While scouring news sites for interesting, data driven stories over the weekend, I came across an intriguing article from the Washington Post. This article analyzed decades worth of marshmallow tests to consider how children’s self-control has changed over time. The main researcher behind this meta-analysis, John Protzko, surveyed 260 experts in cognitive development before commencing his study. In short these experts were not optimistic; fifty-two percent expected to see a decrease in children’s self-control over the last fifty years. However Protzko’s analysis showed an INCREASE. Despite the smartphones and tablets and constant stimulation that today’s children experience, their capacities for self-control have not only prevailed in comparison to past generations but improved! Let’s take a look at Protzko’s analysis for a deeper understanding of this supposed trend.
Protzko’s meta-analysis included thirty-one studies ranging from 1967 to 2017. The Washington Post article states that all of these studies were published. Meanwhile the study itself states that the researchers gathered ALL data possible – both published and unpublished – within this time range. This difference might seem trivial, but it is actually quite important. If the researchers gathered only published studies, the file-drawer effect becomes increasingly worrisome. Past marshmallow tests that deviated from normal results might never have been published – with researchers worried they made some mistake. Yet if the researchers of this meta-analysis gathered all possible data, the question of HOW remains. Detailed methods are not listed in the study making it impossible to discount the file-drawer effect in this scenario as well. Furthermore not all studies are created equal. There is no mention of minimum sample sizes or procedural reviews in order for a marshmallow test study to appear in the meta-analysis.
Nonetheless let us move on from the methods to the data. Nearly all of the marshmallow test studies exhibited an average delay time at or under ten minutes. There were four exceptions to this trend: 11.4 minutes in 2014, 18.2 minutes in 2012, 23.7 minutes in 2000, and the 26.0 minutes in 2000. The gap between 11.4 minutes and the remaining three is large enough to spur some curiosity. The average age of children in the 2012 study was 7.68 years. The average in both of the 2000 studies was 9.2 years. The majority of the studies included in the meta-anlysis had an average age of four years. The three most surprising examples above not only seem to be outliers, but an explanation for their aberration appears to exist as well. Although the analysis identified one of these points as an outlier and stated no difference in the results, I wonder how the results would look if all three points were removed – perhaps if an average age limit was enforced.
Despite my critiques of this analysis, the notion that children’s self-control has increased over time is reasonable. The analysis includes a citation from Flynn in 1984 stating that “all cognitive abilities have undergone secular increases in the past century”. If all realms of human cognition have improved then it would make sense for children’s self-control to improve as well. Not to mention society (for the most part) is progressing and evolving. That must mean that the children of the past have improved the world of their parents, demonstrating some type of cognitive achievement. Regardless, I don’t think this analysis warrants a final say for either end of the argument.