The Persistence Engine: Why the Bond and the Belief Outlast the Reading

Abstract

The Immersion Engine of Explicit Romance describes a machine that bonds the reader to a character and keeps the bond paying out. The Voice in Your Ear describes how a performed voice amplifies that bond. The Persuasion Engine describes how the same immersion rewrites what the reader believes. All three describe the machine while it runs. This document asks what the machine leaves behind, across every way a reader's habit can change.

The intuitive answer is that the effect fades, that an attitude shaped by reading reverts once the reading slows, that a bond to a fictional character dissolves once there is no next chapter. The evidence points the other way.

No reader leaves the engine unchanged. Whether she reads on, reads less, or stops entirely, what the engine built persists.

A reader's habit can move three ways, and the engine wins on all of them. She can continue, and it keeps building on the structures already laid down. She can taper, reading less but still returning to the genre, and the dose that remains keeps those structures live while the residue of the heavier years stays in place. Or she can stop entirely, the hardest test, the one case where the effect has every reason to fade. It does not. A reader who consumed explicit romance compulsively for years does not return to baseline on any of the four layers the engine acts on, even when she quits. The parasocial bond does not dissolve; it goes dormant and reactivates on contact. The attitudes the fiction installed do not decay; they persist, and on the cleanest measure they grow. The cultivated worldview and its sexual scripts harden into automatic defaults that run without the source. And the reward system that drove the compulsion stays sensitized, so craving can rise even in full abstinence rather than fall.

Cessation is the limiting case, not the frame. If the engine's outputs survive a clean break, they survive the far more common trajectories of continued and reduced reading with room to spare. The unifying principle is borrowed from the science of extinction. Stopping a learned behavior does not erase the learning. It installs a new, fragile, context-dependent layer of inhibition on top of an association that remains intact, and the original resurfaces with time, with a change of context, or with a single reminding cue. Abstinence suppresses the engine. It does not dismantle it.

The evidence spans four domains:

1. The persistence and reactivation of parasocial bonds after their object is gone
2. The durability, and the delayed growth, of attitudes changed by narrative
3. The hardening of a cultivated worldview and its sexual scripts into automatic defaults
4. The neurobiology of cessation: extinction as suppression, the incubation of craving, and relapse

Read together they describe a single asymmetry. The bond is easy to form and hard to lose, the belief easy to absorb and hard to revoke, the script easy to learn and hard to overwrite, the craving easy to condition and hard to extinguish. The engine is built to be entered and not built to be left.

1The Bond Does Not Dissolve; It Goes Dormant

The first thing the engine builds is a parasocial bond, and a parasocial bond is not an event that ends with the page. The field draws a hard line between the two phenomena. A parasocial interaction is the in-the-moment response confined to the exposure itself, while a parasocial relationship is an enduring, cross-situational bond that persists between exposures, carried in memory as an ongoing internal representation of the figure ¹. The relationship is, by definition, the part that survives the closing of the book. The question is not whether something persists, but how much, and for whom.

How much is a function of how deep the bond was, and the relationship between depth and durability is direct. When a long-running series ends, the intensity of the prior bond is the dominant predictor of the grief that follows. In a survey of 1,289 fans taken after a finale, the strength of the parasocial bond explained 55 percent of the variance in breakup distress ², and the canonical study of the Friends finale found bond strength the single best predictor of grief at the loss ³. The grief is not metaphorical. Cohen's foundational work found that the anticipated loss of a favorite character produces reactions in kind with the dissolution of a real social relationship ⁴, and a controlled experiment found that closeness, not the fictional status of the relationship, determines the intensity of the grief, so that an intimate parasocial loss is felt as an intimate loss ⁵. A bond the engine built to be intense is a bond whose loss is mourned.

Who suffers the most durable bonds is the same population the engine selects for. Trait anxiety in adult attachment predicts stronger enduring parasocial relationships, while avoidance does not ⁶, and the most anxiously attached respondents anticipate the most intense distress at a character's loss ⁴. Loneliness compounds it. Lonely readers do not necessarily form stronger bonds, but they suffer more acutely when a bond dissolves ³, and the neural signature is measurable: in lonelier individuals the medial prefrontal boundary between real and fictional others is blurred, so the brain files the character closer to a real friend than a non-lonely brain does ⁷. The reader most likely to consume compulsively, the anxious and the lonely, is the reader whose bonds are encoded deepest and surrendered hardest.

The decisive evidence is that the bond reactivates rather than rebuilds. A bond that merely faded would have to be reconstructed from scratch on return. Instead, bonds survive long absences and switch back on at contact. Fans of a band that had been gone for fourteen years greeted its reunion by demonstrating that their parasocial relationship had remained intact across the entire gap, not lapsed and reformed ⁸. Even a temporary loss exposes the same persistence: during a writers' strike, viewers whose shows were only suspended, who knew the characters would return, experienced distress proportional to their bond strength, managing an active sense of loss rather than waiting to restart a cold relationship ⁹. The bond is stored, not spent.

The compulsive reader is the maximal case. She did not form a shallow, few-weeks attachment; she lived inside the genre for years, and her bonds are the high-intensity, deeply encoded, anxiously held kind the evidence ties to the most durable attachment and the most painful loss. For her the bond does not clear with absence. It goes dormant, intact, and waits.

2The Belief Does Not Fade; It May Grow

The second output of the engine is a changed belief, and a changed belief is more durable than the bond. The intuition that persuasion decays once the source is gone is not merely wrong for fiction. On the cleanest measure it is backwards.

The keystone is the absolute sleeper effect. Appel and Richter had readers absorb a fictional story carrying real-world assertions, then measured belief immediately and again two weeks later, and found the persuasive effect was larger after the delay than right after reading ¹⁰. The mechanism is source dissociation. The reader forgets the fictional origin faster than the content, and a claim detached from the label "something I read in a novel" becomes simply "something I know." Time does not wear the belief down. It strips the belief of the one cue that marked it as fiction.

The aggregate evidence confirms that narrative belief does not decay with delay. A meta-analysis of long-term persuasion found the advantage of narrative over non-narrative statistically indistinguishable at delays running to 180 days, and notably found that transportation, which drives the immediate effect, does not mediate the delayed one, implying a separate mechanism sustains the belief after the immersion has passed ¹¹. The most recent and comprehensive synthesis, 77 experiments across more than 24,000 participants, found narrative effects that persist for weeks, months, and in some cases up to a year ¹². And belief learned from fiction is unusually resistant to correction. Readers absorb false assertions from stories and reproduce them later as fact ¹³, warnings that the fiction contains errors fail to suppress the absorption, and highlighting the errors makes readers reproduce them more, not less ¹⁴. Once misinformation has filled a slot in the reader's model of the world, even a correction the reader can recall does not displace it ¹⁵.

The durability sorts by type, and the genre's targets fall on the lasting side. When an attitude does fade, the ones that revert are those tied to a daily behavioral habit, where routine conduct pulls the reader back, while abstract belief-type attitudes persist at twelve months ¹⁶. The attitudes the persuasion engine targets, toward sexuality, monogamy, and identity, are belief-type, not habit-bound: convictions about what is acceptable and normal, not chores under social pressure. They are the kind that endure. And some of the shift does not even surface until later, a sense of personal relevance emerging only at a delayed measure, absent right after reading ¹⁷. The belief can still be settling into place long after the book is closed.

3The Worldview Hardens Into a Default

Beneath any single belief sits the worldview, the background sense of what is common and acceptable, and the worldview is the most entrenched output of all, because repetition converts it into something automatic.

Cultivation, the slow tuning of that background sense by accumulated exposure, is remarkably stable. A meta-analysis spanning five decades and 3,842 effect sizes found the cultivation correlation steady at r = .107, holding across the entire upheaval of the media landscape ¹⁸, and a two-wave panel found that viewing at the first measure predicted a more cultivated belief a year later, with little evidence for the reverse path ¹⁹. The reason heavy exposure persists is that it changes how fast the mind reaches for what the stories supplied. Shrum showed that heavy viewers do not merely hold cultivated estimates; they retrieve them faster, because constant exposure keeps the supplied exemplars at the top of memory, chronically accessible and pulled up by default ²⁰. Chronic accessibility is the doorway to automaticity. A construct made chronically accessible activates without intent and comes to dominate judgment the longer the interval since it was last deliberately primed, applied unintentionally and uncontrollably ²¹. What the genre keeps fresh through weekly dosing does not need the dosing to keep firing once it is chronic.

The same hardening runs through the sexual script in particular. A script, in the foundational theory, governs desire at three levels, the deepest of which is the intrapsychic, the private ordering of images that elicits and sustains arousal ²². A script repeated and rewarded consolidates into a default that feels natural and self-evident, the phenomenological mark of an automatic pattern rather than a chosen one. The acquisition model predicts the durability: a sexual script absorbed from media is most permanent when it is congruent with what the reader already holds, so the genre's heaviest readers, whose priors the genre has itself been shaping for years, are the readers in whom each new script lands as reinforcement and locks in ²³. Longer-term media effects are theorized to persist exactly when they accommodate higher-order schemas, the generic structures that apply across many situations and outlast any specific memory ²⁴. The worldview is not a fact the reader can forget. It is the lens the reader now sees through.

The mechanisms here all run one way, toward entrenchment. Chronic accessibility hardens into automaticity, script consolidation into default, schema accommodation into permanence, and the closest direct evidence finds that earlier-learned sexual scripts resurface even in people who consciously hold the opposite belief. How those sexual scripts are written in the first place is the subject of The Script Engine; once written, they behave like every other output of the machine, hardening rather than fading. A worldview built this way is not a fact the reader can decide to drop. It is the lens she now sees through, and a lens does not wait to be summoned. It returns the moment deliberate attention drops.

4The Reward System Stays Primed

The deepest output of the engine is a conditioned reward circuit, and this is where cessation is least like erasure. The neurobiology of stopping a compulsive behavior is not the neurobiology of forgetting it.

The governing principle is that extinction is new learning, not unlearning. When a rewarded behavior stops being rewarded and fades, the original association is not erased; a separate, inhibitory memory is laid down on top of it, and the original survives intact beneath ²⁵. The proof is that the original response returns through three routes: spontaneously with the passage of time, with a change of context, and with a single reminding cue, each demonstrating that the conditioning memory and the extinction memory coexist ²⁶. Applied to a reader who quits, the model is exact. Abstinence does not delete the years of pairing between the genre and its reward. It builds a fragile, context-bound layer of restraint over an association that is still there, waiting for the context to shift or the cue to arrive.

Worse for the reader who hopes the craving will simply fade, abstinence can make it grow. In the founding demonstration, cue-induced seeking for a withdrawn reward was minimal on the first day of abstinence and climbed steadily to a peak around two months in, the incubation of craving ²⁷, and the same time course has since been shown for a natural reward, palatable food, not only for drugs ²⁸. The wanting outlasts everything else. Incentive-sensitization research finds that sensitized wanting can persist for years after use stops, independent of whether the person still likes the reward or even cognitively wants it, and, far from fading in abstinence, can grow during it ²⁹. The reader who stops and feels the pull intensify over the following weeks is not failing. The system is doing what conditioned reward systems do.

The physical traces of the conditioning persist on the same timescale, and they are laid down by sexual reward specifically. The transcription factor ΔFosB, a molecular switch for compulsive reward-seeking, accumulates with repeated stimulation and persists in the reward circuitry for one to two months after it stops, and it accumulates not only to drugs but to natural rewards including sexual behavior ^{30 31}. Dopamine D2 receptor downregulation, the substrate of tolerance, persists for months into abstinence ³². And the bonding half of the loop is no more erasable than the wanting half. In monogamous voles, a pair bond's behavioral preference outlasts weeks of complete separation ³³, and in humans, prolonged grief activates the reward circuitry toward the lost attachment figure, the nucleus accumbens firing to reminders of the absent person as it would to a craved reward ^{34 35}. A bond, once neurally encoded, behaves like a reward the brain keeps seeking after it is gone, and the cue that fires it can be as slight as a voice, which alone releases the bonding hormone ³⁶.

The clinical label is beside the point; the mechanism is what matters. The wanting-and-liking dissociation that defines compulsive reward has been observed directly in compulsive sexual behavior, greater desire for sexual cues without a proportional rise in liking ³⁷, and the conditioned cues that drive it are unusually hard to extinguish ³⁸. The reward system of a compulsive reader who slows or stops is a primed system, not a reset one.

5The Asymmetry

The four domains close on a single shape, and it is the shape of every prior document in this series read in reverse.

1. The bond the engine builds goes dormant rather than dissolving, stored as an internal representation that reactivates on contact instead of rebuilding from nothing.
2. The belief the engine installs persists across delays and, as the source is forgotten, can grow, and where it targets conviction rather than habit it does not revert.
3. The worldview the engine cultivates hardens into a chronically accessible default that runs automatically, without the exposure that built it.
4. The reward circuit the engine conditions stays sensitized, its craving capable of rising in abstinence, its physical traces persisting for months, its original associations surviving extinction beneath a thin and fragile inhibition.

The reinforcing spiral the persuasion engine describes does not require cessation to break it. A reader pulled in by a mild predisposition is shifted further than she would have moved alone, and the shift pulls her deeper, and the depth shifts her further still. Stopping interrupts the dosing. It does not unwind the turns the spiral has already taken.

This is why extinction is the right frame for the whole machine, not only its reward circuit. Across all four domains the pattern is the same: easy in, hard out. A bond forms in hours and survives years of absence. A belief is absorbed without resistance and resists correction afterward. A script is learned through pleasure and reasserts itself under stress. A craving is conditioned by repetition and incubates in the quiet after it ends. The compulsive reader, dosed weekly for years, is the maximal case of every one of these, the deepest bond, the most integrated belief, the most automatic script, the most sensitized circuit, and therefore the reader for whom no change in the habit, not even stopping outright, undoes what was built.

The industry's craft and the science of immersion explain why the reader cannot put the book down. The science of persuasion and cultivation explains what she believes while she reads. The science of persistence explains the part that should trouble most: that changing how she reads, even putting the book down for good, does not give her back the person she was before she picked it up.

The engine is built to be entered. It is not built to be left, and quitting it only turns it off.

References

Footnotes

1. Schramm, Holger, and Tilo Hartmann. "The PSI-Process Scales: A New Measure to Assess the Intensity and Breadth of Parasocial Phenomena." Communications 33, no. 4 (2008): 385–401. Formalizes the distinction between the in-exposure parasocial interaction and the enduring, cross-situational parasocial relationship that persists between exposures. https://doi.org/10.1515/COMM.2008.025 ↩

2. Gerace, Adam. "When TV Neighbours Become Good Friends: Understanding Neighbours Fans' Feelings of Grief and Loss at the End of the Series." PLOS ONE 19, no. 6 (2024): e0302160. In a survey of 1,289 fans after the finale, parasocial bond strength explained 55 percent of the variance in breakup distress. https://doi.org/10.1371/journal.pone.0302160 ↩

3. Eyal, Keren, and Jonathan Cohen. "When Good Friends Say Goodbye: A Parasocial Breakup Study." Journal of Broadcasting & Electronic Media 50, no. 3 (2006): 502–523. The strength of the parasocial bond was the single best predictor of grief at a favorite character's loss. https://doi.org/10.1207/s15506878jobem5003_9 ↩ ↩²

4. Cohen, Jonathan. "Parasocial Break-Up from Favorite Television Characters: The Role of Attachment Styles and Relationship Intensity." Journal of Social and Personal Relationships 21, no. 2 (2004): 187–202. Anticipated character loss produced reactions in kind with real-relationship dissolution; anxious-ambivalent viewers anticipated the most distress, and weaker bonds dissolved with little. https://doi.org/10.1177/0265407504041374 ↩ ↩²

5. Baker, Aaron C., and Elizabeth L. Cohen. "Relational Closeness, Not Parasociality, Determines the Intensity of Grief Responses to Celebrity Death." Death Studies 48, no. 8 (2024): 873–878. In a controlled experiment, closeness rather than the fictional or real status of the relationship determined grief intensity; intimate parasocial loss was felt as intimate loss. https://doi.org/10.1080/07481187.2023.2276301 ↩

6. Rain, Marina, and Raymond A. Mar. "Adult Attachment and Engagement with Fictional Characters." Journal of Social and Personal Relationships 38, no. 9 (2021): 2792–2813. Trait attachment anxiety predicted stronger enduring parasocial relationships; avoidance did not. https://doi.org/10.1177/02654075211018513 ↩

7. Broom, Timothy W., and Dylan D. Wagner. "The Boundary Between Real and Fictional Others in the Medial Prefrontal Cortex Is Blurred in Lonelier Individuals." Cerebral Cortex 33, no. 16 (2023): 9677–9689. In lonelier people the mPFC distinction between real and fictional others was significantly attenuated. https://doi.org/10.1093/cercor/bhad237 ↩

8. Dibble, Jayson L., and Sarah F. Rosaen, and colleagues, on the New Kids on the Block reunion. "Exploring Relational Maintenance Within the Context of Parasocial Relationships." Journal of Media Psychology 21, no. 4 (2009): 171–182. A thematic analysis of fan postings around a fourteen-year reunion found that bonds had remained committed across the gap rather than lapsing and reforming. https://doi.org/10.1027/1864-1105.21.4.171 ↩

9. Lather, Julie, and Emily Moyer-Gusé. "How Do We React When Our Favorite Characters Are Taken Away? An Examination of a Temporary Parasocial Breakup." Mass Communication and Society 14, no. 2 (2011): 196–215. During a writers' strike, viewers whose shows were only suspended experienced distress proportional to bond strength, and those who watched for companionship suffered most. https://doi.org/10.1080/15205431003668603 ↩

10. Appel, Markus, and Tobias Richter. "Persuasive Effects of Fictional Narratives Increase Over Time." Media Psychology 10, no. 1 (2007): 113–134. An absolute sleeper effect: the persuasive impact of claims in fiction was larger two weeks after reading than immediately after, as the fictional source was forgotten and the content integrated as knowledge. https://doi.org/10.1080/15213260701301194 ↩

11. Oschatz, Corinna, and Caroline Marker. "Long-Term Persuasive Effects in Narrative Communication Research: A Meta-Analysis." Journal of Communication 70, no. 4 (2020): 473–496. Across 14 studies the narrative advantage was statistically indistinguishable at immediate and delayed measurement (delays to 180 days); transportation mediated the immediate effect but not the delayed one. https://doi.org/10.1093/joc/jqaa023 ↩

12. Rahmani, Bardia, Bryan Montano, Dylan W. Groves, and Donald P. Green. "The Persuasive Effects of Narrative Entertainment: A Meta-Analysis of Recent Experiments." Behavioural Public Policy (2025). 77 experiments and more than 24,000 participants; narrative effects persisted for weeks and months, and in some cases up to a year. https://doi.org/10.1017/bpp.2025.10010 ↩

13. Marsh, Elizabeth J., Michelle L. Meade, and Henry L. Roediger III. "Learning Facts from Fiction." Journal of Memory and Language 49, no. 4 (2003): 519–536. Readers spontaneously absorbed false assertions from stories and reproduced them on later general-knowledge tests, even when the errors contradicted well-known facts. https://doi.org/10.1016/S0749-596X(03)00092-5 ↩

14. Fazio, Lisa K., and Elizabeth J. Marsh. "Ironic Effects of Drawing Attention to Story Errors." Memory & Cognition; and Marsh and Fazio, "Learning Errors from Fiction," Memory & Cognition 34, no. 5 (2006): 1140–1149. General warnings did not reduce reliance on errors absorbed from fiction, and highlighting the errors increased their later reproduction rather than reducing it. https://doi.org/10.3758/BF03193260 ↩

15. Lewandowsky, Stephan, Ullrich K. H. Ecker, Colleen M. Seifert, Norbert Schwarz, and John Cook. "Misinformation and Its Correction: Continued Influence and Successful Debiasing." Psychological Science in the Public Interest 13, no. 3 (2012): 106–131. The continued-influence effect: misinformation keeps shaping belief after correction, even when the reader correctly recalls the correction. https://doi.org/10.1177/1529100612451018 ↩

16. Hormes, Julia M., Paul Rozin, Melanie C. Green, and Katrina Fincher. "Reading a Book Can Change Your Mind, but Only Some Changes Last for a Year." Frontiers in Psychology 4 (2013): 778. Most attitudes changed by a book reverted by twelve months, but the reversion was concentrated in habit-linked attitudes, while abstract belief-type attitudes persisted. https://doi.org/10.3389/fpsyg.2013.00778 ↩

17. Moyer-Gusé, Emily, and Robin L. Nabi. "Explaining the Effects of Narrative in an Entertainment Television Program: Overcoming Resistance to Persuasion." Human Communication Research 36, no. 1 (2010): 26–52. Reduced reactance and counterarguing appeared immediately, while a heightened sense of personal vulnerability emerged only at the two-week follow-up, absent right after viewing. https://doi.org/10.1111/j.1468-2958.2009.01367.x ↩

18. Hermann, Erik, Michael Morgan, and James Shanahan. "Television, Continuity, and Change: A Meta-Analysis of Five Decades of Cultivation Research." Journal of Communication 71, no. 4 (2021): 515–544. Overall cultivation effect r = .107 across 3,842 effect sizes from 406 samples, stable across decades. https://doi.org/10.1093/joc/jqab014 ↩

19. Romer, Daniel, Kathleen Hall Jamieson, and Sean Aday, on genre-specific cultivation. "Lagged Associations Between Television Viewing and Fatalistic Beliefs." Viewing at the first wave predicted a more cultivated belief one year later, with little evidence for the reverse path. https://pmc.ncbi.nlm.nih.gov/articles/PMC4297652/ ↩

20. Shrum, L. J. "Psychological Processes Underlying Cultivation Effects: Further Tests of Construct Accessibility." Human Communication Research 22, no. 4 (1996): 482–509; and Shrum, "The Cognitive Processes Underlying Cultivation Effects," Communication Methods and Measures 1, no. 1 (2007): 33–57. Heavy viewers gave higher estimates and retrieved them faster, indicating chronic accessibility; prompting them to consider the source dissolved the first-order effect. https://doi.org/10.1111/j.1468-2958.1996.tb00376.x ↩

21. Bargh, John A., Wendy J. Lombardi, and E. Tory Higgins. "Automaticity of Chronically Accessible Constructs in Person × Situation Effects on Person Perception." Journal of Personality and Social Psychology 55, no. 4 (1988): 599–605. Chronically accessible constructs activated unintentionally and came to dominate judgment as the interval since priming lengthened. https://doi.org/10.1037/0022-3514.55.4.599 ↩

22. Simon, William, and John H. Gagnon. "Sexual Scripts: Permanence and Change." Archives of Sexual Behavior 15, no. 2 (1986): 97–120; and Gagnon and Simon, Sexual Conduct (Chicago: Aldine, 1973). Sexual conduct as learned script operating on cultural, interpersonal, and intrapsychic levels, the last governing the private ordering of images that sustains arousal. https://doi.org/10.1007/BF01542219 ↩

23. Wright, Paul J. "Mass Media Effects on Youth Sexual Behavior: Assessing the Claim for Causality." Annals of the International Communication Association 35, no. 1 (2011): 343–385. The acquisition-activation-application model: media scripts are most durable when congruent with the consumer's preexisting scripts. https://doi.org/10.1080/23808985.2011.11679121 ↩

24. Ecker, Ullrich K. H., and colleagues. "Conceptualizing Long-Term Media Effects on Societal Beliefs." Annals of the International Communication Association 45, no. 1 (2021): 75–93. Media effects persist when they accommodate higher-order schemas, which are generic and apply across many situations and time periods rather than remaining as specific memories. https://doi.org/10.1080/23808985.2020.1857598 ↩

25. Bouton, Mark E. "Context and Behavioral Processes in Extinction." Learning & Memory 11, no. 5 (2004): 485–494. Extinction is a context-modulated inhibitory learning that suppresses but does not erase the original association. https://doi.org/10.1101/lm.78804 ↩

26. Dunsmoor, Joseph E., Yael Niv, Nathaniel Daw, and Elizabeth A. Phelps. "Rethinking Extinction." Neuron 88, no. 1 (2015): 47–63. Spontaneous recovery, contextual renewal, and reinstatement show that conditioning and extinction memories coexist and the original can resurface. https://doi.org/10.1016/j.neuron.2015.09.028 ↩

27. Grimm, Jeffrey W., Bruce T. Hope, Roy A. Wise, and Yavin Shaham. "Incubation of Cocaine Craving After Withdrawal." Nature 412, no. 6843 (2001): 141–142. Cue-induced seeking was minimal on the first day of abstinence and increased to a peak around two months in. https://doi.org/10.1038/35084134 ↩

28. Madangopal, Rajtarun, et al. "Incubation of Palatable Food Craving Is Associated with Brain-Wide Neuronal Activation in Mice." PNAS 119, no. 45 (2022): e2209382119. A natural reward, palatable food, showed the same incubation time course as drugs, peaking at sixty days. https://doi.org/10.1073/pnas.2209382119 ↩

29. Berridge, Kent C., and Terry E. Robinson. "Liking, Wanting and the Incentive-Sensitization Theory of Addiction." American Psychologist 71, no. 8 (2016): 670–679; and Robinson and Berridge, "The Incentive-Sensitization Theory of Addiction 30 Years On," Annual Review of Psychology 76 (2024): 29–58. Sensitized wanting can persist for years and grow during abstinence, independent of liking or cognitive desire, and the 2024 update extends the account to behavioral rewards. https://doi.org/10.1037/amp0000059 ↩

30. Nestler, Eric J., Michel Barrot, and David W. Self. "ΔFosB: A Sustained Molecular Switch for Addiction." PNAS 98, no. 20 (2001): 11042–11046. A stable form of ΔFosB accumulates with repeated reward exposure and persists in the reward circuitry for one to two months after it stops. https://doi.org/10.1073/pnas.191352698 ↩

31. Wallace, Deanna L., et al. "The Influence of ΔFosB in the Nucleus Accumbens on Natural Reward-Related Behavior." Journal of Neuroscience 28, no. 41 (2008): 10272–10277. ΔFosB accumulates in response to natural rewards including sexual behavior, not only to drugs. https://doi.org/10.1523/JNEUROSCI.1531-08.2008 ↩

32. Volkow, Nora D., Joanna S. Fowler, Gene-Jack Wang, Ruben Baler, and Frank Telang. "Imaging Dopamine's Role in Drug Abuse and Addiction." Neuropharmacology 56, supplement 1 (2009): 3–8. Striatal D2 receptor reductions, the substrate of tolerance, persist for months into protracted abstinence. https://doi.org/10.1016/j.neuropharm.2008.05.022 ↩

33. Sadino, Julie M., et al. "Prolonged Partner Separation Erodes Nucleus Accumbens Transcriptional Signatures of Pair Bonding in Male Prairie Voles." eLife 12 (2023): e80517. A pair bond's behavioral partner preference persisted through four weeks of complete separation even as its molecular signature eroded. https://doi.org/10.7554/eLife.80517 ↩

34. O'Connor, Mary-Frances, et al. "Craving Love? Enduring Grief Activates Brain's Reward Center." NeuroImage 42, no. 2 (2008): 969–972. In prolonged grief, reminders of the deceased activated the nucleus accumbens, a reward-craving response to an absent attachment figure. https://doi.org/10.1016/j.neuroimage.2008.04.256 ↩

35. Kakarala, Saren E., et al. "The Neurobiological Reward System in Prolonged Grief Disorder: A Systematic Review." Psychiatry Research: Neuroimaging 303 (2020): 111135. Prolonged grief shows reward-system activation toward the lost figure and altered oxytocin signaling, the same system that formed the bond. https://doi.org/10.1016/j.pscychresns.2020.111135 ↩

36. Seltzer, Leslie J., Toni E. Ziegler, and Seth D. Pollak. "Social Vocalizations Can Release Oxytocin in Humans." Proceedings of the Royal Society B 277, no. 1694 (2010): 2661–2666. A mother's voice alone raised oxytocin in stressed children at levels comparable to physical contact; the bonding response tracks the acoustic cue. https://doi.org/10.1098/rspb.2010.0567 ↩

37. Voon, Valerie, et al. "Neural Correlates of Sexual Cue Reactivity in Individuals with and without Compulsive Sexual Behaviours." PLOS ONE 9, no. 7 (2014): e102419. Individuals with compulsive sexual behavior showed a wanting-versus-liking dissociation paralleling drug addiction: greater desire for sexual cues without proportionally greater liking. https://doi.org/10.1371/journal.pone.0102419 ↩

38. Banca, Paula, et al. "Novelty, Conditioning and Attentional Bias to Sexual Rewards." Journal of Psychiatric Research 72 (2016): 91–101. Compulsive users formed enhanced conditioned associations to cues paired with sexual content that were difficult to extinguish. https://doi.org/10.1016/j.jpsychires.2015.10.017 ↩