i have a specific experimental setup (at least i think so :) It's a replication and i don't know how to correctly determine the false alarm rate for the visual fields since in the article i replicate the FA calculation procedure is not explicite shown. But my impression is that it kind of deviates from a simple 50% noise 50% signal+noise design. Moreover correct d'prime and bias measures depend on the calculated FA% scores. Maybe the design as such is not suitable for calculation of such SDT meassures? I read through the Nacmilian Creelman book (though not very carefully) and am not able to find a solution.

ok it's a lateralized lexical decision study testing the left hemisphere superiority in word recognition ( better word - non(pseudo)word discrimination in the Right vs Left visual field)

  1. overall 256 trials
  2. on every trial 2 letter strings are presented for 100 ms: one string to the left and one to the right of a fixation point. ( xxxxx + xxxxx )
  3. on 128 trials a word is presented randomly to the left and right accompanied by a pseudoword: 64 RVF word presentations (e.g. gilus + virus) and 64 LVF word presentations (e.g. virus + gilus)
  4. on 128 trials pseudwords are presented to both sides-hemifields e.g.(mudo + mumi).

Subjects must press left button if word in LVF right button if word in RVF and spacebar if nonwords are in both hemifields.

Generally i would like to determine the d' or A' and bias measures for each hemifield, the correct calculation of FA% is crucial.

The Hit rate determination for LVF and RVF seems straight forward: correct left and right presses, each divided by 64, since there are 64 word presentations in each hemifield.

Question 1: when it comes to determining the false alarm rate for one hemifield which i would like to use for d' calculation: should i divide the measured FAs on one side by all possible FA occurences on that side i.e. 192 (128 bilateral nonword presentations + 64 nonwords in word-nonword trials)?

Question 2: because the script did not repeat timed out trials should they be considered as misses?

thanks Kuba


Check out chapter 3 of Macmillan and Creelman on rating scales. You have 3 ratings: left arrow (very confident it is on the left), space bar (not sure), right arrow (very confident it is on the right). You also have non-responses which may be equivalent to a space bar "not sure" or might be something else. I would probably want to know more about why they did not respond (and how often) before deciding what to do.

You also have 3 stimulus types (LVF-word, RVF-word, and double non-word) which makes things harder and I don't think is covered in M & C. That said, you can extend their examples to 3 stimuli pretty easily. The index of sensitivity d' is a measure of distance between two distributions. It sounds like you want to know the distance between the LVF-word and double non-word distributions relative to the distance between the RVF-word and double non-word distributions, but maybe you only care about the LVF-word and RVF-word comparison.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.