-differences are reduced or eliminated, but when performance requires switching between control strategies, older adults have more difficulty than younger adults do (e.g., Ardiale & Lemaire, 2012). In order to examine task selection processes in a multitask environment, while eliminating the influence of cue encoding processes and overreliance on environmental cues to guide task selection and switching, this study examines age differences in task switching using the VTS procedure. In the VTS procedure, participants are told to choose randomly which of two tasks to perform on each trial (Arrington & Logan, 2005; Arrington, Rieman, & Weaver, in press; Mayr & Bell, 2006). Switch probabilities are typically less than 50% reflecting a repetition bias. This bias has been attributed to the task representation used on one trial persisting and infiuencing the task selection on the next trial. Probability of switching increases as the time available to make a choice increases, suggesting that probability of switching is sensitive to endogenous control processes (Arrington & Logan, 2004). Successful random task choice requires overcoming the persisting representation from previous trials and/or maintaining a representation of the sequence of previously executed tasks. Task selection processes in younger adults can be influenced by the environment through bottom-up processing in the VTS procedure. Younger adults are more likely to repeat a task when the stimulus repeats from the previous trial compared to when it changes (e.g., Mayr & Bell, 2006; Demanet, Verbruggen, Liefooghe, & Vandierendonck, 2010). To date, only one study has investigated age differences in a VTS environment. Terry and Sliwinski (2012) showed that older individuals were less likely to switch tasks than younger adults. This difference was not observed when the groups were simply asked to generate a series of random "coin-nips," however, supporting the view that older and younger adults have a similar representation of randomness. In addition, older adults' switch probabilities were positively correlated with their task repetition RTs to a greater extent than those of younger adults. This pattem
1025
suggested that older adults adopted an endogenous task selection strategy of maintaining the strength of the current task to support task repetition performance at the cost of lower switch rates, a phenomenon they referred to as a within-set mode of processing. In this view, younger adults were more likely to use a task updating strategy that led to more frequent task reconfiguration to support task switching. The Terry and Sliwinski (2012) study left open the possibility that age differences in the task selection processes may be due to the influence of bottom-up processing on task selection rather than differences in the endogenous processes. Older adult task selection may have been more infiuenced by the environment than the task selection of younger adults. Specifically, in their study one in eight trials, on average, was a repetition of the stimulus that had been presented on the previous trial. Further, the conditions used by Terry and Sliwinski maximized the infiuence of stimulus repetitions. The short response-to-stimulus interval (RSI) of 100 ms that they used provided little time for task selection and task set reconfiguration processes to be executed before the next stimulus was presented. Under such conditions, the stimulus repetition effect is exaggerated (e.g., Weywadt & Butler, 2013). Therefore, according to a bottom-up processing account, it may be that the lower switch probabilities for older adults observed by Terry and Sliwinski were due to a greater influence of the environment on task selection when the stimulus repeated combined with equivalent endogenous task selection processes when the environment could not guide task selection. In this study, we compared older and younger adults' probability of switching separately for stimulus change and stimulus repetition trials. In addition to an RSI of 100 ms, we included RSI intervals of 500 and 1,000 ms to examine age differences in the time-course of task selection under more optimal selection conditions for endogenous control processes. Finally, to examine whether age differences in task maintenance might be exaggerated at longer delays between selection and execution, on the order of those used in previous work, we included a 5,000 ms RSI. VTS experiments routinely use the procedure and tasks that we used with the exception that an RSI greater than 1,500 ms has not been reported in the literature before (see Arrington et al., in press). Therefore, in addition to examining age differences in this condition, it is also of interest to observe if switch probabilities could more closely approximate randomness when more time is given for preparation. If older adults have difficulty inhibiting the infiuence of the environment in a VTS situation, we expect to find larger age differences in switch probabilities when the stimuli repeat compared to when they change. Equivalent endogenous processing under optimal conditions (i.e., at the 500 and 1,000 ms RSIs) is expected to make endogenous task selections more resistant to bottom-up influences as preparation increases and reduce the age differences in the effect of the environment on task selection. If the endogenous task selection processes differ for older (e.g., within-set mode) and younger adults (e.g., task updating mode), but bottom-up processing influences on task selection do not, then the effect of the environment on task selection should be equivalent for older and younger aduhs. Stimulus repetition, therefore, should have the same effect on task selection for older and younger adults. The inclusion of the 5,000 ms RSI condition allows us to evaluate whether older adults might have more difficulty main-
BUTLER AND WEYWADT
1026
taining that selection than younger adults as well. If older adults have more difficulty maintaining their task selection across the 5-second interval, we would expect a larger effect of the stimulus repetition in the 5,000 ms RSI condition for older than younger adults. Contrary to these predictions, we observed that older adults' task selections were less influenced by stimulus repetitions than younger adults across RSIs.
Method Participants Thirty-three younger adults (24 female; M^^^ = 19.9 years, SD = 1.9, age range = 18-24 years; MEJ,,,,^„„ = 13.4 years, SD = 1.3) recruited from introductory psychology classes at the University of New Mexico and 30 older adults (22 female; M^^^ = 75.0 years, SD = 7.1, age range = 61-88 years; M£,,„„,„.„„ = 15.6 years, SD = 2.4) recruited from the Albuquerque, NM area were included in the analyses. Older adults were screened in a telephone interview to be sure they had no history of neurological disorder, including stroke, or history of depression. Older participants provided their own transportation to the lab where they filled out an additional demographic sheet to indicate their current health status, any medical conditions, as well as any medications. Younger adults received course credit for their participation and older adults received $10Aiour. Using previously established exclusion criteria, two younger and three older participants were excluded from the analyses (e.g., Butler, Arrington, & Weywadt, 2011). One younger and one older participant were unable to finish the procedure in the allotted time, one younger participant switched more than 90% of the time, one older participant switched less than 5% of the time, and one made errors on more than 20% of the trials. The exclusion of these participants likely reduced the age difference observed in probability of switching.
Apparatus, Materials and Procedure Stimuli were presented on a Dell Dimension computer on a 17-inch CRT monitor using the E-Prime software package (Schneider, Eschmann, & Zuccolotto, 2002). Responses were made on a Psychological Software Tools, Inc. serial response box. All stimuli were presented in black on a light gray background in 14-point courier new bold font. Stimuli included the digits 1-4 and 6-9. Each digit was 6 mm x 8 mm and was presented just above a fixation cross. Like Terry and Sliwinski (2012), we allowed stimulus repetitions to happen randomly during stimulus presentation. First, participants signed a consent form and completed a demographic sheet that asked for their age, years of education, general health condition, and any current medications. Then they began the task-switching procedure by practicing each task, magnitude and parity, in single task (ST) blocks of 24 trials. The magnitude task required participants to judge a digit as greater or less than 5, while the parity task required participants to judge the digit as odd or even. Participants were then instructed to switch between tasks within blocks of trials. For three blocks of 48 trials, participants saw a randomly selected task cue followed by a stimulus and were told to perform the task associated with the cue.
Following these practice blocks, participants received the VTS instructions to perform each task equally often and in a random sequence (modeled from Arrington & Logan, 2004). During a practice block of 32 trials, participants were given feedback if patterns of task choices were observed; for example, switching between the tasks at regular intervals, repeatedly performing the same task for more than five trials, or performing one task for more than eight trials. Then participants completed eight blocks with 64 experimental trials each. For the VTS trials a fixation cross was present in the center of the screen for a randomly selected interval of 100, 500, 1,000, or 5,000 ms. Previous research using the VTS procedure has often manipulated RSI within blocks (Arrington & Logan, 2004; Butler et al., 2011; Weywadt & Butler, 2013) and research comparing variable and fixed RSI designs has concluded that the processes tapped by VTS are robust regardless of the design used (Liefooghe, Demanet & Vandierendonk, 2009). Following this RSI, the stimulus appeared above the fixation until a response was made. Stimuli and their RSIs were chosen randomly without replacement from a set of 64 trials that included two trials for each of the eight stimuli at each RSI. Each tasks' responses were mapped to either the two right-most or two left-most buttons on the response box and participants were asked to use their index and middle fingers to respond. The task-hand and stimulus-response mappings were counterbalanced across participants. The procedures took approximately 70 minutes for the younger participants and 90 minutes for the older participants.
Results Trials starting a block and trials with RTs greater than 3,000 ms or less than 150 ms were excluded from all analyses (1.5% of trials). Trials were coded to a particular task choice by the hand used to respond. A trial was coded as a task repetition when the task code on the previous trial was the same as the task code on the current trial. Trials were coded as switches when the task on the previous trial was different from the task on the current trial. Error trials and trials following errors were also excluded from analyses of the probability of switching and reaction time analyses (4.2% of trials). After excluding trials, stimulus repetitions occurred on 11% of trials for younger participants (11.2% 100 ms, 10.9% 500 ms, 11.4% 1,000 ms, and 11.1% 5,000 ms trials) and on 10.6% of the trials for the older participants (10.1%, 11.3%, 10.8% and 10.3% at each RSI, respectively). All main effects and interactions were significant at an alpha level of .05 unless otherwise stated.
Probability of Switching Probability of switching p(sw) was calculated by dividing the number of task switch trials by the total number of included trials. When mean p(sw) was calculated across all RSIs and stimulus types, older adults switched less than younger adults did (Ais = .369 for older and .447 for younger adults) as indicated by a univariate analysis of variance (ANOVA) between age groups, F(l, 61) = 5.02, p = .029, 1)1 = .08. As can be seen in Table 1, p(sw) was also calculated for each stimulus type and RSI condition and submitted to a three-way ANOVA with a between-subjects variable of age (younger or older) and within-subjects variables of RSI (100, 500, 1,000, or 5,000 ms) and stimulus type (stimulus
AGE DIFFERENCES IN VOLUNTARY TASK SWITCHING
1027
Table 1 Probability of Switching by Age, RSI, and Stimulus Type RSI
Younger Stimulus Stimulus Older Stimulus Stimulus Note.
100 ms M (SE)
500 ms M (SE)
1,000 ms M (SE)
5,000 ms M (SE)
repeated changed
.334 (.046) .338 (.029)
.379 (.045) .420 (.025)
.409 (.044) .482 (.024)
.534 (.034) .567 (.021)
repeated changed
.316 (.048) .278 (.031)
.392 (.047) .342 (.026)
.428 (.046) .379 (.025)
.491 (.036) .461 (.022)
Values in parentheses are standard errors of the mean.
repeated or changed). Age did not affect p(sw) in this analysis, F = 1.31. As can be seen in Figure 1, the age effect was obscured by the interaction pattem of age with stimulus type, F(l, 61) = 7.14, p = .010, Tip = .11. Similar to previous observations, when the stimulus repeated younger adults were more likely to repeat tasks than when the stimulus changed, i(32) = -2.48, p = .019. Older adults were less likely to switch when the stimulus repeated compared to a stimulus change, although this difference was not significant, i(29) = 1.59, p = .124. There was the expected main effect of RSI indicating that p(sw) increased as individuals had more time to prepare, F(3, 183) = 54.97, p < .001, Ti^ = .474, but this did not interact with age, F = 0.11. No other effects or interactions were significant, all Fs < 0.85.
Reaction Times Two sets of analyses were conducted on the RTs. In the first analysis, age-related differences in the effects of stimulus type on RTs were evaluated. In this analysis, we could only include data from participants with at least one observation in each RSI by transition by stimulus type condition («yo^^g = 26 and n^¡¿^^ = 19). When condition means were calculated across all participants and compared to the means from the reduced sample, the pattems of significant effects were the same. The second set of analyses included all the participants by excluding the stimulus type variable and including only the stimulus change trials.
Probability of Switching by Age Group and Stimulus Type
•80.25 •Q0.20
-Younger -»-Older
2 0.15 0.05 0.00 Change
Repeat
Stimulus Type Figure 1. Probahility of switching by age and stimulus type.
To evaluate the effects of stimulus type on transition RTs for the two age groups we used a four-way mixed effects ANOVA with age as a between-subjects variable and stimulus type, RSI, and transition type (switch or repeat task) as within-subjects variables (see Table 2). Because the effects of the other variables on RTs are better reflected by the larger sample included in the subsequent analysis, and for brevity, only the effects and interactions with age group and stimulus type will be presented. RTs were faster on stimulus repetition than stimulus change trials, F(l, 43) = 29.06, p < .001, T)^ = .40. This effect depended on whether the task chosen was a task repetition or a task switch, F(l, 43) = 47.02, p < .001, T|p = .52. Follow-up tests revealed that individuals were significantly faster to repeat a task if the stimulus repeated, /(60) = —11.516,p< .001. However, stimulus repetitions did not significantly speed responses if an individual chose to switch tasks, r(46) = -.101, p = .920. A marginally significant Stimulus type X Age X RSI interaction reflected less difference in older adults' RTs for each stimulus type than the difference observed for younger adults in the 100, 500, and 1,000 ms RSI conditions, but similar differences in the 5,000 ms RSI, F(3, 129) = 2.50, p = .063, Tip = .055. Stimulus type did not enter into any other interactions with age: Stimulus type X Age, F(l, 43) = 0.003, p = .953, TiJ
Psychology and Aging 2013, Vol. 28. No. 4, 1024-1031
Age Differences in Voluntary Task Switching Karin M. Butler and Christina Weywadt University of New Mexico Task choice processes in older (60-1- years) and younger (18-30 years) adults were compared using a voluntary task switching procedure (Arrington & Logan, 2004). To assess age-related differences in task representation maintenance, preparation times were varied across a large range of responseto-stimulus intervals (100, 500, 1,000, and 5,000 ms) and the environmental influence on task selection was varied by repeating or changing stimuli from trial to trial. Older adults switched less frequently than younger adults and this effect was the same at each RSI. Younger adults were more likely to switch tasks when the stimulus changed than when it repeated suggesting that they used a different process to determine task choices, either endogenous task selection or environmentally supported task repetitions. Older adults' task selection was unaffected by stimulus repetitions indicating that they were less flexible with the processing they used to guide task selection. These findings are consistent with previous observations of age-related increases in goal shielding, but not with age-related deficits in task goal maintenance. Robust age differences in switch costs were observed across RSIs suggesting that task reconfiguration processes are different following endogenous than exogenous task selection. Keywords: aging, task choice, multitasking, task representation, cognitive control
Flexibly moving between tasks in the absence of external cues is important for day-to-day life. As we age, optimal task switching may decline as a result of age-related changes in the ability to inhibit prepotent responses to environmental stimuli (e.g., Butler, Zacks, & Henderson, 1999; Hasher, Zacks, & May, 1999), difficulty maintaining task goals (Braver et al., 2001; Mayr, 2001), or the need for additional time to support task reconfiguration (Salthouse, 1996). In this study we evaluated age differences in task choice in an endogenous task switching environment known as the voluntary task switching (VTS) procedure while manipulating both the time available to prepare a task selection and the presence of an environmental context that supports repetition of the current task rather than task switching. Efficiently moving between tasks is impaired by deficits in maintaining contextually appropriate goals according to the goal maintenance theory of age-related differences in goal-directed behavior (Braver et al, 2001; Braver & West, 2008). Older adults display difficulties maintaining task goals across intervals as short as 5 seconds. For example, in the AX-continuous performance task (AX-CPT), individuals are shown one of four pairs of letters with a delay of 4,900 ms between them (AX, AY, BX, or BY). They are instructed to respond positively only when an "X" probe stimulus
follows an "A" stimulus. On 80% of trials that begin with an "A" stimulus, an "X" probe stimulus follows, biasing the individual to prepare to respond to the probe stimulus. When the "A" is followed by a "Y" probe, false alarms are made often. Older adults made fewer of these errors than younger adults did, though. The result suggests that older adults' goal representations (e.g., prepare to respond to the "X" probe) degrade more quickly than the representations of the younger adults do across interval as short as 4,900 ms. Age-related task maintenance deficits have been demonstrated in a prospective memory task as well. In the context of an ongoing task, if a new goal is introduced that must be maintained across a short 5-second interval (i.e., a prospective memory task), older adults will more often forget to execute that intention than younger adults will (McDaniel, Einstein, Stout, & Morgan, 2003). Age-related deficits in goal representations have also been observed in task switching environments that cued participants to perform a certain task on each trial. The costs associated with being in a task switching environment compared to a single task environment were larger for older than younger adults, particularly when both the relevant stimulus attributes and the response options for each task were the same (Mayr, 2001). Mayr concluded that from one trial to the next older adults do not maintain the previously executed or currently required task as robustly as younger adults do. When a task repetifion is required, this results in more time needed to reactivate the just executed task, as well as greater interference (slowing) when a different task is required but the stimulus or response is the same. The idea that task representations are weaker in older individuals was challenged by Lien, Ruthruff, and Kuhns (2008). They found that older individuals were as affected as younger individuals when task expectancies were violated. Across three experiments, participants used temporal and location cues to predict what
Karin M. Butler and Christina Weywadt, Psychology Department, University of New Mexico. Portions of this work were presented at the 50th Annual Psychonomic Society Conference, Boston, Massachusetts, in 2009. This research was supported by a Ouad-L Eoundation grant to Christina Weywadt. Our thanks to Melakeh Kurdi for her assistance with data collection. Correspondence concerning this article should be addressed to Karin M. Butler, MSC 03 2220; Logan Hall, Department of Psychology, 1 University of New Mexico, Albuquerque, NM 87131-0001. E-mail: kmbutler@ unm.edu 1024
AGE DIFFERENCES IN VOLUNTARY TASK SWITCHING
task to perform on each trial. Occasionally, the stimulus that was presented would not afford the expected task, though. Older adults were slowed by the violation of this expectancy more so than younger adults, suggesting that older adults maintained task information as robustly as younger adults, if not more so. The researchers suggested that the older adults used the predictable environment to establish robust task representations in the procedure that they used. In fact, older adults will rely on the environment to maintain robust task representations even when it is not necessary (Mayr & Liebscher, 2001; Spieler, Mayr, & LaGrone, 2006). After switching tasks in blocks of trials that required following the direction of task cues, older individuals were much more likely than younger adults to look at task-related, but unnecessary cues during single task performance (Spieler et al., 2006). Persisting in processing the cues slowed older adults' responses compared to responses in a single task block performed prior to the cued task switching trials. Older adults who did not have the cues available during single task performance reached the speed of baseline reaction times (RTs) within 60 trials of the single task blocks. Spieler and colleagues (2006) suggested that older individuals are less flexible in their reliance on internal and extemal sources of control. When older adults can rely on a single mode of control (internal or extemal), age-differences are reduced or eliminated, but when performance requires switching between control strategies, older adults have more difficulty than younger adults do (e.g., Ardiale & Lemaire, 2012). In order to examine task selection processes in a multitask environment, while eliminating the influence of cue encoding processes and overreliance on environmental cues to guide task selection and switching, this study examines age differences in task switching using the VTS procedure. In the VTS procedure, participants are told to choose randomly which of two tasks to perform on each trial (Arrington & Logan, 2005; Arrington, Rieman, & Weaver, in press; Mayr & Bell, 2006). Switch probabilities are typically less than 50% reflecting a repetition bias. This bias has been attributed to the task representation used on one trial persisting and infiuencing the task selection on the next trial. Probability of switching increases as the time available to make a choice increases, suggesting that probability of switching is sensitive to endogenous control processes (Arrington & Logan, 2004). Successful random task choice requires overcoming the persisting representation from previous trials and/or maintaining a representation of the sequence of previously executed tasks. Task selection processes in younger adults can be influenced by the environment through bottom-up processing in the VTS procedure. Younger adults are more likely to repeat a task when the stimulus repeats from the previous trial compared to when it changes (e.g., Mayr & Bell, 2006; Demanet, Verbruggen, Liefooghe, & Vandierendonck, 2010). To date, only one study has investigated age differences in a VTS environment. Terry and Sliwinski (2012) showed that older individuals were less likely to switch tasks than younger adults. This difference was not observed when the groups were simply asked to generate a series of random "coin-nips," however, supporting the view that older and younger adults have a similar representation of randomness. In addition, older adults' switch probabilities were positively correlated with their task repetition RTs to a greater extent than those of younger adults. This pattem
1025
suggested that older adults adopted an endogenous task selection strategy of maintaining the strength of the current task to support task repetition performance at the cost of lower switch rates, a phenomenon they referred to as a within-set mode of processing. In this view, younger adults were more likely to use a task updating strategy that led to more frequent task reconfiguration to support task switching. The Terry and Sliwinski (2012) study left open the possibility that age differences in the task selection processes may be due to the influence of bottom-up processing on task selection rather than differences in the endogenous processes. Older adult task selection may have been more infiuenced by the environment than the task selection of younger adults. Specifically, in their study one in eight trials, on average, was a repetition of the stimulus that had been presented on the previous trial. Further, the conditions used by Terry and Sliwinski maximized the infiuence of stimulus repetitions. The short response-to-stimulus interval (RSI) of 100 ms that they used provided little time for task selection and task set reconfiguration processes to be executed before the next stimulus was presented. Under such conditions, the stimulus repetition effect is exaggerated (e.g., Weywadt & Butler, 2013). Therefore, according to a bottom-up processing account, it may be that the lower switch probabilities for older adults observed by Terry and Sliwinski were due to a greater influence of the environment on task selection when the stimulus repeated combined with equivalent endogenous task selection processes when the environment could not guide task selection. In this study, we compared older and younger adults' probability of switching separately for stimulus change and stimulus repetition trials. In addition to an RSI of 100 ms, we included RSI intervals of 500 and 1,000 ms to examine age differences in the time-course of task selection under more optimal selection conditions for endogenous control processes. Finally, to examine whether age differences in task maintenance might be exaggerated at longer delays between selection and execution, on the order of those used in previous work, we included a 5,000 ms RSI. VTS experiments routinely use the procedure and tasks that we used with the exception that an RSI greater than 1,500 ms has not been reported in the literature before (see Arrington et al., in press). Therefore, in addition to examining age differences in this condition, it is also of interest to observe if switch probabilities could more closely approximate randomness when more time is given for preparation. If older adults have difficulty inhibiting the infiuence of the environment in a VTS situation, we expect to find larger age differences in switch probabilities when the stimuli repeat compared to when they change. Equivalent endogenous processing under optimal conditions (i.e., at the 500 and 1,000 ms RSIs) is expected to make endogenous task selections more resistant to bottom-up influences as preparation increases and reduce the age differences in the effect of the environment on task selection. If the endogenous task selection processes differ for older (e.g., within-set mode) and younger adults (e.g., task updating mode), but bottom-up processing influences on task selection do not, then the effect of the environment on task selection should be equivalent for older and younger aduhs. Stimulus repetition, therefore, should have the same effect on task selection for older and younger adults. The inclusion of the 5,000 ms RSI condition allows us to evaluate whether older adults might have more difficulty main-
BUTLER AND WEYWADT
1026
taining that selection than younger adults as well. If older adults have more difficulty maintaining their task selection across the 5-second interval, we would expect a larger effect of the stimulus repetition in the 5,000 ms RSI condition for older than younger adults. Contrary to these predictions, we observed that older adults' task selections were less influenced by stimulus repetitions than younger adults across RSIs.
Method Participants Thirty-three younger adults (24 female; M^^^ = 19.9 years, SD = 1.9, age range = 18-24 years; MEJ,,,,^„„ = 13.4 years, SD = 1.3) recruited from introductory psychology classes at the University of New Mexico and 30 older adults (22 female; M^^^ = 75.0 years, SD = 7.1, age range = 61-88 years; M£,,„„,„.„„ = 15.6 years, SD = 2.4) recruited from the Albuquerque, NM area were included in the analyses. Older adults were screened in a telephone interview to be sure they had no history of neurological disorder, including stroke, or history of depression. Older participants provided their own transportation to the lab where they filled out an additional demographic sheet to indicate their current health status, any medical conditions, as well as any medications. Younger adults received course credit for their participation and older adults received $10Aiour. Using previously established exclusion criteria, two younger and three older participants were excluded from the analyses (e.g., Butler, Arrington, & Weywadt, 2011). One younger and one older participant were unable to finish the procedure in the allotted time, one younger participant switched more than 90% of the time, one older participant switched less than 5% of the time, and one made errors on more than 20% of the trials. The exclusion of these participants likely reduced the age difference observed in probability of switching.
Apparatus, Materials and Procedure Stimuli were presented on a Dell Dimension computer on a 17-inch CRT monitor using the E-Prime software package (Schneider, Eschmann, & Zuccolotto, 2002). Responses were made on a Psychological Software Tools, Inc. serial response box. All stimuli were presented in black on a light gray background in 14-point courier new bold font. Stimuli included the digits 1-4 and 6-9. Each digit was 6 mm x 8 mm and was presented just above a fixation cross. Like Terry and Sliwinski (2012), we allowed stimulus repetitions to happen randomly during stimulus presentation. First, participants signed a consent form and completed a demographic sheet that asked for their age, years of education, general health condition, and any current medications. Then they began the task-switching procedure by practicing each task, magnitude and parity, in single task (ST) blocks of 24 trials. The magnitude task required participants to judge a digit as greater or less than 5, while the parity task required participants to judge the digit as odd or even. Participants were then instructed to switch between tasks within blocks of trials. For three blocks of 48 trials, participants saw a randomly selected task cue followed by a stimulus and were told to perform the task associated with the cue.
Following these practice blocks, participants received the VTS instructions to perform each task equally often and in a random sequence (modeled from Arrington & Logan, 2004). During a practice block of 32 trials, participants were given feedback if patterns of task choices were observed; for example, switching between the tasks at regular intervals, repeatedly performing the same task for more than five trials, or performing one task for more than eight trials. Then participants completed eight blocks with 64 experimental trials each. For the VTS trials a fixation cross was present in the center of the screen for a randomly selected interval of 100, 500, 1,000, or 5,000 ms. Previous research using the VTS procedure has often manipulated RSI within blocks (Arrington & Logan, 2004; Butler et al., 2011; Weywadt & Butler, 2013) and research comparing variable and fixed RSI designs has concluded that the processes tapped by VTS are robust regardless of the design used (Liefooghe, Demanet & Vandierendonk, 2009). Following this RSI, the stimulus appeared above the fixation until a response was made. Stimuli and their RSIs were chosen randomly without replacement from a set of 64 trials that included two trials for each of the eight stimuli at each RSI. Each tasks' responses were mapped to either the two right-most or two left-most buttons on the response box and participants were asked to use their index and middle fingers to respond. The task-hand and stimulus-response mappings were counterbalanced across participants. The procedures took approximately 70 minutes for the younger participants and 90 minutes for the older participants.
Results Trials starting a block and trials with RTs greater than 3,000 ms or less than 150 ms were excluded from all analyses (1.5% of trials). Trials were coded to a particular task choice by the hand used to respond. A trial was coded as a task repetition when the task code on the previous trial was the same as the task code on the current trial. Trials were coded as switches when the task on the previous trial was different from the task on the current trial. Error trials and trials following errors were also excluded from analyses of the probability of switching and reaction time analyses (4.2% of trials). After excluding trials, stimulus repetitions occurred on 11% of trials for younger participants (11.2% 100 ms, 10.9% 500 ms, 11.4% 1,000 ms, and 11.1% 5,000 ms trials) and on 10.6% of the trials for the older participants (10.1%, 11.3%, 10.8% and 10.3% at each RSI, respectively). All main effects and interactions were significant at an alpha level of .05 unless otherwise stated.
Probability of Switching Probability of switching p(sw) was calculated by dividing the number of task switch trials by the total number of included trials. When mean p(sw) was calculated across all RSIs and stimulus types, older adults switched less than younger adults did (Ais = .369 for older and .447 for younger adults) as indicated by a univariate analysis of variance (ANOVA) between age groups, F(l, 61) = 5.02, p = .029, 1)1 = .08. As can be seen in Table 1, p(sw) was also calculated for each stimulus type and RSI condition and submitted to a three-way ANOVA with a between-subjects variable of age (younger or older) and within-subjects variables of RSI (100, 500, 1,000, or 5,000 ms) and stimulus type (stimulus
AGE DIFFERENCES IN VOLUNTARY TASK SWITCHING
1027
Table 1 Probability of Switching by Age, RSI, and Stimulus Type RSI
Younger Stimulus Stimulus Older Stimulus Stimulus Note.
100 ms M (SE)
500 ms M (SE)
1,000 ms M (SE)
5,000 ms M (SE)
repeated changed
.334 (.046) .338 (.029)
.379 (.045) .420 (.025)
.409 (.044) .482 (.024)
.534 (.034) .567 (.021)
repeated changed
.316 (.048) .278 (.031)
.392 (.047) .342 (.026)
.428 (.046) .379 (.025)
.491 (.036) .461 (.022)
Values in parentheses are standard errors of the mean.
repeated or changed). Age did not affect p(sw) in this analysis, F = 1.31. As can be seen in Figure 1, the age effect was obscured by the interaction pattem of age with stimulus type, F(l, 61) = 7.14, p = .010, Tip = .11. Similar to previous observations, when the stimulus repeated younger adults were more likely to repeat tasks than when the stimulus changed, i(32) = -2.48, p = .019. Older adults were less likely to switch when the stimulus repeated compared to a stimulus change, although this difference was not significant, i(29) = 1.59, p = .124. There was the expected main effect of RSI indicating that p(sw) increased as individuals had more time to prepare, F(3, 183) = 54.97, p < .001, Ti^ = .474, but this did not interact with age, F = 0.11. No other effects or interactions were significant, all Fs < 0.85.
Reaction Times Two sets of analyses were conducted on the RTs. In the first analysis, age-related differences in the effects of stimulus type on RTs were evaluated. In this analysis, we could only include data from participants with at least one observation in each RSI by transition by stimulus type condition («yo^^g = 26 and n^¡¿^^ = 19). When condition means were calculated across all participants and compared to the means from the reduced sample, the pattems of significant effects were the same. The second set of analyses included all the participants by excluding the stimulus type variable and including only the stimulus change trials.
Probability of Switching by Age Group and Stimulus Type
•80.25 •Q0.20
-Younger -»-Older
2 0.15 0.05 0.00 Change
Repeat
Stimulus Type Figure 1. Probahility of switching by age and stimulus type.
To evaluate the effects of stimulus type on transition RTs for the two age groups we used a four-way mixed effects ANOVA with age as a between-subjects variable and stimulus type, RSI, and transition type (switch or repeat task) as within-subjects variables (see Table 2). Because the effects of the other variables on RTs are better reflected by the larger sample included in the subsequent analysis, and for brevity, only the effects and interactions with age group and stimulus type will be presented. RTs were faster on stimulus repetition than stimulus change trials, F(l, 43) = 29.06, p < .001, T)^ = .40. This effect depended on whether the task chosen was a task repetition or a task switch, F(l, 43) = 47.02, p < .001, T|p = .52. Follow-up tests revealed that individuals were significantly faster to repeat a task if the stimulus repeated, /(60) = —11.516,p< .001. However, stimulus repetitions did not significantly speed responses if an individual chose to switch tasks, r(46) = -.101, p = .920. A marginally significant Stimulus type X Age X RSI interaction reflected less difference in older adults' RTs for each stimulus type than the difference observed for younger adults in the 100, 500, and 1,000 ms RSI conditions, but similar differences in the 5,000 ms RSI, F(3, 129) = 2.50, p = .063, Tip = .055. Stimulus type did not enter into any other interactions with age: Stimulus type X Age, F(l, 43) = 0.003, p = .953, TiJ
