search

seannyD / StoryTransmission

0 stars 0 forks source link

Distraction task: task #24

Closed fionajordan closed 7 years ago

fionajordan commented 7 years ago

Could we have a 4x4 grid with up to five symbols, chosen from nine? It is way too hard at the moment and I think unnecessarily so.

rewberl commented 7 years ago

I went back into the script (<3 R) and selected the 9 most dissimilar symbols (excluding the "ribbon" one, as I think it's a bit too familiar). The resulting characters are A, B, C, D, I, K, O, U, V. Word file showing the symbols (if you already have the BACS1 font installed) is attached. If not, it is available here: https://osf.io/w4gpc/

BACS_chars2.docx

seannyD commented 7 years ago

I'll adjust this. It's quite tricky to be flexible on the character set, though, so maybe we can test with the current set, then change the characters when we've settled on the number of symbols we want.

seannyD commented 7 years ago

Another issue is the number of rounds. To get usable statistics for people, we want a number of datapoints per player. Currently, there are only 3 rounds, mainly because you need longer to take in more symbols. If we reduce the number of symbols, we can reduce the display time and increase the number of rounds.

rewberl commented 7 years ago

I figure that, out of their two sets of 3 trials (for a total of 6 trials), we have a few options. We could take their highest overall score, average or sum the highest score from each set, average or sum the last score from each set, average all scores, or sum all scores.

In the original VSLT task (Mulac et al. 1991), they had 5 trials and separately summed the 'designs recognized' ("D"), 'positions recalled' ("P"), and 'designs placed in correct positions' ("P&D") scores. They then find P and P&D to be highly correlated and, to reduce redundancy, sum them to get their "Position Learning Index" ("PLI"), which they use along with D. I think that before I had mistakenly interpreted the final score as a sum of P and D. I'll make a new issue to fix this.

Summing them seems to discard information about the distribution of their responses, and taking either the sum or the mean would discard the temporal information. In the end, I don't know that it matters too much; we just want some crude way to control for individual variation in memory. I'd probably lean toward following what the VSLT task did and track P, D, and P&D separately, then sum P and P&D to get PLI and plan to use that score and D for our analyses. Since our task is not identical and could not be compared to "normal" neurological results, I think we skirt any ethics implications of using a real clinical test, but still retain the validity of the design.

I think the two sets of 3 trials is probably fine, as we don't want to overburden people. I definitely start to get bored and frustrated after a couple trials.

seannyD commented 7 years ago

Distraction task is now 4 x 4 grid displaying 5 images out of a total of 9, using the symbols A, B, C, D, I, K, O, U, V. The results file now tracks all variables in the VSLT task, as well as all the actual symbols and locations in each trial.