Both sides of interpersonal emotion regulation

How do different emotion-regulation strategies affect the feelings and neural responses of both people in an interaction?

Published 2026

Journal: Emotion

I contributed across the behavioral and fNIRS analyses, including preprocessing, activation analysis, interbrain synchrony and multivariate classification.

The paired task

Original Figure 1. A and B show the paired fNIRS setup and masked participant photograph. C shows initial and final resting periods around one neutral and four negative-image blocks. For both regulator and target, a negative image and ratings precede an eight-second interaction, followed by the same image and new ratings. Separate rows preserve the cognitive-reappraisal and arithmetic-distraction procedures and all phase timings. — Figure 1, Yang et al., 2026. Thirty-four pairs of women took part. In panel C, CR denotes reinterpreting the image (cognitive reappraisal); DIS denotes mental arithmetic (distraction). The regulator offered the strategy and the target responded; both rated the same negative image before and after the interaction.

Where we measured

Original Figure 2. Frontal and lateral brain views show all sources, detectors and channels across ten regions, including prefrontal, primary motor, primary sensory, Broca and Wernicke regions. The complete color legend identifies the sampled regions. — Figure 2. Simultaneous fNIRS recordings yielded blood-oxygenation time courses from ten cortical regions per person.

Valence improved, especially after reappraisal

Valence ratings were higher after regulation and higher overall for regulators than targets. Reappraisal produced higher postregulation valence than distraction; the strategies did not differ reliably before regulation. The time-by-strategy interaction was significant, while the other valence interactions and all arousal effects were nonsignificant.

Original Figure 3. Four valence distributions show reappraisal and distraction before and after regulation. Individual points, box plots and significance brackets are intact. Postregulation reappraisal has higher ratings than postregulation distraction. — Figure 3. Pre-IER and post-IER denote ratings before and after interpersonal emotion regulation. Higher ratings indicate more positive valence. *** p < .001.

Activation depended on role and strategy

Role-by-strategy interactions appeared in the frontal pole (FP), left and right orbitofrontal cortex (lOFC, rOFC), and right Broca’s area (rBroca). Targets showed higher activation during distraction than reappraisal in all four regions. During reappraisal, regulators showed higher activation than targets in all four; during distraction, targets exceeded regulators in rOFC and rBroca.

Complete original Figure 4, including A frontal pole, B left orbitofrontal cortex, C right orbitofrontal cortex and D right Broca's area. Each panel retains the four role-by-strategy distributions, individual observations, beta-value axes and all reported significance brackets; the shared condition legend remains between the rows. — Figure 4A–D. Task-related oxygenated-hemoglobin activation estimates. * p < .05; ** p < .01; *** p < .001.

Interbrain patterns distinguished strategy and success

Correlations between the two participants’ regional time courses formed a 10 × 10 interbrain-synchronization matrix for each trial. In five-fold within-participant cross-validation, classifiers distinguished the strategy, regulator success and target success above 50% chance (all FDR-corrected p < .001). Success meant a more positive valence rating on the second viewing.

Feature-importance rankings differed across region pairs for all three classifiers. Connections involving the regulator’s frontal pole appeared repeatedly among the most informative features.

Complete original Figure 5. A shows strategy, regulator-success and target-success decoding accuracies with all points, distributions, box plots and the 0.50 chance line. B shows the ten highest-ranked interbrain region pairs for each classifier. Lower average ranks indicate greater importance; each pair names the regulator's region first and the target's region second. — Figure 5A–B. A: all three decoding tests exceeded chance. B: lower average rank indicates greater feature importance; pairs are regulator ROI × target ROI. These tests evaluated held-out trials within participants, rather than generalization to new participants.

Exploratory functional interpretation

Neurosynth analyses linked the four regions to terms from the neuroimaging literature. These associations informed an interpretation in which regulators support monitoring and guidance, while targets engage different demands when accepting a reinterpretation or doing arithmetic. This functional account remains exploratory.

Complete original Figure 6. Four panels retain the brain slices, coordinates and associated term clouds for frontal pole, left and right orbitofrontal cortex and right Broca's area. These are literature-based Neurosynth functional associations, not activation maps measured in this study's participants. — Figure 6A–D. Literature-based functional decoding: brain locations and associated cognitive terms for each region.

The paper’s proposed theoretical model

Original Figure 7, the authors' proposed model. A regulator monitoring-and-guidance system is paired with a target inhibitory-and-executive system, with interbrain synchrony between them and valence outcomes for both roles. Directional arrows express the theoretical interpretation. — Figure 7. The authors’ proposed integration of behavioral, activation and synchrony findings. Its arrows depict a candidate mechanism; the symmetric correlations do not establish causal transmission between brains.

Reported statistical comparisons

Behavior

Valence main effects: role, F(1, 231) = 69.08, η² = .23; time, F = 70.91, η² = .23; strategy, F = 36.68, η² = .14. The time × strategy interaction was F = 22.40, η² = .09 (all p < .001). Other valence interactions had Fs < 3.43, ps > .065. Arousal had no significant main effects or interactions (Fs < 3.63, ps > .058).

Regulators versus targets: t(231) = 8.31, d = 1.01; post- versus preregulation: t = 8.42, d = 1.02; reappraisal versus distraction after regulation: t = 7.63, d = 1.31 (all p < .001). Before regulation, the strategy comparison was t = .94, p = .350, d = .16.

Activation

Role × strategy interactions, all with F(1, 99): FP, 17.20, p < .001, η² = .15; lOFC, 9.30, p = .003, η² = .09; rOFC, 26.09, p < .001, η² = .21; rBroca, 10.95, p = .001, η² = .10. Strategy main effects were also reported in FP (F = 6.31, p = .014, η² = .06) and rOFC (F = 9.45, p = .003, η² = .09).

FP: targets DIS > CR, t(99) = 4.70, p < .001, d = 1.14; regulators > targets under CR, t = 4.02, p < .001, d = .98.
lOFC: targets DIS > CR, t(99) = 2.35, p = .021, d = .57; regulators > targets under CR, t = 3.02, p = .002, d = .78.
rOFC: targets DIS > CR, t(99) = 5.79, p < .001, d = 1.40; targets > regulators under DIS, t = 2.86, p = .005, d = .69; targets < regulators under CR, t = −4.37, p < .001, d = −1.06.
rBroca: targets DIS > CR, t(99) = 3.19, p = .002, d = .77; targets > regulators under DIS, t = 2.54, p = .013, d = .61; targets < regulators under CR, t = −2.14, p = .034, d = −.52.

Synchrony-based decoding

One-sample tests against .50 chance, all with t(33): strategy, 3.78, d = 1.31; regulator success, 7.41, d = 2.58; target success, 4.73, d = 1.65 (all FDR-corrected p < .001). Importance rankings differed across region pairs for strategy, regulator success and target success: χ²(99) = 592.68, 754.58 and 785.33, respectively (all p < .001).

Analysis methods and study scope

After quality checks and preprocessing, activation analyses used task-related oxygenated-hemoglobin estimates. Behavioral and activation models included a dyad random intercept. Trial-level Pearson correlations between ten regions per person supplied the interbrain features. Support-vector machines used five-fold within-participant cross-validation, with oversampling to balance classes when needed; permutation importance ranked the features.

The study recruited 68 female university students in 34 dyads. Its sampling and cortical coverage limit generalization, and the meta-analytic functional account remains a candidate explanation for further hypothesis-driven research.

Read the published article