Science / What we mean by transfer
Does brain training work?
Increasing general intelligence is Synergy IQ’s design purpose. This page explains how we would test that aim: from improvement on a trained task, through independent cognitive measures, to broader intelligence and everyday outcomes. The ambition is clear; the evidence must show how far the gains travel.
That question is called transfer. This page defines the different kinds of transfer we care about, reviews the evidence behind them, and separates established findings from the hypotheses Synergy IQ is built to test.
For AI-supported tasks, we keep another distinction explicit: assisted performance is not the same outcome as what a person can still understand, verify or do independently.
From task practice to construct gain
Near / construct far transfer
Ignore irrelevant alerts
Keep the goal after interruption
Find the relevant signal in dense information
Evidence burden
rises as the claim broadens
The wider research programme
What is established — and what are we proposing?
IQ Mindware's newer adaptive-intelligence vision combines established literatures with emerging evidence and original hypotheses. Keeping those levels separate matters more as the framework expands beyond training tasks into digital environments, social platforms, AI use and real-world agency.
The updated vision does not treat commitment scarcity as a universal human condition. Headroom, cognitive optionality and reality-responsive commitment are organising ideas for research and product design, not validated faculties or a single score. External constraints such as time, resources, support and decision rights must remain distinct from cognitive performance.
For AI use, assisted performance and retained reasoning are different outcomes. IQ Mindware does not assume that AI generally weakens reasoning; the research question is what people understand, judge, verify and can recover when assistance changes or disappears.
Established / substantial evidence base
Psychometric g, Gf and Gc; attention and cognitive control; cognitive offloading; goal pursuit and implementation intentions; situated cognition; and the difficulty of achieving broad transfer from narrow training.
Emerging evidence
Human–AI learning effects; reasoning offload and reliance; supported performance versus retained capability; AI-assisted option generation and decision processes; and the opportunities and costs of algorithmically mediated attention and social learning.
Theoretical synthesis
The high-velocity digital cognitive niche, the proposed Commitment Economy, strategic path entropy and the Generative Cognitive Niche connect older attention, learning and situated-cognition problems to today's platform and AI environment.
IQ Mindware hypotheses
Search-Reopening Asymmetry, Strategic Fluency Illusion, a measurable Simulation–Enactment Gap, Adaptive Path Control, reality-anchored G-loop effects and Asymmetrical Mutualism remain hypotheses to test.
Important: these newer labels are not being presented as established cognitive biases or validated psychological faculties. They organise falsifiable questions for research and product design.
A transfer vocabulary
Six claims that should stay separate.
Transfer is not one yes-or-no outcome. Barnett and Ceci's taxonomy emphasised that distance can vary across task, knowledge, context, time and other dimensions. Synergy IQ therefore labels each level rather than using one broad claim.
Task learning
Performance improves on the exact exercise or backbone task that was practised. This is evidence of learning, but not yet transfer.
Wrapper transfer
The same core operation is required while the display, content, sensory format or emotional wrapper changes. This tests whether performance is tied to surface features.
Construct / capacity transfer
The same underlying capacity improves beyond the trained exercise — for example, attention control on an independently designed attention task or in a different situation that still requires selecting relevant information and resisting distraction.
Network transfer
Training one node changes the later performance or learning trajectory of another node. For example, stronger attention control might alter how efficiently relational memory can subsequently be learned.
Niche transfer
The construct is not merely present in a real-life situation; its use produces a measurable functional difference in the relevant setting, such as fewer missed signals, better recovery after interruption or improved task quality.
Far / broad transfer
Change appears on a more distant outcome such as general reasoning, an intelligence measure or real-world functioning. These claims require independent measures and stronger designs.
Terminology note: “network transfer” and “vertical transfer” below are Synergy IQ research terms. They describe testable product hypotheses rather than established categories in the psychological transfer literature.
1. Backbone
Learn the core operation with clear feedback.
Learn the core operation with clear feedback.
↓
new visual formnew content / modalitynew emotional or contextual wrapper
↓
2. Return
Come back to the backbone and test recovery.
Come back to the backbone and test recovery.
↓
3. Held-out task
Test the target capacity with material that was not trained.
Test the target capacity with material that was not trained.
Principle 01 / Construct transfer
Train the ability — not one display.
Construct-level near transfer is generally more credible than broad far transfer, but even this kind of transfer is selective. Meta-analyses of working-memory training have found reliable improvement on related memory tasks while broader intelligence or real-world outcomes are much less consistent. A 2026 meta-analysis of 55 studies found a small overall cross-paradigm working-memory transfer effect (g = .146), with stronger updating-to-span transfer than span-to-updating transfer.
At the same time, other work shows why this is difficult. Gathercole and colleagues found that transfer was greatest when trained and untrained tasks shared cognitive routines or paradigms. Himi and colleagues varied both working-memory operation and material domain and found no convincing transfer across different materials, even when the nominal operation was the same.
This is why Synergy IQ changes wrappers. A capacity is challenged through different surface forms and then returned to the backbone. The purpose is to reduce the chance that fluency with one visual or procedural routine is mistaken for a capacity gain.
Scientific boundary: using multiple modalities or wrappers is a principled way to test whether learning survives representational change. The current literature does not justify claiming that multimodal training automatically produces construct transfer. Synergy IQ must demonstrate that directly.
CAPACITYCan the cognitive operation be performed under controlled demand?
↓
STRATEGYCan the person recognise when the capacity matters and recover an explicit policy for using it?
↓
NICHECan the capacity and strategy be deployed in the real context, with its cues, tools, incentives and constraints?
Principle 02 / Applied transfer
Capacity → Strategy → Niche.
Synergy IQ uses vertical transfer as shorthand for moving from a trained cognitive operation, to an explicit and accessible strategy, to use in the cognitive niche where it matters. This is not a standard psychometric transfer category; it is an integration of several established learning principles.
Capacity provides the controlled cognitive substrate. Strategy makes the policy explicit: what cue should trigger it, what should I do, and how do I know whether it worked? Niche practice then places that policy into a realistic context rather than assuming that an app-trained skill will spontaneously appear at work, in study or during AI-assisted reasoning.
The supporting literature is broader than cognitive-training research. Meta-analyses find that metacognitive strategy instruction can produce durable academic effects; implementation-intention studies show that explicit cue–action plans can increase use of trained behaviours; and workplace training research shows that peer, supervisor and organisational support predict whether newly learned skills are actually transferred and sustained.
What this supports: strategy accessibility and context matter for deployment. What remains to be tested: whether combining capacity training + strategy instruction + niche practice produces a synergistic effect greater than the components alone.
Principle 03 / Trident-G theory
Construct training builds Gc.
Network training tests Gf.
The Gf → Gc → Gf spiral begins when fluid intelligence searches a novel problem. Structures that survive variation, challenge and delayed re-checking can become reusable Gc: deeply learned cognitive skills, relational structures and control policies, rather than factual knowledge alone.
A focused node is designed for this construct-training route. It aims to make one capacity more robust and portable. The connected network then brings multiple portable capacities back into novel, combined demands. That is the proposed network-training route to Gf: more flexible learning, relational search and problem solving from a richer starting point.
Scientific boundary: Trident-G specifies a theory and testable training architecture. It does not establish that Synergy IQ raises fluid intelligence, IQ or general intelligence. Those claims require independent outcomes, comparison conditions and longitudinal evidence.
Read the Trident G theory preprint by Mark Ashton Smith (2025)
Vocabulary &
knowledge Reasoning Working
memory Processing
speed Visuospatial
ability
knowledge Reasoning Working
memory Processing
speed Visuospatial
ability
Principle 04 / Network transfer
Abilities may help build one another.
Mutualism proposes that the positive correlations among cognitive abilities can emerge partly because abilities interact during development. Van der Maas and colleagues formalised this idea in 2006. Longitudinal studies later found reciprocal coupling between vocabulary and reasoning: higher vocabulary predicted subsequent reasoning gains, while higher reasoning also predicted vocabulary gains. The pattern was reported in adolescents and young adults and replicated in younger children.
Synergy IQ extends this into an asymmetrical mutualism hypothesis. The links need not be equal. Some capacities may exert more downstream leverage than others, and the person's day-to-day cognitive niche may strengthen or constrain those effects.
What the evidence supports: developmental coupling between cognitive abilities is plausible and has longitudinal support. What it does not yet show: that training a Synergy IQ node causes another node to improve, or that the seven product nodes have known causal connection strengths.
What would count as network transfer? A stronger design would train Node A, then test whether learning rate, ceiling, recovery or delayed performance on independently measured Node B changes relative to a suitable comparison condition. Correlated app scores alone would not establish it.
The Synergy IQ evidence ladder
The training score is only the first rung.
A credible transfer architecture should become harder to satisfy as the claim becomes broader. Synergy IQ therefore separates acquisition, wrapper robustness, independent construct measures, cross-node effects and deployment.
Learn
Show a genuine learning curve on the backbone rather than one lucky score.
Challenge + return
Change the wrapper, observe disruption and recovery, then return to the backbone.
Held-out construct test
Use an untrained measure intended to sample the same capacity without reusing the game score.
Cross-node test
Ask whether training one node changes learning or performance elsewhere in the network.
Deploy + delay
Check whether a strategy appears in the relevant niche and whether gains survive a later re-check.
G Track remains a separate measurement layer. Separate measurement reduces circularity, but it does not remove practice effects or by itself establish causal transfer.
Calibration / What would change our mind?
A theory should be able to lose.
Synergy IQ is built around testable claims. The useful question is not whether a story sounds plausible, but what observations would strengthen it, weaken it or force the model to change.
Evidence that would strengthen the model
- Backbone gains that survive wrapper change and delayed re-check.
- Improvement on independently designed measures of the same construct.
- Cross-node changes in learning rate, ceiling or recovery relative to suitable comparisons.
- Real-world effects that track the trained capacity and strategy rather than mere app engagement.
Evidence that would weaken or revise it
- Large game gains with no reliable held-out construct change.
- Wrapper performance that depends on one practised surface or routine.
- No cross-node effect when the mutualist hypothesis predicts one.
- Independent evidence showing that a simpler explanation accounts for the observed change.
Source independence matters. Repeating one originating study across the website, newsletter, social media or AI summaries does not create multiple independent confirmations.
Claims boundary
What the science supports — and what it does not yet.
The strongest scientific position is not “brain training works” or “brain training does not work”. It is to state exactly what changed, how far the test moved from training, and what alternative explanations remain.
Reasonable foundations
- Cognitive abilities can show reciprocal developmental coupling.
- Task learning is common and near transfer can occur under some conditions.
- Transfer can be asymmetric rather than equally likely in both directions.
- Metacognitive strategy instruction can support durable learning.
- Context and social/organisational support matter for real-world training transfer.
Synergy IQ must still demonstrate
- That wrapper variation improves construct transfer compared with a simpler training regime.
- That gains generalise across modalities rather than reflecting learned task routines.
- That one trained node alters another node's subsequent learning trajectory.
- That Capacity + Strategy + Niche produces added value beyond its components.
- That the programme raises IQ or general intelligence. This has not been established.
References
Key sources.
Barnett, S. M., & Ceci, S. J. (2002). When and where do we apply what we learn? A taxonomy for far transfer. Psychological Bulletin, 128(4), 612–637. https://doi.org/10.1037/0033-2909.128.4.612
Blume, B. D., Ford, J. K., Baldwin, T. T., & Huang, J. L. (2010). Transfer of training: A meta-analytic review. Journal of Management, 36(4), 1065–1105. https://doi.org/10.1177/0149206309352880
de Boer, H., Donker, A. S., Kostons, D. D. N. M., & van der Werf, G. P. C. (2018). Long-term effects of metacognitive strategy instruction on student academic performance: A meta-analysis. Educational Research Review, 24, 98–115. https://doi.org/10.1016/j.edurev.2018.03.002
Friedman, S., & Ronen, S. (2015). The effect of implementation intentions on transfer of training. European Journal of Social Psychology, 45(4), 409–416. https://doi.org/10.1002/ejsp.2114
Fu, J., Zhou, Y., Yuan, M., & Cai, Y. (2026). The asymmetry of working memory training transfer: A systematic review and meta-analysis. Psychonomic Bulletin & Review, 33, Article 227. https://doi.org/10.3758/s13423-026-02994-5
Gathercole, S. E., Dunning, D. L., Holmes, J., & Norris, D. (2019). Working memory training involves learning new skills. Journal of Memory and Language, 105, 19–42. https://doi.org/10.1016/j.jml.2018.10.003
Gorman, T. E., & Goldstone, R. L. (2022). An instance-based model account of the benefits of varied practice in visuomotor skill. Cognitive Psychology, 137, 101491. https://doi.org/10.1016/j.cogpsych.2022.101491
Himi, S. A., Stadler, M., von Bastian, C. C., Bühner, M., & Hilbert, S. (2023). Limits of near transfer: Content- and operation-specific effects of working memory training. Journal of Experimental Psychology: General, 152(5), 1305–1333. https://doi.org/10.1037/xge0001328
Hughes, A. M., Zajac, S., Woods, A. L., & Salas, E. (2020). The role of work environment in training sustainment: A meta-analysis. Human Factors, 62(1), 166–183. https://doi.org/10.1177/0018720819845988
Kievit, R. A., Hofman, A. D., & Nation, K. (2019). Mutualistic coupling between vocabulary and reasoning in young children: A replication and extension of the study by Kievit et al. (2017). Psychological Science, 30(8), 1245–1252. https://doi.org/10.1177/0956797619841265
Kievit, R. A., Lindenberger, U., Goodyer, I. M., Jones, P. B., Fonagy, P., Bullmore, E. T., & Dolan, R. J. (2017). Mutualistic coupling between vocabulary and reasoning supports cognitive development during late adolescence and early adulthood. Psychological Science, 28(10), 1419–1431. https://doi.org/10.1177/0956797617710785
Melby-Lervåg, M., Redick, T. S., & Hulme, C. (2016). Working memory training does not improve performance on measures of intelligence or other measures of far transfer: Evidence from a meta-analytic review. Perspectives on Psychological Science, 11(4), 512–534. https://doi.org/10.1177/1745691616635612
Schwaighofer, M., Fischer, F., & Bühner, M. (2015). Does working memory training transfer? A meta-analysis including training conditions as moderators. Educational Psychologist, 50(2), 138–166. https://doi.org/10.1080/00461520.2015.1036274
van der Maas, H. L. J., Dolan, C. V., Grasman, R. P. P. P., Wicherts, J. M., Huizenga, H. M., & Raijmakers, M. E. J. (2006). A dynamical model of general intelligence: The positive manifold of intelligence by mutualism. Psychological Review, 113(4), 842–861. https://doi.org/10.1037/0033-295X.113.4.842
Smith, M. A. (2025). Trident G: A Computational-Level Theory of Intelligent Life [Preprint]. ResearchGate. ResearchGate publication page. https://doi.org/10.13140/RG.2.2.24468.97925
Human intelligence and AI
What does AI use leave you able to do?
Our aim is to use AI in ways that develop and retain human intelligence. Helpful assisted output, independent learning and IQ change are different outcomes.
A survey of knowledge workers found that confidence in generative AI was associated with self-reported critical-thinking effort. It did not measure a change in IQ or establish cognitive decline. This supports asking how people use the tool, rather than assuming that AI use itself improves or harms intelligence.
Lee and colleagues (2025): study and limitations →
The proposed Synergy IQ AI exercises use initial human attempts, hints, challenges and later independent checks. Whether these methods support gains in general intelligence needs direct testing. Separate assessments help investigate progress; controlled comparisons are needed to attribute change to training or AI use.