Millisecond Forums

Stroop with Speech Recognition

https://forums.millisecond.com/Topic3548.aspx

By hilary - 11/18/2009

hi there
i am in the process of setting up a stroop task using the speech recognition software. however i was just wondering what the difference was between using the stroop script with voice recording vs. the stroop script with speech recognition in terms of analysis. also, does this software actually wait for the correct response or does it record the reaction time of any word stated?
thanks! you guys are so helpful!
hilary

By seandr - 11/23/2009

Hi Hilary,


If you use the speech recognition (i.e., /inputdevice = speech), then Inquisit uses the speech recognition engine during the task to determine what response was given on each trial. Only valid responses are accepted (e.g., "red", "green", "blue", "yellow" are accepted, but "bananas" is not). Both the response and the response latency are recorded to the data file. 


If you use voice recording (i.e. /inputdevice = voicerecord), the Inquisit does not attempt to analyze what word was spoken as the task is performed, nor does it distinguish between speech and other sounds such as coughs, screams, etc. The audio input is recorded for each trial, however, so that you can determine whether a correct or valid response was given. You can do this by using the Analyze Recorded Responses command from the Tools menu. The command allows you to enter the valid responses and then run the speech recognition engine on all of the recorded responses so it can automatically determine the response. If the engine doesn't recognize the response, it shows a "?". You can easily play each response that was given to check the recognition engine's accuracy and determine what was said when the recognition results are ambiguous.


Hope this helps,
Sean

By hilary - 11/24/2009

thanks sean!
your explanation was very helpful. i guess i'm just wondering how my results would differ by using the different formats. in the speech recognition format, the response latency recorded would include people self-correcting wrong answers, as this format allows only the correct response to move forward. however, in the voice recording format, the response latency would not include self-correcting and move forward at whatever sound it records. am i understanding the difference correctly? i guess i'm just trying to understand what this difference means for my results? thanks again.
hilary

By seandr - 11/24/2009

Hi Hilary,


If you set "/inputdevice = speech", then the behavior depends on your /validresponse and /correctresponse. Consider the following example:


<trial bluetrial>
...
/ inputdevice = speech
/ validresponse = ("red", "green", "blue", "yellow")
/ correctresponse = ("blue")
...
</trial>


If any of the color names specified in /validresponse are given, the response is accepted and the next trial begins. If the response was "blue", that trial is scored as correct. Otherwise any of the other color names will be scored as an error.


If you want to require that participants give the correct response before moving to the next trial, you would change the responsemode as follows:


<trial bluetrial>
...
/ inputdevice = speech
/ validresponse = ("red", "green", "blue", "yellow")
/ correctresponse = ("blue")
/ response = correct
...
</trial>


If the response was "blue", that trial is scored as correct and the next trial begins. Otherwise, if any of the other color names are given, the trial is scored as an error. The trial waits until participant gives the correct response until moving to the next trial.


With voicekey and voicerecord, Inquisit can't tell whether the response was valid or correct, so it simply accepts any utterance.


-Sean

By hilary - 11/25/2009

awesome! thanks again sean! i don't think i'll be changing the standard script, as this would make the difference between congruent and incongruent sections even larger and would perhaps manipulate things in my favor (as i'm predicting a significant difference between conditions) don't you think?
hilary