Orbitofrontal cortex contributes to the comparison of values underlying economic choices

The study addresses whether the primate orbitofrontal cortex (OFC) contributes causally to value comparison during economic choice. Prior work established that distinct OFC neuron populations encode offer values (inputs) and chosen variables (outputs), suggesting OFC houses a decision circuit. Although offer value signals in OFC were shown to be causal to choice, the causal role of OFC in comparing values remained unproven. Competing hypotheses propose decisions occur in motor systems, through distributed processes, via attention-driven sequential comparison, by hippocampal contributions, or without explicit value comparison. The authors propose a causal test: if a manipulation increases choice variability only when comparison occurs, without degrading input valuation or motor planning, then the affected neural population participates in the comparison process. They aim to identify such a condition in OFC using weak electrical stimulation during sequential-offer choices.

• Microstimulation studies in visual area MT showed low current biases perceptual decisions, while higher current can increase choice variability by making input signals more ambiguous (interpreted as noisier inputs).
• Mouse studies found OFC inactivation increased choice variability via reversion to stereotyped strategies, implying necessity of OFC but not distinguishing valuation versus comparison roles.
• Previous monkey work with higher-current OFC stimulation during sequential offers produced range-dependent and order biases (altered offer values) and increased variability at high current during offer2, but effects could reflect stimulation of fibers or early motor planning. Thus causal evidence specifically for value comparison within OFC was lacking.

Subjects: Two male rhesus monkeys (Macaca mulatta), G (age 8, 9.1 kg) and J (age 7, 10.0 kg). Head restraint and bilateral OFC chambers were implanted. Eye position was tracked at 1 kHz.

Task: In each session, monkeys chose between two juices (A preferred over B) offered in variable amounts. Offers were presented sequentially at screen center; order randomized across AB (A then B) and BA (B then A) trials. Saccade targets’ locations were counterbalanced. Offer types were designed to discourage committing before offer2.

Stimulation: Low-current intracortical microstimulation targeted OFC (central orbital gyrus, areas 13/11). Tungsten electrodes (100–500 kΩ) were placed in gray matter; a second electrode in parallel aided depth confirmation and recording. Monopolar stimulation trains started 0–100 ms after offer onset, lasted 300–600 ms, used biphasic pulses (200 µs each, 100 µs interphase) at 125–200 Hz. Currents were 5–15 µA for this study. Stimulation was delivered during offer1 or offer2 in separate sessions, unilaterally or bilaterally, pseudo-randomly on half of non-forced trials. Stimulation sites spanned both hemispheres in both animals. Parameters were fixed within sessions.

Dataset: 91 sessions total (58 G, 33 J), with N=49 sessions for offer1 stimulation and N=42 for offer2 stimulation at 5–15 µA. Additional comparisons to prior data at 25, 50, 125 µA referenced from earlier work.

Behavioral analysis: Choices were analyzed separately for stimON and stimOFF trials using probit regressions. Primary model estimated: choice B = P(X), X = a0 + a1 log(qB/qA) + a2(orderAB − orderBA). Derived measures: relative value ρ = exp(−a0/a1); sigmoid steepness η = a1 (inversely related to choice variability); order bias ε = a2 (ε<0 favors offer1; ε>0 favors offer2). Control models assessed side bias and choice hysteresis (juice type, side, order) with additional regressors (Eqs. 2–5). Error rates and response times were also compared between stimON and stimOFF.

Effect size comparison across current levels: For each session, Δη = ηstimON − ηstimOFF; Δcv = −Δη to quantify increased variability. Δε = εstimON − εstimOFF with sign rectification for offer2 sessions; range-dependent bias measured as Pearson’s r between Δρ (or Δp) and ΔV = ΔVA − ΔVB. Effects were rectified, averaged across sessions per condition, and normalized by the maximum across current levels (≤15, 25, 50, 125 µA) to produce Fig. 5. Outliers (>3 SD) were excluded only for range-bias analyses. Analyses used Matlab.

• Weak OFC stimulation (5–15 µA) during offer2 selectively increased choice variability without affecting valuation or order bias.
• Example sessions showed unchanged ρ and ε under stimulation during either offer, but a marked decrease in η (steeper to shallower) when stimulating during offer2.
• Population results:
  • Relative value ρ: No systematic change with stimulation during offer1 (Wilcoxon p=0.92; t-test p=0.91; N=49) or offer2 (Wilcoxon p=0.75; t-test p=0.55; N=42).
  • Order bias ε: No systematic change during offer2 (Wilcoxon p=0.35; t-test p=0.48). A marginal increase during offer1 (Wilcoxon p=0.01; t-test p=0.02) without affecting overall conclusions.
  • Sigmoid steepness η (choice variability): No effect during offer1 (Wilcoxon p=0.59; t-test p=0.69); significant reduction during offer2 (Wilcoxon p=0.0011; t-test p=0.0011), indicating increased choice variability only when the comparison occurs.
• Controls:
  • Error rates and response times were not systematically altered by stimulation during either offer (e.g., error rates: offer1 Wilcoxon p=0.72, t-test p=0.99; offer2 Wilcoxon p=0.68, t-test p=0.44).
  • No significant increases in stereotyped behavior: side bias unchanged; choice hysteresis for juice type present but not increased; choice hysteresis for side and order negligible and unaltered.
• Across current levels (normalized effect sizes), choice variability exhibited a U-shaped trend: largest effects at very low (5–15 µA) and high (125 µA) currents, minimal at intermediate (25–50 µA). Range-dependent and order biases appeared at ≥25 µA as valuation effects and were absent at 5–15 µA.
• Together, weak stimulation during offer2 disrupted value comparison specifically, without altering offer value signals or motor planning.

Because value computation occurs during both offers but comparison only upon offer2, the finding that weak OFC stimulation increased choice variability exclusively during offer2 indicates a causal role of OFC in value comparison. The absence of range-dependent biases or order-bias changes at 5–15 µA argues that valuation signals were not perturbed, and unchanged errors/RTs and lack of spatial signals in OFC argue against motor-planning confounds. Mechanistically, low-current stimulation may preferentially activate inhibitory interneurons, perturbing the excitation-inhibition balance necessary for accurate comparison in recurrent networks, thereby reducing decision accuracy. The results challenge action-based models positing decisions as competitions among motor plans, and are inconsistent with sequential models like the attentional drift-diffusion model that predict effects from stimulation during either offer. Instead, they support good-based and distributed-consensus frameworks where comparison occurs in value-representing circuits like OFC. While demonstrating OFC’s involvement, the study does not exclude contributions from other regions (including motor areas).

The study provides causal evidence that neurons in primate OFC participate in value comparison during economic choices. Using weak intracortical stimulation that spares valuation and motor processes, the authors show a selective increase in choice variability when stimulation coincides with the comparison epoch (offer2). Along with prior work showing OFC’s role in value computation, these results constrain theories of decision-making, supporting good-based representations for economic choices. Future work should elucidate the organization of the OFC decision circuit, connectivity among identified OFC cell groups and with other regions, and the functional role of inhibition in comparison.

• The causal manipulation implicates OFC in value comparison but does not rule out contributions from other brain regions, including motor and premotor areas.
• Proposed cellular mechanisms (preferential activation of inhibition at low current; neuronal hijacking at high current) are inferential and were not directly measured.
• The task involves sequential offers of juices in monkeys; generalizability to other species, modalities, or decision contexts may be limited.
• Marginal effects on order bias during offer1 stimulation indicate some subtle valuation-related changes could occur in certain conditions, though they do not alter main conclusions.