Distribution of Foil, Target, Lure in Continuous version

[Kelly Hiersche]

Hello,

My lab is interested in the Continuous version of the task from the Yassa et al. 2010 paper and the Bakker et al 2012 paper. We were wondering what the background is in deciding the distribution of condition types.

The Yassa et al 2010 paper has the breakdown as 

44 Foil

16 Target pairs

16 Lure pairs 

which reduces to a 11:4:4 ratio. 

In this case, the subject would be responding "New" on 76 of the 108 trials (~70% of the time)

In the code that I downloaded from the GitHub, which was for 8 functional runs, the total number of stimulus was 

384 Foil

96 Target pairs

96 Lure Pairs 

which reduces to a 4:1:1 ratio. 

In this case, the subject would be responding "New" on 576 of the 768 trials (75% of the time)

We are interested in reducing the number of foil images, considering the first presentation of the Target and Lure photos are considered "New" to the participant. Has this been done before? 

Any additional information on how this breakdown was developed would be greatly appreciated!

Thank you,

Kelly 

[Craig E.L. Stark]

Funny you should ask...

First, while the ideal rates may be that, remember that the actual rates will vary.  People will call the lures "old" and not "new" (or "similar") and that throws things off a good bit from theory.  Second, if we go back to even the Bakker (2008) paper, being an incidental task, a goal was to have these lure and repeat items slipped in there without drawing too much notice.  When carrying over to the explicit version, we didn't have that goal, but we wanted decent enough lags between initial and subsequent presentations.

Since then, we've learned a few things.  The lag, once your past a handful of items, doesn't matter terribly, which does make for a potentially tighter design.  The GitHub page has a PsychoPy version and the lags generated for it can be shorter using fewer totally random foils.  A few months back, I put together a very tight version that we've been working on to make a much more efficient MST.  We've tested it a good bit in the lab as well and one feature of it is that there are no extra foils.  All foils come from the 1st presentations.  With a few other bits, this makes for a very, very efficient design that can give highly reliable results in a continuous design. We've got some good data on this version, but want some more and we have a few other ideas to try out (that hopefully a recent grant submission will help with).

Craig