Page 7 of 12 · Pillar three

Connect

Link ideas fast. This is the pillar where the claims in the wider literature are boldest and the replications are thinnest, so it is also the pillar where we are most careful about what we say.

As the research review explains, some small relational-training studies reported nonverbal IQ gains, but the evidence has not established a reliable intelligence benefit. Time Gates’ puzzles practise working out relationships directly. Their answers can be checked against the facts in each puzzle.

Relations

Moderate for the drilled skill

"A is more than B. B is the same as C. Is C less than A?" Ten seconds to answer.

Parameters
20 trials per block, two to three blocks. The ladder runs from one relation with same and opposite, to two relations with more and less, to three or more mixed with nonsense syllables in place of real words, to reversed questions. Promotion at 85% across two blocks, demotion below 60%.
Scored
Correct, with a speed bonus under five seconds.
Why this shape
The service generates a relation graph and derives the answer by transitivity, so the key is computed rather than asserted. The nonsense-syllable rung matters: it removes the option of answering from world knowledge instead of from the stated relations.

Analogies with traps

Moderate

"branch : tree :: tributary : ?" with four options, one of which shares a word with the stem but has the wrong relation.

Parameters
15 items, 20 seconds each; the ladder runs from part-whole and category, through function and cause, to abstract relations such as opposite and degree.
Scored
Correct, plus which trap you fell for, which is the more interesting datum.
Why this shape
A 2021 systematic review of reasoning training in children and adolescents concluded it reliably improves fluid reasoning with "less conclusive" evidence for academic gains. The trap options are not decoration: analogy performance depends on inhibiting the surface match as much as on finding the deep relation, which is precisely the overlap between this pillar and Pause.

Three words, one link

Strong as a measureWeak as training

Three cue words - cottage, swiss, cake - and one word that joins all three.

Parameters
12 items, 30 seconds each, in three difficulty bands set by published solution rates: above 70% solved at 30 seconds, 40–70%, and below 40%.
Scored
Exact match against the published key, with half credit for a listed alternative.
Why this shape
There is a normed set of 144 English items with solution rates and times, which makes this the most defensible measurement task in the pillar. Performance correlates around .40 to .50 with insight problem solving. The one training study is a ten-day practice-effect result in college students, so training gets the weaker grade. New items generated for the product are accepted only if every compound is a real dictionary word.

Category sprint

Strong as a measureNo training evidence

Name as many things in a kitchen as you can, in sixty seconds.

Parameters
Two categories per session, rotating through at least twenty so none repeats within four weeks.
Scored
Count of valid unique words, with a bonus for switching between sub-clusters rather than exhausting one.
Why this shape
Total-word reliability is good - an intraclass correlation above .80 over three to four weeks in young people - and clustering and switching sub-scores are moderate. But no study shows that fluency training transfers to anything; scores rise with repetition, which is why alternate categories are mandatory at retest. We use it as a warm-up and a measure and say so.

Word chain

Weak, untested

Get from "ice" to "parliament" in six association steps or fewer.

Parameters
Five chains, 90 seconds each; the allowed step count shrinks and the semantic distance between endpoints grows as skill rises.
Scored
Steps used, and plausibility of each step checked against a free-association dataset.
Why this shape
No intervention studies exist. It is an engaging wrapper for association retrieval that happens to be checkable against real data, which makes it a good delivery format and not a claimed treatment.

Other uses

Moderate for the trained task

List uses for a brick. Then list uses that must involve water.

Parameters
Two prompts, two minutes each. The constraint on the second prompt exists to force flexibility rather than fluency.
Scored
Fluency and originality reported separately, with originality scoring calibrated once against a hand-rated sample.
Why this shape
Five decades of meta-analyses give creativity training moderate-to-large effects, largest on originality and largest when the training teaches explicit strategies rather than exhorting people to be creative. Transfer to everyday flexibility is unproven. Automated scoring of this task now reaches about r = .81 against human raters, which is what makes it practical at home.

Build a map

Strong for learningUntested for association

Eight to twelve concepts from a short passage; draw the links and label every one.

Parameters
Eight minutes; fewer concepts are given as skill rises, so you have to supply your own.
Scored
Correct labelled links against a reference map, with unusual but valid links flagged rather than marked wrong.
Why this shape
A meta-analysis of 142 effect sizes across 11,814 students found learning with concept maps at g = 0.58, and constructing your own map (g = 0.72) beating studying a ready-made one (g = 0.43). That is evidence about learning content, not about general associative ability, and the two grades above keep those apart.
Next: the Recall exercises →

Version 133 · Privacy · Terms of use · Appearance