Visuomotor Adaptation — Research Log
A running timeline of the visuomotor rotation (VMR) + EEG study — experiments, analyses, and decisions recorded as they happen. Newest updates appear first.
View the project page-
Latest
Data collection is complete
Today we completed the data collection phase. With final exams approaching for university students, the study has reached 50 participants.
For now, we will stop recruiting new participants and move into the next stage: data analysis and manuscript writing. I hope everything goes smoothly from here.
At this point, the data appear to be consistent with the reference model and broadly in line with our hypotheses. The more specific differences in strategy use, however, will have to wait until the analysis is complete.
Wish us luck!
-
Thirty-plus participants, and the pseudo-implicit problem
After an intense couple of weeks of data collection, the group finally has some encouraging results to show for it. The task is now fully aligned with the paradigm.
One thing became clear along the way: making sure participants genuinely understand the task strategy is critical. If a participant can't grasp the strategy, they very easily turn into a pseudo-implicit participant — looking implicit on the surface while something quite different is going on underneath.
The group has now collected data from 30+ university students, and the analysis starts soon. :)
-
Eight targets, and the trouble with instructions
After rebuilding the task to follow Experiment 4 of the reference paper — the eight-target version — a fresh set of problems showed up.
- Too few counter-rotation trials. With only a handful of reverse-rotation trials, I'm caught between two failure modes at once: too few to properly disrupt the original adaptation, yet also too few for participants to genuinely learn the new, opposite adaptation. Either way, by the time we reach the error-clamp (EC) phase there isn't enough competition left between the two memories to produce a clear spontaneous recovery.
- The instructions are the whole task. The part I've underestimated most is the wording of the instructions. Explaining what a participant has to do — and getting them to actually understand it — turns out to be genuinely hard. Time and again someone assures me they've understood, then behaves during the task in a way that makes it obvious they haven't. And the constraint runs both ways: the instructions can't reveal that anything about the task will change (the rotation switch has to stay hidden), so I can't simply tell them what's coming. Saying enough to be understood without saying too much is a very narrow path.
For now it's slow going — the valid-data rate is sitting at around 25%. The next pass is mostly going to be about the wording of the instructions and the trial counts in the counter-rotation block. :)
-
Daily note
21 participants done. It took about a week to tune the task paradigm, and there are still plenty of spots where it's easy to go wrong:
- In the VMR task, to cleanly separate an individual's implicit and explicit components, you have to use a landmark-based report paradigm. Our earlier data had no reports, so the two were hard to dissociate. The trade-off: once reports are added, everything climbs steeply — recording what each participant reports, explaining the task to them clearly, and the difficulty of the task itself. I still need to find a better way to handle this.
- Whether the task uses a single direction or multiple directions also shapes the results. A single direction gives stronger spontaneous recovery and is easy for participants to grasp, but it's easily confounded with a directional bias. A multi-direction task fixes that, yet participants then struggle to understand the R1 / R2 rotation directions.
For now the EEG experiment is on hold — the behavioral task alone is already hard enough. Hopefully I can work through the issues above and get the follow-up experiments running smoothly. :)