Summary: How do strangers converge on a shared habit or workplace norm? New research suggests the process is neither simple imitation nor complex calculation. Instead, people first sample different behaviors and then commit decisively once a clear internal threshold is reached.
That threshold follows the Tolerance Principle, a compact mathematical rule originally formulated to explain how children learn grammar. The same principle appears to predict when people treat repeated social behavior as a rule, despite occasional exceptions.
Key Facts
- Two-stage process: Rather than immediately copying others or always choosing the statistically optimal action, people explore options probabilistically and switch to a stable rule once their experience passes an internal threshold.
- The Tolerance Principle: This simple equation defines how much consistency a person needs to observe before accepting a pattern as a rule, even when some conflicting evidence exists.
- One mechanism across domains: The cognitive mechanism that helps children learn grammatical rules—accepting regular patterns like “-ed” while tolerating irregular verbs—also appears to guide how adults adopt social conventions.
- Predicting tipping points: The model estimates how large a dissenting minority must be to overturn a dominant norm, offering a mathematical view of how social change and flips in collective behavior occur.
Source: CUNY
A new paper in the Proceedings of the National Academy of Sciences (PNAS) provides a clear, simple answer to a long-standing puzzle: how people learn and stabilize shared social conventions, from everyday habits to workplace norms.
Researchers from the CUNY Graduate Center, the University of Pennsylvania, and Stanford University show that people do not primarily learn by rote imitation or by performing complex optimization. Instead, learning follows a categorical, two-stage pattern: an initial sampling phase marked by probabilistic choices, followed by a stable commitment once enough evidence accumulates.
That commitment point is captured by the Tolerance Principle, a parameter-free mathematical rule that predicts when observers consider a behavior consistent enough to be treated as a rule despite occasional exceptions.
Originally developed to explain how children discover grammatical patterns, the Tolerance Principle also predicts adult behavior in social coordination tasks and explains how competing norms can displace one another. The same cognitive rule that helps children accept “walked” and “talked” while tolerating exceptions such as “went” appears to underlie adults’ shift from exploration to rule application in social contexts.
“People often assume social learning is about imitation or careful optimization,” said Spencer Caplan, Linguistics professor at the CUNY Graduate Center and co-lead author. “We find a more basic pattern: people sample options until one pattern reaches a ‘good enough’ threshold, then they commit and largely stick with that rule even in the face of some disagreement.”
To test this idea, the team developed computational models representing alternative learning strategies and compared their predictions with data from coordination experiments. These included previously published studies and new experiments in which participants attempted to align on shared choices—such as agreeing on a name for an unfamiliar face—while interacting through social networks. Small rewards for matching others’ responses allowed researchers to track how conventions emerged and stabilized over time.
Across multiple experiments, participants consistently diverged from the two dominant theories of social learning. They rarely copied only the most recent behavior they observed, and they did not simply maximize statistical likelihood. Instead, participants behaved probabilistically during early interactions and then transitioned sharply to consistent choices once their accumulated experience crossed the Tolerance threshold.
Compared with imitation or optimization-based models—including several Bayesian approaches—the threshold-based model more accurately reproduced observed learning dynamics. It also better predicted how many dissenters were required to overturn an established convention, offering a quantitative account of social tipping points.
The researchers further validated the approach in a preregistered dyadic experiment that required participants to infer nonlinguistic patterns in noisy environments. In that controlled setting, the Tolerance Principle model outperformed competing models at matching human learning rates.
These results suggest a single, simple cognitive mechanism may account for learning across domains—from grammatical rules in language acquisition to the spread and stabilization of social norms. The model explains why certain workplace cultures are resistant to change: once a practice surpasses the Tolerance threshold, people stop sampling and begin applying the rule, making them less responsive to isolated counterexamples. To flip a deeply rooted culture, a dissenting minority must be large enough to shift the majority’s experience back below that threshold.
The authors emphasize that future work should explore how this threshold-based mechanism operates in complex, real-world settings where identity, status, and power also shape conventions.
Key Questions Answered:
A: Humans sample before committing. You avoid locking in a behavior that could be a temporary fluke. Your mind waits until observed consistency meets the Tolerance threshold. After that point you tend to adopt the behavior as a rule and will often persist even if you later see exceptions.
A: Yes. Once a norm exceeds the threshold, people apply it automatically and discount conflicting evidence. To change the culture you need enough dissenters to alter the majority’s experience and push it back below the threshold.
A: It’s the same mathematical idea. Children hear many examples of regular forms (e.g., “walked,” “talked”) and, after enough exposure, infer the rule “-ed” for past tense. Irregulars like “went” are tolerated as exceptions once the rule has sufficient weight.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- Journal paper reviewed in full.
- Additional context added by our staff.
About this social neuroscience research news
Author: Shawn Rhea
Source: CUNY
Contact: Shawn Rhea – CUNY
Image: The image is credited to Neuroscience News
Original Research: Closed access. “A Simple Threshold Captures the Social Learning of Conventions” by Douglas Guilbeault, Spencer Caplan, and Charles Yang. PNAS
DOI: 10.1073/pnas.2508061123
Abstract
A Simple Threshold Captures the Social Learning of Conventions
A central question in cognitive and social science is how people learn social conventions from limited, noisy observations across diverse actors without explicit instruction. Dominant theories like imitation and optimization often fail to match real behavior observed in coordination experiments.
Across multiple studies, participants systematically deviate from both imitation and optimization. Instead, they follow a two-stage pattern: initial probabilistic behavior while gathering information, followed by a categorical switch when a mental threshold is reached, after which choices stabilize.
We estimate this threshold using the Tolerance Principle (TP), a parameter-free equation developed to model how children infer linguistic rules. Threshold-based agents produce social learning patterns that align more closely with human data than agents based solely on imitation or optimization, and the model also captures how a critical mass of dissenters can overturn established conventions.
The TP model outperforms a range of optimization methods, including Bayesian approaches. In a preregistered dyadic experiment involving noisy signals, TP better reproduced human learning rates than competing models. These results support the claim that a simple mathematical threshold can underlie both individual rule learning and collective convention formation.