Slow neural oscillations explain temporal fluctuations in distractibility

Human environments comprise various sources of distraction, which often occur unexpectedly in time. The proneness to distraction (i.e., distractibility) is posited to be independent of attentional sampling of targets, but its temporal dynamics and neurobiological basis are largely unknown. Brain oscillations in the theta band (3 – 8 Hz) have been associated with fluctuating neural excitability, which is hypothesised here to explain rhythmic modulation of distractibility. In a pitch discrimination task (N = 30) with unexpected auditory distractors, we show that distractor-evoked neural responses in the electroencephalogram and perceptual susceptibility to distraction were co-modulated and cycled approximately 3 – 5 times per second. Pre-distractor neural phase in left inferior frontal and insular cortex regions explained fluctuating distractibility. Thus, human distractibility is not constant but fluctuates on a subsecond timescale. Furthermore, slow neural oscillations subserve the behavioural consequences of a hitherto largely unexplained but ever-increasing phenomenon in modern environments – distraction by unexpected sound.


Introduction
Selective attention enables humans to focus on relevant information at the expense of 47 distraction. The brain prioritizes representations of relevant events while filtering out task-48 irrelevant distractors (Desimone & Duncan, 1995;Picton et al., 1971). Recent research posited 49 that distractor processing is not merely collateral to attentional sampling of targets but may 50 follow its own dynamics (Schneider et al., 2018;Wöstmann et al., 2019Wöstmann et al., , 2020. The  Here, we ask if the brain spontaneously alternates between states of higher and lower 84 distractibility and whether such fluctuations have the potency to explain behavioural 85 consequences of distraction. If so, we would expect to observe a brain-behaviour relation 86 between the pre-distractor brain state and the distractor-induced detriment in task performance. 87 To this end, we employed a pitch discrimination task wherein an auditory distractor could occur 88 at variable and unexpected times in-between two target tones. We probed this research question 89 in the auditory modality as temporal information is especially important to auditory attentional 90 selection (Shamma et al., 2011). During the task, participants had to identify whether the two 91 target tones were the same or different in pitch (Fig. 1A). The distractor was a fast-varying, 25-92 Hz modulated sequence of tones that differed in pitch, which allowed us to extract its evoked 93 25-Hz neural response (Ding & Simon, 2009). 94 We used behavioural sensitivity and distractor-evoked neural response as the 95 behavioural and neural proxies of distraction. Behaviourally, perceptual sensitivity was 96 calculated as an indirect measure of distraction: The more distracted, the lower the sensitivity 97 in pitch discrimination should be. Neurally, we extracted the amplitude of the distractor-evoked 98 event-related potential (ERP) at 25 Hz, which corresponded to the modulation rate of the 99 frequency-modulated distractor tone sequence. Although these post-distractor measures may 100 not solely reflect distractibility but also other distractor-related processes (e.g., suppression), 101 we should observe temporal fluctuations of the two measures if distractibility exhibits temporal 102 dynamics. If there is a brain-behaviour relation in the temporal fluctuations in distraction, we 103 should also observe temporal co-fluctuations between the two measures. Furthermore, we 104 aimed to unveil the neural origins of the distractibility dynamics. If a brain region is involved 105 in the momentary changes in distractibility, we should observe a relationship between the pre-   Thirty participants (20 females, 10 males; mean age = 23.67, SD = 3.56) took part in the EEG 117 experiment. They provided written informed consent and were compensated by either €10/hour 118 or course credit. Participants were right-handed according to the Edinburgh Handedness 119 Inventory (Oldfield, 1971) (mean score = 92), with self-reported normal hearing, normal or 120 corrected-to-normal vision, and no psychological or neurological disorders. All procedures of 121 the current study were approved by the ethics committee of the University of Lübeck.

123
Participants performed a pitch discrimination task wherein they decided whether the first (tone 124 1) and the second (tone 2) target tones in a trial were the same or different in pitch. Prior to the 125 experiment, they were instructed to answer as accurately and as fast as possible. The target 126 tones were 75 ms long pure tones with 5 ms rise and fall periods. In each trial, the frequencies 127 of tone 1 were randomly selected between musical note A#3 (233 Hz) and G#5 (830.6 Hz), 128 while that of tone 2 was either the same (50%) or different (higher or lower, 25% each) in 129 frequency compared to tone 1.

130
The pitch difference between tone 1 and tone 2 was titrated for each participant with an 131 adaptive task (see below). The offset-to-onset interval between tone 1 and tone 2 was 1550 ms.

138
In-between the two target tones, a distractor was presented in 50% of trials (distractor-139 present condition) and no distractor was presented in the remaining trials (distractor-absent 140 condition). The inclusion of distractor-absent trials serves two purposes. First, we could verify 141 that the distractors had the potency to distract by comparing behavioural performance for 142 distractor-present versus distractor-absent trials . Second, participants 143 could not anticipate whether or when a distractor would occur in a given trial, which eliminated 144 7 potential effects of such anticipation on behavioural performance (Grabenhorst et al., 2021) or 145 pre-stimulus neural activity (Dürschmid et al., 2018;Herbst et al., 2022;Stefanics et al., 2010). 146 The distractor was a tone sequence which consisted of 10 40-ms tone pips, thus creating a 25-147 Hz temporal structure (total duration: 400 ms).

148
In the distractor-present condition, the distractor was presented at one of 24 distractor 149 onset times (0 ms to 1150 ms, 50-ms steps, relative to the offset of tone 1), which was selected 150 at random on each trial. The distractor-absent trials were randomly assigned to the 24 distractor 151 onset times. Specifically, for each distractor-absent trial, a distractor onset time was assigned 152 as in the distractor-present trial. A distractor is however not presented during stimulus 153 presentation in the distractor-absent condition. Time 0 in a distractor-absent trial, therefore, 154 refers to the time when a distractor would have been presented in that trial.

155
After the offset of target tone 2, participants had a 2000 ms response time window.

156
After the presentation of tone 2, a prompt is shown on the screen asking if the two target tones 157 were the same or different in pitch ("same" or "different"?). Participants were only allowed to 158 respond after the presentation of tone 2. Any button press beforehand was thus not recorded.

159
To avoid potential temporal predictability effects of the onset of the next trial, the inter-trial 160 intervals were randomly selected from a truncated exponential distribution (mean = 1460 ms), 161 ranging between 730 and 3270 ms.

162
The trial order was pseudo-randomized with no repetition in probe tone frequency and 163 distractor onset for any two consecutive trials. In total, there were 12 trials for each unique 164 condition (distractor-present/absent x distractor onset x same/different target pitch) and 1152 165 trials for the whole experiment. All auditory materials were presented via Sennheiser 166 headphones (HD 25-1 II). Responses were made using a response box (The Black Box Toolkit).

167
The assignment of buttons to the response options ("same" or "different") was counterbalanced  Prior to the main experiment, each participant's threshold for the pitch discrimination task was 173 titrated using an adaptive staircase procedure, implemented in the Palamedes toolbox (Prins & 174 Kingdom, 2018) for Matlab. For the initial 11 participants, the threshold was titrated to an 175 8 approximate accuracy of 70.7%. As the overall accuracy was relatively high even after the 176 adaptive staircase procedure for these 11 participants (mean = 79.59%, SD = 10.43%), the final 177 16 participants performed an adaptive procedure altered to yield approximately 65% accuracy 178 instead. Due to technical issues, performance of the remaining three participants was tracked 179 at 35% accuracy. As all relevant statistical analyses in the present study are within-subject, and 180 as paired t-tests (2-tailed) comparing the behavioural performance between distractor-absent 181 and distractor-present conditions were significant with (t29 = 8.11, p < .001) and without (t26 = 182 9.41, p < .001) these participants, their data were included in the final analysis.

183
Each participant went through the adaptive staircase procedure two to three times, 184 depending on the stability of the tracked threshold. The range of frequencies used in the 185 adaptive task was the same as that used in the main experiment (i.e., 233 Hz to 830.6 Hz).

186
There were in total 30 trials for each run of the adaptive staircase procedure with an initial pitch    Hit rate was defined as the "different" response when the two tones were different in 212 pitch, and false alarm rate the "different" response when the two tones were the same in pitch.   dependence (see Fig. S5).

343
Second, when doing the quadratic fit analysis on one fixed time window only, the 344 quadratic fit may fail to capture the phasic relationship that manifests as a sine wave function. 345 We circumvented this potential caveat by running a time-resolved quadratic fit analysis. By

365
One may ask how the dynamics in distractibility relate to rhythmic attentional sampling.

366
As the current study mainly focuses on distractibility dynamics, we did not use the dense      (Fig. 2D). 435 We then ran a linear mixed model with the cross-correlation coefficient as the outcome measure 436 and sine-and cosine-transformed time lag as predictors. Pearson's r = -0.08; Spearman's r: t29 = -2.13, p = 0.04, mean Spearman's r = -0.10).

444
As a control analysis, the same analysis pipeline was run on the data in the distractor-445 absent condition by randomly assigning a "distractor onset" for each distractor-absent trial, 446 which did not reveal any significant co-modulation (Fig. S7)   To test whether the significant cluster overlaps with sources of auditory-evoked activity 495 in auditory cortex regions, we compared its source with the source of distractor-evoked inter-496 21 trial phase coherence (ITPC) at 3 -7 Hz (shown also in Fig. 2B, bottom panel). Importantly, 497 although the two effects were localized in proximal cortical regions (Fig. 3C, bottom panel), 498 their core regions were mostly non-overlapping. 499 For control, we conducted the same analysis on the distractor-absent trials, which 500 revealed no significant cluster (Fig. S10). We also tested the relationship between the pre-501 distractor neural phase and the post-distractor neural measure of distraction (i.e., 25-Hz 502 amplitude of the distractor-evoked ERP), which did not reveal a significant effect (Fig. S11). 503 Lastly, we tested whether there is also a quadratic relationship between pre-tone 2 neural phase 504 and behavioural sensitivity. While there was no significant cluster in the distractor-present 505 condition (all p > .06), a significant positive cluster was found in the distractor-absent condition 506 (p = .01), but in different neural regions (i.e., left lingual gyrus and right inferior frontal cortex) 507 from the left insular/inferior frontal origins found for distractibility dynamics. Importantly, in 508 both distractor-present and distractor-absent trials, there was no significant cluster in the left 509 insular or inferior frontal cortex, suggesting that pre-tone 2 neural phase in these regions does 510 not explain fluctuations in behavioural sensitivity (Fig. S12).   The current study aimed to unravel the temporal dynamics of distractibility, using a pitch 533 discrimination task with auditory distractors. The eventual degree of distraction and the neural 534 processing of distractors were respectively quantified by distractor-evoked performance 535 detriments and neural responses in the human electroencephalogram (EEG). We made a series 536 of interesting observations.   We did not observe a corresponding phasic relationship between pre-tone 2 neural 644 phase and behavioural sensitivity in the current study. Nevertheless, the absence of evidence 645 here should not be taken as the evidence of absence as we did not manipulate target onset time 646 after phase reset, which is crucial for unveiling any dynamics in auditory attentional sampling  Against what might have been expected, pre-distractor theta phase did not predict 652 fluctuations in the distractor-evoked neural response (Fig. S11). There are two salient reasons 653 why the present data do not bear out a direct correspondence between pre-distractor neural The present study demonstrates that human proneness to distraction is not uniformly distributed 667 across time but fluctuates on a subsecond timescale in cycles of ~3 to 5 Hz. In the brain, time 668 windows of higher distractibility are coined by stronger neural responses to distractors.