The design choice worth dwelling on is what sat in the control arm. In 54 of the trials the comparison students got no ethics lessons at all [4], and the authors are explicit that this measures how much moral reasoning students acquire anyway, from parents and from ordinary schooling [5]. So the reported gain is an increment on top of a childhood, not a rescue of children who had none. That is the harder test, and it is the one a curriculum authority is actually being asked to rule on.
Now the arithmetic. Across 92 trials and 12,693 students [2], the average trial enrolled about 138 students [19], which is two or three classes. Trials that size measure a reasoning outcome noisily, so the pooling is doing most of the work. The remaining 38 trials compared one kind of ethics instruction against another [20], and that is what gives the delivery finding its footing: a randomised head-to-head comparison of discussion against lecture [10] is better evidence about method than a correlation across studies would be.
The summary gives you a count, not a magnitude. "Four in five trials showed meaningful positive effects" [7] counts trials, not effect sizes, and the write-up neither reports a pooled estimate nor names the instruments used to score moral judgment [23]. It is also unclear whether four in five refers to all 92 trials, about 74 of them, or only to the 54 with controls, about 43 [21]. With students from elementary school through university [2], across 20 countries and nearly five decades [3], a Year 4 discussion group and an undergraduate ethics unit are averaged into the same number.
This also leaves behaviour unaddressed: whether the children act differently. The authors say they did not measure long-run behaviour [8]. The link from reasoning capacity to helping others, resisting peer pressure and lower juvenile delinquency is an association carried over from other literature [9], which is a reasonable prior and not a result of these trials.
The Australian mechanics are where the evidence bites. Ethics has been in the national curriculum for more than a decade as a general capability sitting alongside literacy and numeracy [11], yet Australian research finds even teachers who value those capabilities struggle to teach them explicitly, because attention follows what gets tested [12], and ethics has no dedicated time of its own [13]. New South Wales solves this with Primary Ethics, which reaches roughly 40,000 students a week in about 500 schools using trained volunteers [15], an average of 80 students per school [22], in the slot otherwise used for religious instruction [16]. Queensland and Western Australia keep that slot but offer no secular alternative, so those students get "other instruction" [17].
David Gonski's recent call was framed around social cohesion [18]. The review supports something narrower and more useful: timetabled small-group discussion reliably moves individual reasoning scores. Cohesion is a collective outcome measured in behaviour, and 92 trials of individual reasoning do not reach it. My view is that the timetable argument is now an evidence claim curriculum designers have to answer on its merits, while the cohesion argument remains rhetoric, and welding the two together makes the strong one easier to dismiss the first time a program fails to deliver the weak one.
