Abstract
Researchers on terror-management theory (TMT) often obtain effects on dependent variables, such as worldview assertion, after a delay following mortality salience (contemplating death) but not immediately. As justification, TMT researchers invoked a post hoc assumption: Death thoughts are immediately suppressed following mortality salience but rebound after a delay. In contradiction, Trafimow and Hughes and Rife et al. found that death thoughts are more accessible immediately following mortality salience than after a delay. The contradiction is so problematic that ignoring it trends toward degenerative science. TMT research might exemplify a larger problem in psychology.
Keywords
According to terror-management theory (TMT), people’s thoughts about their eventual death can cause terror. There are defensive strategies people can employ against that terror, such as asserting one’s worldview or others’. To test these, typical TMT experiments involved randomly assigning participants to a mortality-salience manipulation in which they think about their eventual death (to make death thoughts accessible) or a control group, such as thinking about dental pain. Either immediately or after a delay, participants complete a dependent variable, such as indicating how much they favor their own worldview or others’. The prediction is that mortality salience should increase death-thought accessibility, thereby also increasing worldview assertion or other TMT effects. However, a complication is that TMT effects often occur only after a delay following mortality salience but not immediately (e.g., Arndt et al., 2002; J. Greenberg et al., 2000, 2001).
To address the complication, TMT researchers proposed that death thoughts are suppressed immediately after mortality salience but later rebound to become more accessible, the death-thought-suppression-and-rebound assumption (DTSaRA; T. Greenberg et al., 1994). However, Trafimow and Hughes (2012) noted a large literature that concept accessibility following a priming manipulation usually decreases rather than increases over time. They published a set of six experiments to test the DTSaRA and obtained contrary results: Death thoughts were more accessible immediately than after a delay. Like Trafimow and Hughes, Rife et al. (2025) emphasized that whether death thoughts are more accessible immediately following mortality salience or after a delay is crucial for the TMT literature. They considered Experiment 3 as the most central one by Trafimow and Hughes because it included both a mortality-salience manipulation and a delay manipulation. Consequently, they attempted to replicate and expand Experiment 3 using participants from 22 laboratories in 11 countries (total N = 3,447). The Rife et al. attempt succeeded regardless of whether they used a word-generation task or a word-fragment-completion task to measure death-thought accessibility. Thus, both the Trafimow and Hughes findings and the Rife et al. findings contradict the DTSaRA; death thoughts are less accessible after a delay than immediately following mortality salience. The goal of the present commentary is to integrate Rife et al. and Trafimow and Hughes in service of making a conceptual point that hopefully will have an influence on TMT research and psychology research more generally.
In many but not all cases, as TMT researchers have reported, their predictions were disconfirmed except when there was a delay between the mortality-salience manipulation and the dependent variable (e.g., T. Greenberg et al., 1994). The DTSaRA is a post hoc assumption to justify these findings. If the DTSaRA is false, then this falsity is an immense problem for TMT because death thoughts do not rebound after a delay but, rather, are less accessible after a delay than immediately. The falsity of the DTSaRA implies a lack of justification for the many TMT findings that occur after a delay but not immediately following mortality salience.
A counter is that perhaps Trafimow and Hughes (2012) and Rife et al. (2025) did not use a sufficiently long delay. However, these researchers employed similar delays as in the experiments showing TMT effects after a delay but not immediately. For example, Trafimow and Hughes used the standard PANAS-X delay (a 60-item positive- and negative-affect schedule), commonly used in TMT research to (a) provide a delay in which death thoughts rebound and (b) control for the possibility that mortality salience influences affect. Rife et al. used a news story and had participants answer four questions following the news story. These can be considered standard delays roughly equivalent to the delays used to obtain effects in the TMT literature.
The length-of-delay argument, to save the DTSaRA, actually worsens the case for TMT. Suppose one momentarily accepts the length-of-delay argument while remembering that Trafimow and Hughes (2012) and Rife et al. (2025) used delays approximately as long as those used to obtain typical TMT findings. If the length of the delay was too small for death thoughts to rebound and become accessible, then according to the theory, TMT effects should not have occurred. That TMT researchers nevertheless obtained them is inconvenient for the theory. Or, denying the length-of-delay argument brings one back to the theory-refuting fact that many TMT effects occur when death thoughts are less accessible following a delay and not when death thoughts are more accessible immediately following mortality salience. Either way, the theory is in trouble.
That death thoughts do not become more accessible under the standard delay is highly problematic for TMT, and yet extraordinarily little has been done about it other than the length-of-delay argument, which is problematic even were it true. Two directions are necessary. First, the many published studies showing TMT effects after standard delays need reinterpreting. Second, the reinterpretations, whatever they are, need empirical testing. That 13 years have passed since the publication of the Trafimow and Hughes (2012) article with little in the way of movement in these directions suggests the unpalatable possibility that TMT has become or is in danger of becoming degenerative (Lakatos, 1978).
Lakatos (1978) eschewed naive falsification in which a single contrary finding disconfirms the theory (or research program more generally) because he understood that predictions come from both theory (or central programmatic assumptions) and auxiliary assumptions and that it might be an auxiliary assumption that is to blame for the empirical defeat. Including new auxiliary assumptions that help the program account for the data is one way to save it from contrary findings. This research program can be generative or degenerative. If new auxiliary assumptions expand explanatory scope or predictive power, the research program is generative. But if not, and their only function is to protect against contrary findings, then the program is degenerative. The way the DTSaRA currently functions to protect the TMT program without increasing explanatory scope or predictive power propagates in the direction of degeneracy.
The present goal is to do more than criticize TMT but, rather, use it as an example of a larger problem pertaining to psychology theories. According to the late Meehl (1978), “Most of them [psychology theories] suffer the fate that General MacArthur ascribed to old generals—They never die, they just slowly fade away” (p. 807). Unfortunately, what Meehl said more than half a century ago still applies. And TMT research exemplifies why. With very few exceptions, TMT researchers have not faced, squarely, the negative implications of the blatantly problematic finding that death thoughts are not more accessible following standard delays, such as completing the PANAS-X.
In conclusion, both Trafimow and Hughes (2012) and Rife et al. (2025) found that death thoughts following mortality salience are less accessible after a delay than immediately; the DTSaRA is false, and this falsity is damaging for the theory. Alternatively, if the DTSaRA is deemed saved by the length-of-delay argument, the theory remains in trouble because of the implication that TMT effects occurred in the presence of insufficient delay for death thoughts to rebound and become more accessible. Rather than ignoring a serious difficulty, as they have previously, TMT researchers ought to accept the challenge. One direction would be to address a question: Acknowledging the DTSaRA is false, how is it possible to explain or reinterpret the totality of TMT-relevant findings immediately following mortality salience or after a delay?
Either a new theory is needed or more creative auxiliary assumptions are needed to account for all the TMT-relevant literature and lead to new predictions. That TMT researchers have mostly not explored new theories or new auxiliary assumptions since the Trafimow and Hughes (2012) publication is suggestive of a degenerative program and exemplifies Meehl’s (1978) objection pertaining to psychological research. My hope is not just that TMT researchers will take note but that psychology researchers across areas will so that the Meehl quotation no longer applies.
