David M. Huberdeau
This code analyzes data obtained for a behavioral visuomotor association (VMA) study. The study has two experiment variants: a recall-variant and a statistical learning-variant.
The primary objective of these experiments were to determine if learning the associations between arbitrary visual cues and movement goals would facilitate movement preparation, and if so if that facilitation would be as effective as when goals were more directly predictive (such as presenting the target location ahead of the actual appearance of that target). Furthermore, we sought to determine how learning is done in the VMA task, especially when the associations had to be learned through experience. Finally, we used a novel experimental technique to eliminate the potential ambiguity inherent in using Reaction Times, where we eliminate the speed-accuracy tradeoff.
The recall-variant visuomotor association study has human participants train in the Timed-Response paradigm, and then gives them symbol-target associations to memorize. Those symbols appear in the study as cues presented ~3 seconds prior to the appearance of an actual target. On some trials, instead of the symbol, a target itself appears as a cue, and sometimes no cue is provided. All cues are 90% reliable; when a cue is unreliable, a different target (chosen at random from among the remaining targets) is presented instead without the participant's prior awareness.
The statistical learning-variant visuomotor association study proceeds similary to the recall-variant, only participants must learn the associations between arbitrary symbols and targets de novo. Specifically, human participants are trained in the Timed-Response paradigm, and then engage in a task where, much like it's complementary recall-variant, symbolic cues are presented that uniquely indicate the location of a target. In this variant of the task, multiple symbols were associated with each target.
There were three conditions in this study that differed in terms of the number of targets, number of symbols, and type of symbols used. In Experiment 2.1, three targets were used with two symbols per target. In version 2.2, six targets were used and 12 symbols, and in version 2.3 four targets were used and 12 symbols (three symbols per target). This parameterization was attempted to calibrate the parameters to use in an fMRI version of the task so that learning is reasonable over a 45 - 60 minute study duration.
These studies also answer other important questions. Do changes in a movement plan require more time to plan than forming a plan de-novo? Is the quality of movement execution the same when a movement is prepared at short-notice compared to when there is sufficient prior notice? How (i.e. by what mechanisms) are symbol-target associations learned? Do participants memorize the associations directly, is there a slower, non-recall-based process at work, or some combination? Furthermore, by carefully controlling processing time, this study investigates VMA learning without the potential confound of having both reaction time and choice accuracy simultaneously changing, i.e. because of the existence of a speed-accuracy tradeoff function (SAF) between processing time and choice accuracy, observing changes in both processing time and accuracy makes it unclear if participants are moving along an existing SAF or experiencing a shift in their SAF.
Primary study outcomes in addition to empirically determining appropriate parameters for future imaging studies was to observe whether pre-planning with symbols would become as efficient as when pre-cued directly with the target (e.g. by measuring movement variability) and to ascertain the probability of symbol recall as a function of the number of occurrences of that symbol.