- Split View
-
Views
-
Cite
Cite
Kiran K. Vadaga, Mervin Blair, Karen Z. H. Li, Are Age-Related Differences Uniform Across Different Inhibitory Functions?, The Journals of Gerontology: Series B, Volume 71, Issue 4, July 2016, Pages 641–649, https://doi.org/10.1093/geronb/gbv002
- Share Icon Share
Abstract
In the current experiment, we examined the relative age-sensitivity of 3 inhibitory functions: access, deletion, and restraint by taking into consideration their underlying control processes: proactive and reactive control.
The 3 inhibitory functions were measured using a sequential flanker task. Young (age: 18–35, n = 24) and older adults (age: 60–75, n = 25) first memorized a series of 8 animal words in a fixed order. In the test phase, these stimuli were presented randomly either singly or with flankers and participants responded “yes” or “no” based on the prelearned sequence. In the access trials, flankers were either ahead of the current target or unrelated. In the deletion trials, flankers were previous target items. In the restraint trials, the flanker cues (XXXX) prompted the participants to withhold responses occasionally. Unflanked trials served as the baseline condition.
Age-related differences in the magnitude of inhibition effects were largest in restraint, followed by deletion. No age-related differences were observed in access.
Our findings suggest that the magnitude of age-related differences in inhibitory functions is contingent on the degree of proactive control recruited by a given inhibitory function.
The current literature provides mixed evidence on the relative magnitude of age effects across three inhibitory functions: access, deletion and restraint. We examined the proposal that the magnitude of age effects should increase to the extent that proactive control processes are required, given the assertion that proactive control declines more so than reactive control with age ( Braver, Gray, & Burgess, 2007 ). To this end, we varied the requirement for proactive control within and across our tests of access, deletion and restraint. To maintain comparable task characteristics across measures, we operationalized the three inhibitory functions in terms of their underlying control processes within a single cognitive task.
Age-Related Differences in Three Inhibitory Functions
According to the inhibitory deficit hypothesis (IDH; Hasher, Lustig, & Zacks, 2007 ; Hasher, Zacks, & May, 1999 ), inhibitory-based executive control processes constrain the working memory (WM) system by suppressing goal-irrelevant information. To this end, Hasher and colleagues proposed three distinct inhibitory functions. The access function prevents the entry of irrelevant information (e.g., environmental distraction) into WM. The deletion function suppresses no-longer relevant information from WM by resisting proactive interference (PI) from previous tasks, and the restraint function suppresses goal-irrelevant habitual responses.
One major assumption of the IDH is that older adults show declines in all three inhibitory functions ( Hasher et al., 2007 ). Typically, the access function is measured by selective attention tasks where individuals focus on target information amongst distractors. To illustrate, Connelly, Hasher, and Zacks (1991) showed that in a reading with distraction task, older adults were slower and made more comprehension errors when the distractor information was semantically related to the target text. Similarly, in visual attention tasks older adults are typically slower to detect targets among distractors and make more errors as the number of distractors increases (e.g., Madden, 1983 ). The evidence for age-related differences in the deletion function comes from directed-forgetting (e.g., Andrés, Van der Linden, & Parmentier, 2004 ), updating (e.g., De Beni & Palladino, 2004 ), and WM tasks (e.g., May, Hasher, & Kane, 1999 ). For example, Blair, Vadaga, Shuchat, and Li (2011) showed that in a sequential updating task, older adults made more intrusion errors on items that were former targets, presumably reflecting age-related difficulty in suppressing no-longer relevant information. Support for age-related differences in the restraint function comes from response inhibition tasks such as the antisaccade task ( Butler, Zacks, & Henderson, 1999 ), stop-signal task ( Kramer, Humphrey, Larish, & Logan, 1994 ), and Stroop task ( Davidson, Zacks, & Williams, 2003 ) in which participants withhold habitual or overlearned responses.
Although much of the evidence points to age-related declines in inhibitory efficiency, we note that age-equivalent effects are reported in access, as measured by a problem solving task ( Feyereisen & Charlot, 2008 ), deletion as measured by a task switching flanker paradigm ( Li & Dupuis, 2008 ), and restraint as measured by go/no-go tasks ( Vallesi & Stuss, 2010 ). Together, the extant literature reveals age-sensitivity across many operationalizations of inhibition, but does not suggest a categorical pattern in which one inhibitory function is age-invariant while other functions decline substantially with age. Our approach is therefore to understand the underlying processes within each function that better predict where age-related decline is observed.
The Three Inhibitory Functions: Differing Mechanisms and Issues With Measurement
While the IDH explicitly outlines the function of each inhibitory process based on the type of irrelevant information being targeted, a more recent conceptualization has differentiated inhibitory processes based on other characteristics. For instance, Braver and colleagues (2007) have suggested that the different inhibitory effects may be a consequence of dual mechanisms of cognitive control (DMC). According to the DMC account, interference generated by irrelevant information is resolved by either proactive or reactive control processes. Broadly defined, proactive control is a form of preparatory attention and resolves interference by early correction methods such as anticipation and prior knowledge. Therefore, a proactive control strategy requires that the goal-related representations are actively maintained in WM. In contrast, reactive control resolves interference after it occurs. Thus, reactive control processes are recruited when there is no predictive information to resolve interference, and are engaged only as needed on a just-in-time basis ( Braver et al., 2007 ). With regard to age-related differences in proactive and reactive control processes, evidence from multiple studies suggests that reactive control is generally preserved in older adults, but proactive control, due to inherent WM demands, decline with age ( Braver, Paxton, Locke, & Barch, 2009 ; Braver, Satpute, Rush, Racine, & Barch, 2005 ; Braver et al., 2001 ; Paxton, Barch, Racine, & Braver, 2008 ).
To better understand how proactive and reactive control processes underlie different inhibitory functions, a more careful process-driven operationalization of tasks is required. We propose that access employs both proactive and reactive control processes depending on the nature of irrelevant information (i.e., proactive for anticipated distraction and reactive for novel information), while deletion and restraint engage preparatory attention via proactive control. We further propose that the degree of proactive control required can vary within and between inhibitory functions. For instance, varying degrees of proactive control can be seen in the deletion function, where PI from previous task representations is suppressed. If PI is small on any given trial (i.e., early on in a succession of items or lists), then a smaller degree of proactive control is required. Among different inhibitory functions, restraint elicits the largest degree of proactive control, as overriding of habitual responses requires sustained goal-relevant representations in WM.
While we acknowledge that proactive and reactive control processes are distinct mechanisms, we note that any given complex cognitive task employs a blend of reactive and proactive control processes. For example, a prototypical selective attention task that requires participants to attend to targets and ignore distractors may involve both proactive and reactive control. This task would require reactive control process if the distractor information (i.e., content, location, frequency) is unpredictable. However, the superordinate task instructions—respond to targets and ignore distractors—can be seen as employing a certain degree of proactive control. Therefore, we prefer not to conceptualize any given task as categorically proactive or reactive. This also implies that one cannot easily classify access, deletion, and restraint functions as categorically or orthogonally requiring proactive or reactive control. Instead, we argue that it is more fruitful to assume that the magnitude of age differences in inhibitory performance will depend on the degree to which proactive control processes are required.
We believe that the integration of the DMC framework with the original IDH offers a fruitful approach to understanding age-related differences in different inhibitory functions. To this end, we ask whether the magnitude of age-related decline in inhibitory functions is dependent on the extent to which proactive control processes are recruited. Based upon this process-oriented approach, one can then revisit the issue of the relative magnitude of age-related decline in different inhibitory functions.
As mentioned earlier, access, deletion, and restraint are typically measured by selective attention tasks, updating tasks, and response inhibition tasks, respectively. Therefore, the relative magnitude of age-related differences across the three inhibitory functions is partly confounded by variations in measurement characteristics. To alleviate this measurement problem, other researchers (e.g., Friedman & Miyake, 2004 ) have employed latent variable analyses to statistically extract the common variance among multiple tasks chosen to tap inhibitory processes. Although this approach succeeds in controlling for task-specific variance, we note that the choice of tasks included in the analysis will influence outcomes and conclusions. For example, if all the tasks chosen to measure deletion involve updating and WM demands, then the latent variable will capture both the inhibitory and WM processes (e.g., Friedman & Miyake, 2004 ). Therefore, if one were to examine the relative age-related differences across inhibitory functions, it would be advisable to measure each inhibitory function under comparable task conditions.
Current Study
To examine whether there are uniform age-related differences across access, deletion, and restraint functions of inhibition, the sequential flanker task (SFT: Li & Dupuis, 2008 ) was modified to measure each inhibitory function while holding the task-specific characteristics (e.g., retrieval, updating, response mappings) constant. Participants memorized a fixed sequence of word stimuli and responded to the items based on the learned sequence. In the current study, flanked stimuli were presented in access, deletion, and restraint trials. Under the assumption that access recruits both reactive and proactive control processes, we measured access in two different ways. One type of access trial contained novel flankers (to recruit reactive control), whereas a second type of access trial involved flanker words that were future targets (to recruit proactive control). To measure varying degrees of proactive control in the deletion function, we had two types of flankers in the deletion trials: (a) flankers that were targets in the immediately preceding trial (low PI), and (b) flankers that were targets in earlier trials, but were not immediately preceding (high PI). The operationalization of restraint was modeled after the go/no-go paradigm ( Donders, 1969 ). Given that the participants in the SFT make a speeded response to every presented word, it was predicted that after many trials this response pattern would become well-learned, or pre-potent. Therefore, the restraint trials included a flanker cue (XXXX) prompting the participants to occasionally withhold their response. In our operationalization of the three inhibitory functions, the restraint trials are assumed to recruit the largest degree of proactive control as the participants have to actively maintain task instructions (no response to XXXX flanker cue) in WM.
Taken together, we assumed that access recruits a combination of proactive and reactive control processes; deletion recruits varying degrees of proactive control, while restraint recruits the largest degree of proactive control. We predicted that our operationalization of restraint should be the most age-sensitive, while the novel flanker variant of our access operationalization should be the least age-sensitive. Further, we expected that in the intermediate case of deletion, there should be larger age-related differences in the trials involving high PI than those involving immediate suppression.
Method
Participants
Twenty-four young ( Mage = 22.75 years, SD = 3.77) and 25 older adults ( Mage = 66.50 years, SD = 3.50), were recruited from the Psychology Department at Concordia University, and the Montreal community, respectively. Participants were excluded if they reported any conditions that might impair perceptual abilities, concentration, or fine motor performance. Young adults were compensated with partial course credit, whereas older adults were compensated with a $20 honorarium. Older adults had more years of formal education ( M = 16.43, SD = 2.51) compared to young adults ( M = 15.25, SD = 1.32), t (47) = 2.04, p < .05. Both groups, however, were similar in general health status (young: M = 3.91, SD = 0.63; older: M = 3.86, SD = 0.69), t (47) = .23, p =.81, with options 1 through 5 representing poor , fair , good , very good , and excellent , respectively.
Materials
Background measures
To better describe the cognitive abilities of our sample, tests of psychomotor speed (WAIS digit-symbol substitution test; Wechsler, 1981 ), interference control (Stroop task; adapted from Spreen & Strauss, 2001 ), WM capacity (modified reading span task; Daneman & Carpenter, 1980 ), and language abilities (Extended Range Vocabulary Test; ERVT Form V2; Educational Testing Service, 1976 ) were given to the participants.
In the WAIS digit-symbol substitution test, participants copied the symbols corresponding to each of the randomly ordered digits, according to the key shown at the top of the worksheet. The dependent variable was the number of symbols substituted correctly within 90 s. In the Stroop task baseline condition, participants indicated verbally the ink color of 112 neutral stimuli (XXX) printed on a single sheet. In the incongruent condition, participants were instructed to name the ink color of each word (e.g., “red”) while avoiding the tendency to read the word (e.g., “blue”). The dependent variable was the proportional slowdown (i.e., correct words per second) from the neutral to the incongruent condition.
The modified reading span task ( Daneman & Carpenter, 1980 ) was used to measure WM capacity. The task was computerized and programmed with Superlab V. 4.7. The task comprised short sentences presented on a desktop monitor in Black, 22 point Times New Roman font on a white background. Sentences were presented one at a time, and participants were asked to read them out loud and make a key press response indicating whether they made sense or not. The task began with sets of two sentences, and after every two trials the set size increased by one sentence, up to six sentences per set. After the completion of each set, the participants were cued to recall the last word of each sentence in the order they were presented. The dependent variable was the total number of words recalled correctly.
Sequential flanker task (SFT)
The stimuli for the SFT consisted of eight animal words presented on a desktop monitor in yellow, 22 point Times New Roman font on a dark blue background. Each animal word was presented either singly, or with flankers (irrelevant information). On flanked trials, three items appeared on the screen in a column 2cm high, and spacing between them was kept constant. In keeping with previous studies ( Li & Dupuis, 2008 ), we followed the methodology of Shaw’s (1991) study of aging and flanker effects in terms of visual angle and viewing distance. On any flanked trial, the flanker words were presented above and below the middle item and were always identical, but always different from the target item. Given that it is easier to ignore visual distraction presented at the same location ( Carlson, Hasher, Zacks, & Connelly, 1995 ; Li & Dupuis, 2008 ; Li, Hasher, Jonas, Rahhal, & May, 1998 ), the stimuli were presented randomly at five different screen locations to maximize the effects the flankers. The inter-stimulus interval (ISI) was kept constant throughout the experiment at 1500ms.
At the beginning of the task, participants were asked to memorize eight animal words (i.e., butterfly, camel, tiger, ladybug, zebra, wolf, bird, elephant) in a predefined order. Once they were able to recall all the words in the set order without errors, they were given further instructions about the task. In the SFT, the participants’ objective was to determine whether the presented animal word was a target or a distractor based on the pre-learned sequence. A typical trial would begin with participants looking for the first animal word, butterfly (target condition). If the target word was found, then the participants responded by pressing the “yes” key; otherwise they pressed the “no” key (distractor condition). Right and left arrow keys on the standard keyboard were assigned for “yes” and “no” responses and were counterbalanced across participants. Once the participants responded to the first target, then they looked for second target (i.e., camel) and so on, until they reached the final target word of the learned sequence. A completed sequence or “run” included 16 trials with equal numbers of targets and distractors. After each run there was a short pause (5 s), followed by a cue indicating the beginning of a new run. Since the stimuli in a run were presented in scrambled order, the participants had to constantly update and activate the next target word in WM.
Operationalization of inhibitory functions
Access was operationalized with two types of trials: extra-list flanker trials and positive lag flanker trials. In the extra-list flanker trials, flankers were from outside the current set of eight animal words and were drawn from different categories (e.g., valley, cloud, guitar, etc.), controlled for frequency and word length ( Battig & Montague, 1969 ). In the positive lag flanker trials, flankers included all the animal words that were ahead of the presented word. For example, in the predefined sequence shown above, if the participants were looking for the fourth animal word, ladybug, and then wolf would be a lag +2 flanker. The latencies on these trial types were compared against an unflanked (baseline) condition to assess the efficiency of access. We included two different types of irrelevant information to acknowledge the possibility that the access function operates by both reactive and proactive control processes. For instance, when the positive lag flankers are presented, it is assumed that the top-down control processes prepare the participants to maintain the set order of the targets in their WM to effectively deal with the distraction. In contrast, when novel items are presented such as extra list flankers, bottom-up attentional processes via reactive control are assumed to address the distraction.
The deletion function was operationalized in two ways that varied in terms of proactive control requirements: lag −1 flankers and pooled negative lag flankers . In the example given, tiger would be a lag −1 flanker, as it was responded to in the preceding trial. All the previously relevant items other than the just preceding item (i.e., lag −1) are referred to as pooled negative lag flankers. The latencies on the deletion trial types were compared against the baseline unflanked condition. While the lag −1 flankers represent immediate PI, the pooled negative flankers represent a buildup of PI across trials, and require sustained deletion over time.
The restraint function refers to the suppression of habitual or over-learned responses. The participants were instructed to make a speeded response to every presented word. It is assumed that after many trials, this response pattern would become habitual. In 20% of all trials, the flankers were “XXXXX”, indicating that participants should withhold their response. Thus, the inability to avoid responding (i.e., commission errors) was considered a failure of restraint.
In sum, the trials for all three inhibitory functions were flanked and had similar perceptual and retrieval demands. To increase the stability of the data, the correct RTs for both the target and distractor conditions (yes and no responses) were pooled. Similarly, response errors on the restraint trials included both the target and distractor trials.
Trial construction
Overall there were 928 trials: 58 runs × 16 stimuli, grouped into 7 blocks. The first block was considered practice, and these data were not included in the statistical analyses. Both the targets and distractors were equally represented (464 trials each), and the baseline trials (unflanked condition, 16%), access trials (positive lag flankers, 15%; extra list flankers, 15%), deletion trials (lag −1 flankers, 19%; pooled negative lag flankers, 15%), and restraint trials (XXXXX flankers, 20%) had comparable representation.
Sequences of trials were constructed with the constraint that there could be no more than three consecutive “yes” or “no” responses. A similar constraint was used for flanked and unflanked trials. Following errors of omission or commission, error screens indicated that an error had occurred and oriented participants to the next sequence item. Figure 1 illustrates one partial trial and includes examples of flanked, unflanked, access, deletion, and restraint trials.
General Procedure
The participants were tested in the Adult Development and Aging Laboratory at Concordia University. A consent form and demographic questionnaire (age, years of education, general health status, and current medications) were given early in the session. Participants were then familiarized with the SFT and completed eight practice trials. A short break was provided after the first three test blocks, during which the digit-symbol substitution task and Stroop task were administered. This was followed by three more blocks of the SFT, the reading span task and ERVT. Participants were then debriefed and compensated. Each session lasted approximately 90min.
Data Reduction
In the SFT data, data from error trials, trials following errors, and restraint trials were excluded from the RT analyses. To account for general performance on the SFT task (i.e., processing speed, updating, and retrieval demands), median RTs for the access and the deletion trials were transformed into residual scores by partialling out the RT variance from the baseline, unflanked trials. To examine the relative magnitude of age-related differences in each inhibitory function, we estimated the effect size (Cohen’s D) by means of the formula d = 2t/√ df. Participants were classified as outliers if they exceeded more than 3 SDs on the SFT (RTs, errors) and at least one of the background neuropsychological measures. Two older adults were classified as outliers and their data excluded from further analysis.
Results
The background measures and all the trial types on the SFT were approximately normally distributed with acceptable values of skewness and kurtosis of less than 3 and 10 respectively ( Kline, 2009 ). Descriptive statistics for the background measures are shown in Table 1 . As expected, age-related differences were observed in processing speed, interference proneness and WM favoring the young group. However, older adults’ verbal knowledge was superior to that of young adults. See Table 2 for descriptive statistics and reliability estimates for all trial types in the SFT.
Age group . | Digit symbol a, * . | Stroop b, * . | WM c, * . | ERVT d, * . |
---|---|---|---|---|
Young | 69.66 (8.91) | 0.29 (0.12) | 23.66 (4.34) | 07.23 (3.64) |
Older | 58.83 (8.75) | 0.45 (0.19) | 20.50 (3.03) | 14.78 (6.20) |
Age group . | Digit symbol a, * . | Stroop b, * . | WM c, * . | ERVT d, * . |
---|---|---|---|---|
Young | 69.66 (8.91) | 0.29 (0.12) | 23.66 (4.34) | 07.23 (3.64) |
Older | 58.83 (8.75) | 0.45 (0.19) | 20.50 (3.03) | 14.78 (6.20) |
Notes : WM = working memory. Values reflect average score per group; standard deviations are shown in parentheses.
a Wechsler Adult Intelligence Scale-Revised (WAIS-R) Digit Symbol Substitution. The values reflect number of items completed in 90 s.
b The values reflect proportional slowdown in the Stroop task, from the neutral to incongruent condition.
c The values reflect recall scores on the reading span task.
d Extended Range Vocabulary Test—Form V2. The values reflect correct items minus ¼ point deduction for errors.
*Age group differences were statistically significant in independent-samples t -tests ( p < .05).
Age group . | Digit symbol a, * . | Stroop b, * . | WM c, * . | ERVT d, * . |
---|---|---|---|---|
Young | 69.66 (8.91) | 0.29 (0.12) | 23.66 (4.34) | 07.23 (3.64) |
Older | 58.83 (8.75) | 0.45 (0.19) | 20.50 (3.03) | 14.78 (6.20) |
Age group . | Digit symbol a, * . | Stroop b, * . | WM c, * . | ERVT d, * . |
---|---|---|---|---|
Young | 69.66 (8.91) | 0.29 (0.12) | 23.66 (4.34) | 07.23 (3.64) |
Older | 58.83 (8.75) | 0.45 (0.19) | 20.50 (3.03) | 14.78 (6.20) |
Notes : WM = working memory. Values reflect average score per group; standard deviations are shown in parentheses.
a Wechsler Adult Intelligence Scale-Revised (WAIS-R) Digit Symbol Substitution. The values reflect number of items completed in 90 s.
b The values reflect proportional slowdown in the Stroop task, from the neutral to incongruent condition.
c The values reflect recall scores on the reading span task.
d Extended Range Vocabulary Test—Form V2. The values reflect correct items minus ¼ point deduction for errors.
*Age group differences were statistically significant in independent-samples t -tests ( p < .05).
Trial type . | . | Extra list a . | Positive lag a . | Lag −1 a . | Pooled negative lag a . | . |
---|---|---|---|---|---|---|
Inhibitory function . | Unflanked a . | Access: reactive . | Access: proactive . | Deletion: proactive (low PI) . | Deletion: proactive (high PI) . | Restraint: proactive b . |
Young | 644 (71) | 707 (73) | 691 (72) | 710 (71) | 731 (75) | 5.0% (0.03) |
Older | 785 (74) | 863 (90) | 835 (80) | 862 (92) | 907 (94) | 9.1% (0.04) |
Reliability Estimatesc | ||||||
Young | 0.91 | 0.92 | 0.91 | 0.93 | 0.91 | 0.46 |
Older | 0.94 | 0.95 | 0.90 | 0.93 | 0.91 | 0.58 |
Trial type . | . | Extra list a . | Positive lag a . | Lag −1 a . | Pooled negative lag a . | . |
---|---|---|---|---|---|---|
Inhibitory function . | Unflanked a . | Access: reactive . | Access: proactive . | Deletion: proactive (low PI) . | Deletion: proactive (high PI) . | Restraint: proactive b . |
Young | 644 (71) | 707 (73) | 691 (72) | 710 (71) | 731 (75) | 5.0% (0.03) |
Older | 785 (74) | 863 (90) | 835 (80) | 862 (92) | 907 (94) | 9.1% (0.04) |
Reliability Estimatesc | ||||||
Young | 0.91 | 0.92 | 0.91 | 0.93 | 0.91 | 0.46 |
Older | 0.94 | 0.95 | 0.90 | 0.93 | 0.91 | 0.58 |
Notes : PI = proactive interference.
a Median reaction times in milliseconds.
b Percentage of commission errors on the restraint trials. Standard deviations are in parentheses.
c Reliability was calculated by adjusting split-half correlations with the Spearman–Brown formula. All the reliability estimates were statistically significant ( p < .05).
Trial type . | . | Extra list a . | Positive lag a . | Lag −1 a . | Pooled negative lag a . | . |
---|---|---|---|---|---|---|
Inhibitory function . | Unflanked a . | Access: reactive . | Access: proactive . | Deletion: proactive (low PI) . | Deletion: proactive (high PI) . | Restraint: proactive b . |
Young | 644 (71) | 707 (73) | 691 (72) | 710 (71) | 731 (75) | 5.0% (0.03) |
Older | 785 (74) | 863 (90) | 835 (80) | 862 (92) | 907 (94) | 9.1% (0.04) |
Reliability Estimatesc | ||||||
Young | 0.91 | 0.92 | 0.91 | 0.93 | 0.91 | 0.46 |
Older | 0.94 | 0.95 | 0.90 | 0.93 | 0.91 | 0.58 |
Trial type . | . | Extra list a . | Positive lag a . | Lag −1 a . | Pooled negative lag a . | . |
---|---|---|---|---|---|---|
Inhibitory function . | Unflanked a . | Access: reactive . | Access: proactive . | Deletion: proactive (low PI) . | Deletion: proactive (high PI) . | Restraint: proactive b . |
Young | 644 (71) | 707 (73) | 691 (72) | 710 (71) | 731 (75) | 5.0% (0.03) |
Older | 785 (74) | 863 (90) | 835 (80) | 862 (92) | 907 (94) | 9.1% (0.04) |
Reliability Estimatesc | ||||||
Young | 0.91 | 0.92 | 0.91 | 0.93 | 0.91 | 0.46 |
Older | 0.94 | 0.95 | 0.90 | 0.93 | 0.91 | 0.58 |
Notes : PI = proactive interference.
a Median reaction times in milliseconds.
b Percentage of commission errors on the restraint trials. Standard deviations are in parentheses.
c Reliability was calculated by adjusting split-half correlations with the Spearman–Brown formula. All the reliability estimates were statistically significant ( p < .05).
Access
The residual scores for the access function were subjected to a 2×2 mixed factorial ANOVA with age group (young, older) as the between-subjects factor and flanker type (extra list [access reactive], positive lag [access proactive]) as the within-subjects factor. There was a nonsignificant main effect of flanker type, F (1, 44) = 0.58, p = .79, MSE = 0.34, η 2 = .00, a nonsignificant effect of age group, F (1, 44) = 2.3, p = .13, and a nonsignificant interaction of age group and flanker type, F (2, 88) = 0.57, p = .81, η 2 = .00. The effect sizes between the two age groups were small for access reactive ( d = −.28) and access proactive ( d = −.45), see Table 3 .
Inhibitory function . | Young . | Older . | Age-related effect size (Cohen’s D) . | Cohen’s D effect size interpretation . |
---|---|---|---|---|
Access (reactive) | −0.05 (0.92) | 0.22 (0.92) | −0.28 | Small |
Access (proactive) | −0.05 (0.67) | 0.30 (0.89) | −0.45 | Small |
Deletion (proactive; low PI) | −0.03 (0.84) | 0.25 (0.87) | −0.32 | Small |
Deletion (proactive; high PI) | −0.17 (0.86) | 0.38 (0.90) | −0.63 | Medium |
Restraint (proactive) | −0.35 (0.68) | 0.39 (1.17) | −0.80 | Large |
Inhibitory function . | Young . | Older . | Age-related effect size (Cohen’s D) . | Cohen’s D effect size interpretation . |
---|---|---|---|---|
Access (reactive) | −0.05 (0.92) | 0.22 (0.92) | −0.28 | Small |
Access (proactive) | −0.05 (0.67) | 0.30 (0.89) | −0.45 | Small |
Deletion (proactive; low PI) | −0.03 (0.84) | 0.25 (0.87) | −0.32 | Small |
Deletion (proactive; high PI) | −0.17 (0.86) | 0.38 (0.90) | −0.63 | Medium |
Restraint (proactive) | −0.35 (0.68) | 0.39 (1.17) | −0.80 | Large |
Note : PI = proactive interference. Standardized residual scores were derived by partialling out the variance from the unflanked trials. Standard deviations are in parentheses. A standardized residual score is interpreted similarly as a Z -score. For example, a negative standardized residual score of −0.17 for young adults in the deletion proactive (high PI) condition represents faster RTs of 0.17 SD units, after partialling out the variance from the unflanked trials.
Inhibitory function . | Young . | Older . | Age-related effect size (Cohen’s D) . | Cohen’s D effect size interpretation . |
---|---|---|---|---|
Access (reactive) | −0.05 (0.92) | 0.22 (0.92) | −0.28 | Small |
Access (proactive) | −0.05 (0.67) | 0.30 (0.89) | −0.45 | Small |
Deletion (proactive; low PI) | −0.03 (0.84) | 0.25 (0.87) | −0.32 | Small |
Deletion (proactive; high PI) | −0.17 (0.86) | 0.38 (0.90) | −0.63 | Medium |
Restraint (proactive) | −0.35 (0.68) | 0.39 (1.17) | −0.80 | Large |
Inhibitory function . | Young . | Older . | Age-related effect size (Cohen’s D) . | Cohen’s D effect size interpretation . |
---|---|---|---|---|
Access (reactive) | −0.05 (0.92) | 0.22 (0.92) | −0.28 | Small |
Access (proactive) | −0.05 (0.67) | 0.30 (0.89) | −0.45 | Small |
Deletion (proactive; low PI) | −0.03 (0.84) | 0.25 (0.87) | −0.32 | Small |
Deletion (proactive; high PI) | −0.17 (0.86) | 0.38 (0.90) | −0.63 | Medium |
Restraint (proactive) | −0.35 (0.68) | 0.39 (1.17) | −0.80 | Large |
Note : PI = proactive interference. Standardized residual scores were derived by partialling out the variance from the unflanked trials. Standard deviations are in parentheses. A standardized residual score is interpreted similarly as a Z -score. For example, a negative standardized residual score of −0.17 for young adults in the deletion proactive (high PI) condition represents faster RTs of 0.17 SD units, after partialling out the variance from the unflanked trials.
Deletion
The residual scores for the deletion function were subjected to a 2×2 ANOVA with age group as the between-subjects factor, and flanker type (lag −1 [deletion proactive, low PI], pooled negative lag [deletion proactive, high PI]) as the within-subjects factor. There was no significant main effect of flanker type, F (1, 44) = 0.18, p = .67, MSE = 53.45, η 2 = .00. The main effect of age group was also not significant, F (1, 44) = 1.93, p = .17, but the interaction of age group and flanker type was significant, F (2, 88) = 3.98, p = .04, η 2 = .10. Bonferroni corrected independent samples t -tests revealed age-invariance in the low PI condition, t (44) = −1.09, p = .28 with a small effect size ( d = −.32), and a significant age-related difference favoring the young adults in the high PI condition, t (44) = −2.11, p = .04, with a medium effect size ( d = −.63).
Restraint
The commission errors on restraint trials served as an index of restraint-type inhibition. Compared to young adults, older adults committed more errors on restraint trials, t (44) = −3.71, p < .01. To control for individual and age differences in overall error rates, we computed standardized residual restraint scores by partialling out the variance from other commission errors (i.e., “yes” or “no” errors on all other trial types) in the SFT. With this conservative approach, older adults ( M = 0.38, SD = 1.14) nevertheless had higher Restraint scores compared to young adults ( M = −0.34, SD = 0.65), t (44) = −2.66, p < .01. The effect size between the two age groups for the restraint-type commission errors (residual scores) was large ( d = −.80).
Discussion
Our main objective was to examine the relative age-sensitivity of access, deletion, and restraint functions of inhibition, taking into consideration the underlying control processes. We hypothesized that the older adults should be at a relative disadvantage in those inhibitory functions that require a larger degree of proactive control. Consistent with these predictions, we found the largest and smallest age-related differences in restraint and access (reactive), respectively. Moreover, moderate age-related differences were observed in the deletion condition that involved high PI than in a condition that involved low PI. Overall, the current findings support the contention that varying degrees of proactive control is a key determinant of age effects in inhibition.
A key finding from the current experiment was that, compared to young adults, older adults showed the largest declines in restraint ( d = −.80). We note that this age difference remained even after controlling for overall errors. The current restraint findings are consistent with previous research ( Feyereisen & Charlot, 2008 ) in which similar age-related effect sizes were observed in the Stroop task ( d = −.80) and the Hayling task ( d = −.98). Age-related differences in the restraint function have also been reported elsewhere (e.g., Butler et al., 1999 ; Davidson et al., 2003 ; Kramer et al., 1994 ).
However, to our knowledge, this is the first study to report significant age-related differences in a go/no-go task (cf., Vallesi & Stuss, 2010 ). The age differences presently observed may be due to our operationalization of restraint in the SFT. That is, our participants withheld their response to the restraint trials in addition to updating the target information in their WM. By Braver and colleagues’ DMC framework, a go/no-go paradigm with WM updating demands would require a greater degree of proactive control than a more basic variant of the paradigm (e.g., Vallesi & Stuss, 2010 ). Consistent this argument, we found a negative correlation between the commission errors on the restraint trials and reading span WM for both age groups ( Table 4 ).
Measures . | 1 . | 2 . | 3 . | 4 . | 5 . | 6 . |
---|---|---|---|---|---|---|
1.Access (reactive) | — | 0.23 | 0.70** | 0.66** | 0.19 | −0.23 |
2.Access (proactive) | 0.41* | — | 0.50** | 0.15 | 0.28 | −0.34 |
3.Deletion (proactive; low PI) | 0.62* | 0.38* | — | 0.71** | 0.10 | −0.30 |
4.Deletion (proactive; high PI) | −0.11 | −0.34 | 0.34 | — | 0.16 | −0.17 |
5.Restraint (proactive) | −0.08 | −0.11 | −0.11 | −0.18 | — | −0.46* |
6.WM | −0.20 | 0.15 | −0.18 | 0.12 | −0.41* | — |
Measures . | 1 . | 2 . | 3 . | 4 . | 5 . | 6 . |
---|---|---|---|---|---|---|
1.Access (reactive) | — | 0.23 | 0.70** | 0.66** | 0.19 | −0.23 |
2.Access (proactive) | 0.41* | — | 0.50** | 0.15 | 0.28 | −0.34 |
3.Deletion (proactive; low PI) | 0.62* | 0.38* | — | 0.71** | 0.10 | −0.30 |
4.Deletion (proactive; high PI) | −0.11 | −0.34 | 0.34 | — | 0.16 | −0.17 |
5.Restraint (proactive) | −0.08 | −0.11 | −0.11 | −0.18 | — | −0.46* |
6.WM | −0.20 | 0.15 | −0.18 | 0.12 | −0.41* | — |
Notes : PI = proactive interference, WM = working memory. Intercorrelations for young adults ( n = 24) are presented above the diagonal, and intercorrelations for older adults ( n = 23) are presented below the diagonal. Variables 1–4 are residual scores derived after partialling the variance from the baseline, unflanked condition. The values reflect recall scores on the reading span task.
* p < .05. ** p < .05.
Measures . | 1 . | 2 . | 3 . | 4 . | 5 . | 6 . |
---|---|---|---|---|---|---|
1.Access (reactive) | — | 0.23 | 0.70** | 0.66** | 0.19 | −0.23 |
2.Access (proactive) | 0.41* | — | 0.50** | 0.15 | 0.28 | −0.34 |
3.Deletion (proactive; low PI) | 0.62* | 0.38* | — | 0.71** | 0.10 | −0.30 |
4.Deletion (proactive; high PI) | −0.11 | −0.34 | 0.34 | — | 0.16 | −0.17 |
5.Restraint (proactive) | −0.08 | −0.11 | −0.11 | −0.18 | — | −0.46* |
6.WM | −0.20 | 0.15 | −0.18 | 0.12 | −0.41* | — |
Measures . | 1 . | 2 . | 3 . | 4 . | 5 . | 6 . |
---|---|---|---|---|---|---|
1.Access (reactive) | — | 0.23 | 0.70** | 0.66** | 0.19 | −0.23 |
2.Access (proactive) | 0.41* | — | 0.50** | 0.15 | 0.28 | −0.34 |
3.Deletion (proactive; low PI) | 0.62* | 0.38* | — | 0.71** | 0.10 | −0.30 |
4.Deletion (proactive; high PI) | −0.11 | −0.34 | 0.34 | — | 0.16 | −0.17 |
5.Restraint (proactive) | −0.08 | −0.11 | −0.11 | −0.18 | — | −0.46* |
6.WM | −0.20 | 0.15 | −0.18 | 0.12 | −0.41* | — |
Notes : PI = proactive interference, WM = working memory. Intercorrelations for young adults ( n = 24) are presented above the diagonal, and intercorrelations for older adults ( n = 23) are presented below the diagonal. Variables 1–4 are residual scores derived after partialling the variance from the baseline, unflanked condition. The values reflect recall scores on the reading span task.
* p < .05. ** p < .05.
A second key finding was the pattern of small to moderate age-related declines found in the deletion conditions. As mentioned previously, deletion refers to the suppression of no-longer relevant information from WM by resisting PI from previous task-relevant information. We operationalized deletion in terms of lag −1 flankers (low PI) and pooled negative lag flankers (high PI), expecting that a larger degree of proactive control is required in the pooled negative lag flanker condition than in the lag −1 condition to resolve the buildup of PI. Accordingly, we found longer RTs overall in the pooled negative lag flanker condition compared to the lag −1 flanker condition ( Table 1 ). Consistent with our predictions, we observed age invariance on the lag −1 flanker condition, replicating previous research (e.g., Li & Dupuis, 2008 ), and an age group difference in the pooled negative lag flanker condition favoring the young adults. The age-related effect size found in the pooled negative flanker condition ( d = −.63) parallels the one reported by Feyereisen and Charlot (2008) , who used a listening span task ( d = −.68) to measure deletion. Together, the pattern of deletion findings suggests that older adults may have sufficient inhibitory strength to handle distraction from immediately preceding trials, but this inhibitory strength declines more quickly over time for older adults than younger adults. Future work on the time course differences between young and older adults is needed to address this possibility directly.
The third key finding of the current experiment was that no age-related differences were found in either operationalization of access-type inhibition. We argued that access operates both by proactive and reactive control processes depending on the type of irrelevant information encountered. To test this assumption, we measured access by using two types of irrelevant information, positive lag and extra list flankers, to elicit proactive and reactive control processes, respectively. With regard to age-equivalent effects in extra-list flanked trials, our findings are consistent with past research ( Paxton et al., 2008 ) in that reactive control is largely preserved in older adults. With regard to proactive control, we predicted age-related differences in the positive lag flanker condition, but found no statistically significant age-related differences. Nevertheless, observing a larger age-related effect size ( d = −.45) in the positive lag condition than in the extra-list condition ( d = −.28) suggests that proactive control is more age-sensitive than reactive control.
To our knowledge this is the first study to examine the relative age-sensitivity across different inhibitory functions within a single cognitive task. Assuming that complex cognitive tasks recruit a blend of proactive and reactive control processes, our findings suggest that the age-sensitivity of different inhibitory functions is influenced by the extent to which proactive control is required, and not categorically predicted by the type of material or action to be suppressed. The WM demands that are added to these control process requirements may further influence the magnitude of age differences in any given inhibitory task. This view suggests that under different task conditions, the effect sizes of age-related differences in different inhibitory functions are likely to change. However, we believe that the relative magnitudes across access, deletion, and restraint should be preserved as long as the task-specific characteristics are held constant across conditions.
The taxonomy of three inhibitory functions has been foundational in describing and examining older adults’ susceptibility to different types of irrelevant information. The DMC framework proposed by Braver and colleagues (2007) offers a different, noncategorical, approach to explaining how age-related differences in inhibitory efficiency can vary as a function of a single control process namely, proactive control. The results from the current experiment suggest that it is not the presence or absence of proactive control per se, but the relative degree of proactive control required in a given task, that is critical in explaining age-related differences in inhibition.
In the interest of parsimony, with the present results, we argue that a fruitful direction of research is to consider age-related differences in inhibitory efficiency in terms of the specific blend of control and WM demands imposed within any given task. At an empirical level, the current experiment provides a step forward in understanding why age-related differences are not uniform across different inhibitory functions. Finally, on a more methodological level, the current experiment highlights the utility of measuring different inhibitory functions within a single cognitive task.
Funding
This work was supported by grant funds from the Natural Sciences and Engineering Research Council (NSERC: 238627-2011) awarded to K. Z. H. Li and Canadian Institute of Health Research (CIHR) and Fonds de recherche du Québec – Nature et technologies (FQRNT) Master’s scholarships awarded to K. K. Vadaga.
References
Author notes
Decision Editor: Bob G. Knight, PhD