Aging and task design shape the relationship between response time variability and emotional response inhibition

ABSTRACT Intra-individual variability (IIV) refers to within-person variability in behavioural task responses. Several factors can influence IIV, including aging and cognitive demands. The present study investigated effects of aging on IIV of response times during executive functioning tasks. Known age-related differences in cognitive control and emotion processing motivated evaluating how varying the design of emotional response inhibition tasks would influence IIV in older and younger adults. We also tested whether IIV predicted inhibitory control across task designs and age groups. Older and younger adults (N = 237) completed one of three versions of a stop-signal task, which all displayed happy, fearful, or neutral faces in Stop trials. An independent group of older and younger adults (N = 80) completed a go/no-go task also employing happy, fearful and neutral faces. Results showed older adults had more consistent responses (lower IIV) than younger adults in the stop-signal task, but not the go/no-go task. Lower IIV predicted more efficient emotional response inhibition for fear faces in the stop-signal task, but only when attention to emotion was task-relevant. Collectively, this study clarifies effects of aging and task design on IIV and illustrates how task design impacts the relationship between IIV and emotional response inhibition in younger and older adults.

Researchers and clinicians often use neurocognitive testing to assess cognitive and brain functioning. The degree of within-person variability or inconsistency in these tests can provide valuable additional information about individuals' cognitive functioning beyond interpretation of mean response time or performance (MacDonald & Trafimow, 2013). Several factors impact within-person variability, or intra-individual variability (IIV), in behavioural performance including participant characteristics as well as present cognitive demands. For instance, greater IIV is observed in individuals with age-related neurodegeneration, traumatic brain injury and attention deficit hyperactivity disorder (ADHD) (Dixon et al., 2007;Duchek et al., 2009;Fozard et al., 1994;Hultsch et al., 2002Hultsch et al., , 2008Hultsch & MacDonald, 2004;Vaurio et al., 2009). IIV also corresponds negatively with performance in several cognitive domains including working memory, episodic memory, executive functioning and attention in cognitively unimpaired adults (Hultsch et al., 2002;MacDonald et al., 2006).
There are several inferences about why increased IIV corresponds with poorer performance on tasks of cognition. Attentional control is required to preserve stimulus information in working memory so that information relevant to task goals can be assessed and applied. Difficulty maintaining attentional control can be attributed to inconsistent implementation of cognitive processes, consequently producing high variability in response times across trials within a task Unsworth, 2015). In the domain of executive functioning, several studies have demonstrated that high variability in trial-to-trial response times (i.e. increased IIV) is associated with poorer response inhibition, which is one's ability to suppress actions that interfere with their goals (Bellgrove et al., 2004;Joly-Burra et al., 2018;MacDonald et al., 2006;Rochat et al., 2013). The strong positive relationship between response inhibition and response consistency is also supported by neuroimaging evidence that higher IIV corresponds with greater activation bilaterally in prefrontal executive control regions during successful response inhibition (Bellgrove et al., 2004). Taken together, evidence suggests poorer response inhibition in individuals with greater IIV is attributed to difficulty sustaining the attentional control needed to successfully override a response.

Effects of task design on response consistency
Effects of cognitive demands on IIV can be investigated further by varying the features of task design.
A key distinction in cognitive demands between types of response inhibition tasks is whether they implement proactive or reactive inhibitory control, which each have differing neural and behavioural bases, and thus may yield differing relationships to IIV. Neurally, the temporal dynamics of proactive and reactive inhibition tasks differ, as demonstrated in electroencephalography (EEG) and electromyography (EMG) activity declining significantly faster across each trial in reactive than proactive inhibition tasks (Liebrand et al., 2017;Raud et al., 2020). Behaviourally, proactive inhibition relies on action restraint, whereas reactive inhibition requires action cancellation (Aron, 2011). In an action restraint paradigm, the individual is presented with a stimulus and must proactively choose whether to respond by pressing a designated key on a computer keyboard. In an action cancellation paradigm, the individual has been primed to respond, and after the presentation of a stimulus they must reactively cancel the initiated response (Krämer et al., 2013;Schachar et al., 2007) to not press the key. Previous studies have demonstrated that behavioural performance can change for one type of response inhibition task independently from the other, such that individuals may have good performance on one type of response inhibition task yet poor performance on another (Krämer et al., 2013;Littman & Takács, 2017;Raud et al., 2020).
Proactive and reactive inhibition also utilise different degrees of automaticity. In proactive inhibition tasks, the go stimulus is repeatedly mapped onto a single response that remains constant throughout the task, thereby engaging more automatic inhibition. However, in reactive inhibition tasks, the go stimulus may be associated with varied responses throughout the task (e.g. choice of two response keys) and require more controlled inhibitory processes. Consequently, reactive inhibition tasks require a more consistent implementation of executive control processes to maintain task performance than proactive inhibition tasks require (Patterson et al., 2016;Verbruggen & Logan, 2008a). Thus, reactive inhibition tasks may impose greater cognitive demands and yield less consistent responding.
Two of the most frequently chosen tasks of proactive and reactive response inhibition are the go/no-go and stop-signal tasks, respectively (Aron, 2011). Although IIV of response times can be computed equivalently for both tasks, the outcome measures of response inhibition differ by task. In the proactive inhibition go/nogo task participants are instructed for which stimuli they should make a response (go to targets) versus withhold their response (no-go to non-targets), and they initiate their action (or not) according to the instruction (Raud et al., 2020). Inhibition performance is typically assessed as proportions of errors of commission to non-target no-go trials (i.e. false alarms) and errors of omission to target go trials (i.e. misses) (Littman & Takács, 2017;Raud et al., 2020). In contrast, reactive inhibition requires an individual to interrupt and stop an ongoing response upon presentation of a cue (Meyer & Bucci, 2016;Smittenaar et al., 2015). Notably, the stop-signal taska test of reactive inhibitionmeasures response stopping efficiency, the time needed to stop an ongoing response, which is a different construct from the go/no-go task's inhibition outcome measure of proportion of errors (Aron, 2011;Verbruggen & Logan, 2008b). The differing measurements of the construct of response inhibition may yield distinct relationships to response consistency.
Neuroimaging techniques reveal that the stopsignal task and go/no-go task utilise different neural mechanisms to implement response inhibition (Meyer & Bucci, 2016;Raud et al., 2020). FMRI research shows that although the stop-signal task and go/nogo task both activate the right inferior frontal gyrus and the dorsomedial prefrontal cortex (pre-SMA), the stop-signal task additionally distinctly activates the basal ganglia, while the go/no-go task additionally activates occipital regions, putamen and the left premotor cortex (Aron, 2011;Simmonds et al., 2008;Verbruggen & Logan, 2008b). EEG and EMG studies have identified distinctions between proactive and reactive inhibition in prepotent motor activity and frontal-midline theta activity, and also relatively earlier timing of reactive inhibitory mechanisms (Messel et al., 2021;Raud et al., 2020). Taken together, evidence suggests that proactive and reactive response inhibition tasks are likely to have distinct relationships to consistent (i.e. efficient) implementation of cognitive and neural processes (as measured through IIV), such that IIV is likely to be greater while implementing reactive than proactive inhibition. Comparing IIV between response inhibition tasks can offer insights about the implications of fluctuating cognitive consistency on proactive versus reactive inhibition.

Effects of attention to emotion on response consistency
In addition to demands of implementing proactive versus reactive inhibition, engagement with emotional information can also significantly influence moment to moment cognitive processing and its relationship to inhibitory control performance (Ding et al., 2020;Pessoa et al., 2012;Tyng et al., 2017;Waring et al., 2019;Williams et al., 2020). Emotional stimulifacial expressions, words, or images of emotional contenthave often been employed in tasks of inhibitory control to identify how the presence of affective information impacts response inhibition, i.e. emotional response inhibition (Ding et al., 2020;Elliott et al., 2000;Patterson et al., 2016;Pessoa et al., 2012;Schulz et al., 2007;Verbruggen & De Houwer, 2007). Literature evaluating the effect of emotion on response inhibition is mixed, with studies finding that emotion either impairs or aids performance, depending on the cognitive mechanism employed (Ding et al., 2020;Hoffmann et al., 2021;Levens & Phelps, 2008;Mancini et al., 2020;Verbruggen & De Houwer, 2007).
Response consistency and inhibition during a task featuring emotional stimuli may be further modulated by task design. For example, researchers presented younger adults with negative, neutral, or no stimuli before go, no-go and stop trials. The results showed that exposure to negative stimuli before each trial affected performance on the stop-signal task but not on the go/no-go task (Littman & Takács, 2017). The authors speculated that the unique effects of negative stimuli on reactive inhibition could be attributable to signalling of differing neurotransmitters between the two tasks (namely norepinephrine) or the ability of negative stimuli to interfere with and slow the cognitive resources required for topdown inhibition (Littman & Takács, 2017). Both of these accounts for mechanisms of emotional response inhibition describe factors that are also likely to influence trial-to-trial response consistency, e.g. arousal and attentional mechanisms.
Task instructions can affect allocation of attention to emotional information, potentially impacting ability to maintain attentional control on the task goals (Duchek et al., 2009). One example is the degree to which overt attentional focus on emotional information is required to correctly follow task instructions, i.e. the task relevance of emotional attributes of the stimulus. We previously reported that emotional stimuli significantly impacted response inhibition (operationalised as stop signal response time, the time needed to stop one's response) in older and younger adults only when attention to emotion was task-relevant (Williams et al., 2020). These results and others showing that emotional stimuli elicit attention only when task performance relies on the emotional attributes of the stimulus (Mancini et al., 2020;Tannert & Rothermund, 2020;Victeur et al., 2020) support the contingent capture hypothesis, which states that attentional capture is dependent on task demands that influence attentional control settings (Folk et al., 1992;Victeur et al., 2020). Moment-tomoment fluctuation in attending to and implementing task goals is revealed in measures of response consistency. Taken together, this work indicates IIV may be lower when participants are instructed to attend to the emotional attributes of task stimuli because greater attentional control is applied in these circumstances, compared to when direct attention to emotion is not required to follow task instructions.

Relationships among aging, response consistency and response inhibition
Aging may particularly affect response consistency (IIV) and the relationship of IIV to emotional response inhibition. Consistent effective engagement of cognitive resources typically declines with aging (Dykiert et al., 2012;Fozard et al., 1994;Hultsch et al., 2002Hultsch et al., , 2008Nesselroade & Salthouse, 2004; but see Myerson et al., 2007;Nicosia & Balota, 2021;Shammi et al., 1998), yet whether age-related changes in response consistency predict emotional response inhibition performance is an open question. Nearly all prior research comparing IIV to response inhibition tested effects in children with ADHD (Vaurio et al., 2009). Emotion regulation improves with age, while fluid cognition including cognitive control declines.
Although older adults are typically slower to inhibit responses to neutral than emotional stimuli, they are more accurate (versus younger adults) during trials employing positive stimuli (Bloemendaal et al., 2016;Waring et al., 2019;Williams et al., 2020). In our previous work, we observed that emotional response inhibition in younger adults was impaired by negative stimuli in a stop-signal task, and by positive stimuli in a go/no-go task (Waring et al., 2019;Williams et al., 2020). In older adults, positive information facilitated emotional response inhibition when attending to emotional attributes of task stimuli was task-relevant, while negative and positive stimuli had no significant effect on emotional response inhibition when attending to emotion was not task-relevant (Williams et al., 2020). Results of emotional response inhibition tasks appear inconsistent across the adult lifespan and are substantially impacted by elements of the tasks employed. Research of agerelated differences in emotional response inhibition is notably limited and existing literature does not offer a clear indication of how aging may influence the relationship between response consistency and emotional response inhibition (i.e. interaction between aging and IIV to predict task performance). However, taken together, our prior work and other research on effects of emotion on attention indicate that task instructions to focus attention on emotional (vs non-emotional) stimulus attributes are likely to modulate this relationship.
Assessing younger and older adults' data from two emotional response inhibition tasks can reveal effects of aging and task demands on cognitive consistency trial-to-trial (IIV) and whether it predicts emotional response inhibition performance for either age group or under specific task demands. Evaluating IIV in this manner can provide a clearer characterisation of the implementation of proactive and reactive emotional response inhibition in non-clinical aging populations. Although the existing literature has taken multifaceted approaches to studying factors influencing IIV or response inhibition independently, the present study additionally clarifies the effects of aging and attention to emotion on IIV and informs the relationship between IIV and emotional response inhibition.

Present study
The overarching purpose of the present study was to identify factors significantly impacting consistent effective engagement of cognitive resources (i.e. IIV) during emotional response inhibition tasks and understand if those factors interact with IIV to predict inhibitory control performance in younger and older adults. The present study used data we collected previously first to assess whether aging and task design would impact trial-to-trial consistency in response times (i.e. IIV) during proactive (i.e. go/nogo) and reactive (i.e. stop-signal task) emotional response inhibition tasks displaying positive, negative and neutral stimuli. Prior publications reporting results from these datasets did not evaluate response consistency (Greif & Waring, 2018;Waring et al., 2019;Williams et al., 2020). We hypothesised that IIV (standard deviation of response times) would be greater in the stop-signal task compared to the go/no-go task because the stop-signal task requires additional executive control processes and limits automatic processing to a greater extent than the go/no-go task (Aron, 2011;Verbruggen & Logan, 2008a). We also hypothesised that aging would negatively impact response consistency during the stop-signal and go/ no-go tasks such that older adults would have significantly higher IIV than younger adults. We expected this directionality due to age-related neurodegeneration and cognitive control declines. We additionally evaluated if task instructions that directed visual attention to specific stimulus attributes differentially impacted IIV. We hypothesised that stopsignal task instructions that required overt focus on facial expressions shown in the task would positively impact response consistency (i.e. lower IIV) because participants would be applying greater attentional control to successfully complete the task.
Lastly, we examined whether IIV would predict response inhibition performance. We predicted a significant interaction of IIV with age group and task instruction to predict emotional response inhibition in the stop-signal task. We likewise conducted additional analyses on the go/no-go task to test our expectation that IIV would interact with age group to predict emotional response inhibition. The limited relevant literature prevented a prediction of direction of these interactive effects.

Stop-Signal task methods
This is a secondary data analysis of data published in Williams et al. (2020). Briefly, analyses included three versions of a stop-signal task that used images of human faces as stop cues. Each version differed in its instructions regarding the Stop cue, such that participants were instructed to either withdraw their responses for All Faces, a specific Gender, or a specific Emotion. Institutional Review Boards of Saint Louis University and Washington University in St. Louis approved the research protocol in accordance with the Declaration of Helsinki.

Participants
Recruitment. Recruitment approaches and participant screening are described fully in Williams et al. (2020). To summarise briefly, younger adults were recruited from Saint Louis University through an online recruitment portal (SONA Systems, Bethesda, MD, USA) and older adults were recruited from the greater St. Louis area through fliers, advertisements and the Washington University Volunteers for Health Recruitment Enhancement Core's Research Participant Registry (https://sites. wustl.edu/wuvfh/). Participants were screened for exclusion criteria of present or prior diagnosis or treatment of any psychiatric or neurological conditions, uncorrected vision or hearing problems, colorblindness and severe head injury. Inclusion criteria for younger adults were ages 17-30 years and for older adults were age 60 or above. Sample sizes and demographic characteristics are reported in Table 1. Younger adults were offered credits toward psychology course requirements and older adults were offered monetary compensation for their time.

Procedure
Participants first provided written informed consent with HIPAA authorisation, then provided demographic information and completed neuropsychological measures to assess cognitive functioning. Participants were assigned to one of three versions of the stop-signal task (i.e. a between-subjects factor), which differed in their instructions (as described below). Participants heard the stop-signal task instructions (described below) for the respective task version assigned and had the opportunity to complete a practice task to assure the instructions were understood and retained. Once participants were comfortable with the task practice, they completed the emotional stop-signal task protocol.

Emotional stop-signal task design
Complete description of task stimuli and design is available in Williams et al. (2020) and is summarised briefly below.
Stimuli. The set of 180 greyscale images of human faces used in the task were replicated from Pessoa et al. (2012), which were drawn from four published stimulus sets (Ekman & Friesen, 1976;Ishai et al., 2004;Lundqvist et al., 1998;Tottenham et al., 2009). The facial expressions were balanced, such that there were 60 fear, 60 happy and 60 neutral faces shown in the task. The gender of the faces was also balanced across types of facial expressions.
All faces stop cue instruction. The stop-signal task required participants to complete a button-press response during Go trials where circles and squares were presented. Participants identified the shape on Go trials with a button-press (left arrow, right arrow). Presentation of Go stimuli (circle or square) was counterbalanced between participants for a given trial. The Go stimulus was presented on the computer screen for 1000ms, with a 1000ms blank screen as inter-trial interval. In Stop trials, a face appeared on the screen inside the shape (circle or square) for 500ms ( Figure 1 displays task schematic) cuing participants to stop their response underway and withhold a button press. In the All Faces task version participants were instructed to withdraw their in-progress response upon presentation of any face (i.e. the stop signal) inside the shape. In Stop trials, the stop signal was presented inside the shape following a delay after onset of the Go stimulus, known as the stop-signal delay (SSD). The SSD was adaptive in 50ms increments to maintain task accuracy near 50% (i.e. if inhibition was not successful, the SSD decreased by 50ms to encourage success in the next trial. If inhibition was successful, SSD increased by 50ms to make the subsequent trial more challenging). SSD was adaptive independently for each of the 3 face conditions so there were unique SSDs computed for the fear, happy and neutral faces. (For description and calculation of outcome measure of response inhibition see Measures.) The stop-signal task contained 900 experimental trials that were divided into six blocks of 150 trials each. Stop trials were present 20% of the time (30 per block, 180 total) to establish responding to the Go stimuli as an over-learned response. A practice block was administered to participants before the experimental task.
Gender stop cue instruction. Task design was kept constant from the All Faces version, apart from task Table 1. Stop signal task and go/no-go task participant demographics and go trials mean response times.  instruction. Participants were told to stop their ongoing responses based on the gender of the face in the stop signal, e.g. "Stop your response to male faces, and press the button for female faces." Therefore, one gender, as well as circles and squares, served as a Go signal in this task version. Consequently, the Gender stop cue required a greater focus on the details of the faces than the All Faces instruction. A practice block was administered to participants before the experimental task and when the stop cue changed (e.g. stop for female faces instead of male faces).
Facial expression stop cue instruction. Task design was kept constant from All Faces and Gender task versions. The instructions were modified to ask participants to withhold responses based on the facial expression of the stop-signal face, e.g. "Stop your responses to happy faces, and press the button for neutral or fear faces." Go signals for these instructions consisted of the two non-target facial expressions (i.e. neutral and fearful faces for the above example), as well as circles and squares. This is the only task instruction that required an overt focus on the emotional attributes of the faces, and therefore attention to the emotional expressions on the faces was task-relevant for this task instruction exclusively. A practice block was administered to participants before the experimental task and each time the stop cue changed (e.g. stop for fear faces).

Measures
IIV was computed for each individual as their withinperson standard deviation of response times for Go trials across all task blocks (see Supplemental Materials for discussion and analyses of coefficient of variation as an alternative measure of IIV). Stopsignal reaction time (SSRT) was the outcome measure used to index the efficiency of stopping responses (i.e. response inhibition). SSRT was computed by subtracting the mean SSD for each stop signal condition (i.e. fear, neutral and happy faces) from the Go trial median response time (RT), e.g. fear SSRT = median Go RTmean fear SSD (median Go RT provides a less biased estimation of stopping time than mean Go RT; for discussion of SSRT computations see Verbruggen et al., 2013Verbruggen et al., , 2019. Several neuropsychological measures were collected to confirm normative scores for executive functioning. Full description and results of these measures are reported in Williams et al. (2020).

Go/no-go task methods
This is a secondary analysis of data published in Waring et al. (2019). The study recruited younger and older adults to complete a go/no-go task displaying human faces with three expressions (happy, neutral and fear

Procedure
Participants first provided written informed consent with HIPAA authorisation, then they provided demographic information and completed neuropsychological measures to assess cognitive functioning. Participants learned the go/no-go task instructions and completed a brief practice task assure the instructions were understood and retained. Once participants were comfortable with the practice task, they completed the go/no-go task protocol.

Go/No-go task design
Complete description of task stimuli and design is available in Waring et al. (2019) and are summarised briefly below.
Stimuli. Task stimuli comprised 36 greyscale human faces from the NimStim Set (Tottenham et al., 2009). The faces depicted an equal number of neutral, happy and fearful expressions. The gender and racial identities of the faces were balanced so that there were 12 each of African American, Asian and White identities.
Task design. The go/no-go task design required participants to press a button on a computer keyboard after presentation of the target facial expression on the computer screen. They were asked to withhold their response if any other facial expression was presented, i.e. non-target trials. Thus, focus on the facial expressions was task-relevant. Each facial expression served as a target and non-target across 6 blocks of target/non-target pairs: Fear/Neutral, Neutral/Fear, Happy/Fear, Fear/Happy, Happy/Neutral and Neutral/ Happy. Each of the 6 blocks contained 35 targets and 13 non-targets. Non-targets were presented approximately 25% of the time to ensure that responding to the targets was over-learned. The faces were shown on the computer screen for 500ms, with a variable duration inter-stimulus fixation cross of 1-2.5 sec. Figure 2 displays the task schematic.

Measures
As in the stop-signal task, the go/no-go task IIV was computed for each individual as their within-person standard deviation of response times to Go trials (i.e. target stimuli trials) across all task blocks (see Supplemental Materials for discussion and analyses of coefficient of variation as an alternative measure of IIV

Data analysis plan
Analyses were conducted in R (https://cran.r-project. org/) within RStudio Version 2022.07.01 + 554 (RStudio, PBC, Boston, MA). Participants were grouped as younger or older adults to assess age differences in effects. First, to test for main or interactive effects of response inhibition task type (proactive inhibition for go/no-go task, reactive inhibition for stop-signal task) and age group (younger adults, older adults) on IIV, we conducted a 2 × 2 between-subjects ANOVA. Posthoc two-tailed independent samples t-tests comparing results between age groups for each task were used to clarify interactive effects of task type and age group. Levene's tests indicated no significant difference in variance of IIV between age groups for either task (all Fs < 1, all ps > .39).

Stop-Signal task analyses
To examine if task instruction or age group impacted IIV, we computed a 3 × 2 between-subjects ANOVA with factors of task instruction (All Faces, Gender, Facial Expression) and age group (younger adults, older adults). Post-hoc two-tailed independent samples t-tests were completed to clarify main effects of task instructions. To assess if IIV predicts reactive emotional response inhibition, three separate linear regressions were conducted for the stop-signal task, one for each type of stop-signal condition (i.e. neutral, fear and happy). In each of the three linear regression models, age group (younger adults, older adults) and task instruction (All Faces, Gender, Facial Expression) were included as additional predictors to assess whether they had significant interactive effects with IIV to predict SSRT. Only main and interactive effects of IIV on SSRT will be discussed. Follow-up simple slopes tests were used to clarify significant interactive effects.

Go/no-go task analyses
We employed linear regression models to assess if IIV predicts proactive emotional response inhibition, measured with the go/no-go task. Four of the six target/non-target blocks were evaluated to specifically assess effects of emotional faces (happy, fearful) as the target versus non-target (i.e. fear/neutral, happy/ neutral, neutral/fear, neutral/happy) (Waring et al., 2019). In each of the four linear regressions, age group (younger adults, older adults) was included as a predictor to test for significant interactive effects with IIV to predict false alarm rate. Only main and interactive effects of IIV on false alarm rates will be discussed. One younger adult's false alarm rate was identified as a statistically significant outlier (z's > 3.5) in the Neutral/Fear and Fear/Neutral blocks. We removed this participant from regression analyses for the respective blocks and retained their data for analyses on task blocks in which they were not an outlier.

Results
The ANOVA of task type and age group on IIV showed no main effects of task type (F[1, 313] = 2.41, p = .12; Thus, variability in response times for Go trials significantly differed between age groups for the stop-signal task, but not the go/no-go task. Mean response times for Go trials are reported in Table 1.

Go/no-go task results
Only one of the four linear regression models examining if IIV interacted with age group to predict false .20, p = .84). Furthermore, age group did not significantly interact with IIV to predict false alarms in any of these blocks (Bs < .002, ts > 1.85, ps ≥ .067). Together, these analyses demonstrate that variability in response times on Go trials did not have a strong relationship with correctly withholding responses to the facial expressions shown in the go/no-go task. (For full results of linear regression models, see Supplemental Table S3.)

Discussion
The present study assessed if aging or task design impact trial-to-trial consistency in Go trials response times (IIV) in two types of emotional response inhibition tasks, and whether response consistency could predict inhibitory control in younger and older adults. To briefly summarise results, we observed that older adults had greater response consistency than younger adults in the stop-signal task, but not the go/no-go task. There was no mean difference in response consistency between the tasks. Participants who completed the stopsignal task with the Facial Expression instruction had more consistent response times than those with the Gender or All Faces instructions. Response consistency predicted efficiency of inhibition of responses (SSRT) to fearful faces in the stop-signal task only with the Facial Expression task instructions; more consistent responses across Go trials predicted more efficient response inhibition to negative stimuli when the task instructions required focus on emotional stimulus attributes. There were no interactive effects of age group and response consistency to predict emotional response inhibition in either task.
Older adults have more consistent responses than younger adults in the stop-signal task, but not the go/no-go task The stop-signal and go/no-go tasks use fundamentally different response inhibition mechanisms, engaging proactive inhibition in the go/no-go task and reactive inhibition in the stop-signal task (Aron, 2011;Krämer et al., 2013;Liebrand et al., 2017;Littman & Takács, 2017;Raud et al., 2020;Schachar et al., 2007). Directly comparing IIV between tasks clarifies how inhibitory control demands impact how well consistency of mental processes is sustained from trial-to-trial. Prior literature suggested that reactive inhibition tasks (e.g. stop-signal task) require greater executive functioning than proactive inhibition tasks (e.g. go/no-go task) (Verbruggen & Logan, 2008b). However, contrary to hypotheses, those differences in task demands did not yield mean differences in trial-to-trial response consistency in the present study.
While there were no mean differences in IIV between tasks, older and younger adults behaved differently in each task. Younger adults had greater IIV than older adults in the stop-signal task, reflecting that younger adults have less trial-to-trial consistency in response times. Crucially, the significant age effects in the stop-signal task but not in the go/no-go task were still upheld when accounting for mean Go response times in the measure of response variability (e.g. coefficient of variation; CoV. See Supplemental Materials). These additional analyses indicate that effects of aging in the stop-signal task are in fact due to variability in responding, and not an incidental effect caused by ceiling effects in older adults' mean response times from age-related slowing. In summary, our hypotheses that aging would negatively affect IIV was not supported in either task of emotional response inhibition.
Previous literature examining the effect of aging on IIV offers inconclusive results. Many studies have identified significantly greater IIV in older adults compared to younger adults (Dykiert et al., 2012;Fozard et al., 1994;Hultsch et al., 2002Hultsch et al., , 2008Nesselroade & Salthouse, 2004), while others have not detected an age-related increase in response variability (Myerson et al., 2007;Shammi et al., 1998). For example, Nicosia and Balota (2021) reported significantly greater IIV in younger adults compared to older adults while completing a modified Sustained Attention to Response (SART) task. The authors concluded that their results may have been due to a greater propensity of younger adults to engage in mind-wandering compared to older adults (Nicosia & Balota, 2021). Mind-wandering is significantly related to IIV (Unsworth, 2015) and is also more likely to occur when people complete easier tasks (Giambra, 1989;Seli et al., 2018) or have lower motivation (Seli et al., 2021). Accordingly, in the present study, younger adults might have perceived the stop-signal task to be less cognitive demanding or had less motivation than the older adults did, leading to greater mind-wandering, and consequently lower consistency in response times trial-to-trial as they switched back and forth between mind wandering and task responses. Future studies would be enhanced by collecting ratings of subjective task difficulty and motivation from participants to directly inform the possibility of differences in perceived difficulty or motivational factors. Another consideration is that during adolescence and the period of emerging adulthood individuals employ an adaptive strategy of increasingly testing possible courses of action and exploring the outcomes of various choices. Others have speculated that higher IIV in younger adults may reflect testing hypotheses and brainstorming strategies during experimental tasks (Hultsch et al., 2008;Vandermorris et al., 2013). In the context of the current study, this possibility could manifest in relatively greater variability in response times among the younger adult sample. The present study offers new insights into age-related differences in dynamic implementation of inhibitory processes over emotional information.

Overt attention to emotion in the stop-signal task evokes greater response consistency
In support of hypotheses, stop-signal task instructions to focus directly on the facial expressions presented produced significantly greater response consistency (i.e. lower IIV) than either the instructions to focus on the Gender of the face or on All Faces. One possibility is that the latter two versions were less cognitively taxing than the former. The Facial Expression task instruction required greater discernment, as there were 3 possible facial expressions presented (i.e. happy, neutral, fear), whereas the All Faces or Gender instruction only required participants to discern between two options (i.e. face/no face appeared or male/female, respectively). Consequently, the Facial Expression instruction may have required closer attention to the faces shown than the other instructions required. It is also possible that greater cognitive engagement is granted in the condition overtly eliciting attention to emotional information. Previous studies concluded that directing attention to an emotional stimulus using instructions such as "attend to emotional stimuli", compared to a passive viewing condition, may result in deeper evaluation of the emotion-evoking material (Diers et al., 2014;Mancini et al., 2020). Therefore, it is possible that higher degrees of cognitive engagement are needed to significantly increase response consistency. Another consideration is the difference in task switching demands among the three instructions of the stop-signal task. It is possible that task instructions requiring switching between stop cues twice during the Facial Expressions condition (e.g. fear to happy to neutral faces, in turn) may have facilitated sustained alertness across the six task blocks and consequently favourably impacted IIV in that condition versus the task instructions that required switching to a different stop cue only once or not at all (i.e. the Gender or All Faces instruction, respectively). However, there was no significant difference in IIV between the go/no-go taskrequiring switching after each block (i.e. switching 5 times across the 6 task blocks)and the stop-signal task, suggesting that degree of task switching cannot fully account for the observed effects of task instructions on IIV. Taken together, the varying demands for attention to emotion is a more likely explanation for the pattern of observed IIV results than a task switching account.
There was no interaction of task instructions with age on IIV, which demonstrates that instruction to focus on the facial expression of the task stimuli did not uniquely impact response consistency for younger versus older adults. This may be another instance where tasks that highly constrain cognitive processing are less likely to reveal age differences in emotion processing (e.g. positivity bias) than are tasks that do not constrain cognitive processing and where older adults can employ their natural information processing preferences (Reed et al., 2014;Reed & Carstensen, 2012). Analyses using coefficient of variation as the outcome measure of response consistency (i.e. taking mean response time into account) broadly yielded the same conclusions; there were no main or interactive effects of age, and response consistency was significantly greater during the Facial Expression than Gender task instructions (see Supplemental Materials).
Attention to negative emotion fosters positive relationship of IIV and inhibition in stop-signal tasks IIV predicted efficiency of stopping (SSRT) only when participants were directly instructed to stop their responses to fearful faces in the stop-signal task. The participants in the Facial Expression stop cue version of the task with greater response consistency (i.e. lower IIV) also more efficiently stopped their responses to fear faces (i.e. lower SSRT). Notably, this effect was unique to stopping for fearful faces; response consistency did not predict participants' stopping efficiency to neutral or happy faces. Compared to neutral and happy faces, fearful faces (i.e. negative stimuli) retain an advantage in garnering attention that might have been amplified when participants were instructed to specifically allocate their attention to the emotional expression of the face (Hartikainen et al., 2000;Peeters & Czapinski, 2011;Pratto & John, 1991). The present finding that IIV significantly predicted response inhibition only when stopping specifically for negative stimuli is supported by the contingent capture hypothesis because attention to the emotional characteristics of the stimuli was induced by task instructions (Tannert & Rothermund, 2020;Victeur et al., 2020). Another consideration is prior evidence that response consistency corresponds with increased activation in pregenual anterior cingulate cortex (pgACC; Johnson et al., 2015), which is a region also recognised for its strong involvement in processing emotion conflict, a form of spontaneous emotion regulation (Etkin et al., 2011). Although speculative, one intriguing possible explanation for the relationship between IIV and response inhibition only for negative stimuli is degree of pgACC engagement. For example, individuals with greater pgACC activation to facilitate conflict resolution between stop and go processes may have both more efficient stopping for the highly arousing negative stimuli and also more consistent responses across trials (lower IIV). While future neuroimaging studies can shed light on the plausibility of this hypothesis, the present study adds further specificity to the behavioural literature reporting association between lower response variability and greater inhibitory success (Bellgrove et al., 2004;Joly-Burra et al., 2018;Rochat et al., 2013). Consequently, our hypothesis that task instruction and age group would interact with IIV to predict emotional response inhibition in the stop-signal task was partially supported because task instruction (but not age group) significantly interacted with IIV.
Our parallel hypothesis that response consistency would interact with age group to predict response inhibition in the go/no-go task was not supported. In the go/no-go task attending to negative emotion, either as the go or no-go stimulus (Fear/Neutral or Neutral/Fear conditions), did not evidence the significant positive predictive relationship of response consistency to response inhibition seen within the stopsignal task. However, when mean Go response times were accounted for in the measure of response variability (e.g. using CoV), age interacted with CoV to predict false alarms in several of the target/nontarget blocks of the go/no-go task. Specifically, CoV positively predicted false alarm rates in younger adults but not older adults (Supplemental Materials, Table S4, Figures S1-S4), suggesting that lesser trialto-trial consistency in response times (i.e. increasing IIV) has a stronger correspondence to poorer response inhibition in younger than older adults when typical age-related psychomotor slowing is considered.

Descriptive comparison of differences between go/no-go and stop-signal task results
Although we could not directly compare performance on the go/no-go task to stop-signal task statistically due to differing outcome measures of response inhibition (see Introduction for details), descriptive comparisons can illuminate important distinctions in how IIV predicts proactive and reactive inhibition. In the stop-signal task we evaluated the stopping efficiency (SSRT) for the three stimulus types (fear, happy and neutral faces) individually. In the go/no-go task the false alarm rate to non-target stimuli was the measure of response inhibition. These outcome measures have different implications. There is a greater range of possible values in the SSRT measure, in contrast with the restricted range of accuracy and possibility of floor effects for false alarm rate. We speculate that the resolution of a continuous measure with a larger range, such as SSRT, is more sensitive to reflecting the relationship between response consistency and emotional response inhibition. Moreover, accuracy cannot be an outcome measure in stop-signal task, as it is an adaptive task where stimulus timing (stop signal delay; SSD) adjusts to maintain performance accuracy around 50%. Therefore, although it is not possible to directly compare the response inhibition measures to one another, they can each inform the relationship of IIV to response inhibition and indicate significant moderating effects of aging and task demands. Collectively, the results of this study imply that the relationship between response inhibition and consistency differs between the two task types, advancing the literature distinguishing the cognitive demands of proactive and reactive inhibition (Aron, 2011;Littman & Takács, 2017;Raud et al., 2020).
Taken together, although both the go/no-go task and the Facial Expression instruction of the stopsignal task required task-relevant attention to emotion, IIV predicted emotional response inhibition only when the stop-signal task instructionsbut not go/no-go task instructionsrequired overt attention on negative emotion. The stop-signal task of reactive inhibition may have been more demanding than the go/no-go task of proactive inhibition (Patterson et al., 2016;Verbruggen & Logan, 2008a), so the combination of focus on negative emotion specifically during a more demanding executive functioning task may have impacted cognitive resources sufficiently to commensurately modulate trial-to-trial response consistency.

Limitations and future directions
One limitation of the present study is that the stopsignal task and go/no-go task evaluated included only three types of facial expressions: fear, happy and neutral. We observed that stopping for fearful faces revealed a unique relationship to response consistency, yet it is unclear whether effects are due to the arousal level or valence of the fearful faces (i.e. high-arousal negative). Future studies could investigate whether effects similar to those reported here are evoked by low-arousal negative stimuli, like sadness. Ding et al. (2020) suggested that lowarousal stimuli impair response inhibition in a unique way from high-arousal stimuli, and so it would be worthwhile to test whether the relationship between IIV and emotional response inhibition differs for sad faces and other low-arousal stimuli.
A future study investigating within-subject individual differences in response consistency between proactive and reactive emotional response inhibition tasks across varying task instructions would also be highly informative. The present study was a between-subject study design, as participants were assigned to one of three stop-signal task instructions (All Faces, Gender, Facial Expression) or the go/no-go task to minimise participant time burdens. A multiday study where participants complete all study versions, while distributing the time burdens to minimise effects of fatigue, would also advance understanding of the relationship of response consistency to cognition. Additionally, although the go/no-go task was a within-subject design where all participants completed all task blocks, task demands were not systematically varied to evaluate attention to emotion as task-relevant versus not task-relevant. A fully crossed design where the stop-signal task and go/no-go task both employ all 3 task instructions manipulations within-subjects would be an optimal experimental design to implement in future studies.
Other potential future directions include expanding recruitment to test effects across the adult lifespan and beyond samples with typical cognition. Broadening the sample age range to include a mid-life cohort in addition to younger and older adults would expand understanding of how IIV is affected by age as a continuous variable. Comparisons to clinical samples with known executive functioning defects (e.g. individuals with mild cognitive impairment or ADHD) could clarify how relationships between IIV and emotional response inhibition vary in samples with differing neural and cognitive functioning.

Conclusion
This study offers new evidence of how aging and task design (e.g. inhibitory control demands, task instructions) influence response consistency and of how response consistency interacts with task instructions to predict emotional response inhibition. In summary, in contrast to previous literature studying age-related differences in IIV, we found that older adults were more consistent than younger adults in the reactive emotional inhibition task. This effect was not present in the proactive emotional inhibition task. Additionally, task instructions explicitly directing participants' attention to emotional attributes of the facial expressions promoted more consistent trial-to-trial response times for younger and older adults. We also identified that trial-to-trial response consistency positively predicts emotional response inhibition only when task demands require directly focusing on negative facial expressions in the context of a reactive inhibition task. The present study clarifies the relationship of cognitive consistency and aging in the context of cognitively demanding executive functioning tasks, highlights important distinctions between cognitive processing during proactive and reactive inhibition tasks, and extends understanding of the impact of attention to emotion on response consistency.