Negative Verbal Probabilities Undermine Communication of Climate Science
- Mar 11
- 4 min read
IPCC's "unlikely" terminology undermines consensus perception vs. "small probability" across 8 experiments (N=4,150), weakening when projections exceed expectations.
A December 2025 study in Nature Climate Change by Marie Juanchich and colleagues demonstrates that the IPCC's recommended use of negative verbal probability terms (such as "unlikely") to describe low-probability climate outcomes inadvertently undermines public perception of scientific consensus and evidence quality. Across eight preregistered experiments with 4,150 UK participants, the researchers found that outcomes described with negative low-probability terms are perceived as reflecting lower scientific consensus than probabilistically equivalent but positively framed terms (such as "a small probability").
The effect persists after controlling for climate change beliefs, IPCC familiarity, and political orientation, although it weakens when projected values exceed participants' personal expectations. Participants also associate negative low-probability terms more strongly with extreme outcomes beyond model ranges and judge them as less evidence-based than positive counterparts, suggesting that IPCC's current uncertainty terminology may inadvertently fuel climate skepticism by implying scientific disagreement.
Key Findings: How Word Choice Shapes Consensus Perception
Negative Terms Signal Disagreement Despite Equivalent Probabilities
Experiments 1 and 2 established that "unlikely" and "the likelihood is low" (IPCC-recommended terms for <33% probability) convey similar numerical probabilities as positive alternatives ("there is a small probability/possibility"), with all four terms peaking around 15-20% probability estimates. However, participants associated negative terms with reasons against the outcome occurrence (91% for "unlikely," 86% for "likelihood is low") versus reasons supporting occurrence for positive terms (43% and 21% respectively).
In consensus perception tasks, scientists using "unlikely" were perceived as disagreeing with colleagues 53% of the time versus only 26% for "small probability"—a perception gap that persisted across multiple experimental contexts including newspaper headlines and IPCC-attributed statements.
Negative Probabilities Linked to Extreme, Out-of-Range Outcomes
When asked to select which sea-level rise magnitude was "unlikely" from IPCC projection graphs, 61% of participants selected outcomes outside model confidence intervals (>1m or <0.1m), compared to only 21-23% for positive probability terms. This extremity association held even when 86% of participants correctly answered graph comprehension questions, demonstrating the effect was not due to misunderstanding data distributions.
The tendency to select extreme outcomes correlated with perceiving disagreement: 46% of those selecting out-of-range values believed the speaker disagreed with other scientists, versus 30% of those selecting in-range values. Experiment 5 showed this pattern extends to numerical probabilities (10-33%), which like positive verbal terms directed attention toward outcome occurrence (79%) rather than non-occurrence (39% for "unlikely").
Evidence Perception Weakens with Negative Framing
Experiment 3 found participants judged "unlikely" temperature projections as having lower scientific evidence (scale 0-100) than "small possibility" projections for Australian temperature increase (significant difference, Cohen's d = 0.77) though not for African precipitation decrease.
Experiment 4 revealed this evidence perception gap exists for both "unlikely" high-value projections and their "likely...not" complements: participants believed more scientists agreed with "likely that temperature will not reach 3°C" projections (mean 40.35 out of 100) than with equivalent positive framings (mean 59.42-61.79), despite formal logical equivalence. This undermining of perceived evidence quality occurred even when statements were explicitly attributed to IPCC (Experiment 5), suggesting the terminology's implicit meaning overrides source credibility.
Why This Matters: Prior Beliefs and Boundary Conditions
Context and Expectations Moderate Effects: Experiment 6 demonstrated the directionality effect depends critically on how projections align with prior beliefs. For low-emission temperature projections (where participants expected more warming than projected), negative versus positive framing showed large consensus perception gaps (F(1,754) = 19.74, ηP² = 0.03 for interaction). However, for high-emission scenarios exceeding expectations, directionality effects disappeared or reversed, with both framings perceived similarly. This pattern suggests negative probabilities most strongly undermine consensus perception when describing outcomes people expect to occur—precisely the scenarios where communicating scientific agreement matters most for motivating preparedness.
Positive Terms Maintain Consensus Without Sacrificing Accuracy: The complementary "likely" framing recommended as an alternative performs poorly on consensus perception: "it is likely that global warming will not reach 3°C" generated similar low consensus estimates (40.35 out of 100 scientists) as "unlikely to reach 3°C" (42.10), both significantly lower than positive low-probability framings (59.42 for "small probability," 61.79 for "10-33% probability"). This occurs because negations ("will not reach") carry negative directionality regardless of the probability term. Positive affirmative low-probability expressions align with numerical probability ranges, are regarded as more scientific and trustworthy, and avoid directing attention away from potentially impactful low-probability events that require preparedness despite technical "smallness."
Individual Differences Show Limited Protection: The effects proved robust across individual differences in climate change belief, IPCC knowledge, education level, and political conservatism (Experiments 4-6). Self-assessed climate expertise did not attenuate directionality effects in representative UK samples, though further research should test generalization to climate scientists and policymakers. The consistency across demographics suggests the implicit pragmatic meaning of negative verbal probabilities—that speakers use them to voice disagreement or describe unsupported claims—operates at a fundamental linguistic level that overrides explicit knowledge about probability communication conventions.
Beyond Technical Accuracy: Pragmatic Implications
The research reveals a critical gap between technical probability equivalence and pragmatic communication effects. In everyday discourse, negative verbal probabilities signal disagreement with over-confident speakers: a 30% probability was more often described as "unlikely" when following a 50% estimate but as "a chance" when following a 10% estimate. This conversational function persists when negative terms enter scientific communication, where they inadvertently suggest outcomes are exceptional, lack evidentiary support, and represent minority scientific positions.
The IPCC Sixth Assessment Report used "likely" 26 times more often than "unlikely," suggesting authors intuitively recognize these issues, but this creates its own problems: focusing on complementary high-probability outcomes ("likely that severe drought will not occur") neither raises awareness nor triggers appropriate concern about the low-probability hazards themselves. The authors recommend using affirmative low-probability expressions that refer directly to target events, avoid negations, and align with numerical probability ranges while maintaining comparable uncertainty levels without undermining perceptions of scientific consensus and evidence.
Sign up for our newsletter or connect with us on social media to stay up-to-date with our latest posts and permaculture inspiration.
Explore our inspiring series and posts:
Love the post? Share it with your circle, inspire your people.