Measures of resting-state functional connectivity have been of great interest in the field of neuroimaging for predominantly two reasons: Firstly, studies have provided new insight in understanding the human brain including possible overlaps between functional and structural connectivity (Koch et al., 2002; Greicius et al., 2009). Secondly, studying functional connectivity with resting-state data is relatively straight-forward to apply to studies of brain development and in clinical research. This may lead to improved compliance and lower training demands as compared to other approaches like task-based functional Magnetic Resonance Imaging (fMRI) (Fair et al., 2007). Functional connectivity is typically computed using low-frequency (<0.1 Hz) blood-oxygen-level-dependent (BOLD) signal fluctuations, allowing inference of neuronal activity between regions of interests (ROIs) (Biswal et al., 1995). These low-frequency fluctuations are typically obtained during condition of resting, when the brain consequently being in a resting state. This is thought to represent spontaneous neuronal activation as well as unconstrained mental activity, such as “day-dreaming” or “mind-wandering” (Patriat et al., 2013).

Functional connectivity studies of the resting brain have revealed consistent large-scale spatial distributions of coherent signals which we refer to as Resting State Networks (RSNs). It is now confirmed that these networks reflect functional brain systems supporting core perceptual and cognitive processes (Dijk et al., 2010). Interestingly, although RSNs are mostly mentioned in relation to BOLD fluctuations that emerge spontaneously during resting, the same networks also show synchronous fluctuations in task-paradigms and in altered states of consciousness (Dijk et al., 2010). For example, the default network (DMN), which is the most studied RSN, shows strong correlations during task conditions but at an attenuated level relative to rest (Fransson, 2006). Therefore, these networks may be described as either “task-positive” or “task-negative,” in terms of the direction of correlation between mean network activity and the timing of stimulus-events during a task (Hugdahl et al., 2019). Some studies suggest that performance of a cognitive task might lead to “carry-over effects” onto subsequent resting periods, influencing RSN-connectivity. For example, a study by Waites et al. (2005) on functional connectivity analysis of DMN-regions during resting-periods before and after solving an orthographic lexical retrieval task revealed relatively small increases and decreases in functional connectivity between the resting periods before and after task execution. Pyka et al. (2009) presented evidence that a cognitively challenging working-memory task was followed by increased DMN-activity compared to a non-challenging letter-matching task. This might be interpreted as a functional correlate of self-evaluation and reflection during resting from the preceding task, or as relocation of cognitive resources representing recovery from challenging cognitive demands (Pyka et al., 2009). Additionally, Fox and Raichle (2007) showed that coherent spontaneous brain activity could explain trial−to−trial variance in task paradigms. Hence it is not unreasonable to conclude that intrinsic neuronal activity during resting might have an influence on subsequent task-related behavior and corresponding brain activity. Moreover, Sadaghiani et al. (2010) suggested that ongoing activity is organized in a functional architecture. The evoked neural responses are embedded into an underlying functional architecture which cannot be fully understood in isolation from the context established by ongoing activity.

These observations provide a back-drop for the present study, where the focus was to examine and compare RSN-data acquired during a single prolonged resting period with data acquired during repeated brief resting-periods in between task-processing periods (Riemer et al., 2020). Previous studies have rather investigated how well resting-state activity seen in task-processing experiments reflect resting-state activity, and subsequent RSNs seen in classic resting-state DMN experiments (Fair et al., 2007; Ganger et al., 2015). In our study, we used a standard fMRI block-design, and asked the question of how RSN-activity observed during periods of absences of tasks (OFF-blocks) is different from RSN-activity observed during a prolonged resting-period with no interspersed tasks. Hence, we explored similarities and differences in RSN-activity in two resting-state conditions, an intermittent resting-state condition and a continuous resting-state condition. In particular, we focused on similarities and differences in the Default Mode Network (DMN) between the two experimental conditions since it is one of the most well-known and researched RSNs. We hypothesize that although DMN-activity in principle should be similar between the two conditions, differences in mental states shown in these conditions may influence the recorded DMN-activity and connectivity across the two experimental conditions. In order to compare resting-state data between the two conditions, we first extracted BOLD volumes from the OFF-blocks to create the intermittent resting condition dataset. Secondly, we extracted RSNs from both conditions using Independent Component Analysis (ICA) and identified DMN components that were compared statistically. Thirdly, we compared ROI-to-ROI functional connectivity for the DMN with the regions belonging to selected atlas-networks, using permutation statistics and paired t-tests.

The current study examined similarities and differences in brain networks with data collected using two different experimental conditions, an intermittent resting-state condition and a continuous resting-state condition, with a focus on the Default Mode Network (DMN). Intermittent resting-state data were extracted from rest-periods (OFF-blocks) in between active task-processing periods (ON-blocks). Continuous resting-state condition data were from 5 min with task-absence, i.e., a classic resting-state period. We first used spatial ICA to identify the DMN network, and then examined differences in the overall spatial architecture and connectivity between the ROIs of the DMN and ROIs belonging to other main functional networks of the brain. Our results showed that the DMN network was present in both the intermittent and continuous resting-state condition. Although the overall DMN activity-pattern was similar between the two conditions, there were also some differences. Comparing functional connectivity between ROIs of the DMN and ROIs belonging to other functional networks revealed that the DMN in the intermittent resting-state condition was stronger connected with the SN compared with the corresponding analysis for the DMN observed in the continuous resting-state condition. This is to be expected considering that the DMN in the former condition is switched on and off when switching from active to passive processing (Hugdahl et al., 2019; Riemer et al., 2020).

To our knowledge there are very few studies that have attempted to compare DMN activity from two resting-state conditions. The few studies that did attempt to compare between intermittent and continuous resting-state conditions have mainly focused on the methodological aspect of the comparison (Fair et al., 2007; Ganger et al., 2015). Although this approach is valid for assessing how well different methods perform in extracting “continuous-like” resting-state networks from task-fMRI, it does not address the theoretical issue whether the DMN seen in intermittent versus continuous resting-state reflect two sides of the same network, or two different networks. The current study therefore provides a unique opportunity to compare resting-state networks across two conditions in the same subject acquired in a single MRI session. In fact, a recent study by Zhao et al. (2023) compared fMRI-data from a resting-state condition with data from three different fMRI-tasks. Interestingly, they found that functional connectivity during the task-condition correlated better with behavior than functional connectivity during the resting-condition. The authors concluded that task-based fMRI paradigms may better reflect behavior changes than resting-state fMRI, which we support on the basis of the present data. Although the data collected for the current study were not ideal due to differences in duration between the two conditions, it nevertheless allows for a unique comparison on a within-participants basis, which is a strength of the study. Increasing the duration of the blocks in the block-design to have an equal duration of the two conditions would solve the problem of the duration differences but at the same time might introduce additional challenges. The longer the duration of the acquisition periods, the higher the probability that the participants would learn the tasks. This may in turn impact the level of activation of task-positive networks which are supposedly antagonistic to the DMN, such as the EMN Hugdahl et al. (2015). As was suggested by Hugdahl et al. (2015), the EMN is upregulated during active task-processing, therefore it can be suspected that learning may attenuate this effect. Moreover, one could argue that prolongation of the OFF-blocks would improve data quality in comparison with a continuous resting condition. However, long OFF-blocks are generally not recommended. As Henson (2007) writes: “For two alternating conditions, for example, block lengths of more than 50 s would cause the majority of signal (i.e., that at the fundamental frequency of the square-wave alternation) to be removed when using a high- pass cut-off of 0.01 Hz. In fact, the optimal block length in an on-off design, regardless of any high-pass filtering, is approximately 16 s” (Henson, 2007). Thus, although not an ideal design, we nevertheless believe that the current design for comparison between an intermittent and continuous resting situation did produce valid data, inviting to further future research.

The results showed that the overall network connectivity patterns were in line with the results of the few previous studies (Fair et al., 2007; Ganger et al., 2015) despite the different methods used to address the comparison between resting conditions. When comparing ICA components that included DMN activity from both conditions, there was a clear overlap in the overall pattern of activity between the two conditions in both anterior and posterior brain regions. The overlap was present regardless of the method used for the comparison (conjunction or probability maps) and remained the same even after inclusion of ICA-maps characterized by a weaker correlation with the template DMN-map. The highest similarity between the two DMN-maps was observed in the posterior parts of the brain, which also converge with the key DMN-nodes in the precuneus and posterior cingulate cortex, which was expected.

The intermittent resting-state condition resulted in reduced activity in the medial prefrontal cortex (MPFC), and the first intermittent ICA-component (1_I) did not include the classic DMN frontal node (Shulman et al., 1997). The second ICA-component (2_I) showed weak correlation with the template DMN, and weakly overlapped with frontal activation components that were obtained in the continuous resting-state condition. This is also seen in the percentage probability maps in Figure 5. When comparing the two ICA-components with strongest correlation to the DMN template in the two conditions (1_C, 2_C, 1_I, and 2_I), the activity in frontal region was less similar between the intermittent and continuous resting-state conditions. This could mean that the second component (2_I) obtained from the intermittent resting-state condition is driven by different underlying mechanisms than the corresponding continuous resting-state component.

Activity in the MPFC is involved in self-related processing (referencing information to oneself), retrieving personal knowledge such as autobiographical memories and simulate personal future events or social interactions (Spreng et al., 2009; Andrews-Hanna, 2012; Denny et al., 2012; D’Argembeau, 2013; Moran et al., 2013). This may reflect brain correlates of mind-wandering, which is expected to be dominant in a continuous resting-state condition compared to a intermittent resting-state condition. It may be hypothesized that in an intermittent resting-state state, participants will not achieve a deep state of relaxation and mind wandering, comparable to a continuous resting-state period. Activity during ON-blocks may therefore “carry over” and interfere with activity during OFF-blocks (Riemer et al., 2020). As a result, there would be reduced activity in the MPFC. An interesting question is the degree to which the current results, and particularly the MPFC results, can be generalized to other domains beyond the cognitive domain. The three tasks used in the current study all primarily target cognition, especially attention, working memory and executive control. We are not aware of any studies that have investigated whether similar network relationships would appear for tasks tapping affective domains. It could be speculated that affective tasks might activate basal ganglia nodes to a higher degree than cognitive tasks. What we do know, however, is that an auditory task, dichotic listening to speech sounds, results in correspondingly up- and down-regulation of the DMN and task-positive networks across ON- and OFF-blocks as shown by Hugdahl et al. (2019).

The comparison of DMN functional connectivity between the two resting-state conditions is a test of how network dynamics vary between conditions. There was a significant difference for the correlation between the DMN and the SN networks, with stronger connectivity between the DMN and SN in the intermittent resting-state condition compared to the connectivity in the continuous resting-state condition. This finding supports the hypothesis that the SN is modulating both the central executive network (CEN) and the DMN. The salience network (SN) includes activity in the anterior cingulate cortex (ACC) and the ventral anterior insular cortices, as well as the amygdala, hypothalamus, ventral striatum, thalamus, and specific brainstem nodes. SN is activated in response to diverse cognitive tasks and conditions, suggesting a task non-specific function (Seeley, 2019), also consistent with what Hugdahl et al. (2015), labeled the extrinsic mode network (EMN). The hypothesis that the salience network (SN) plays a significant role in switching between the CEN and DMN has been suggested in previous studies (Mesulam, 1998; Craig, 2002; Critchley et al., 2004; Goulden et al., 2014). The ACC and the insula have known reciprocal connectivity with one another and are also connected to motor and sensory areas of the brain. This makes these regions ideal to receive inputs required to initiate switching between resting and task-processing networks, like switching between the DMN and CEN. These regions are also co-activated in different cognitively demanding tasks (Dosenbach et al., 2006). Therefore, the higher functional connectivity for the DMN with nodes of the SN in the current study may reflect that the SN is dynamically modulating brain activity, while constantly transitioning between active and resting periods during the experiment.

Understanding the role of the SN in modulation of the CEN and DMN networks may be critical for understanding underlying neuronal mechanisms of mental disorders. Dysfunctional “carry-over” effects from one network to another may be characteristic of brain connectivity dynamics in, e.g., schizophrenia (Manoliu et al., 2013). Manoliu et al. (2013) demonstrated increased functional connectivity between the DMN and CEN in schizophrenia patients and related this to severity of hallucinations. Additionally, Hugdahl et al. (2020) suggested that abnormal switching between networks in schizophrenia patients may be the reason for the fluctuating pattern seen in auditory verbal hallucinations, with intermittent periods of presence and absence of “voices.”

Despite all research done on the dynamics of brain activity and DMN activity, we still do not know the functional role of this network and what exact role that different nodes of the DMN play for the dynamics of the network as a whole. Although the DMN was present in both conditions in the current study, there were also differences in overall DMN activity and connectivity between the two conditions. This has also been pointed out in previous studies, e.g., Pyka et al. (2009), who used temporal ICA and showed that the DMN was not only more down-regulated during performance of more demanding tasks (working memory 2-back compared to 1-back task), but that it was also more activated after performance of such tasks. This led to the hypothesis that cognitive load influences DMN-activity during task performance, and in particular after performance of challenging or demanding tasks (Pyka et al., 2009). The fact that similar “carry-over” effects were seen in the present study may indicate that the DMN during brief intermittent resting-periods may be more easily disrupted than the “classic” DMN seen in prolonged resting-periods without any intermittent task-processing periods. At the same time, “classic” resting-state designs with 10–20 min of uninterrupted resting while in the scanner, may not reflect a typical situation in daily life since such long resting periods are unusual in modern urban life. Experimental designs that are distant from everyday life dynamics may be a hindrance to progress of understanding the neurobiology of the brain and brain-related mental disorders.

The DMN network observed during OFF-blocks in the intermittent resting-state condition seems to share a similar pattern with the DMN network obtained during a continuous resting-state condition. However, there were also some differences between the conditions, mainly in frontal areas of the network. The functional connectivity analysis between DMN nodes and the whole brain volume revealed further differences between the two conditions. From inspection of Figure 7, it seems that the DMN in the intermittent resting-state condition had significantly stronger connections with parts of the SN, being previously associated with modulating switching between the DMN and CEN. The observed differences may reflect a “carry over”-effect from task-processing ON-blocks to a following OFF-blocks. However, further research is needed to fully understand the extend of the differences between these two conditions, and whether the DMN is essentially similar in an intermittent and a continuous resting-state situation. It could be of interest to examine the dynamics of changes in the DMN network over longer periods of activity and resting. It could also be of interest to investigate if the “carry-over”- effect is task-dependent, by varying task difficulty and cognitive load.

The data analyzed in this study is subject to the following licenses/restrictions: Due to restrictions posed by the ethics approval, data are not allowed shared other than after request to the corresponding author, and after written permission from the ethics committee. Requests to access these datasets should be directed to KH, hugdahl@uib.no.

The studies involving human participants were reviewed and approved by the Regional Committee for Medical Research Ethics in Western Norway (2014/1641/REK Vest). Written informed consent was obtained from all participants. The patients/participants provided their written informed consent to participate in this study.

JB development of the fMRI tasks, data collection, analysis of the data, manuscript writing, and figure preparation. FR assisted in data analysis. KK development of fMRI tasks and data collection. LE and KH conceived and conducted the experiments. AC contributed to analysis of the data and read and commented on the manuscript. RG planned and conceived the study and conducted the experiment. All authors reviewed the manuscript.