Elsevier

NeuroImage

Volume 65, 15 January 2013, Pages 456-465
NeuroImage

Similar neural mechanisms for perceptual guesses and free decisions

https://doi.org/10.1016/j.neuroimage.2012.09.064Get rights and content

Abstract

When facing perceptual choices under challenging conditions we might believe to be purely guessing. But which brain regions determine the outcome of our guesses? One possibility is that higher-level, domain-general brain regions might help break the symmetry between equal-appearing choices. Here we directly investigated whether perceptual guesses share brain networks with other types of free decisions. We trained an fMRI-based pattern classifier to distinguish between two perceptual guesses and tested whether it was able to predict the outcome of similar non-perceptual choices, as in conventional free choice tasks. Activation patterns in the medial posterior parietal cortex cross-predicted free decisions from perceptual guesses and vice versa. This inter-changeability strongly speaks for a similar neural code for both types of decisions. The posterior parietal cortex might be part of a domain-general system that helps resolve decision conflicts when no choice option is more or less likely or valuable, thus preventing behavioural stalemate.

Highlights

► Medial parietal activation patterns cross-predict free decisions and guesses. ► These results speak for a similar neural code for both types of decisions. ► Medial parietal cortex might be part of a system that resolves decision conflict.

Introduction

Perceptual decision-making is essential for our every-day life and seems to happen effortlessly when sufficient sensory information is available. Perceptual decision-making under varying sensory conditions has been intensively studied using single-cell recordings in monkeys (Britten et al., 1996, Gold and Shadlen, 2007, Shadlen and Newsome, 2001) and functional magnetic resonance imaging (fMRI) and electroencephalography (EEG) in humans (Heekeren et al., 2004, Heekeren et al., 2008, McKeeff and Tong, 2007, Philiastides and Sajda, 2006, Philiastides and Sajda, 2007, Serences and Boynton, 2007). One challenging question, however, is how we make decisions in the absence of sufficient information when stimuli are very difficult to see. This problem is exacerbated by the absence of “priors” in many perceptual tasks, where choice alternatives are equally likely and rewarding. Participants are thus frequently faced with a ‘symmetry situation’: How do we decide between equal options without getting stuck like Buridan's mythical donkey, starving in the middle between two haystacks? One possibility is that we rely on the same sensory neural systems as for veridical perception but that random noise fluctuations in the sensory system determine trial-by-trial choices (Shadlen et al., 1996, Swets, 1961). Another possibility is that domain-general regions are used to break the symmetry between equal options (Deco and Romo, 2008).

Using fMRI in combination with pattern classification techniques (Haynes and Rees, 2006, Norman et al., 2006) we recently assessed which brain regions predict perceptual choices under high and low visibility. We found that under low visibility, perceptual choices about objects could be decoded from medial posterior parietal cortex, but not from sensory brain regions (Bode et al., 2012). This parietal region was clearly distinct from sensory areas in the lateral-occipital complex (LOC) in ventral visual cortex, which encoded sensory information about object categories (Haxby et al., 2001, Williams et al., 2007). The medial parietal area strongly overlapped with a network that was recently found to be involved in free decision-making (Soon et al., 2008). Thus, we reasoned that guessing and free decisions might share similar brain networks, a hypothesis that is yet to be tested.

Here, we conducted an event-related fMRI study in which participants made category choices about pianos and chairs using three different conditions. In all conditions participants saw brief, repeated mask–target–mask sequences. (1) Perceptual choices under high visibility: On each trial the sequence contained an object image (piano or chair) that was clearly visible. Participants were asked to identify the category of the presented object. (2) Perceptual choices under “low visibility” (perceptual guesses): Participants were given the same task but the timing of the stimuli was chosen to yield strong masking and the sequence only contained a neutral noise image instead of an object. The participants were unaware of this manipulation (as confirmed with post-experimental questionnaires and interviews) and thus believed to be making normal (albeit difficult) perceptual guesses. Thus, this condition was in fact a zero-visibility condition. (3) Free choices: The presentation was identical to the “low visibility” condition, again ensuring that no visual bias could drive the decisions. Participants were asked to spontaneously and freely choose either “piano” or “chair”, whatever category came first to their minds. They were instructed to ignore the visual stimulus display (even though it did not contain any differential visual information). We then used a “searchlight” decoding approach (Haynes et al., 2007, Kriegeskorte et al., 2006) to search for brain regions that encoded participants' category choices in all three conditions. In order to identify potential neural representations of the choices, we identified regions that exhibit different response patterns to the different choice options. This classification-based approach is frequently used in order to identify information-coding neural response patterns. The crucial analysis used activation patterns associated with decision outcomes from perceptual guessing to cross-predict free decisions, and vice versa. This analysis aimed to search for any brain region in which the choice-predictive activation patterns were inter-changeable between the two conditions. Finding such a brain region would not provide conclusive evidence for a shared neural code for both decisions, as also non-decision related processes could systematically differ between the choice options. It would be a strong indicator, however, that such pattern similarity may reflect important similarities in decision-making.

Section snippets

Participants

Sixteen right-handed participants with normal or corrected to normal visual acuity gave written informed consent and participated in the study. The experiment was approved by the local ethics committee and was conducted according to the Declaration of Helsinki. One participant indicated that he had noticed the missing object images and was excluded. The data from the remaining 15 participants (7 female; mean age 25 years; range 21–28) were used for the analyses.

Stimuli and experimental procedure

The stimuli were 24 pictures of

High visibility condition

The average hit rate for both visible pianos (85.2%) and chairs (91.5%) was very high, indicating that the sensory information was used to make decisions (Table 1). Second, the searchlight classification analysis revealed that the category of highly visible object stimuli was encoded in the bilateral visual cortex/LOC (accuracy 61%; Table 4). Third, replicating our earlier findings (Bode et al., 2012), choices under high visibility could exclusively be decoded from the LOC (p < .05, FWE-corrected

Discussion

In summary, we demonstrated that a region in the precuneus, extending laterally to the inferior parietal cortex and medially to the posterior cingulate cortex, encoded guesses when participants were required to make perceptual decisions about invisible objects. Supporting our a priori hypothesis, we showed that activation patterns for guesses and free decisions in these regions allowed for the prediction of outcomes from one type of decision from activation patterns associated with the other.

Conclusion

Our study is the first to demonstrate that perceptual guesses can be predicted from neural patterns for free decisions and vice versa. Whilst it is not possible to interpret the similarity in fine-grained activation patterns as unambiguously and directly representing the neural code of the decisions, the most likely interpretation of our results is that it points towards a common mechanism for internal decisions when external input cannot be used to resolve a decision conflict. Thus, we

Acknowledgments

The authors thank Sabrina Walther, Anna H. He and Martin Hebart for valuable support with data acquisition and discussions. This work was funded by the Max-Planck-Society, the German Research Foundation (DFG-Grant HA-5336/1-1), the Bernstein-Computational-Neuroscience-Program (BMBF-Grant 01GQ0411) and the Excellency Initiative (DFG Grant GSC86/1-2009) of the German Federal Ministry of Education and Research.

References (54)

  • F. Pereira et al.

    Machine learning classifiers and fMRI: a tutorial overview

    NeuroImage

    (2009)
  • M.E. Raichle et al.

    A default mode of brain function: a brief history of an evolving idea

    NeuroImage

    (2007)
  • L. Reddy et al.

    Reading the mind's eye: decoding category information during mental imagery

    NeuroImage

    (2010)
  • F. Schüür et al.

    What are self-generated actions?

    Conscious. Cogn.

    (2011)
  • P.L. Smith et al.

    Psychology and neurobiology of simple decisions

    Trends Neurosci.

    (2004)
  • C.K. Abbey et al.

    Classification image analysis: estimation and statistical inference for two-alternative forced-choice experiments

    J. Vis.

    (2002)
  • D. Badre et al.

    Is the rostro-caudal axis of the frontal lobe hierarchical?

    Nat. Rev. Neurosci.

    (2009)
  • S. Bode et al.

    Tracking the unconscious generation of free decisions using ultra-high field fMRI

    PLoS One

    (2011)
  • M. Brett et al.

    Region of interest analysis using an SPM toolbox

    NeuroImage

    (2002)
  • K.H. Britten et al.

    A relationship between behavioral choice and the visual responses of neurons in macaque MT

    Vis. Neurosci.

    (1996)
  • A.E. Cavanna et al.

    The precuneus: a review of its functional anatomy and behavioural correlates

    Brain

    (2006)
  • R.M. Cichy et al.

    Imagery and perception share cortical representations of content and location

    Cereb. Cortex

    (2011)
  • J. Duncan

    An adaptive coding model of neural function in prefrontal cortex

    Nat. Rev. Neurosci.

    (2001)
  • J.K. Fitzgerald et al.

    Generalized associative representations in parietal cortex

    Nat. Neurosci.

    (2011)
  • J.I. Gold et al.

    The neural basis of decision making

    Annu. Rev. Neurosci.

    (2007)
  • K. Grill-Spector et al.

    The dynamics of object-selective activation correlate with recognition performance in humans

    Nat. Neurosci.

    (2000)
  • D.A. Gusnard et al.

    Searching for a baseline: functional imaging and the resting human brain

    Nat. Rev. Neurosci.

    (2001)
  • Cited by (35)

    • Free Wally: Where motor intentions meet reason and consequence

      2019, Neuropsychologia
      Citation Excerpt :

      Specifically, self-paced wrist extensions, finger flexions or button presses (for an overview: see Saigle et al., 2018). Although some studies investigate spontaneous decisions rather than movements (e.g. Bode et al., 2013; Soon et al., 2013), it is the general element of spontaneity that seems most important for actions to be considered free (Deutschländer et al., 2017; Haggard, 2019). However, it is also the element of spontaneity that makes these actions most trivial (Banks and Pockett, 2007).

    • Freely chosen and instructed actions are terminated by different neural mechanisms revealed by kinematics-informed EEG

      2019, NeuroImage
      Citation Excerpt :

      In these comparisons, volition has been associated with brain structures, especially in the medial prefrontal cortex, that exhibit elevated activity when an action is self-selected relative to when that action is instructed (Krieghoff et al., 2011; Brass et al., 2013; Zapparoli et al., 2017). However, these findings have been controversial and the topic of an extensive debate due to the difficulty in specifying and interpreting how actions were “self-selected” in these experiments and the numerous confounding differences (Lau et al., 2004; Botvinick et al., 2004; Nachev et al., 2005; Nachev and Husain, 2010; Passingham et al., 2010; Schüür and Haggard, 2011; Bode et al., 2013; Brass et al., 2013). An example of such a confounding difference is the sensory and perceptual demand to identify a stimulus-defined instruction that has no equivalent for freely selected actions.

    • Perceived freedom of choice is associated with neural encoding of option availability

      2018, NeuroImage
      Citation Excerpt :

      However, the T1 structural image was co-registered to the mean of the realigned functional images and then spatially normalised to MNI space using the segment process of SPM8 to obtain normalisation parameters for the MVPA accuracy maps. MVPA was performed using custom scripts run in MATLAB (Bode et al., 2013) in conjunction with SPM8. A first-level general linear model (GLM) was first estimated for each individual participant based on motion corrected, non-normalised, unsmoothed data to produce parameter estimates for each voxel in each run and each condition.

    View all citing articles on Scopus
    View full text