The Journal29 August 20267 min read
The effect twenty-one laboratories did not find
Terror management theory and its registered replication
In the late 1980s a researcher told a contact at a court in the south-west of the United States that she was studying the relation between personality, attitudes and bail decisions. The contact then handed out envelopes to colleagues at the courthouse; her own data were not included in the analysis. Inside the envelopes were questionnaires, a case file and a form. The file was laid out as court services present one every day: charge of prostitution, address, employment history, length of residence, and with it the prosecution’s recommendation that the defendant not be released. Of 24 packets handed out, 22 came back, fifteen from men and seven from women.
Half of them contained an additional sheet, headed as a new projective personality measure. It asked two open questions: what will happen to you when you physically die? And what feelings does the thought of your own death arouse in you? The judges who had answered those two questions set bail for the same defendant at 455 dollars on average. The others set 50.
Abram Rosenblatt, Jeff Greenberg, Sheldon Solomon, Tom Pyszczynski and Deborah Lyon published this in 1989 as the first of six studies. It is one of the most frequently retold experiments in social psychology, and it is the reason this file is open.
What the theory claims
Terror management theory comes from Greenberg, Pyszczynski and Solomon and draws, as the authors themselves write, above all on Ernest Becker. People know of their vulnerability and their death; cultural conceptions of reality buffer the anxiety that follows from that knowledge. To believe in values, groups, religions, nations or works that outlast an individual life is to hold on to something that does not die with you. From this follows the testable prediction: when mortality is called into awareness, the defence of that order should increase — friendlier towards those who uphold it, harsher towards those who violate it.
The question itself is older than any laboratory. Epicurus wanted to disarm the fear: as long as we are here, death is not; when death is here, we are not. Heidegger saw in a conscious relation to one’s own death a way of reclaiming life from anonymous habit. Viktor Frankl stressed that meaning becomes urgent not despite limitation but partly through it. What terror management theory added is not the insight but the operationalisation: two open questions, a control group, a measurable judgement.
What the meta-analysis counted — and why the effect needs a pause
This arrangement grew into a research programme with hundreds of variants. In 2010 Brian Burke, Andy Martens and Erik Faucher added up what had become of it: 164 papers with 277 experiments. Across all of them they found a medium-sized effect of mortality salience on defence of worldview and self-esteem, an effect size of r = .35.
What is remarkable about this meta-analysis is less the average than the conditions under which it came out larger. The effect was stronger with US participants, stronger with students, stronger when more time lay between the reminder and the measurement, and stronger when the instrument recorded judgements about people. That is a narrow base: a finding that appears most clearly in college samples from one country is either culturally embedded or methodologically dependent — and either one limits the reach usually attributed to the theory.
The third of these conditions is not a curiosity of the analysis but theory. In 1999 Pyszczynski, Greenberg and Solomon set out a dual-process model of defence. As long as the thought of death stands in consciousness, immediate defences take hold: one pushes it away, declares oneself not at risk, moves the event into a distant future. Only once the thought has left attention but is still accessible are the symbolic defences the theory is about supposed to set in.
In practice that means a distraction belongs between the two questions and the measurement. The effect is precisely not supposed to appear at once. This is a theoretically motivated but empirically awkward requirement, because it makes the absence of a finding hard to interpret — was the prediction wrong, or the pause too short?
What twenty-one laboratories found
Many Labs 4 was meant to clear exactly this objection out of the way. Richard Klein, Corey Cook, Charles Ebersole and colleagues repeated a classic experiment of the theory, Greenberg and colleagues from 1994, in a preregistered design with a test built into it: one half of the laboratories worked from a protocol the original authors had helped shape and had approved, the other half from a version produced without any involvement of the original authors. The question was whether proximity to the inventors of an effect improves its repeatability.
Twenty-one laboratories collected data and more than 2,200 people took part; after the exclusions specified in advance, 17 laboratories with 1,550 participants entered the main analysis. The planned comparison of the two protocols proved impossible in the end — not because it failed, but because there was nothing to compare. Under none of the conditions did an effect of mortality salience on worldview defence appear. The authors draw two possible conclusions from this, and their abstract leaves both open:
This suggests that the original finding was either a false positive or the conditions necessary to obtain it are not yet understood or no longer exist
The contrast between r = .35 and this null result is the actual case. A meta-analysis summarises what has been published; a preregistered multi-laboratory study generates its own data and fixes in advance how they will be analysed. Where the two procedures diverge this far, the difference is not noise but a finding about the literature.
| Test | Scope | Result |
|---|---|---|
| Rosenblatt et al., founding experiment | 1989 | mortality salience changes the judgement |
| Greenberg et al., the delay | 1994 | the effect needs a pause between stimulus and measurement |
| Burke et al., meta-analysis | 277 experiments | average effect r = .35 |
| Many Labs 4, preregistered | 21 laboratories, 2,220 people | no effect under any condition |
An explanation that demands less
Terror management theory is not the only reading of its own findings. In 2006 Steven Heine, Travis Proulx and Kathleen Vohs proposed a more economical model: people need meaning — a structure of expected relations through which they perceive the world. When that structure is disturbed, they affirm other relations in its place, including relations in quite different domains. The authors call this fluid compensation and observe it after very different threats: attacks on self-esteem, uncertainty, rejection — and mortality among them.
If that is right, death is not the root but one case among several. Being able to explain the same finding in two ways is not yet to have decided between them; the neighbourhood of cognitive dissonance is visible here, where a disturbed order is likewise restored, if necessary somewhere else.
Anxiety, meaning and the limits of interpretation
The experience at issue remains independent of the theory. Existential anxiety is not automatically a disorder; it can indicate that something valuable is at stake. Where it becomes persistently overwhelming and leads to panic, compulsion or severe avoidance, professional help is called for. Philosophy does not then take the place of assessment.
The reverse move — inferring a person from their stance towards death — carries no better. Humour may be defence, intimacy or cultural habit; silence may mean fear as readily as privacy. To unmask religious conviction, equanimity or activism as “fear of death” is to apply a theory exactly where its evidence is weakest and its sound most tempting.
What remains
The bail study of 1989 stands as what it is: a single experiment with 22 people whose difference only just crossed the threshold of significance: F(1, 20) = 4.70, p < .05. Elegantly devised, often retold, never repeated under the same conditions. The theory behind it has not been refuted — it is unsupported at the point where it was asserted most strongly. What holds is the observation that people affirm their order after threats to it. In dispute is whether death occupies a special place among those threats or is one threat among many. And it is open whether the field can demonstrate the basic effect once more in a way that lets other laboratories find it again; the authors of Many Labs 4 hold precisely that to be the condition for continuing to trust this area of research.
In everyday life the difference shows in the direction of the argument. To say that mortality makes life more urgent is to say something about oneself, and it may well be right. To say that a particular political attitude is really repressed fear of death is to invoke a laboratory finding — and as things stand, that finding is no solid ground for reading a person from a distance. Finitude poses a question nobody can delegate: if time is not property but a condition, what should it be enough for? An answer to that is not an effect size.
Sources, and why they are here
Greenberg, J., Pyszczynski, T., & Solomon, S. (1986). The causes and consequences of a need for self-esteem: A terror management theory. In Public Self and Private Self. Springer.
The first formulation of the theory — needed to show that it began as an explanation of self-esteem and not as research on death.
Becker, E. (1973). The Denial of Death. Free Press.
The text the theory takes its basic idea from. It is cultural philosophy, not an experiment — and the article keeps the two apart.
Rosenblatt, A., Greenberg, J., Solomon, S., Pyszczynski, T., & Lyon, D. (1989). Evidence for terror management theory: I. Journal of Personality and Social Psychology, 57(4), 681–690.
The founding experiment with the judges and the bail — the scene every account of the field begins with.
Greenberg, J., Pyszczynski, T., Solomon, S., Simon, L., & Breus, M. (1994). Role of consciousness and accessibility of death-related thoughts in mortality salience effects. Journal of Personality and Social Psychology, 67(4), 627–637.
The finding on the delay — and exactly the experiment Many Labs 4 repeated.
Pyszczynski, T., Greenberg, J., & Solomon, S. (1999). A dual-process model of defense against conscious and unconscious death-related thoughts. Psychological Review, 106(4), 835–845.
The theoretical justification for why the effect needs a pause — an auxiliary assumption that makes the theory harder to test.
Burke, B. L., Martens, A., & Faucher, E. H. (2010). Two decades of terror management theory: A meta-analysis of mortality salience research. Personality and Social Psychology Review, 14(2), 155–195.
The summary figure of r = .35 from 277 experiments — one side of the contrast at issue.
Klein, R. A., Cook, C. L., Ebersole, C. R., et al. (2022). Many Labs 4: Failure to replicate mortality salience effect with and without original author involvement. Collabra: Psychology, 8(1), 35271.
The other side, preregistered and with the original authors involved in the design — and the source of the quotation.
Heine, S. J., Proulx, T., & Vohs, K. D. (2006). The meaning maintenance model: On the coherence of social motivations. Personality and Social Psychology Review, 10(2), 88–110.
The explanation that demands less: not fear of death but any disturbance of the coherence of meaning. It explains the same findings with one assumption fewer.
Frankl, V. E. (1946/2006). … trotzdem Ja zum Leben sagen [Man's Search for Meaning].
The clinical counter-voice: meaning as a task, not as a defence. The article carries it because the theory declares it a protective function.
Heidegger, M. (1927). Sein und Zeit [Being and Time].
The philosophical prehistory of being-towards-death — the figure of thought the theory claims to test experimentally.