Age differences in inhibitory and working memory functioning: limited evidence of system interactions

ABSTRACT

Debate persists regarding the nature of age-related deficits in inhibition, and whether inhibitory functioning depends on working memory systems. The current research aimed to measure age-related differences in inhibition and working memory, characterize the relationship between inhibitory functions and working memory performance, and determine how these relationships are affected by age. Toward these ends, we measured performance on a range of established paradigms in 60 young adults (18–30 years) and 60 older adults (60–88 years). Our findings support age-related increases in reflexive inhibition (based on the fixation offset effect and inhibition of return) and age-related decrements in volitional inhibition (based on several paradigms: antisaccade, Stroop, flanker, and Simon). This evidence of stronger reflexive inhibition combined with weaker volitional inhibition suggests that age-related deterioration of cortical structures may allow subcortical structures to operate less controlled. Regarding working memory, older adults had lower backward digit scores and lower forward and backward spatial scores. However, of the 32 analyses (16 in each age group) that tested for dependence of inhibitory functioning on working memory functioning, only one (in young adults) indicated that inhibition performance significantly depended on working memory performance. These results indicate that inhibition and working memory function largely independently in both age groups, and age-related working memory difficulties cannot account for age-related declines in inhibitory control.

PUBLIC SIGNIFICANCE STATEMENT

• This study indicates that, compared to young adults, older adults have difficulties with both working memory and volitional inhibition but show stronger reflexive inhibition.

• The pattern of age differences shown in this study suggests that the balance of power shifts from cortical to subcortical brain structures as we age, which may reflect cortical brain regions deteriorating more rapidly as we age, thus limiting their ability to control subcortical brain regions.

• In contrast to current theory, in this study the difficulties that older adults showed with volitional inhibition could not be explained by failing working memory systems, and both young and older adults showed very limited evidence that inhibitory functioning depends on working memory.

KEYWORDS:

• Aging
• selective attention
• inhibition
• memory systems
• cognitive functioning

Inhibition enables humans to prevent reflexive responses in favor of intentional behaviors and facilitates effective responses in the presence of distracting information (Easton, 1972). Less effective inhibitory mechanisms can compromise the ability to strategically control behavior (Friedman & Miyake, 2004; Houghton & Tipper, 1996) and can adversely affect the performance of daily activities (Anstey et al., 2005, 2009; Hoffman et al., 2005). A number of studies have indicated that inhibition tends to decline with adult aging, affecting a variety of functions (Zanto & Gazzaley, 2014). For example, compared to young adults, older adults tend to have more difficulty inhibiting distracting identity-based information during a flanker task (Colcombe et al., 2005; Machado et al., 2009), inhibiting distracting spatial information during a Simon task (Castel et al., 2007; van der Lubbe & Verleger, 2002), and inhibiting word reading during a Stroop task (Henik et al., 1993; Troyer et al., 2006). In addition, older adults tend to have more difficulty suppressing reflexive eye movements toward a visual stimulus appearing in the periphery (prosaccade) and take longer to initiate eye movements away from a visual stimulus appearing in the periphery (antisaccade; Brett & Machado, 2017; Chen & Machado, 2016, 2017).

Other studies, however, have reported a lack of age differences in inhibition. For example, Pratt et al. (1997) found that the magnitude of the fixation offset effect (FOE) did not differ between young and older adults. Likewise, in Hartley and Kieley (1995), inhibition of return (IOR) – an inhibitory mechanism for optimizing visual search – was similar in young and older adults (see also Faust & Balota, 1997; McCrae & Abrams, 2001). Moreover, a meta-analysis by Rey-Mermet and Gade (2018) investigating age differences on 10 tasks commonly used to measure inhibition only found age-related impaired performance on two of them (go/no-go and stop-signal), which led the authors to question the theory of a general inhibition deficit in older adults (for more studies that observed no age differences in measures of inhibitory control, see Borella et al., 2009; Eenshuistra et al., 2004; Fernandez Duque & Black, 2006; Langley et al., 2005; Rey-Mermet et al., 2018; Verhaeghen, 2011; Waszak et al., 2010).

Bugg et al. (2007) suggested that the observed age differences in some tasks (e.g., Stroop) could be attributed to a general slowing in processing speed, in addition to age-related changes in inhibitory control. It has also been suggested that aging only affects volitional inhibition, hence IOR and FOE may be unaffected because they are non-conscious in nature (Nieuwenhuis et al., 2000). In a similar vein, in relation to adult aging being associated with atrophy of the frontal lobes in particular (e.g., see Hedden & Gabrieli, 2004; Vinke et al., 2018), it has been suggested that performance on frontally mediated volitional inhibition tasks would be more adversely affected by adult aging than performance on more reflexive inhibition tasks (Andres et al., 2008; Kramer et al., 2001). Given that performance on tasks that tap volitional inhibition tends to rely more on cortical structures (e.g., Stroop, flanker; Chambers et al., 2009; Tsuchida & Fellows, 2013), whereas performance on tasks that tap reflexive inhibition tends to be more subcortically mediated (e.g., FOE and IOR paradigms; Machado & Rafal, 2004a; Munoz & Fecteau, 2002; Sapir et al., 1999, 2002), differences in age-related deterioration of cortical versus subcortical structures (Aanerud et al., 2012; Walhovd et al., 2011) could help explain why we do not see an across-the-board decline in inhibitory functioning. Thus, to fully characterize age-related changes in inhibitory functioning, assessment of a wide range of measures is needed, with attention to whether the relevant inhibitory process is considered to be predominantly cortically or subcortically mediated.

There is evidence to suggest that inhibitory processes depend on working memory (WM) functioning (e.g., Andrés et al., 2004; Borella et al., 2007; Conway et al., 1999; Gazzaley et al., 2007; Roberts & Pennington, 1996; Roberts et al., 1994; Verhaeghen & Basak, 2005). WM is a system for temporarily processing task-relevant information during complex cognitive tasks and involves the manipulation of information in short-term memory (STM) stores (Baddeley, 1992). According to Eenshuistra et al. (2004) and McCabe et al. (2005), inhibitory control uses a significant amount of WM capacity, and successful inhibition is dependent on the WM demands of the task at hand and the available WM resources needed to execute the appropriate response (see also Boucher et al., 2021; Spangler & Friedman, 2017). WM abilities could therefore determine inhibitory control performance. Indeed, there is evidence to suggest that individuals with better WM functioning are able to use more attentional resources and therefore outperform those with worse WM functioning on tasks that require inhibition (Kane & Engle, 2003; Kane et al., 2001; Rosen & Engle, 1998; Unsworth et al., 2004). For example, studies assessing young and middle-aged adults with high versus low WM capacity reported that, compared to participants with high WM capacity, participants with low WM capacity experienced more interference from distractors and were slower and less accurate on measures of both reflexive and volitional inhibition (Kane & Engle, 2003; Kane et al., 2001; Kiefer et al., 2005; Minamoto et al., 2010).

In relation to this, it has been suggested that age-related difficulties with inhibitory processes are due to age-related declines in the WM system, rather than impaired inhibitory control in its own right (Eenshuistra et al., 2004; McCabe et al., 2005), as WM functioning also tends to deteriorate with adult aging (Braver & West, 2008; Cansino et al., 2013; Hurley & Machado, 2018). For example, in Borella et al. (2007) study on age differences in WM and inhibition, they reported that older adults showed worse WM performance than young adults, and furthermore that old-old (75–86 years) participants performed significantly worse than young-old participants (65–74 years), which suggests progressively worsening WM performance with advancing age. In addition, compared to young and young-old adults, old-old adults suffered from significantly more intrusion errors, indicating inhibition difficulties. Regarding potential interrelationships between WM and inhibition systems, age could be a relevant factor (e.g., see Eenshuistra et al., 2004). Given the evidence of age-related declines in WM and inhibition systems, it may be that their interrelationships change as we age, presumably with shifts toward stronger interrelationships in older adults due to more limited WM resources, assuming inhibitory functioning depends on the WM system.

Another factor that needs to be considered is that the relationship between WM and inhibition can be domain-dependent. WM can be broken down into subsystems distinguishable by the nature of the content being processed. The most common distinctions in the literature are between verbal and spatial material, with verbal WM linked to the left hemisphere and spatial WM to the right hemisphere (Hurley & Machado, 2018; Owen et al., 2005; Ramos et al., 2020; Ramos & Machado, 2021). Regarding associations with inhibitory functioning, different types of WM appear to be linked to different types of inhibition. For example, Castel et al. (2003) discovered that IOR is partially mediated by spatial, but not verbal, WM. Similarly, using a dual-task paradigm, Chou and Yeh (2008) found that IOR was affected by a spatial but not a non-spatial WM secondary task. Clouter et al. (2015) discovered that the Simon effect is affected by both verbal and spatial WM secondary tasks, although the extent of the effect was much larger during the spatial dual-task conditions. In contrast, in McNab et al. (2008) the flanker effect was strongly associated with spatial but not verbal WM. It is therefore important to assess different types of WM to better understand how each of them relates to different inhibitory functions, especially as age-related deficits in WM appear to be domain-specific (Jenkins et al., 2000), with performance on spatial WM tasks deteriorating at a faster rate than performance on verbal WM tasks (Hale et al., 2011).

Present study

The aims of the current study were to: 1) measure age-related differences in inhibition and WM function with consideration of inhibition type (reflexive/subcortical vs. volitional/cortical) and WM domain (verbal vs. spatial); and 2) explore the relationship between each domain of WM and each type of inhibition, and whether these relationships are influenced by adult aging. By comparing the performance of samples of young and older adults on a range of tasks that differentially tap verbal versus spatial WM and reflexive/subcortical versus volitional/cortical inhibition, the current study aimed to provide a detailed characterization of age-related changes in inhibitory functioning and to shed further light on possible interrelationships between WM and inhibition. To these ends, we tested groups of young and older adults using established paradigms to yield a comprehensive assessment of the effects of healthy aging on inhibitory and WM functioning.

To measure reflexive/subcortical inhibitory functioning, we used fixation offset and IOR tasks. Regarding the FOE, the presence of a stimulus at fixation suppresses the eye movement system, and this stimulus-induced suppression must be overcome in order to make a saccade, resulting in longer latencies to commence movement compared to when no stimulus is present at fixation (Machado & Rafal, 2000a, 2000b). Weaker inhibition is associated with difficulty initiating saccades during active fixation, resulting in a larger FOE. Regarding IOR, this mechanism makes visual search more efficient by biasing against the return of attention (and gaze) to previously attended locations (Lupianez et al., 2006; Machado & Rafal, 2004c). Weaker inhibition is associated with a smaller IOR effect, rendering individuals more likely to revisit previously attended locations, which results in a less efficient visual search (Hartley & Kieley, 1995).

To measure volitional/cortical inhibitory functioning, we used flanker (Eriksen & Eriksen, 1974), Simon (Hommel, 2011), and Stroop (Stroop, 1935) tasks. In these tasks, selectively ignoring distractors requires participants to inhibit processing of the distracting information (Machado & Devine, 2019; Machado et al., 2007, 2011, 2013; Wyatt & Machado, 2013a, 2013b). The additional time it takes to inhibit processing associated with the distracting information is reflected in the compatibility effect (Castel et al., 2007; Colcombe et al., 2005), with larger effects indicating weaker inhibitory control (evidenced by the absence of negative compatibility effects at longer distractor-target delays; Machado et al., 2009). In addition, we measured the rate of reflexive errors during an antisaccade task as another indicator of cortically mediated inhibitory functioning. This task requires participants to inhibit a reflexive response to a peripheral stimulus and initiate a voluntary saccade to the opposite side. Inhibition of this reflexive response is cortically mediated by the frontal eye fields (Machado & Rafal, 2004a; Neggers et al., 2012; Rafal et al., 2000; Vernet et al., 2014), and reflexive errors toward the target indicate a failure of volitional inhibition.

We used forward, backward, and addition digit span tasks (based on the subscale of the Wechsler Memory Scale III) to measure the functioning of the verbal WM system, and forward and backward spatial span tasks to measure the functioning of the spatial WM system (Forsyth et al., 2016; White et al., 2018, 2019, 2020). Although performance on forward span tasks is often considered a measure of STM and performance on backward and addition span tasks a measure of WM as these require manipulation of information in STM stores (Hester et al., 2004), it has been argued that all of these span tasks tap components of the WM system (albeit to different degrees; Hale et al., 2011). For this reason, and to fit with previous research comparing inhibitory functioning in participants with varying levels of WM ability, in the current study we used combined forward and backward scores as a measure of WM ability (Kiefer et al., 2005). Consistent with previous findings, we expected performance deficits in older compared to young adults to be greater for the spatial than verbal WM tasks (Hale et al., 2011).

Finally, regarding potential links between WM and inhibition, we anticipated that higher WM scores would be associated with better performance in inhibition tasks. In consideration of age-related declines in both WM and inhibitory functioning, we also postulated that functional declines in the relevant systems could impact interrelationships, as evidenced by WM-inhibition interactions that differ by age group. In addition, with respect to aging particularly affecting cortical structures, we postulated that possible interactions between WM and inhibition would be more pronounced on tasks that require volitional as opposed to reflexive inhibition, as both WM and volitional inhibition rely more on cortical structures (Chambers et al., 2009; Zaehle et al., 2011), whereas reflexive inhibition relies more on subcortical structures (Machado & Rafal, 2004a; Sapir et al., 1999, 2002).Figure 1 provides a visual representation of the proposed conceptual framework and Figure 2 illustrates the categorization of each cognitive measure in the current study.

Figure 1. Conceptual framework for working memory and inhibition system interactions.

Arrows indicate system interactions suggested in the literature. Dashed lines indicate a weaker link. Orange demarcates functions expected to show age-related deterioration, with the thicker box outline indicating greater decline.

Figure 2. Categorization of the cognitive measures in relation to the conceptual framework.

Working memory is cortically mediated.

Method

Transparency and openness

The University of Otago Human Ethics Committee (#03/171) approved this study, and participants provided written informed consent before taking part. We describe all methods in detail but did not pre-register the study. Code and data are available upon request, in accordance with the ethical approval.

Participants

Sixty neurologically healthy young adults (mean age in years = 21, SD = 3, range = 18–30; 22 males; 55 right-handed; mean years of education = 15) recruited at the University of Otago participated in association with a course. Sixty neurologically healthy older adults (mean age in years = 71, SD = 6, range = 60–88; 22 males; 54 right-handed; mean years of education = 14) recruited from the local community participated and received NZ$20 reimbursement. The young and older adult groups did not significantly differ in years of education (p > .3) or proportion of right-handed participants (p > .7). All older adults scored at least 25 out of 30 on the Mini-Mental State Examination (Folstein et al., 1975), suggesting none had dementia. Two additional young adults and four additional older adults participated but were excluded due to computer issues (n = 2), eye-tracking difficulties (n = 3), or a history of neurological issues (n = 1). These six participants will not be discussed further.

Materials and procedure overview

The session took approximately 1.5 h, during which participants performed 12 cognitive tasks that stemmed from seven paradigms: fixation offset (pro- and antisaccade tasks), IOR (central and peripheral saccade signal tasks), flanker, Simon, Stroop, digit span (forward, backward, and addition tasks), and spatial span (forward and backward tasks). The order of the tasks was counterbalanced across participants within each age group with the constraint that the following were yoked together due to methodological overlap: fixation offset and IOR (both involved eye tracking); flanker and Simon (both involved the same stimulus-response mapping); digit span forward, backward, and addition; and spatial span forward and backward. During testing, participants sat 57 cm from a computer monitor in a dimly lit room, with viewing distance maintained via a chinrest (Applied Science Laboratories, Bedford, MA). For all tasks, stimuli were presented against a black background via MATLAB (The MathWorks, Natick, MA) and the Psychophysics Toolbox (Brainard, 1997; Pelli, 1997). Except for the memory tasks, participants were instructed to respond as quickly as they could without compromising accuracy; all conditions were equally probable, counterbalanced, and randomized; and an error was recorded and a 1000 Hz tone sounded for 500 ms if the reaction time (RT) was less than 80 ms or more than 1500 ms, or an incorrect response was made.

Fixation offset and inhibition of return paradigms

Participants wore a lightweight head-mounted infrared scleral reflectance device (Model 310; Applied Science Laboratories, Bedford, MA), which measured eye position at an approximate rate of 1100 Hz using the Data Acquisition Toolbox (The MathWorks, Natick, MA). An eye movement response was defined as the first time since the onset of the saccade signal that eye velocity exceeded 50°/s for at least 2.5°. To determine the start of the eye movement response, the program searched backward from this point to the point at which eye velocity reached 10°/s. Reaction time (RT) was calculated as the timing delay between saccade signal onset and the start of the eye movement response. An error was recorded and a 1000 Hz tone sounded for 500 ms if the eye position was not within 3° of the fixation dot at the start of the eye movement, RT was less than 80 ms or more than 1500 ms, or the eyes moved in the incorrect direction. In addition to computer analysis of eye movement responses, the experimenter monitored the participant’s eye position overlaid on the task display on a separate screen. During the practice trials, the experimenter provided verbal error feedback, and during the test trials the experimenter marked trials in which the participant blinked or failed to maintain fixation until the target appeared or an external disruption occurred. All trials marked as erroneous by the automated program or rejected by the experimenter were excluded from RT analyses, which included correct responses only. A black screen displayed for 500 ms between trials.

Fixation offset tasks

Two light gray unfilled 1° squares were displayed 8° to the left and right of the center, throughout these tasks. Each trial began with a 0.3° white central fixation dot displayed at center for 500 ms, after which a white cue appeared at center, occluding the fixation dot. The cue was either a single arrowhead pointing toward the impending saccade signal or a double arrowhead pointing in both directions. After 200, 400, or 600 ms, a white filled 0.5° square (saccade signal) appeared in the center of one of the two light gray squares for 1500 ms. In the fixation offset condition, the central cue disappeared when the saccade signal appeared. In the fixation overlap condition, the central cue remained on screen until the end of the trial. Figure 3 illustrates two example trials. There were two tasks: pro- and antisaccade. In the prosaccade task, participants were instructed to move their eyes to the white square. In the antisaccade task, participants were instructed to move their eyes away from the white square to the opposite box. Eye position recording commenced when the cue appeared and continued until the trial ended. Participants completed 10 practice trials followed by 96 test trials of each task (pro- and antisaccade).

Figure 3. Example trial sequences from the fixation offset paradigm.

Note. After the fixation period, a cue appeared (either a single arrowhead pointing toward the impending saccade signal or a double arrowhead pointing in both directions) followed by the saccade signal (filled white square). In half the trials, the cue disappeared when the saccade signal appeared (offset trials), and in the other half, the cue remained present (overlap trials). In the prosaccade task, participants were instructed to move their eyes to the saccade signal. In the antisaccade task, participants were instructed to move their eyes away from the saccade signal to the opposite box.

Inhibition of Return Tasks

Two light gray unfilled 1° squares were displayed 8° to the left and right of the center, throughout these tasks. Each trial began with a 0.3° white central fixation dot displayed for 500 ms, after which the lines of one of the light gray squares thickened and turned white for 200 ms. Following cue offset, the fixation dot remained on screen for a further 700, 900, or 1100 ms before one of the two possible saccade signals appeared, depending on the task: a central arrowhead pointing in the direction to saccade (central saccade signal task) or a peripheral white filled 0.5° square (centered in one of the two light gray squares) indicating the saccade goal (peripheral saccade signal task). Figure 4 illustrates an example trial from each task. For each task, on 50% of the trials the saccade goal was the cued square (cued), and on the other 50% of the trials the saccade goal was the other square (uncued). For each of the two tasks (central and peripheral saccade signals), participants completed 10 practice trials followed by 60 test trials.

Figure 4. Example trial sequences from the inhibition of return paradigm.

Note. After the initial fixation period, a cue appeared (either the left or right box brightened) followed by a second fixation period and then the saccade signal (filled white square in the peripheral saccade signal task; central arrowhead in the central saccade signal task). In the peripheral saccade signal task, participants were instructed to look at the filled white square. In the central saccade signal task, participants were instructed to look at the box that the arrowhead points toward. The saccade goal was cued on half the trials.

Flanker and Simon paradigms

For flanker and Simon, participants responded by pressing a button using a two-button DirectIN Response Box (Empirisoft, New York). To determine the stimulus-response mapping, for each participant, the computer program randomly selected three consonants and assigned two of them to the response buttons (e.g., “F” might be assigned to the left button and “P” to the right button), to be used for flanker and Simon; the stimulus-response mapping was depicted on the screen at the beginning of the practice and test trials. The third consonant, which was not assigned to a response button, could appear as a distractor during the flanker task.

Flanker Task

Each trial began with a 0.3° white central fixation dot displayed at center for 500 ms, after which two white uppercase consonants (Arial font, 1° in height) appeared simultaneously: a target consonant in the center of the screen (occluding the fixation dot) and a distractor consonant 1° above or below the target (to avoid effects of spatial compatibility with the response buttons; Machado & Franz, 2004). Participants were instructed to indicate the identity of the central letter by pressing the assigned button. The target could be either of the consonants assigned to a response button, and the distractor could be any one of the three consonants. Thus, the identity of the distractor could match the identity of the target (compatible), not match the identity of the target and not be associated with a response (neutral), or not match the identity of the target and be associated with the incorrect response (incompatible). Figure 5 shows examples displays. A black screen displayed for 1000 ms between trials. Participants completed 10 practice trials followed by 84 test trials.

Figure 5. Example displays illustrating the three conditions in the flanker paradigm.

Note. In this example, K was assigned to the left button, and S was assigned to the right button. Participants were instructed to indicate the identity of the central letter by pressing the assigned button. The distractor letter was positioned either above or below the central letter. The compatibility conditions reflect the identity of the distractor.

Simon Task

Each trial began with a 0.3° white central fixation dot displayed at the center for 500 ms, after which a white uppercase consonant (Arial font, 1° in height) appeared either in the center of the screen, 2° to the left of center, or 2° to the right. Participants were instructed to indicate the identity of this letter by pressing the assigned button. The position of the consonant varied randomly in relation to the side of the correct response, thus making its position either compatible, neutral (central letter), or incompatible with the side of the correct response. Figure 6 shows examples displays. A black screen displayed for 1000 ms between trials. Participants completed 10 practice trials followed by 84 test trials.

Figure 6. Example displays illustrating the three conditions in the Simon paradigm.

Note. In this example, K was assigned to the left button, and S was assigned to the right button. Participants were instructed to indicate the identity of the letter by pressing the assigned button. The compatibility conditions refer to the position of the letter in relation to the side of the correct response.

Stroop paradigm

Each trial began with a 0.3° white central fixation dot displayed at the center for 500 ms, after which a word (Arial font, 1° in height) appeared in the center of the screen for 2000 ms. Participants were instructed to verbally report the color of the font. The possible font colors were red, orange, yellow, green, blue, and purple. The word could be a color word (“RED,” “ORANGE,” “YELLOW,” “GREEN,” “BLUE,” or “PURPLE”) or a neutral word (“ONE,” “ELEVEN,” “TWELVE,” “SEVEN,” “FOUR,” or “NINETY”), such that the word was compatible (e.g., the word “BLUE” in blue font), incompatible (e.g., the word “RED” in blue font), or neutral (e.g., the word “ONE” in blue font) with the color of the font. Figure 7 shows examples displays. When participants responded, a Cedrus SV-1 Smart Voice Key and I/O recorded RT; the experimenter entered the first letter of the participant’s response (r, o, y, g, b, or p) or “e” if an error occurred (such as a cough triggered the voice key). A black screen displayed for 500 ms between trials. Participants completed 5 practice trials followed by 72 test trials.

Figure 7. Example displays illustrating the three conditions in the stroop paradigm.

Note. Participants were instructed to verbally report the color of the font. A voice key was used to record reaction times. The compatibility conditions refer to the identity of the word in relation to the font color.

Digit span paradigm

Based on the digit span subscale of the Wechsler Memory Scale III, the three tasks (forward, backward, and addition) differed only in the response instructions and the specific digits in the sequences. Each sequence began with a 0.3° white central fixation dot displayed at the center for 500 ms, after which a sequence of white digits (Arial font, 1° in height) appeared one after the other in the center of the screen. Each digit in the sequence appeared for 1 s with no interstimulus interval. At the end of each sequence, a 400 Hz tone sounded for 500 ms and the white fixation dot reappeared, cueing participants to respond verbally. Participants were instructed for the forward task to report the digits in the same order in which they appeared, for the backward task to report the digits in the reverse order to which they appeared, and for the addition task to report the sum total of all digits in the sequence. After the experimenter keyed in the response, a black screen displayed for 1000 ms. Two sequences of each length were presented, starting with a length of two digits. If participants responded correctly to at least one of the sequences of a particular length, the sequence length increased by one digit. If participants responded incorrectly to both of the sequences of a particular length, the task terminated. The maximum sequence length was nine digits.

Spatial span paradigm

Based on the Corsi block-tapping test (Corsi, 1972; Kessels et al., 2000) the two tasks (forward and backward) differed only in the response instructions and the specific sequences. The display (see Figure 8) consisted of nine gray filled 4° squares positioned as per Kessels et al. (2000). Prior to each sequence, a 400 Hz tone sounded for 500 ms, and then a subset of the squares turned white for 1 s, one at a time, with no interstimulus interval. At the end of each sequence, a 400 Hz tone sounded for 500 ms, cueing the participant to respond by touching the relevant squares using a KeyTech MagicTouch touch screen; squares turned white when touched. Participants were instructed for the forward task to touch the squares in the same order in which they lit up, and for the backward task to touch the squares in the reverse order to which they lit up. After the participant responded, a black screen was displayed for 1000 ms. Two sequences of each length were presented, with the first comprising two squares. If participants responded correctly to at least one of the two sequences of a particular length, the sequence length increased by one square, else the task terminated. The maximum sequence length was nine squares.

Figure 8. Organization of the squares in the spatial span paradigm.

Note. Each sequence entailed a subset of the squares turning white for 1 s. Participants were instructed for the forward task to touch the squares in the same order in which they lit up, and for the backward task to touch the squares in the reverse order to which they lit up.

Statistical analyses

The measured variables of interest in the inhibition tasks were median RT for correct responses and, for the antisaccade task only, percentage of reflexive errors. The measured variable of interest in the verbal and spatial WM tasks was score, which was calculated by multiplying span (the longest sequence length for which the participant responded correctly to at least one of the two sequences; maximum = 9) by the number of correctly recalled sequences (maximum = 16), resulting in a maximum score of 144 (Kessels et al., 2000).

All statistical analyses were performed using SPSS version 25.0. To investigate age-group differences in inhibition, for fixation offset, IOR, flanker, Simon, and Stroop median correct RTs, we conducted mixed analyses of variance (ANOVAs) with trial type (offset, overlap or uncued, cued or compatible, incompatible) as the within-subjects variable and age group (young, old) as the between-subjects variable, done separately for each task, and for antisaccade reflexive errors we conducted a univariate ANOVA with age group (young, old) as the between-subjects variable. We also subjected the correct RT data to log transformations to help confirm the observed aging effects (see Bush et al., 1993; Faust et al., 1999). To investigate age-group differences in WM function, for each WM task we conducted a one-way ANOVA with age group as the between-subjects variable. Analyses in G*Power (Faul et al., 2007, 2009) indicated that with a sample size of 120 we had 80% power to detect medium-sized (η_p² ≥0.06) main effects of age group and trial type, as well as Trial Type × Age Group interaction effects, which could be expected based on past research utilizing similar tasks to investigate inhibition and WM, as well as age differences in these functions (Andres et al., 2008; Bos & Machado, 2013; Castel et al., 2007; Hester et al., 2004; Kemtes & Allen, 2008). The effect sizes were interpreted as small (η_p² ≥ .01), medium (η_p² ≥ .06), and large (η_p² ≥ .14), as is typical in the literature based on Cohen (1988) (see Richardson, 2011).

To explore WM as a variable of interest regarding performance on the inhibition tasks, we first generated for each age group composite scores for verbal and spatial WM by averaging across tasks (Song et al., 2013), resulting in a verbal WM composite score (forward + backward) and a spatial WM composite score (forward + backward). We did not include digit addition score in the verbal WM composite as the results of the initial age-group comparisons suggested that crystalized intelligence (e.g., knowledge, ability, and skills) was likely confounded with age group and had a strong influence on scores (see Discussion section for further information about this). Based on recommendations by Schneider et al. (2015), we grand-mean centered each WM composite score by subtracting the group mean score from each participant’s individual score ($x_i-\stackrel{\mathit{\hspace{0.17em}}}{x_i}$) before entering it as a covariate (Schneider et al., 2015). For correct RTs, we conducted repeated-measures ANCOVAs with trial type as a within-subjects variable, and for antisaccade reflexive errors we conducted univariate ANCOVAs. ANCOVAs were conducted separately for young and older adults, as WM scores differed between the young and older adults (i.e., the data did not meet the assumption that the expected value of the covariate at each level of the between-subjects factor is the same; Schneider et al., 2015). For repeated-measures ANCOVAs, interaction effects indicate that the slope of the line relating the covariate (e.g., verbal WM composite score) to the dependent measure (e.g., Stroop RT) depends on the level of the within-subjects factor (e.g., compatible vs. incompatible; Schneider et al., 2015). Analyses in G*Power (Faul et al., 2007, 2009) indicated that with sample sizes of 60, we had 80% power to detect medium-to-large effect sizes (η_p² ≥0.11), which could be expected based on past research (Borella et al., 2007; Kiefer et al., 2005; Unsworth et al., 2004). Effects reaching statistical significance (p < .05) or showing at least medium effect sizes (η_p² ≥0.06) were explored by visually inspecting scatter graphs with a regression line plotting the relationship between the relevant WM composite score and the relevant inhibition task measure (for RTs, each trial type was plotted separately), as recommended by Schneider et al. (2015).

Results

Mean of the median correct RTs and error percentages for the inhibition measures (except antisaccade reflexive errors) are shown in Table 1. Tables 2 and 3 show a summary of the mixed and one-way ANOVAs, respectively. Table 3 also shows descriptive statistics for antisaccade reflexive errors and WM scores.

Table 1. Reaction times (ms) and excluded trials (%), separated by inhibition task and trial type, in young and older adults.

	Young adults (n = 60)			Older adults (n = 60)
Paradigm	M	SD	%E	M	SD	%E
Fixation offset
Prosaccade
Offset	226	54	6.84	350	72	8.68
Overlap	305	64	6.49	365	69	9.27
Effect	38	−	−	15	−	−
Antisaccade
Offset	304	46	7.81	408	71	13.30
Overlap	329	53	7.15	423	75	12.78
Effect	25	−	−	14	−	−
Inhibition of return
Peripheral saccade signal
Cued	290	59	3.28	395	67	8.83
Uncued	262	58	5.11	311	52	6.94
Effect	−28	−	−	−64	−	−
Central saccade signal
Cued	372	60	4.56	487	77	8.72
Uncued	373	59	5.00	471	82	7.28
Effect	1	−	−	−16	−	−
Flanker
Compatible	466	68	1.37	615	105	1.85
Incompatible	489	81	2.26	659	118	4.46
Effect	23	−	−	44	−	−
Simon
Compatible	429	63	0.54	548	90	0.60
Incompatible	480	62	2.62	633	111	2.32
Effect	51	−	−	85	−	−
Stroop
Compatible	832	105	4.44	989	166	10.42
Incompatible	945	107	6.18	1222	170	17.71
Effect	113	−	−	233	−	−

Note. M = mean. SD = standard deviation. %E = percent excluded trials. Effect = difference between trial types. Fixation offset effect = overlap − offset. Inhibition of return = uncued – cued. Flanker, Simon, and Stroop effects = incompatible – compatible.

Table 2. A summary of mixed ANOVAs examining the effect of age groups (young versus older adults), trial types, and their interactions on raw and log-transformed reaction times in the inhibition tasks.

	Age Group			Trial Type			Age Group × Trial Type
Inhibition measure	F	p	ηp^2	F	p	ηp^2	F	p	ηp^2
Fixation offset prosaccade RT	39.89	<.001	0.25	72.52	<.001	0.38	14.16	<.001	0.11
Fixation offset prosaccade log	46.01	<.001	0.28	87.75	<.001	0.43	22.52	<.001	0.16
Fixation offset antisaccade RT	79.25	<.001	0.40	67.88	<.001	0.37	4.91	.029	0.04
Fixation offset antisaccade log	85.56	<.001	0.42	90.77	<.001	0.44	13.45	<.001	0.10
IOR peripheral RT	71.07	<.001	0.38	185.58	<.001	0.61	28.55	<.001	0.20
IOR peripheral log	74.46	<.001	0.39	102.88	<.001	0.47	11.75	<.001	0.09
IOR central RT	71.68	<.001	0.38	9.18	.003	0.07	12.11	.001	0.09
IOR central log	80.59	<.001	0.41	7.08	.009	0.06	6.88	.010	0.06
Flanker RT	87.46	<.001	0.43	109.89	<.001	0.49	10.60	.001	0.08
Flanker log	99.37	<.001	0.46	114.56	<.001	0.48	4.28	.041	0.04
Simon RT	83.45	<.001	0.41	368.79	<.001	0.76	24.06	<.001	0.17
Simon log	93.93	<.001	0.44	395.67	<.001	0.77	5.04	.027	0.04
Stroop RT	82.18	<.001	0.41	348.30	<.001	0.75	42.24	<.001	0.26
Stroop log	83.73	<.001	0.42	361.31	<.001	0.75	22.89	<.001	0.16

N = 120, df = 1, 118. RT = reaction time. IOR peripheral = inhibition of return task with a peripheral saccade signal. IOR central = inhibition of return task with a central saccade signal. Trial types for the different measures were as follows: fixation offset (overlap, offset), inhibition of return (cued, uncued), and Stroop, flanker, and Simon (compatible, incompatible).

Table 3. A summary of one-way ANOVAs examining the effect of age groups (young versus older adults) on antisaccade reflexive errors and working memory scores.

	Young adults		Older adults		Age Group
Measure	M	SD	M	SD	F	p	ηp^2
Antisaccade reflexive errors	2.81%	3	6.18%	6	16.51	<.001	0.12
Digit forward score	67	26	59	27	2.74	.100	0.02
Digit backward score	42	21	32	22	7.31	.008	0.06
Digit addition score	87	41	109	38	9.55	.002	0.08
Spatial forward score	68	20	39	16	76.01	<.001	0.39
Spatial backward score	59	26	32	18	45.63	<.001	0.28

N = 120, df = 1, 118.

Inhibition

Mixed ANOVAs on the raw RT data indicated large age-group effects (p < .001, η_p² ≥0.25),medium-to-large trial-type effects (p ≤ .003, η_p² ≥ 0.07), and interaction effects (p ≤ .029, η_p² ≥ 0.04). After log transformations, all of the main effects and interaction effects remained significant: age group main effects (p < .001, η_p² ≥0.28) and trial type main effects (p ≤ .009, η_p² ≥ 0.06), as well as the Age Group × Trial Type interaction effects (p ≤ .041, η_p² ≥0.04), thus confirming the observed age-group effects (see Bush et al., 1993; Faust et al., 1999). The main effects of trial type reflect slower RTs on trials associated with inhibition (i.e., overlap for fixation offset, cued for IOR, and incompatible for flanker, Simon, and Stroop). The significant Age Group × Trial Type interaction effects for FOE and IOR reflect older adults exhibiting considerably smaller FOEs and larger IOR than young adults (see Table 1), indicating stronger reflexive inhibition in older adults. In contrast, the significant Age Group × Trial Type interaction effects for flanker, Simon, and Stroop RT and the significant main effect of age group for antisaccade reflexive errors reflect older adults exhibiting considerably greater slowing on incompatible relative to compatible trials (i.e., larger compatibility effects) and a higher percentage of reflexive errors on the antisaccade task compared to young adults, collectively indicating significantly weaker volitional inhibition in the older adults. Taken together, these results indicate stronger reflexive inhibition and weaker volitional inhibition in older adults.

Working memory

One-way ANOVAs investigating the main effect of age group on WM measures produced mixed results. The verbal WM tasks showed no significant effect of age group on digit forward score (p = .100, η_p² = 0.02), but there were significant effects of age group on digit backward score (p = .008, η_p² = 0.06) and digit addition score (p = .002, η_p² = 0.08). These, respectively, reflect lower backward, but higher addition, scores in older compared to young adults. Regarding spatial WM, the significant effects of age group on both forward (p < .001, η_p² = 0.39) and backward (p < .001, η_p² = 0.28) scores reflect the considerably lower scores in older compared to young adults.

Inhibition and working memory

The results of the repeated-measures ANCOVAs exploring WM as a variable of interest regarding correct RT performance on fixation offset, IOR, flanker, Simon, and Stroop are shown in Table 4. Note that verbal and spatial WM were considered separately, and that we analyzed raw RTs (rather than transformed) as ANCOVAs were conducted separately for young and older adults due to group differences in WM scores (Schneider et al., 2015). As can be seen in Table 4, there was very limited evidence of WM being a variable of interest in inhibitory functioning in either age group. There was a significant Trial Type × Verbal Working Memory interaction in older adults for fixation offset antisaccade RT (p = .017, η_p² = 0.09). However, visual inspection of scatter graphs, with a regression line plotting the relationship between verbal WM scores and antisaccade RTs (done separately for offset and overlap trials) showed that, although higher verbal WM scores were weakly associated with faster RTs for both trial types, the contribution of verbal WM to antisaccade RTs was greater on offset trials (i.e., the slope was steeper, r² = .053) compared to overlap trials (r² = .008). Two other interactions showed medium effect sizes, both also in the older adults (p = .066, η_p² = 0.06, in both cases), one reflecting a small tendency for higher verbal WM scores to be associated with faster RTs on incompatible (r² = .024) but not compatible (r² = .000) trials of the Stroop task, and the other reflecting a small tendency for higher spatial WM scores to be associated with faster RTs on cued (r² = .027) but not uncued (r² = .000) trials of the peripheral saccade signal IOR task. No other interactions have reached statistical significance or shown at least medium effect size in either age group (p ≥ .073, η_p² ≤0.05).

Table 4. A summary of repeated-measures ANCOVAs exploring the contribution of working memory to performance on measures of inhibition in each age group.

	Trial Type × Verbal Working Memory										Trial Type × Spatial Working Memory
	Young adults (n = 60)						Older adults (n = 60)				Young adults (n = 60)						Older adults (n = 60)
Inhibition measure	F	p	ηp^2			F		p	ηp^2	F		p	ηp^2			F		p	ηp^2
Fixation offset prosaccade RT	1.70	.197	0.03			0.03		.872	0.00	0.19		.664	0.00			0.30		.585	0.01
Fixation offset antisaccade RT	0.73	.398	0.01			6.03		.017	0.09	2.74		.103	0.05			0.52		.475	0.01
IOR peripheral RT	0.19	.662	0.00			0.33		.570	0.01	2.62		.111	0.04			3.51		.066	0.06
IOR central RT	0.66	.420	0.01			0.12		.727	0.00	0.01		.930	0.00			0.01		.909	0.00
Flanker RT	2.67	.108	0.04			1.63		.206	0.03	0.64		.428	0.01			0.87		.356	0.02
Simon RT	0.38	.540	0.01			0.73		.396	0.01	1.32		.255	0.02			3.33		.073	0.05
Stroop RT	2.33	.132	0.04			3.50		.066	0.06	0.03		.864	0.00			2.17		.146	0.04

df = 1, 58 for both age groups. RT = reaction time. For each age group, the working memory scores were centered by subtracting the overall mean working memory score from each participant’s individual working memory score prior to being submitted to the ANCOVA (Schneider et al., 2015). IOR peripheral = inhibition of return task with a peripheral saccade signal. IOR central = inhibition of return task with a central saccade signal. Trial types for the different measures were as follows: fixation offset (overlap, offset); IOR (cued, uncued); and Stroop, flanker, and Simon (compatible, incompatible).

Results of the univariate ANCOVAs exploring WM as a variable of interest regarding antisaccade reflexive errors showed in the young adults a significant effect of spatial WM score on antisaccade reflexive errors, F(1, 58) = 5.85, p = .019, η_p² = 0.09, and visual inspection of a scatter graph indicated higher spatial WM scores were associated with fewer reflexive errors (r² = .092). The young adults also showed a medium effect size for higher verbal WM scores to be associated with fewer reflexive errors (p = .065, η_p² = 0.06, r² = .057). In contrast, the results in the older adults showed no evidence linking verbal or spatial WM scores to antisaccade reflexive errors (p ≥ .478, η_p² ≤0.01).

Discussion

The aims of the current study were to investigate age-related differences in inhibition and WM function with consideration of inhibition type (reflexive/subcortical vs. volitional/cortical) and WM domain (verbal vs. spatial) and to characterize relationships between each WM domain and inhibition type in both young and older adults. Compared to young adults, older adults showed significantly stronger reflexive inhibition and significantly weaker volitional inhibition. Regarding the WM tasks, aside from mental arithmetic performance, older adults tended to show worse performance, especially for the spatial tasks, which showed quite large effect sizes for age-group differences. Surprisingly, we found very limited evidence of relationships between WM and inhibition, with only one significant effect (in the young adults) supporting WM ability as a determinate of inhibition performance. These results will be discussed in more detail in subsequent sections.

Age-group differences in inhibition

Our results indicate age-related increases in reflexive inhibition, as evidenced by FOE and IOR, and age-related decrements in volitional inhibition demonstrated across several measures (antisaccade reflexive errors and flanker, Simon, and Stroop effects). Analyses of the transformed data confirmed the observed age–group interaction effects (see Bush et al., 1993; Faust et al., 1999). These patterns of results are consistent with previous studies reporting, in older compared to young adults, a smaller FOE (Bos & Machado, 2013), stronger IOR (McLaughlin et al., 2010; Poliakoff et al., 2007), more antisaccade reflexive errors (Brett & Machado, 2017), and larger flanker (Machado et al., 2009; Zeef & Kok, 1993; Zeef et al., 1996), Simon (Castel et al., 2007; van der Lubbe & Verleger, 2002), and Stroop effects (Spieler et al., 1996; Troyer et al., 2006). The observation of these age-group differences in the current study but not in some past studies in most cases may relate to the larger sample sizes in the current study (60 versus 20 or fewer per age group in past studies), although differences in the details of the tasks utilized to measure inhibitory functioning may be a factor as well and may help explain the lack of age-related inhibitory deficits in some meta-analyses (e.g., Rey-Mermet & Gade, 2018) as previously argued in relation to WM tasks (Hurley & Machado, 2018). It should be noted that in the current study the age-group differences in inhibition occurred in the absence of older adults exhibiting lower error rates, thus indicating that the older adults did not adopt a more cautious strategy (in contrast to Vallesi et al., 2021).

The observed combination of weaker volitional inhibition and stronger reflexive inhibition suggests that age-related deterioration of cortical structures may allow subcortical structures to operate under a reduced level of cortical control. For example, the superior colliculus is a subcortical structure that mediates reflexive oculomotor responses to both central visual signals, which stimulate the fixation reflex to help anchor the eyes, and peripheral visual signals, which inhibit the fixation reflex (Munoz & Istvan, 1998; Munoz & Wurtz, 1993). These subcortical mechanisms allow for efficient reflexive responding, but they require cortical areas to impose control over them to implement goal-directed behavior (Machado & Rafal, 2004a, 2004b; van Koningsbruggen et al., 2009). Age-related deterioration of cortical structures may allow the superior colliculus to operate under less control, thus resulting in stronger attentional and/or oculomotor capture by peripheral stimuli and weaker volitional inhibition of prepotent responses (Machado & Rafal, 2004b; Nieuwenhuis et al., 2000). Indeed, in addition to the present findings, there is evidence that older adults experience stronger attentional capture by peripheral cues (Juola et al., 2000), they are less successful at suppressing a reflexive eye movement toward a sudden stimulus in the periphery (Brett & Machado, 2017; Chen & Machado, 2016; Olincy et al., 1997), and have more difficulty ignoring spatial information triggered by a target located in the periphery (Castel et al., 2007). This suggests that older adults may experience hyper-responsiveness to peripheral visual stimulation.

Age-group differences in working memory

Regarding verbal WM, digit forward scores were not significantly worse in older compared to young adults, but older adults had significantly lower scores for digits backward. In contrast, older adults had significantly higher digit addition scores, which we suspect likely reflect factors unrelated to WM (namely, crystallized intelligence, which has been shown to remain stable or even improve with healthy aging; Salthouse, 2012). In relation to this, a cohort effect may have contributed to the age-group differences in digit addition performance (Harada et al., 2013), as the older adults were more likely to have more experience adding numbers to their heads due to more training in school followed by many years of practice. By comparison, the young adults may not have received as much mental arithmetic training or engaged in as much practice, due in part to greater availability and higher reliance on electronic calculators (e.g., on their mobile phones). These conjectures fit with reports in the United States of America (Geary & Lin, 1998; Geary & Wiley, 1991; Geary et al., 1993).

Regarding the spatial WM tasks, compared to young adults, older adults had substantially lower scores for both forward and backward recall. In addition, as predicted, the effect sizes for age differences in WM were much larger for the spatial tasks than for the equivalent verbal tasks (digit forward and backward), indicating greater age-related deficits in spatial WM performance. These findings support those of previous studies reporting a more pronounced age-related decline in spatial compared to verbal domains (e.g., Cansino et al., 2013; Hale et al., 2011), which interestingly has also been reported in association with Parkinson’s disease (Ramos & Machado, 2021).

Inhibition and working memory

Despite robust age-group differences in measures of both inhibition and WM, we found very limited evidence that WM abilities influenced performance on the tasks measuring inhibition. Of the eight inhibition measures (four reflexive/subcortical and four volitional/cortical) considered in each age group, very few showed the predicted relationships with WM measures (verbal or spatial). In the older adults, of the 16 analyses run, one yielded a statistically significant effect and two yielded medium effect sizes that did not reach significance (p = .066 in both cases). With regard to the latter, higher verbal WM scores tended to be associated with faster RTs on incompatible (but not compatible) Stroop trials, which fits with the conceptual framework regarding verbal WM influencing volitional inhibition (see Figure 1); however, the association was fairly weak (r² = .024) and limited to this measure of volitional inhibition. In addition, higher spatial WM scores tended to be associated with faster RTs on cued (but not uncued) trials of the peripheral saccade signal IOR task, which fits with the conceptual framework regarding spatial WM influencing reflexive inhibition; however, again the association was fairly weak (r² = .027) and limited to this measure of reflexive inhibition. The only statistically significant evidence in the older adults linking WM to inhibition task performance stems from higher verbal WM scores being associated with faster fixation offset antisaccade RTs, but this relationship was stronger on offset trials (r² = .053) and bordered on negligible on overlap trials (r² = .008), which does not fit with the model as it is the overlap trials that require inhibition to overcome the effects of the fixation stimulus in order to initiate the saccade. In the young adults, of the 16 analyses run, one yielded a statistically significant effect and one yielded a medium effect size that did not reach significance (p = .065). Respectively, these provided evidence for an association between higher spatial and verbal WM scores and fewer antisaccade reflexive errors, consistent with the model linking verbal and spatial WM to volitional inhibition; however, none of the other measures yielded evidence consistent with the conceptual framework regarding WM influencing inhibition in young adults.

Regarding the tendency in the older adults for spatial WM scores to be associated with faster cued peripheral saccade signal RTs (IOR paradigm), although consistent with the model depicted in Figure 1, this result was somewhat surprising in the context of the lack of WM links with volitional inhibition measures, as IOR is a reflexive subcortically mediated mechanism (Sapir et al., 1999, 2002) and we expected to see effects of WM ability primarily on volitionally mediated inhibition in relation to both relying on cortical structures (Chambers et al., 2009; Zaehle et al., 2011). Nevertheless, this finding is consistent with the concept that inferior cortical functioning is associated with hyperactive subcortical functioning, and fits with previous reports that IOR is partially mediated by spatial WM (Castel et al., 2003; Chou & Yeh, 2008).

The only link in the older adults between WM and volitional inhibition that reached the threshold of an at least medium effect size stemmed from the Stroop task: higher verbal WM scores tended to be associated with faster correct RTs on incompatible trials. As reading becomes increasingly automatic over the years, which can in part account for larger Stroop effects in older than young adults (Lien et al., 2006), older adults with poorer verbal WM ability may have found it particularly difficult to inhibit reading the words due to having fewer resources available for conflict resolution, relative to older adults with superior verbal WM abilities (Aschenbrenner & Balota, 2015). Greater automaticity of word reading among older adults might help explain why evidence of a relationship with WM ability emerged for this volitional inhibition measure but not the others. The indication that verbal WM may be linked to Stroop performance more so in older than young adults fits with a previous study reporting on a lifespan sample (Aschenbrenner & Balota, 2015; see also McCabe et al., 2005). However, the lack of a relationship in young adults contrasts with past research linking superior verbal WM performance to less interference from incompatible Stroop words in university students (Hutchison, 2011; Kane & Engle, 2003; Long & Prat, 2002; Meier & Kane, 2013). This difference does not appear to be related to sample size as the current study involved a larger sample than some of these previous studies. The discrepancy may relate to differences in the WM tests and/or the Stroop tests. In contrast to the current study, in assessing WM these previous studies interjected a second task (reading or operation/sentence/symmetry judgments) between items in the memory set, and in assessing Stroop performance they manipulated the frequency of compatible versus incompatible trials (mostly compatible or mostly incompatible), which was a determinant of the WM-inhibition links and thus potentially the main factor underpinning the different outcomes.

The findings in the young adults linking higher WM scores to fewer reflexive errors during the antisaccade task fit with past reports in young adults (Kane et al., 2001; Unsworth et al., 2004). However, there was virtually no evidence of links between WM scores and reflexive errors during the antisaccade task in the older adults, which goes against what one would expect if inhibitory control depended on deteriorating WM systems (see Eenshuistra et al., 2004). That is, given that older adults showed both poorer WM performance and more reflexive errors during the antisaccade task, based on the model illustrated in Figure 1 one would expect the latter to be linked to the former in older adults. The absence of this link calls into question the concept that inhibitory functioning depends on WM systems.

Overall, our findings largely fail to support previous conjectures that WM mediates performance on tasks that require inhibitory control (e.g., Castel et al., 2003; Kramer et al., 2001; Roberts et al., 1994) and that participants with low WM capacity experience more interference from distractors and are slower and less accurate on inhibition tasks than participants with high WM capacity (e.g., Kane & Engle, 2003; Kane et al., 2001; Long & Prat, 2002; Minamoto et al., 2010). The discrepant findings likely reflect differences in the methodological details of the tasks used to assess WM and inhibition. Further to key methodological differences already discussed in relation to past Stroop studies in young adults observing links with verbal WM not observed here, many of the previous studies reporting WM-inhibition links used a combined task that tapped both WM and inhibition (e.g., Blair et al., 2011; Borella et al., 2007; Castel et al., 2003; Chou & Yeh, 2008; Clouter et al., 2015; Roberts et al., 1994) else intermixed the WM and inhibition tasks and measures (Borella et al., 2008). Consequently, in contrast to the current study, the results of these studies cannot speak to whether inhibition and WM systems function interdependently when the two systems are not expressly co-activated.

Limitations and suggestions for future research

One limitation of the current study relates to the assumptions of the categorization scheme. For example, tasks categorized as indexing reflexive/subcortical inhibition are not purely subcortically mediated and performance relies on many factors beyond inhibition. As an example of the latter, the magnitude of a flanker effect reflects not only the extent to which distractor processing was inhibited but also the extent to which target processing was facilitated (Wyatt & Machado, 2013a). Thus, differences in effects may reflect differences in other task-relevant functions, which would have added extraneous variation that could have led to erroneous conclusions regarding the profile of age-related decline and may have compromised our ability to detect WM-inhibition relationships. Covariance matrices for inhibition measures supported the grouping of the variables categorized as volitional/cortical inhibition, but only partially supported the grouping of the variables categorized as reflexive/subcortical inhibition (see Supplementary Material A). In a related note, the age-related deterioration of cortical regions is not uniform, and many subcortical regions are also adversely affected (Sele et al., 2020; Vinke et al., 2018). Although we used the broad terms cortical and subcortical, the former was primarily in reference to the prefrontal cortex, which is known to atrophy with adult aging (e.g., see Hedden & Gabrieli, 2004) and the latter in reference to the superior colliculi, which we have presumed to undergo less age-related atrophy, however this is based on limited data (Aanerud et al., 2012; Bellot et al., 2016; Luft et al., 1999; Walhovd et al., 2011). Moreover, although in the current study all of the inhibition measures categorized as volitional/cortical indicated reduced inhibition in the older adults, it should be noted that different cortical inhibitory control processes rely on distinct neural networks (e.g., Zhang et al., 2017) and, in relation to this, the different volitional/cortical inhibition measures in the current study are not assumed to be tapping a unitary inhibition construct, consistent with past findings (Rey-Mermet & Gade, 2018; Rey-Mermet et al., 2018).

Another limitation of the current study relates to all of the older participants being grouped together despite their ages spanning a fairly wide range (60–88 years). According to Persad et al. (2002), such designs fail to account for the continued age-related decline among older adults (see also Borella et al., 2007). Previous studies classifying older participants as “young-old” (60–74 years) and “old-old” (75 years and older) found that old-old participants performed significantly worse on a range of neuropsychological tasks compared to their younger counterparts (Christensen et al., 1994; Palladino & De Beni, 1999; Persad et al., 2002). Unfortunately, in the current study, we were unable to separate the older adults into subgroups as 48 of the 60 older adults were aged 74 or under. Future research seeking to characterize age-related differences in WM and inhibitory functioning should consider utilizing a study design that allows for the comparison of young, young-old, and old-old adults. A final limitation of the current study worthy of note relates to the sample size. Although we had relatively large sample sizes and power analyses indicated that we had 80% power to detect medium-to-large effects (which could be expected based on previous studies), larger samples would nevertheless have been preferable to improve the power of the current study. While it seems unlikely that adding more participants would change the conclusions, given that in some instances the direction of effects opposed the prediction that inhibitory functioning depends on WM ability, inclusion of larger sample sizes would allow for the use of more sophisticated statistical models, such as structural equation modeling (Westland, 2010). This could enable more direct testing of the conceptual framework proposed here, resulting in much stronger conclusions.

Conclusions

In summary, the results reported here support age-related increases in reflexive inhibition (based on the fixation offset effect and inhibition of return) and age-related decrements in both volitional inhibition (based on performance on the antisaccade, Stroop, flanker, and Simon paradigms) and WM performance (verbal and spatial). This pattern of results provides novel evidence suggestive of an age-related shift in the balance of power from cortical to subcortical structures, as indicated by weaker volitional inhibition and poorer WM performance but stronger reflexive inhibition in older than young adults. This behavioral evidence of hyperactive subcortical functioning presumably reflects age-related deterioration of cortical structures, leaving subcortical structures to operate under less cortically orchestrated control. Functionally, this would result in older adults behaving in a relatively more reflexive and less volitional manner, effectively losing some of the earlier life gains in strategic control (e.g., Fischer et al., 1997; Munoz et al., 1998); however, it should be noted that the trajectory of these shifts varies among individuals and much can be done via lifestyle choices to promote a more favorable trajectory (Machado, 2021). Regarding potential links between WM ability and inhibition, in contrast to current theory, our findings suggest that inhibition and WM are largely independent and that WM limitations cannot account for age-related declines in inhibitory functioning.

Acknowledgments

The authors wish to thank Andrew Webb for assisting with preparing the software used for cognitive testing and Natalie Wyatt for assisting with data collection.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

Data and code will be made available upon request.

Additional information

Funding

This research was supported by University of Otago Research Grants