The functional neuroanatomy of event-based and time-based prospective memory, and its improvement Authors: Jessica Peter, Marta Menéndez-Granda, Nadine Schmidt, Nathanael Jost, Katharina Klink, Yosuke Morishima, Sina Straub, Sebastian Horn, Matthias Kliegel, Michael Orth Published: 2026 ABSTRACT Prospective memory anchors upcoming activities in future time and space either cued externally (event-based prospective memory) or internally (time-based prospective memory). It is key to remain autonomous and can be enhanced using incentives. To examine the neuronal underpinnings of prospective memory and any performance improvement, we randomly assigned 58 healthy older participants (60-75 years old, 36 women) to time-based and focal event-based prospective memory tasks using 7T-fMRI, with or without incentives. Event-based and time-based prospective memory differ in their functional neuroanatomy. The focal event-based task required activity important for the perception of relevant changes in the environment and for adapting the focus of attention to then respond swiftly. The time-based task, in contrast, involved regions supporting cognitive control or serving as a multimodal information buffer. Incentives had no effect in time-based prospective memory but recruited brain regions important for attention thus improving cue recognition in event-based prospective memory. This highlights the differences between these two prospective memory types and the need for differential approaches depending on the type of memory demand. Introduction Future actions can be planned in the present. This requires prospective memory that is the ability to anchor upcoming activities in future time and space while being engaged in other activities1. For event-based prospective memory an external cue triggers the retrieval of the what and how of the planned activity, which then needs to be executed. Time-based prospective memory relies on the internal perception of the passing of time until it is time to act including the what and how, similar to event-based prospective memory1. Prospective memory is an important cognitive ability, an impairment of which can disrupt routine responsibilities such as managing appointments, taking medications, or household chores2–4. It is possible to improve prospective memory performance using an off-loading approach with reminders or calendars as external aides for memory5. In addition incentives, in particular loss-related incentives, can also improve prospective memory performance in laboratory tasks6. A key question concerns the neuronal underpinnings of prospective memory and those underlying any performance improvements. A better understanding of its functional neuroanatomy in healthy people is important as it can contribute to the development of additional, more focused treatment strategies in the clinical context. This can be particularly valuable for elderly people in whom an impairment of prospective memory can result in losing their autonomy with subsequent dependence on others2–4. Several brain regions seem to be involved when remembering what to do following an event-based cue while performing an ongoing task. If the ongoing activity is sufficiently similar to facilitate cue detection no additional voluntary monitoring for the prospective memory cue may be necessary7–9. This may involve the ventral frontoparietal network which is activated when behaviourally relevant stimuli occur rather unexpectedly. Active strategic monitoring may be more important for cues that are distinct from the ongoing activity10,11. This involves the dorsal frontoparietal network12 which mediates top-down voluntary allocation of attention. Less is known about brain networks involved in time-based prospective memory13–15. The decisions to check the clock and execute the time-based intention are self-initiated (i.e. volitional) and do not involve reacting to an external cue, as in event-based tasks. Free choice decision-making processes involve activity in a network prior to action selection that includes the prefrontal cortex and the supplementary as well as pre-supplementary motor area16. The internal perception of the passing of time can be supported with checking a clock. Clock-checking itself may be another free choice action, a view supported by activations in the prefrontal cortex and the pre-supplementary motor area in a time-based task15. Very few studies have examined focal event-based prospective memory in healthy older adults using fMRI17,18 and only one has assessed time-based prospective memory19. This study reported that similar brain networks were active during both tasks (i.e. frontal and parietal brain regions, the insula and the thalamus). However the time-based task in that study was very similar to an event-based task since a clock was always in plain view, so the participants did not need to remember the time-based intention by themselves. Instead they were reminded by an external cue (i.e. the clock) which was constantly visible. Therefore it could be that brain activity was similar in both tasks just because the task setup was very similar. Given the many different brain regions reported to be involved there may not be a single region or network responsible for prospective memory, either event-based or time-based. It is therefore important to develop an understanding of the functional neuroanatomy of the different types of cognitive efforts involved in prospective memory. This multitasking challenge entails keeping task instructions in mind integrating external (cues) or internal (time perception) information, attention, and the ability to switch between different tasks. Prospective memory deficits could arise as a consequence of dysfunction in any of the involved systems. Therefore a better understanding in healthy individuals of how these systems contribute to the construction of prospective memory could be valuable in clinical contexts to help infer the underlying dysfunctional mechanism. We therefore aimed to disentangle the functional neuroanatomy of event-based prospective memory (using a focal task), monitoring for cue detection, time-based prospective memory and clock-checking while performing an ongoing task using ultra-high field neuroimaging in healthy older adults. We hypothesised that this involves prefrontal regions involved in cognitive control, attention, and time perception, but also regions implicated as important hubs for regulating the allocation of cognitive resources necessary given the task at hand, for instance, parietal regions or the cingulate cortex20. These regions may not serve a single function but may synergistically support the integration of attentional temporal and contextual information necessary for prospective memory to operate. In addition we tested whether event-based or time-based prospective memory performance could be modulated with incentives, and, if so, if this corresponded to a functional neuroanatomy signature. Such a dynamic understanding is important for rethinking for instance non-pharmacological treatment approaches to improve prospective memory in its multitasking context. We show that event-based and time-based prospective memory differ in their functional neuroanatomy. The focal event-based task involved the perception of relevant changes in the environment adapting the focus of attention between ongoing task activity and cue perception to then respond swiftly in line with the action that was associated with a given stimulus. The time-based task in contrast, involved regions supporting cognitive control or serving as a multimodal information buffer. Incentives had no effect in time-based prospective memory but recruited brain regions important for attention thus improving cue recognition in event-based prospective memory. Results Behavioural findings Participants in the incentive group (M = 97.59, SD = 5.77) had more prospective memory hits (i.e. their task accuracy was higher) in the focal event-based task than participants in the no-incentive group (M = 86.55, SD = 23.34) (t(56) = 2.47, p = 0.017, d = 0.65; Table 1 and Fig. 1). We found no significant improvement in time-based task accuracy (incentive group, M = 59.66, SD = 34.79; no-incentive group, M = 56.21, SD = 33.96) (t(56) = 0.38, p = 0.704, d = 0.10, BF10 = 0.282; Table 1 and Fig. 1). With incentives (M = 1.45, SD = 0.41) participants responded slightly, but not significantly, faster in the focal event-based task than without incentives (M = 1.61, SD = 0.35) (t(55) = 1.56, p = 0.122, d = 0.42, BF10 = 0.741; Table 1). With incentives (M = 23.72, SD = 13.99), they did not check the clock significantly more often in the time-based task than without incentives (M = 24.14, SD = 9.79) (t(56) = 0.131, p = 0.897, BF10 = 0.267, d = 0.034, BF10 = 0.267; Table 1). Since the hit rate for the event-based prospective memory task was close to the ceiling neither the normality assumption nor the assumption of variance homogeneity may hold. We therefore repeated the statistical analysis assuming non-parametric distribution of the data (Mann–Whitney U test) or variance inhomogeneity (Welch’s test). The results were very similar (U = 270.50, p = 0.006) or (t = 2.471, p = 0.019).Fig. 1Task accuracy (in %) in an event-based or a time-based prospective memory task.Healthy older adults (N = 58) were able to receive an incentive (red) or not (blue) during an event-based (A) or a time-based (B) prospective memory task. Box plots show individual data points (circles), the median (solid line), 25th–75th percentiles (box), and min to max values (whiskers).Table 1Sociodemographic information and task performance (mean ± standard deviation) in a sample of healthy older adults, assigned to an incentive or a no-incentive group. We used t-tests or Χ2-tests for statistical analysisNo-incentive (n = 29)Incentive (n = 29)p-valueAge in years67.76 ± 3.9466.72 ± 3.560.300Sex (f/m)18/1117/120.999Education in years16.62 ± 4.4817.35 ± 3.910.391Retired (number yes/no)23/624/50.999Monthly income in CHF4758.62 ± 2246.515431.03 ± 1989.810.233CogTel total score40.16 ± 7.0542.37 ± 6.350.215Accuracy event-based task in %86.55 ± 23.3497.59 ± 5.770.017Accuracy time-based task in %56.21 ± 33.9659.66 ± 34.790.704Response time event-based task in s1.61 ± 0.351.45 ± 0.410.129Number of clock-checks24.14 ± 9.7923.72 ± 13.990.897n number, f/m number female/number male, CHF Swiss Francs, CogTel the cognitive telephone screening, s seconds. Task accuracy (in %) in an event-based or a time-based prospective memory task. Healthy older adults (N = 58) were able to receive an incentive (red) or not (blue) during an event-based (A) or a time-based (B) prospective memory task. Box plots show individual data points (circles), the median (solid line), 25th–75th percentiles (box), and min to max values (whiskers). Table 1Sociodemographic information and task performance (mean ± standard deviation) in a sample of healthy older adults, assigned to an incentive or a no-incentive group. We used t-tests or Χ2-tests for statistical analysisNo-incentive (n = 29)Incentive (n = 29)p-valueAge in years67.76 ± 3.9466.72 ± 3.560.300Sex (f/m)18/1117/120.999Education in years16.62 ± 4.4817.35 ± 3.910.391Retired (number yes/no)23/624/50.999Monthly income in CHF4758.62 ± 2246.515431.03 ± 1989.810.233CogTel total score40.16 ± 7.0542.37 ± 6.350.215Accuracy event-based task in %86.55 ± 23.3497.59 ± 5.770.017Accuracy time-based task in %56.21 ± 33.9659.66 ± 34.790.704Response time event-based task in s1.61 ± 0.351.45 ± 0.410.129Number of clock-checks24.14 ± 9.7923.72 ± 13.990.897n number, f/m number female/number male, CHF Swiss Francs, CogTel the cognitive telephone screening, s seconds. Sociodemographic information and task performance (mean ± standard deviation) in a sample of healthy older adults, assigned to an incentive or a no-incentive group. We used t-tests or Χ2-tests for statistical analysis n number, f/m number female/number male, CHF Swiss Francs, CogTel the cognitive telephone screening, s seconds. Brain activity when preparing for action in a working memory task In real life one needs to keep a prospective intention in mind while being engaged in day-to-day activities. In a laboratory, prospective memory tasks intend to mirror real-life engagements with an ongoing task as the background activity to the actual prospective memory task. In the working memory-related 1-back task brain activity is prompted by a specific working memory cue (Fig. 2). We found that this was associated with higher activity in the right insula the left precuneus, the left posterior cingulate gyrus, and in areas related to visual processing (Table S3, Figs. 3 and 4 upper panel).Fig. 2Event-based and time-based tasks in our study.Participants needed to keep in mind event-based or time-based prospective intentions while working on a 1-back working memory task (ongoing task). For the event-based task they needed to press a button when an animal appeared (left panel). For the time-based task they needed to press a button when a minute had passed (right panel). To find out how much time really had passed they were able to check a clock by pressing another button.Fig. 3Brain activity in a 1-back working memory or an event-based prospective memory task.The upper panel shows brain activity associated exclusively with working memory. The lower panel shows brain activity that was higher with event-based prospective memory button presses than with working memory. The statistical threshold was set at p < 0.05 FWE-corrected at a cluster level, with the cluster size estimated at p < 0.005, uncorrected. Reaction times (rt) were used as parametric modulators in each model.Fig. 4Brain activity in a 1-back working memory or a time-based prospective memory task.The upper panel shows brain activity associated exclusively with working memory. The lower panels show brain activity that was higher with preparing clock-checks or time-based prospective memory (PM) button presses than with working memory. The statistical threshold was set at p < 0.05 uncorrected at a cluster level, with the cluster size estimated at p < 0.005, uncorrected. Event-based and time-based tasks in our study. Participants needed to keep in mind event-based or time-based prospective intentions while working on a 1-back working memory task (ongoing task). For the event-based task they needed to press a button when an animal appeared (left panel). For the time-based task they needed to press a button when a minute had passed (right panel). To find out how much time really had passed they were able to check a clock by pressing another button. Brain activity in a 1-back working memory or an event-based prospective memory task. The upper panel shows brain activity associated exclusively with working memory. The lower panel shows brain activity that was higher with event-based prospective memory button presses than with working memory. The statistical threshold was set at p < 0.05 FWE-corrected at a cluster level, with the cluster size estimated at p < 0.005, uncorrected. Reaction times (rt) were used as parametric modulators in each model. Brain activity in a 1-back working memory or a time-based prospective memory task. The upper panel shows brain activity associated exclusively with working memory. The lower panels show brain activity that was higher with preparing clock-checks or time-based prospective memory (PM) button presses than with working memory. The statistical threshold was set at p < 0.05 uncorrected at a cluster level, with the cluster size estimated at p < 0.005, uncorrected. Brain activity when preparing for action in event-based or time-based prospective memory tasks Comparable to the working memory task, in event-based prospective memory tasks, participants also need to respond to a cue (i.e. an animal). For both tasks this means they must be ready to perceive an image, compute its meaning in the context of the tasks and prepare for action (Fig. 1). There was no additional brain activity when contrasting ‘1-back hit’ working memory-related brain activity with that related to ‘event-based prospective memory’ (contrast 1, Table S4). However there was more activity with event-based prospective memory in the left medial superior and middle frontal gyrus, left precuneus, right thalamus and right postcentral gyrus than with working memory (contrasting brain activity related to ‘event-based prospective memory’ more than ‘1-back hits’, contrast 2, Table S4). This indicates that no additional brain activity was needed when doing a working memory task and at the same time, being ready for an event-based prospective memory task coming up. Doing the prospective memory task however involved brain regions not active in the working memory task (Fig. 3 lower panel). In the time-based prospective memory task participants engaged in the same background 1-back working memory task as with the event-based prospective memory task. In contrast to the event-based prospective memory task participants had to monitor time internally and press a button every minute. Time monitoring could be complemented at will by checking a clock for how much time has elapsed (Fig. 1). The respective brain activity for action preparation of clock-checking, prospective memory task execution, and the ongoing 1-back task can be contrasted to each other to tease out any brain activity that was specific to the respective behaviour. When contrasting ‘1-back hit’ working memory-related brain activity with that of ‘clock-checking and time-based prospective memory’ (contrast 3, Table S5), we found increased brain activity in the right middle frontal gyrus and the right superior parietal lobe. With clock-checking there was more activity in the left superior frontal gyrus and the left occipital fusiform gyrus (Table S5 and Fig. 4 middle panel) than with working memory performance (contrasting the time immediately before they decided to ‘check the clock’ with ‘1-back hits’, contrast 4). Preparing and executing the time-based prospective memory task involved more activity in the left occipital gyrus the left superior frontal gyrus and the right superior temporal gyrus (Table S5 and Fig. 4 lower panel) than with working memory performance (contrasting brain activity immediately before the decision to press the ‘time-based button’ with that of ‘1-back hits’, contrast 5). When contrasting brain activity immediately before ‘clock-checking’ with that immediately before pressing the ‘time-based button’ (contrast 6, Table S5) clock-checking behaviour involved more activity in the left cingulate gyrus, the left precentral gyrus and the right inferior occipital gyrus than preparing and executing time-based prospective memory responses (Fig. 4 middle panel). Brain activity with or without an incentive With an incentive regions that were active when participants were responding to event-based prospective memory cues without incentives (in comparison to ‘1-back hits’, see contrast 2 above), were now active already when participants were in the prospective memory monitoring phase during working memory activity (i.e. when contrasting ‘1-back hits’ to ‘event-based prospective memory’, see contrast 1 above) (Table S6). In line with this observation with an incentive there was no longer any additional brain activity with the event-based prospective memory task compared to the working memory task (Table S6 and Fig. 5 upper panel). This suggests that participants were prepared for the event-based prospective memory task even when processing the working memory cue. Consequently they became significantly more accurate in their responses to event-based prospective memory cues (p = 0.017; Fig. 1).Fig. 5Brain activity in an event-based prospective memory task or a time-based prospective memory task in which participants either received an incentive or not.Green indicates that more activity was found during the monitoring phase of an event-based or time-base prospective memory (PM) task while orange indicates that more activity was found after participants perceived a prospective memory cue. The statistical threshold was set at p < 0.05 uncorrected at a cluster level, with the cluster size estimated at p < 0.005, uncorrected. Brain activity in an event-based prospective memory task or a time-based prospective memory task in which participants either received an incentive or not. Green indicates that more activity was found during the monitoring phase of an event-based or time-base prospective memory (PM) task while orange indicates that more activity was found after participants perceived a prospective memory cue. The statistical threshold was set at p < 0.05 uncorrected at a cluster level, with the cluster size estimated at p < 0.005, uncorrected. In the time-based prospective memory task regions that were active during clock-checking and time-based prospective memory performance without an incentive were with an incentive also already active during the monitoring phase, during working memory behaviour (see contrast 3 above, Table S7 and Fig. 5 lower panel). The same region but on the other hemisphere (i.e. the right superior frontal gyrus) was also more active with incentives when participants decided to check the clock (Table S7 and Fig. 5 lower panel). However this did not relate to a significant change in task performance. With an incentive participants did not check the clock more often, nor did they respond more accurately in the time-based task than without an incentive. Brain-behaviour correlations We found a significant negative correlation between activity in the left middle temporal gyrus and event-based prospective memory accuracy in participants receiving no incentive (r = −0.41, p < 0.039), while the correlation was positive (though not significant) in participants receiving an incentive (r = 0.37, p < 0.055) (Fig. 6). For time-based prospective memory accuracy activity in the right superior temporal gyrus was significantly negatively correlated with task accuracy only in participants receiving no incentive (r = −0.26, p < 0.019; Fig. 7A). In contrast higher activity in the left superior frontal gyrus was significantly negatively correlated with the number of clock-checks in participants receiving an incentive (r = −0.57, p < 0.007) (Fig. 7B).Fig. 6Correlation between activity in the left middle temporal gyrus and event-based prospective memory task accuracy (in %).Participants either received an incentive (red) or not (blue). Significant at p < 0.05* or at trend 0.10†.Fig. 7Correlation between task performance and brain activity.A Correlation between time-based prospective memory accuracy (in %) and activity in the right superior temporal gyrus. (B) Correlation between number of clock-checks and activity in the left superior frontal gyrus. Participants either received an incentive (red) or not (blue). Significant at p < 0.05* or p < 0.01**. Correlation between activity in the left middle temporal gyrus and event-based prospective memory task accuracy (in %). Participants either received an incentive (red) or not (blue). Significant at p < 0.05* or at trend 0.10†. Correlation between task performance and brain activity. A Correlation between time-based prospective memory accuracy (in %) and activity in the right superior temporal gyrus. (B) Correlation between number of clock-checks and activity in the left superior frontal gyrus. Participants either received an incentive (red) or not (blue). Significant at p < 0.05* or p < 0.01**. Discussion Conceptionally current models for prospective memory processes mainly distinguish event-based and time-based prospective memory1. Both types of prospective memory are embedded in the many other ongoing cognitive demands of day-to-day life. In event-based prospective memory people need to intentionally encode that a particular future event should be followed by a specific behaviour. This suggests differential cognitive effort is required for remembering task instructions monitoring for event-based cues, continuous engagement with ongoing activities and switching attention from the ongoing activity to the prospective memory-related task, and back. According to the multi-process theory7 and the dynamic multi-process framework21 remembering prospective memory task instructions is complemented with effortful monitoring for relevant cues. Such monitoring may be dynamic and thus be switched on in a specific context in real life rather than be continuous as in a laboratory task. When a prospective memory task is focal which means the processing of ongoing activities facilitates the processing of prospective memory cues, retrieval may rely on automatic processes that require no additional voluntary monitoring, a view supported by our data. Participants activated regions associated with working memory while engaging in the working memory task but no additional regions that might have been engaged in monitoring for an event-based prospective memory cue. We did not find any activity in the dorsal frontoparietal network the ventral frontoparietal network or the rostral prefrontal cortex, which have been associated with monitoring, either strategic or not10–12,22,23. This suggests that participants fully focused on the ongoing working memory task and did not at the same time, expend cognitive effort preparing themselves for the event-based prospective memory cue. This does not exclude the possibility that participants monitored outside of their engagement with the working memory task. During working memory task engagement however, monitoring efforts likely dropped substantially or ceased completely. In the focal prospective memory task we used, the working memory task and prospective memory behaviour both resulted in the activation of brain regions involved in visual perception, decision making and motor planning for the button press. However with the response to the event-based prospective memory cue, there was more activity in the left dorsal posterior cingulate cortex, the left middle and superior frontal gyrus, the right thalamus and the right primary motor cortex than during working memory behaviour. In keeping with their presumed role for stimulus-driven attention24 this likely reflects the perception of behaviourally relevant changes in the environment, followed by adapting the focus of attention to rapidly change cognitive state and then respond swiftly to those changes25. Working memory task performance involves a narrow focus with low posterior cingulate cortex activity. Switching the focus from the working memory to the prospective memory task activated the posterior cingulate cortex. This is in keeping with its proposed role as a regulator of whole brain metastability25. Posterior cingulate cortex activity would then be low when focused on a given task in our case working memory and high in the transition from the synchronised network activity required in that task to the synchronized network activity needed to engage in another task, in our case event-based prospective memory. The activated regions in the superior frontal gyrus were close to the frontal eye field (Brodmann area 8) and its associated regions (see Table S4). Frontal eye fields are mainly involved in the voluntary control of eye movements in particular saccades26. There is however emerging evidence in non-human primates to suggest a role in attentional and motor-related responses, but also in the integration of sensory information for decision making27. Activity in the premotor cortex relates to motor sequencing and movement planning. Activity in the thalamus with its nuclei is usually related to sensory integration motor planning and attention28. The peak voxels in our study suggest an involvement of medial nuclei as well as mediodorsal nuclei (e.g. medial pulvinar nucleus; see Table S4). These nuclei have reciprocal connections with the prefrontal cortex, the cingulate cortex, the hippocampus and the insula (among others) and have been related to executive functions, memory, multisensory integration and selective attention. In summary, the activity we observed following event-based cues was thus related to movement planning, decision making, and responding to external cues. None of this suggests any involvement of strategic monitoring processes. The additional activity we observed with the event-based prospective memory task may thus reflect the effort necessary when changing the brain state assumed in one task to that assumed in another with a transition phase in between. The cognitive processes underlying time-based prospective memory are less clear. Remembering task instructions, the engagement in ongoing activities, and the need to switch from the ongoing to the prospective memory task are the same as for event-based prospective memory. If based on a clock with an alarm a time-based task may become an event-based task with the alarm being the cue. In the absence of such a cue time-based prospective memory requires some internal perception of the passing of time to know when it is time to act. To examine the functional neuroanatomy of those fundamental differences between event-based and time-based prospective memory we designed experiments using a time-based prospective memory task used in many behavioural studies. The conceptual difference between event-based and time-based prospective memory is supported by our data. Working memory performance involved no additional brain regions when contrasted to event-based prospective memory, indicating that no monitoring for a cue was needed. However when contrasting working memory-related brain activity with time-based prospective memory, we did observe higher neuronal activity in the right dorsolateral prefrontal cortex (Brodmann area 9), the right anterior prefrontal cortex (Brodmann area 10) and the right angular gyrus (Brodmann area 39) (see Table S5). Since the contrast included a mask for all working memory-related activities that additional neuronal activity is not linked to working memory processes but has to do with the computational demands involved with multitasking between working memory, time-based prospective memory, time perception, and the optional clock-checking. This involves a considerable amount of cognitive control and attentional resources likely involving the dorsolateral prefrontal cortex29. In addition it requires a multimodal buffer for the information that is received from the outside, in our case, working memory cue presentation, and the task instructions that are retained internally for the time of the experiment, a role that has been proposed for the angular gyrus30. These regions were not more active when preparing the time-based prospective memory button press. This suggests that the preparatory efforts that go into the prospective memory behaviour, for instance remembering task instructions and judging the timing, have taken place before participants prepared the motor part involved in button presses. In our study the internal tracking of time was complemented by the option to check how much time had passed (i.e. check the clock). Clock-checking behaviour does not depend on any cue and is internally driven as a decision of choice. Clock-checking may thus be related to free choice paradigms as participants can freely decide whether and when to initiate an action (i.e. press a button to make a clock appear)16,31. The ‘when’ component in free-will paradigms has been associated with higher activity in the motor cortex, the middle frontal gyrus and the inferior parietal lobe, while the ‘whether’ component was associated with higher activity in the cingulum, the inferior orbital frontal gyrus and the insula32. Since we masked other actions (e.g. responses in the ongoing task that also require finger movement), increased activity in the superior frontal gyrus seems to be associated with ‘when’ to check the clock rather than reflecting movement preparation. The expectation of receiving visual information from the clock may have engaged relevant visual brain regions such as the fusiform gyrus that is known to be relevant for processing of higher order visual information and to contribute to the integration of multiple stimuli. Specifically receiving information from the clock will update computations for time estimation processes that entails the decision to act, i.e. proceed to the time-based prospective memory task, or not to act because it is too early and return full attention to the ongoing task until the next clock check. Time tracking involves a frontoparietal network including the prefrontal cortex, the superior parietal lobe, and the angular gyrus33, regions that were active in our time-based task already during working memory behaviour (see Table S5). The nature of the time-based task likely changed the closer a clock-check is to the end of the one-minute interval. Hence time-based responses may become event-based responses elicited by a cue (a close-enough to 1-min clock check). Immediately before deciding to execute time-based button presses increased activity was observed in the right superior occipital gyrus, the right superior temporal gyrus (planum temporale) as well as the left superior frontal gyrus (Brodmann area 6, motor cortex; see Table S5). The right superior temporal gyrus (and particularly the planum temporale) may contribute to rule learning or pattern learning34 as well as the decision to direct attention to a particular stimulus and non-verbal auditory processing including of scanner noise35. The latter may explain a negative association of right superior temporal gyrus activity with time-based task accuracy perhaps reflecting the level of distraction by scanner noise. The main neuronal activity when preparing the time-based prospective memory action involved motor regions. This suggests that the preparatory efforts that go into the prospective memory behaviour, for instance remembering task instructions and judging the timing, have taken place beforehand. Clock-checking and time-based prospective memory button presses differed at the neuronal level as there was significantly more activity in the left ventral posterior cingulate gyrus (Brodmann area 23), the precentral gyrus, and the right inferior occipital gyrus (Brodmann area 18) with clock-checking than with time-based button presses (see Table S6). Activations in the cingulate gyrus have been previously associated with remembering what needs to be done, while the precentral gyrus has been associated with the preparation of a motor response31. We (and others) have previously shown that it is possible to improve prospective memory performance in healthy older adults using loss-related incentives or practice (i.e. task repetition)6,36,37. In the current study we went on to examine the functional neuroanatomy underlying any performance improvements due to reward-mediated reinforcement of learning. We reproduced our previous findings that incentives enhanced event-based but not time-based prospective memory performance36,37. According to the motivational-cognitive model of prospective memory38,39 there are two different pathways to enhance performance with an incentive in prospective memory tasks. First, incentives can influence how well participants monitor for cues. Second, incentives may improve encoding and subsequent maintenance of intentions (e.g. deeper encoding or facilitated access to intentions). Regions that were involved in the execution phase of event-based button presses without incentives (i.e. dorsal posterior cingulate gyrus, middle temporal gyrus and thalamus) were already active in the monitoring phase with incentives. Thus our data indicate that incentives enhance monitoring efforts, by strategic monitoring, goal prioritisation, or by increased response readiness. Activity in the left middle temporal gyrus may reflect semantic processing that is required to categorise stimuli and to match the semantic properties of a stimulus with the stored prospective intention (the action associated with animal vs. non-animal)40. This is supported by a case study of semantic dementia (with atrophy in the left temporal lobe) with impaired event-based prospective memory in which the semantic relationship between a cue and an action had to be evaluated41. The middle temporal gyrus’ role may be relevant in any variation of an event-based prospective memory task in which the semantic qualities of a stimulus need to be appreciated to then result in the previously associated action behaviour. In our case, in the absence of an incentive, higher activity in the middle temporal gyrus was associated with poorer task accuracy, suggesting inefficient semantic processing that was not reliably translated into intention execution. In contrast when incentives were present, higher activity was correlated with more accurate responses, consistent with effective semantic categorisation and successful matching of the stimulus to the intention. These findings suggest that the middle temporal gyrus supports prospective memory through semantic cue processing but its functional contribution may depend on the motivational state. With incentives the event-based intention became goal relevant. More activity in the middle temporal gyrus was thus beneficial for cue detection and participants became better at detecting them. Consequently participants may have adjusted their strategy to focus on recognising and responding to an event, with the speed of response becoming a secondary concern as accuracy tends to be prioritised over speed in general42. It is worth noting that incentives did not lead to greater activations of brain regions typically described in incentive paradigms, like the monetary incentive delay task43–45. Instead they modulated regions relevant to prospective memory itself. Although the present study focused on healthy older adults the observed dissociation between event-based and time-based prospective memory aligns well with findings from younger adults and lesion studies. Research in younger adults has consistently shown that focal event-based prospective memory relies more strongly on spontaneous retrieval and ventral attention networks whereas non-focal and time-based prospective memory places greater demands on sustained cognitive control and prefrontal systems12,46. Lesion studies further support this distinction showing disproportionate impairments in time-based prospective memory following prefrontal damage47,48, and in event-based prospective memory following medial temporal or posterior cingulate damage46,49. The present findings suggest that this functional architecture remains detectable in old age but that its modifiability differs depending on the task at hand, with event-based mechanisms being more responsive to incentives than time-based mechanisms. Our findings may inform theories of age-related prospective memory decline by showing that older adults retain the capacity to flexibly upregulate preparatory control when the task structure allows it (event-based prospective memory) whereas time-based prospective memory appears less responsive to motivational modulation. This supports the view that age-related prospective memory deficits are heterogeneous and mechanism-specific rather than reflecting a global loss of resources. Limitations Our study had some limitations. First the use of a time-based task in which participants have to press another button to make a clock appear increased movement. Hence we had to exclude more participants than in the event-based prospective memory experiments. Second the anticipation of the incentive during the prospective memory task was only addressed at the beginning of the task during the instructions and not repeatedly during the task, as this would serve as a reminder for the prospective memory task. Therefore we cannot be sure that the participants were aware of the incentive throughout the task. Third the regressor of ongoing working memory behaviour included correct hits, but it did not differentiate between later remembered or forgotten prospective memory button presses. This was not possible for the event-based task as performance was close to the ceiling. However we tried this for time-based prospective memory, and the results and interpretations were similar. Another limitation may be that we did not include a group of younger adults to examine ageing effects. We also need to acknowledge that our explicit masking approach may limit the ability to identify regions that are modulated continuously across task demands. However our goal was to prioritise specificity for prospective memory-related activity rather than to capture all regions showing modulating effects of task load. Conclusion In summary the results of the present study that includes both event-based and time-based prospective memory in its design, argue that event-based and time-based prospective memory performance require differential cognitive effort. At the functional neuroanatomy level focal event-based prospective memory seems a variation of cued-driven behaviour. Our results emphasise the constant need to perceive behaviourally relevant changes in the environment, in this case, focal event-based prospective memory cues, followed by adapting the focus of attention to rapidly change cognitive state and then respond swiftly to those changes25. We show that this involves the posterior cingulate cortex in its role as a regulator of whole brain metastability so that the synchronized activity in task relevant networks can be alternated between similar tasks. A previous study reported similar brain activity in event-based and time-based prospective memory in older adults19. However in the time-based task used in that study, a clock was visible on the screen throughout the entire experiment. Thus participants were able to use the clock to know when the prospective memory task was to be performed. Hence the time-based task in that study may in fact have been more like an event-based task. We used a time-based prospective memory task that is well-established in behavioural studies in which the computational demands are much higher when multitasking between a working memory, time-based prospective memory, and the optional clock-checking. We show that this involves a considerable amount of cognitive control and attentional resources involving the dorsolateral prefrontal cortex but also the angular gyrus as a multimodal buffer for the information that is received from the outside and the task instructions that are retained internally for the time of the experiment30. One must be cautious when extrapolating from a laboratory task to real life. However these laboratory insights into the functional neuroanatomy of prospective memory may support recommendations for a well-structured daily routine consisting of a series of event-based tasks with many cues provided by alarm clocks, timers, or agendas. To some extent incentives can improve event-based prospective memory performance even further. This cannot be said for time-based prospective memory. Hence one approach could be to transform as many time-based prospective memory chores into event-based chores. Any tools to help keep an eye on time such as timers or alarms set at specific times, could help transform a time-based prospective memory task into an event-based task. This strategy could then also involve incentives and ultimately compensate for any time-based prospective memory behaviour difficulties. Methods Participants We recruited N = 58 healthy older participants (60–75 years of age, 36 women; Table 1). All participants were fluent in German and had normal or corrected-to-normal vision. They reported no history of psychiatric or neurological disorders or injuries no current or lifetime seizures, no alcohol or drug abuse, and no use of medications that interfered with cognition (e.g. antiepileptics, anticholinergics, psychotropics, opioids, sedatives, and benzodiazepines). In addition they did not report past or current head injuries. None of the participants showed signs of cognitive impairment50 or clinically relevant depressive symptoms (geriatric depression score ≤ 5)51. Study procedure We randomly assigned participants to an incentive or a no-incentive group (between-subject factor ‘group’), using a formula in Excel based on sex and age. Both groups were similar in age, sex, education, and other sociodemographic variables (see Table 1). Participants performed a focal event-based and a time-based prospective memory task (within-subject factor ‘task type’) in an MR scanner. The order of the tasks was counterbalanced across participants. The study was performed in accordance with the Declaration of Helsinki and approved by the local ethics committee (project number 2019-01779). All ethical regulations relevant to human research participants were followed. All participants provided written informed consent before study participation. Structural and functional MRI data acquisition We acquired structural and functional MRI data using a 7-Tesla ultra-high-field Magnetom Terra scanner (Siemens Healthineers, Erlangen, Germany) using a 32-channel head coil. T1-weighted images were obtained using a compressed-sensing accelerated MP2RAGE sequence (TR = 6 s, TE = 2.06 ms, TI 1 = 800 ms, TI 2 = 2700 ms, flip angle 1 = 4°, flip angle 2° = 5, voxel size 0.6 × 0.6 × 0.6 mm, acquisition time = 7:40 min). For task-based fMRI we used a fat-suppressed whole-brain 2D gradient-echo EPI sequence with simultaneous multi-slice acceleration (TR = 1 s, TE = 24 ms, voxel size = 1.4 × 1.4 × 1.4 mm, flip angle = 90°, slice acceleration factor = 4, in-plane acceleration = 3, acquisition time = 28:21 min, volumes = 1680). We presented prospective memory tasks with PsychoPy (v2021.1.2). The tasks consisted of four blocks: during the first and the third blocks participants performed a 1-back working memory task as an ongoing task (Fig. 1). During the second and fourth blocks we added a prospective memory task to the ongoing task. One block contained the event-based task and the other block contained the time-based task. The order was counterbalanced across participants. Stimuli appeared for 2–3 s on the screen with an inter-stimulus interval of 1–2 s. No feedback was given. Button presses were assigned to the different types of responses. This assignment did not change across task blocks or participants. Further details can be found in our study protocol52. In the focal event-based task participants were asked to press a button whenever an animal appeared on the screen (Fig. 1). Every answer that occurred within 5 s following the presentation of an animal was considered a prospective memory hit53,54. Task accuracy was defined as the percentage of prospective memory hits out of all possible hits (i.e. 0–100%). In the time-based task participants were asked to press a button whenever one minute had passed (Fig. 1). To monitor the time they were able to press another button to make a clock appear for 3 s in the upper left corner of the screen. The display was blank until participants pressed that button to check the time. Previous studies considered responses within ± 2.5 s of each target time as a prospective memory hit53,54. This would have led to a limited number of events for our neuroimaging analysis or to an exclusion of 5 participants in our time-based sample. Hence we considered responses within 2 standard deviations from the mean deviation time (in our sample 8.32 s) as a prospective memory hit. Task accuracy was defined as the percentage of prospective memory hits out of all possible hits (i.e. 0–100%). Each prospective memory block consisted of 158 trials. Of these 54 trials were 1-back hits (i.e. trials in which an image was identical to the previous one), and 10 prospective memory cues (i.e. one every minute). Each prospective memory block lasted approximately 11 min. Participants in the incentive group started with a credit of 5 Swiss Francs for each prospective memory task. For each missed prospective memory cue they lost 0.50 Swiss Francs (i.e. 0.5 Swiss Francs were deduced from their credit). They were also informed that half of the money they had at the end of the experiment would be donated to an organisation of their choice (Doctors without borders, UNICEF, or WWF) as in our previous studies36,37. We used loss-related incentives rather than gain-related incentives as this had the best motivating effect in our previous studies in older healthy people36,37. Behavioural data analysis We first evaluated whether incentives improved accuracy or reaction times in the event-based task or accuracy or clock-checks in the time-based task using t-tests (two-sided). Then we examined whether incentives improved accuracy or reaction times in the ongoing task using t-tests (two-sided; results reported in the Supplement). Next we tested for brain-behaviour associations. To this end we used Pearson correlations of behavioural measures (event-based or time-based task accuracy, or number of clock-checks) and measures of brain function (second-level beta contrast estimates extracted from each neuroimaging general linear model). fMRI data processing and analysis Pre-processing of fMRI data was done using statistical parametric mapping (SPM12; Wellcome Trust Centre for Neuroimaging, London, UK), implemented in Matlab R2019b (MathWorks, Natick, MA, USA). It included realignment slice-time correction, co-registration to the skull-stripped structural image, normalisation and smoothing with a 6 mm FWHM Gaussian kernel. To remove low-frequency noise from pre-processed data we applied a high-pass filter using SPM12’s default settings. For first-level analyses we used a GLM that models time-series as a sequence of events convolved with the canonical hemodynamic response function. We analysed physiological data using the CompCor approach55 implemented in TAPAS56. We used the Computational Anatomy Toolbox (CAT12.8) for segmentation of T1 images57. For first-level analysis we fitted a GLM for the working memory task alone (i.e. the ongoing only block) and for each prospective memory task. In each of these models, we included six nuisance regressors for movement parameters and eight regressors for physiological data (see Supplementary material for more details on the regressors). The ongoing-only model included two regressors: one for ‘1-back hits’ (duration 0 s) and one for ‘1-back correct rejections’ (duration 0 s). We then created two contrasts for first-level analysis. One examined 1-back hits > baseline and 1-back hits > 1-back correct rejections. The focal event-based prospective memory model included two regressors: one for 1-back trials during the prospective memory block (duration 0 s) and one for prospective memory trials (duration 4 s). We added response times as parametric modulators. We then created two contrasts for first-level analysis. One examined thinking about the prospective memory task while busy working on the 1-back task (i.e. 1-back hit > event-based prospective memory cue). The second contrast examined the detection of an event and the preparation of a response (i.e. button press; Fig. 1). To this end we contrasted brain activity from when participants saw the image of the 1-back task to when they responded with a button press to that from when they saw the event-based cue (i.e. the animal) to when they responded (i.e. event-based prospective memory cue > 1-back hit). We used explicit masks to partial out any brain activity associated with 1-back working memory alone since we were interested in the preparatory actions associated with prospective memory. We know that participants must have been preparing the 1-back task response from when they saw the image to when they pressed the button and in addition, they must have been prepared for an animal (i.e. the event-based prospective memory cue). On the other hand they must have been preparing the event-based response from when they saw the animal image to when they pressed the button. Any differential brain activity can then be attributed to either behaviour. Activity larger in ‘1-back hit’ preparation than ‘event-based prospective memory’ preparation (contrast 1) relates to ‘monitoring the environment for event-based cues’ while activity larger in ‘event-based prospective memory’ preparation than ‘1-back hit’ preparation (contrast 2) is very likely specific to action preparation for event-based prospective memory cues. We applied exclusive masking to ensure that brain activity was associated with the respective action preparation and not with working memory behaviour. The time-based prospective memory model included three regressors: one for 1-back trials during the prospective memory block (duration 0 s) one for clock checks (onset −3 s, duration 3 s), and one for time-based prospective memory trials (onset −3 s, duration 3 s). The onset was set to −3 s since we were interested in the preparatory processes just before voluntary actions58,59. We additionally included two covariates (i.e. regressors of interest) accounting for the timing of clock checks within 1 min and the progression of time, as cognitive processes may differ depending on how close the current time is to the target time. We then defined four contrasts for each participant. We contrasted brain activity when responding with a button press to the images of the 1-back task with that in the 3 s immediately preceding the button press for clock-checking and time-based prospective memory button press. We know that participants must have been preparing the 1-back task response from when they saw an image to when they pressed the button; conversely, we know that participants must have thought about clock-checking (or time-based prospective memory button presses) in the 3 s preceding a button press. Any differential brain activity can then be attributed to either behaviour. Activity larger in ‘1-back hits’ preparation than ‘clock-checking’ and ‘time-based prospective memory’ preparation (contrast 3) relates to monitoring (i.e. thinking about the time-based task while working on a 1-back task) while activity larger in ‘clock-checking’ preparation than ‘1-back hits’ preparation is very likely specific to brain activity when internally preparing clock-checking behaviour (contrast 4). Likewise activity larger in ‘time-based prospective memory button press’ than ‘1-back hits’ is very likely specific to brain activity when preparing time-based prospective memory button presses (contrast 5). To examine whether clock-checking and time-based button presses differ at the neuroanatomical level we created a contrast between ‘clock-checking’ and ‘time-based prospective memory button presses’ (contrast 6). Again we applied exclusive masking to ensure that brain activity was associated with the respective action preparation and not with working memory behaviour. For second-level analyses we first conducted one-sample t-tests on first-level contrasts to examine which brain regions were active during event-based or time-based prospective memory in the no-incentive group. Then we conducted a two-sample t-test to compare the incentive and no-incentive group. The statistical threshold was set at p < 0.05 FWE-corrected at cluster level with cluster size estimated at p < 0.005 uncorrected60. This cluster-level correction considers both the ‘strength’ or ‘height’ of activation peaks and the spatial extent of the activation (i.e. the cluster size)61. In addition we considered results surviving a less stringent statistical threshold of p < 0.05 uncorrected at the cluster level (cluster-size estimated at p = 0.005, uncorrected). Statistics and reproducibility We used R (version 4.2.1) with RStudio (version 2022.02.3) for statistical analyses and GraphPad Prism (version 9.0.0; USA) for visualisation of the results. We considered p < 0.05 statistically significant. For non-significant behavioural findings we calculated Bayes Factors using BIC-approximation62,63. No priors need to be specified with this approach. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Supplementary Information Reporting Summary