You Look Calm: The Illusion of Transparency and Speech Anxiety

By Noam Shemla · The Studies

TL;DR Nervous speakers are sure their nerves are written all over their face. Kenneth Savitsky and Thomas Gilovich showed that they are not: speakers rated themselves as looking more nervous than the people watching them did. Then they told a group of speakers exactly that, just before they spoke, and those speakers gave better talks, in their own eyes and in the eyes of observers who did not know what they had been told.

• Paper: Savitsky, K., & Gilovich, T. (2003). The illusion of transparency and the alleviation of speech anxiety. Journal of Experimental Social Psychology, 39(6), 618–625.
• Participants: Study 1: 40 Cornell students speaking in pairs, plus a replication with 17 Williams College speakers and 25 observers. Study 2: 77 Cornell speakers and 40 observers.
• Design: Study 1 compared how nervous speakers thought they looked with how nervous they looked. Study 2 randomly assigned speakers to be told about the illusion, to be reassured, or to be told nothing, and had blind observers rate the videotaped talks.
• Talks: Three minutes, unprepared in Study 1 and after five minutes of preparation in Study 2 (on race relations at Cornell).
• Access: Read in full from the copy the authors posted on Cornell’s psychology site, preserved by the Internet Archive.

Five years earlier, Gilovich, Savitsky and Victoria Medvec had named the effect this paper builds on: the illusion of transparency, the belief that what we feel shows more than it does.1 In their first study, people telling a lie in a group game guessed that about half the group (48.8%) would spot them; a quarter (25.6%) did, no better than chance. The 2003 paper took the idea to the podium, and then asked a practical question: if nervous speakers knew about the illusion, would they speak better?2

For the wider story of the effect, see our explainer on why your nerves are less visible than they feel.

Study 1: you look calmer than you think

Forty students came to the lab in pairs. Each stood at a podium, drew a topic from a card and spoke for three minutes with no preparation. Afterwards, in separate cubicles, each rated from 0 to 10 how nervous they had felt, how nervous they thought they had looked, and how nervous their partner had looked.

A partner who has just spoken, or is about to, may not be the most attentive judge, so the authors ran the study again with observers who had nothing else to do. Seventeen students were filmed speaking alone and groups of five observers rated the tapes. The result held: speakers thought they had looked 5.71 out of 10 nervous; observers saw 4.24 (t(16) = 2.94, p < .01; a within-speaker effect of about 0.71 standard deviations, our calculation).

The sharpest evidence is a single question at the end of Study 1: who looked more nervous, you or your partner? If people judged themselves fairly, half would say "me". Twenty-seven of forty did.

Even at the bottom of that interval, more than half of speakers believed they came across as the more nervous one. In eight of the twenty pairs both speakers said so; in only one pair did both say the other looked worse.

Study 2: the truth as a treatment

The authors reasoned that the illusion does more than mislead. Part of what frightens people about public speaking is the fear that the fear will show. A speaker convinced it shows gets more nervous, and then worries that the extra nerves show too.

To test it, 77 students were each given five minutes to prepare a three-minute talk on race relations at Cornell, delivered to a video camera and an experimenter who did not know their condition, with more observers supposedly watching through a one-way mirror. Before speaking, each heard one of three things:

• Informed (39 speakers): the experimenter acknowledged that being nervous about looking nervous is common, then explained the illusion of transparency: speakers feel their nerves are obvious, but audiences pick up far less than speakers expect.
• Reassured (18): the same opening, then encouragement not to worry about what others think, with no mention of the illusion.
• Control (20): nothing extra.

The reassured group is what makes the study informative. If the informed speakers improved only because someone told them to relax, the reassured speakers should improve too.

Afterwards the speakers rated themselves, and each videotaped talk was rated by five observers who did not know the conditions.

Mean ratings by condition, Study 2 (1 to 7, higher is better)

Three rows carry the finding. Observers who had no idea what the speakers were told rated the informed speakers as more composed and their talks as better. The reassured speakers were no better than the controls on anything. And the self-ratings moved in the same direction as the observers', so the improvement was not only in the speakers' heads.

How big were the effects?

The paper tests each measure with a planned contrast, comparing the informed group with the average of the other two. From each reported F we can recover a standardised effect size that needs nothing else from the data.

The speakers themselves were least moved on how relaxed they felt; that contrast did not reach significance. The instruction changed how they spoke and how they believed they came across more than how anxious they felt inside.

One more detail is easy to miss. Knowing about the illusion did not make it go away: informed speakers still overestimated how nervous they looked, and the gap between their prediction and the observers' ratings was not significantly smaller than in the other groups. What changed was not the bias but its grip. Freed from the belief that their nerves were on display, the speakers gave better talks.

Has it held up?

The first half of the finding, that speakers overestimate how visible their nerves are, was replicated in a real classroom. Cara MacInnis and colleagues had 93 students in a public speaking course give speeches to audiences of about fifteen classmates, who rated each speaker's anxiety on a six-item scale.3

The replication also echoed the Cornell oddity: more than one speaker in three (36%) believed the audience had seen more anxiety than they felt. The authors traced it to a shared script, in which everyone is expected to be terrified of public speaking. The second half, that learning about the illusion improves the talk, has less behind it. A 2011 master’s thesis at Minnesota State University, Mankato, repeated the three conditions with 31 students who had scored high on fear of negative evaluation, the people the instruction should help most, and its abstract reports results that were not consistent with the original.4 It is one small study, but it is the only direct test we found, and it did not confirm the benefit.

What this study cannot tell you

• Whether the benefit lasts. The talks followed the instruction within minutes. The authors say plainly they cannot know whether it would endure.
• Whether it helps everyone. It should only help people whose fear includes the fear of looking afraid. For someone who is afraid of forgetting their material, it answers the wrong worry.
• Whether it works in a clinic. The speakers were students, not people diagnosed with social anxiety.
• What a live room does. Study 2's audience was a camera, one experimenter and an empty one-way mirror; observers watched tapes.

What a speaker can take from it

The nervousness you feel is not the nervousness they see. What you are experiencing from the inside, the racing heart, the dry mouth, the shake in your hands, reaches the audience as less than it feels, often much less.

The useful part is not reassurance, which did nothing here, but the reason behind it. You feel your nerves more than anyone can see them because you are the only one with a view from the inside. In the Cornell study, knowing that was enough to let speakers stop watching themselves and start talking to the room.

This article is built from the full text of Savitsky & Gilovich (2003), in the copy the authors posted, and the full texts of Gilovich, Savitsky & Medvec (1998) and MacInnis, MacKinnon & MacIntyre (2010). The 2011 replication is described from its published abstract. Every mean, test statistic and condition size is the papers’ own. The exact binomial probability, the Wilson interval, the partial eta squared, the within-speaker effects, the contrast correlations and their intervals, and the standardised mean differences are our calculations from those numbers.

Measured

• How nervous speakers felt, and thought they looked, on 0 to 10 and 1 to 7 scales
• How composed and how good independent observers judged the speakers to be

Inferred, not measured

• That the informed speakers improved because the instruction broke the spiral of worrying about looking worried: the authors’ interpretation, supported by the failure of plain reassurance
• Effect sizes: derived by us from the reported F and t values

References

1. Gilovich, T., Savitsky, K., & Medvec, V. H. (1998). The illusion of transparency: biased assessments of others’ ability to read one’s emotional states. Journal of Personality and Social Psychology, 75(2), 332–346. https://doi.org/10.1037/0022-3514.75.2.332
2. Savitsky, K., & Gilovich, T. (2003). The illusion of transparency and the alleviation of speech anxiety. Journal of Experimental Social Psychology, 39(6), 618–625. https://doi.org/10.1016/S0022-1031(03)00056-8
3. MacInnis, C. C., MacKinnon, S. P., & MacIntyre, P. D. (2010). The illusion of transparency and normative beliefs about anxiety during public speaking. Current Research in Social Psychology, 15(4), 42–52. https://crisp.org.uiowa.edu/sites/crisp.org.uiowa.edu/files/2020-04/15.4.pdf
4. Gloth, C. (2011). The illusion of transparency and public speaking: a study of social anxiety. Master’s thesis, Minnesota State University, Mankato. https://cornerstone.lib.mnsu.edu/etds/27/