Article activity feed

  1. Author Response

    Reviewer #3 (Public Review):

    The paper considers two brain measures in younger and older adults, EEG and pupil size fluctuations. Although the relationship of both measures to the reaction time variability is described separately in great detail, the findings of both measures are not combined: for instance, it is not clear if and how their contributions to the behavioural variability interact, whether they explain different aspects of the behavioural variability, etc. In my view, the paper would improve from adding a coherent picture of how these two measures contribute to the behavioural variability together.

    We have now added these analyses to the manuscript. We found that single trial CNV amplitude and the single trial amplitude of pupil dilation responses were correlated and that the correlation coefficients were stronger after adjusting each signal for the slow ongoing signal fluctuations. However, the amount of reaction time variance they together explain was significantly higher than the amount of variance each one of the evoked responses explained on its own suggesting that they contribute independently towards reaction time variability. See Results Section pages 24/25 and Discussion page 30/31.

    The main component of the EEG signal that the authors look at is the amplitude of the Contingent Negative Variation (CNV). The main analysis window for the CNV amplitude is 1-1.5 sec post-cue onset (see for example the grey bar in Fig2A). A clear motivation for choosing this particular window is lacking, leaving open the possibility that the reported results are dependent on this particular analysis window that was chosen.

    This time window was chosen to include the period of highest amplitude of the preparatory response while avoiding any activity related to target processing and therefore positioned just before the earliest target onset (1.5 s after cue onset). We chose a 500 ms window as a compromise between averaging over time, and therefore reducing the inevitable ‘noise’ associated with single trial responses and being close enough to the onset of the target stimulus. This might have biased the results towards lower frequencies, i.e., a hypothetical additive relationship between the amplitude at higher frequencies and variability in the evoked responses might not be detected because that variability was averaged out in the 500 ms time window. Nevertheless, given the 1/f property of the frequency spectra of the signals any effect of higher frequencies was always going to have reduced impact in comparison with the effect of low frequencies with higher amplitude.

    We did not do a systematic search for the “best” time window to use for our analyses. We did do some exploratory analyses regarding the association between the preparatory responses and reaction time and noticed that the association was stronger closer to the time of target onset.

    The motivation for choosing this particular window is now better explained in the Methods Section page 34 and legend of Figure 2.

    The authors distinguish between two factors that contribute to variability in evoked responses: differences in brain state, or a simple summation of two independent signals (fluctuating baseline plus evoked response). They argue for the latter explanation for their data, for good reasons. However, I would like to point out that many studies on pupil size suggest that fluctuations in pupil size are caused by fluctuating brain states (e.g. Pais-Roldan et al, PNAS 2020; Reimer et al, Nat Comm 2016; Yuzgec et al, Curr Biol 2018). The authors could use the Discussion section of the paper to explain how they integrate these findings with their own results on simple summation of ongoing and evoked signals.

    Now that we also include the study of the relationship between ongoing pupil signal and reaction time showing that pupil baseline correlates with reaction time, we can say that, as has been demonstrated in previous studies, our findings also support the idea that pupil fluctuations reflect changes in brain state that affect task performance. We discuss this topic by arguing that the ongoing pupil signal is likely comprised of different components some of which might not be associated with task performance. See Discussion Section pages 28/29.

    Was this evaluation helpful?
  2. Evaluation Summary:

    This manuscript is of interest to cognitive neuroscientists working in the field of (fMRI or EEG) resting-state fluctuations. The role of these fluctuations is compellingly demonstrated in solving an existing mystery about brain variability and ageing; namely, that older adults exhibit increased behavioural variability but reduced neural variability. The work should be of interest to cognitive neuroscientists using fMRI and EEG to study neural noise and inter-individual and inter-group differences, particularly in the realm of aging and age-related disorders.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #3 agreed to share their name with the authors.)

    Was this evaluation helpful?
  3. Reviewer #1 (Public Review):

    This paper sets out to address a conundrum in the literature on cognitive aging - that older participants tend to exhibit increased behavioural variability on cognitive tasks despite having decreased neural signal variability. Here, the authors tested the theory that this apparent discrepancy might reflect the influence of background slow oscillations that change with age but without necessarily being tied to the observed changes in behaviour. Data were collected from a group of older and younger participants who performed a Go/No-Go task in which the onset of the critical stimuli was foreshadowed by a warning cue. Older participants exhibited increased raw RT variability on the task but equivalent coefficients of variations. ERP analyses centered on the CNV signal which builds in anticipation of the critical stimulus. Older adults had reduced CNV amplitude variability and a relationship between ERP amplitude and RT was observed, although this was focussed over left frontal electrodes rather than the frontocentral electrodes where the CNV is focussed. EEG and pupillometry analyses showed that older subjects had lower slow oscillation amplitudes in both modalities. When these differences were controlled for, the group difference in CNV and pupil dilation amplitude variability disappeared and stronger relationships with RT was observed. Similarly, a significant relationship between pupil dilation and RT was only evidence after controlling for the slow oscillations. These results suggest that behaviourally-irrelevant slow oscillations can potentially confound comparisons of aging effects on ERP and pupil measurements as well as their relationships with behaviour.

    This is a very interesting paper that addresses a long overlooked topic in the neurobiology of aging. Where there has been growing awareness of the impact of age-related vascular changes on BOLD responses in fMRI research, equivalent issues have rarely been considered in the EEG literature. This study therefore highlights an important issue that should be of broad interest and relevance to the field, with implications not only for studies of aging but any study comparing individuals or groups.

    Some questions do remain. The contention that the ERP-RT relationships observed in this study pertain to the CNV specifically are questionable since the reported effects are focused over left frontal areas that may be more closely tied to preparation of the right handed button clicks participants were making.

    Was this evaluation helpful?
  4. Reviewer #2 (Public Review):

    This paper is very well written and the analyses are very competently carried out. It reveals a very interesting aspect concerning the interplay between evoked responses, aperiodic activity and behavior. A strong aspect about the paper is that the results are quite convincing, and it is clear that aperiodic activity should be taken into account in future studies working on differences in the variability of evoked signals. A potential weakness is that this study feels more like a technical report, demonstrating an important "confound", rather than it teaching us something on its own (which might partly be the consequence of the focus of the paper).

    Was this evaluation helpful?
  5. Reviewer #3 (Public Review):

    Significance:

    The paper provides a rigorous analysis of the EEG and pupil data, and the results of this analysis sufficiently explain the apparently contradictory findings that the paper set out to investigate, namely that older adults exhibit increased behavioural variability but reduced neural variability. As such, I feel that these findings will be of great interest to researchers in the field of neuroscience and ageing, and more generally, the methods employed on the data might help EEG researchers in the resting-state activity field.

    Strengths:

    - The key to explaining the reduced neural variability in older compared to younger adults lies in the fact that their ongoing neural activity is fluctuating less. The authors do a very thorough analysis of this ongoing EEG and pupil signal, separating the aperiodic from the periodic component, checking the PSD, and computing the phase of both of these signals at cue onset. This in-depth analysis inspires confidence in their findings and could be of interest to other EEG researchers in the resting-state field.

    Weaknesses:

    - An important aim of the paper is to explain the increased behavioural variability in older adults. However, only a limited part of the behaviour, namely the reaction times on the employed go/nogo task, is being reported and analysed. It is easily imaginable that there are differences also between younger and older adults in terms of hit - misses - false alarms - correct rejections. It would be helpful if the paper provided a more complete picture of the behaviour.

    - The paper considers two brain measures in younger and older adults, EEG and pupil size fluctuations. Although the relationship of both measures to the reaction time variability is described separately in great detail, the findings of both measures are not combined: for instance, it is not clear if and how their contributions to the behavioural variability interact, whether they explain different aspects of the behavioural variability, etc. In my view, the paper would improve from adding a coherent picture of how these two measures contribute to the behavioural variability together.

    - The main component of the EEG signal that the authors look at is the amplitude of the Contingent Negative Variation (CNV). The main analysis window for the CNV amplitude is 1-1.5 sec post-cue onset (see for example the grey bar in Fig2A). A clear motivation for choosing this particular window is lacking, leaving open the possibility that the reported results are dependent on this particular analysis window that was chosen.

    - The authors distinguish between two factors that contribute to variability in evoked responses: differences in brain state, or a simple summation of two independent signals (fluctuating baseline plus evoked response). They argue for the latter explanation for their data, for good reasons. However, I would like to point out that many studies on pupil size suggest that fluctuations in pupil size are caused by fluctuating brain states (e.g. Pais-Roldan et al, PNAS 2020; Reimer et al, Nat Comm 2016; Yuzgec et al, Curr Biol 2018). The authors could use the Discussion section of the paper to explain how they integrate these findings with their own results on simple summation of ongoing and evoked signals.

    Was this evaluation helpful?