Task-irrelevant sounds unexpectedly deviating from a predictable sequence induce an orienting response characterized by specific electrophysiological responses and behavioral distraction (Schröger, 1996, 2005; Escera et al., 1998; Berti and Schröger, 2001; Horváth et al., 2008; Parmentier et al., 2008; Berti, 2013; Parmentier, 2014; Rosburg et al., 2018). This is typically studied in simple forced-choice categorization tasks involving task-irrelevant sounds (Escera et al., 1998; Pacheco-Unguetti and Parmentier, 2014; Parmentier et al., 2014, 2018), though it is not specific to that task and has been observed in others such as duration judgment, go-no go, visual matching, serial recall, or gap detection tasks (Berti and Schröger, 2001; Hughes et al., 2005; Bendixen et al., 2010; Li et al., 2013; Röer et al., 2015, 2017, 2018; Pacheco-Unguetti and Parmentier, 2016; Körner et al., 2017; Vachon et al., 2017; Volosin et al., 2017; Volosin and Horváth, 2020).

In cross-modal oddball tasks, target stimuli are typically visual, each preceded by a task-irrelevant sound which is repeated on most trials and unexpectedly replaced by a different sound on some trials. These unexpected changes typically entail longer response times (RT) and sometimes a small reduction in response accuracy (see Parmentier, 2014, for a review). This effect is thought to reflect the transient inhibition of motor actions (Wessel and Aron, 2013; Parmentier, 2016; Wessel, 2017; Vasilev et al., 2019, 2021; Wessel and Huber, 2019) and the involuntary shift of attention to and from the unexpected sound (Schröger, 1996; Parmentier et al., 2008). These behavioral manifestations are thought to reflect an adaptive interruption of ongoing actions, the assessment of whether a new action plan must be selected and, if not, the reactivation of the relevant task set (Corbetta and Shulman, 2002; Verbruggen et al., 2014; Wessel and Aron, 2017). Importantly, unexpected sounds distract because they violate the cognitive system’s predictions, not because they are rare per se (Schröger et al., 2007; Parmentier et al., 2011; Volosin and Horváth, 2014; Schröger and Roeber, 2021). Consequently, distraction lessens or disappears when unexpected sounds are made predictable (e.g., Sussman et al., 2003; Horváth and Bendixen, 2012; Parmentier and Hebrero, 2013), but it appears immune to the predictability of our own behavior (Parmentier and Gallego, 2020).

Several studies have examined distraction by unexpected sounds in young and older adults. In studies using uni-modal tasks, aging does not appear to modulate deviance distraction. While one study using an auditory duration discrimination task reported an age-related increase in distraction by rare and unexpected changes in pitch (Berti et al., 2013), a subsequent study using a larger sample found no difference between the two age groups (Leiva et al., 2015a). Using a similar task in children, young and older adults, Horváth et al. (2009) found no difference between young and older adults with respect to behavioral deviance distraction. Similar degrees of behavioral distraction were also observed in young and high-performing older adults (Getzmann et al., 2013).

In contrast to uni-modal studies, cross-modal oddball studies consistently show an age-related increase in distraction by unexpected sounds. Using a visual digit categorization task, Andrés et al. (2006); see Parmentier and Andrés, 2010, for a replication) reported a twofold increase in distraction in older compared to young adults as measured from RT. A study comparing the two age groups in uni-modal and cross-modal oddball tasks confirmed a selective effect of aging on distraction in the cross-modal task (Leiva et al., 2015b). Converging findings were also reported in a cross-sectional study comparing children, young and older adults with respect to deviance distraction, working memory capacity and response inhibition (Leiva et al., 2016). The exact origin of the selective effect of aging on deviance distraction in the cross-modal as opposed to uni-modal tasks has not clearly been identified yet. The cross-modal task differs from the uni-modal in that it involves attention crossing sensory boundaries, which has been hypothesized as a one contributor to the effect (Parmentier et al., 2008). Hence, one reasonable hypothesis is that aging selectively affects the deployment of attention across sensory modalities while within-modality shifts are relatively unscathed (Leiva et al., 2015b).

Evidence shows that, relative to a repeated and predictable sound, unexpected sounds distract participants from the task at hand and yield longer RT. Interestingly, when this task is visual, distraction is significantly larger in older than in young adults (Andrés et al., 2006; Parmentier and Andrés, 2010; Leiva et al., 2015b, 2016). However, one factor has been overlooked in these studies: the general slowing exhibited by older adults. Indeed, while these studies reported age-related differences in RT between unexpected and standard sound conditions, the baseline response speed of participants was not considered. In other words, one cannot rule out the possibility that this age-related increase of distraction may simply be a proportional effect.

It is well-established that variations in cognitive and response speed constitute a key general factor in accounting for age-related changes in memory and cognition (Shimamura, 1994). Gradual slowing is a well-documented hallmark of aging (Cerella, 1985; Rogers and Fisk, 1990; Salthouse and Meinz, 1995; Salthouse, 1996, 2000). Not surprisingly, older adults were overall slower than young adults in all four cross-modal studies reviewed above. Hence, before concluding that unexpected sounds selectively affect older adults in cross-modal oddball tasks, it is necessary to examine whether the proportional increase in RT in the unexpected sound condition relative to the standard condition is greater for older than for young adults.

Here, we reanalyzed the data from the four cross-modal oddball studies on aging (Andrés et al., 2006; Parmentier and Andrés, 2010; Leiva et al., 2015b, 2016) to test for an age-related increase in distraction while controlling for baseline response speed. To examine this issue, we compared young and older adults using a proportional metric of distraction whereby distraction (RT unexpected – RT standard) was pitched against baseline response speed (RT standard). We then used Bayesian estimation on the combined data set to derive point-estimates and 95% high density intervals (95% HDI) of the mean difference in proportional distraction and its effect size. The hypotheses were simple: If the age-related difference in RTs between unexpected and standard sound trials reflect a genuine increase in distraction, then (1) we should find significant differences in proportional distraction effects between the two age groups and (2) Bayesian estimation should reveal credible value ranges for the age group difference in proportional distraction and its effect size that exclude zero. In contrast, if the age-related increase in distraction reported in previous studies is a mere reflection of a proportional slowing of responses, then (1) we should find no difference between the two age groups when controlling for baseline response speed, and (2) Bayesian estimation should highlight zero as a credible value for the mean age-related difference in distraction and its effect size.

The sample included a total of 204 participants (148 females and 56 males) forming two age groups: 108 young adults (M age = 21.62, SD = 3.68) and 96 older adults (M age = 67.11, SD = 8.38). We present in Table 1 (Panel A) a descriptive breakdown of age and sex per study. For an average effect size of d = 0.963 (mean of the d values for the effect of aging on deviance distraction in the four studies we are revisiting), the statistical power of a t-test for this sample size, a Type I error of 0.05, and a one-tailed hypothesis, the statistical power of our study was >0.999. Put differently, given our sample size, the minimum effect size affording a power of 0.95 was d = 0.463.

The cross-modal oddball task presented the same general structure in all four studies, though with small variations with respect to the exact stimuli, proportion of unexpected sound trials, number of test trials, duration of stimuli, inter-stimuli and inter-trial intervals. For clarity, we present a descriptive comparison of the task characteristics in Table 1 (Panel B). In all cases, the participants’ task was to perform a binary categorization of a visually presented stimulus (a digit in three studies, a cartoon dog in one). A task-irrelevant sound preceded each visual target. The duration of the sounds, sound-to-target, target, and trial-to-trial intervals was fixed within each study and varied slightly across studies. In the majority trials (70–90%, depending on the study), a 600 Hz sinewave tone served as the standard sound while unexpected sounds (white noise or environmental sounds) was used in the remaining trials. Participants were instructed to attend the visual task and endeavor to respond as quickly as possible while trying to make no error. All participants were tested individually in a sound attenuated laboratory. Sounds were presented through headphones with a level of approximately 75 dB SPL.

For clarity, we present the full list of statistical results for our first analysis in Table 2 (Panel A). As visible from these results, both non-parametric and Bayesian statistics supported our experimental hypothesis. Deviance distraction was significantly and credibly greater in older adults in all four studies for RT, and in two studies for the PC (while in the remaining two, the results were inconclusive). As visible in Table 2 (Panel B), the BANOVAs carried out on the combined sets of data provided strong (PMD_PC) and extreme (PMD_RT) evidence of the effect of aging on our proportional measures of distraction. While the evidence favored the conclusion that PMD_PC did not vary across studies, the BF₁₀ of 0.346 was inconclusive. In contrast, extreme evidence supported the absence of variation of PMD_RT across studies (BF₁₀ = 0.003). Importantly, the results supported the absence of interaction between Study and Age Group: BF₁₀ = 1.243/7.371 = 0.169, and BF₁₀ = 7308.905/48004.861 = 0.152, for PMD_PC and PDM_RT, respectively. The PMD_PC and PMD_RT values for participants in each age group are presented in Figure 1. The results can be summarized as follows. The strength of the evidence of the impact of aging on PMD_PC varied across individual studies but was strong in the combined data set. More importantly, the data clearly supported the hypothesis that aging increases PMD_RT, with moderate to strong evidence in individual studies, and extreme evidence in the combined data set.

Bayesian estimation was carried out on PMD_RT and PMD_PC to generate credible posterior parameters given our data. All Gelman-Rubin diagnostic values were equal to 1 (confirming that convergence was reached), and the effective sample sizes varied between 18,579 and 62,535 (i.e., superior to 10,000, the recommended value for accurate and stable estimates of the 95% HDI bounds, Kruschke, 2015).

We report the complete set of results for every parameter in Table 3 and focus here on the key relevant findings. For PMD_RT, the point estimate for the mean difference between young and older adults was 0.035 (95% HDI: 0.022–0.076), indicating that unexpected sounds produce a proportional increase of RT that is 3.5% greater in older than in young adults. The point estimate of the mean effect size for this difference was 0.8 (95% HDI: 0.496–1.114), a large effect (Cohen, 1988; Lakens, 2013). Of importance, neither of the 95% HDIs included the zero value. It is worth pointing out that this 95% HDI does not overlap with a Region Of Practical Equivalence (ROPE) corresponding to a small effect size (d = 0.2) centered around zero (Kruschke, 2011, 2018). Hence, we can be confident that the effect of aging on PMD_RT is not equivalent to a null or a small effect.

The analysis of PMD_PC revealed results in the same direction as PMD_RT. The point estimate for the mean difference between the two age groups was 0.010, suggesting that, relative to their performance in the standard condition, older adults see their PC decrease by 1% more than young adults (95% HDI: 0.002–0.019). While this difference is small in absolute terms, its statistical effect size estimate was small to medium (0.419). The percentage overlap between the 95% HDI (0.074–0.760) and ROPE (as defined as above) was 3.214, suggesting that the probability that the effect size estimate is equivalent to a small effect is 0.032.

The evidence above shows that aging increases deviance distraction in cross-modal tasks. As we pointed out earlier, past work using uni-modal auditory tasks contrasts with cross-modal studies in reporting no effect of age on deviance distraction. However, these conclusions were based on absolute differences in RT and did not consider age-related variations in response speed. Hence, as a supplementary analysis, we sought to confirm the selective effect of aging on deviance distraction by using PMD_RT to compare cross-modal and uni-modal (auditory) tasks. We included data from three of our cross-modal oddball studies (Andrés et al., 2006; Parmentier and Andrés, 2010; Leiva et al., 2016) and two past studies using an auditory uni-modal task (Leiva et al., 2015a, b). The uni-modal tasks of Leiva et al. (2015b) was identical to its cross-modal version but involved the parity categorization of an auditory (as opposed to a visual) digit. Note that because the study by Leiva et al. (2015b) compared cross-modal and uni-modal tasks within-participants, we could not include both in our analysis and opted to include the uni-modal data set. From the study by Leiva et al. (2015a), we used the data from the auditory task in which 84 participants (42 young adults, M_age = 20.9, and 42 older adults, M_age = 62.8) performed a binary tone duration judgment task in which 88% of trials involved a standard 1,000 Hz tone, and the remaining 12% involved a deviant tone. The sample included in our analysis comprised of 150 young and 138 older adults. We conducted a 2 (Age group) × 2 (task type: cross- vs. uni-modal) BANOVA on PMD_RT. The results revealed strong evidence for the main effects of age group (BF₁₀ = 12.087) and task type (BF₁₀ = 13.927). More importantly, extreme support was found for the interaction between these factors (BF₁₀ = 2454.720). Bayesian Mann-Whitney tests were carried out to analyze this interaction (under the hypothesis of greater distraction in older than in young adults). In line with the results reported in the sections above, older adults exhibited greater PMD_RT than young adults (W = 4,601, BF₁₀ = 1127.591) in cross-modal tasks. In contrast, we found strong support for the absence of a difference between the two age groups in the uni-modal tasks (W = 1,714, BF₁₀ = 0.090).