Millisecond Forums

Latencies of incorrect trials

https://www.millisecond.com/forums/Topic16512.aspx

By lg021905 - 6/16/2015

Hi everybody,

does anybody know how the incorrect-trial latencies are recorded?

I have used the provided script of the ST-IAT (Inquisit 3) and modified it in terms of the stimuli and the color settings. Given that the required answer format is "correct", meaning that participants are supposed to correct any wrong answer before proceeding, I just wanted to make sure that I am right in assuming the following:

>> latency of correct trials = response time to press correct key
>> latency of incorrect trials = response time to press incorrect key + response time to press correct key

If so, the incorrect-trial latencies would contain an in-built error penalty...

I have checked the descriptive statistics and, clearly, the incorrect latencies are significantly larger than the correct latecies. However, the inquisit 4 ST-IAT script says that the latency refers ro the "the latency of the final (correct) response". Does anybody know if this applies to the Inquisit 3 script as well?

Thanks so much in advance!
CB
By Dave - 6/16/2015

Yes, that is correct. The latencies for incorrect trials reflect when the correct response was given (not when the initial incorrect response was given), i.e., they include a built-in error penalty.
By lg021905 - 6/16/2015

Ah, great! Thanks so much for you quick response, Dave! Does it say so somewhere in the script? Your information is very valueable to me: I have to rethink my analysis strategy then, because I have used the D-2SD and the D-600 algorithm so far which has been recommended for lab studies by Glashouwer et al. (2013).

Do you, by any chance, also know how Inqusit treats the onset that it takes until the error message is displayed? I think it is 200ms according to the script:
"/ errormessage = true(error,200)"


By Dave - 6/16/2015

I frankly don't recall whether it says so explicitly in the script, but that's how /response=correct always behaves.

Also, AFAIK, you can safely calculate alternative D-variants because in all those variants the actual latencies of error trials (i.e., the point in time at which the incorrect response occured) is never used in the calculations / is not needed.

There is no way to determine the *onset* of the error message in a given trial -- it will be displayed directly upon submitting an incorrect response. That would be equivalent to the response latency for the (initial) incorrect response, however you don't know that value (due to the /response=correct requirement). The "200" in the /errormessage attribute merely reflects its minimum display duration.

Hope this helps / clarifies.