TMS-fMRI evidence that magnitude representations in IPS causally drive risky choice
Poster No:
907
Submission Type:
Abstract Submission
Authors:
Marius Moisa1, Gilles de Hollander2, Christian Ruff3
Institutions:
1University of Zurich, Zürich, Zurich, 2UZH, Zürich, Switzerland, 3University of Zurich, Zurich, Zurich
First Author:
Co-Author(s):
Introduction:
Risk appetite refers to the extent to which decision-makers are willing to choose options that increase potential returns at the cost of a higher variance of those returns. A recent study showed that individuals with a noisier approximate number sense (ANS) are less consistent in making risky choices and, notably, tend to be more risk-averse. This result highlights a crucial influence of perceptual processes on apparent risk appetite that defies mainstream economic theory(1). Here, we aimed to elucidate the causal relevance of the parietal ANS by perturbing it using theta-burst TMS2 over intraparietal cortex (IPS).
Methods:
78 participants performed a risky choice task while undergoing fMRI, choosing between a sure option with a fixed payout and a risky option with a 55% probability of a larger payout. We estimated a numerical population receptive field model (nPRF) in parietal cortex (1, 3) quantifying which voxels in IPS are reliably tuned to a preferred numerosity (Fig. 2A). Thirty-five participants showed reliable numerosity fields, no random behavior, and no adverse effects to TMS. These subjects underwent 2 more sessions, where rTMS was applied either over their individual nPRF cluster in IPS, or over vertex, immediately before performing risky choices.
Results:
We observed no difference in the raw risky choice proportion between the IPS and vertex stimulation conditions (Fig. 1A). However, when we split choices by the order in which the risky and safe options were presented, IPS stimulation increased the proportion of risky choices where the safe option was presented first from 56.3% to 61.6% (F(1, 34)=5.00, p=0.032, Fig. 1B).
Choice proportions can be modulated by both a shift in preference or in the noisiness of the response. To differentiate between these two accounts, we also modeled choice data using a psychometric model, yielding for each subject both (1) an indifference point (a measure of risk appetite) and (2) a slope (a measure of choice consistency) (Fig. 1C). This analysis shows that subjects after IPS stimulation are generally more risk-seeking (pmcmc=0.022) and marginally more noisy (pmcmc=0.066). Crucially, we show that both effects are driven by the trials on which the safe options were presented first. That is, only in trials where the safe option was presented first, both risk appetite (pmcmc=0.001) and choice consistency (pmcmc=0.013) were significantly decreased by IPS stimulation. The interaction effect between presentation order and stimulation condition was significant for both risk appetite (pmcmc=0.002) and choice consistency (pmcmc=0.0338).
A Bayesian model of risky choice can explain the interaction between order and TMS stimulation as an increase in noise for small, but not larger payoff magnitudes after TMS. This is consistent with the large majority of nPRFs in parietal cortex having preferred numerosities that are relatively low compared to the risky payoff magnitudes.
We also probed the effect of TMS on neural magnitude representations in IPS. We found significantly lower nPRF amplitudes after IPS stimulation (t(33) = 2.71, p=0.011) compared to vertex stimulation (Fig. 2B), but no difference in the average preferred numerosity (t(33) = 1.32, p=0.19) or the dispersion of the nPRFs (t(33) = 0.82, p=0.418). This highlights a specific effect of TMS on the amplitude of the nPRF. Accordingly, an inverted nPRF model showed higher decoding accuracy in the vertex vs the IPS condition (mean r=0.142 versus r=0.092, F(1,34) = 4.99, p=0.032, Fig. 2C), consistent with the noisier behavior of the subjects.
Choice proportions can be modulated by both a shift in preference or in the noisiness of the response. To differentiate between these two accounts, we also modeled choice data using a psychometric model, yielding for each subject both (1) an indifference point (a measure of risk appetite) and (2) a slope (a measure of choice consistency) (Fig. 1C). This analysis shows that subjects after IPS stimulation are generally more risk-seeking (pmcmc=0.022) and marginally more noisy (pmcmc=0.066). Crucially, we show that both effects are driven by the trials on which the safe options were presented first. That is, only in trials where the safe option was presented first, both risk appetite (pmcmc=0.001) and choice consistency (pmcmc=0.013) were significantly decreased by IPS stimulation. The interaction effect between presentation order and stimulation condition was significant for both risk appetite (pmcmc=0.002) and choice consistency (pmcmc=0.0338).
A Bayesian model of risky choice can explain the interaction between order and TMS stimulation as an increase in noise for small, but not larger payoff magnitudes after TMS. This is consistent with the large majority of nPRFs in parietal cortex having preferred numerosities that are relatively low compared to the risky payoff magnitudes.
We also probed the effect of TMS on neural magnitude representations in IPS. We found significantly lower nPRF amplitudes after IPS stimulation (t(33) = 2.71, p=0.011) compared to vertex stimulation (Fig. 2B), but no difference in the average preferred numerosity (t(33) = 1.32, p=0.19) or the dispersion of the nPRFs (t(33) = 0.82, p=0.418). This highlights a specific effect of TMS on the amplitude of the nPRF. Accordingly, an inverted nPRF model showed higher decoding accuracy in the vertex vs the IPS condition (mean r=0.142 versus r=0.092, F(1,34) = 4.99, p=0.032, Fig. 2C), consistent with the noisier behavior of the subjects.
·Figure 1
·Figure 2
Conclusions:
We provide evidence for a crucial causal role of neural magnitude representations in the IPS in risky choices. The results corroborate a neurocognitive account of risk aversion where risk preferences are determined by noisy, biased perception of magnitudes. This highlights the relevance of 'low-level' perceptual processes and their neurocognitive substrates for economic choice theory.
Brain Stimulation:
TMS 2
Higher Cognitive Functions:
Decision Making 1
Space, Time and Number Coding
Keywords:
Modeling
Transcranial Magnetic Stimulation (TMS)
Other - Numerosity
1|2Indicates the priority used for review
Provide references using author date format
2. Y. Huang et al. Theta burst stimulation of the human motor cortex, Neuron, 2005
3. B. Harvey & S. Duomolin, Can responses to basic non-numerical visual features explain neural numerosity responses?, Neuroimage, 2017