Facial Feedback Hypothesis: Research and Replication

Author: Angela Bogdanova · Published: October 13, 2026 · Editorial Policy

Research cluster: Emotion & Affect Science · Written by Artificial Sapiens

The facial feedback hypothesis proposes that movements of the face can influence emotional experience. It asks whether smiling, frowning, or producing another facial configuration can cause a change in feeling, rather than merely express a feeling that already exists. The strongest contemporary answer is qualified: some experimentally induced facial movements produce small changes in reported emotion, but the size and reliability of the effect depend on how the movement is created and how feeling is measured.

In 2016, 17 laboratories failed to reproduce the famous pen-in-mouth effect. A 2019 meta-analysis nevertheless found small and variable overall facial-feedback effects, while the 2022 preregistered Many Smiles Collaboration found evidence for increased reported happiness after deliberate or mimicked smiling, with less conclusive findings for the pen method. These results make the hypothesis a valuable example of how psychology separates an appealing theory from the precise predictions that survive replication. Wagenmakers et al. (2016); Coles, Larsen, and Lench (2019); Coles et al. (2022).

What Is the Facial Feedback Hypothesis?

Facial feedback refers to the proposed causal contribution of facial activity to feeling or emotion-related judgment. Consider two sequences. In one, a funny event produces amusement and the person smiles. In another, the person is instructed to activate smiling-related muscles, and the experiment asks whether amusement or happiness subsequently changes. Only the second sequence directly tests a feedback effect.

The claim belongs to embodied approaches to emotion, which examine the interactions among experience, bodily processes, appraisal, attention, and behavior. It does not imply that the face is the sole origin of an emotion. Nor does it require every smile to be pleasant or every frown to be unpleasant. Facial movements arise in social interactions, tasks, habits, communication, and deliberate performances as well as emotional episodes.

Scientists distinguish a strong claim, that a particular facial action reliably creates a particular emotion, from a weaker claim, that some facial actions can sometimes modify emotional ratings. The experimental evidence bears more convincingly on the qualified second claim. Even when an intervention increases average reported happiness, the change can be modest, context-dependent, and different across individuals.

Emotion, feeling, mood, and expression

An emotion is commonly studied as a coordinated or interacting set of subjective, cognitive, physiological, motivational, and behavioral processes; the precise theoretical architecture remains debated. Feeling refers to subjective experience, while mood ordinarily concerns a more diffuse and often longer-lasting affective state. An expression is an observable action, such as a smile, that may be part of emotion and may serve other functions.

This matters because most facial-feedback studies measure brief self-reports or stimulus judgments. Higher ratings of cartoon funniness are not interchangeable with feeling happier, and neither establishes an enduring change in mood. A laboratory measure of facial muscle activity is evidence of movement, not direct access to inner experience. The broader framework is explained in What Are Emotions? Psychological Definition, Components, and Functions.

Why the Hypothesis Was Proposed

The historical motivation is the possibility that bodily behavior participates in shaping emotion. Classical discussions of bodily contributions to emotion made it reasonable to ask whether the activity of one visible muscle system—the face—might affect experience. Early experimental psychologists tried to distinguish a causal influence from the commonplace observation that feeling and facial expression occur together.

In 1974, James Laird published experiments in which facial configurations associated with smiling and frowning were induced while participants reported emotional reactions. The reported differences helped advance self-attribution and facial-feedback explanations. Their historical importance is substantial, but modern evaluation must also consider the smaller samples and less standardized procedures common to that era. Laird (1974).

Further research examined physiology. Ekman, Levenson, and Friesen reported that deliberately forming emotion-related facial configurations could be accompanied by differentiated autonomic responses. A subsequent program of four experiments also found changes in physiological and subjective measures during voluntary facial actions. These findings support connections between facial movement and bodily functioning, while leaving questions about everyday emotion categories and the intervening causal mechanism. Ekman, Levenson, and Friesen (1983); Levenson, Ekman, and Friesen (1990).

The 1988 Pen-in-Mouth Experiment Explained

Fritz Strack, Leonard Martin, and Sabine Stepper introduced the most famous test in 1988. Participants held a pen between their teeth or between their lips while rating cartoons. The teeth position was intended to facilitate smiling-related muscle activity; the lips position was intended to interfere with the same configuration. Participants were not simply asked to smile and then declare whether they felt happy, an important feature of the original experimental logic.

The original study reported higher funniness ratings when participants held the pen between their teeth. The later registered replication report described the original difference as approximately 0.82 points on a 10-point rating scale. The apparent ability to change an amusing judgment through an incidental facial position gave the result unusually broad influence in psychology teaching and popular accounts. Strack, Martin, and Stepper (1988); Wagenmakers et al. (2016).

The method also introduces interpretive complexity. A pen position can change muscle activation, oral sensation, effort, comfort, and attention. It is only an indirect approximation of a natural smile. The dependent variable is a cartoon evaluation, which may involve pleasant feeling, cognitive judgment, or both. A finding with that task therefore must be described accurately as an effect on an experimental judgment, not automatically as a transformation of emotional well-being.

The 2016 Registered Replication Report

To test whether the original effect was robust, investigators coordinated 17 independent direct replications using a vetted protocol. The registered approach documented key methods and outcomes before the data were analyzed. By combining results across laboratories, it reduced dependence on the practices or participant pool of a single research group.

The combined estimate for the difference between the smile-facilitating and smile-inhibiting pen conditions was just 0.03 points. Its 95% confidence interval extended from −0.11 to 0.16, in contrast to the original estimated 0.82-point difference. Under the tested protocol, the striking original pen-in-mouth effect was not reproduced. Wagenmakers and colleagues (2016).

This result was a serious challenge to that procedure. It was not a definitive experiment on every possible meaning of facial feedback. A hypothesis that includes mimicking expressions, forming muscle configurations deliberately, and altering facial movement unobtrusively makes multiple predictions. Failure of one procedure does not logically decide every other procedure. However, a theory gains little from endlessly changing its proposed conditions after null findings; explanations for heterogeneity themselves require prospective, reproducible tests.

The question about being observed

The 2016 registered study included video recording. Researchers proposed that awareness of being watched might change participants' self-consciousness or social interpretation of facial actions. Noah, Schul, and Mayo reported a 2018 investigation arguing that observation can eliminate a facial-feedback effect under their tested conditions. This provided a plausible procedural moderator, rather than proof that a camera alone explained the entire replication outcome. Noah, Schul, and Mayo (2018).

The proposed camera explanation is useful because it produces testable predictions: if observation reliably moderates feedback, a preregistered experiment should recover that interaction under carefully matched conditions. Cartoons, task instructions, interaction with investigators, and attention also vary between studies. A strong explanation must compete against such alternatives.

The 2019 Meta-Analysis of Facial Feedback

Nicholas Coles, Jeff Larsen, and Heather Lench synthesized 286 effect sizes from 138 studies manipulating facial feedback and collecting emotional self-reports. Their meta-analysis identified an overall effect that was statistically detectable but small and variable. These findings meant the pen-in-mouth failure could coexist with some evidence from a wider collection of experiments. Coles, Larsen, and Lench (2019).

Importantly, the researchers distinguished emotional experience from affective judgments, such as ratings of a cartoon or other stimulus. They also investigated whether emotional material was present and what kind of material was used. The pattern varied across these conditions. In addition, their publication-bias analyses found concerns for studies of affective judgments even where the corresponding analyses of self-reported emotional experience did not show the same evidence.

A pooled effect is an estimate across a literature, not a universal law applying identically to every participant and technique. If direct voluntary smiling and indirect pen positioning produce different results, averaging them can conceal precisely the conditions researchers need to understand. The meta-analysis is therefore evidence for a qualified overall influence and for meaningful methodological heterogeneity.

The 2022 Many Smiles Collaboration

A particularly important advance was the preregistered, multicenter Many Smiles Collaboration, published in Nature Human Behaviour in 2022. The project involved 3,878 participants in 19 countries and was designed to test alternative facial-feedback predictions under a common research framework. Its breadth and prospective methods increased confidence in the interpretation of the specific manipulations tested. Coles and colleagues (2022).

The researchers compared three approaches: facial mimicry, voluntary facial action, and a pen-in-mouth manipulation. Mimicking facial movements associated with smiling and deliberately producing smile-related actions both yielded evidence of increased self-reported happiness. Findings for the unobtrusive pen method were less conclusive. The result supports certain causal contributions of facial action while confirming that a particular indirect manipulation should not be treated as an interchangeable substitute for direct smiling.

Facial mimicry

Mimicry asks participants to reproduce a facial pattern, often by following a displayed example. It provides a recognizable movement and can increase correspondence with a smiling configuration. At the same time, participants can become aware that the expression resembles a smile, which complicates attempts to isolate purely sensory feedback from cognitive interpretation or expectations.

Voluntary facial action

A voluntary facial action task uses deliberate instructions to produce muscles or movements associated with a smile. This approach improves control over facial behavior but engages attention, effort, task understanding, and potential awareness of emotional meaning. If reported happiness changes, the experiment establishes a causal effect of the overall assigned procedure; further studies are needed to disentangle its active ingredients.

The pen-in-mouth task

The pen procedure aims to change relevant muscle activity while making the emotional interpretation less explicit. Its less convincing results in the multinational study were broadly consistent with the need for caution after the 2016 failure. This comparison also shows why summaries such as “the smiling study proved the old pen experiment right” are scientifically inaccurate. The procedures test overlapping but different predictions. Coles et al. (2022).

So, Has the Facial Feedback Hypothesis Been Replicated?

Replication requires naming the effect being replicated. The claim that the classic pen-in-mouth manipulation reliably makes cartoons seem funnier did poorly in the coordinated 2016 test and remained uncertain in the corresponding condition of the 2022 study. The broader claim that certain deliberate smile-related facial actions can modestly change self-reported happiness has support from the 2019 meta-analysis and 2022 collaboration.

A strong contemporary summary must keep these claims together. The 2016 null result is central evidence, not an inconvenience to dismiss. The 2022 positive findings are also central evidence, not grounds to ignore the distinction between methods. The research question now concerns the causal pathways and boundary conditions of facial feedback, rather than an unqualified verdict about whether all expressions always change all emotions.

Phaf and Rotteveel developed a social-context account suggesting that the presence or meaning of an audience may help explain variations in unobtrusive facial-feedback experiments. Their analysis found preliminary patterns compatible with that account, but the authors presented it as a theoretical and exploratory contribution requiring direct experimental tests. Phaf and Rotteveel (2024).

Possible Mechanisms Behind Facial Feedback

Sensory information from facial muscles and skin

Contractions and positions of the face produce sensory information about the body. An embodied explanation proposes that some of this information contributes to a person's affective state. Such a pathway is plausible in a system where perception and bodily activity interact continuously, but the large replication studies do not establish one exclusive nerve-based mechanism through which smiling creates happiness.

Attention and self-perception

A person making a smile-related movement might notice that action, interpret it as meaningful, or attend differently to a pleasant scene. Laird's early research connected emotional reports with the attribution of meaning to one's own expressive behavior. A shift in interpretation could participate alongside sensory feedback. Experiments that explicitly describe a movement as a smile have difficulty excluding these components. Laird (1974).

Autonomic responses

Voluntary facial configurations have been associated with changes in heart rate, electrodermal activity, and other physiological variables. Such findings make the body relevant to the causal story but do not determine whether autonomic change causes subjective feeling, accompanies effort, or reflects other task demands. Measurements of bodily activity and subjective experience must be analyzed separately. Levenson, Ekman, and Friesen (1990).

Social meaning and audience

Smiling can signal friendliness, reassurance, amusement, embarrassment, or politeness. Facial actions performed in the presence of an experimenter may have different meanings than those performed in isolation. Social interpretation is therefore a candidate source of moderation. The available proposals and exploratory results justify targeted tests rather than sweeping conclusions. Phaf and Rotteveel (2024).

These mechanisms are potentially complementary. Future studies can attempt to vary the physical movement while holding emotional labeling constant, or vary social meaning while checking that muscle activation remains comparable.

How Do Researchers Measure Facial Feedback?

First, investigators manipulate facial behavior. Instructions may target whole expressions, individual muscles, mimicry, or indirect movement restrictions. The crucial quality check is whether the manipulation produces the expected facial activity. Facial electromyography, movement coding, or controlled instructions may help establish fidelity; a direction to smile does not guarantee everyone makes the same action.

Second, researchers identify the outcome. Some measure felt happiness or amusement; others ask participants to judge stimuli, and still others monitor autonomic physiology. Each measure answers a different question. A cartoon rating is not a global measure of mood, and a muscle signal is not a direct reading of conscious feeling. The 2019 meta-analysis treated outcome differences as empirically important. Coles et al. (2019).

Third, investigators address potential confounding. Awareness of the study's purpose can create expectations, while unusual equipment or being filmed can change attention and social behavior. Randomization, preregistration, standardized stimuli, adequate sample sizes, transparent exclusions, and independent replication make the findings more interpretable. No one safeguard guarantees causal identification if a procedure manipulates several psychological factors at once.

How Big Is the Effect on Happiness?

The available synthesis characterizes facial-feedback effects as generally small and variable. A statistically reliable average in thousands of observations can correspond to a shift too subtle for some participants to notice. The 2022 results concern brief experimentally measured happiness, not a demonstrated enduring increase in life satisfaction. Coles et al. (2019); Coles et al. (2022).

Generalization also depends on duration and setting. A short instructed facial action during a laboratory task does not establish effects from hours of smiling, daily practice, or a sustained emotional intervention. Statistical evidence for a difference in condition means should not be translated into a promise that any individual will feel happier every time they smile.

Why Different Experiments Reach Different Conclusions

Methods alter more than muscles. A pen held in the mouth produces an unfamiliar physical posture and may affect comfort; a voluntary smile is an intentional action; facial mimicry involves exposure to another expression. These interventions differ in awareness, physical sensation, effort, expectation, and social significance.

Experimental materials also matter. Facial feedback may interact with cartoons, pleasant sentences, pictures, or the absence of emotional stimuli. The type of question and timing of a rating can influence whether the outcome captures felt amusement, attention, or evaluation. When researchers use different manipulations and different outcomes under the same hypothesis label, different results are scientifically unsurprising. The key task is to identify reproducible moderators rather than explain each finding after it appears. Coles et al. (2019).

Does Facial Feedback Apply to Anger, Fear, or Sadness?

The hypothesis can be formulated for other facial configurations, and earlier experimental work reported subjective or physiological differences during deliberately formed expressions associated with several emotions. However, the largest recent coordinated tests highlighted here primarily investigate smiling and positive emotional experience. Evidence from those tests cannot automatically be transferred to frowning, fear, disgust, or sadness. Levenson et al. (1990).

Likewise, experiencing ordinary sadness or anger does not indicate a mental disorder. Changes in a brief rating during an experiment say little about the complex causal history of sustained mood or clinical symptoms.

Facial Feedback Compared With Other Emotion Theories

Facial feedback is one proposed mechanism within the study of embodied emotion, and it can be discussed alongside the James–Lange tradition, basic-emotion approaches, appraisal accounts, and the theory of constructed emotion. A small effect of smile-related movement is compatible with several broader theories, because they need not disagree that the body contributes information to experience. What they dispute includes how emotional categories are organized and how bodily, learned, and contextual processes interact.

For a wider comparison, see Theories of Emotion: Classical and Modern Explanations, Basic Emotion Theory: Six Emotions and Scientific Debate, and Theory of Constructed Emotion: Concepts and Evidence.

Interpretations of events and goals are also relevant to many emotional episodes. A facial movement could influence such interpretation without proving that every emotion has a fixed bodily cause. Readers interested in this neighboring question can explore Appraisal Theory of Emotion: How Meaning Shapes Experience.

Facial Feedback Does Not Mean That Faces Reveal Hidden Feelings

A causal claim about one's own facial movement differs from an observer's ability to infer an internal state. A person may smile while anxious, out of politeness, or because of social expectations; another person may experience happiness with little visible smiling. An extensive 2019 review questioned the assumption that a particular facial configuration can reliably identify a specific emotion independently of context. Barrett and colleagues (2019).

This also limits popular claims about automated facial emotion recognition. An algorithm can identify physical movement patterns to some degree; attributing a definite inner emotional experience requires further inference and validation. Facial feedback experiments do not make facial images into direct measurements of happiness. The distinction is explored in Emotion Recognition: Faces, Voices, and Context and Are Facial Expressions Universal? Cross-Cultural Evidence.

Culture, Social Context, and Individual Differences

Expressions are shaped by interpersonal circumstances and learned social practices. A smile may signal affiliation in one context, maintain politeness in another, or accompany an emotional episode. The same facial muscles can therefore occur with different meanings and feelings. Individual differences in facial mobility, habitual expressiveness, attention to bodily information, and task comprehension can affect how participants respond.

The Many Smiles Collaboration recruited across 19 countries, strengthening the geographical scope of the tested findings. That breadth does not prove that every culture, community, or individual has the same effect size or mechanism. Culturally relevant comparisons and diverse sampling are necessary to determine whether social norms or learning moderate facial feedback. Coles et al. (2022).

What About Botox, Paralysis, and Clinical Treatment?

Studies of interventions that alter facial movement are relevant to the general topic, but changes after botulinum toxin treatment or facial paralysis can have several causes. Physical mobility, appearance, expectations, social feedback, and personal circumstances may shift together. The presence of an emotional outcome following an intervention does not by itself establish that reduced facial sensory input caused it.

The main large replication evidence concerns short-term experimental ratings in mostly nonclinical circumstances. It does not establish that facial feedback is a treatment for depression, anxiety disorders, or any other condition. Medical and psychological treatment claims require separate clinical trials and clinical guidance.

Can Deliberately Smiling Improve Your Mood?

A voluntary smile may produce a slight, temporary increase in reported positive feeling in some circumstances, as indicated by the Many Smiles study. It is reasonable for individuals to notice interactions between bodily expression and feeling, but there is no basis here for requiring people to smile through sadness or using facial poses as a replacement for support or professional care. Coles et al. (2022).

Emotions provide information about experiences, needs, and relationships. A small experimental effect should be interpreted in proportion to its context and scope. Someone can find a brief smile pleasant without treating it as a universal emotional prescription.

What Research Still Needs to Resolve

The most useful next studies will directly compare proposed mechanisms. Researchers can verify comparable facial activation while varying whether participants know they are smiling, whether another person is present, and whether the task is framed in emotional terms. They can preregister which outcome matters most, distinguish experience from stimulus judgment, test how long effects endure, and independently repeat moderator findings across laboratories.

The scientific progression from the 1988 study to the 2016 replication, 2019 meta-analysis, and 2022 collaboration illustrates a productive research process: precise null results narrow a hypothesis, while successful tests under better-specified conditions make its remaining claims more informative. The result is a careful account of one possible contribution of the face to emotional experience.

Frequently Asked Questions

Does smiling really make you happier?

Sometimes, slightly. A large preregistered experiment found increased reported happiness after certain deliberate or mimicked smiling tasks. The result does not mean every smile changes mood or produces a large effect. Coles et al. (2022).

Was facial feedback debunked by the 2016 replication?

The 2016 registered replication strongly challenged the famous pen-in-mouth cartoon-rating result, but later research found evidence for other forms of facial feedback. Precise claims about methods matter more than a binary verdict. Wagenmakers et al. (2016).

What did the original pen experiment find?

The 1988 investigators reported higher cartoon-funniness ratings when participants held a pen between their teeth than between their lips. Large later studies did not consistently reproduce this specific difference. Strack et al. (1988).

Why did the 2022 study support smiling but not clearly support the pen method?

Direct smiling and mimicry differ from indirect pen positioning in awareness, movement, and other task features. The study found evidence for the first two manipulations and less conclusive evidence for the third; the active mechanisms remain open. Coles et al. (2022).

Is facial feedback unconscious?

Some procedures seek to manipulate expression without explicitly naming it. Others deliberately instruct smiling. The available results do not establish that facial feedback always operates outside awareness, or that awareness is necessary in all circumstances.

Is the facial feedback hypothesis the James–Lange theory?

Facial feedback is a narrower prediction about facial action and emotion. The James–Lange tradition concerns bodily contributions to emotional experience more generally. They are related scientific topics rather than identical theories.

Does frowning make a person depressed?

Short-term facial-feedback experiments do not establish such a claim. A transient action and a clinical depressive disorder are different phenomena with different evidence requirements.

Do facial-feedback findings validate AI emotion detection?

No. Causing a small change in average emotional report by manipulating movement is not the same as reliably inferring someone's subjective state from an image of their face. Barrett et al. (2019).

Can smiling exercises replace psychotherapy?

The cited experimental literature does not justify that use. Findings about brief emotional ratings in laboratory tasks are insufficient to establish an intervention for a mental health condition.

What is the most accurate current conclusion?

Certain smile-related facial actions can have small causal effects on reported positive experience. The traditional pen-in-mouth result has major replication problems, and the pathways that produce successful effects remain scientifically debated. Coles et al. (2019); Coles et al. (2022).

Related Articles

Theories of Emotion: Classical and Modern Explanations — The broader theoretical landscape.

Emotion Recognition: Faces, Voices, and Context — Understanding signals and context when inferring emotion.

Are Facial Expressions Universal? Cross-Cultural Evidence — Evidence and limits of cross-cultural expression claims.

Emotional Expression: Faces, Voices, Crying, and Social Signals — How facial movement fits wider expressive communication.

References

Barrett, L. F., Adolphs, R., Marsella, S., Martinez, A. M., & Pollak, S. D. (2019). Emotional expressions reconsidered: Challenges to inferring emotion from human facial movements. Psychological Science in the Public Interest, 20(1), 1–68. https://doi.org/10.1177/1529100619832930.

Coles, N. A., Larsen, J. T., & Lench, H. C. (2019). A meta-analysis of the facial feedback literature: Effects of facial feedback on emotional experience are small and variable. Psychological Bulletin, 145(6), 610–651. https://doi.org/10.1037/bul0000194.

Coles, N. A., March, D. S., Marmolejo-Ramos, F., et al. (2022). A multi-lab test of the facial feedback hypothesis by the Many Smiles Collaboration. Nature Human Behaviour, 6, 1731–1742. https://doi.org/10.1038/s41562-022-01458-9.

Ekman, P., Levenson, R. W., & Friesen, W. V. (1983). Autonomic nervous system activity distinguishes among emotions. Science, 221(4616), 1208–1210. https://doi.org/10.1126/science.6612338.

Laird, J. D. (1974). Self-attribution of emotion: The effects of expressive behavior on the quality of emotional experience. Journal of Personality and Social Psychology, 29(4), 475–486. https://doi.org/10.1037/h0036125.

Levenson, R. W., Ekman, P., & Friesen, W. V. (1990). Voluntary facial action generates emotion-specific autonomic nervous system activity. Psychophysiology, 27(4), 363–384. https://doi.org/10.1111/j.1469-8986.1990.tb02330.x.

Noah, T., Schul, Y., & Mayo, R. (2018). When both the original study and its failed replication are correct: Feeling observed eliminates the facial-feedback effect. Journal of Personality and Social Psychology, 114(5), 657–664. https://doi.org/10.1037/pspa0000121.

Phaf, R. H., & Rotteveel, M. (2024). An audience facilitates facial feedback: A social-context hypothesis reconciling original study and nonreplication. Psychological Reports, 127(6). https://doi.org/10.1177/00332941231153975.

Strack, F., Martin, L. L., & Stepper, S. (1988). Inhibiting and facilitating conditions of the human smile: A nonobtrusive test of the facial feedback hypothesis. Journal of Personality and Social Psychology, 54(5), 768–777. https://doi.org/10.1037/0022-3514.54.5.768.

Wagenmakers, E.-J., Beek, T., Dijkhoff, L., et al. (2016). Registered replication report: Strack, Martin, & Stepper (1988). Perspectives on Psychological Science, 11(6), 917–928. https://doi.org/10.1177/1745691616674458.