CoChat
Sign Up
Cited by 36Cognitive Abilities and TestingCognitive and developmental aspects of mathematical skillsCognitive Functions and Memory

Near transfer to an unrelated N-back task mediates the effect of N-back working memory training on matrix reasoning

Open access
Pahor, Anja, Seitz, Aaron R., Jaeggi, Susanne M.·2022·Nature Human Behaviour
DOI 10.1038/s41562-022-01384-wPMC PMC12305750
Open this paper in CoChat — free accountOpen PDF

Abstract

The extent to which working memory training improves performance on untrained tasks is highly controversial. Here we address this controversy by testing the hypothesis that far transfer may depend on near transfer using mediation models in three separate randomized controlled trials (RCTs). In all three RCTs, totalling 460 individuals, performance on untrained N-back tasks (near transfer) mediated transfer to Matrix Reasoning (representing far transfer) despite the lack of an intervention effect in RCTs 2 and 3. Untrained N-back performance also mediated transfer to a working memory composite, which showed a significant intervention effect (RCT 3). These findings support a model of N-back training in which transfer to untrained N-back tasks gates further transfer (at least in the case of working memory at the construct level) and Matrix Reasoning. This model can help adjudicate between the many studies and meta-analyses of working memory training that have provided mixed results but have not examined the relationship between near and far transfer on an individual-differences level.

AI snapshot

Written from the paper’s full text.

TL;DR

In three RCTs totaling 460 participants, N-back training’s effect on Matrix Reasoning was mediated by near transfer to untrained N-back tasks, even when overall intervention effects on Matrix Reasoning were absent in RCTs 2 and 3. Untrained N-back also mediated transfer to a working memory composite in RCT 3, but not to inhibitory control or a fluid intelligence composite.

Key findings

  • Across all three RCTs, post-test untrained N-back performance significantly mediated training-group effects on Matrix Reasoning: whole sample indirect effect (IE) = 0.19, 95% CI 0.09–0.31; RCT 1 IE = 0.19, 95% CI 0.03–0.40; RCT 2 IE = 0.21, 95% CI 0.07–0.36; RCT 3 IE = 0.26, 95% CI 0.10–0.43. Total effects on Matrix Reasoning were significant in the whole Retrospective sample (TE = 0.23, p = 0.02) and RCT 1 (TE = 0.40, p = 0.048), but not RCT 2 (TE = 0.17, p = 0.15) or RCT 3 (TE = 0.07, p = 0.51).
  • The parallel mediator Countermanding did not significantly mediate transfer to Matrix Reasoning in the whole sample (IE = −0.02, 95% CI −0.06–0.01), RCT 1 (IE = −0.02, 95% CI −0.12–0.05), RCT 2 (IE = −0.02, 95% CI −0.07–0.02), or RCT 3 (IE = 0.04, 95% CI −0.004–0.11).
  • In RCT 3, untrained N-back mediated transfer to WM Composite 2 (IE = 0.20, 95% CI 0.07–0.30), and WM Composite 1 including N-back mediated Matrix Reasoning (IE = 0.11, 95% CI 0.02–0.22). There was no evidence of mediation via the inhibitory control composite, and no evidence of an intervention effect or mediation for the Gf composite (BFincl = 0.22).
  • Training groups showed near-transfer gains on untrained N-back (ES = 0.80–0.92 in RCTs 1–2; ES = 1.52 in RCT 3), small Matrix Reasoning benefits (ES = 0.43–0.49 in RCTs 1–2; ES = 0.46 in RCT 3), and inhibitory-control benefits in RCT 2 (ES = 0.44) and RCT 3 (ES = 0.39).

Methods

  • Design: Retrospective study with two RCTs (RCT 1 N = 94; RCT 2 N = 177) using no-contact controls, plus a pre-registered, double-blind Replication RCT 3 with an active general-knowledge control (analytical N = 189; 128 N-back training, 61 active control). Total analytical sample across RCTs = 460.
  • Tasks: N-back training for 10 working days in RCTs 1–2 and 20 sessions in RCT 3; near transfer was an untrained N-back task (average 2-back/3-back accuracy); far transfer included Matrix Reasoning (UCMRT in RCTs 2–3); inhibitory control was measured with a Countermanding task; RCT 3 added WM, inhibitory-control, and spatial-reasoning measures to form composites.
  • Analysis: Mediation models using ordinary least squares regression via PROCESS, with training group as the independent variable, post-test untrained N-back and Countermanding as parallel mediators, post-test Matrix Reasoning as the outcome, and pre-test scores as covariates; bootstrap confidence intervals based on 5,000 samples. RCT 3 also used Bayesian repeated-measures ANOVAs on composite scores.

Limitations

  • Stated in the paper: RCT 3 was conducted online due to COVID-19, so it could not fully match the laboratory conditions of RCTs 1–2; participants used their own devices at home, supervision differed, and distractions were likely greater. The sample was young across all RCTs, limiting generalizability, and no follow-up assessments were conducted, so duration of transfer effects is unknown.
  • Stated in the paper: RCTs 1–2 used no-contact controls, so non-specific expectancy/contact effects cannot be ruled out; near and far transfer were measured with single tasks, so results may partly reflect task-specific learning; serial mediation models in RCT 3 were post-hoc and not pre-registered.
  • Stated/evident in the text: In RCTs 1–2, mediator and outcome were measured in the same session, so causality between mediator and outcome cannot be established unequivocally; RCT 3 addressed this by measuring near transfer one day before Matrix Reasoning. The paper also makes no claims about transfer to fluid intelligence at the construct level because the Gf composite showed no effect.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41562-022-01384-w

Cites

107 of 107 references verified0 citation errors · 0 not checkable

Statistics

3 tests recomputed · 0 with the reported and recomputed p on opposite sides of .05

Claims the paper makes

32 of 72 cited sources checked · 20 supported · 2 flagged · accuracy not yet measured

  • Premisecitation

    Training on N-back improves performance on N-back tasks that were not trained.

    “While there is little doubt that N-back training leads to improved performance on untrained N-back tasks13,25–27”
    Quote in the paperSubstantial claim2 of 4 sources checked
    • cites [13] Soveri, A., Antfolk, J., Karlsson, L., Salo, B. & Laine, M. Working memory training revisited: a multi-level meta-analysis of n-back training studies. Psychon. Bull. Rev. 24, 1077–1096 (2017). source
    • cites [25] Soveri, A., Karlsson, E. P. A., Waris, O., Grönholm-Nyman, P. & Laine, M. Pattern of near transfer effects following working memory training with a dual N-back task. Exp. Psychol. 64, 240–252 (2017). source
    • cites [26] Buschkuehl, M., Hernandez-Garcia, L., Jaeggi, S. M., Bernard, J. A. & Jonides, J. Neural effects of short-term training on working memory. Cogn. Affect. Behav. Neurosci. 14, 147–160 (2014). source
    • cites [27] Küper, K. & Karbach, J. Increased training complexity reduces the effectiveness of brief working memory training: evidence from short-term single and dual n-back training interventions. J. Cogn. Psychol. 28, 199–208 (2016). source
  • Premisecitation

    Mediated effects can be statistically significant in the absence of a total effect in certain situations.

    “There is evidence to suggest that in certain situations, mediated effects can be statistically significant in the absence of a total effect.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [52] O’Rourke, H. P. & MacKinnon, D. P. Reasons for testing mediation in the absence of an intervention effect: a research imperative in prevention and intervention research. J. Stud. Alcohol Drugs 79, 171–181 (2018). source
  • Premisecitation

    Transfer to untrained N-back tasks is often not reported, so information about what participants learned is often missing.

    “Transfer to untrained N-back tasks is often not reported therefore information is often missing about what participants learned, and thus, the extent of learning on measures of transfer remains unknown.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [13] Soveri, A., Antfolk, J., Karlsson, L., Salo, B. & Laine, M. Working memory training revisited: a multi-level meta-analysis of n-back training studies. Psychon. Bull. Rev. 24, 1077–1096 (2017). source

      The Directions for future studies section says that closely related tasks should be analyzed systematically to clarify what drives transfer effects. But the paper does not say that transfer to untrained n-back tasks is often unreported; its Main analyses report a moderate transfer effect on those tasks. The claimed reason that information about learning is often missing is therefore not established by this source. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [19] Pergher, V. et al. Divergent research methods limit understanding of working memory training. J. Cogn. Enhanc. https://doi.org/10.1007/s41465-019-00134-7 (2019). source

      The Training Task Features section says that important procedural details are often unreported, and the Discussion and Future Directions section says the mechanisms and extent of transfer after N-back training remain unclear. The paper also notes that not all studies assessed both types of transfer, but the passages do not establish that transfer to specifically untrained N-back tasks is often unreported or that this omission is why information about participants’ learning is missing. Page numbers were not available in our copy. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
  • Premisecitation

    In another meta-analysis of N-back training studies, only 37% reported using N-back tasks as an outcome measure.

    “In another meta-analysis of N-back training studies, only 37% of N-back training studies reported using N-back tasks as an outcome measure19”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [19] Pergher, V. et al. Divergent research methods limit understanding of working memory training. J. Cogn. Enhanc. https://doi.org/10.1007/s41465-019-00134-7 (2019). source

      The Introduction and Transfer Task Features sections say that N-back tasks were among the outcome measures used in the reviewed training studies. They do not report that 37% used N-back outcome tasks. The paper also describes itself as a qualitative review, not a meta-analysis. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
  • Premisecitation

    The power of mediation is likely greater than the power of the total effect test because the mediator is more closely related to the outcome and measured closer in time than the intervention.

    “the power of mediation is likely greater than the power of the test of the total effect because the mediator (untrained N-back) is more closely related to the outcome measure (Matrix Reasoning) than the intervention, and was measured more closely in time with the outcome measure than the intervention52.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [52] O’Rourke, H. P. & MacKinnon, D. P. Reasons for testing mediation in the absence of an intervention effect: a research imperative in prevention and intervention research. J. Stud. Alcohol Drugs 79, 171–181 (2018). source
  • Classified as probably not a substantial claim. 57
  • May combine two claims. 39

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Transfer to sequencing and/or complex WM tasks has been reported by some N-back training studies but not others.

    “Transfer to sequencing and/or complex WM tasks has been reported by some57–63, but not other N-back training studies58,64–70.”
    Quote in the paperNot a substantial claim5 of 10 sources checked
    • cites [57] Li, W. et al. Dual n-back working memory training evinces superior transfer effects compared to the method of loci. Sci. Rep. 11, 3072 (2021). source
    • cites [58] Lilienthal, L., Tamez, E., Shelton, J. T., Myerson, J. & Hale, S. Dual n-back training increases the capacity of the focus of attention. Psychon. Bull. Rev. 20, 135–141 (2013). source
    • cites [59] Chein, J. M. & Morrison, A. B. Expanding the mind’s workspace: training and transfer effects with a complex working memory span task. Psychon. Bull. Rev. 17, 193–199 (2010). source

      The Discussion (page not provided) identifies this study’s training task as an adaptive complex working-memory (CWM) span task, not an N-back task; it describes N-back as a different training paradigm used by earlier studies. This paper therefore cannot serve as an example of an N-back training study that failed to show the stated transfer. (LLM, conf=0.88)

      Checked by a reasoning model — may be wrong.

      Experimental — accuracy not yet measured
    • cites [60] Borella, E., Carretti, B., Riboldi, F. & De Beni, R. Working memory training in older adults: evidence of transfer and maintenance effects. Psychol. Aging 25, 767–778 (2010). source
    • cites [61] Schwarb, H., Nail, J. & Schumacher, E. H. Working memory training improves visual short-term memory capacity. Psychol. Res. 80, 128–148 (2016). source
    • cites [62] Stepankova, H. et al. The malleability of working memory and visuospatial skills: a randomized controlled study in older adults. Dev. Psychol. 50, 1049–1059 (2014). source
    • cites [63] Pahor, A. et al. Multisensory facilitation of working memory training. J. Cogn. Enhanc. https://doi.org/10.1007/s41465-020-00196-y (2020). source
    • cites [64] Linares, R., Borella, E., Lechuga, M. T., Carretti, B. & Pelegrina, S. Nearest transfer effects of working memory training: a comparison of two programs focused on working memory updating. PLoS ONE 14, e0211321 (2019). source
    • cites [65] Blacker, K. J., Negoita, S., Ewen, J. B. & Courtney, S. M. N-back versus complex span working memory training. J. Cogn. Enhanc. 1, 434–454 (2017). source

      Experiment 2’s Results and General Discussion (page unavailable) report no significant transfer from dual n-back training to the Ospan complex-span task, supporting the ‘not others’ portion. The paper discusses cross-modality transfer in prior work, but the passages do not establish that other N-back training studies found transfer to sequencing or complex WM tasks. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
    • cites [66] Chooi, W.-T. & Thompson, L. A. Working memory training does not improve intelligence in healthy young adults. Intelligence 40, 531–542 (2012). source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    There is no consensus on transfer to inhibitory control measures, with some WM training studies reporting benefits and others not.

    “there is no consensus regarding transfer to inhibitory control measures, with some WM training studies reporting training-related benefits59,60,71,72, yet others failing to find evidence of transfer51,73.”
    Quote in the paperNot a substantial claim4 of 6 sources checked
    • cites [59] Chein, J. M. & Morrison, A. B. Expanding the mind’s workspace: training and transfer effects with a complex working memory span task. Psychon. Bull. Rev. 17, 193–199 (2010). source
    • cites [60] Borella, E., Carretti, B., Riboldi, F. & De Beni, R. Working memory training in older adults: evidence of transfer and maintenance effects. Psychol. Aging 25, 767–778 (2010). source
    • cites [71] Schweizer, S., Hampshire, A. & Dalgleish, T. Extending brain-training to the affective domain: increasing cognitive and affective executive control through emotional working memory training. PLoS ONE 6, e24372 (2011). source
    • cites [72] Jones, M. R., Katz, B., Buschkuehl, M., Jaeggi, S. M. & Shah, P. Exploring N-back cognitive training for children with ADHD. J. Atten. Disord. https://doi.org/10.1177/1087054718779230 (2018). source

      The citation is attached to the studies said to have failed to find transfer to inhibitory control. In the Discussion (page unavailable), this paper instead reports improved inhibitory control after n-back training in children with ADHD, albeit with small effects. Its background section describes mixed findings from other studies, but its own result is not a failure to find transfer. (LLM, conf=0.86)

      Checked by a reasoning model — may be wrong.

      Experimental — accuracy not yet measured
    • cites [51] Dahlin, E., Neely, A. S., Larsson, A., Bäckman, L. & Nyberg, L. Transfer of learning after updating training mediated by the striatum. Science 320, 1510–1512 (2008). source
    • cites [73] Maraver, M. J., Bajo, M. T. & Gomez-Ariza, C. J. Training on working memory and inhibitory control in young adults. Front. Hum. Neurosci. 10, 588 (2016). source

  • May combine two claims.

    Premisecitation

    N-back is an updating task that shows a substantial correlation with higher-level cognitive measures, including Matrix Reasoning.

    “One of the most common tasks used in these WM training studies is the N-back, an updating task that correlates significantly with measures of higher-level cognition, such as Matrix Reasoning20–24.”
    Quote in the paperSubstantial claim2 of 5 sources checked
    • cites [20] Burgess, G. C., Gray, J. R., Conway, A. R. A. & Braver, T. S. Neural mechanisms of interference control underlie the relationship between fluid intelligence and working memory span. J. Exp. Psychol. Gen. 140, 674–692 (2011). source
    • cites [21] Gray, J. R., Chabris, C. F. & Braver, T. S. Neural mechanisms of general fluid intelligence. Nat. Neurosci. 6, 316–322 (2003). source
    • cites [22] Hockey, A. & Geffen, G. The concurrent validity and test–retest reliability of a visuospatial working memory task. Intelligence 32, 591–605 (2004). source
    • cites [23] Kane, M. J., Conway, A. R. A., Miura, T. K. & Colflesh, G. J. H. Working memory, attention control, and the N-back task: a question of construct validity. J. Exp. Psychol. Learn. Mem. Cogn. 33, 615–622 (2007). source
    • cites [24] Jaeggi, S. M., Buschkuehl, M., Perrig, W. J. & Meier, B. The concurrent validity of the N-back task as a working memory measure. Memory 18, 394–412 (2010). source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    In working memory training studies, near transfer effects are easier to observe and larger than far transfer effects.

    “several meta-analyses of working memory training studies have demonstrated that near transfer effects are both more readily observable and show larger effect sizes than far transfer effects13,14.”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites [13] Soveri, A., Antfolk, J., Karlsson, L., Salo, B. & Laine, M. Working memory training revisited: a multi-level meta-analysis of n-back training studies. Psychon. Bull. Rev. 24, 1077–1096 (2017). source

      The Training effects by cognitive domain section (page unavailable) reports a much larger effect for untrained n-back tasks than for far-transfer domains, and the Main analyses call it the only notable transfer effect. But effects on other working-memory tasks were not significantly different from those on the other domains, and effects in all domains were statistically significant. The claim therefore overgeneralizes the near-versus-far contrast and its suggestion that far transfer is less readily observable. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
    • cites [14] Weicker, J., Villringer, A. & Thöne-Otto, A. Can impaired working memory functioning be improved by training? A meta-analysis with a special focus on brain injured patients. Neuropsychology 30, 190–212 (2016). source

  • May combine two claims.

    Premisecitation

    Previous studies typically compared low versus high training gain subgroups, and this responder analysis can be problematic.

    “Previous studies have examined the relationship between training gain and far transfer, typically by comparing subgroups of participants who show low versus high training gain31,38,39; however, this type of responder analysis can be problematic40.”
    Quote in the paperSubstantial claim2 of 4 sources checked
    • cites [31] Jaeggi, S. M., Buschkuehl, M., Jonides, J. & Shah, P. Short- and long-term benefits of cognitive training. Proc. Natl Acad. Sci. USA 108, 10081–10086 (2011). source
    • cites [38] Zinke, K. et al. Working memory training and transfer in older adults: effects of age, baseline performance, and training gains. Dev. Psychol. 50, 304–315 (2014). source
    • cites [39] Rudebeck, S. R., Bor, D., Ormond, A., O’Reilly, J. X. & Lee, A. C. H. A potential spatial working memory training task to improve both episodic memory and fluid intelligence. PLoS ONE 7, e50431 (2012). source
    • cites [40] Tidwell, J. W., Dougherty, M. R., Chrabaszcz, J. R., Thomas, R. P. & Mendoza, J. L. What counts as evidence for working memory training? Problems with correlated gains and dichotomization. Psychon. Bull. Rev. 21, 620–628 (2014). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Elevated COVID-19-related anxiety is associated with poorer WM performance on N-back tasks.

    “Elevated levels of COVID-19-related anxiety have also been associated with poorer WM performance as measured by N-back tasks56.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [56] Fellman, D., Ritakallio, L., Waris, O., Jylkkä, J. & Laine, M. Beginning of the pandemic: COVID-19-elicited anxiety as a predictor of working memory performance. Front. Psychol. 11, 576466 (2020). source

  • May combine two claims.

    Premisecitation

    In a meta-analysis of WM training studies, 35 experiments measured fluid intelligence, 28 assessed WM, but only 9 included untrained N-back tasks.

    “For example, in a meta-analysis of WM training studies13, 35 experiments measured fluid intelligence, 28 experiments assessed WM, including tasks that correlate poorly with the N-back24, but only 9 experiments also included untrained N-back tasks.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [13] Soveri, A., Antfolk, J., Karlsson, L., Salo, B. & Laine, M. Working memory training revisited: a multi-level meta-analysis of n-back training studies. Psychon. Bull. Rev. 24, 1077–1096 (2017). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    It is plausible that transfer occurs when training and transfer tasks engage overlapping processes and neural circuits.

    “Extending this to the neural level, it is plausible that transfer occurs if the training and transfer tasks engage overlapping processes and neural circuits26,51.”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites [26] Buschkuehl, M., Hernandez-Garcia, L., Jaeggi, S. M., Bernard, J. A. & Jonides, J. Neural effects of short-term training on working memory. Cogn. Affect. Behav. Neurosci. 14, 147–160 (2014). source
    • cites [51] Dahlin, E., Neely, A. S., Larsson, A., Bäckman, L. & Nyberg, L. Transfer of learning after updating training mediated by the striatum. Science 320, 1510–1512 (2008). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    A recent comparison of lab and remote videoconferencing cognitive assessment and training studies revealed similar compliance and performance rates.

    “a recent comparison of cognitive assessment and training studies conducted in the lab and remotely via videoconferencing revealed similar compliance and performance rates78.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [78] Collins, C. L. et al. Video-based remote administration of cognitive assessments and interventions: a comparison with in-lab administration. J. Cogn. Enhanc. https://doi.org/10.1007/s41465-022-00240-z (2022). source

      The Data Quality/Participant Performance section reports very similar training progress in the lab and remotely. But the Participant Compliance section reports that remote participants took significantly longer to complete training and were less compliant with the assigned schedule. The claim is supported for performance, not for similar compliance rates; page numbers were not available in our copy. (LLM, conf=0.92)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    UCancellation is a timed selective-attention and concentration measure with validation against the D2 Test of Attention.

    “A timed test of selective attention and concentration that has been validated against the D2 Test of Attention108,109.”
    Quote in the paperNot a substantial claim2 of 2 sources checked
    • cites [108] Conway, A. R. A. et al. Working memory span tasks: a methodological review and user’s guide. Psychon. Bull. Rev. 12, 769–786 (2005). source

      The VALIDITY section discusses working-memory span tasks and their relationship to attention, not UCancellation. Searches for UCancellation, the D2 Test of Attention, and cancellation-test validation found no passage addressing the claimed measure or its validation; page numbers were not available in our copy. (LLM, conf=0.75)

      Experimental — accuracy not yet measured
    • cites [109] Pahor, A. et al. UCancellation: a new mobile measure of selective attention and concentration. Behav. Res. Methods https://doi.org/10.3758/s13428-021-01765-5 (2022). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    In non-invasive brain stimulation and other intervention studies, non-responders are routinely excluded or treated differently.

    “in non-invasive brain stimulation studies84,85 or other intervention studies, such as those focusing on physical exercise or nutrition86–89, ‘non-responders’ are routinely excluded from data analyses or treated differently.”
    Quote in the paperNot a substantial claim3 of 6 sources checked
    • cites [84] López-Alonso, V., Cheeran, B., Río-Rodríguez, D. & Fernández-Del-Olmo, M. Inter-individual variability in response to non-invasive brain stimulation paradigms. Brain Stimul. 7, 372–380 (2014). source
    • cites [85] Bailey, N. W. et al. Responders to rTMS for depression show increased fronto-midline theta and theta connectivity compared to non-responders. Brain Stimul. 11, 190–203 (2018). source
    • cites [86] Reid, G. et al. Responders and non-responders to probiotic interventions: how can we improve the odds? Gut Microbes 1, 200–204 (2010). source
    • cites [87] Salas-Pérez, F. et al. Differentially methylated regions (DMRs) in PON3 gene between responders and non-responders to a weight loss dietary intervention: a new tool for precision management of obesity. Epigenetics https://doi.org/10.1080/15592294.2021.1873629 (2021). source

      The Study design section (page unavailable) says this dietary-intervention study separated responders from non-responders for analysis, supporting the claim that they were treated differently in this instance. It did not exclude non-responders: it excluded participants with poor dietary adherence and those in the intermediate response tertile. One study also cannot establish that this practice is routine, and this paper does not address brain stimulation. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
    • cites [88] Ostojic, S. M. Short-term GAA loading: responders versus nonresponders analysis. Food Sci. Nutr. 8, 4446–4448 (2020). source

      The Methods and Results (pages unavailable) classify participants in a GAA nutrition intervention as responders or nonresponders and compare the groups, supporting the claim that nonresponders were treated differently in this analysis. The paper does not establish that excluding or separately treating nonresponders is routine across intervention studies, and it does not report excluding them from analysis. (LLM, conf=0.82)

      Experimental — accuracy not yet measured
    • cites [89] Pickering, C. & Kiely, J. Do non-responders to exercise exist—and if so, what should we do about them? Sports Med. 49, 1–7 (2019). source

      The “Did Not Respond”—Potential Interventions section describes exercise non-responders receiving a modified training intervention, so it supports the possibility of treating them differently. It does not say that non-responders are routinely excluded from data analyses or routinely treated differently in intervention studies. This source addresses exercise, not the brain-stimulation studies cited separately in the sentence. (LLM, conf=0.82)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    There is some evidence that such transfer effects are long-term.

    “there is some evidence to suggest that such effects are long-term76,77.”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites [76] Ramani, G. B. et al. Racing dragons and remembering aliens: benefits of playing number and working memory games on kindergartners’ numerical knowledge. Dev. Sci. 23, e12908 (2020). source
    • cites [77] Zhang, Q. et al. The malleability of executive function in early childhood: effects of schooling and targeted training. Dev. Sci. 22, e12748 (2019). source

  • May combine two claims.

    Findingown data

    Across three randomized controlled trials with 460 participants, performance on untrained N-back tasks mediated transfer to Matrix Reasoning, even when trials 2 and 3 showed no intervention effect.

    “In all three RCTs, totaling 460 individuals, performance on untrained N-back tasks (near transfer) mediated transfer to Matrix Reasoning (representing far transfer) despite the lack of an intervention effect in RCTs 2 and 3.”
    • Also stated in: § Discussion · plain
      “all RCTs, involving a total of 460 participants, showed that the extent of transfer to untrained N-back tasks determines transfer to Matrix Reasoning.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The two N-back training groups showed small benefits on Matrix Reasoning (ES = 0.43-0.49), while control groups showed negligible change (ES < 0.20).

    “The two N-back training groups showed small performance benefits on Matrix Reasoning (ES = 0.43–0.49), whereas the control groups showed negligible change in performance (ES < 0.20).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Post-test N-back mediated the relationship between training group and post-test Countermanding in the whole sample and in RCT 2, but not in RCT 1.

    “This was found in the whole sample (IE = −0.23, SE = 0.09, 95% CI = −0.42 to −0.08) and in RCT 2 (IE = −0.28, SE = .12, 95% CI = −0.55 to −0.08), but there was no statistically significant evidence that post-test N-back mediated the relationship between training group and post-test Countermanding in RCT 1 (IE = −0.19, SE = 0.12, 95% CI = −0.46 to 0.01).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    No-contact control groups showed little or no improvement on Near Transfer, while N-back training groups showed a substantial increase (ES = 0.80-0.92).

    “no-contact control groups showed little or no improvement at post-test relative to pre-test on Near Transfer, whereas N-back training groups showed a substantial increase in accuracy on this task (ES = 0.80 – 0.92) (Table 2).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The total effect was demonstrated for the whole sample but separately only for RCT 1.

    “Note that the total effect was demonstrated for the whole sample but when examined separately for each of the two RCTs, only for RCT 1.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Bayesian repeated measures ANOVAs showed significant Time × Group interactions favoring N-back training for WM Composite 1, WM Composite 2, and the IC composite.

    “Statistically significant interactions in favor of the N-back training group were obtained for WM Composite 1 (BFincl > 100), WM Composite 2 (BFincl = 2.46), IC composite (BFincl = 9.59)”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The composite scores validly measure the constructs of working memory, inhibitory control, and fluid intelligence.

  • May combine two claims.

    Findingown data

    A significant total intervention effect on Matrix Reasoning was observed for the whole sample in the Retrospective study but not in the Replication study.

    “While a significant total (intervention) effect on Matrix Reasoning was observed for the whole sample in the Retrospective study, this was not observed in the Replication study (RCT 3).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The Replication study showed a significant intervention effect and significant mediation to the WM composite.

    “the Replication study showed a significant intervention effect as well as significant mediation to the WM composite”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The General knowledge training control does not engage fluid intelligence or WM.

  • May combine two claims.

    Findingown data

    Training curves for RCT 2 and RCT 3 are highly similar and both show slightly lower average training performance than RCT 1.

    “the training curves for RCT 2 and RCT 3 are highly similar, yet both show slightly lower average training performance compared to RCT 1 (Extended Data Fig. 2)”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The identified factors were: an inhibitory-control factor loaded by Countermanding and UCancellation; a reasoning factor loaded by UCMRT, ETS Surface Development, ETS Form Board, and DAT Space Relations; and a working-memory factor loaded by Near Transfer N-back, Sequencing, Simple Span, and Complex Span.

    “The following factors were identified: inhibitory control (Countermanding, UCancellation), reasoning ability (UCMRT, ETS Surface Development Test, ETS Form Board Test, DAT Space Relations), and working memory (Near Transfer N-back, Sequencing, Simple Span, and Complex Span)”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    Running Span correlated only with Sequencing.

    “except for Running Span, which correlated only with Sequencing”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    At pre-test, all test-battery measures correlated significantly with one another except Running Span.

    “Pre-test correlations of performance on the test battery revealed significant correlations among all measures (Supplementary Table 5) except for Running Span”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The PCA yielded three factors accounting for 58.5% of variance.

    “A three-factor solution was obtained, which explained 58.5% of the variance (Supplementary Table 9).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data

    Post-test performance on untrained N-back preceded Matrix Reasoning by one day, satisfying temporal precedence for the mediator.

    “Moreover, post-test performance on untrained N-back preceded performance on Matrix Reasoning by one day, hence satisfying the recommendation that the mediator must intervene between the predictor and the outcome37.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [37] Tate, C. U. On the overuse and misuse of mediation analysis: it may be a matter of timing. Basic Appl. Soc. Psych. 37, 235–246 (2015). source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Neuroimaging evidence indicates that the frontostriatal system mediates near transfer, while dorsolateral and ventrolateral prefrontal regions use domain-general networks that support broader learning.

    “neuroimaging research suggests that the frontostriatal system mediates near transfer effects whereas the dorso- and ventrolateral prefrontal cortices appear to employ domain-general networks that facilitate learning more wide5.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [5] Salmi, J., Nyberg, L. & Laine, M. Working memory training mostly engages general-purpose large-scale networks for learning. Neurosci. Biobehav. Rev. 93, 108–122 (2018). source

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationown data

    The paper demonstrates alternative ways of analyzing transfer data that avoid dichotomization and may reveal concealed associations and mechanisms.

    “Here we demonstrate that there are alternative ways of analyzing transfer data that do not involve dichotomization and may reveal concealed associations and mechanisms to explain the effects of training.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    These results imply that N-back training does not merely enhance task-specific strategies but may change underlying WM components such as focus of attention capacity or updating, which can support Matrix Reasoning.

    “These results imply that N-back training does not seem to merely enhance task-specific strategies, but may lead to changes in underlying components of WM such as the capacity of the focus of attention41 or updating42, which in turn can support performance on Matrix Reasoning problems43.”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [41] Chuderski, A. & Necka, E. The contribution of working memory to fluid reasoning: capacity, control, or both? J. Exp. Psychol. Learn. Mem. Cogn. 38, 1689–1710 (2012). source
    • cites [42] Friedman, N. P. et al. Not all executive functions are related to intelligence. Psychol. Sci. 17, 172–179 (2006). source
    • cites [43] Halford, G. S., Cowan, N. & Andrews, G. Separating cognitive capacity from knowledge: a new hypothesis. Trends Cogn. Sci. (Regul. Ed.) 11, 236–242 (2007). source
    • Rests on an assumption: The mediation models assume that there are no unmeasured confounders of the mediator–outcome relationship, so the estimated indirect effects can be interpreted causally.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    The present results align with research showing that updating and/or executive attention have strong relations with Matrix Reasoning performance.

    “Our results are therefore in line with research showing that updating42 and/or executive attention49 show strong relations with Matrix Reasoning performance.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [42] Friedman, N. P. et al. Not all executive functions are related to intelligence. Psychol. Sci. 17, 172–179 (2006). source
    • cites [49] Engle, R. W. Working memory and executive attention: a revisit. Perspect. Psychol. Sci. 13, 190–193 (2018). source

  • May combine two claims.

    Study design and conductown data

    The Replication study was preregistered, double-blind, and included an active control group.

    “we pre-registered a double-blind Replication study and included an active control group”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationreasoning

    Span-based WM tasks might not share enough variance with Matrix Reasoning to explain individual differences in transfer, or their gains might be too small to affect transfer.

    “Span-based WM tasks might not share enough variance with Matrix Reasoning to explain individual differences in transfer, or the gain on these tasks might not be meaningful enough to have an observable effect on transfer.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    The serial mediation models were not pre-registered and were added post-hoc.

    “Note that serial mediation models were not pre-registered and were added post-hoc.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The COVID-19 pandemic was associated with widespread emotional and mental distress and increased risk for psychiatric illness.

    “It is now well documented that the pandemic was associated with widespread emotional and mental distress and increased risk for psychiatric illness54,55.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [54] Twenge, J. M. & Joiner, T. E. Mental distress among U.S. adults during the COVID-19 pandemic. J. Clin. Psychol. 76, 2170–2182 (2020). source
    • cites [55] Pfefferbaum, B. & North, C. S. Mental health and the COVID-19 pandemic. N. Engl. J. Med. 383, 510–512 (2020). source

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    These factors may explain the lack of an intervention effect on Matrix Reasoning in the Replication study.

    “This may explain the lack of an intervention effect on Matrix Reasoning in the Replication study (RCT 3).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The young-adult college student sample experiences sleep deprivation, stress, anxiety, and disengagement as task novelty wears off.
    • Rests on an assumption: Remote testing in the Replication study likely produced more distractions than a controlled laboratory setting.

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    The Replication study was conducted remotely with participants completing sessions at home during COVID-19 social distancing.

    “the pre-registered Replication study was conducted remotely with participants completing sessions in their home environments during social distancing amid the COVID-19 pandemic.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    In the sample, limited personal space, home responsibilities, and distractions may have increased the proportion of participants motivated mainly by payment and not performing their best.

    “In our population it was typical for students to have limited personal space, and numerous home responsibilities and distractions which may have increased the proportion of participants that were motivated to complete the study so that they would receive payment, but perhaps were not motivated or able to perform their best on the assigned tasks.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The young-adult college student sample experiences sleep deprivation, stress, anxiety, and disengagement as task novelty wears off.
    • Rests on an assumption: Remote testing in the Replication study likely produced more distractions than a controlled laboratory setting.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Recent evidence indicates that the relationship between WM capacity and Matrix Reasoning can be explained by individuals' ability to maintain how they solved previous problems in memory.

    “Recent evidence also challenges these accounts, showing that the relationship between WM capacity and Matrix Reasoning can be explained by individuals’ ability to maintain how they solved previous problems in memory48.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [48] Harrison, T. L., Shipstead, Z. & Engle, R. W. Why is working memory capacity related to matrix reasoning tasks? Mem. Cogn. 43, 389–396 (2015). source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductreasoning

    Participants in the Replication study completed the study on their own devices at home and were instructed to find a quiet environment, but more distractions likely occurred than in a lab.

    “Participants completed the study on their own devices at home and were instructed to find a quiet environment for each session; however, it is likely that more distractions occurred compared to a controlled laboratory setting.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Remote testing in the Replication study likely produced more distractions than a controlled laboratory setting.

  • Classified as probably not a substantial claim.

    Study design and conductown data

    Participants in the Replication study were randomly assigned to one of three groups.

    “Participants were randomly assigned to one of three groups: Gamified N-back training, Non-gamified N-back training, and General knowledge training (control group)”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Alternate versions of Near Transfer N-back, Running Span, and Matrix Reasoning were administered at pre- and post-test in counterbalanced order.

    “Alternate versions of Near Transfer N-back, Running Span and Matrix Reasoning were administered at pre- and post-test in a counterbalanced order (A-B, B-A).”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Pre-test performance on the three tasks was entered as covariates to control for pre-existing individual differences in cognitive ability.

    “To control for pre-existing individual differences in cognitive ability, we entered performance for the three aforementioned tasks as covariates.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductown data

    Participants in the Retrospective study were randomly assigned to N-back training or no-contact control.

    “Participants were randomly assigned either to N-back training (N = 191) or to no-contact control (N = 92)”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Inter-individual differences in responsiveness to WM training interventions exist.

    “it is not surprising that inter-individual differences in responsiveness to WM training interventions exist82,83”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [82] Katz, B. et al. in Cognitive Training (eds Strobach, T. & Karbach, J.) 157–166 (Springer International, 2016); https://doi.org/10.1007/978-3-319-42662-4_15 source
    • cites [83] Jaeggi, S. M., Buschkuehl, M., Shah, P. & Jonides, J. The role of individual differences in cognitive training and transfer. Mem. Cogn. 42, 464–480 (2014). source

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    All variables were z-transformed based on pre-test scores separately for each RCT to control for RCT effects.

    “All variables were z-transformed (based on pre-test scores, separately for each RCT) prior to regression-based analyses to control for RCT effects.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    Supervision differed across studies: all training sessions were supervised in the Retrospective study, while only the first training session was supervised in the Replication study (or more if needed).

    “The level of supervision was different across studies given that all training sessions were supervised in the Retrospective study and only the first training session was supervised in the Replication study (or more, if needed).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationown data

    Other factors than supervision, such as small changes in game mechanics or adaptive algorithm after RCT 1, may have affected these results.

    “it is possible that other factors than the extent of supervision, such as small changes in the game mechanic or adoptive algorithm implemented after RCT 1 had concluded, affected these results.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data

    The sample in all three RCTs was young, limiting generalizability to populations with diverse cognitive needs.

    “Another limitation is the young age of the sample in all three RCTs. Therefore, the extent to which the current results can be generalized to the many studies examining the effects of cognitive training in populations with diverse cognitive needs remains to be clarified.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    Six outliers were removed based on Near Transfer and Matrix Reasoning performance, and six additional N-back training participants were removed due to software error.

    “Six outliers were removed from the sample based on performance on Near Transfer and Matrix Reasoning (|zaccuracy| > 3 at pre- or post-test separately for each RCT). In the N-back training group, 6 additional participants were removed due to a software error in the training program.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    Due to COVID-19, the Replication study could not be conducted in the lab and did not completely match the Retrospective study conditions.

    “due to the COVID-19 pandemic, we were not able to conduct it in the lab, therefore we were not able to completely match the conditions of the Retrospective study (RCT 1 and RCT 2).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    Multicollinearity, heteroscedasticity, and independent errors were checked and found in appropriate ranges.

    “We checked for multicollinearity, heteroscedasticity, and independent errors using a Durbin-Watson test100, and found these to be in appropriate ranges.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [100] Field, A. in Discovering Statistics Using IBM SPSS Statistics (ed. Carmichael, M.) 293–356 (Sage, 2013).

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    All assessment sessions in the Replication study were supervised via video conferencing by a researcher blind to training condition.

    “All assessment sessions were supervised via video conferencing by a researcher who was blind to the training condition the participant had been assigned to.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Simple Span measures spatial working memory storage, whereas Complex Span measures working memory storage together with processing.

    “Simple Span was used to assess spatial WM storage while Complex Span was used to measure WM storage-and-processing106.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [106] Broadway, J. M. & Engle, R. W. Validating running memory span: measurement of working memory capacity and links with fluid intelligence. Behav. Res. Methods 42, 563–570 (2010). source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductreasoning

    Different experimenters supervised pre/post-test versus training sessions to ensure experimenter blinding, and participants were unaware of training conditions.

    “Different experimenters supervised pre/post-test sessions versus training sessions to ensure experimenter blinding; likewise, participants were not aware that there were multiple training conditions and which one they had been assigned to.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Target sample size was 64 per condition based on power analysis for two-mediator models using Monte Carlo simulations in MPLUS.

    “with a target sample size of 64 participants per condition as determined by a power analysis for two-mediator models using Monte Carlo simulations in MPLUS.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    These results can inform studies of neuroplastic changes, predicting weaker far transfer in individuals with little to no near transfer.

    “these results can inform studies examining neuroplastic changes associated with WM interventions, predicting weaker far transfer effects in individuals who show little to no near transfer, both in terms of behavior and related plastic neural changes.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Mann-Whitney tests showed no significant post-test performance differences by test version in either group.

    “Mann-Whitney tests indicated that there were no significant performance differences at post-test as a function of test version (different versus the same) in the no-contact (U = 40, z = −.54, p = .88) and N-back training (U = 624, z = −1.01, p = .27) groups.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    A relevant contribution is that near transfer after WM training mediates far transfer, and this is not challenged by the null Gf composite effect.

    “a relevant contribution of the paper is that near transfer after WM training mediates far transfer, which is not challenged by the fact that the Gf composite did not show an effect.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationown data

    Training-related improvement in inhibitory control could be a separate form of far transfer, with processes somewhat independent.

    “training-related improvement in inhibitory control could be seen as a separate form of (far) transfer, and the processes that underly this change seem to be somewhat independent.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductreasoning

    In the Replication study, Near Transfer performance preceded post-test Matrix Reasoning and other Gf tasks by 1 day, making it a plausible causal agent.

    “in the Replication study, Near Transfer performance preceded post-test performance on Matrix Reasoning and other Gf tasks by 1 day, hence it represents a plausible causal agent for these outcome measures”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisereasoning

    Matrix Reasoning and the WM composite are considered far transfer because they are substantially different from the training task.

    “We consider both, the Matrix Reasoning task, and the WM composite as far transfer because they consist of tasks that are substantially different from the training task.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Six of the 201 study completers were excluded for ETS and DAT administration errors.

    “Out of 201 participants who completed the study, 6 participants were removed due to administration error on ETS and DAT tests.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    Near Transfer N-back versions were counterbalanced across individuals and sessions, except 40% of RCT1 participants completed the same version twice due to technical error.

    “The versions were counterbalanced across individuals and sessions, except for 40% of participants in RCT 1, who completed the same version twice due to a technical error.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    Alternate versions of ETS Surface Development, ETS Form Board, and DAT Space Relations were grouped and counterbalanced across participants.

    “Alternate versions51 of ETS Surface Development Test, ETS Form Board Test, and DAT Space Relations were grouped (As and Bs together) for administration at pre- and post-test, the order of which was counterbalanced across participants.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [51] Dahlin, E., Neely, A. S., Larsson, A., Bäckman, L. & Nyberg, L. Transfer of learning after updating training mediated by the striatum. Science 320, 1510–1512 (2008). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    In the Retrospective study, multiple participants completed sessions in the same room, so that environment was not distraction-free.

    “in the Retrospective study, multiple participants completed assessment and training sessions in the same room, thus, even this environment wasn’t distraction-free.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The Replication study was preregistered on AsPredicted (#40240).

    “The experiment was preregistered on AsPredicted (#40240)”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    A data point was treated as an outlier when it fell at least three SDs from the mean at pre-test or post-test.

    “Outliers were removed if a given data point was 3 or more standard deviations from the mean at pre- or post-test.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationown data

    Every measure had common variance with at least some other measures.

    “thereby confirming that each measure shared some common variance with other measures”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductown data

    For the PCA, sampling adequacy was supported (KMO = .79), Bartlett's sphericity test was significant (χ2(45) = 387.55, p < .001), all antiimage diagonals exceeded .7, and all communalities exceeded .3.

    “The sample size for PCA was adequate (KMO = .79), Bartlett’s test of sphericity was significant (χ2 (45) = 387.55, p < .001), the diagonals of the antiimage correlation matrix were all over .7, and the communalities were all above .3”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    After outlier removal, the sample size was 189, with 128 in N-back training and 61 in active control.

    “hence the resulting sample size was 189 (128 N-back training, 61 Active control) (Table 3)”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The second analytical dataset contained 170 participants, 115 in N-back training and 55 in active control.

    “This second analytical dataset, which consisted of 170 participants (115 N-back training, 55 Active control) was used for the additional mediation analyses”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Eleven participants were excluded for failing to comply with the Sequencing and Spatial Span tasks.

    “11 participants who showed non-compliance on Sequencing and Spatial Span tasks”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductown data

    Running Span was excluded from subsequent analyses.

    “hence it was excluded from subsequent analyses”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    Near Transfer and Matrix Reasoning had no outliers removed.

    “No outliers were removed for Near Transfer or Matrix Reasoning”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Six outliers were removed because of Countermanding speed.

    “6 outliers were removed based on Countermanding speed”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Additional outliers were removed for Running Span, eight in total.

    “additional outliers were removed for Running Span (N=8)”
    Quote in the paperNot a substantial claim
76 of 76

  • The paper relies on this without support.

    The mediation models assume that there are no unmeasured confounders of the mediator–outcome relationship, so the estimated indirect effects can be interpreted causally.

    “These results suggest that the causal effect of training group (No-contact control/N-back training) on Matrix Reasoning is mediated by the extent to which participants improve on near transfer.”

  • The paper relies on this without support.

    The composite scores validly measure the constructs of working memory, inhibitory control, and fluid intelligence.

    “we expanded the test battery to extend the analyses to other outcome measures and to be able to generate composite scores for each domain.”

  • The paper relies on this without support.

    The General knowledge training control does not engage fluid intelligence or WM.

    “This task was selected for the control intervention because it taps into crystallized intelligence and should not engage fluid intelligence or WM.”

  • The paper relies on this without support.

    UCMRT taps mainly into abstract reasoning and logical problem solving.

    “While UCMRT requires some spatial visualization, it is thought to tap mainly into abstract reasoning and logical problem solving.”

  • The paper relies on this without support.

    The young-adult college student sample experiences sleep deprivation, stress, anxiety, and disengagement as task novelty wears off.

    “Based on subjective observation of the young-adult college student population that is represented in the sample, we note that they are at times sleep deprived, experience stress and anxiety, and adopt strategies of disengagement as the novelty of tasks wears off.”

  • The paper relies on this without support.

    Remote testing in the Replication study likely produced more distractions than a controlled laboratory setting.

    “it is likely that more distractions occurred compared to a controlled laboratory setting.”
6 of 6

Statistics that don’t add up

Each reported p-value recomputed from its test statistic and degrees of freedom.

  • Inconsistent p-value: z = −.54, p = .88

    Minor

    The paper reports p = .88 here; recomputing from the test statistic and its degrees of freedom gives p = .589 (two-tailed). The check cannot tell whether the statistic or the p-value is the misprint.

    Reported p = .88 · recomputed p = .589

1 of 1

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Doesn’t add upz = -0.54reported p = .88 · recomputed p = .589

    “z = −.54, p = .88”

  • Consistentz = -1.01reported p = .27 · recomputed p = .312

    “z = −1.01, p = .27”

    The paper states an exact rank test

  • Consistentχ²(45) = 387.55reported p < .001 · recomputed p < .001

    “χ2 (45) = 387.55, p < .001”

3 of 3

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Self-check results appear here once an independent check has confirmed them.

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Pahor, A., Seitz, A. R., & Jaeggi, S. M.
Affiliations
  • University of California, Riverside (US)
  • University of California, Irvine (US)
Corresponding author
Pahor, Anja
ORCID
Venue
Nature Human Behaviour
Publisher
Springer Science and Business Media LLC
Published
2022-06-20
Volume
6
Issue
9
Pages
1243-1256
ISSN
2397-3374 (ISSN-L)
DOI
10.1038/s41562-022-01384-w
PMC
PMC12305750
OpenAlex
W4283163023
Cited by
34
References
106
Citations by year
2026: 6 · 2025: 8 · 2024: 12 · 2023: 6 · 2022: 1
FWCI
6.20
Type
journal article
Language
en
Update notices
correction (2023-01-31)
Open access
Licence
Locations
  • Nature Human Behaviour (publishedVersion)
  • PubMed (publishedVersion)
  • PubMed Central (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
Funders
  • U.S. Department of Health & Human Services | NIH | National Institute of Mental Health (R01MH111742)
  • U.S. Department of Health & Human Services | NIH | National Institute on Aging (1K02AG054665)
Topics
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Cognitive Abilities and Testing
  • Physical Sciences › Mathematics › Statistics and Probability › Cognitive and developmental aspects of mathematical skills
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Cognitive Functions and Memory
Keywords
working memory training, n-back task, far transfer, near transfer, matrix reasoning, mediation analysis, individual differences, cognitive training
MeSH terms
Humans, Learning, Memory, Short-Term, Problem Solving

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-08.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-08.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review