The stability of thought: using experience sampling and brain imaging to determine the contextually bound nature of human cognition

Reviewer #1 (Public review):

Summary:

This paper suggests an alternative model for the function of the multiple demand network. Specifically, its role is not necessarily to sustain cognitive control and maintain task sets, but rather to "stabilize task-appropriate modes of thought". Evidence for this would be that the MD network is responsible for maintaining a particular thought state during a task. To investigate this, they use a combination of fMRI brain data during a set of 14 tasks and experience sampling in a different set of participants performing the same tasks. Using dimensionality reduction, they reduced the space of task features (and brain systems) to a smaller, more tractable set of dimensions and examined whether stability in specific thought components was related to recruitment of specific brain systems during the task.

Strengths:

Overall, this is an interesting and creative study with strong analytic methods that do a good job accounting for confounds or alternative explanations (save one I mention below).

Weaknesses:

I have mostly minor comments and one major one.

Major:

The principal finding is that tasks that evoke brain activity patterns that resemble the MDN also had more stable "deliberate task focus" features. While all the analysis and controls are impressive, I'm still left with the sense that this is reifying something we already know or that alternative explanations are more parsimonious than the MDN induces stability in "thought".

I thought an example might be easiest to understand my point: If I gave participants a series of working-memory-related tasks. Some of these are the crème de la crème, and others are sloppy and poorly designed. Then suppose I assess them on measures related to deliberate thought; I'd likely find the "good" tasks elicit more consistent/reliable deliberate task focus. I also would bet money that these same tasks would evoke canonical WM and MDN activity patterns more than the sloppy tasks. This isn't evidence of MDN stabilizing patterns, but rather that both stable thought patterns and activity in the MDN share a common cause. Thus, I would predict that with my thought experiment, your analysis would find the same result. So it strikes me as a strong alternative possibility for these results is that tasks that reliably evoke deliberate task focus are also those that more strongly and consistently evoke working-memory demand (i.e., Figure 2 shows that they are primarily driven by the executive/WM tasks).

Minor:

(1) The PCA was reviewed previously, and I don't want to relitigate a prior method, but I had one minor concern. It would be useful to know how the principal results are based merely on the "deliberate" or "focus" items specifically. Is the thought space necessary, or do the individual items that likely drive the "deliberate task focus" PC essentially replicate the main result?

(2) I struggled with the motivation for projecting the task data onto a resting state FC analysis that focuses on "gradients". I understand that with 14 tasks activation maps, data reduction is a good thing. But as someone who isn't as enmeshed in this work, I didn't follow why this specific "atlas" was chosen over any other (parcellations, meta-analytic maps of canonical networks, etc.). Maybe a brief sentence saying why this and not that would help readers who find themselves in my shoes.

Reviewer #2 (Public review):

Summary:

The study's aim was to establish whether stability in thought patterns relates to the particular thought pattern, the task context, or their interaction. And further, whether stable thought patterns could be linked with distinct brain patterns.

Strengths:

The core reliability framing is novel, and the trait/state/interaction decomposition is a fruitful way to pose the question, leading to the finding that stability is neither a pure trait nor a pure task property, but emerges from their interaction, which is a solid contribution.

The aim to characterise aspects of stability across individuals and across tasks was achieved and is well supported.

Weaknesses:

While the paper makes excellent use of existing data sets, the independent samples, i.e., one study sample for the cognitive/thought-sampling data, and several different study samples to generate the brain maps, do limit the brain-behaviour conclusions that can be drawn.