Illusion of Transparency

​

What is the Illusion of Transparency?

The Illusion of Transparency is the tendency to overestimate the extent to which other people can discern your internal state. You feel the state from the inside, so you treat a small outward sign as if it were a clear signal.

People often mistakenly believe that their internal states “leak out” more than they really do.

That sentence is the abstract of Gilovich, Savitsky, and Medvec (1998). The authors’ account is anchoring and adjustment. The anchor is your own experience of the feeling. The adjustment is the step that asks how much of that feeling another person can actually see. Like other adjustments from a strong anchor, this one tends to stop too soon. The result is a private state that feels public. A useful picture is a window you are standing behind. You can see your own face in the glass and the room behind you, so the pane feels transparent. The person outside mostly sees a reflection. Your nervousness, your disgust, or the lie you are telling can be as loud as a shout from your side of the glass and still be a murmur from theirs. The sibling bias is the Spotlight Effect: overestimating how much other people notice your appearance or your actions. Gilovich and Savitsky separated the two in Current Directions in Psychological Science in 1999. Spotlight is about what is on the outside being watched. The Illusion of Transparency is about what is on the inside leaking out. A stained shirt and a pounding heart are different errors, even when both make you self-conscious. The Curse of Knowledge is a third cousin: the trouble of setting aside a fact you know. The 1998 authors argued that their lie results were not well explained by that cousin, or by a general theory that lies are easy to catch. This page does not treat those three names as aliases.

​

Illusion of Transparency in 3 Depths

• Beginner: Notice the moment you are sure your face has given you away: the shaky answer, the dislike you tried to hide, the worry you think the whole room can see. The feeling is real. The broadcast is the part to doubt.
• Practitioner: Before a talk, a hard meeting, or a first question in class, say the specific finding, not a vague reassurance. Audiences are worse at reading a speaker’s nerves than the speaker expects. Then do the task.
• Advanced: The fear of being read can feed the state it fears. Savitsky and Gilovich (2003) call speech anxiety a self-exacerbating loop: you feel nervous, you think it shows, that thought makes you more nervous. The loop is about leakage, not about whether the talk is actually any good. Breaking it requires naming the leak, and it still will not help a fear that has nothing to do with being seen.

​

Origin

The name and the first experimental set belong to Thomas Gilovich at Cornell, Kenneth Savitsky at Williams College, and Victoria Husted Medvec at Northwestern. Their paper is “The illusion of transparency: Biased assessments of others’ ability to read one’s emotional states,” Journal of Personality and Social Psychology 75(2), August 1998, pages 332–346, DOI 10.1037/0022-3514.75.2.332. The abstract describes three sets of studies. Liars overestimated how detectable their lies were (Studies 1a, 1b, and 1c). People who tasted a foul drink thought their disgust was more apparent than it was (Studies 2a and 2b). A last pair (Studies 3a and 3b) looked at reluctance to step in during an emergency. All three sets were offered as consistent with anchoring and insufficient adjustment. The emergency procedures were not re-read for this page, so they are not described here. In the lie work, the paper’s text reports that across eight sessions liars estimated that 48.8 percent of the other participants would pick them as the liar, when 25.6 percent did, a rate the authors call indistinguishable from chance. In a follow-up, forty Cornell students in eight groups of five each lied once across five rounds. Liars estimated that 50 percent would select them; 27 percent did, paired t(7) = 6.56. A further lie study had liars predict that 44.3 percent of fellow players would detect the deception, against 32.4 percent who did, paired t(7) = 2.91, p < .05. The same paper reports a check that did not show the illusion: truth-tellers estimated that 63 percent of observers would say they were telling the truth, and 73 percent did. The authors say that analysis should be read with caution, because the figures were obtained by reverse-scoring the participants’ estimates. In one tasting study, tasters who tried to keep a neutral face estimated that 4.91 observers would correctly pick the foul drink. That estimate sat above a chance rate of 3.33, t(14) = 3.97, p < .005. The observers’ actual accuracy was not reliably above chance, t < 1. The paper does not, in the passage used here, print the observers’ mean count, so this page does not invent one. The review that sets this bias next to the spotlight effect is Gilovich and Savitsky, “The Spotlight Effect and the Illusion of Transparency: Egocentric Assessments of How We Are Seen by Others,” Current Directions in Psychological Science, December 1999. The abstract of that review was read; the chapter-length PDF was not. A later negotiation paper, Van Boven, Gilovich, and Medvec, Negotiation Journal 19, 2003, pages 117–131, reports that negotiators overestimated how well a partner could read preferences they were hiding, and also preferences they were trying to send. That abstract was read. Its percentages are not reprinted here.

In anxiety-provoking situations, it is often very difficult for people to believe that, despite feeling highly nervous, they do not appear highly nervous.

Savitsky and Gilovich (2003, page 619) quote that sentence from Miller and McFarland (1991, page 310). Miller and McFarland’s chapter was not opened. The sentence is used here as the 2003 paper prints it.

​

Key Points

Four distinctions keep the Illusion of Transparency from turning into “nobody can tell anything.”

​

Applications

These four moves use the finding as a check on a specific fear: that a private state is already public. They are not a treatment plan.

​

Case Study

In 2003, Kenneth Savitsky and Thomas Gilovich published two studies of public speaking in the Journal of Experimental Social Psychology, volume 39, pages 618–625. The question was whether speakers think their nerves are more visible than they are, and whether knowing that fact changes the speech. Study 1 used forty Cornell students in pairs. Each stood at a podium and spoke for three minutes, one on “The Best and Worst Things About Life Today” and the other on “The Most Important Moral Lessons to Teach One’s Children.” Topic order was counterbalanced. After both had spoken, each rated, on a scale from 0 (not at all) to 10 (very), how nervous they had felt, how nervous they thought they had looked, and how nervous their counterpart had looked. Speakers felt somewhat nervous, mean 5.95. They rated their own appearance at 6.55, and their counterpart rated that same appearance at 5.25. A repeated-measures test gave F(1, 19) = 16.38, p < .001, for self versus other, with no other reliable effect. On a forced choice, 27 of the 40 speakers, 67.5 percent, said they had looked more nervous than the other person. In eight pairs both speakers thought this about themselves. In only one pair did both think it about the other person. The authors then answered an obvious objection: the “observer” in Study 1 was another speaker, busy with their own talk. They replicated the comparison with 42 Williams College students, 17 of them speakers and the rest observers who did not speak. Speakers were videotaped on “The Best and Worst Things About Life Today” and rated how nervous they thought they looked, again from 0 to 10. Groups of five observers watched three or four speeches each. Speakers’ mean was 5.71. Observers’ mean was 4.24, t(16) = 2.94, p < .01. Study 2 asked whether telling people about the illusion would change the speech. One hundred seventeen Cornell students took part, 77 as speakers and 40 as observers. Part of the data from 9 speakers was misplaced, so some tests use all 77 speakers and some use 68. Each speaker had five minutes to prepare a three-minute speech on race relations at Cornell, then spoke to a camera in front of a second experimenter who did not know the condition. Speakers were told that other researchers were behind a one-way mirror. No one was. Control speakers (n = 20) got no extra instructions. The other speakers were told that anxiety was natural and that people are often nervous about looking nervous. Informed speakers (n = 39) then heard a description of the Illusion of Transparency: research had found that audiences cannot pick up anxiety as well as speakers expect, because one’s own emotion feels so strong. Reassured speakers (n = 18) heard that they should not worry about other people’s impressions, without the illusion being named. About half of the informed group heard the explanation before preparing, and half after preparing but before speaking. Those two timings did not differ reliably, so they were combined. Five observers rated each tape, on scales that the paper reverse-scored so that higher numbers are more favorable, from 1 to 7. The table of means is: self-rated speech quality 3.50 informed, 3.04 control, 2.83 reassured; how relaxed the speaker felt 3.94, 3.47, and 3.25; the quality the speaker expected observers to give 3.54, 3.00, and 3.15; how relaxed the speaker thought they looked 4.20, 3.35, and 2.69. Observers’ composed-appearance index was 4.65, 3.90, and 3.94. Observers’ speech-quality index was 4.23, 3.50, and 3.62. Planned contrasts put the informed condition against the other two. Speakers informed about the illusion rated their own speeches higher, F(1, 65) = 4.47, p < .05, and thought they looked more relaxed, F(1, 65) = 12.30, p < .001. Their own relaxed-feelings index moved in the same direction but was not significant, F(1, 65) = 2.43. The expected-quality contrast was F(1, 65) = 3.70, p < .06. Observers rated the informed speakers as more composed, F(1, 74) = 9.49, p < .005, and as giving better speeches, F(1, 74) = 7.94, p < .01. Control and reassured did not differ reliably on any measure. Prior talk about the topic, and usual comfort with public speaking, also did not differ across conditions. The lesson is narrow. A sentence that states the Illusion of Transparency, given minutes before a short lab speech, changed both the speaker’s forecast and the observers’ ratings. A sentence that only said not to worry did not. The boundary is in the paper’s own discussion. The manipulation would be expected to help people whose fear includes being seen as nervous, the speeches were given immediately, and the authors say they do not know whether the effect would last, or whether it would help a clinical group. These means are not a workplace error rate, a course grade, or a result for any later year.

​

Boundaries and Failure Modes

The Illusion of Transparency fails as a slogan in three places. First, it is a claim about an estimate, not a claim that nothing ever leaks. In the tasting study above, concealment kept observers near chance. That is not a promise that a strong emotion is invisible. A later disgust study, Holder and Hawkins, Basic and Applied Social Psychology 29(3), 2007, pages 235–243, still found the overestimate, and its abstract reports that women were less successful than men at hiding disgust, with no sex difference in the illusion itself. That paper’s PDF was not opened, so no further numbers from it are used here. If someone tells you what they saw, their report is evidence. The bias is a reason to doubt your own certainty, not a reason to dismiss theirs. Second, the 2003 authors limit the remedy. Informing someone about the illusion should lower anxiety only when part of the anxiety is the belief that the nerves are apparent. Some speech anxiety is not about that. The effect they measured was immediate. They write that they have no evidence it would last past the session, or that it would serve as a clinical treatment. Do not export the Cornell means, the Williams means, or the 1-to-7 table onto a keynote, a job interview, a classroom you teach, or another year. Third, the truth-teller check in the 1998 paper did not show the same gap. Using the illusion to explain every mismatch between a private state and a public reading overreaches the studies. So does using it to stay silent in a real emergency on the grounds that “someone must already see that I am alarmed.” The emergency studies are cited by the authors as a place the bias can matter. They are not a finding that concern is invisible, and this page does not describe their design.

​

Common Misconceptions

Three shortcuts flatten a specific estimate error into a personality claim.

​

Related Concepts

These six pages sit next to the Illusion of Transparency. Each one is a different mistake.

​

One-Line Takeaway