Thanks
@bobbler for your useful analysis.
Here's a bit more:
The Effort-Expenditure for Rewards Task (EEfRT)
So, yes,
Method in the Walitt et al study
for each choice, the participant chooses one of two tasks:
the easy task involving 30 button presses in seven seconds with the dominant index finger
the hard task involving 98 button presses in 21 seconds with the non-dominant little finger
Each choice of task has a reward of a specified value and a probability of winning it, if the task was completed it. And it sounds as if the probabilities are kept the same within a choice, but vary between choices. It isn't clear to me if, in the Walitt et al study, the probabilities tended to increase towards the end, or if participants were told that they would (both of which seems to be the case with the test validation study that bobbler linked).
So, the first choice might look like (and I'm just making this up)
Choice 1a: Easy task reward $2.00; probability of winning it if easy task completed - 25%
Choice 1b: Hard task reward $3.50; probability of winning it if hard task completed - 25%
Once Choice 1 is done and the participant finds out if they won the reward, the participant goes on to other choices, until 15 minutes is done.
At the end, two rewards are chosen randomly, and the participants receives the sum of those two rewards.
From bobbler's posts, it appears that the test in the Walitt et al study differed from the standard method in crucial ways, including the outcome measured i.e.
- ratio of hard tasks to easy tasks, or
- a measure that takes into account variation in the propensity to choose the hard tasks based on the probability of receiving the reward
It is not at all clear to me yet if or how probabilities and rewards varied between tasks, if participants were told about the variation and how any such variation was accounted for in the analysis.
Results
View attachment 21176
We aren't shown the actual data in Figures 3a-c, just confidence intervals. I think that's a problem. The statistics used to produce the confidence intervals could be worth looking at. If the probability of choosing a hard task is much lower (and it was especially towards the end of the trial), there should be less data in Fig 3c, and I would have thought that less data would make the confidence intervals bigger. But, they are not.
Figure 3 a - Probability of choosing the hard task
The probability of choosing the hard task was a bit lower in the ME/CFS group than in the healthy group, throughout the trial. It wasn't actually that big of a difference. The difference didn't change over time, which I think the authors concluded meant that there wasn't more fatigue in the ME/CFS group. But actually, if your dominant index finger was getting tired, you might choose an occasional hard task using another finger, to give the dominant index finger a break for a while.
Figure 3b - Button press rate for easy tasks
What they mean there is that people with ME/CFS got slower with the button pressing for the easy tasks over time, but were still fast enough to reliably complete it. That is what the authors see as 'pacing'. And that approach makes good sense, because the participants were not rewarded on the number of tasks that were completed - however many they did, the computer would only pick two rewards at random to pay out on. The only reason you might want to get through the easy tasks fast is to get to later tasks that might have a higher reward.
(I think it's questionable how big an incentive something like $5 would be to a participant with ME/CFS who probably wanted to make sure that they could get through all of the studies, many of which probably seemed a lot more important than making choices about button pushing. I assume it was clear to the participants that this wasn't meant to be a test of fine motor control. So, would you really go all out, concentrating and pressing buttons for 15 minutes, in that context, after you had completed the two tasks needed to qualify for some reward, knowing that this was a government study and so it was very unlikely that later rewards would be a lot more than what you had already secured? There was a lot less on the line for the healthy participants, and so they probably felt freer to fully engage in the game. In terms of effort preference then, I think the people with ME/CFS may well have been expressing a very reasonable effort preference, given the circumstances. )
It is possible that fatigue played a part in the decline in button pushing rate and the substantial difference in button pushing rate between the two groups seen in later easy tasks - I don't see how the authors can rule that out.
Figure c - Button press rate for hard tasks
Most people undertaking the hard tasks pressed the button fast enough to complete them. Both groups showed a very slight increase in press rate with experience. But Figure 3a tells us that the people with ME/CFS were unlikely to choose to do the hard task, especially towards the end, so I expect that there wasn't much data behind that finding.
(See my note about the confidence intervals above - I think they are generalised in a way that conceals the paucity of data at different trial numbers. It would have been better to show the button press rates as plotted data points, with the x axis being time. Maybe such a chart can be constructed using the source data.)
****
Frankly, I find this task a joke in terms of providing insight into the supposed pathological psychology of people with ME/CFS. I feel appalled that conclusions are being made about the 'effort preference' of people with ME/CFS on the basis of this 15 minute study with 15 ME/CFS participants.
