Dear all,
I'm currently investigating the relationship between response accuracy and latency when predicting a participant's performance. In many previous publications, response latencies from incorrect responses were excluded from the analysis. However, I've yet to find a rationale given. I do understand, that you want to reduce the influence of response times that stem from a guessed response. However, in many tasks, the participants have a 50% chance of guessing the item correctly, therefore only excluding the responses from incorrect trials removes only half of the problem. Furthermore, important information is lost (e.g. is the accuracy maybe lower for quicker responses?).
I wanted to ask specifically if there are any sources out there, that I could read for a rationale of this procedure, as there are never any references...
Thank you in advance,
Daniel
Daniel Schulz I have voiced these very concerns in a few previous (and current) papers of mine, urging researchers to not omit error trials from analyses! Many researchers looking at RT data are only concerned with 'correct' trials, since "we can't infer why an error occurred." The problem with this mentality is twofold: (1) it's also true for 'correct' trials! We try to account for 'guessing' by using various experimental designs or outlier detection methods, but we can never be certain if a correct response is due to the effect of interest or guessing.
(2) We are losing a lot of important information by not examining error data - you are correct! Patterns in error data can help us understand why participants make more errors for one kind of stimulus relative to another, or help reveal particular strategies that participants may be using (see Ratcliff, 1993).
*You won't likely find a source justifying omission of error data because there really isn't one. Now, if participants displayed ceiling effects in an experiment I could see why a researcher might not examine error RTs because there simply isn't enough data to make any meaningful inferences concerning participant errors.
Cheers
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments