Advertisement
Research Article

When the Choice Is Ours: Context and Agency Modulate the Neural Bases of Decision-Making

  • Birte U. Forstmann mail,

    b.u.forstmann@uva.nl

    Affiliation: Amsterdam Center for the Study of Adaptive Control in Brain and Behaviour, Universiteit van Amsterdam, Amsterdam, The Netherlands

    X
  • Uta Wolfensteller,

    Affiliation: Max Planck Institute for Human Cognitive and Brain Sciences, Department of Cognitive Neurology, Leipzig, Germany

    X
  • Jan Derrfuss,

    Affiliations: Max Planck Institute for Human Cognitive and Brain Sciences, Department of Cognitive Neurology, Leipzig, Germany, Research Center Juelich, Institute of Medicine, Juelich, Germany

    X
  • Jane Neumann,

    Affiliation: Max Planck Institute for Human Cognitive and Brain Sciences, Department of Cognitive Neurology, Leipzig, Germany

    X
  • Marcel Brass,

    Affiliation: Max Planck Institute for Human Cognitive and Brain Sciences, Department of Cognitive Neurology, Leipzig, Germany

    X
  • K. Richard Ridderinkhof,

    Affiliation: Amsterdam Center for the Study of Adaptive Control in Brain and Behaviour, Universiteit van Amsterdam, Amsterdam, The Netherlands

    X
  • D. Yves von Cramon

    Affiliation: Max Planck Institute for Human Cognitive and Brain Sciences, Department of Cognitive Neurology, Leipzig, Germany

    X
  • Published: April 02, 2008
  • DOI: 10.1371/journal.pone.0001899

Reader Comments (2)

Post a new comment on this article

Referee Comments: Referee 1

Posted by PLoS_ONE_Group on 07 Apr 2008 at 21:28 GMT

Referee 1's Review:

Review of the first revised manuscript:
In their revision, the authors have clarified a number of points regarding the analyses included in the paper and the relationship between the results and the conclusions drawn. However, I am still somewhat unclear about the responses to some of the methodological comments. These remaining points could be further clarified.

(1) In response to the comment that region by effect interactions were necessary to draw distinctions between the regions, the authors provided a reliable three-way interaction between agency, context, and ROI. However, the distinction between these regions is not only in these conditions but also between cue onsets, correct? If this is true, doesn't this mean that the appropriate interaction is a four-way interaction between agency, context, on-set, and ROI. Alternatively, given that a four way interaction may be difficult to obtain, two three-way interactions of the form already reported (agency x context x ROI) should be reported for each onset. Also, these region by effect interactions should be reported in the text, and the pattern of activation in each ROI should be plotted for both onsets.

(2) I was somewhat confused by the response to the comment regarding unbiased ROIs. The authors response to the question of whether the ROIs were defined in a neutral and unbiased fashion was that the ROIs were based on the whole brain analysis. Relying on a whole brain analysis does not preclude unbiased definition of ROIs. The question was which contrast was used to define the ROIs. In particular, the authors state in the methods that ROIs were defined from the "relevant contrast". It is unclear what this means. Was the BA 10 ROI defined based on the interaction contrast between Self-agency and External agency? If so, then this is not a neutral contrast, and so the interaction statistic derived from these data is not a valid test of this interaction nor is it appropriate to use in the region by effect interaction. Can the authors be explicit about the contrast used to define the ROIs and why these contrasts are neutral with respect to the conditions tested using that ROI. An example of a neutral contrast would be all conditions across both onsets versus baseline.

(3) The authors response regarding lowering the threshold for the second onset was also unclear. 100 trials and 21 subjects seem like a reasonably powered study, and these were sufficient numbers to provide statistical significance at a .001 threshold in BA 10. So, why would a different threshold be necessary for the RCZ? Also, perhaps I simply misunderstood the authors' justification, but reducing a statistical threshold does not seem an appropriate method of compensating for low power. Also, the authors did not respond to my question regarding the reduction in cluster size along with a reduction in threshold for the second onset. If the authors are using cluster correction, shouldn't the cluster size threshold increase with a lower alpha threshold in order to guard against multiple comparisons? How are the authors setting their cluster correction size?

(4) In response to a comment regarding choice, the authors included the comparison of DF2 and DF3 and note that it is not reliable. This is an important addition to the results. However, the authors have also now added a line of discussion that says "Finally, this modulation is independent of the degrees of freedom of choice revealing an all-or-none choice pattern which is in line with our previous study (10)" (pg. 7). Is this in reference to the null result between DF2 and DF3? If so, isn't this something of a null inference?

Review of the second revised manuscript:
In their revision, the authors have computed separate ANOVAs for each onset, they have defined ROIs neutrally to condition, with the exception of onset, and they have provided a consistent cluster corrected threshold across contrasts. Thus, the authors have satisfied my major concerns.

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements.