Stories and the AT Field

Why does a story get through where advice bounces off?

People often dislike mirrors they did not ask for. An observation about a habit can arrive as a judgment about the person who has it. Even accurate advice may require its recipient to concede something uncomfortable before considering whether it helps.

Neon Genesis Evangelion gives this resistance an image: the AT field, the barrier that makes each person a separate self. Taken as a metaphor, it suggests a boundary that both protects a person and keeps other people out. A story is also a mirror, but it can reflect a fault while barely touching the reader’s ego. The person making the mistake is someone else. Why does a story get through where direct advice bounces off?

The central hypothesis is that stories can postpone the need to defend the self. A reader follows another person’s decisions and recognizes their consequences before deciding whether the lesson applies personally. Narrative theories of identification and reduced persuasive intent support parts of this explanation. Experiments find a small average reduction in resistance to narrative messages. Moyer-Gusé (2008); Ratcliff and Sun (2020).

Consider being told that you interrupt people. The remark asks you to accept a verdict about yourself before you can weigh it. A film in which a character keeps interrupting lets a viewer notice the habit first and decide later whether it applies to them. It also leaves room to ask whether the character’s circumstances resemble the viewer’s own.

The messenger adds another complication. A recipient may distrust the speaker’s judgment or resent being corrected by them. Information about a source can be relevant to credibility, but it can also distract from the claim. A story may change this encounter without changing the quality of the evidence. Sperber et al. (2010).

What is the wall, and is ego the right word?

The wall is a shorthand for several kinds of resistance. Calling all of them ego makes an experience recognizable, but leaves its cause unclear. A person defending a belief about their competence may react differently from someone resisting an instruction because it restricts their freedom.

Self-affirmation research concerns threats to self-integrity. Identity-protective cognition concerns the protection of socially valued identities. Brehm’s reactance theory concerns threatened freedom; Dillard and Shen studied its expression through anger and negative thoughts during persuasion. Counterarguing is something a listener does, with several possible motives. The terms separate the source of resistance from the response it produces. Cohen and Sherman (2014); Kahan et al. (2007); Brehm (1966); Dillard and Shen (2005).

Freud’s ego belongs to a different theoretical system. It is an agency that negotiates between drives and the demands of reality and conscience, and parts of it can be unconscious. Everyday references to a large ego usually mean pride or self-importance. Popular spiritual usage often means a self or identity that should be transcended. None of these meanings identifies a single psychological variable that can simply be lowered until learning improves. Freud Museum London.

The mind as a selective receiver is a more useful model if its filters are specified. Existing knowledge affects what seems plausible. Trust affects whose testimony receives attention. A concern about belonging may affect which conclusions feel acceptable. These processes can operate at different stages: noticing information is separate from understanding it, and understanding it is separate from accepting it. A claim can remain intelligible while being excluded from the beliefs that guide a decision. Sperber et al. (2010); Kahan et al. (2007).

The receiver also protects. Sperber and colleagues call scrutiny of communicated information epistemic vigilance. People need ways to assess both a message and the reliability of its source. A filter that admits everything would admit deception along with correction. Learning requires discrimination; the problem is discrimination that tracks threats to identity more closely than the strength of evidence. Sperber et al. (2010).

This model has precedents in research on selective exposure and source evaluation. It need not imply a passive mind or a fixed setting. Choosing what to read can alter the information available for later judgment. The empirical question concerns how selection changes under different conditions. Hart et al. (2009).

Why do stories get past it?

Stories can reduce resistance, but the average advantage is small. Several mechanisms could contribute, and they concern different parts of learning. Watching a decision provides an example. Identifying with a character makes a perspective easier to inhabit. Absorption can leave less attention for arguing with the message. These explanations are compatible without being interchangeable. Ratcliff and Sun (2020).

Bandura’s observational-learning experiments separate acquiring a response from choosing to perform it. In a 1965 study, children watched a filmed model whose aggressive actions were rewarded, punished or left without consequences. Those who saw punishment initially reproduced fewer actions. Once all groups received incentives to reproduce what they had seen, the performance differences disappeared. The children could learn from another person’s experience even when they initially withheld the behavior. The study compared consequences for a model, rather than stories with advice. Bandura (1965).

Mar and Oatley propose that fiction simulates social experience. A reader follows what a character wants, the choices available and the consequences of acting. A simplified situation can make a relationship easier to examine than it would be amid the distractions of ordinary life. The consequences in a novel are selected by a writer, so a convincing fictional outcome still needs checking against the world. Mar and Oatley (2008).

Transportation is absorption in the story world. Green and Brock found that more transported readers expressed more story-consistent beliefs and detected fewer implausible elements. Slater and Rouner proposed that following the narrative competes with counterarguing against its embedded message. Moyer-Gusé added a second possibility: entertainment may provoke less reactance because the intention to persuade is less conspicuous. A reader can devote considerable thought to what happens while giving less thought to whether someone is trying to change their mind. Green and Brock (2000); Slater and Rouner (2002); Moyer-Gusé (2008).

Kaufman and Libby’s mirror experiment gives self-focus a direct test. In Study 3, 123 undergraduates read with either a mirror’s reflective surface or its back facing them. Experience-taking, the temporary adoption of a character’s position, averaged 5.52 with the reflection and 6.47 without it on a nine-point scale. Incorporation of the character’s traits also weakened. General transportation and story memory showed no significant differences. Kaufman and Libby (2012).

The manipulation increased self-focus without accusing anyone of a fault. Its relevance to unsolicited advice depends on the proposed connection between self-focus and defence.

Ratcliff and Sun pooled nine experiments comparing narrative and non-narrative messages. Resistance was lower for narratives: d = −.213, with a 95% confidence interval from −.318 to −.107. This is about one fifth of a standard deviation. Publication-bias adjustment reduced the estimate. The measures included counterarguing and perceived freedom threat; the comparisons were broader than unsolicited self-help. Ratcliff and Sun (2020).

Persuasion is a separate outcome. Braddock and Dillard pooled narrative effects on beliefs and behavior, including comparisons with no message and changes measured before and after exposure. Their estimates therefore include more than the benefit of rewriting equally informative advice as a story. Braddock and Dillard (2016).The estimates below are after outlier removal. Participant sets overlap across outcomes.

Here, r is an effect-size metric, not the percentage of people persuaded. The behavior estimate rests on only five effects. Van Laer and colleagues measured another relationship: transportation correlated with attitudes at corrected ρ = .44 and with critical thoughts at ρ = −.20. Those associations describe differences in absorption; they cannot be read as the causal benefit of changing a message’s format. Braddock and Dillard (2016); Van Laer et al. (2014).

Fiction and narrative also need separating. A biography can tell a story, while a novel can lecture. Green and Brock found that fiction labels generally made little difference. Appel and Malečkar likewise found no reliable persuasion difference between fiction and nonfiction labels. A fiction label did outperform a fake label, d = .48. Openly invented events can retain persuasive force; information presented as fabricated evidence faces a different credibility problem. These experiments changed labels, not every feature of a genre. Green and Brock (2000); Appel and Malečkar (2012).

The case for automatic empathy is weaker. Kidd and Castano’s 2013 experiments reported an immediate advantage for literary fiction on tests of inferring other people’s mental states. Panero and colleagues found no reliable advantage in a replication with 792 participants. A coordinated replication also failed to reproduce the first experiment, a failure Kidd and Castano acknowledged while defending other results. Kidd and Castano (2013); Panero et al. (2016); Camerer et al. (2018); Kidd and Castano (2018).

A 2024 synthesis of 70 experiments estimated a small overall cognitive benefit of fiction reading, g = .14, with a 95% confidence interval from .06 to .21. Significant outcome categories were empathy and mentalizing, though the overall estimate combines different tasks and comparison activities. Wimmer et al. (2024).

Teaching stories show how indirect recognition can be arranged. In 2 Samuel 12, Nathan describes a rich man taking a poor man’s lamb. David condemns the action before Nathan identifies him as the target. 2 Samuel 12:1-13.

Meadows uses the blind men and the elephant similarly in Thinking in Systems (Earthscan, 2009 edition, p. 7). Each observer encounters a part and mistakes it for the whole. The elephant represents the system; its behavior depends on relationships among parts, which an inventory of isolated elements leaves out. Pooling the observers’ descriptions is a reasonable further lesson about collective inquiry. Even a combined inventory, however, still needs an account of how the parts interact.

Meadows calls this version a Sufi story. Its teaching function explains its place in a systems primer: a visible error makes an abstract limitation easier to consider.Idries Shah’s Tales of the Dervishes collects stories in this tradition, including the elephant tale, with notes on their origin and use. The Idries Shah Foundation’s book page provides access to the collection.

Can readers supply the meaning or forget the messenger?

A story can feel acceptable because the reader has partly made its meaning, or because an idea later arrives without its source attached. Both routes can make an idea feel less external. Their connection to reduced defensiveness is a hypothesis; the evidence directly concerns interpretation and memory.

Readers can make a message congenial to prior beliefs. LaMarre and colleagues studied 332 viewers of The Colbert Report. Conservatives more often thought Colbert meant his ostensibly conservative remarks; liberals more often understood them as satire. The groups did not differ reliably in how funny they found him. Political ideology was measured, so the relationship is correlational. Enjoying the same performance was compatible with taking different lessons from it. LaMarre et al. (2009).

Projection here means bringing expectations and personal concerns to interpretation.It need not mean the psychoanalytic defence of attributing an unwanted trait to someone else. The reader’s contribution may produce recognition, but it can also preserve the belief a story was intended to challenge. A useful interpretation must account for the details that resist it, as well as the details that feel familiar.

Memory creates another opening. Johnson, Hashtroudi and Lindsay’s source-monitoring framework describes how people judge where remembered information came from. Content can survive while information about its origin is lost or misassigned. An idea encountered in a book may later feel like an unprompted thought. Johnson et al. (1993).

Brown and Murphy demonstrated cryptomnesia, or inadvertent reuse of remembered material as if it were new. Participants generated examples of categories, then recalled their own contributions and tried to produce fresh ones. Some reproduced other people’s answers despite instructions to avoid doing so. The feeling of originality was an unreliable guide to origin. These experiments support unconscious borrowing; they leave room for later reasoning to combine borrowed material into something different. Brown and Murphy (1989).

The sleeper effect concerns delayed persuasion. A message initially discounted because of a credibility cue can become more persuasive later under restricted conditions. Kumkale and Albarracín’s meta-analysis found that strong initial effects of both message and cue mattered. The cue’s influence can weaken relative to the message without complete forgetting. An idea may consequently lose part of the reason it was first distrusted. Applying that result to a forgotten book or a disliked adviser requires checking the initial discounting and the later change in belief. Kumkale and Albarracín (2004).

Can a story make a false idea feel true?

Reduced resistance can also help falsehood. The mechanisms that make an idea easy to entertain provide no guarantee that it deserves acceptance.

Marsh, Meade and Roediger embedded correct and incorrect information in fictional stories, then tested general knowledge. Readers reused misinformation as well as facts. Sometimes they believed information from the story had been known before the experiment. Knowing that a work is fiction offers incomplete protection against learning incidental claims from it. The failure can occur without a reader consciously deciding that an invented plot is historical fact. Marsh et al. (2003).

The feeling of discovery can add credibility of its own. Laukkonen and colleagues used anagram tasks alongside factual propositions. Among 268 retained participants, reported Aha experiences within the anagram condition accompanied higher truth ratings, including for false propositions. The Aha versus no-Aha contrast was d = .629. But Aha itself was not randomized, and the overall randomized comparison between task conditions was nonsignificant. The result links experienced insight with felt truth; it leaves the stronger causal account uncertain. Laukkonen et al. (2020).

A lesson discovered while following a character may seem especially persuasive because the reader has helped infer it. That extension to narrative is a hypothesis, not a result of the anagram experiment. It suggests a practical distinction between understanding why an explanation fits and checking whether it is correct. A coherent ending supplies the former more easily than the latter.

For a factual claim, the check is evidence beyond the story. For an analogy, it is the relationship being carried across: whether the consequences depend on the same conditions in both cases. A tale can make exploitation recognizable while giving a poor account of how often it occurs or what would prevent it. Those questions require evidence the plot may never contain.

Can the wall lock, and can it be loosened?

Resistance can reinforce itself without becoming an impermeable barrier. The distinction matters because a permanent lock predicts failure of correction, whereas selective resistance permits changes in some beliefs under some conditions.

Hart and colleagues’ meta-analysis found a general preference for information congenial to existing beliefs. Accuracy motives could alter that preference: challenging information could attract attention when it helped with a task. Selection therefore changes what reaches the receiver, and it varies with what the person is trying to accomplish. Hart et al. (2009).

Interpretation can filter a message after exposure. Kahan and colleagues found that cultural worldviews helped explain differences in risk judgments within demographic groups. Their identity-protective account proposes that people assess information partly in ways that preserve valued social identities. This observational evidence supports studying identity-related selectivity, rather than treating every refusal as a deliberate decision to remain ignorant. Kahan et al. (2007).

Together, selection and interpretation could form a self-reinforcing loop. Agreeable material receives more attention, and ambiguous material supplies apparent confirmation. The loop could persist without a conscious decision to build it. Calling it a single ego setting conceals those separate processes. Its strength would have to be tested by varying the information or the conditions under which it is considered.

The backfire effect places a limit on the strongest version of the wall. It predicts that a factual correction can increase the false belief it targets. Wood and Porter tested corrections on 52 issues in five experiments involving more than 10,100 participants. No correction reliably produced factual backfire. A person can dislike a correction while learning from it, and can revise a factual belief while retaining a political identity. Wood and Porter (2019).

Self-affirmation offers a way to reduce the cost of reconsideration. In this research, people reflect on a valued part of themselves separate from the threatened belief. The purpose is to make an error less consequential for their overall worth. It differs from praising the disputed position or telling someone that every view is equally valid. Cohen and Sherman (2014).

Cohen, Aronson and Steele found that affirmation improved evaluations of opposing evidence, produced attitude change in one study and increased criticism of an advocate on the participant’s own side in another. The scientific reports used as evidence were fictitious. The studies therefore demonstrate increased openness under those conditions, not a general improvement in accuracy. They also show why receptivity need not require eliminating the self: supporting it can reduce defence of a particular belief. Cohen et al. (2000).

Ego dissolution names a different experience: weakening of the ordinary felt self or its boundary with the world. Nour and colleagues developed an inventory from retrospective reports by 691 people covering 1,828 drug experiences. The eight-item measure distinguished dissolution from ego inflation. It measures reported experience, rather than permanent loss of identity or reliably better reasoning. Nour et al. (2016).

Carhart-Harris and Friston’s REBUS model, short for relaxed beliefs under psychedelics, proposes reduced precision of high-level prior beliefs.Precision here means the weight a belief receives in constraining interpretation. Lower weight would allow previously discounted signals more influence. REBUS is a theoretical account of psychedelic action; the correspondence with a story loosening a belief remains speculative. A less constraining prior could admit useful evidence or a misleading suggestion. Carhart-Harris and Friston (2019).

Can believing oneself beyond ego become another defence?

A person can sincerely feel unbiased while judging selectively. The beholding paradox is that a person may regard their own apparent transcendence from within the same self-evaluation processes they think they have escaped.

Pronin, Lin and Ross found a bias blind spot: participants attributed more bias to others than to themselves, even when they had opportunities to observe their own biased judgments. Looking inward for a sincere motive gave a different answer from evaluating behavior. A conviction that other people are narrow-minded can therefore coexist with an unexamined selectivity of one’s own. Pronin et al. (2002).

Practices intended to quiet the self are not automatically exempt. Gebauer and colleagues studied 93 yoga students and 162 meditators. Self-enhancement was greater immediately after practice than in comparison conditions. Measures included favorable assessments of oneself relative to others. These were immediate outcomes among the studied practitioners; they cannot settle the effects of lifelong practice. Gebauer et al. (2018).

Trungpa’s Cutting Through Spiritual Materialism offers a related warning from Buddhist teaching. Spiritual practice can become a possession or achievement through which the self asserts its importance. This is a conceptual account, distinct from the controlled comparisons. It describes how an identity organized around having less ego might become another identity to protect. (Trungpa, Cutting Through Spiritual Materialism, Shambhala, 1973.)

The paradox becomes useful when it changes the test. Openness should be judged by treatment of unwelcome evidence, including willingness to apply the same standard to a favored claim and an opposing one. Declaring every improvement another illusion would make the theory impossible to disprove. A person who notices a mistake and corrects it has shown a specific capacity, even if complete freedom from bias remains beyond demonstration.

How far does the power of stories go?

A popular claim holds that stories are more powerful than money. No empirical ranking supports that as a general rule, because power needs an outcome. Changing a belief differs from buying someone’s time. Money can alter the options available; a story can alter which options seem desirable. Their relative effects require a specified choice and a comparison that holds other conditions steady. Narrative meta-analyses measure persuasion, not this universal contest. Braddock and Dillard (2016).

A story also differs from an ideology. Stories represent events and their relations, often through people pursuing goals. Ideologies organize beliefs about how a social world works and how it should be arranged. A worldview can be expressed as principles without a plot, while a story can portray conflicting worldviews without choosing between them. A promotion story might support a belief in deserved success or invite scrutiny of inherited advantages. Ideology helps determine the reading. Jost et al. (2009).

Stories often work through emotion, and emotion shapes judgment and choice in its own right, sometimes helpfully and sometimes harmfully. Lerner et al. (2015).

Harari argues that shared stories enable large-scale cooperation. His account makes common beliefs central to institutions and collective action. It is a historical argument, not a conclusion of the narrative experiments. If story includes all shared belief, money’s dependence on common acceptance becomes part of the definition. A comparison between stories and money then compares a broad category with one of its proposed instances. Harari (2018).

Smith and colleagues studied the social role of storytelling among Agta hunter-gatherers in the Philippines. Stories conveyed cooperative norms; camps with more skilled storytellers were more cooperative, and skilled storytellers were preferred as campmates. The evidence is largely correlational. It documents contemporary social relationships without locating the prehistoric origin of storytelling. Smith et al. (2017).

Explicit reflection on narrative influence is ancient. In Republic, Book II, 377b to 378e, Plato discusses controlling the stories children hear because the beliefs they form can persist. This supplies an early written argument about narrative education. It cannot date humanity’s first recognition of storytelling’s power; awareness can precede a surviving written record. Plato, Republic II.

The familiar claim about a good story has a verifiable fictional speaker. In Game of Thrones, season 8, episode 6, The Iron Throne (2019), Tyrion Lannister says, “There is nothing in the world more powerful than a good story.” The screenplay is credited to David Benioff and D. B. Weiss. Tyrion is arguing for Bran as king.

The Steve Jobs version names the storyteller as the most powerful person. Bernadette Jiwa’s 2013 post attributes it to a 1994 conversation at NeXT, crediting Tomas Higbey’s recollection. A contemporary recording or transcript remains unverified. William Golding’s 1983 Nobel lecture provides an earlier related claim: writers’ words might become the world’s most powerful instrument. Its conditional wording and subject differ from Tyrion’s sentence. Jiwa (2013); Golding, official site.

Jonathan Gottschall’s The Storytelling Animal offers a general introduction to the argument that stories help people simulate social life and shape shared beliefs. Its broad account can be read alongside the smaller effects and replication failures in the experimental literature. For the question of resistance specifically, Ratcliff and Sun provide the more focused starting point. Gottschall, book description; Ratcliff and Sun (2020).

Are there three levers on the same gate?

Stories, self-affirmation and ego dissolution may each loosen some constraint on revision, but a shared mechanism remains untested. A story may shift attention away from defending the current self. Affirmation may make an error less costly to self-worth. Ego dissolution changes the reported experience of a bounded self, with REBUS proposing altered weighting of prior beliefs. These are distinct descriptions with distinct measures. Moyer-Gusé (2008); Cohen and Sherman (2014); Nour et al. (2016); Carhart-Harris and Friston (2019).

A first experiment could cross narrative versus direct presentation with values affirmation versus a neutral task. Content and evidential strength would remain matched. Measures would separate perceived threat from counterarguing, then test whether belief revision tracks evidence quality. Asking readers what they thought the message meant would detect selective reinterpretation; testing source memory would distinguish acceptance from lost attribution.

If both methods sufficiently reduce the same obstacle, combining them should add little. Independent mechanisms might produce additional benefits, although ceiling effects can imitate redundancy and a shared mechanism can still produce additive gains before saturation. Ego dissolution would require a separate suitable study. The decisive outcome would be better discrimination between strong and weak evidence, including when the more appealing story carries the weaker case.

Appendix

Method

Sources were checked in September 2026 against the papers, publisher pages or author copies linked in the text. Effect sizes and sample sizes are quoted from the papers. Some things could not be verified: Steve Jobs’s exact words, an independent replication of the mirror experiment, and any joint test of the three levers in the final section, which remains a proposal.

Related

• Red Trees shares cognitive bias
• From A to B shares memory