🧬 biology

Fear Recognition in Healthy and Clinical Adults: A Systematic Review and Meta-analysis

This systematic review and meta-analysis of 40 studies reveals that while pooled fear recognition accuracy in healthy adults is highly variable and largely driven by methodological differences in response options, the consistent confusion between fear and surprise across cultures suggests it may serve as a valuable differential indicator for future standardized clinical assessment.

Original authors: A. T. Martins, F. R. Carvalho, J. Russo, C. Campos, J.A. Camacho-Ruiz, C.M. Galvez-Sánchez, J. Rodrigues, A. Gomes

Published 2026-08-13

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: A. T. Martins, F. R. Carvalho, J. Russo, C. Campos, J.A. Camacho-Ruiz, C.M. Galvez-Sánchez, J. Rodrigues, A. Gomes

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine your brain is a super-fast security guard stationed at the front door of your mind. Its main job? To scan the faces of everyone walking by and instantly shout out what they are feeling: "Happy!" "Angry!" "Sad!" This ability, called facial emotion recognition, is like the universal language of human connection. It helps us know when a friend needs a hug, when a stranger is dangerous, or when to laugh at a joke. For decades, scientists have believed that our faces have a set of "basic" emotions—like a standard keyboard with six main keys (happiness, sadness, anger, fear, disgust, and surprise)—that everyone, everywhere, presses to send the same message. But lately, researchers have started to wonder: is this keyboard really universal, or are some keys just harder to press than others? And what happens when our security guard gets tired, sick, or confused?

This is where a new, massive detective story comes in. A team of scientists from Portugal and Spain decided to act as "meta-detectives." Instead of running their own experiments, they gathered and analyzed 40 different studies published between 2016 and 2025. They wanted to solve a specific mystery: How good are adults at recognizing the emotion of fear? Fear is a special case because it's the emotion that screams "DANGER!" It's the one our brains are wired to spot the fastest to keep us safe. But, strangely, it's also the hardest one to get right. The researchers wanted to know: Is the difficulty real, or is it just a result of the questions we ask? And when people get it wrong, what do they think they are seeing instead?

The Great Fear Confusion

The researchers looked at data from over 1,300 healthy adults (the "control group" who have no brain or mental health issues) and found something wild. If you asked these people to identify fear, they got it right only about 54.5% of the time. That's barely better than flipping a coin! But here's the twist: the results were all over the place. In some studies, people got it right 94% of the time; in others, they only got it right 19% of the time.

Why such a huge gap? The paper suggests it's mostly a result of the game, not a lack of skill. It turns out that the number of choices you give someone changes the score.

The "Easy Mode" Trap: In studies where people only had to choose between fewer than six options (like "Is this fear or not?"), the accuracy jumped to 80.7%.
The "Hard Mode" Reality: In studies where people had to choose from six or more options (like "Is this fear, surprise, anger, sadness, disgust, or happiness?"), the accuracy dropped to 47.8%.

The authors explain that this isn't because people are suddenly bad at spotting fear when there are more choices. It's because the extra choices include surprise. When you have to pick between fear and surprise, people get confused a lot. It's like trying to tell the difference between a "Startled Cat" and a "Scared Cat." Both have wide eyes and raised eyebrows! The study found that when healthy adults did get fear wrong, they confused it with surprise 32.3% of the time. This confusion was rock-solid; it happened the same way in different cultures and with different types of pictures, suggesting that our brains genuinely struggle to tell these two "wide-eyed" emotions apart until they see more clues (like the mouth).

The Clinical Clues

The paper also peeked into how this works for people with certain health conditions. They found that people with Parkinson's disease, amyotrophic lateral sclerosis (ALS), and major depression consistently struggle to recognize fear more than other emotions. It's like their security guard is missing a specific tool for spotting that particular danger signal.

However, the story gets even more interesting with schizophrenia. For these individuals, the confusion isn't just between fear and surprise. Sometimes, they mix up fear with disgust or anger. This suggests their "emotional keyboard" might be rearranged in a different way than healthy people's. On the flip side, people with PTSD (Post-Traumatic Stress Disorder) have a different problem: they don't necessarily miss fear; instead, they see fear (or anger) in faces that are actually neutral. It's like their security guard is so on high alert that they think a calm face is a threat.

What This All Means

The big takeaway from this massive review is that we can't just look at a single number and say, "People are 54% good at fear." The way we ask the question matters immensely. If we want to truly understand how the brain handles fear, we need to stop just counting correct answers and start looking at where people get confused.

The authors suggest that the "fear vs. surprise" mix-up is a reliable clue. Because it happens so consistently, it could be a useful tool for doctors. If a patient's brain starts mixing up fear and surprise in a weird new way, or if they stop mixing them up entirely and start confusing fear with something else (like disgust), it might signal a specific type of brain change.

In short, recognizing fear is a tricky business. Our brains are wired to spot the "wide eyes" of danger, but that same feature makes us stumble when we have to choose between "scared" and "surprised." By understanding these little glitches in our emotional radar, scientists hope to build better tools to help diagnose and treat conditions where the brain's security guard starts to malfunction. The paper doesn't claim to have solved the mystery of fear, but it has definitely handed us a much better map of where the confusion lies.

Technical Summary: Fear Recognition in Healthy and Clinical Adults

Problem Statement
Facial emotion recognition (FER) is a critical sociocognitive capacity, yet the literature regarding the recognition of fear remains fragmented due to significant methodological and cultural heterogeneity. While fear is evolutionarily distinct as a signal of imminent threat and is associated with specific neural pathways (e.g., amygdala activation), it is paradoxically the most difficult basic emotion to recognize under standardized conditions. A persistent issue in the field is the systematic confusion between fear and surprise, likely due to overlapping facial action units (AUs). Previous reviews have focused on specific clinical conditions or cross-cultural comparisons but have not simultaneously synthesized fear recognition accuracy, confusion patterns, and methodological moderators across both healthy and clinical adult populations. This review addresses the lack of a unified quantitative synthesis that accounts for the variability in recognition rates and the specific nature of misclassification errors.

Methodology
This study is a PRISMA 2020-compliant systematic review and meta-analysis covering literature published between 2016 and 2025.

Search & Selection: A search of PubMed/MEDLINE and Scopus yielded 2,835 records. After removing duplicates and screening titles/abstracts, 46 full-text articles were assessed. The final sample included 40 studies (21 contributing to the quantitative meta-analysis of healthy controls).
Inclusion Criteria: Studies must assess facial fear recognition in adults (≥18 years), report quantifiable fear metrics (accuracy, hit rates, etc.), and use original empirical methods. Studies were excluded if they focused solely on children, lacked a healthy control group (unless comparing clinical subgroups), or relied exclusively on non-facial modalities.
Risk of Bias: A hybrid PROBAST/AXIS tool was applied to all 40 studies. No studies were rated "LOW" risk; 33 were "UNCLEAR" and 7 were "HIGH." A continuous Quality Score (QS) was derived for each study.
Statistical Analysis:
- Meta-analysis: A DerSimonian–Laird random-effects model with logit transformation was used to pool fear accuracy in healthy controls (k=21, n=1369).
- Subgroup Analyses: Pre-specified moderators included stimulus modality (static vs. dynamic), cultural background (Western vs. East Asian), publication era (pre- vs. post-2020), and the number of response options (<6 vs. ≥6).
- Confusion Analysis: Pooled rates of fear misclassification (e.g., fear → surprise) were calculated using data from studies reporting complete confusion matrices (k=3).
- Sensitivity: Publication bias was assessed via funnel plots and Egger's regression. Leave-one-out analyses and quality-restricted pools were conducted to test robustness.

Key Contributions

Quantification of Heterogeneity: The study provides the first pooled estimate of fear recognition accuracy in healthy adults, revealing a global mean of 54.5% with extreme heterogeneity ( $I^2 = 87.6\%$ ).
Identification of the Primary Moderator: The analysis identifies the number of response options as the dominant methodological driver of accuracy variance. Paradigms with ≥6 options (typically including "surprise") yielded significantly lower accuracy (47.8%) compared to those with <6 options (80.7%).
Stability of Confusion Patterns: Despite the massive variability in absolute accuracy, the direction of misclassification is highly stable. Fear is consistently confused with surprise at a rate of 32.3% ( $I^2 = 4\%$ ) across different methods and cultures.
Clinical Synthesis: The paper narratively synthesizes fear recognition deficits across various clinical populations (e.g., Parkinson's disease, schizophrenia, MDD, PTSD), highlighting distinct patterns such as the "threat interpretation bias" in PTSD (misinterpreting neutral faces as threatening) versus the fear-surprise confusion in healthy controls.

Results

Overall Accuracy: The pooled fear accuracy in healthy controls is 54.5% (95% CI: 45.8–62.9). The observed range spans from 19.4% to 94.3%, indicating that raw accuracy is heavily dependent on experimental design rather than intrinsic ability alone.
Response Options Effect: The gap between paradigms with fewer than six options (80.7%) and those with six or more (47.8%) is approximately 33 percentage points. This suggests that the inclusion of "surprise" as a response option creates a specific error pathway (fear → surprise) that lowers raw accuracy scores.
Confusion Matrix: In the three studies with complete matrices, fear → surprise was the predominant error (32.3%). Other confusions (fear → disgust, sadness, anger, happiness) were negligible (<9%).
Clinical Findings:
- Parkinson's, ALS, MDD: Consistent deficits in fear recognition are reported.
- Schizophrenia: Deficits are present, but the confusion pattern may shift (e.g., fear confused with disgust rather than surprise), suggesting a pathological reorganization of the emotional perceptual space.
- PTSD: The deficit manifests as a bias where neutral expressions are misidentified as threatening (fear/anger/disgust), rather than a failure to recognize fear itself.
- Stimulus Modality: Dynamic stimuli (videos/morphs) generally yield higher accuracy and may attenuate deficits observed with static stimuli in clinical groups.

Significance and Claims
The paper argues that the "difficulty" of recognizing fear is not an intrinsic property of the emotion but a structural property of the perceptual space shared with surprise. The stability of the fear → surprise confusion (32.3%) across diverse methodologies suggests that both expressions share early, biologically salient signals (upper-face AUs: brow raise, eye widening) that are transmitted before lower-face cues (lip stretching) can disambiguate them.

The authors claim that:

Aggregate accuracy pools are misleading: Due to high heterogeneity ( $I^2 \approx 88\%$ ), global averages should be interpreted with caution. Subgroup analyses, particularly regarding response options, are more informative.
Methodological Standardization is critical: Raw accuracy is an inadequate metric for cross-study comparison. Future research must report unbiased hit rates and complete confusion matrices to distinguish between true recognition deficits and methodological artifacts.
Clinical Utility: The fear-surprise confusion serves as a potential differential marker. While the confusion is stable in healthy controls, deviations from this pattern (e.g., fear confused with disgust in schizophrenia) or the presence of threat-interpretation biases (in PTSD) may offer specific diagnostic insights.
Ecological Validity: The pandemic era (post-2020) introduced facial masks as a variable, with studies showing significant reductions in fear recognition under masked conditions, necessitating this as a factor in future ecological studies.

The paper concludes that fear recognition is a clinically sensitive ability, but its assessment requires standardized procedures and detailed error analyses to be useful for diagnosing and monitoring neuropsychiatric disorders.

The Great Fear Confusion

The Clinical Clues

What This All Means

Technical Summary: Fear Recognition in Healthy and Clinical Adults

More like this