Aging-related changes in cognitive flexibility: fMRI meta‐analysis

Abstract

Cognitive flexibility—the ability to adaptively shift between different mental processes—is essential for human functioning. This meta-analysis examines age-related changes in neural correlates of cognitive flexibility using two common assessments: the Wisconsin Card Sorting Test (rule-discovery) and Task-Switching Paradigm (rule-retrieval). We synthesized findings from 85 articles comprising 118 experiments with 2246 participants across young, middle-age, and older adult groups. Activation Likelihood Estimation analyses revealed an age-related decrease in neural involvement, particularly in posterior regions, with an anterior shift in older adults. Younger adults exhibited bilateral activation patterns while older adults showed left-dominant activity, indicating neural circuit redistribution. Rule-retrieval tasks consistently engaged left-lateralized frontoparietal regions across all age groups, with middle-age adults additionally recruiting the right cerebellum and medial frontal gyrus. For rule-discovery tasks, age-related changes were observed in bilateral frontoparietal regions, with older adults showing unique activation in the left inferior frontal gyrus. These findings highlight differential aging trajectories for rule-retrieval versus rule-discovery processes, reflecting changes in neural mechanisms with aging. Furthermore, middle-age adults recruited additional regions related to conflict monitoring, whereas older adults relied more on planning-related areas, suggesting strategy differences. Our study provides critical insights into the neural underpinnings of cognitive flexibility and its age-related changes, emphasizing the need for research on mechanisms and task-specific age trajectories.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12993-025-00301-1.

Introduction

Flexible thinking generally enables us to adjust our cognitive performance to new circumstances, approach phenomena from different standpoints, devise novel solutions, and generate new ideas, thereby constituting a key element of creativity, problem solving and behavioral adaptation [41], [60, 105, 118, 140, 147]. One of its critical constituents is cognitive flexibility—the ability to switch between pieces of information being processed, including switching between attributes of tangible objects and between two or more cognitive tasks, concepts, or strategies [30] . Functional magnetic resonance imaging (fMRI) meta-analyses have extensively explored the neural correlates of cognitive flexibility in young adults, implicating key regions within the frontoparietal and cingulo-opercular networks as well as a range of subcortical structures [15, 23], [59, 111]. However, age-related changes in brain correlates of cognitive flexibility remain poorly understood, particularly when considering the distinct cognitive demands of rule-discovery and rule-retrieval. Recent meta-analytic evidence in young adults highlights divergent neural activation patterns associated with these processes, underscoring the need to examine them separately when investigating lifespan changes [23]. The current meta-analysis addresses this gap by exploring the brain correlates of cognitive flexibility across young, middle, and older adulthood, explicitly distinguishing between rule-discovery and rule-retrieval processes to elucidate age-related patterns in their neural mechanisms.

Cognitive flexibility is often referred to as switching and set shifting. As such, it comprises four cognitive components including (1) salience detection and attention, (2) working memory, (3) inhibition, and (4) mental set reconfiguration, which operate in concert to ensure flexible behavior [26]. Furthermore, external factors, such as task demands, play a crucial role in switching ability since they define a set of cognitive components necessary for any flexible adjustment [50]. More specifically, switching between familiar tasks requires attention directed to their unique attributes as well as efficient suppression of one task when switching to another one. This becomes possible through the mental set reconfiguration. Besides, task switching relies on working memory to hold accessible relevant pieces of information while switching between tasks for operating on them [26]. However, when task uncertainty increases (e.g., playing chess, brainstorming), we have to additionally rely on our prior experience (knowledge representations) that subsumes strategies potentially useful for the identification of flexible and optimal solutions [133].

In psychological research, the two switching types—namely, switching in case of uncertainty and switching in pre-determined conditions—are typically measured via the Wisconsin Card Sorting Test [11] and Task-Cued Switching Paradigm [112], respectively. Both tasks examine cognitive flexibility; however, while the TSP assesses rule switching based on memory retrieval, the same behavior in WCST is estimated via trial and error based on feedback [70]. In other words, in TSP participants are given a cue and asked to retrieve from memory the rule they need to apply (i.e., rule-retrieval), whereas in the WCST participants are asked to discover the rule after negative feedback (i.e., rule-discovery) under conditions of uncertainty.

Neuroimaging studies have examined brain correlates of performance on these tasks in young adults, showing both rule-discovery and rule-retrieval activation in several left-lateralized frontoparietal regions, as well as mostly right-hemispheric cingulo-opercular areas such as insula and cingulate gyrus [15, 23], [59, 111]. Direct comparisons of brain responses to these two tasks have revealed that rule discovery in WCST is associated with extensive bilateral activation spanning into frontoparietal, cingulo-opercular, and subcortical regions (most notably, thalamus), while activation related to rule retrieval is detected only in the left medial frontal gyrus close to the cingulate gyrus [23]. Notably, similar thalamocortical activation was observed for making decisions under uncertainty [63], while activation in the frontopolar cortex (BA10) linked to rule discovery was registered during strategizing and planning [22, 122]. These data collectively show how task demands (i.e., cue-based switching or switching in uncertainty) alter brain responses in young adults. The neural underpinnings of cognitive flexibility in older adults, and especially middle-age individuals, remain inadequately understood. Behavioral studies have demonstrated that performance on cognitive flexibility tasks begins to deteriorate between the age of 30 and 49, highlighting the importance of investigating the brain correlates of cognitive flexibility from middle adulthood onward. This age-related decline also underscores the necessity of examining neural mechanisms across the adult lifespan rather than focusing exclusively on comparisons between young and older populations [38].

The studies that did examine brain correlates of cognitive flexibility in middle-age adults, particularly with regard to rule-discovery switching, produced somewhat mixed results. Several studies highlighted cerebellar activation [44, 51, 52, 67, 71] while others generally demonstrate enhanced activation in the frontoparietal cortex, cingulate gyrus, and subcortical regions—similar to patterns observed in young adults [83, 84, 108]. In addition, middle-age adults show bilateral activation in frontoparietal regions and the right insula during rule-retrieval switching [10, 44, 67]. With regard to subcortical structures, some research implicates the involvement of thalamus [44, 67], caudate nuclei, and putamen [10, 44], while other studies fail to support these findings [51, 52, 94, 119]. Overall, extant results demonstrate that there is considerable variability in the lateralization and topographical specificity of these activations across different brain structures in middle-age population.

Bilateral activation and additional involvement of subcortical regions for rule-retrieval in middle age, as opposed to predominantly left-lateralized cortical activation in young adults, may be attributed to compensatory processes [110]. Indeed, rule-retrieval switching relies heavily on conflict monitoring processes [78] and working memory updating due to continuous cue presentation, revealing age-related changes that begin to emerge in middle age [2, 103]. This may be the reason why successful performance on rule-retrieval task switching is accompanied by recruitment of additional brain areas. However, it is uncertain whether this more extensive activity distribution results from the putative compensatory processes, from age-related reduction of brain response selectivity [77] , or from other factors. As a result, a satisfactory explanatory mechanism for both rule-discovery and rule-retrieval specific activations is still lacking. It remains difficult for the same reason to establish general patterns of brain correlates of cognitive flexibility in middle age, leaving this question open for further research.

In older adults, rule-discovery switching is associated with multiple bilateral frontal activations, especially in frontopolar area [54, 81, 82]. Involvement of subcortical regions [81] as well as insular and parietal cortices has also been shown [7, 54, 82]—notably, similar patterns of bilateral frontal activations, along with subcortical involvement, have been observed in young adults [23]. These discrepancies further complicate the assessment of age-related changes in older individuals. One potential explanation for these findings is offered by the Posterior-Anterior Shift in Aging (PASA) theory, which postulates that, with age, older adults demonstrate a shift from posterior (e.g., occipital, parietal) to anterior (e.g., prefrontal) brain regions during cognitive tasks, potentially serving as a compensatory mechanism for age-related decline in posterior brain functions [28]. The PASA phenomenon could potentially help explain the observed variations in neural activations, particularly the increased engagement of frontal areas during cognitive flexibility tasks.

Finally, even less consistent results are reported in studies of rule-retrieval switching in older adults. Some studies have found left [61, 90, 142], or bilateral [31, 36, 126] frontoparietal activations. In addition, the temporal cortex, sub-cortical regions [31], cingulate gyrus [68], cerebellum, and amygdala [126] also featured in several studies. So, overall, existing findings exhibit considerable variability, and it is also plausible that additional cortical regions may be implicated in rule-retrieval switching in older age.

Although brief, the literature overview above highlights significant inconsistencies among neuroimaging studies which leave crucial questions about age-related changes in the brain mechanisms underlying cognitive flexibility largely unanswered. To fill this gap, the present meta-analysis aims to identify and characterize the hemodynamic brain correlates of general switching ability across different stages of adulthood as documented in the existing fMRI literature. Additionally, we seek to determine age-specific brain correlates of both rule-retrieval and rule-discovery switching processes, as well as to identify commonalities and differences between these processes, particularly in relation to aging. Our general hypothesis predicts that general switching ability would exhibit age-related decreases in activation within posterior brain regions accompanied by significant increases in frontal regions, consistent with the Posterior-Anterior Shift in Aging (PASA) model [28], previously confirmed for other executive functions (see meta-analysis [132]). Given that switching under uncertain and predetermined conditions involves distinct processes (rule-retrieval and rule-discovery), we assume different aging trajectories for each process. Based on the available evidence above, we might expect left-lateralized frontoparietal and cingulo-opercular involvement for rule-retrieval with notable posterior decreases across all ages, and more bilateral involvement of these regions with age-related posterior decreases for rule-discovery switching.

Method

Literature search and article selection criteria

A comprehensive literature search was conducted in PubMed, Web of Science, BrainMap, Neurosynth, and PsycInfo databases to identify relevant articles published prior to October, 2024. The following specific search terms were used: (“cognitive flexibility” OR “set-shifting” OR “task-switching”) AND (fMRI) resulting in 5882 articles. Two additional articles were received manually by scrutinizing the references of one previous meta-analysis [59]. After duplicate removal, 3502 publications remained for the screening procedure based on titles and abstracts. Of these, 2988 studies were excluded because they (a) did not utilize fMRI methodology, (b) did not examine cognitive flexibility, (c) were not original studies (i.e., they were reviews or meta-analyses), or (d) were single-case studies. Figure 1 shows the PRISMA flow chart providing detailed information about the literature screening process.

Fig. 1.

Prisma flowchart for identification and eligibility of articles. n = number of articles. *Three articles included data for two subject groups with different age groups [82, 31, 36]

Full-texts of the remaining 514 articles were further excluded after applying the following criteria:

1. Studies must be published in English;

2. Studies must involve human beings;

3. The reports must include whole-brain analyses;

4. The findings must be obtained from general linear model analyses (e.g., Psychophysiological interaction, Principal Component Analyses were excluded; deactivations, correlations, conjunction, and connectivity analyses were also excluded);

5. Studies must employ the classical blood-oxygenation level-dependent (BOLD) subtraction method (e.g., task switching vs. control task) in WCST (only task-related feedback version) or TSP tasks. Studies were excluded from analysis if they used emotional stimuli or if they included reward-based manipulation task design to avoid biases introduced by such designs. Contrasts comparing different cognitive domains (such as switching versus inhibition) were also excluded;

6. Coordinates must be reported in standardized stereotactic Talairach or Montreal Neurological Institute (MNI) space;

7. Studies must incorporate data from typically developing participants older than 18 years. For studies involving patients, data from healthy control groups were included if they could be isolated.

The final set of 85 articles contained the total of 118 experiments (i.e., some articles reported more than one experiment) that altogether involved 2246 participants. There were 91 experiments with young adults, 12 experiments involving middle-age samples, and 15 experiments with older adults (see Table 1 for sample ages). Datasets prepared for all age groups included data obtained from WCST and TSP separately. Notably, empirical simulation with coordinate data has demonstrated that the sensitivity and power of Activation Likelihood Estimation (ALE) are optimal for meta-analyses with 17 or more experiments [37]. While our middle-age sample and separate WCST/TSP analyses fall below this threshold, we applied recommended robust thresholding criteria [37] and interpret these results as indicative trends that may inform future research and meta-analytic investigations. Given the substantially larger young adult sample compared to middle-age and older groups, we also conducted supplementary analyses to verify that sample size disparities did not influence outcomes, particularly for between-group contrasts and conjunctions. Results from three randomly selected subgroups are presented in the supplementary materials (Table S1). Because methodological factors such as switch probability, preparatory versus stimulus-based activity and cue-to-task assignments have been shown to influence results [13, 29, 56, 89], we examined these factors finding comparable distributions across all age groups. The literature search and article screening were conducted by two authors (ZC and AFi) who thoroughly read each identified article, applied the eligibility criteria, and categorized the articles' characteristics into subcategories for the purpose of data extraction. An additional screening by other the co-authors (AFa and MA) was performed in instances of uncertainties regarding data analysis, contrast, or eligibility for inclusion criteria.

Table 1.

Information on source datasets included in the meta-analysis for young adults

No.	Study	N	Mean age ± SD (range)	F	Task	Contrast
1	Aizawa et al. [1]	30	21.4 ± 1.5	14	WCST	RNF > RPF
2	Armbruster et al. [3]	20	23.5 (20–32)	15	TSP	Switch > Repeat
3	Asari et al. [6]	16	27 ± 5 (20–37)	66	WCST	Dimensional change > no‐change
4	Barber and Carter [8]	13	20–35	4	TSP	Switch > Repeat
5	Braem et al. [12]	35	26 ± 6	7	TSP	Task switch > Task repetition
6	Braver et al. [14]	13	21 (19–26)	3	TSP	Switch > Repeat
7	Buss et al. [16]	20	23.8 ± 3.8	5	TSP	Switch > Repeat trials
8	Buss et al. [16]	20	23.8 ± 3.8	7	TSP	Shifting dimensions > Repeat dimensions
9	Calcott and Berkman [17]	19	22.63 ± 3.59 (19–30)	1	TSP	Switch > Non-Switch
10	Chiu and Yantis [21]	16	20–32	3	TSP	Rule Switch > Rule hold
11	Chiu and Yantis [21]	16	20–32	6	TSP	Attention Shift > Attention hold
12	Crone et al. [24]	20		2	TSP	Univalent switches > Repetitions
13	Crone et al. [24]	20		23	TSP	Bivalent switches > Repetitions
14	Cubillo et al. [25]	30	24 ± 2	3	TSP	Switch > Repeat
15	Dang et al. [27]	16	25 ± 2.5	6	TSP	Object shift > No shift
16	De Baene and Brass [29]	19	22 ± 1.8	5	TSP	Task-switch > Cue-repeat
17	DiGirolamo et al. [31]	8	25 (20–30)	68	TSP	Switch > Non-switch
18	DiGirolamo et al. [31]	8	25 (20–30)	31	TSP	Switch > Fixation
19	Dove et al. [33]	16	21–29	13	TSP	Task switch > Task repetition
20	Dreher and Grafman [34]	8	25 (20–31)	14	TSP	Switching > Repeat
21	Dreher et al. [35]	8	25 (20–31)	9	TSP	Switching > Repeat
22	Eich et al. [36]	71	26.10 (20–31)	5	TSP	Switch > Repeat
23	Fuentes-Claramonte et al. [39]	28	24.21 ± 4.08 (19–32)	20	TSP	Switch > Repeat
24	Graham et al. [43]	18	21 (19–25)	17	WCST	MNF > MPF
25	Graham et al. [43]	18	21 (19–25)	14	WCST	2 + NF > 2 + PF
26	Gu et al. [45]	21	24.8 ± 3.7	18	TSP	Switch > Repeat
27	Hakun and Ravizza [46]	20	21.47 ± 2.95 (18–29)	3	TSP	RULE-Catch Switch > Repeat
28	Hedden and Gabrieli [48]	17	21.6 (18–28)	1	TSP	Neutral Shifting > Incongruent Non-Shifting
29	Hippmann et al. [49]	23	23 (19–29)	5	TSP	Switch > Repeat
30	Jamadar et al. [51, 52]	18	25 ± 7	9	TSP	Informative switch > Repeat
31	Kim et al. [59]	16	23.6 ± 2.9 (18–35)	20	TSP	Switch > Non switch
32	Konishi et al. [62]	16	19–35	9	Modified WCST	Dimensional change > no‐change
33	Lie et al. [73]	12	24 ± 5 (19–36)	10	WCST	A – HLB
34	Liston et al. [74]	19		7	TSP	Shift > Repeat (color and motion)
35	Liston et al. [74]	19		2	TSP	Shift > Repeat (color)
36	Liston et al. [74]	19		2	TSP	Shift > Repeat (motion)
37	Liu et al. [76]	48	20.65 ± 2.1	5	TSP	Switch > Repeat
38	Methqal et al. [82]	20	24.85 ± 3.85 (19–35)	12	Word-matching task	MNF > MCF
39	Methqal et al. [82]	20	24.85 ± 3.85 (19–35)	8	Word-matching task	MNF > MPF
40	Monchi et al. [85]	11	24 (18–31)	30	WCST	RNF > RCF
41	Monchi et al. [85]	11	24 (18–31)	6	WCST	MNF > MCF
42	Muhle-Karbe et al. [86]	44	1 study: 21.1 (21.1); 2 study: 23.3 (18–32)	13	TSP	Switch > Repeat (Task switch)
43	Muhle-Karbe et al. [86]	44	1 study: 21.1 (21.1); 2 study: 23.3 (18–32)	12	TSP	Switch > Repeat (Full Switch)
44	Muhle-Karbe et al. [86]	44	1 study: 21.1 (21.1); 2 study: 23.3 (18–32)	10	TSP	Switch > Repeat (SR Switch)
45	Nagahama et al. [87]	6	27.4 ± 8.1	14	Modified WCST	Set shifting > reversal
46	Nagano-Saito et al. [88]	19	22.6 ± 2.2 (18–27)	29	WCST	MNF > MCF
47	Nagano-Saito et al. [88]	19	22.6 ± 2.2 (18–27)	24	WCST	RNF > RCF
48	Nir-Cohen et al. [91]	43	25.05 ± 2.5	5	TSP	Switch > Repeat
49	Orr and Banich [93]	28	21.6 ± 3.8	19	TSP	Switch > Repeat
50	Parris et al. [96]	22	25 (21–52)	20	TSP	Flip—Hold
51	Perianez et al. [100]	19	26.8 ± 1.3	9	TSP	Switch > Repeat
52	Philipp et al. [101]	23	26	2	TSP	Stimulus‐categorization switch > Stimulus‐categorization repetition
53	Philipp et al. [101]	23	26	1	TSP	Response-modality switch > Response-modality repetition
54	Piguet et al. [102]	18	24.9 ± 5.46	5	TSP	Switch > Repeat
55	Pollmann et al. [104]	11	22–27	6	TSP	Switch > Repeat
56	Ravizza and Carter [106]	14	27.14	3	TSP	Rule Shift > Rule repetition
57	Ravizza and Carter [106]	14	27.14	3	TSP	Perceprual shift > Perceprual repeat
58	Rubia et al. [113]	22	28 ± 6 (20–43)	7	TSP	Switch > Repeat
59	Ruge et al. [114]	18	25.5 (21–35)	2	TSP	Task switch > Task repeat
60	Rushworth et al. [115]	10	19–31	2	TSP	Switch > Stay
61	Rushworth et al. [115]	10	19–31	4	TSP	Switch > Stay
62	Sali et al. [116]	24	26.25 ± 5.4 (21–39)	6	TSP	Switch > Repeat
63	Sato et al. [117]	15	22 ± 3	5	WCST	RNF > RPF
64	Sekutowicz et al. [121]	108	26.26 ± 3.75	22	TSP	Task switch > Repetition
65	Shi et al. [124]	32	25.5 ± 3.9 (20–36)	21	TSP	Task switching > Task repetition
66	Simard et al. [125]	14	26 ± 2.29 (22–31)	36	WWST	RNF > RCF
67	Simard et al. [125]	14	26 ± 2.29 (22–31)	30	WWST	MNF > MCF
68	Simard et al. [125]	14	26 ± 2.29 (22–31)	27	WWST	MNF > MPF
69	Simard et al. [125]	14	26 ± 2.29 (22–31)	32	WWST	FNF > RPF
70	Smith et al. [128]	20	28.8 ± 7(20–43)	10	TSP	Switch > Repeat
71	Sohn et al. [130]	12	18–36	4	TSP	Repetition and switch x Scan
72	Stelzel et al. [134]	48	Female (22 ± 1.99); male (22.6 ± 1.99)	5	TSP	Task switch > Task repetition
73	Stelzel et al. [134]	48	Female (22 ± 1.99); male (22.6 ± 1.99)	13	TSP	Hand switch > Hand repeat
74	Stelzel et al. [135]	18	Female (25.6 ± 2.8); male (26.5 ± 6.7)	4	TSP	Rule switch > Rule repeat
75	Stelzel et al. [135]	18	Female (25.6 ± 2.8); male (26.5 ± 6.7)	7	TSP	Hand switch > Hand repeat
76	Tei et al. [137]	24	31 ± 6.6 (20–46)	10	TSP	Switch > Repeat
77	Tsumura et al. [138]	27	18–23	40	TSP	Switch > Repeat
78	Vallesi et al. [139]	31	23 (21–30)	10	TSP	Task switching > Repeat (Spatial)
79	Vallesi et al. [139]	31	23 (21–30)	8	TSP	Task switching > Repeat (verbal)
80	Vatansever et al. [141]	28	26.8 ± 2.8 (22–34)	5	WCST	Task > Control
81	Ward et al. [143]	30	22 (18–32)	2	TSP	Switch Only > Repeat
82	Weissberger et al. [144]	19	20.45 ± 1.9	6	TSP	Switch trials > Repeat (color-shape)
83	Whitmer and Banich [145]	27	20.04 ± 2.73	6	TSP	Switch > Repeat
84	Witt and Stevens [146]	83	22 ± 2.7 (18–31)	23	TSP	Switch > Non switch
85	Wylie et al. [148]	13	24.5 ± 4.4	11	TSP	Color switch > Color repeat (cues)
86	Wylie et al. [148]	13	24.5 ± 4.4	2	TSP	Speed switch > Speed repeat (cues)
87	Wylie et al. [148]	13	24.5 ± 4.4	19	TSP	Color switch > Color repeat (targets)
88	Wylie et al. [148]	13	24.5 ± 4.4	4	TSP	Speed switch > Speed repeat (targets)
89	Xu et al. [149]	18	26.4 ± 4.5	10	TSP	Switch > Switch go
90	Yeung et al. [151]	15	19–24	12	TSP	Switch > Repeat
91	Yin et al. [152]	26	21.3 (21–25)	15	TSP	Switch > Repeat

RNF  receiving negative feedback; RPF  receiving positive feedback; MNF  matching after negative feedback; MPF matching after positive feedback; 2 + NF  receiving second negative feedback; 2 + PF receiving second positive feedback; A-HLB  no instruction of dimension (A)—High-level baseline

Data extraction

The data collected from each experiment included the following information: (1) authors, (2) publication date, (3) sample size, (4) demographics, (5) number of foci, (6) task, and (7) experimental contrasts. A summary of the extracted data for each experiment is represented in Tables 1, 2, and 3—for young, middle-age, and older adult groups separately. All studies were reviewed twice by ZC and AFi.

Table 2.

Information on source datasets included in the meta-analysis for middle-age adults

No.	Author	N	Mean age ± SD (range)	F	Task	Contrast
1	Berberat et al. [10]	15	38 ± 12 (22–55)	13	TSP	Switch > Repeat
2	Gray et al. [44]	33	37 ± 11.5	43	TSP	Shift > Repeat
3	Jamadar et al. [51, 52]	11	37.42 ± 8.17	10	TSP	Switch > Repeat
4	Kuptsova et al. [67]	36	34.83 ± 8.95 (21–48)	18	TSP	Task switching > Control condition
5	Kuptsova et al. [67]	34	32.12 ± 6.6 (21–48)	11	TSP	Task switching > Control condition
6	Lao-Kaim et al. [71]	32	34 ± 8 (22–46)	5	WCST	Efficient Shift > First correct
7	Monchi et al. [83]	9	54 ± 5 (47–68)	17	WCST	RNF > RCF
8	Monchi et al. [84]	7	51.1 ± 4 (47–60)	13	MCST	Continuous shift > Control
9	Page et al. [94]	11	34.1 ± 10.1	3	TSP	Switch > Repeat
10	Ren et al. [108]	14	34.07 ± 14.4	6	WCST	2 + NF > 2 + PF
11	Ren et al. [108]	14	34.07 ± 14.4	12	WCST	MNF > MPF
12	Schmitz et al. [119]	12	39 ± 6 (18–52)	9	TSP	Switch > Repeat

Table 3.

Information on source datasets included in the meta-analysis for older adults

No.	Author	N	Mean age ± SD (range)	F	Task	Contrast
1	Au et al. [7]	17	60.5 ± 9.2	7	Modified WCST	MCST—NF
2	DiGirolamo et al. [31]	8	69 ± 4 (63–75)	47	TSP	Switch > Fixation
3	Eich et al. [36]	175	65.34 (60–71)	12	TSP	Switch > Repeat
4	Jimura and Braver [53]	14	73 ± 7 (65–87)	3	TSP	Switch > Repeat
5	Kimura et al. [61]	21	70.9 ± 4.10 (60–79)	11	TSP	Switch > Repeat
6	Lamar et al. [68]	10	63 ± 5.3 (56–73)	7	TSP	Switch > Repeat
7	Martins et al. [81]	10	62 ± 8 (55–75)	22	WWST	MNF > CM
8	Martins et al. [81]	10	62 ± 8 (55–75)	16	WWST	MNF > MPF
9	Martins et al. [81]	10	62 ± 8 (55–75)	12	WWST	FNF > RCF
10	Methqal et al. [82]	20	69.45 ± 4.54 (63–80)	12	Word-matching task (based on computerized WCST)	Switch rule—Control matching
11	Methqal et al. [82]	20	69.45 ± 4.54 (63–80)	7	Word-matching task (based on computerized WCST)	Switch rule—Maintain rule
12	Nieuwhof et al. [90]	26	71.2 ± 5.3	11	TSP	Switch > Repeat
13	Skolasinska et al. [126]	129	71.5 ± 4.5	7	TSP	Switch > Repeat
14	Yoon et al. [54]	26	68.6 ± 6.1 (60.1–80.9)	14	WCST	RNF > RPF
15	Yoon et al. [54]	26	68.6 ± 6.1 (60.1–80.9)	8	WCST	MNF > MPF

Data categorization

Since the aim of this work was to distinguish cognitive flexibility tasks that require switching between known or unknown rules at different ages, we grouped the data according to (a) rule-derivation type and (b) age. Consistent with our previous work [23], tasks with cued rule switching were categorized as rule retrieval, represented by the TSP. Correspondingly, tasks where switching rules needed to be discovered via feedback-based trial-and-error rather than using a provided cue were categorized as rule discovery, represented by the WCST.

To investigate age-related brain correlates of cognitive flexibility, we categorized the data into three age groups: young adults (20–30 years), middle-age adults (31–60 years), and older adults (60–80 years). Given age-range variability between studies, particularly as some studies employed wide age ranges, fixed categorization was sometimes challenging; we assigned such cases to the corresponding age groups based on the mean age and standard deviation. Despite this effort, some overlap between age ranges could still not be fully ruled out. To address such potential age overlap, we identified 24 young adult experiments with mean ages between 20–30 years but age ranges still extending beyond 30 years and ran a comparative analyses excluding these experiments. This yielded results consistent with the full dataset, confirming that age-range variations did not substantially impact our findings (Table S2). Our choice to use age 30 years as the maximum age for young adults, and 31 years as the minimum for the middle-age groups was based mainly on practical reasons for preserving experiments for middle-age group that allowed us to analyze indicative trends in this age group. Further, empirical research suggest that features extracted from resting state fMRI data have been shown to accurately classify brain age [153] reporting high accuracy in distinguishing young adults (21–30 years) from individuals aged 31–40 years [131]. Behaviorally, older adults consistently demonstrate poorer cognitive flexibility compared to younger participants [66]. Although further dividing the middle-age group (31–60 years) into narrower subgroups could provide additional insights, the limited number of available articles precluded this approach. Therefore, we opted for defining the age group ranges broadly based on data extracted from selected studies and their associated brain correlates as follows: young adults (20–30 years), middle-age adults (31–60 years), and older adults (60–80 years), as specified above.

Coordinate grouping

Coordinates were grouped to examine cognitive flexibility in young, middle and older age groups and sub-grouped to examine rule-retrieval (TSP) and rule-discovery (WCST) tasks in young, middle and older age groups. A detailed summary of conducted analyses and its purposes are presented in Table 4.

Table 4.

Summary of analyses

Name	Description	Analysis
Age-specific analysis	Separate analyses of each age group reflecting age-related brain correlates of cognitive flexibility	Young adults: TSP + WCST Middle-age adults: TSP + WCST Older adults: TSP + WCST
Conjunction analysis of three age groups	Conjunction of cognitive flexibility tasks for all three age groups reflecting concordant brain correlates across groups	Young ∩ Middle ∩ Old
Task-specific analysis for each age group	Separate analyses of rule-retrieval (TSP) and rule-discovery (WCST) for young, middle, and older adult groups, reflecting task-specific brain correlates	TSP: young, middle, older adults WCST: young, middle, older adults
Conjunction analysis of cognitive flexibility tasks for each age groups	Conjunction of the cognitive flexibility tasks for young, middle and older adult groups reflecting common brain correlates between tasks for each group	Young adults: TSP ∩ WCST Middle-age adults: TSP ∩ WCST Older adults: TSP ∩ WCST
Contrast analysis between cognitive flexibility tasks within age groups	Contrast analysis between cognitive flexibility tasks for different age groups reflecting task-specific (i.e., rule-retrieval and rule-discovery) neural correlates in young, middle, and older adults	Young adults: TSP vs WCST Middle-age adults: TSP vs WCST Older adults: TSP vs WCST
Conjunction analysis of age groups within cognitive flexibility tasks	Conjunction of the age groups for each task reflecting communalities between young, middle and old age for rule-retrieval and rule-discovery types of switching	TSP: Young ∩ Middle, Young ∩ Old, Middle ∩ Old TSP: Young ∩ Middle ∩ Old WCST: Young ∩ Middle, Young ∩ Old, Middle ∩ Old WCST: Young ∩ Middle ∩ Old
Contrast analysis between age groups within cognitive flexibility tasks	Contrast analysis between young, middle and old age reflecting age-specific brain correlates for rule-retrieval and rule-discovery types of switching	TSP: Young vs Middle, Young vs Old, Middle vs Old WCST: Young vs Middle, Young vs Old, Middle vs Old

Activation likelihood estimation meta-analysis

All meta-analyses were conducted using a coordinate-based Activation Likelihood Estimation (ALE) method using GingerALE software, version 3.0.2 (https://brainmap.org/ale/). ALE evaluates the overlap of coordinates (foci) gathered from various studies to identify clusters that are most likely to be consistently active. The latest version of this algorithm enhances statistical power by incorporating all relevant experiments while reducing within-group effects [37]. All coordinates were standardized to a common MNI space, using Lancaster transform method [69]. For individual analyses, findings were thresholded at p < 0.05 cluster-level family-wise error (cFWE) corrected for multiple comparisons and cluster-forming threshold at p < 0.001 [37]. For contrast and conjunction analyses, the threshold was set to p < 0.01 (10,000 permutations, 200 mm³ minimum volume, e.g., [5, 150]).

Results

Age-specific analysis

This analysis included data from articles employing the WCST and/or TSP to assess cognitive flexibility in different age groups (see Table 5 and Fig. 2).

Table 5.

Age-related brain correlates of the cognitive flexibility

Cluster	Volume (mm3)	Region	BA	ALE Value	x	y	z
Young adults (TSP + WCST)
1	12,736	L Inferior Frontal Gyrus	9	0.079	− 46	8	32
		L Middle Frontal Gyrus	9	0.055	− 44	32	30
		L Middle Frontal Gyrus	46	0.050	− 44	22	26
2	12,152	L Superior Parietal Lobule	40	0.076	− 34	− 52	48
		L Inferior Parietal Lobule	40	0.051	− 44	− 42	50
		L Precuneus	7	0.046	− 26	− 70	40
		L Postcentral Gyrus	2	0.027	− 50	− 28	52
		L Postcentral Gyrus	2	0.027	− 46	− 24	38
3	10,592	L Medial Frontal Gyrus	8	0.089	− 4	18	46
		R Medial Frontal Gyrus	8	0.046	4	30	38
		L Superior Frontal Gyrus	6	0.039	− 2	6	58
		L Superior Frontal Gyrus	6	0.035	− 2	0	70
		R Cingulate Gyrus	32	0.024	6	16	36
4	4160	L Middle Frontal Gyrus	6	0.064	− 28	8	60
		L Middle Frontal Gyrus	6	0.057	− 28	− 4	56
5	3976	R Superior Parietal Lobule	7	0.033	38	− 58	54
		R Inferior Parietal Lobule	40	0.033	44	− 44	48
		R Precuneus	19	0.028	32	− 68	38
		R Inferior Parietal Lobule	7	0.028	34	− 58	42
6	3576	R Insula	13	0.063	34	24	− 4
		R Lentiform Nucleus		0.031	18	20	− 4
7	3376	L Precuneus	7	0.056	− 8	− 70	48
8	2816	R Middle Frontal Gyrus	46	0.038	46	34	24
9	2704	L Insula	13	0.056	− 32	24	0
10	2168	R Thalamus		0.040	10	− 12	8
		R Lentiform Nucleus		0.029	14	2	2
11	2016	R Middle Frontal Gyrus	9	0.041	44	10	32
		R Middle Frontal Gyrus	9	0.025	52	22	36
12	1904	R Middle Frontal Gyrus	6	0.041	28	2	58
13	1424	L Thalamus		0.059	− 10	− 16	8
14	1408	L Lentiform Nucleus		0.038	− 16	6	2
		L Lentiform Nucleus		0.028	− 16	6	10
Middle-age adults (TSP + WCST)
1	2696	R Medial Frontal Gyrus	8	0.020	2	24	44
		L Medial Frontal Gyrus	8	0.019	− 6	18	46
		R Superior Frontal Gyrus	8	0.018	2	16	48
		L Superior Frontal Gyrus	6	0.018	− 4	14	48
		R Medial Frontal Gyrus	8	0.018	4	18	44
2	2264	R Superior Parietal Lobule	7	0.021	38	− 56	48
		R Inferior Parietal Lobule	40	0.021	46	− 40	44
3	1664	L Inferior Parietal Lobule	40	0.020	− 36	− 52	44
		L Inferior Parietal Lobule	40	0.017	− 44	− 46	44
4	1168	R Insula	13	0.023	32	24	0
5	1152	L Inferior Frontal Gyrus	9	0.023	− 44	4	30
6	896	L Inferior Occipital Gyrus	18	0.017	− 32	− 88	− 4
		L Middle Occipital Gyrus	18	0.016	− 28	− 94	4
7	824	R Middle Occipital Gyrus	18	0.019	34	− 86	− 2
8	792	L Middle Frontal Gyrus	46	0.018	− 44	30	24
		L Middle Frontal Gyrus	46	0.014	− 44	38	26
Older adults (TSP + WCST)
1	2488	L Middle Frontal Gyrus	9	0.026	− 42	20	28
		L Inferior Frontal Gyrus	9	0.014	− 40	8	30
2	1960	L Medial Frontal Gyrus	8	0.025	− 2	24	46
3	1280	R Middle Frontal Gyrus	10	0.023	34	58	6
4	1176	L Inferior Parietal Lobule	40	0.016	− 38	− 52	46
		L Inferior Parietal Lobule	7	0.015	− 32	− 56	40
5	1104	L Insula	13	0.021	− 32	22	2
6	968	L Inferior Frontal Gyrus	10	0.020	− 40	54	2

Coordinates are in MNI space; R = right; L = left; BA = Brodmann Areas; Vol = volume

Fig. 2.

ALE maps for cognitive flexibility in (A) Young adults, (B) Middle-age adults, (C) Older adults, (D) Conjunction of three age groups. R = right; L = left

For young adults, results showed concordance in bilateral fronto-parietal, cingulo-opercular areas, and subcortical regions, including thalamus and lentiform nucleus. The largest clusters of concordance were found in the left frontal and parietal cortices.

For middle-age adults brain concordance was observed in the bilateral medial frontal gyri (Brodmann area (BA) 8–6), as well as fronto-parietal regions (BA 9/46, 40), occipital cortices, and right insula. The largest cluster of activation was observed in bilateral medial frontal and superior frontal gyri with a max ALE value in the right medial frontal gyrus.

For older adults, results showed convergence in frontal and parietal areas mostly in the left hemisphere. The largest cluster of activation was detected in the left middle frontal gyrus, extending into the left inferior frontal gyrus. Activation was also found in the right middle frontal gyrus, left medial frontal gyrus, left inferior parietal lobule and left insula.

Conjunction analysis across three age groups for both tasks

Our conjunction analysis considering both tasks revealed common brain areas across three age groups—in the right insula cortex, left inferior parietal lobule, and bilateral medial frontal gyri (see Table 6).

Table 6.

Common brain correlates among young, middle-age, and older adults

Young adults ∩ middle-age adults ∩ older adults (TSP + WCST)
Cluster	Volume (mm3)	Region	BA	ALE Value	x	y	z
1	121	L Medial Frontal Gyrus	8	0.014	− 4	20	46
		R Medial Frontal Gyrus	8	0.012	3	25	45
2	73	R Anterior Insula	13	0.012	34	24	− 3
3	49	L Inferior Parietal Lobule	40	0.015	− 38	− 50	44

Coordinates are in MNI space; R  right; L  left; BA  Brodmann Areas; Vol  volume

Task-specific analysis

Rule-retrieval (TSP)

Results associated with rule-retrieval (i.e., TSP) for young adults showed concordance in fronto-parietal areas in the left hemisphere, with an exception for the right middle frontal gyrus, and bilateral cingulo-opercular regions (Table S3, Fig. 3). Results for middle-age adults showed concordance mainly in left fronto-parietal areas (Table 7, Fig. 3). The largest cluster of activation peaked in the left superior frontal gyrus extending into right superior frontal gyrus and left medial frontal gyrus. There was also activation in the left inferior frontal and parietal areas, as well as right cerebellum. Rule-retrieval switching in older adults showed one cluster in the left inferior parietal lobule (Table 7, Fig. 3).

Fig. 3.

ALE maps for rule-retrieval switching type in young, middle-age and older participants. R = right; L = left

Table 7.

Task-specific brain correlates of cognitive flexibility in middle-age and older groups

Rule-discovery (WCST)
Cluster	Volume (mm3)	Region	BA	ALE value	x	y	z
Middle-age adults
1	1528	R Medial Frontal Gyrus	8	0.018	0	26	44
2	784	R Insula	13	0.012	34	24	− 2
3	696	R Middle Frontal Gyrus	9	0.011	46	30	26
Older adults
1	1736	R Middle Frontal Gyrus	10	0.023	34	58	6
2	1600	R Superior Frontal Gyrus	8	0.016	6	26	48
		L Medial Frontal Gyrus	8	0.015	− 4	22	46
3	1552	L Insula	13	0.021	− 30	22	2
4	1512	L Middle Frontal Gyrus	9	0.015	− 44	24	30
5	1400	L Inferior Frontal Gyrus	10	0.019	− 40	54	2
6	728	R Insula	13	0.017	34	22	− 2

Rule-retrieval (TSP)
Cluster	Volume (mm3)	Region	BA	ALE value	x	y	z
Middle-age adults
1	1576	L Superior Frontal Gyrus	6	0.018	− 4	14	48
		R Superior Frontal Gyrus	6	0.017	2	14	48
		L Medial Frontal Gyrus	8	0.017	− 6	18	46
2	648	L Inferior Parietal Lobule	40	0.017	− 36	− 50	44
3	592	R Anterior Lobe.Culmen		0.015	34	− 66	− 28
4	592	L Precentral Gyrus	9	0.016	− 44	4	32
Older adults
1	920	L Inferior Parietal Lobule	40	0.015	− 38	− 50	46

Coordinates are in MNI space; R = right; L = left, BA = Brodmann Areas; Vol = volume

Rule-discovery (WCST)

Cognitive flexibility associated with rule discovery (i.e., WCST) in young adults showed concordance in bilateral fronto-parietal, cingulo-opercular areas, as well as in the subcortical regions of the right hemisphere (Table S3, Fig. 4). In the middle-age group, concordance was found in the right frontal cortex. Specifically, activation was observed in the right middle and medial frontal gyri, as well as in the right insula (Table 7, Fig. 4). Rule-discovery switching type in older adults showed activation in frontal and insular cortices bilaterally: areas with higher ALE score were detected in right middle frontal gyrus, bilateral insular and right inferior frontal gyrus. Other areas included right superior frontal gyrus and left middle and medial frontal gyri (Table 7 and Fig. 4).

Fig. 4.

ALE maps for rule-discovery switching type in young, middle and old age groups. R = right; L = left

Age specific analysis within cognitive flexibility tasks

Rule-discovery (WCST)

Contrast analysis

Young adults showed increased concordance in fronto-parietal areas (BA 9/46, 7/40) when compared with older adults, whereas older adults showed increased concordance in the left inferior frontal gyrus (BA 46). No differences were observed in comparisons with middle-age adults (Fig. 5, Table 8).

Fig. 5.

ALE maps for rule-discovery (A) Older adults > Young adults (green), (B) Young adults > Older adults (red), and (C) Older adults ∩ Young adults (blue). R = right; L = left

Table 8.

Age-specific brain correlates within cognitive flexibility tasks

Rule-discovery (WCST)
Cluster	Volume (mm3)	Region	BA	ALE value	x	y	z
Young adults > older adults
1	1712	R Superior Parietal Lobule	7	3.540	34	− 48	46
		R Superior Parietal Lobule	7	2.697	34	− 52	57
		R Superior Parietal Lobule	7	2.442	38	− 52	56
2	1360	L Superior Parietal Lobule	7	3.121	− 28	− 60	58
		L Inferior Parietal Lobule	40	3.011	− 40	− 53	53
		L Superior Parietal Lobule	7	2.820	− 32	− 53	53
3	800	R Middle Frontal Gyrus	46	3.291	48	42	18
		R Superior Frontal Gyrus	9	3.195	48	42	22
		R Middle Frontal Gyrus	9	3.036	40	40	16
Older adults > young adults
1	328	L Inferior Frontal Gyrus	46	3.156	− 46	46	2
Young adults > middle-age adults
No suprathreshold clusters
Middle-age adults > young adults
No suprathreshold clusters
Middle-age adults > old adults
No suprathreshold clusters
Older adults > middle-age adults
No suprathreshold clusters

Rule-retrieval (TSP)
Cluster	Volume (mm3)	Region	BA	ALE value	x	y	z
Young adults > middle-age adults
No suprathreshold clusters
Middle-age adults > young adults
1	576	R Cerebellum, Culmen		3.891	32	− 63	− 26
		R Cerebellum, Culmen		3.353	37	− 66	− 28
2	240	R Medial Frontal Gyrus	32	3.156	6	12	44
Young adults > older adults
No suprathreshold clusters
Older adults > young adults
No suprathreshold clusters
Middle-age adults > older adults
No suprathreshold clusters
Older adults > middle-age adults
No suprathreshold clusters

Coordinates are in MNI space; R = right; L = left, BA = Brodmann Areas; Vol = volume

Conjunction analysis

Young and middle-age adults shared common areas in right medial and middle frontal gyri, as well as in the right insula (Table 9, Fig. 5). Middle-age and older adults shared similar common activity with an exception of the right middle frontal gyrus (Table 9). Conjunction analysis of young and older adults showed bilateral frontal and insular activation (Table 9, Fig. 5). All three groups showed common (albeit small) clusters in cingulo-opercular regions (Table 9).

Table 9.

Shared brain correlates of three age groups within cognitive flexibility tasks

Rule-discovery (WCST)
Cluster	Volume (mm3)	Region	BA	ALE value	x	y	z
Young adults ∩ middle-age adults
1	1280	R Medial Frontal Gyrus	8	0.018	0	26	44
2	744	R Insula	13	0.012	34	24	− 2
3	624	R Middle Frontal Gyrus	9	0.011	46	30	26
Middle-age adults ∩ older adults
1	752	L Medial Frontal Gyrus	6	0.013	− 2	24	46
		R Medial Frontal Gyrus	8	0.013	4	26	44
2	584	R Insula	13	0.012	34	24	− 2
Young adults ∩ older adults
1	1336	R Superior Frontal Gyrus	8	0.016	6	26	48
		L Medial Frontal Gyrus	6	0.015	− 4	22	46
2	832	L Insula	13	0.021	− 30	24	2
3	824	L Middle Frontal Gyrus	9	0.015	− 44	24	30
4	728	R Insula		0.017	34	22	− 2
5	360	L Middle Frontal Gyrus	10	0.015	− 38	54	4
Young adults ∩ middle-age adults ∩ older adults
1	94	L Medial Frontal Gyrus	8	0.013	-2	24	46
		R Medial Frontal Gyrus	8	0.012	3	25	45
2	73	R Insula	13	0.012	34	24	-2

Rule-retrieval (TSP)
Cluster	Volume (mm3)	Region	BA	ALE value	x	y	z
Young adults ∩ middle-age adults
1	1416	L Medial Frontal Gyrus	6	0.018	− 4	14	48
		R Medial Frontal Gyrus	6	0.017	2	14	48
		L Medial Frontal Gyrus	6	0.017	− 6	18	46
2	632	L Inferior Parietal Lobule	40	0.017	− 36	− 50	44
3	528	L Precentral Gyrus	6	0.016	− 44	4	32
Middle-age adults ∩ older adults
1	392	L Inferior Parietal Lobule	40	0.015	− 38	− 50	44
Young adults ∩ older adults
1	856	L Inferior Parietal Lobule	40	0.015	− 38	− 50	46
Young adults ∩ middle-age adults ∩ older adults
1	49	L Inferior Parietal Lobule	40	0.015	− 38	− 50	44

Coordinates are in MNI space; R = right; L = left, BA = Brodmann Areas; Vol = volume

Rule-retrieval (TSP)

Contrast analysis

Compared with the young-aged group, the middle-age group exhibited higher activations in the right cerebellum and right medial frontal gyrus (Table 8; Fig. 6). Other contrasts were not statistically different.

Fig. 6.

ALE maps for rule-retrieval (A) Middle-age adults > Young adults (green), and (B) Young adults ∩ Middle-age adults (red). Note. R = right; L = left

Conjunction analysis

Young and middle-age groups shared common activation areas in medial frontal gyrus and parietal regions mainly in the left hemisphere (Table 9). Paired conjunction analysis of young and older adults, as well as middle- and older-adult groups showed a single common cluster in the left inferior parietal lobule. All three groups shared activation in the left inferior parietal lobule (BA 40).

Task specific analysis within age groups

Conjunction analysis of cognitive flexibility tasks within age groups

WCST (rule-discovery) and TSP (rule-retrieval) in young adults revealed common clusters in the left frontoparietal regions, as well as in bilateral insula (Table S4). In middle-age and older adults, conjunction analyses revealed no suprathreshold clusters.

Contrast analysis between cognitive flexibility tasks within age groups

Contrast analyses in young adults revealed increased concordance for rule-discovery (WCST) than with rule retrieval (TSP) in prefrontal (BA 9, 46, 10), parietal (BA 40, 7, 39) as well as insular and subcortical clusters (Table S4). No suprathreshold clusters were found associated with rule-retrieval compared with rule-discovery. For middle-age adults, rule-discovery (WCST), when contrasted with rule-retrieval, showed increased concordance in the right medial frontal gyrus. Rule-retrieval (TSP) did not reveal any suprathreshold clusters in comparison with the rule-discovery one (Table S4). For older adults, rule-discovery specific activation was detected in bilateral frontopolar areas (BA 10). No suprathreshold areas associated with rule-retrieval > rule-discovery contrast were detected (Table S4).

Discussion

This meta-analysis examined brain regions associated with cognitive flexibility across adulthood, revealing two key findings. First, cognitive flexibility demonstrated an age-related decrease in the number of implicated brain regions with a shift toward left-hemispheric dominance. Specifically, we observed age-related reductions in activation in posterior regions and increased left–right asymmetry, particularly pronounced in older adults. Second, rule-discovery (WCST) and rule-retrieval (TSP) tasks showed distinct age-related changes in neural correlates, albeit using a relatively small number of articles. For TSP, middle-age adults, as opposed to young ones, exhibited increased involvement of the right cerebellum and medial frontal gyrus (BA 8), while for WCST, older adults showed greater engagement of the left inferior frontal gyrus (BA 46). Furthermore, activation patterns for rule-retrieval switching remained predominantly left-lateralized across all age groups. In contrast, rule-discovery switching engaged both hemispheres, with the exception of middle-age adults who displayed activation exclusively in the right hemisphere. Notably, older individuals showed specific activation in the left inferior frontal gyrus (BA 46), whereas young adults demonstrated bilateral concordance in parietal areas and right-lateralized frontal regions. These findings are briefly discussed in more detail below within frameworks of compensatory theories and constructivist theories of age-associated neurocognitive changes and in light of existing behavioral evidence.

Cognitive flexibility throughout adulthood

Overall, cognitive flexibility tasks appear to engage widespread brain areas in both hemispheres, with common regions in medial frontal gyri, the anterior insula, and the inferior parietal lobule. Although the number of clusters in young adults is more extensive, middle-age adults showed similar bilateral concordance, whereas older adults showed concordance primarily in the left hemisphere. These changes may be attributed to general cerebral changes that accompany normal aging [107, 109]. It has been demonstrated, for example, that frontoparietal regions—the core regions for cognitive flexibility—undergo substantial changes in older adults, with notable gray matter loss and white matter shrinkage [107, 109] as well as reduction in functional connectivity [20]. Three brain regions appear to be consistently involved in cognitive flexibility across all considered age groups and irrespective of the task: bilateral medial frontal gyri, right anterior insula, and left supramarginal gyrus. These regions can thus be considered as critical nodes for the general cognitive flexibility function.

Furthermore, we observed age-related left–right asymmetry in posterior regions, as well as age-related posterior-anterior shift. Both patterns have been demonstrated and discussed in the literature within the frameworks of the Right Hemi-Aging Model [32] and Posterior-Anterior Shift Hypothesis (PASA [28]), respectively. Particularly, PASA refers to age-related reduction in occipital activity coupled with increased activation in lateral prefrontal cortex. This phenomenon is considered to reflect compensatory processes, suggesting that older adults' inability to engage specialized neural mechanisms within posterior regions is compensated by the activation of the frontal lobe to help ensure optimal cognitive functioning [28].

Differences in concordance may reflect changes in cognitive strategies, given that older adults typically require more time for problem-solving [19, 47]. Age-related changes in general cognitive resources, such as mental attentional capacity, may influence the resources available for problem-solving tasks, representing a key component of working memory function [99]. Indeed, the theory of constructive operators suggests that adulthood involves transitions in mental-attentional capacity across developmental periods [97, 98], with specific changes aligning with molecular aging processes [123]. This theoretical framework is supported by evidence attributing cognitive reserve as a driving factor for preserving cognitive flexibility in particular, and cognitive efficiency in general, with age [40, 47]. Additionally, clinical research has demonstrated altered activation patterns in neurodegenerative disorders such as Parkinson's Disease [83] and in preclinical conditions like Mild Cognitive Impairment [88]. These convergent findings from psychological, neuroscience, and molecular studies support the view that aging involves transitions in neurocognitive resource allocation across adulthood.

Regarding left–right asymmetry, our findings indicate that bilateral concordance, characteristic of young individuals, gradually gives way to the left-hemispheric concordance. One possible explanation for this shift is that it may result from changes in gray and white matter with aging in the right hemisphere documented in some previous studies [32, 42, 72]. This age-related neuroanatomical tissue atrophy is consistent with the observation that activations in posterior right hemisphere (e.g., in tasks involving visuospatial processing, attention, and monitoring) are more susceptible to age-related cognitive changes [32]. Consequently, these anatomical and functional changes may lead to a greater involvement of the left hemisphere in order to compensate for the declining function in areas subserving task-related performance at younger ages thus allowing for maintaining relatively consistent performance levels over time. Moreover, this left-lateralized shift may be additionally explained by utilization of specific strategies by older individuals during task performance. For instance, it has been shown that verbal self-instruction can enhance task-switching performance in older adults [58, 65]; employing a verbal self-instruction strategy would likely involve the left hemisphere, given its dominant role in the language function.

Hemispheric activation patterns may also reflect dynamic trade-offs between task demands and available mental-attentional capacity, as proposed by the right-left–right hypothesis [4,154]. According to this framework, familiar but non-automatized cognitive schemes (such as those in TSP tasks) impose lower executive demands and preferentially engage left-hemisphere networks, while novel executive schemes requiring discovery (such as in WCST tasks) primarily recruit right-hemisphere resources [4, 23, 99]. However, additional research is necessary to fully elucidate these hemispheric specialization patterns.

Rule-retrieval switching across adulthood

Rule-retrieval switching, assessed via the Task-Switching Paradigm (TSP), exhibited left-dominant concordance in frontoparietal regions across all age groups. Consistent with the general pattern of age-related changes in cognitive flexibility, rule-retrieval switching also showed a reduction in the number of recruited brain areas in middle-age and older adults compared to younger participants. Notably, only one brain region demonstrated consistent activation across all three age groups: the left inferior parietal lobule (supramarginal gyrus)—an area traditionally associated with multiple functions, including those involved in speech comprehension, production, and reading [18, 92, 127]. This area was also identified as the sole and common region for switching between stimulus categorizations and response modalities, suggesting its critical role in the selection of action rules [101]. This finding indicates that the left supramarginal gyrus is implicated in mental set reconfiguration that implies a selection of the relevant task set. Based on the previous findings and the present meta-analysis, it may therefore be the case that this region constitutes one of the key nodes in the age-resistant cognitive flexibility network.

Age-group comparisons also revealed an additional involvement of the right cerebellum and right medial frontal gyrus (BA 32) in the middle-age group, in contrast to the younger one. The role of the right medial frontal gyrus, part of the anterior cingulate cortex (BA 32), in cognitive flexibility has been well-established. This region is involved in conflict monitoring, ensuring adaptive and flexible responses during changing task demands, which happens often when a cue is provided. The right cerebellum is functionally connected to the left prefrontal cortex (via the cerebrocerebellar loop; [9]), and is implicated in conflict resolution, especially when multiple competitive mental sets are activated [120].

Given behavioral differences in switching ability across young, middle-age, and older adults [19, 38], these areas appear to play a role in the strategies used to ensure flexible behavior through active cue monitoring and subsequent identification of a relevant task. Presumably, the regions associated with conflict monitoring and conflict resolution, such as the right cingulate gyrus and left dorsolateral prefrontal cortex (BA 9) [75, 79] may be insufficient to counteract the onset of aging-related cognitive decline, leading to the recruitment of additional areas with similar functions. Critically, in the case of the medial frontal gyrus, the middle-age group engages the same homologous area in the opposite hemisphere, resulting in bilateral activation of medial frontal gyri. This finding aligns with the "dedifferentiation" hypothesis that proposes age-related non-specificity of brain areas involvement that is not observed in young adults [95], taking the form of contralateral recruitment in this medial frontal gyrus case. Conversely, the cerebellar activation fits the description of unique recruitment, which proposes involvement of distinct brain areas that are not homologous to the relevant areas active in young adults [95]. Thus, such dedifferentiation through the recruitment of additional regions from both homologous and alternative networks serves as a neural compensation mechanism, which can be efficiently utilized for maintaining cognitive functions [80].

Rule-discovery switching across adulthood

Rule-discovery switching based on the WCST demonstrated alterations in brain correlates taking place with agе. Specifically, young adults showed greater activation in bilateral parietal lobules and right dorsolateral prefrontal cortex (BA 9,46) as compared with older adults. Age-related under-activation of frontoparietal areas can be attributed to structural atrophy or to the employment of inefficient or underutilized solving strategies [110]. Notably, our analysis revealed a single brain area associated with the older adult group, namely the left inferior frontal gyrus (BA 46). The involvement of BA 46 can be explained by its role in higher-order cognitive functions such as hypothesis testing [129]. Furthermore, this region is related to the action-selection episodic control [57]. These functions are interconnected and implemented during switching in WCST, since a rule is not provided but rather needs to be discovered. Thus, participants have to initially consider all relevant alternatives, then select one based on specific criteria, and implement it. These functions are susceptible to aging-related changes [64], apparently leading to additional involvement of the related brain area.

Finally, conjunction analysis of the three age groups revealed convergence in bilateral medial frontal gyrus (BA 8) and right insula cortex. One recent lesion study showed a critical role of medial frontal gyrus in decision-making and learning, attention and working memory [55, 136]. As a result, it is challenging to define its specific role in rule-discovery switching as this area is implicated in multiple cognitive processes. The bilateral activation of these regions appears to be crucial for decision-making under uncertainty, given that the WCST employs multivalent stimuli (cards can be categorized based on color, shape, and number), while the insula, crucial for initiating neural switching [26], facilitates mental set reconfiguration.

Considerations

Whilst producing important novel insights, the study is not without some limitations in its methodology. Our sub-analyses were conducted on cognitive flexibility subtypes across three age groups; however, they possessed limited statistical power, as per the Ginger ALE recommendation of at least 17 experiments [37]. Consequently, these findings should be interpreted with caution. Despite the suboptimal nature of these analyses, the decision was made to examine data from the limited number of experiments within each sub-category, as they suggest trends that could inform future research which will be needed to validate and extend the present findings.

Conclusions

This meta-analysis elucidates age-related changes in brain areas associated with cognitive flexibility in general and with the WCST and TSP in particular. Our findings confirm the notion of age-related changes in the extent and type of neural recruitment for cognitive flexibility, presumably induced by anatomical, functional, and strategy differences. Furthermore, our analysis shows that some cognitive flexibility-related activations become more anterior and more left-lateralized with age, supporting in part the hemi-aging model and the Posterior-Anterior Shift in Aging (PASA) theory which posit neuroanatomical redistribution of cognitive functions with age. We also observed distinct aging trajectories for rule retrieval (TSP) and rule-discovery switching (WCST). Specifically, rule retrieval was associated with age-related narrowing in the neural networks underlying cognitive flexibility and additional recruitment of the cerebellum and medial frontal gyrus in middle age beyond that seen in young adults. Concurrently, demonstrating alterations in implicated neural networks across the lifespan, rule-discovery also revealed brain areas associated with young and old age separately—bilateral fronto-parietal activation and activation in inferior frontal gyrus in the left hemisphere, respectively. Additional activation detected for both switching types was attributed to the dedifferentiation mechanisms, characterized by contralateral and unique brain regions recruitment.

Author contributions

ZC, AFa, and MA designed research; ZC, AFa, and AFi performed research; ZC and AFa analyzed data and prepared figures; ZC and AFa wrote original draft; ZC, Afa, AM, YS and MA interpreted results; AM, YS and MA reviewed manuscript; AM, YS and MA supervised.

Funding

This article is an output of a research project implemented as part of the Basic Research Program at the National Research University Higher School of Economics (HSE University). Affiliation: Cognitive Health and Intelligence Centre.

Data availability

All data generated or analyzed during this study are included in this published article.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.