Re: [hddm-users] Is it logical to use changes in drift rate to model duration-dependent changes in performance?

[Michael J Frank]

I think there is some validity to both views:

 It's true that the drift rate is a constant in the standard DDM, such that once it begins accumulating the rate is fixed (+gaussian noise).

But accumulation doesn't necessarily start right at stimulus onset: the non-decision time reflects a period of time for perceptual encoding before accumulation begins (and also motor execution after). So it is plausible that the animal would first perceive the stimulus during the non decision time (which is likely > 50 or 100 ms, so enough to encompass the short duration stims), requiring reliance on degraded iconic memory to accumulate after that. So then accumulation could be steeper / more efficient for stimuli that were processed more robustly (or continue to be presented after non decision time) compared to those that rely on noisy memory.  And there are certainly lots of cases where higher drifts correspond to stimuli that are easier to encode (higher contrast, numerosity differences, etc). 

Another possibility is that a better model would be a drift rate that accelerates with stimulus duration (sort of like the ornstein uhlenbeck model but with an acceleration instead of leak), which would require using LANs in HDDM or HSSM. But even if that would be more accurate, the DDM might still roughly approximate that with different slopes per duration. In that case you could acknowledge that the DDM is still useful for your purposes in approximating a different dynamic. 

 

Michael J Frank, PhD | Edgar L. Marston Professor

Director, Carney Center for Computational Brain Science 

Laboratory of Neural Computation and Cognition  

Brown University
website 

On Thu, May 2, 2024 at 12:23 PM Chase Mackey <seem...@gmail.com> wrote:

Hi all, we are trying to publish a paper in which animals displayed differences in discrimination performance as a function of duration. We modeled these changes by allowing drift rate to vary as a function of stimulus duration. One of our reviewers insists this is logically flawed, and we cannot get the paper published. The reviewer says the drift rate is set at the beginning of the stimulus, and therefore cannot be different for different stimulus durations, unless the animal had "predictive knowledge of the stimulus" which they didn't. I believe R1 is incorrect here. 

 Are we missing something here? If we are correct, can someone help me better articulate why this reviewer is incorrect? Below I share our main DDM figure with one of the reviewer's comments below it. 

"The authors argue that "The drift rate is not fixed at the onset of the stimulus but is instead a measure of how the diffusion process changes over the entire decision period (between the onset of the stimulus and the decision being made)." This statement is incorrect - see the mdel equation - the drift rate has a constant value that takes effect at the onset of the stimulus and only apaplies during the stimulus. The output of the accumulation process increases over time, and 'lingers' after stimulus offset. But the drift rate is a simple coefficient that is a constant during the stimulus. "

--
You received this message because you are subscribed to the Google Groups "hddm-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hddm-users+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hddm-users/a70107d5-c78a-4d13-b6c9-c1fd43475c54n%40googlegroups.com.

[Chase Mackey]

This is a really helpful explanation. Thank you Michael. I also found this paper by Ratcliff, which seems to help our case that what we are modeling isn't totally illogical: https://pubmed.ncbi.nlm.nih.gov/12120790/

While we do not have all the details about the processing underlying the task performance, it seems the DDM does provide a potential explanation, which is how we frame it in the paper. 

Chase

To view this discussion on the web visit https://groups.google.com/d/msgid/hddm-users/CAGYaZ%2BdxnuZahb-PBOY4WGM6FZmHJB5rp_WoTY%3Dqs0GaPt7p4Q%40mail.gmail.com.