Toward an Integrative Theory of Psychological Defense

Terror management theory

TMT’s central tenet describes people as motivated to avoid the terror of inevitable death by envisioning themselves as “good” members of a meaningful, existentially satisfying universe. The theory was originally devised to explain humans’ ardent and assiduous striving for self-esteem and tendency toward intergroup conflict, proposing that self-esteem (feeling good about oneself) and internalized cultural worldviews (meaningful systems of belief, which differ across groups) palliate death concerns by providing a sense of symbolic and/or literal immortality (see Fig. 1, left panel). People can “live on” in heaven or nirvana or simply by leaving a legacy—a distinctive personal imprint—in an enduring cultural community.

OPEN IN VIEWER

Fig. 1.

A structural comparison of terror management theory (TMT), epistemic equilibrium, and the security system. These are rough conceptual schematics intended to highlight key structural differences between theories. They are not comprehensive or strictly definitive. Arrows signify associative relations between constructs and do not represent a statistical model. Red arrows signify approximate functional equivalence. Relational and self-focused theories (not pictured) cast belongingness and self-esteem concerns, respectively, as the centerpiece of defensiveness.

According to TMT’s dual-process model (Pyszczynski, Greenberg, & Solomon, 1999), self-esteem and worldview defenses are distal mechanisms—they are not always logically related to death concerns per se, and they tend to occur when death thoughts are active but subconscious. Indeed, part of their efficacy lies in operating without conscious death-related affect; they keep death concerns submerged by focusing cognitive resources on themes that are symbolically incompatible with death (e.g., potency, purpose). However, when mortality is consciously salient, people must instead use proximal defenses, which are obviously and logically oriented toward minimizing the threat of death (e.g., denial, rationalization, suppression; resolving to live a healthy lifestyle), allowing distal defenses to resume operation once conscious death thoughts are suppressed.

TMT’s allure is in potentially explaining proclivities ranging from the mundane (e.g., reading the Bible) to the monumental (e.g., building a cathedral) to the monstrous (e.g., genocide). Its main weakness is the difficulty of definitively confirming or falsifying the proposition that self-esteem and worldview pursuits stem from complex psychodynamic efforts to deny death. In other words, most critics do not take issue with the bulk of TMT’s insights about self-esteem and worldview striving, or with the empirical findings showing that mortality salience causes self-esteem and worldview striving, but rather with the notion that death concerns, per se, figure as a primary distal impetus for those strivings (e.g., Muraven & Baumeister, 1997).

TMT’s history is a veritable exhibition of its strength and weakness. Indeed, even detractors would probably concede the elegance of a theoretical solution that unifies under the self-esteem and worldview rubrics such diverse phenomena as

prejudice, nationalism, social judgments, interpersonal attraction, romantic love, charitable giving . . . support for preemptive wars and suicide bombings . . . stereotyping, evaluations of one’s mother, attraction to the physical aspects of sex, reactions to simple physical sensations, cancer prevention, and health promotion behavior, risk taking, legal decision making, material consumption, attributional biases, and other forms of behavior. (Pyszczynski, Greenberg, Solomon, & Maxfield, 2006, p. 329)

Moreover, a recent meta-analysis confirms abundant empirical support for the causal influence of mortality salience on a vast range of psychological outcomes (Burke, Martens, & Faucher, 2010). This research supports the mortality salience hypothesis—that reminders of their own mortality increase people’s desire to augment psychological structures that palliate death anxiety. Typical studies involve randomly assigning research participants to respond to open-ended questions about their own death or a control topic, followed by a delay or distraction and some measure of self-esteem striving or worldview defense (e.g., evaluating a person who disparages participants’ affiliations or ideologies). Another body of research supports the complementary anxiety buffer hypothesis: that bolstering self-esteem or worldviews lessens the effects of death awareness, whereas threatening self-esteem or worldviews increases susceptibility to death anxiety. These studies typically use experimental manipulations that raise or lower participants’ self-esteem or praise or condemn their beliefs, followed by either a measure of the accessibility of death thoughts (see Schimel, Hayes, Williams, & Jahrig, 2007) or a mortality salience paradigm.

Yet before these two research lines had gained much mass, reviewers and other critics argued that mortality salience effects were more parsimoniously explained by some constituent factor, such as general negative affect or anxiety (e.g., Pelham, 1997) or the feeling of not having control over events (e.g., Snyder, 1997). These critics had a structural advantage because TMT was a sophisticated theory about notoriously difficult-to-observe unconscious processes being tested using comparatively primitive laboratory methods. The mortality salience paradigm was the most straightforward test of TMT’s tenet that death concerns underlie self-esteem and worldview striving—that is, ask study participants to think about death, then measure strivings—but the bigger challenge lay in devising suitable comparison conditions. Death threatens everything in life. What other dismal subject has matching properties but does not result in death?

This challenge led to experiments comparing mortality salience with a myriad of other upsetting thoughts, ranging from dental pain to physical paralysis, none of which seemed to produce the same effects as mortality salience (Greenberg et al., 1995). TMT’s authors do not believe that only mortality salience will cause people to defend their self-esteem and worldviews (in fact, they deem that hypothesis “absurd”; Pyszczynski et al., 2006, p. 331). But in establishing the uniqueness of mortality salience effects, they have occasionally appeared to endorse this strict form of the “uniqueness” hypothesis, for instance, by stating that “mortality salience effects seem to result exclusively from thoughts of death” (Greenberg et al., 1995, p. 418).

To some degree, this central dynamic resulting from TMT’s early battles with critics still defines the territory being contested in laboratories. New theories seem to crop up perennially, laying claim to the laboratory effects generated under TMT’s auspices by attempting to explain them in simpler, more intuitive ways, which usually involves positing that the true cause of defensiveness is not death but something that death represents or that death thoughts instill. The disagreements have played out elsewhere, so I will not fully reiterate them here but instead conclude from it all that

The second conclusion is pivotal. If mortality salience effects were absolutely unique, TMT’s critics and competitors would have much less traction. But even without theoretical adversaries, threats beside death might eventually have been found to cause defensiveness, because that prediction can be derived from TMT itself. As noted above, TMT’s anxiety buffer hypothesis predicts that threatening terror management structures (self-esteem and worldviews) should raise death concerns; in turn, the mortality salience hypothesis predicts that heightened death concerns will instigate self-esteem and worldview defenses. At the very least, then, self-esteem threats should cause worldview defense and vice versa, even if TMT is correct in every detail.

Indeed, some threats that cause defensiveness do increase the accessibility of death thoughts (e.g., worldview or epistemic threats, such as uncertainty salience, and self-esteem or relationship threats; see Hayes, Schimel, Arndt, & Faucher, 2010). But what is puzzling is that TMT experiments still frequently include such threats and find that they do not produce effects parallel to mortality salience. Not only do these results appear to contradict TMT’s own hypotheses, but they also contradict the fact that each theory that can be considered a TMT alternative seems to come with a basket of studies suggesting that nonmortality threats can cause the same defensiveness as mortality salience.

In sum: ambiguity reigns. Later in the article, I will unpack these issues further; for now, it is enough to say that TMT’s principal authors and detractors are understandably concerned about the exclusivity of mortality salience effects, because if defensiveness stems from diverse threats, it seems near impossible for TMT to cleanly win the argument, to paraphrase Ernest Becker. Before we leap to conclusions, though, let us consider some theoretical perspectives that cover similar ground as TMT.

Relational theories: Attachment and belonging

One of psychological science’s axioms is that people share an innate need for close interpersonal relationships. Although frequently treated as a unitary motive (e.g., the “need to belong”; Baumeister & Leary, 1995), belongingness needs probably comprise distinct behavioral systems, such as attachment, affiliation, and sexual mating (Bowlby, 1969/1982; Weiss 1998). Nevertheless, all relational theories assume that humans’ interpersonal needs are fundamental and pressing and that these needs exert broad influence on psychological functioning. Most relevant to defensiveness, it is clear that being liked, accepted, included, or loved fosters positive self-regard and psychological well-being (e.g., Baumeister & Leary, 1995; Denissen, Penke, Schmitt, & van Aken, 2008; Srivastava & Beer, 2005), whereas being disliked, rejected, excluded, or hated fosters the opposite (e.g., Gerber & Wheeler, 2009). Similarly, in the closest relationships, being consistently loved and supported contributes to dispositional attachment security, whereas inconsistency or rejection contributes to insecurity (anxiety and/or avoidance; Bowlby, 1969/1982; Mikulincer & Shaver, 2007a). In sum, satisfied relational needs promote equanimity, and threatened needs engender adverse psychological outcomes.

Given that death represents a permanent separation from loved ones, it threatens relational needs. Therefore, mortality salience experiments, which do not typically include a control condition threatening participants’ relationships, might be interpreted as demonstrating that people bolster their self-esteem to augment their perceived social value (cf. sociometer theory; see Leary & Baumeister, 2000, which argues that self-esteem functions as an internal gauge of a person’s social desirability) or defend their worldviews as a way of ingratiating themselves (cf. coalitional psychology; Navarrete & Fessler, 2005). If so, these would be logical, socially oriented responses to social threats—no big surprise. Indeed, mortality salience activates attachment defenses (e.g., proximity seeking; see Mikulincer, Florian, & Hirschberger, 2003), and attachment threats, such as thinking about separation from a loved one, activate TMT-style worldview and self-esteem defenses (Hart et al., 2005), whereas attachment security attenuates these defenses (Hart, Shaver, & Goldenberg, 2013; Mikulincer & Shaver, 2001).

Are relational theories sufficient to explain defensive phenomena? Some evidence suggests not. First, it is unclear how primitive meaning threats—such as exposure to anomalous word pairings (Randles, Proulx, & Heine, 2011)—or control threats (i.e., thinking about being powerless) could be construed as relational (i.e., interpersonal); yet both meaning threats and control threats produce defensiveness. Second, attachment threats increase the accessibility of death thoughts (Florian, Mikulincer, & Hirschberger, 2002), which is consistent with (though not directly supportive of) a model whereby relational threats cause compensatory self-esteem and worldview defenses because they elicit mortality salience.

Of course, it is possible that different threats cause defensiveness for different reasons. And one could argue that attachment and belongingness concerns are paramount because they emerge early in the life span—well prior to death awareness. These issues will remain unsettled in the near future; but for the meantime, psychological defense theorists should attend carefully to attachment theory’s well-developed and densely researched insights, which have somewhat underappreciated potential to contribute to defensiveness theories. Attachment theory is a mature theory, built on extensive biological and comparative (i.e., ethological) foundations, with over 40 years of systematic research in developmental, personality, social, and clinical psychology. In some ways, it is the original psychological defense theory of the modern era, describing how young children literally defend themselves from threats using, essentially, an emotion-regulation strategy: the “turning to” supportive others for security when distressed. The attachment system’s mechanics continue into adulthood, and though they are far from being precisely identified (e.g., via neurobiology), they are relatively well delineated (see Mikulincer & Shaver, 2007a).

To its credit, TMT has already begun to integrate attachment theory by weaving it into a developmental analysis of terror management processes. Specifically, TMT posits that young children’s relationship with caregivers creates the template for regulating existential distress; initially, children regulate anxiety by turning to supportive caregivers, but caregivers’ provision of affection becomes increasingly conditional during socialization, and children eventually learn to associate safety and security (in an attachment sense) with good behavior and self-esteem (e.g., Solomon, Greenberg, & Pyszczynski, 2004). Later, cultural worldviews provide an additional source of self-esteem, in addition to addressing existential concerns more directly, making them analogous to a caregiver in at least two ways.

Hence, according to TMT, the palliative function of self-esteem and worldviews is developmentally rooted in attachment processes. And research integrating attachment theory and TMT reveals that among adults, close relationships serve a terror management function, and attachment style is a key personality trait influencing terror management processes (Mikulincer et al., 2003). In fact, attachment theory’s thorough accounting of individual differences may be particularly illustrative of how it can contribute to defensiveness theories. Attachment anxiety and avoidance (and their combinative opposite, security) reflect different modes of attachment system functioning resulting from experiences in close relationships. Attachment anxiety reflects chronic hyperactivation of the attachment system due to experiences with close relationship partners who are unpredictable or ambivalent caregivers, leading to anxiety about abandonment and a needy or “clingy” interpersonal style. By contrast, attachment avoidance reflects chronic deactivation of the attachment system due to experiences with close relationship partners who are rejecting, leading to discomfort with intimacy and excessive self-reliance. In addition to sometimes producing qualitatively different patterns of behavior in response to the same circumstances (e.g., in romantic relationships; see Mikulincer & Shaver, 2007a, for a review), these attachment styles influence the likelihood and mode of defensiveness (Hart et al., 2005; McGregor, Nail, Marigold, & Kang, 2005; Mikulincer et al., 2003). Such personality differences, or ones like them, may help resolve some inconsistencies in defensiveness research, which often emphasizes normative processes and neglects consideration of individual differences.

Some terror management theorists have leaned toward interpreting attachment’s anxiety-buffering properties as proximal defenses, or as subsidiary to self-esteem and worldviews, rather than as an independent symbolic mechanism (e.g., Greenberg, 2012). As with many areas of conceptual disagreement, data are mixed regarding this point, with some evidence suggesting that attachment defenses can trump self-esteem or worldview ones, implying partial autonomy (e.g., Hirschberger, Florian, & Mikulincer, 2003). In my view, attachment’s developmental precedence should afford it extra deference in TMT’s structural hierarchy. In adulthood, attachment involves drawing emotional strength not only from close relationship partners but also from symbolic mental representations of their loving support (Mikulincer & Shaver, 2007b). This lack of literalness persuades me that it should be counted as a distal, symbolic defense, not just a proximal one.

In sum, attachment theory and other relational theories may help illuminate the structure and process of psychological defense, both normatively and as moderated by personality variables. Although it appears that defensiveness is not wholly reducible to an interpersonal process, there can be no doubt that relational concerns play an important role in it.

Self-focused theories: Self-affirmation and control

If the developmental precedence of relational needs puts them near the core of defensiveness, then surely it is plausible that defensiveness revolves around some motive even more central to the psychological self. Self-affirmation theory (Steele, 1988) seems to assert this most directly. The concept of self-integrity as an overarching motivation echoes TMT’s description of a “valuable member of a meaningful universe” (Solomon et al., 1991, p. 97) or the compound conception that one is good (self-esteem) in relation to cultural norms (i.e., worldviews). And self-affirmation research has established that affirming cherished values mitigates defensiveness (e.g., Sherman & Cohen, 2002), consistent with terror management research. Similarly, Tesser (2000) argued that the “self-zoo” of theories pertaining to self-esteem maintenance could be neatly integrated if one allows that, notwithstanding superficial differences, diverse self-esteem regulation mechanisms (e.g., self-affirmation, social comparison, and even dissonance reduction) are essentially interchangeable and reflect a single motive oriented toward “augmenting, protecting, or repairing” (p. 290) self-esteem.

Does the assumption that human beings are innately driven to maintain something like self-integrity or self-esteem obviate TMT? Death threatens self-esteem by undermining a person’s apparent significance and capacity to demonstrate enduring value, and though a few TMT studies include control conditions designed to threaten self-esteem, they are by far the exception. The self-esteem solution requires only that one reinterpret attachment/belongingness and worldview defenses as oriented toward protecting self-esteem. That argument is easily made, because people’s close relationships can provide self-esteem (Leary & Baumeister, 2000), as can their beliefs and identifications (Tajfel & Turner, 1986). Yet similar rebuttals apply here as to the case of relational theories: Meaning threats via exposure to “uncanny” stimuli, such as absurd Monty Python humor or peculiar word pairings, cause self-esteem, worldview, and relational defenses (e.g., Proulx, Heine, & Vohs, 2010), but they are difficult to construe as threatening self-esteem directly. Moreover, self-esteem threats increase the accessibility of death thoughts (Hayes, Schimel, Faucher, & Williams, 2008), consistent with TMT’s analysis that self-esteem motivation is related to death denial (and perhaps difficult to explain otherwise).

Possibly, then, rather than self-esteem or self-integrity on the whole, people seek control (i.e., a sense of agency) per se, which is why they become defensive when exposed to existential, relational, self-esteem, worldview, or meaning threats (e.g., Snyder, 1997; cf. the compensatory control model; Kay, Gaucher, Napier, Callan, & Laurin, 2008). Arguably, each of these threats undermine individuals’ sense of personal control, and subsequent defensiveness could be understood as efforts to restore that sense (directly or vicariously). Indeed, recent studies showed that thoughts of uncontrollable death caused worldview defense, but thoughts of controllable (self-determined) death did not (Fritsche, Jonas, & Fankhänel, 2008). Although both controllable and uncontrollable death thoughts increased the accessibility of death thoughts, only the uncontrollable-death manipulation increased implicit control motivation. However, this latter finding was essentially a manipulation check—there was no direct evidence that implicit control motivation accounted for the effects of thoughts of uncontrollable death—and otherwise, one could interpret these findings as suggesting that increasing the sense of control over one’s death buffers mortality salience effects. ³ Perhaps this indicates a boundary condition for mortality salience effects, but it cannot definitively establish that mortality salience effects are attributable to a control threat.

Again, then, we reach an impasse because both TMT and a “deathless” version of it—one that construes worldview defense as self-esteem defense or compensatory control—make largely the same predictions; where their predictions differ, a dearth of research prevents confident conclusions. However, research on relational theories, self-esteem maintenance, and control motivation suggests two conclusions worth articulating here in addition to those asserted earlier:

Agreement with these conclusions has inspired yet another integrative effort that views psychological defense through an epistemic lens (Proulx et al., 2012); I now turn to consider this kind of perspective.

Epistemic theories: Meaninglessness, inconsistency, and anxious uncertainty

In some respects, contemporary defensiveness research was presaged by cognitive dissonance theory (Festinger, 1957). If death disrupts people’s sense of the way the world works (i.e., dissonance), and self-esteem and worldview defenses confer a sense of “rightness” (i.e., consonance between cognitions or between expectations and reality), then dissonance theory works reasonably well as a predictor of the defensiveness effects generated by TMT and its theoretical relatives (save, perhaps, that mortality salience effects seem to occur without palpable arousal).

Although dissonance theory has not been invoked much in this realm recently (but see Gawronski, 2012, and Proulx et al., 2012), the meaning maintenance model (MMM; Heine et al., 2006) similarly posits that the integrity of mental representations plays a central motivational role in defensive phenomena. According to the MMM, people want to maintain meaning, defined as expected relations between mental representations; threats like mortality salience upset meaning, and self-esteem and worldview defenses aim to restore it. Likewise, uncertainty compensation theories (e.g., Tritt, Inzlicht, & Harmon-Jones, 2012; van den Bos, 2001) and inconsistency compensation theories (Proulx et al., 2012) seek to explain defensiveness as “palliative responses to a basic, biologically based pattern of aversive arousal that follows from any given prediction error” (Proulx et al., 2012, p. 285). Despite the different terminology used by these theories (i.e., “meaning” vs. “uncertainty” vs. “inconsistency”), they share several author–researchers, and there is little substantive difference between them—so for brevity’s sake, I treat them as a single epistemic equilibrium perspective (see Fig. 1, middle panel). (Proulx and Inzlicht [2012] acknowledged that despite semantic differences across some of their articles, they intend to refer to the same underlying process, which they have mirthfully termed “disanxiousuncertilibrium.”)

That these theories pose the strongest challenge to TMT is attested by the relative frequency with which TMT studies include uncertainty manipulations and by a meta-analysis suggesting that the effect of meaning threats on defensiveness rivals the effect size of mortality salience (Burke et al., 2010). The theories are evolutionarily plausible because organisms must reliably interface with the environment to reproduce; for more complex animals, this means building clear, coherent, and consistent representations of reality. If organisms could not adapt to a reality that continually changes in unpredictable ways, neither could they function in a stable reality if their mental representations were chaotic. A motivation to construct practicable mental representations and to maintain them by becoming attentive and making swift adjustments if they seem lacking would therefore be advantageous across species and developmental stages. Even infants who have yet to develop more complex behavioral systems like attachment could benefit from a rudimentary preference for certainty and consistency (e.g., a newborn who has come to expect the steady pressure of contact with a caregiver’s body will cry when set down and consequently get picked up again). Finally, epistemic equilibrium theories have made generative connections to biological and neurological literature, namely, by identifying threat responses with a general behavioral inhibition–anxiety system that detects threats and initiates anterior cingulate cortex activation and physiological arousal, followed by compensatory behavioral approach systems governed by the dorsalateral prefrontal cortex (e.g., Tritt et al., 2012).

Insofar as mortality salience violates expectations (of continued life) or elicits uncertainty (e.g., about what happens after death), epistemic theories seem capable of explaining many TMT findings. Yet there are pitfalls in the temptation to construe all defensiveness as reflexive responses to expectancy violations, particularly because most epistemic equilibrium views are explicitly indiscriminating about the content of expectancies or of worldviews in a broader sense: “inconsistency compensation cannot be about any specified content (e.g., the self). . . . Rather, [threat-compensation] behaviors are palliative responses to the aversive arousal that follows from any experience that is inconsistent with expected relationships” (Proulx et al., 2012, p. 288).

Ergo, death violates expectations by disrupting the network of associations connecting oneself to the world (Heine et al., 2006). But following this logic to its conclusion leads to the ironic predictions that (a) people will respond defensively when their expectancies are violated in a salutary way (I expected that attractive woman not to date me, but she did; I expected a 3% salary raise, but it was 4% instead; despite expectations, the tumor on my thyroid turned out to be benign), and (b) people will not respond defensively when their negative expectancies are confirmed, as when my belief that the New York Jets would get trounced by the San Francisco 49ers one weekend turned out to be magnificently consistent with the outcome. (Try convincing my wife that this did not cause defensiveness.)

In fairness, some research does suggest that people show signs of threat when their positively valenced expectations are violated (e.g., Townsend, Major, Sawyer, & Mendes, 2010) and that they appear to take some comfort in negatively valenced expectation affirmations (e.g., Plaks & Stecher, 2007; Swann, Wenzlaff, Krull, & Pelham, 1992; Townsend et al., 2010). Yet other research points to opposite conclusions. In one example, atheists exposed to information supporting the existence of an afterlife became less defensive in response to a subsequent mortality salience prime (Heflick & Goldenberg, 2012). In another study, mortality salience increased belief in “culturally alien” supernatural agents (Norenzayan & Hansen, 2006).

Epistemic equilibrium theorists acknowledge that the content of mental representations can influence defensive processes but only as a moderator of defense against inconsistency, something that alters the process but does not drive it (Proulx, 2012; Proulx & Inzlicht, 2012). For example, perhaps intrinsically religious individuals do not respond defensively to mortality salience (Jonas & Fischer, 2006) because, even though death disrupts mental representations connecting them to the world, it is nevertheless consistent with their religious beliefs—death is “part of God’s plan,” just another stage in an everlasting spiritual existence.

In my view, this stance toward mental representations misplaces the role of specific motivational systems that are integrally relevant to defensiveness and that care very much about representational content—systems that mediate, or drive, the defensive process, instead of influencing it from outside. Regarding attachment, for example: by definition, individuals higher in attachment anxiety or avoidance both harbor negative relationship expectations (e.g., as evidenced by greater physiological response to stressors when their romantic partners are present versus absent; Carpenter & Kirkpatrick, 1996). This does not mean that they are not hopeful that close relationship partners will unconditionally love and support them. Hence, it is not surprising, from an attachment perspective, that thoughts of separation—despite ostensibly confirming (negative) expectations—cause both anxious and avoidant individuals to defend their worldviews or self-esteem (Hart et al., 2005). Nor is it surprising that being subliminally primed with intimacy words or led to think about sensitive, responsive, and loving others—ostensibly violating insecure individuals’ expectancies about how close others typically behave—mitigates both anxious and avoidant individuals’ worldview defense (e.g., Mikulincer & Shaver, 2001). An epistemic equilibrium perspective does not seem to straightforwardly account for such findings, in which desires, not expectations, appear to be the mechanisms driving defensive processes. ⁴

We do not yet know whether the content of mental representations is better understood as a mediator or as a moderator of defensiveness. I argue that meaning is important to defensiveness but is not an exclusive (nor always the most important) mediator; other motivations are centrally involved in maintaining equanimity. In other words, I am skeptical that defensiveness can always be understood as a consistency-maintenance phenomenon. It seems that a more comprehensive defensiveness theory is needed, one that explicitly incorporates the various motives and goals that other theories suggest people would be likely to pursue and defend, in addition to meaning maintenance.

McGregor and colleagues (2010; Nash, McGregor, & Prentice, 2011) have advanced a reactive approach motivation (RAM) theory that comes close to such a resolution by specifying that defensiveness occurs when meaning violations and inconsistency specifically threaten “active goals,” arousing anxious uncertainty and, in turn, defensive approach motivation. This version of epistemic equilibrium helpfully allows for the role of specific motivational systems that are sensitive to conflict with reality.

But the worldview content problem remains, because the RAM theory appears largely silent about why individuals carry around with them the specific goals they do. It is also unclear how the RAM concept explains defensive avoidance (not approach) goals, as when low self-esteem individuals try to preserve rather than augment self-esteem after mortality salience (Landau & Greenberg, 2006) or when individuals high in neuroticism respond to mortality salience by avoiding physical sensations (Goldenberg et al., 2006).

In sum, epistemic equilibrium perspectives offer impressive explanatory potential, scholarship, and insight. However, they seem to fall short of a full accounting of defensive phenomena, especially when they imply that expecting trumps wanting. Magical thinking, superstition, and similar well-documented human quirks seem to suggest that at least sometimes, people choose senseless-but-comforting meanings over sensible-but-terrifying ones.

If I am correct that defensiveness is about compensating for the violation of desired, equanimity-providing meanings, in addition to meanings per se, then epistemic equilibrium theories will be most successful if they more explicitly preserve and assimilate insights from related theories to account for why people hold the representations they do and why, at least sometimes, defensiveness seems more a reflection of how people think reality ought to be than how they expect it to be. This brings me to my next conclusion:

The security system

My formal input to this “zoo” of defensiveness theories may offer some small advantages. A relatively accommodating integration of insights from TMT, attachment theory, and consistency theories, among others, the security system model (Hart et al., 2005; see also Hart, 2013) hypothesizes that the core of defensiveness is insecurity, or “apprehension and anxiety about personal vulnerability” (Hart et al., 2005, p. 999). That is, people are motivated to be vigilant about various threats, and appraising a threat—to health or well-being or to the system’s constituent subsystems that maintain close relationships, self-esteem and agency, or meanings and worldviews—potentiates anxiety, activating the system and its supportive elements and motivating compensatory responses. Insecurity is not fear (i.e., an acute response to danger) but rather the potential for full-blown felt anxiety (cf. Greenberg et al., 2003), a lack of confidence that “everything will be okay” (which can certainly be triggered by fearful experiences). Security, then, reflects psychological durability in the face of threats.

Conceptually patterned on the attachment system, the security system includes an assemblage of more basic emotion-regulation and self-protective behavioral systems and an elaborated symbolic associative network comprising attachment, self-esteem, and worldview-related representations. These elements are organized around the security motivation, related to such an extent as to be practically inextricable (not to say totally indissociable) in the context of defensiveness. Because of attachment’s putative role in early “terror management” (i.e., regulation of distressing emotions) and self-esteem processes, attachment and self-esteem are seen as largely fungible. Consequently, in contrast to TMT, “symbolic immortality” does not always fully explain self-esteem augmenting defenses; instead, similar to sociometer theory, self-esteem is seen, in part, as a proxy for close relationships (“good me” = loved and secure; close relationships, in turn, are a source of self-esteem—see also Cox & Arndt, 2012). The same applies to worldviews, which, in addition to directly addressing existential concerns (i.e., by providing order, consistency, predictability—meaning—as well as symbolic and literal immortality), also contextualize self-esteem and close relationships, even as worldviews and meaning are themselves constructed from close relationships (e.g., via socialization) and individuals’ own proclivities (i.e., people tend to place epistemic emphasis on domains where they think they excel; Dunning, Perie, & Story, 1991).

This “fluid compensation” view (Allport, 1943) is consistent with epistemic equilibrium theories but departs from them in suggesting that no one of the system’s interacting constituent elements (e.g., meaning or consistency) sits atop a hierarchy above the others; rather, it casts the system’s elements as roughly on the same plane (see Fig. 1, right panel)—moderated by individual differences and context, as I describe below. The “worm at the core” of defensive efforts to shore up relationships, self-esteem, and worldviews is appraised threat itself and resulting insecurity. Threats to relationships, self-esteem or agency, and meaning undermine perceptions of security; and robust relationships, self-esteem and agency, and meaning themselves form a bulwark against such threats. (Death—the ultimate security threat—may well be “centrally” represented within this network of terrifying concepts: aloneness, humiliation, weakness, epistemic chaos.)

This perspective explains why it is easy to construe close relationships and self-esteem as integral aspects of meaningful worldviews—they are—or to construe close relationships and worldviews as relevant to self-esteem and agency concerns—they are—or to construe self-esteem and worldview concerns as having relational implications: To weaken one is to weaken all; to strengthen one is to strengthen all, albeit indirectly. Perhaps it also explains why there is evidence to support each of the defensiveness theories. They are all correct, to an extent; they fall short when they insist on making their favorite construct the whole explanation.

I might be accused of the same, but the security concept is encompassing rather than exclusive; it unites the least tenuous parts of other theories. The notion of an overarching, adaptive motive to feel secure and free of appraised threats has ample precedent in neurobiology (e.g., Halpern & O’Connell, 2000; McNaughton & Corr, 2004; Tritt et al., 2012; Woody & Szechtman, 2011). Indeed, “insecurity” seems similar to Proulx and colleagues’ (2012) aversive arousal (see also Tritt et al., 2012), McGregor and colleagues’ anxious uncertainty (e.g., McGregor et al., 2010; McGregor, Prentice, & Nash, 2012), and Holbrook and colleagues’ unconscious vigilance (Holbrook, Sousa, & Hahn-Holbrook, 2011).

A final aspect to recommend the security system concept is that it conforms to my view that a unified defensiveness theory should explicitly preserve the best elements of its progenitors. Specifically, it borrows from other theories a taxonomy of defensive motives, an analysis of their development and evolutionary origins, a theory of cognitive architecture, and predictions about individual differences.

Individual differences

In a similar vein, though defensiveness is frequently discussed in normative terms, individual differences need to be considered. Many personality traits could be brought to bear; TMT research has examined how mortality salience interacts with trait self-esteem, attachment style, and worldview differences such as political orientation and personal need for structure. (Neuroticism is also examined, especially in the context of TMT research on health and the body [see Goldenberg & Arndt, 2008].)

In short, security maintenance processes are sometimes influenced by personality variables relevant to one or more of the system’s components. For example, attachment avoidance predicts defensive self-esteem augmentation and evasion of intimacy (e.g., Hart et al., 2005; Mikulincer, 1998). Similarly, high-self-esteem individuals have been shown to respond to mortality salience by trying to boost self-esteem even when there is a risk of failure; by contrast, low-self-esteem individuals responded by avoiding risk and thus preserving self-esteem (Landau & Greenberg, 2006).

According to the security system model, attachment style, self-esteem, and variables like personal need for structure reflect essential dimensions of a person’s overall psychological security across domains. These traits generally influence (a) proneness to threat and (b) selection of preferred defenses, although there is empirical inconsistency on the first point, with some studies suggesting that dispositional insecurity increases the likelihood of defensiveness (e.g., Mikulincer et al., 2003) and others suggesting that dispositional security does (e.g., high self-esteem; McGregor et al., 2005). Some evidence suggests that insecurity increases avoidance-oriented responses, whereas security increases approach-oriented responses to threat (e.g., Cavallo, Fitzsimons, & Holmes, 2009; consistent with Murray, Holmes, and Collins’s [2006] risk regulation theory). Still other evidence suggests that implicit insecurity combined with explicit security heightens the likelihood of defensiveness (e.g., McGregor et al., 2005; Schmeichel et al., 2009). Obviously, no single conclusion can neatly tie up all the relevant findings, making this one of the key areas for future research (see Table 2).

OPEN IN VIEWER

Table 2.

Summary of Methodological Goals, Questions to Be Answered, and Recommendations

Questions to Be Answered (1) Worm at the core: One or more?  a) Does defensiveness reflect a unitary process or multiple distinct processes?  b) If unitary, what is the best way to characterize what is at the core of defensiveness? (2) Mediating mechanisms  a) Which structures and processes mediate defensiveness and how are they related?  b) Does defensiveness always involve consistency violations?  c) To what extent does defensiveness involve autonomic nervous system arousal and affective processes? What is the role of consciousness?  d) Which brain structures and systems control defensiveness?  e) How do defensive processes unfold over time?  f) To what extent are basic emotion systems (e.g., fear, anger, sadness) involved? (3) Boundary conditions and other moderating factors  a) When, why, and how do individual differences and context moderate defensiveness?  b) How does defensiveness develop and change across the life span? Methodological Goals (1) Develop more reliable and valid measures of unconscious constructs. (2) Systematically validate experimental protocols.  a) Create more targeted, confound-free manipulations and measurements.  b) Conduct manipulation checks more routinely. (3) Expand research to include larger-scale studies and more diverse methodological approaches. Recommendations (1) Prioritize methodological goals. (2) Design studies that pit theoretical predictions against one another. (3) Coordinate efforts to answer overarching questions and “drill down” into nuanced findings.

Manipulations, measurements, and mechanisms

Whence comes defensiveness? As much of the foregoing suggests, the primary conceptual issues to be resolved concern mechanisms (i.e., mediators) and moderating conditions. The most vexing question, perhaps, is the mechanistic one: What is it that people are defending against?

I have argued that the answer is insecurity stemming from a range of threats to equanimity-providing structures and processes that constitute a coordinated security system. Other frameworks specify a more distinct psychological concern around which defensive processes are organized. But we have yet to see definitive evidence of a singular process underlying defensiveness, a shortcoming that I attribute to methodological problems.

Consistent with TMT’s contention that death concerns lie at the root, several studies find that the accessibility of death-related thoughts corresponds to defensiveness. Specifically, threats to close relationships, self-esteem, or worldviews elevate death-thought accessibility, whereas boosting those structures reduces or prevents death-thought accessibility after mortality salience (see Hayes et al., 2010, for a review). However, measures of death-thought accessibility almost never control for negative words unrelated to death, so many of the relevant findings may simply reflect the accessibility of negative themes in general, thus calling into question conclusions from the vast majority of death-thought accessibility studies.

Just as troubling, to my knowledge only a single, small study (N = 26) has found that the effect of mortality salience on a defensive outcome was statistically mediated by the accessibility of death thoughts (Vail, Arndt, Motyl, & Pyszczynski, 2012, Study 4). Statistical mediation may be a high bar, because implicit measurements can make conscious the very constructs they are measuring—which may disrupt defensive processes (cf. Spencer, Zanna, & Fong, 2005). I suspect a bigger problem: Measures of construct accessibility are not very good, particularly when used to detect remote associations.

A good handful of examples from my laboratory file drawer support these concerns. Similarly, Hughes (2013) meticulously conducted numerous death-thought accessibility studies using the most common measure, a word-fragment completion task in which some word stems can be completed either with death or nondeath words. In four of the studies, mortality salience produced no hint of an expected increase in the accessibility of death thoughts (i.e., after a delay period). In two of the studies, compared with dental pain salience, mortality salience did not increase the accessibility of death thoughts with or without a delay. Several additional studies using alternative indirect measures also found no effect of mortality salience on the accessibility of death thoughts. Lexical decision tasks do not seem much better; in one (unpublished) study, I used Schimel et al.’s (2007) measure and gathered over 100 participants per experimental cell, finding only a nonsignificant trend—death-thought accessibility was highest in the mortality salience condition—and the trend also occurred for negative, nondeath words.

To be clear, I am not saying that thought-accessibility measures are completely unfruitful (for particularly impressive examples, see Kosloff, Greenberg, Sullivan, & Weise, 2010, Study 3; and Schimel et al., 2007). But implicit measures have a long history of questionable reliability and low replication rates in experimental studies (LeBel & Paunonen, 2011). Indeed, if such methods were sensitive and reliable, there would be no reason for defensiveness researchers not to use them in every study to provide direct support for mediating processes. (The word-fragment measures are particularly easy to administer and score.) That this is far from the norm further suggests that the measures are wanting and that it may be futile to apply existing construct-accessibility measures to solving the “worm at the core” riddle.

Difficulties also plague attempts to tap mechanisms by manipulating them—which is the approach most experimental defensiveness studies use (e.g., by manipulating mortality salience or meaninglessness). As noted earlier, how can researchers exclusively target death concerns when mortality salience threatens nearly everything people care about? Moreover, if the security system model is accurate, then it is also true that threats to attachment will threaten self-esteem and meaning and elicit death thoughts; threats to self-esteem will threaten attachment and meaning and elicit death thoughts; and threats to meaning will threaten attachment and self-esteem and elicit death thoughts!

If so, then it might matter less that researchers usually fail to ensure that their manipulations are free of confounds (Leary, Terry, Allen, & Tate, 2009) and have the intended impact (e.g., as confirmed by manipulation checks). Yet these are widespread omissions: Those of us who use open-ended prompts to prime particular themes (e.g., death, uncertainty) almost never code participants’ narrative responses, either to screen for the intrusion of unrelated topics or to examine other properties of the narratives (e.g., length, complexity, distinctiveness). Any of these properties, mundane or not, could at least sometimes help explain experimental effects (see Hart & Burns, 2012) or, just as important, null findings.

A similar problem applies to measuring defensive outcomes, which is typically done using instruments that assess multiple concerns. Measures of worldview defense involve self-esteem and relational (e.g., ingroup) concerns; measures of relationship striving also seem to necessarily entail self-esteem and meaning concerns; and so on.

We defensiveness researchers habitually make a related internal validity assumption that qualitatively different threat manipulations are equivalently robust. Comparing mortality salience with uncertainty salience to “control” for uncertainty assumes that thinking about uncertainty outside a specified context is as psychologically affecting as thinking about uncertainty related to death. There is little or no evidence to support this assumption. Perhaps some threats simply need to be more strongly instantiated to produce defensive effects. (A pervasive dearth of “true,” no-treatment control conditions compounds this problem.)

Nor should we assume equivalence among threats within a given domain. Thinking about a relationship breakup is not necessarily the same as thinking about a permanent breakup (Mikulincer, Florian, Birnbaum, & Malishkevich, 2002), and neither of these are the same as being excluded or ostracized (cf. Gerber & Wheeler, 2009), to say nothing of actually enduring a breakup or rejection experience. Likewise, reflecting on past (known) experiences may not be psychologically equivalent to reflecting on future (possible) ones.

The problems discussed above are not limited to defensiveness research, though they are perhaps especially encumbering there, given the narrow grounds on which theoretical disagreements play out. Exceptions can be found, but on the whole, neither our manipulations nor measurements of defensiveness-relevant constructs have much claim to discriminant validity, which makes parsing them essentially impossible. Even more realistic or well-controlled manipulations may have different weight across different kinds of threats or insufficient strength to produce an effect, and our measures of mechanisms seem lacking. Consequently, existing data are open to a wide range of interpretations, sustaining the circular firing squad of competing theories.

Moderating factors and the evidentiary morass

Given the widespread methodological problems, it is no wonder that many process issues remain unresolved. It is easier to understand why one research group’s findings contradict another’s.

Methodological problems may also help explain ongoing confusion regarding moderating factors: why sometimes high self-esteem intensifies threat-induced defensiveness (see Burke et al., 2010), whereas sometimes low self-esteem does instead (e.g., Das, Bushman, Bezemer, Kerkhof, & Cermeulen, 2009; Harmon-Jones et al., 1997); why sometimes high explicit self-esteem combined with low implicit self-esteem best predicts defensiveness (e.g., McGregor et al., 2005) or why, as often as not, self-esteem does not influence defensiveness (e.g., Hart et al., 2005). They may help explain why sometimes specific goals must be experimentally primed before a threat elicits defensiveness (Nash et al., 2011), whereas most studies do not require such priming; or why threats to individuals’ sense of control cause different defenses than mortality salience (Kay et al., 2008; Shepherd, Kay, Landau, & Keefer, 2011) even though mortality salience is ostensibly a control threat (Fritsche et al., 2008).

What boundaries limit defensive processes? In particular, how do individual differences influence defensiveness? These issues inform process questions, too—defensiveness might rely on different mechanisms or work differently in different contexts depending on personality—but they, too, remain opaque. Frequently, personality differences moderate threat responses to such an extent that if they were not accounted for, no effects would be observed. In addition to the examples of high or low self-esteem, studies show that political liberalism (Greenberg, Simon, Pyszczynski, Solomon, & Chatel, 1992), intrinsic religiosity (Jonas & Fischer, 2006), belief in symbolic immortality (Florian & Mikulincer, 1998), attachment security (Mikulincer & Florian, 2000), and a sense of self-control (Gailliot, Schmeichel, & Baumeister, 2006) all completely eliminate certain mortality salience effects. But if these findings always hold, how are the effects ever obtained when these variables are not measured, particularly in studies with small sample sizes? After accounting for liberals, secure individuals, the intrinsically religious, those who have a sense of symbolic immortality or self-control, and individuals with either high or low self-esteem, how many people are left?

Clearly, personality and contextual factors sometimes influence the extent of interchangeability, the selection of different defenses, and their efficacy. Yet many studies show robust defensive effects without accounting for personality differences. I am led to conclude that there is no satisfying understanding of why personality differences sometimes matter and sometimes do not. Presumably, methodological factors are partly to blame.

As long as manipulations remain messy and unchecked, mediating mechanisms unreliably or invalidly measured, and procedures heterogeneous (or homogeneous but unscrutinized), such empirical contradictions seem bound to persist (and Daniel Patrick Moynihan’s dictum will seem to be violated: We can have our own opinions and our own facts)! Fortunately, methodological problems help explain an evidentiary morass; it seems likely that differences in the strength of manipulations or subtle contextual factors or sample characteristics are responsible for some empirical inconsistencies. The bad news is that the inconsistencies obscure answers to many pressing questions: Which defenses do people use (and who uses them), when do they use them, and how do they work?