The process of cognitive control supports adaptive responses and inhibits automatic ones. It is linked to the working memory not only by function but also by location in the brain—in the prefrontal cortex (for review, see: Miller, 2000). Cognitive control is also enormously involved in decision-making processes to obtain appropriate responses to changing environmental conditions. A model of simple, two-choice decisions that has gained popularity in recent years is the drift-diffusion model (DDM; Ratcliff, 1978; Ratcliff and McKoon, 2008). It describes the decision-making process as an accumulation of evidence about a stimulus from perceptual organs or memory, leading to a reaction (most often a motor one) when the evidence exceeds a certain threshold. The studies investigating neural underpinnings of decision-making focus mainly on the prefrontal areas but increasingly also on the prefrontal-basal ganglia loop (e.g., Bogacz and Gurney, 2007; Sieveritz et al., 2019). Moreover, the interaction between the basal ganglia and the frontal regions has been proven in working memory access control: the basal ganglia detects the appropriate context for a motor response to a stimulus stored in memory (McNab and Klingberg, 2008; Baier et al., 2010; Guo et al., 2018).

Working memory is currently viewed as a multi-component system consisting of three subsystems and a central executive one (Baddeley, 2003). These subsystems include the visuospatial sketchpad, which enables storage of visual information, the phonological loop involved in auditory and verbal information storage, and the episodic buffer, which integrates information from other components keeping a continuous sequence. For many years, researchers attempted to understand the neural correlates of information remembering and retrieving. The hippocampus is a neural structure whose participation in both long-term and working memory has been confirmed in many studies (e.g., Olson et al., 2006; Yonelinas, 2013; Libby et al., 2014). It is also well-known that the human memory is prone to errors (Loftus, 1979; Schacter et al., 1998), a fact that motivates the investigation of memory distortion (i.e., false memories) as a byproduct of the memory system, attempting to reveal their nature and mechanism (for meta-analysis on fMRI studies, see: Kurkela and Dennis, 2016). The prevalent paradigm for studying false memories is the DRM (Deese-Roediger-McDermott) paradigm (Deese, 1959; Roediger and McDermott, 1995). Investigating the neural mechanism of false recognition of verbal or non-verbal stimuli is relevant for both encoding (Kim and Cabeza, 2006) and retrieval phases (Schacter et al., 1998). Regarding the false alarms (i.e., when participants incorrectly claim that a new, similar stimulus has appeared previously), most fMRI studies have been using verbal material (e.g., Kim and Cabeza, 2006; Atkins and Reuter-Lorenz, 2011), while the studies using visual objects are less frequent (Lewandowska et al., 2019; Sikora-Wachowicz et al., 2019, 2021). Neuroimaging studies with pictorial material revealed increased activation in the anterior cingulate cortex (Sikora-Wachowicz et al., 2021) and frontal, parietal, and visual cortices (Slotnick and Schacter, 2004; Garoff-Eaton et al., 2006; Gutchess and Schacter, 2012) related to false recognitions. In our study, we used two tasks requiring global and local information processing, respectively. The previous neuroimaging research revealed hemispheric differences in activation of brain regions during these two processing strategies, namely global information processing in the right hemisphere and local—in the left hemisphere, and are processed by other networks in these hemispheres (Fink et al., 1996; Martinez et al., 1997; Weissman and Woldorff, 2004).

According to Borbély’s two-process model, circadian processes, the endogenous oscillatory pacemaker, and the homeostatic sleep pressure, which increases with time spent awake, regulate overall human performance and behavior during the 24 h cycle (Borbély, 1982; Daan et al., 1984; Borbély et al., 2016). Indeed a large number of studies consistently revealed differences in the cognitive domain according to the time of day (for review, see: Schmidt et al., 2007). Alertness, attention, executive functions, among many others, can be affected by the circadian clock. Also, both short- and long-term memory might be modulated by the time of day (Fabbri et al., 2013; Schmidt et al., 2015).

The present study aims to find neural correlates of encoding and retrieval and the diurnal activity of those correlates in the modified DRM paradigm with abstract, visual objects using a new analysis method—non-linear correlation implemented to the task environment. A typical approach to establish a functional proxy for brain connectivity is to calculate the Pearson linear correlation between the brain’s blood oxygenation level dependent (BOLD) time series and a given stimulus of interest. The alternative used here is motivated by the fact that BOLD fluctuations of a relatively large amplitude capture most of the information (Tagliazucchi et al., 2011, 2012; Liu and Duyn, 2013; Petridou et al., 2013). In consequence, bursts of correlated activity across the brain may be efficiently described by a point process consisting of few discrete events (Cifre et al., 2020). The result was subsequently observed with related methods (Liu et al., 2013; Allan et al., 2015; Karahanoğlu and Van De Ville, 2015), and co-activation patterns driven by the point process were studied also in the task paradigm (Jiang et al., 2014; Chen et al., 2015) and in the clinical context (Amico et al., 2014; Li et al., 2014). Following that work, the present correlation approach focuses on the analysis of the brief instances of large-amplitude signals (so-called events); such amplitude thresholding acts as a simple non-linear filter (similarly, to e.g., the effect of using sigmoid activation functions in artificial neural networks), a feature that by design increases the signal-to-noise ratio significantly. This method and its related implementations mentioned above, has been proven effective in analyzing the brain functional connectivity in resting-state conditions (Tagliazucchi et al., 2011, 2012; Cifre et al., 2021), where it is essential to identify the most significant events from the spontaneously fluctuating signals. The present study is the first to implement a similar approach to characterize the responses evoked by a task.

The current study investigated false memory formation in short-term memory. Employing a novel approach, it revealed features of the neural mechanism behind memory distortions. Although the method has been applied before to resting-state data, according to our knowledge, the present results are the first attempt to employ this technique to task fMRI data. We successfully demonstrate the use of the non-linear correlation method on short-term memory tasks requiring global and local visual processing as well as search for diurnal differences in correlations.

These results indicate some time-of-day effects on the neural mechanism of false memories formation. Previous studies revealed time-of-day differences in neural activity during cognitive tasks (for a review, see: Gaggioni et al., 2014). Marek et al. (2010) confirmed diurnal variations in neural activity of orienting the attentional system during a Stroop-like task. Schmidt et al. (2015), using the n-back paradigm with different cognitive load, demonstrated decreased activity in the ventrolateral prefrontal cortex and premotor areas from the morning to the evening hours for the higher cognitive load. The interaction of time-of-day with BOLD correlations that we observe are in general weaker than other ones (e.g., correlation × probe). In the global processing task, the overall correlations with stimuli predict a higher proportion of “no” responses in the evening and lower in the morning. However, other interactions (with phase and probe), which appeared significant only in the analysis of the local processing task, might confound this effect. Nevertheless, our result with diurnal variation of responses is consistent with the recent work of Tandoc et al. (2021), which revealed an increased generalization process that leads to increased false memory formation in the morning—an effect explained by lower inhibition at morning hours. Another study on resting-state data suggested the less efficient brain networks organization in the first hours after waking, which could be an effect of sleep inertia (Farahani et al., 2021). The stronger effect for local processing, visible in Figure 5, reads that in the evening the increased correlation of the whole brain’s activity with encoding stimulus predicts a lower proportion of “no” responses to a positive probe (more correct responses), and similarly, the increased correlation with retrieval stimulus predicts a lower proportion of “no” responses to a lure probe (more incorrect responses).

Regions that showed significant differences in correlations in both tasks overlapped in the retrieval phase (see: Tables 2, 3), which strengthens the result and demonstrates the effectiveness of the new method of analysis. Regarding the task requiring global information processing, we observed differences in correlations in the orbitofrontal region for the memorizing phase of the task and hippocampal and parahippocampal areas for the retrieval phase. The orbitofrontal cortex is involved in the process of decision making (Steiner and Redish, 2012) and encoding the new visual stimuli (Frey and Petrides, 2000; O’Doherty, 2007). The middle temporal cortex, which includes the hippocampus and parahippocampal gyrus, plays a crucial role in remembering and retrieving events, facts, and details. Moreover, using auditory verbal material, Cabeza et al. (2001) suggested that the parahippocampal gyrus can distinguish between false and true recognition. Garoff-Eaton et al. (2006) confirmed that capacity of the right parahippocampal gyrus with a procedure that uses abstract visual stimuli. The current study supports these results, showing that correlations of the right hippocampus and parahippocampal gyri with the positive and lure probe differentially predict subjects’ responses. Higher correlation in all these areas consistently predicted more “no” responses to positive probe and fewer for lure (both increasing the proportion of incorrect responses, see Figure 3). The opposite trend was seen in supplementary motor area, which might be due the different nature of this region, not involved specifically in memory processing, but in motor planning (Welniarz et al., 2019).

There are more differences in correlations in the task requiring local (detailed) information processing, especially in frontal, cingulate, and temporal cortices. For the encoding phase, one sees the activations of the anterior, middle, and posterior cingulate cortex, which are part of the limbic system responsible for regulating emotion, learning, and memory (Rolls, 2019). Extensive psychological studies on the functional organization of the brain revealed the hemispheric functional separation in a way that the left hemisphere is engaged in language processing and the right is responsible for visuospatial functions (e.g., Gazzaniga et al., 1965; Milner, 1971; Corballis et al., 1999, 2002; Zuanazzi and Cattaneo, 2017). Our task employing abstract objects located in space seems to engage the right hemisphere more, which can be seen in differences in correlations in right temporal cortices.

We have limited the current study by the choice of the AAL atlas, which did not allow the examination of fine-grained structures such as the hippocampus. The choice was made due to partly exploratory character of the present analysis and for the sake of presenting a proof of concept. Nonetheless, there are no technical limitations to apply the present approach to a much finer atlas or even to specific voxels (especially with a hypothesis driven analytic design, where not all brain areas are of interest).

It is worth mentioning that the type of correlation analysis used here is appealing, given the recent results indicating that BOLD infraslow signal fluctuations throughout the brain are coherent with arousal fluctuations (Raut et al., 2021). These results on humans fMRI show that ongoing arousal fluctuations are correlated with global waves of activity slowly propagating in parallel through the neocortex, thalamus, striatum, and cerebellum. We could speculate that the observed differences between time-of-day may be causally related with these high amplitude waves, corresponding with different degrees of arousal and thus of cognitive performance, a point that deserves further consideration.

This study was the first attempt at implementing the new non-linear co-activations method in the task environment. The limitation of the current application of the method is the possible delay of the BOLD response to the stimuli with respect to the timing of the stimuli themselves. If the dominant peak of the response is 1–5 TRs delayed, then one would observe a negative non-linear correlation with an ROI that, in fact, responds positively to a stimulus, unless the anticipatory vasodilation comes into play (Sirotin and Das, 2009) and synchronizes the BOLD activations. The precise interpretation of an ROI’s role following such statements as “higher correlation with an ROI predicts fewer correct responses” is thus contingent on that delay. Incorporating delays into the non-linear correlation analysis is actually possible, as shown in Cifre et al. (2021). In this study, we provide only a proof of concept of that additional feature, a topic that deserves further investigation.

In conclusion, the models generally explained less variance in the global visual-feature processing task data than the local one. The dependence of false memories formation on time-of-day was generally present but weaker than other effects, currently not allowing to pinpoint any particular ROI affected by it. However, diurnal variation of responses could be explained by lower cortical inhibition immediately after sleep, according to the synaptic homeostasis hypothesis. On the other hand, there was enough evidence to find significant differences in processing the positive and the misleading stimuli in specific brain areas. Most notably, we found that peaks in the BOLD signal in the supplementary motor areas immediately after presenting stimuli in the retrieval phase consistently predicted correctness of the following response (“no” for lure and “yes” for positive), and the reverse was true for the hippocampal regions (peaking BOLD at the time of retrieval consistently predicted incorrect responses).

The study demonstrates that non-linear fMRI correlations can be applied effectively to the task paradigm. They were found informative as predictors in generalized linear models, where the interaction terms with atlas-based ROIs indicated specific loci associated with producing responses to the tasks. The method allows finding brain areas related to processing the stimuli and opens new possibilities for analyzing other cognitive tasks. Our results on the influence of the time-of-day on the correlations between the stimulus and neuronal activity and, consequently, on the formation of memory distortions show that the diurnal factor is crucial in various cognitive functions and cannot be ignored when designing research procedures. In addition to scientific research, an important general conclusion from this study is that the human brain functions differently depending on the time of day, which is confirmed in an increasing number of psychological domains.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

The studies involving human participants were reviewed and approved by the Research Ethics Committee at the Institute of Applied Psychology at the Jagiellonian University, Krakow, Poland. The participants provided their written informed consent to participate in this study.

MF, BS-W, KL, and TM conceptualized the experiment and fMRI tasks procedure. AC, JO, IC, and DC conceptualized the method of the analysis. MF, BS-W, and KL collected the behavioral and imaging data. AC performed the preprocessing steps. AC and JO analyzed the fMRI data and prepared figures. DC provided feedback on results’ interpretation. AC, JO, IC, and DC wrote the first draft of the manuscript. All authors reviewed the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.