Home Forums TCPW forum Bump attractors for working memory impairments in schizophrenia (Oct 6th 2015)

This topic contains 6 replies, has 4 voices, and was last updated by  John Murray 4 years ago.

  • Author
  • #347

    Quentin Huys

    Questions and comments on the meeting on this topic on October 6th 2015.

  • #348

    Quentin Huys

    Theoretically, distractibility in delayed working memory task should also imply impaired integration of information over time. Has this been tested?

  • #436

    Hi, great talk. Can I ask whether you think other perturbations in the model may be able to produce similar patterns to the E/I imblanace. For example, if the angle was represented more crudely wouldn’t this produce a similar effect?

  • #437

    Quentin Huys

    To what extent can these models be fitted to individual subject performance?

  • #438

    Martin Paulus

    How specific is the WM distractor dysfunction to schizophrenia?

    • #440

      John Murray

      Martin: As far as I know, the specificity of WM distractor dysfunction (or WM impairment more generally) to schizophrenia is very much an open question. First there is the need for more cross-diagonstic studies. But I also suspect we will need more informative tasks and analyses, which for my collaborators and me have been improved through consideration of the computational models. In the case of distractibility, the model guided us to have parametric variation of target-distractor similarity, and this feature of task design provided better characterization of the specific nature of distractibility, beyond its existence, which we hope can in turn inform mechanistic hypotheses. Perhaps such approaches can provide greater specificity.

      But perhaps aspects of cognitive impairment are indeed shared across disorders. One thing we learned from the model is that multiple distinct parameter perturbations can converge in their impact on the circuit function, leading to the similar net impairments in these WM behavioral measures.

  • #439

    John Murray

    Quentin: There has been very little application of biophysically-based circuit models to fit individual level performance. I think this is currently a major limitation of circuit modeling approaches. Partly this is due to the computational cost of simulation, especially for spiking circuit models. Partly this is due the parameterization of circuit models: the dependence of model behavior on parameters can be complex, with many parameters one can vary in principle, and particular combinations of parameters (e.g. relative E/I ratio) having stronger impact. As such, biophysically-based circuit models have primarily been used for more qualitatively purposes, to understand the mechanisms by which one parameter affects an aspect of model behavior.

    That said, there is promise in developing circuit models in that direction. I know of one recent paper that successfully used a (mean-field reduced) circuit model to quantitatively fit individual-level performance on a cognitive task:

You must be logged in to reply to this topic.