Updates to HCP dataset

Hi all,

Three components have just been added to the HCP dataset:

  • Aggregate (run-wise) behavioral measures for the in-scanner tasks
  • Original IDs that would allow you to link the processed data to other HCP derivatives
  • Synthetic demographic measures from a model trained on the original dataset and real demographics

Please see the Data Loader notebook for information about how to use.


nice this might be real useful.
link to data note books in NMA github:

Thank you for these additions – is there anyway that data (accuracy and reaction time for example) could be retrieved for individual trials?

Trialwise accuracy should be obtainable from the EV files (correct vs. error). For trialwise RTs, you would have to download the e-prime output files from the HCP servers (after sign-up, as one would have to do to download unrestricted demographics and behavior). But in my opinion it’s unclear whether RTs are interesting/useful for these specific tasks included in the HCP.

1 Like

Can we get the gender and age of the subjects ??

Yes, I believe that is included in unrestricted demographic information from HCP


We downloaded the data from OSF to see if we could find trial-wise accuracy, but had no luck. The behavioral data appear to be aggregated. Is there any way to recover the trial-wise behavioral data for the task. We are attempting to use tfMRI to predict responses.

Here is an example of what the EVT file contains:

1 Like

Apparently, no separate package containing only the e-prime output files is available in HCP servers. As we found out, the trial-wise behavioral data for each task is only obtainable from a giant-size package containing all the fMRI data of the corresponding task. Has anyone managed to find the trial-wise behavioral data as a separate dataset?

1 Like

That sounds right; I don’t know if it’s available in any independent format. I would say that trial-level behavior from a block design — which most HCP tasks used — is likely to have relatively small utility for fMRI analyses. I think part of the reason that the detailed trial-level data is not easily accessible is that there isn’t going to be too much you can do with it, unfortunately.

1 Like