You may recall the delightful panther irony when we looked at experimental studies which showed that Other Guys would go High WATT when they consumed sugar. This led to the observation that the Lifestyle Drum and Bugle Corps needed to give Other Guys high doses of Cream Soda along with their strong scientific Arguments against . . . Cream Soda. The science behind this effect turned on basic brain physiology as the explanation and launched a series of studies that manipulated sugar in some form then looked at decision making, delay discounting, and willingness to pay or to work as outcomes. Well, there’s been enough work in this Sweet Science to produce a meta-analysis. And, things are a lot more complicated for a lot of complicated reasons. Much to ponder, my friend.
Start with the meta. Researchers scoured the peer review lit for all the studies on the topic. Once located, they carefully coded each experiment for a variety of theory-relevant variables along with the primary outcomes (decision making, delay discounting, etc.) Then, and this is the interesting and complicated part of the meta, the researchers identified four different theories that explain why Other Guys do what they do depending upon sugar.
In the following sections, we describe four theories, two of which are examples of the constraint view — ego depletion (Gailliot et al., 2007) and dual systems theory (Kahneman, 2011)—and two of which are exemplars of the signal view — optimal foraging theory (Kacelnik & Bateson, 1997) and the theory of visceral influences (Loewenstein, 1996). A key difference between the two types of models is how they treat food as opposed to other sorts of rewards.
After a thorough and fair discussion of the four theories, the researchers then create a simple table which shows how the four theories differ in their predictions about outcomes (decision making, delay discounting, etc.). Here you go.
And, now you can see how this meta is more complicated than the usual meta we dig into. Those typical metas focus upon a primary X causes Y hypothesis as with interventions on Health Beliefs or Fear Appeals or whatever. This meta goes beyond the basic X -> Y prediction and tests it under the microscope of theories. Thus, it’s not just the Windowpane, but how the Windowpane fits or fails different theory predictions. Here’s what the researchers expect among the four theories.
In sum, the ego depletion and the dual systems model, because they are proposed to be general in their applicability, do not predict a difference for food-related and nonfood-related tasks. They predict that those with low levels of glucose will show, generally, greater willingness to pay, less willingness to work, greater time discounting, and greater reliance on heuristic decision making. The visceral influences view, as well as the optimal foraging view, predicts that willingness to pay and work for nonfood items will, in contrast, decrease for those with lower levels of glucose. The predictions for discounting and decision style are unclear for optimal foraging theory. Finally, the visceral influences view makes the (debatable) prediction that under low blood glucose time discounting for food will be stronger, but not for nonfood items.
Pretty cool, huh? We’re not only opening Windowpanes on effects, we are assembling those Windowpanes according to theory to see which theory/theories fit the Windowpanes best. Sounds simple and it is if the data break clean and exact, like homemade peanut brittle on a cookie sheet out of the oven and cooled: Just whack the sheet with a little hammer and snap, clean and separate pieces. Let’s start counting the change. Like I said at the top, this is complicated.
Note my red highlights. The first row provides effect sizes and other statistics for all 42 experiments in the meta. The column with the “r” header contains the attenuated correlation while the column with the “ρ” header contains the same correlation, but not attenuated (and we’ll get to the attenuation process later). As you look at the average effect size for all 42 studies you see a trivial -0.04 correlation, less than half a Small Windowpane, about a 48/52 effect for sugar on all the outcomes. This looks like what we find with all those bad interventions with Framing or Fear or Health Beliefs. Not much going on. But!
Remember, the four theories make different predictions under different conditions. The average effect is not that interesting and in fact is misleading in this context. You need to break out the 42 experiments for their individual theory implications.
And, to do that, you must read the paper because that break out is incredibly complicated. Each of the four theories succeeds and fails on some of the predictions. Here’s how the researchers conclude the analysis.
The results of the meta-analysis lead us to believe that peripheral glucose levels do seem to exert effects, in some way, on a number of decision-making tasks. Having said that, we view no extant proposal as having survived the analysis; in each case, at least one prediction fails. The view from optimal foraging theory was not falsified, but that framework, in our view, offers the fewest predictions in terms of the work that we reviewed.
It is, of course, possible, even likely, that the systems in question are sufficiently complex that there will be no single explanation that can handle the diversity of effects of blood glucose on decision making. This would not be very surprising insofar as caloric state is a crucial biological variable, and the internal measurement of the organism’s current need for food is likely to serve as an input to any number of decision-making systems, upregulating some, down-regulating others, and so on. We look forward to additional work addressing these complex issues.
That’s a complicated way of saying we’ve got a mess on our hands. Sugar appears to have important and practical effects on the way Other Guys think, feel, and act on outcomes like decision making, delay discounting, and willingness to work or pay. But, these effects vary widely and no one theory covers everything. It also appears that each theory is true and useful under specific conditions. Put Other Guys in this Box and you get one practical effect. Put Other Guys in that Box and you get a different practical effect.
Part of the mess stems from Sweet Science in the mechanics of both the experiments and the meta. I’ll give you an example of each. First, the mechanics of doing a sugar experiment. The researchers quickly discover that different researchers use different methods for manipulating sugar. Here’s a table.
You can see that manipulation makes quite a difference ranging from an effect near zero (0.06) with the Hunger Score approach to a near Medium effect (0.28) with either direct glucose measurement or giving the Other Guy a blast of Cream Soda. Realize that each research team had a good reason to manipulate sugar in the Other Guys with these different tactics and none is wrong. But, it makes a big difference in outcomes.
Now, the second complication with the mechanics of the meta. The researchers took the raw data from each of the 42 studies and then ran various transformations, alterations, and adjustments. This is analogous to all those “adjustments” so infamous with Tooth Fairy Tales. You don’t just take the raw data that is the most basic marker, measure, or indicator of the variable; you diddle that thing with various legitimate, widely used, and widely defended adjustments, almost always a math trick.
According to psychometric meta-analysis, imperfect construct validity of study i attenuates the observed effect size r relative to the true effect size ρ proportional to the square root of the reliability rxx:
The square root of the reliability rxx is referred to as the artifact multiplier ai and is a more direct way of seeing the expected attenuation. Studies in which blood glucose, for instance, is operationalized via food deprivation have an average reliability of rxx = .50. The artifact multiplier ai is thus √.50 = .71, which means that food deprivation studies on average result in effect size estimates that are 71% of the true effect size ρ; that is, the effect size which would be obtained using an operationalization with perfect construct validity such as glucose administration.
You just hit the Brooks Effect, didn’t you? Eyes glazed over with the Greek symbols and the a to the ith notations. That’s the nature of adjustments. Most scientific Other Guys go Brooks on this stuff. What can you expect from a Child of the Night just looking to run a Box and Play with Mountain Dew?
This is a standard transformation in meta analysis and it simply says that if measurement is unreliable, then the effect size is underestimated and would be higher if the measurement was perfect. So the formula corrects the unreliable measurement and provides a perfect estimate of the perfect effect if you had a perfect measurement.
I understand both the math and the justification; I just don’t trust it. Unreliable measurement means you need to work harder at measuring things, not playing math games that look perfect on paper. And, in this meta, there are more coding and math transformations that also make me wary about conclusions. My wariness arises from past metas that do things like this and often create more noise than music. An infamous series of metas back in the 1980s with the ELM employed many of these same coding transformations to prove perfectly that the ELM was stupid, wrong, and biased. Of course, almost no one remembers any of these perfect metas and the ELM just rolls on as the most cited persuasion theory (although that Nobel Prize seems to have helped make the weaker System 1 and 2 idea more popular!).
And, now you realize that Professor Poopypants has snuck up on you and you’ve lost your main reason for reading this post. Hey! Does sugar change WATTage? Here’s how the authors summarize that.
The analysis revealed a significant positive effect of blood glucose levels on decisions style meaning that low blood glucose increases the tendency to make intuitive rather than deliberate decisions on tasks that are not food related.
That confirms earlier sugar studies we looked at. When you hit Other Guys with Cream Soda they will have High WATTage and can travel the Central Route with your strong scientific Arguments that they should stop drinking Cream Soda! But only if the TACT doesn’t involve food! Double irony! Sugar makes it easier for Other Guys to understand the science against sugar, but since sugar is food, they won’t care!!!
Orquin, J. L., & Kurzban, R. (2016). A meta-analysis of blood glucose effects on human decision making. Psychological Bulletin, 142(5), 546-567.
doi:10.1037/bul0000035