OMXUS Press
CONSTRUCTED GUILT
2026
This thesis began with a phone call.
26,146 words
~104 min read
31 chapters
OMXUS Press
2026
This thesis began with a phone call.
This thesis examines the mechanisms through which guilt is produced by the criminal justice system. Drawing on semiotics, philosophy of language, critical legal theory, cognitive psychology, and empirical criminology, the analysis proceeds across seven institutional sites: pre-interrogation detention, police interrogation, legislative language, courtroom proceedings, media framing, jury processes, and the specific position of neurodivergent populations.
The analysis documents the following findings:
1. Legal language does not describe pre-existing facts but constitutes institutional realities through performative speech acts. The verdict "guilty" is not a finding but a constitution.
1. The behavioural cues that trained investigators and lay observers use to assess credibility operate at chance levels (54.1% accuracy) and are empirically inverted: the behaviours interpreted as indicators of deception are more strongly associated with truthful communication.
1. Post-event linguistic manipulation alters witness memory in approximately 22% of cases (d = 0.72). Testimony is a product of the interaction between memory and the linguistic environment of questioning.
1. False confessions occur in 12–30% of documented exonerations. Pre-interrogation detention elevates suggestibility by 80–120% above baseline. The legal concept of voluntariness does not account for these neurobiological effects.
1. Neurodivergent individuals—including those with autism, FND, PTSD, and CPTSD—present authentically in ways that systematically trigger credibility-reducing inferences. Their innocence is structurally illegible to assessment instruments calibrated to neurotypical baselines.
1. The system produces these outcomes through its ordinary operation. The architecture serves institutional interests in conviction rates. The presumption of innocence operates as doctrine, not as practice.
These findings have implications for the evidentiary weight that should be accorded to investigator credibility assessments, confession evidence, witness testimony elicited through suggestive questioning, and verdicts in cases involving neurodivergent defendants or significant pre-trial publicity.
Keywords: criminal justice, credibility assessment, deception detection, false confessions, memory distortion, interrogation, neurodivergence, autism, functional neurological disorder, Signal Inversion Effect, presumption of innocence, voluntariness doctrine
The criminal justice system does not find guilt. It builds it.
Alex Applebee
L. N. Combe
Independent Research
March 2026
This thesis examines the mechanisms through which guilt is produced by the criminal justice system. Drawing on semiotics, philosophy of language, critical legal theory, cognitive psychology, and empirical criminology, the analysis proceeds across seven institutional sites: pre-interrogation detention, police interrogation, legislative language, courtroom proceedings, media framing, jury processes, and the specific position of neurodivergent populations.
The analysis documents the following findings:
These findings have implications for the evidentiary weight that should be accorded to investigator credibility assessments, confession evidence, witness testimony elicited through suggestive questioning, and verdicts in cases involving neurodivergent defendants or significant pre-trial publicity.
Keywords: criminal justice, credibility assessment, deception detection, false confessions, memory distortion, interrogation, neurodivergence, autism, functional neurological disorder, Signal Inversion Effect, presumption of innocence, voluntariness doctrine
This thesis began with a phone call.
In January 2025, the Department of Communities and Family Services in Western Australia issued a clearance letter regarding a child protection matter. The clearance stated, in effect: no fault, no concern. The matter was closed.
Five months later, the person who had been cleared was arrested. The charge: failure to provide care — for the same incident that had already been cleared. The arrest was made by officers of the Child Abuse Squad. The person was placed in a cell with vomit on the walls. Their books were taken. Their teddy bear was taken. They were called by their legal name, not the name they had used for decades. When they told staff in the cell that their treatment was unethical, they were told: "Everyone says that."
The jail psychologist assessed the person's symptoms of functional neurological disorder — a documented medical condition involving seizures, tremor, and movement difficulties — and noted, in the clinical record, that the symptoms appeared to be fake.
The person was, at the time of their arrest, at a shopping centre helping a victim of domestic violence.
This thesis does not tell that story. It tells the story underneath it: the story of how a system designed to construct guilt can take a cleared person, arrest them, process them, and produce the appearance of criminality from the raw material of innocence.
Every mechanism documented in this thesis — the pre-interrogation degradation, the behavioural credibility heuristics, the Signal Inversion Effect, the neurodivergent credibility gap, the performative voluntariness doctrine — operated in the case described above. The case is not anomalous. It is the system functioning as designed.
The evidence in this thesis is drawn from peer-reviewed empirical research, meta-analyses, government reports, and publicly available datasets. The analysis is the authors' own. The conclusions are uncomfortable. They are also, we believe, correct.
This work is dedicated to every person who has been processed by a system that cannot distinguish their innocence from guilt — and to the ones who couldn't survive the processing.
Alex Applebee
L. N. Combe
March 2026
Part I: Foundations
Part II: The Architecture of Constructed Guilt
Part III: Synthesis
Appendices
| Figure | Description | Source |
|---|---|---|
| 1.1 | The Criminal Justice Pipeline—Seven Stages of Guilt Construction | Original |
| 2.1 | The Signal Inversion Effect—What Observers Believe vs. What Evidence Shows | Original |
| 2.2 | The Behavioral Adaptation Feedback Loop | Original |
| 2.3 | Neuroimaging: Alert vs Stressed Brain States | Arnsten (2015), CC BY |
| 2.4 | Threat Regulatory Neurocircuitry | Fenster et al. (2018), CC BY |
| 2.5 | Healthy vs PTSD Threat Circuits | Fenster et al. (2018), CC BY |
| S.1 | 91% Inversion Rate—Belief vs Reality | Original analysis |
| S.2 | Forest Plot—Believed vs Actual Deception Cues | Original analysis |
| S.3 | Belief-Reality Matrix Scatter Plot | Original analysis |
| S.4 | Linguistic Effect Sizes in Trial Testimony | Original analysis |
| C.1 | Cross-Cultural Variation in Truthful Speech | Original analysis |
| C.2 | Cross-Cultural Classifier False Positive Rates | Original analysis |
| Table | Description |
|---|---|
| 2.1 | The Inversion Pattern: Behaviour, Actual Signal, Perceived Signal |
| 2.2 | Deception Detection Accuracy—Meta-Analytic Summary |
| 2.3 | Memory Distortion Effect Sizes |
| 2.4 | False Confession Rates Across Exoneration Datasets |
| 2.5 | Suggestibility Elevation Under Pre-Interrogation Conditions |
| 2.6 | Neurodivergent Presentation vs System Interpretation |
| 4.1 | The Nine Steps of the Reid Technique |
| 5.1 | Legal Terms as Contested Constructions |
| A.1 | Four-Pillar Convergent Validity Synthesis |
| C.1 | Cultural Variation in Linguistic Features (Kruskal-Wallis) |
As of 2024, Australia imprisons approximately 42,000 people at any given time — a rate of 160 per 100,000 population. Aboriginal and Torres Strait Islander Australians are imprisoned at approximately 2,300 per 100,000 — a rate higher than any identified population on earth, including the Black incarceration rate in the United States (Weatherburn & Holmes, 2017).
In the United States, approximately 2.2 million people are currently incarcerated, with a further 4.5 million under community supervision (parole or probation). The Innocence Project has secured over 375 exonerations through post-conviction DNA testing since 1992. The National Registry of Exonerations documents over 3,300 exonerations since 1989.
These are the people who have been proved innocent — after conviction, after imprisonment, after years or decades of their lives consumed by a system that processed them and produced an output labelled "guilty."
The question this thesis asks is not how many innocent people are in prison. It is whether the system that put them there has any reliable mechanism for distinguishing the innocent from the guilty.
The answer, across four independent methodological pillars and twenty-three behavioural cues examined, is no.
# PART I: FOUNDATIONS
# Chapter 1: Introduction
On any given day in any jurisdiction operating under the common law tradition, a person may be arrested, stripped of their clothing and personal possessions, subjected to an invasive physical search by strangers, confined in a small and deliberately austere cell, and subsequently placed in a closed room from which they are not free to leave—all prior to any finding of guilt, and under the formal constitutional protection of presumptive innocence.
What follows in that closed room will be called an "interview." The person's responses, their silences, their eye contact or lack thereof, their composure or distress, will be interpreted by trained investigators and subsequently by courts, juries, and the public as evidence bearing upon guilt or innocence.
This thesis argues that every element of this sequence is linguistically and institutionally constructed—and that the construction is not incidental but essential.
The argument begins with a deceptively simple proposition: language does not reflect reality. It constructs it.
This is not a novel philosophical claim; it has been foundational to semiotics, philosophy of language, and critical social theory for well over a century (de Saussure, 1916/1983; Wittgenstein, 1953). What has been insufficiently applied to criminology is the full institutional and coercive weight of this claim.
When a person is designated "guilty," "suspicious," "a flight risk," "of prior bad character," or simply "Australian" in a court of law, these are not descriptive statements. They are speech acts that perform classifications, assign subject positions, and activate institutional machinery. The word "Australian" tells us nothing about a person's neurological architecture, their evolutionary history, or their material needs—but it tells us everything about which legal system has jurisdiction over their body.
Language, in law, is not the medium of truth. It is the mechanism of power.
A second proposition follows necessarily: if language constructs guilt, then guilt is, in principle, constructable from any body.
This thesis does not argue that all convicted persons are innocent, nor does it argue that there is no such thing as harmful conduct warranting social response. It argues something more precise and more disturbing: that the evidentiary and procedural architecture of the criminal justice system is structured such that guilty narratives can be assembled around any individual, regardless of what that individual actually did.
The empirical literature on false confessions makes this case with unusual clarity. Kassin and Gudjonsson (2004) documented that between 14% and 25% of exonerated individuals had falsely confessed—a finding that is not anomalous but is rather the predictable output of an interrogation methodology designed to produce confessions rather than truth.
Elizabeth Loftus's decades of research on memory malleability (Loftus, 1979, 2005) demonstrated that the specific linguistic choices of an interrogator or cross-examining attorney can alter not merely a witness's account but the witness's actual memory of events.
The justice system does not merely describe what happened. It authors it.
A third proposition is institutional and structural: the principle of presumptive innocence, while real as legal doctrine, is practically negated at every procedural stage at which it should operate.
This thesis traces that negation through its full sequence:
The argument is not that individual actors within the system are malicious, though some may be. It is that the system's architecture—its physical design, its procedural rules, its linguistic conventions, its institutional incentives—systematically produces guilt-presumptive outcomes while maintaining the rhetorical apparatus of neutrality.
This thesis introduces a fourth proposition, developed empirically across multiple research traditions: the Signal Inversion Effect.
The behaviours that trained investigators and lay observers interpret as indicators of deception—gaze aversion, hedging, fragmented narrative, disfluency, expressions of uncertainty—are, empirically, more strongly associated with truthful communication than with lying.
Conversely, the behaviours interpreted as indicators of honesty—steady eye contact, confident assertion, fluent narrative—are the hallmarks of rehearsed, performed, and strategically deployed speech.
The people who sound most guilty are most likely to be telling the truth.
This inversion is not a marginal effect. Meta-analytic evidence establishes that 91% of the behavioural cues people use to assess credibility are either empirically invalid or directionally inverted (original analysis, this thesis).
A fifth proposition addresses a population that has received insufficient attention in the criminological literature: neurodivergent individuals face compound vulnerability within the criminal justice system.
Autistic individuals, those with functional neurological disorder (FND), ADHD, PTSD, and CPTSD present authentically in ways that the folk psychology of credibility assessment systematically misreads as deception:
The system is not failing neurodivergent people. It is processing them exactly as designed.
This thesis is situated within critical criminology and draws substantially on the philosophy of language, semiotics, and Foucauldian discourse theory. It is empirically grounded in the psychological literature on interrogation, memory, and decision-making, and engages with doctrinal legal analysis where institutional mechanisms require close reading.
The argument has antecedents in:
It departs from some of these traditions in its emphasis on language as the primary site of analysis, while recognising that class, race, and economic structure are inseparable from the linguistic construction of guilt.
This thesis proceeds in ten substantive chapters:
Part I: Foundations
Part II: The Architecture of Constructed Guilt
Part III: Synthesis
Appendices provide the full statistical analysis supporting the empirical claims.
The analysis is primarily contextualised within common law jurisdictions, with particular reference to Australian criminal procedure. Where research from other jurisdictions—particularly the United States and United Kingdom—is drawn upon, the relevance to Australian practice is addressed.
This thesis makes extensive use of terms such as "guilt," "innocence," "suspect," "offender," and "victim." These terms are employed critically throughout—as objects of analysis rather than transparent descriptions. Where these terms are used without qualification, they refer to their institutional or colloquial usage. Where the analysis turns on their constructed character, this is made explicit.
# Chapter 2: Theoretical Framework
The foundational theoretical claim of this thesis—that language constructs rather than reflects reality—derives from the structuralist linguistics of Ferdinand de Saussure.
In the posthumously published Course in General Linguistics, de Saussure (1916/1983) established the sign as a two-part structure comprising the signifier (the sound-image or written mark) and the signified (the concept produced in the mind of the listener or reader). Crucially, de Saussure argued that the relationship between signifier and signified is arbitrary: there is no natural, necessary, or pre-given connection between a word and the concept it evokes.
"Dog" means what it means because a community of speakers agrees, tacitly and historically, that it does—not because the word contains or resembles the animal.
Terms such as "guilty," "reasonable," "intent," "consent," "Australian," and "criminal" do not describe pre-existing conditions in the world. They are classifications produced through linguistic practice and institutionally enforced.
The designation "guilty" does not identify a property of the defendant; it is a verdict—a speech act, in Austin's (1962) terminology, that changes the institutional status of a person in the world.
Before the word is spoken, there is a person.
After it, there is a convict.
The word does not describe a transformation. It performs one.
Roland Barthes (1957/2009) extended Saussure's framework to identify a second order of signification he termed myth. Where first-order signification involves the direct relationship between signifier and signified, second-order signification involves taking an already-complete sign and using it as the signifier in a new system.
This is the mechanism by which ideologically loaded meanings come to appear natural and self-evident.
In the context of criminal justice, "prior criminal record" is a first-order sign denoting documented legal history. At the level of myth, it becomes a second-order signifier for "the kind of person who does these things"—a naturalised narrative that transforms contingent institutional history into essential character.
The defence attorney who cannot suppress the jury's knowledge of a defendant's prior convictions is not fighting against evidence; they are fighting against a mythological structure.
Ludwig Wittgenstein's later philosophy provides a complementary framework for the analysis of legal language.
In Philosophical Investigations, Wittgenstein (1953) introduced the concept of the language game—a set of linguistic practices embedded in a form of life, where meaning is constituted by use rather than by reference to any underlying reality.
"The meaning of a word is its use in the language." (§43)
For Wittgenstein, there is no metalanguage, no view from nowhere, no final description. There are only language games, each with its own grammar, and each game-community defining the rules of its own discourse.
The legal system constitutes a language game in Wittgenstein's sense—one with exceptionally high stakes and coercive enforcement mechanisms.
Legal language is not an attempt to describe the world in ordinary terms; it is a specialised grammar with its own rules for:
The "reasonable person" standard in negligence law, for instance, is not a description of any actual human being but a grammatical device within the legal language game—a standard that judges (themselves participants in the game) calibrate through practice and precedent.
When a jury is asked to determine whether a defendant acted as a "reasonable person" would have acted, they are not being asked to consult reality; they are being asked to participate in a language game whose rules they do not fully understand and have not chosen.
In On Certainty, Wittgenstein (1969) argued that doubt presupposes a background of certainty—that one can only meaningfully question some things by taking other things for granted.
Legal proceedings construct their background certainties through procedural rules, evidentiary standards, and institutional authority. What is treated as self-evident in a courtroom—that the proceedings are neutral, that the law is clear, that the jury understands their instructions—is not self-evident at all.
It is the background against which the foreground drama of guilt or innocence is played out, and it is a background that systematically advantages the prosecution.
Michel Foucault's contribution to this analysis is substantial and pervasive. Two bodies of work are of particular relevance:
For Foucault, discourse is not merely language. It is the ensemble of practices, institutions, rules, and statements that produce particular objects of knowledge and particular kinds of subjects.
Criminal law, in this analysis, is a discursive formation that does not merely describe the criminal but produces the criminal as a category of person.
The criminal is not someone who has committed a prohibited act; the criminal is a particular kind of subject—one with a psychology, a history, a propensity—who exists as an object of knowledge within the discursive field of criminological, legal, psychiatric, and carceral practice.
Foucault's concept of the "delinquent" in Discipline and Punish captures this precisely: the judicial system ostensibly addresses acts, but the penal system addresses persons. The shift from punishing a deed to managing a type of subject is the historical achievement that Foucault traces through the transformation of penal practice.
For Foucault, power does not merely suppress or prohibit; it produces. It produces knowledge, it produces subjects, it produces truth.
The expert witnesses, forensic psychologists, criminologists, police investigators, and legal professionals who participate in the criminal justice process are not neutral truth-tellers applying objective methodologies. They are nodes in a network of power/knowledge that produces the official account of what happened and who is responsible.
The authority of this account derives not from its correspondence to reality but from the institutional positions of those who produce it. A detective's interpretation of a suspect's body language carries evidentiary weight not because detectives have reliable access to truth but because detectives occupy an institutional position that authorises their interpretations.
Discipline and Punish provides a genealogy of the prison that anticipates the argument of this thesis. Foucault argues that the modern penal system did not emerge from humanitarian concern with justice but from the administrative need to manage populations, and that surveillance, normalisation, and the construction of the "dangerous individual" are its primary mechanisms.
The pre-interrogation detention regime described in this thesis is, in this analysis, a technology of normalisation: a process of stripping and repositioning the subject that has its genealogical roots in the disciplinary institutions of the nineteenth century.
The holding cell is not an unfortunate logistical necessity; it is a chamber of subject production.
J.L. Austin's How to Do Things with Words (1962) introduced the concept of the performative utterance—a speech act that does not describe a state of affairs but constitutes one.
"I hereby sentence you to ten years imprisonment"
This is not a description of an event; it is the event.
Performative utterances require institutional conditions—felicity conditions, in Austin's terminology—to succeed: the speaker must occupy an authorised position, the circumstances must be appropriate, and the conventions must be recognised by the relevant community.
Legal language is saturated with performatives. The verdict, the sentence, the charge, the warrant, the caution—each of these is a speech act that transforms the institutional status of a person in the world.
This means that the legal determination of guilt is not a cognitive act of recognition (identifying someone as guilty) but a social act of constitution (making someone guilty).
Once this is understood, the question "was the verdict correct?" becomes philosophically complex. The verdict is not correct or incorrect in the way a factual description might be. It is, like a baptism or a contract, a social fact produced by the performance of a specific speech act under specific institutional conditions.
John Searle's (1969) development of Austin's framework extended the analysis to indirect speech acts—utterances whose illocutionary force differs from their literal content.
The literature on interrogation tactics is replete with indirect speech acts that disguise coercion as inquiry. The felicity conditions for valid confession evidence—that the statement be voluntary, uncoerced, and accurate—are systematically undermined by a discursive architecture built on precisely these indirect speech acts.
This thesis proposes that legal language operates at what might be called a double remove from reality.
The first remove is the general Saussurean point: no word corresponds directly to the thing it purports to describe.
The second remove is specific to legal and institutional language: legal descriptions of events are accounts of accounts—retrospective narrativizations of past states of affairs, produced under institutional pressures and interpreted through cognitive and cultural schemas that are largely invisible to those who deploy them.
This double remove is the epistemological condition within which the justice system operates, yet the system consistently presents its outputs as though they were simple observations of fact.
The empirical psychology of memory provides the most direct evidence for the second remove.
Loftus and Palmer (1974) demonstrated in a seminal experiment that the use of different verbs to describe the same event—"smashed" versus "hit" versus "contacted"—produced significantly different estimates of vehicle speed and, one week later, different reports of whether broken glass had been present (it had not).
The word changed the memory.
Subsequent decades of research have established the misinformation effect as robust and replicable (Loftus, 2005): post-event information, including information embedded in the linguistic formulation of questions, integrates with and alters the memory of the original event.
Legal testimony, produced through questioning—often leading, repetitive, and adversarial questioning—is therefore not a report of memory. It is a product of an interaction between memory and the linguistic environment in which recall occurs.
Cross-examination is explicitly designed to exploit the instability of memory. An effective cross-examination does not simply challenge what a witness says; it reconstructs what the witness believes they saw.
The opposing counsel who elicits the admission "I suppose it could have been red" from a witness who initially reported a blue car has not discovered truth; they have manufactured a new memory that serves the narrative they are constructing.
The jury, who will later deliberate on which story to believe, will not have access to the original event. They will have access to two competing linguistic constructions of it, produced under adversarial institutional conditions, and they will be asked to decide which is true.
The critical legal studies (CLS) movement of the 1970s and 1980s developed a sustained critique of legal formalism—the doctrine that legal decisions are determined by the application of fixed rules to established facts—and argued that law is fundamentally indeterminate: the same legal materials can support contradictory conclusions, and outcomes are shaped by political and ideological factors that legal reasoning conceals (Unger, 1983; Kennedy, 1997).
If legal rules do not determine outcomes, then what does?
The argument of this thesis is that narrative, institutional power, and the linguistic construction of character and credibility do.
Duncan Kennedy's (1997) analysis of the ideological dimensions of legal argument is particularly illuminating. Kennedy argued that legal reasoning is not a technical practice insulated from political commitments but a form of rhetoric that frames contested value choices as though they were determinations of neutral principle.
The reasonable person standard, the balancing test, the doctrine of proportionality—each of these constructs presents what is in fact a politically loaded choice as a factual or logical conclusion.
In the context of criminal procedure, this rhetorical operation is most clearly visible in the treatment of confession evidence. Courts routinely hold that confessions obtained after hours of psychologically coercive interrogation, from a person in an acute stress state following hours of pre-interrogation detention, are "voluntary."
The word "voluntary" is doing substantial ideological work—concealing a coercive reality behind a legal fiction that immunises the system from accountability for its own methods.
The theoretical framework assembled across the preceding sections provides a multi-levelled account of how guilt is constructed within the criminal justice system:
| Theoretical Tradition | Contribution |
|---|---|
| Saussurean semiotics | Words do not report pre-existing facts but constitute new institutional realities |
| Barthes (myth) | Constructions come to appear natural and self-evident, making ideological operations invisible |
| Foucault (discourse) | Reveals the institutional and power-laden conditions within which legal knowledge is produced |
| Foucault (Discipline and Punish) | Locates the contemporary interrogation chamber within a genealogy of disciplinary practice |
| Austin/Searle (speech acts) | Identifies the specific performative mechanisms through which institutional transformation—from suspect to convict—is accomplished linguistically |
| Memory research | Demonstrates that linguistic construction extends backward in time: words rewrite the past events taken as evidence |
| CLS theory | Shows that formal legal reasoning actively enables construction by converting political choices into the appearance of neutral determinations |
The body, no less than the mouth, is a site of legal construction.
The people who sound most guilty are most likely to be telling the truth.
This section documents what this thesis terms the Signal Inversion Effect: the systematic pattern whereby authentic cognitive and linguistic behaviours are misidentified as indicators of deception, while performed or rehearsed behaviours are mistakenly interpreted as signals of honesty.
The core paradox is deliberately counterintuitive—and it is supported by converging evidence from multiple independent research traditions.
| Behaviour | What It Actually Signals | What Observers Believe |
|---|---|---|
| Hedging ("I think," "maybe") | Genuine memory retrieval; honest uncertainty | Evasion; hiding something |
| Breaking eye contact | Cognitive effort; accessing memory | Shifty; untrustworthy |
| Fragmented narrative | Authentic trauma recall; real memory is messy | Incoherent; fabricating |
| Saying "I don't know" | Accurate self-knowledge; genuine limits of memory | Incompetent; evasive |
| Confident, fluent delivery | Rehearsed narrative; possible deception | Honest; credible |
| Impersonal pronouns ("it," "that") | Distancing from false narrative (false confessions) | Not typically noticed |
| Spontaneous corrections | Genuine memory retrieval; commitment to accuracy | Inconsistency; changing story |
| Reported confusion | Deep engagement; authentic processing | Failure to understand |
The foundational claim of police interrogation methodology is that trained investigators can identify deception through behavioural observation. If this claim is true, then investigator assessments of guilt carry epistemic weight—they are, at least partially, observations of reality. If the claim is false, investigator assessments are no more reliable than chance, and the guilty narrative they produce is constructed rather than observed.
The empirical literature on deception detection constitutes one of the most thoroughly replicated bodies of research in applied psychology. Its findings are consistent, well-powered, and directly devastating to the foundational claim of interrogation methodology.
| Study | N (Judges) | Accuracy % | 95% CI | Population |
|---|---|---|---|---|
| Ekman & O'Sullivan (1991) | 509 | 52.8 | [49.1, 56.5] | Law enforcement, judges, psychiatrists |
| Vrij & Graham (1997) | 156 | 53.2 | [48.0, 58.4] | Police officers, UK |
| Meissner & Kassin (2002) | 4,435 | 54.0 | [52.5, 55.5] | Meta-analysis: civilians & investigators |
| Bond & DePaulo (2006) | 24,483 | 54.3 | [53.7, 54.9] | Meta-analysis: 247 studies |
| Vrij (2008) — overall | ~5,000 | 54.0 | [53.0, 55.0] | Mixed professional/civilian |
| Hartwig & Bond (2011) | ~3,000 | 53.9 | [52.1, 55.7] | Law enforcement |
| CHANCE BASELINE | — | 50.0 | [50.0, 50.0] | Theoretical maximum with zero information |
| WEIGHTED MEAN | ~37,500 | 54.1 | [53.6, 54.6] | Advantage over chance: 4.1 percentage points |
The weighted mean accuracy of 54.1% across approximately 37,500 judgements represents a statistically significant deviation from chance (p "Exposure to uncontrollable stress rapidly evokes chemical changes in brain that impair the higher cognitive functions of the PFC while strengthening primitive brain reactions. This flip from reflective to reflexive brain state may have survival value when we are in danger, but it can be ruinous."
Application: A person who has been arrested, stripped, searched, and confined in isolation presents with measurably reduced prefrontal cortex function. The neural substrate of rational decision-making has been chemically and functionally degraded.
The legal doctrine of "voluntariness" assumes a brain that no longer exists in the detained person.
[Figure 2.4 — Threat Regulatory Neurocircuitry. Fenster et al. (2018), Neuropsychopharmacology, PMC8617299, CC BY.]
Source: Fenster et al. (2018), Neuropsychopharmacology. PMC8617299. CC BY.
Caption: Human brain anatomy highlighting regions involved in threat learning, extinction, avoidance, cognitive regulation, and contextual modulation. Key structures: vmPFC (ventromedial prefrontal cortex), dlPFC (dorsolateral prefrontal cortex), amygdala, hippocampus, dACC (dorsal anterior cingulate cortex).
Application: These are the neural structures that produce autistic social processing differences, PTSD-related threat dysregulation, and the credibility judgments that observers make. The differences are neurological, not behavioral choices.
[Figure 2.5 — Healthy vs PTSD Threat Circuits. Fenster et al. (2018), Neuropsychopharmacology, PMC8617299, CC BY.]
Source: Fenster et al. (2018), Neuropsychopharmacology. PMC8617299. CC BY.
Caption: Panel (A) shows healthy threat circuitry with intact connectivity between dlPFC, vmPFC, hippocampus, and amygdala. Panel (B) shows PTSD-related threat circuitry where the dlPFC, vmPFC/IL, and hippocampus show impaired functioning with PTSD, whereas the amygdala and dACC/PL are enhanced.
Application: PTSD produces measurable changes in the brain circuits responsible for threat processing and memory. The fragmented recall, emotional dysregulation, and avoidant presentation characteristic of trauma are neurological symptoms, not credibility indicators.
Observers who interpret these presentations as deception are misreading brain states.
Despite two decades of research investment and the most sophisticated brain imaging technology available, fMRI-based lie detection does not work reliably enough for forensic use.
| Study / Review | Accuracy | Critical Finding |
|---|---|---|
| Meta-analysis (Nature Reviews Neuroscience) | 75% | Best case in controlled laboratory conditions |
| UC Berkeley 2024 (PNAS) | 79% | Confounded by selfishness — neural signatures of deception identical to self-interest |
| Mock crime paradigms | 69% sensitivity | Low specificity — high false positive rates |
| Medial PFC region analysis | 71% | Best single region; no region worked across all individuals |
| Applied Cognitive Psychology 2026 | — | "Not suited for use as a lie detector" |
The 2024 UC Berkeley study (Wills Neuroscience Institute / Haas) identified the fundamental problem:
"One reason it's so hard to isolate signals of deception is that lying is a complex process that isn't housed in a single part of the brain, and it's challenging to separate activity linked to lying from that reflecting anxiety, self-interest, or other factors."
The brain states associated with deception overlap extensively with the brain states associated with:
If fMRI—which directly images brain activity—cannot reliably distinguish deception from innocence under stress, then behavioural heuristics—which attempt to infer internal states from external presentation—are necessarily less reliable still.
| Method | What It Measures | Accuracy | Forensic Validity |
|---|---|---|---|
| fMRI | Direct brain activity | 69-79% | Not reliable |
| Polygraph | Physiological arousal | 65-70% | Not reliable |
| Trained investigators | Behavioural observation | 54% | Chance level |
| Untrained observers | Behavioural intuition | 54% | Chance level |
The most sophisticated neuroimaging cannot reliably detect deception. The behavioural heuristics used by the criminal justice system perform at chance level. There is no valid method currently employed.
The neuroimaging research confirms the thesis argument: there is no clean neural signal of deception that can be separated from the neural signatures of innocence under stress.
An innocent person who is:
will produce brain activation patterns indistinguishable from a guilty person who is lying.
The neural overlap between deception and innocent-under-stress is not a limitation that better technology will resolve. It reflects the fundamental reality that the cognitive processes involved in lying are also involved in being falsely accused, being stressed, being anxious, and being neurodivergent.
If the brain itself cannot be reliably read for deception, the claim that external behaviour can be read for deception is necessarily false.
Sources: UC Berkeley 2024, Applied Cognitive Psychology 2026, Nature Reviews Neuroscience, PMC Review 2024
# PART II: THE ARCHITECTURE OF CONSTRUCTED GUILT
# Chapter 3: The Body Before the Interview
There is a sequence of events that precedes every police interrogation and that the literature on interrogation almost entirely ignores.
A person is stopped, typically without warning. Hands are placed on their body. They are told they are under arrest. They may be forced to the ground. Their arms are restrained behind them. They are moved—into a vehicle, into a building—without being asked. They are taken to a room and told to remove their clothing. Strangers examine their body, sometimes roughly. Their personal possessions—wallet, phone, keys, watch, the material coordinates of identity and social connection—are taken and bagged. They are given a paper or cloth garment, or nothing at all, and placed in a small room. The room is bare. It is typically painted in a colour specifically chosen to minimise stimulation. It smells of institutional cleaning agents and, often, of other people's fear. The door is locked. Time passes.
Only after some portion of this sequence is complete does the "interview" begin.
This chapter argues that the pre-interview sequence is not preamble. It is the first stage of the guilt-construction process, and its effects on the subsequent linguistic exchange are so profound that the very concept of a voluntary statement, as deployed by Australian and other common law courts, is rendered empirically incoherent.
The presumption of innocence is a foundational principle of common law criminal procedure. In Australia, it operates as:
Its doctrinal content is clear: a person is to be treated as innocent until guilt is established by a court. The corollary is equally clear: prior to such establishment, the state is not entitled to administer punishment.
The pre-interrogation detention regime described above is, by any substantive account, punishment.
It is experienced as punishment. It produces in the detained person the physiological and psychological states that punishment is designed to produce—fear, subordination, disorientation, and the acute awareness of institutional power over the body.
That it is not called punishment, that it is classified as "administrative procedure" or "custody management," is itself a significant linguistic operation—one that will be returned to in Chapter 5 in the context of legislative language and legal fiction.
For present purposes, the point is structural: the justice system inflicts an experience of punitive treatment on persons who are, by the system's own formal declaration, innocent.
This contradiction is not an oversight. It is not a gap between legal ideal and operational reality that could be closed by better regulation. It is built into the architecture. The conditions of pre-interrogation detention are maintained because they are functional—they serve the goal of producing statements.
The physiological response to arrest—to the sudden, forceful, and involuntary seizure of the body by strangers—is a textbook activation of the hypothalamic-pituitary-adrenal (HPA) axis. Cortisol and adrenaline are released. Heart rate and blood pressure elevate. Glucose is mobilised.
These are adaptive responses to acute threat, mediated by the amygdala and coordinated across the autonomic nervous system (McEwen, 2007).
The prefrontal cortex—the region most centrally involved in executive function, working memory, rational deliberation, impulse control, and the inhibition of automatic responses—is significantly impaired by acute stress (Arnsten, 2009).
This is not a subtle effect. Arnsten's (2009) review documented that even moderate stress levels produce measurable degradation in prefrontal cortical function, with consequent impairment of the cognitive capacities that underpin rational decision-making.
The detained person who is placed in an interrogation room following the arrest and detention sequence is, neurobiologically, not the same person who existed before the sequence began.
Their capacity to:
has been materially diminished.
To these acute effects must be added the consequences of prolonged confinement:
The person who enters the interview room has been, through the pre-interview sequence, rendered more suggestible, more compliant, more desperate for social connection, and less capable of rational resistance than any baseline measure of their cognitive capacities would suggest.
Erving Goffman's (1961) analysis of total institutions—prisons, asylums, military barracks, convents—identified a characteristic process he termed the "mortification of self": a systematic dismantling of the identity resources through which individuals maintain a sense of continuous, autonomous selfhood in the social world.
The mortification process operates through specific institutional techniques:
These techniques are not incidental to the institution's operations; they are constitutive of them. They produce a subject whose prior identity has been suspended and who is thus available for institutional redefinition.
The pre-interrogation detention sequence maps precisely onto Goffman's mortification process.
The confiscation of personal possessions removes the material anchors of social identity—the phone through which relationships are maintained, the wallet containing the cards and documents that attest to who one is in the institutional world, the watch that marks one's place in the shared temporal framework of social life.
The removal and search of clothing submits the body—the most intimate site of self—to institutional examination.
The assignment of a cell number, a booking reference, a case file reduces the person to an administrative object.
By the time the interview begins, the person has been systematically repositioned, through physical and institutional processes, from a citizen to a suspect—from a subject of rights to an object of inquiry.
The physical environment of the holding cell is not accidental. Institutional design research has established that environmental features systematically affect the psychological states of those who inhabit them (Evans, 2003).
The characteristic features of police holding cells—limited space, minimal natural light, plain or specifically coloured walls, hard surfaces, the absence of any material that would allow productive activity—are, from an environmental psychology perspective, a collection of stressors.
The colours used in detention facilities—the grey-beige-green spectrum characteristic of institutional interiors—is not the result of aesthetic indifference. Institutional colour choice in carceral environments has historically been informed by the goal of minimising stimulation and maintaining order, with the consequence of producing environments experienced by occupants as oppressive, disorienting, and dehumanising (Kwallek et al., 1996).
The cell is a designed environment, and its design serves the institution's goals.
From a semiotic perspective, the cell communicates something specific to its occupant:
The cell, prior to any question being asked, has already begun the work of positioning the person as a certain kind of subject: guilty, contained, available for interrogation.
Australian courts apply the common law principle that a confession is admissible only if it was made voluntarily—that is, without threats, inducements, or oppressive conduct that overbore the will of the accused (R v Lee, 1950; Uniform Evidence Acts, s 84).
These doctrinal protections rest on a model of the person as a rational agent capable of making free choices about whether to speak, and capable of accurately reporting facts when they do.
The empirical literature reviewed in this chapter establishes that this model is false for the person who has been through the pre-interrogation sequence.
The model assumes a baseline of cognitive capacity, emotional regulation, and autonomous agency that the detention process is specifically designed to degrade.
Courts routinely admit confessions obtained in these conditions as "voluntary" because:
The coercive architecture of the hours preceding the interview is treated as legally irrelevant—a prior administrative matter, not part of the interrogation.
This doctrinal gap is not a failure of judicial attention. It is the predictable consequence of a legal framework that defines voluntariness in formal rather than substantive terms.
A formally voluntary statement—one made without explicit compulsion—can be, in the substantive sense relevant to the purposes of the voluntariness doctrine, profoundly involuntary.
The person who confesses to a crime they did not commit after eight hours in a holding cell, in an acute stress state, with impaired prefrontal function, heightened suggestibility, and a desperate need for the social approval of the only human beings in their immediate environment, has made a "free choice" only in the most impoverished sense of that phrase.
The law's insistence that this constitutes voluntary statement is itself a speech act—one that constitutes a legal reality that diverges fundamentally from the experiential and neurobiological reality of the person concerned.
This chapter has argued that the pre-interrogation detention sequence constitutes a first stage of guilt construction that has been systematically neglected in both the legal doctrine of voluntariness and the scholarly literature on interrogation.
The sequence—arrest, bodily search and exposure, confiscation of identity materials, confinement in a designed environment of minimal resource and maximal institutional control—produces, through neurobiological, psychological, and semiotic mechanisms, a subject who is measurably less capable of free and rational communication than they were prior to detention.
The legal concept of voluntariness, which governs the admissibility of the statements subsequently obtained, does not account for these effects.
The conclusion is not that reforms to the voluntariness doctrine would solve the problem—though such reforms would be an improvement—but that the conditions of pre-interrogation detention are structurally functional: they serve the system's interest in producing statements, and their preservation reflects that interest.
# Chapter 4: The Interview Room
You are seated in a small room. Across from you sits a trained investigator. Between you is a table. The door, behind you and to one side, is closed. You have been told you are free to leave, but this statement—required by law to preserve the legal fiction of voluntariness—has been delivered in a context that makes its sincerity implausible.
You were brought to this room from a cell. You do not have your phone. You do not have your wallet. You are wearing your own clothes, or paper clothes, depending on jurisdiction and procedure. You have not eaten recently, or slept well, or spoken to a lawyer for as long as the system could lawfully delay that access.
The investigator leans forward. They are friendly. They understand. They are just trying to get your side of the story.
This chapter examines what happens next.
The focus is on the dominant interrogation methodology in use across common law jurisdictions: the Reid Technique, developed by John E. Reid and Fred Inbau in the mid-twentieth century and codified in the manual Criminal Interrogation and Confessions (Inbau et al., 2013).
The Reid Technique is not merely an interrogation method. It is a system for the linguistic and behavioural construction of guilt.
The Technique proceeds in two formal stages:
The structure itself encodes a presumption: the BAI assesses whether the person is lying; the interrogation is then used to extract a confession from those assessed as deceptive.
The possibility that the BAI assessment is wrong—that an innocent person has been assessed as deceptive—is not procedurally accommodated. There is no step in the Technique designed to consider this possibility; the entire structure points forward to confession.
The BAI's assessment methodology has been subjected to substantial empirical scrutiny, and the findings are consistently damaging.
The Technique's claim that trained investigators can reliably detect deception through behavioural observation—noting gaze aversion, postural shifts, self-grooming behaviours, verbal hedging, and similar cues—is not supported by the research evidence.
A meta-analysis by Vrij (2008) found that trained investigators perform at rates only marginally above chance, with a mean accuracy rate of approximately 54% against a chance baseline of 50%.
Crucially, investigators tend to be more confident in their assessments than their accuracy warrants—a pattern that Kassin et al. (2005) termed "confidence without accuracy."
A technique that produces high confidence and low accuracy is, in the context of an interrogation that will proceed to extraction of confession based on that assessment, precisely the architecture required to generate false confessions from innocent people.
The behaviours that the Reid Technique treats as indicators of deception—gaze aversion, postural shifting, delayed responses, verbal qualifications—are also among the most reliably produced responses to acute stress in innocent people (Vrij et al., 2006).
An innocent person, suddenly accused of a serious offence in a formal institutional setting by a trained investigator after hours of pre-interrogation detention, will typically be frightened.
Frightened people:
The Reid-trained investigator observes these behaviours and records: deceptive.
The innocent person's innocence is being used as evidence against them.
The interrogation begins with the investigator directly telling the suspect that the evidence conclusively establishes their guilt. This is a direct statement, made with certainty and authority, regardless of the actual state of the evidence.
From a speech act perspective, this is not a description of evidentiary reality; it is a performative—an attempt to constitute, through confident assertion, a social reality in which the suspect's guilt is already established.
The innocent suspect who responds with confusion, distress, or vigorous denial is, within the Technique's interpretive framework, demonstrating resistance rather than innocence.
The investigator presents a morally minimising narrative—a "theme"—that offers the suspect a face-saving account of why they committed the act.
Themes are typically constructed to shift moral responsibility: the offence was understandable given the circumstances; it was a momentary lapse; the victim was partly responsible; anyone might have done the same thing.
The function of theme development is to reduce the psychological cost of confession. Kassin and McNall (1991) demonstrated that minimisation significantly increases confession rates—including, critically, among innocent suspects.
The mechanism is straightforward: the suspect is offered a choice between a bearable story (the theme) and an unbearable situation (continued interrogation in conditions of pre-punitive detention). Many innocent people choose the bearable story.
The Technique instructs investigators to actively prevent the suspect from completing their denials—to interrupt, to redirect, to maintain forward momentum toward confession.
This is presented in the training literature as preventing the suspect from "reinforcing" their denials. From a conversational and linguistic perspective, it is the systematic suppression of the suspect's counter-narrative.
The suspect who attempts to tell their own story—to construct their own account of where they were, what they did, and why the investigator's account is wrong—is procedurally blocked from doing so.
The conversation is a monologue with interruptions, not a dialogue. The investigator's narrative is the only one with structural permission to develop.
The implications for an innocent suspect are severe. The innocent person has, by definition, an alternative account of events. The Reid Technique's procedural suppression of their denials prevents them from presenting this account with the coherence and completeness it requires to be persuasive.
The alternative question is among the most analytically significant elements of the Technique.
The investigator presents the suspect with a choice between two versions of how the act occurred—one more morally serious, one less—and invites the suspect to choose:
"Did you plan this out, or did it just happen in the heat of the moment?"
"Was this about money, or was it something personal?"
Both options presuppose the act.
The question is framed as an inquiry into motivation and circumstance, but its deep structure is a presuppositional trap: answering either alternative constitutes an admission of the act itself.
Austin's (1962) analysis of presupposition is precisely applicable here. The alternative question embeds guilt as a presupposition and then offers a choice that cannot be taken without accepting that presupposition.
Courts have held that explicit promises of leniency render confessions involuntary; the alternative question achieves the same psychological effect through implication rather than explicit statement, and thus typically survives voluntariness analysis (Leo, 2008).
The linguistic indirection of the alternative question is not an incidental feature of its design. It is the mechanism that allows the Technique to produce the psychological effects of explicit coercion while maintaining the legal appearance of voluntary interaction.
Once an admission has been obtained—typically through the alternative question—steps eight and nine involve expanding the oral admission into a detailed narrative and then reducing that narrative to a written statement.
This sequence introduces a further layer of linguistic construction. The investigator who assists in elaborating the confession narrative, or who drafts the written statement, introduces their own language, their own emphases, their own causal and temporal frameworks.
Research on contamination of confession evidence has documented numerous cases in which written confessions contain details that the confessor could not have known—details that could only have been introduced by investigators who possessed information about the crime that had not been disclosed to the suspect (Leo & Ofshe, 1998; Kassin et al., 2010).
The written confession, presented in court as the suspect's own account, is often substantially the investigator's account—a narrative constructed by institutional power and attributed to the person whose guilt it is then used to prove.
The empirical literature on false confessions provides the most direct evidence for the thesis that guilty narratives can be constructed from innocent behaviour.
Gross et al. (2005), in an analysis of 340 exonerations in the United States, found that false confessions were documented in approximately 15% of cases.
Brandon Garrett's (2011) analysis of DNA exonerations found that 27 of the first 250 exonerees—all of whom were factually innocent, established by post-conviction DNA testing—had falsely confessed.
These are not people who confessed under duress and were later shown to have actually committed the offence. They are people who confessed to crimes they demonstrably did not commit.
Saul Kassin's (2017) review identified three categories of false confession:
The internalised false confession is the most disturbing category theoretically: it represents not merely a person who lied to end the interrogation, but a person whose memory of events was rewritten by the interrogation process itself.
The thesis advanced in Chapter 1—that the architecture of the interrogation room is structured such that any behaviour can be narrated as evidence of guilt—is now demonstrable in its full specificity:
Australian jurisdictions, including the Australian Federal Police and most state forces, have formally adopted PEACE-aligned frameworks in their investigative interviewing guidelines.
The persistence of Reid-influenced practices at the operational level, despite formal policy alignment with PEACE principles, is itself a significant institutional phenomenon—one that speaks to the depth of the system's interest in confession production over truth production.
The neurodivergent individual in the interrogation room faces a compound vulnerability:
Autistic individuals may:
Individuals with FND may:
Individuals with PTSD/CPTSD may:
The Reid Technique's interpretive framework contains no mechanism for distinguishing neurodivergent presentation from deceptive presentation. The same behaviours are coded identically regardless of their actual cause.
This chapter has examined the Reid Technique as a system for the linguistic and behavioural construction of guilt in the interrogation room.
The Technique's architecture—presuming guilt, suppressing denial, embedding guilt as presupposition, reducing the psychological cost of false confession, and exploiting the suggestibility produced by pre-interrogation detention—is structured to produce confessions from the persons subjected to it, regardless of their actual culpability.
The empirical false confession literature documents the output of this system. The comparison with the PEACE model establishes that the Technique's guilt-production outcomes are not inevitable features of interrogation but choices embodied in a specific institutional architecture.
# Chapter 5: Legislative Language and Legal Fiction
Before there can be a guilty person, there must be a crime. And before there can be a crime, there must be a word. The sequence is rarely stated this plainly in criminal law scholarship, because criminal law scholarship tends to operate within the assumption that statutes describe pre-existing wrongs rather than create new ones. This chapter argues the reverse: that criminal legislation is a performative rather than descriptive enterprise, that the categories it establishes are not reflections of natural moral facts but political constructions with specific historical and ideological genealogies, and that the key terms on which guilt or innocence turns — ’intent,’ ’consent,’ ’reasonable,’ ’recklessness,’ ’dishonesty’ — are semantically unstable in ways the law persistently conceals. The person convicted of an offence defined by these terms has not been found to have done a thing that was always and obviously wrong. They have been found to have satisfied the conditions of a definitional network constructed by legislators, interpreted by judges, and applied by juries who received the key terms with minimal instruction and without any of the philosophical context that would reveal how much work those terms are quietly doing.
When a parliament enacts a statute criminalising an act, it does not discover that the act was already criminal. It makes it criminal. This is Austin’s (1962) performative utterance at the level of sovereign institutional power: the legislation is the speech act that constitutes the offence. Prior to the statute, the act existed in the world. After the statute, the act is a crime. The word — the legislative text — is the mechanism of transformation. This is so obvious at the level of legislative creation that it tends to escape notice; but its implications extend into every subsequent stage of criminal proceedings. If criminal categories are created rather than discovered, then they are contingent. Cannabis possession has been criminal and non-criminal within the same jurisdictions across the span of a few decades. Marital rape was legally impossible in Australia until the late twentieth century — not because it did not occur, but because the legal category that would have made it criminal did not exist. Homosexual acts between consenting adults were criminal offences in all Australian states until the 1970s and 1980s. The content of the criminal law is not a map of natural harms. It is a historically specific construction, produced by specific institutional actors with specific interests, reflecting specific political settlements — and those political settlements are themselves the product of power relations that are, as Foucault (1977) argued, neither neutral nor stable. The constructed character of criminal categories is most visible at the margins, where definitional contests occur explicitly. But the same construction is present at the centre, in the definitions of offences so familiar that their contingency has become invisible. ’Murder’ appears to name a natural moral category — the wrongful killing of a human being — but the legal definition immediately introduces contestable constructions: ’unlawful’ killing (as opposed to lawful killings in warfare, policing, or capital punishment), ’with intent to kill or cause grievous bodily harm’ (introducing a mental state element that requires inference from observable behaviour), and ’of a human being’ (a category whose edges, at the beginning and end of life, are legally contested). The word ’murder’ appears to describe a thing that obviously happened. It actually denotes the satisfaction of a complex definitional matrix, each element of which is a legal construction.
Among the most consequential legal fictions in the common law criminal system is the figure of the ’reasonable person.’ The reasonable person appears throughout criminal law: as the standard against which provocation is assessed, as the benchmark for self-defence, as the reference point for negligence and recklessness, and as the implicit model against which a defendant’s conduct is evaluated in innumerable contexts. The reasonable person is presented as an objective standard — a neutral reference point that transcends the particular perspectives of the parties before the court. The reasonable person has never existed. They are a normative construct — a legal fiction whose function is to translate contested value judgments into the apparently neutral language of objective fact. When a court asks whether a ’reasonable person’ would have been provoked to kill in the circumstances before it, or whether a ’reasonable person’ would have perceived a threat justifying self-defensive force, it is asking: does this defendant’s response fall within the range of responses that this legal community is prepared to treat as excusable or justified? The reasonable person is the vehicle through which that collective normative judgment is expressed as though it were a factual observation. The political content of this fictional figure has been extensively documented in feminist legal scholarship. Schneider (1992) argued that the reasonable person standard historically encoded a specifically male response to perceived threat, systematically disadvantaging women who had used defensive force against intimate partners whose violence was cumulative and contextual rather than immediate and acute. The battered woman who kills a sleeping abuser does not act as a ’reasonable person’ would act in the moment of attack — because the attack, for her, is not happening in the moment. It has been happening for years. The standard’s failure to accommodate this reality was not a neutral technical limitation; it was the reflection of a normative framework constructed from a specifically male experiential baseline (Stubbs & Tolmie, 1999). Australian jurisdictions have partially responded to this critique through legislative reforms to self-defence provisions (Model Criminal Code, s 10.4; Crimes Act 1900 (NSW), s 418), but the reasonable person remains operative as a standard in numerous contexts, carrying its historically loaded normative content into each application. The reasonable person also encodes racial and class assumptions that have been documented with particular acuity in the critical race legal scholarship. The question of whether conduct was ’reasonable’ is answered by a judicial system and jury pool that are, in Australian terms, disproportionately white and middle-class. Research on implicit bias in legal decision-making (Rachlinski et al., 2009; Richardson & Goff, 2012) has established that the application of apparently neutral standards is systematically affected by the racial identity of defendants and victims. The reasonable person is not a universal figure. They are a culturally specific figure whose particular rationality is validated by the institutional structures that deploy them.
Criminal liability in most common law jurisdictions requires proof of a mental element — the mens rea — alongside proof of the prohibited act. The mental element varies by offence: murder requires intent or knowledge; manslaughter may be established by recklessness; some offences require only negligence. This requirement of mental state is presented as a fundamental safeguard — the principle that people should be punished for what they meant to do, not merely for what happened as a result of their actions. It is, in this framing, the law’s recognition of the moral distinction between the deliberate wrongdoer and the unfortunate accident. The safeguard is, however, largely nominal. Mental states are not observable. Intent, knowledge, recklessness, and negligence are all inferences drawn from observable behaviour and circumstance. The jury that is asked to find whether a defendant intended to kill is not being asked to report an observation. They are being asked to make an inference — to take the observable facts (the act, the circumstances, the defendant’s statements and conduct before and after) and to construct a narrative about what was happening inside the defendant’s head. This inference is not a neutral cognitive operation. It is shaped by the same cognitive schemas, cultural assumptions, and narrative frameworks that govern all human judgment, and it is performed by people who have already heard the prosecution’s account of events and been subjected to the full rhetorical architecture of courtroom presentation. Research on attribution theory and criminal judgment has consistently demonstrated that people apply different causal and intentional explanations to the same behaviour depending on the social identity of the actor (Graham & Lowery, 2004; Goff et al., 2014). The behaviour that is read as impulsive and reckless when attributed to one social group is read as calculated and intentional when attributed to another. The mental element requirement, far from providing a safeguard against the arbitrary attribution of guilt, provides a site at which social biases are institutionally reproduced through the language of objective inquiry. The jury is not asked ’did this person intend to kill?’ — a question about the actual contents of another human mind. They are asked a question they cannot answer, and they answer it using whatever cognitive resources they have available — which include their cultural schemas for who is the kind of person who does this kind of thing.
Few legal terms carry as much contested political weight as ’consent,’ particularly in the context of sexual offences. The legal definition of consent — and the evidentiary rules governing how it is established or negated — has been the site of sustained feminist legal reform advocacy across common law jurisdictions over the past five decades. In Australia, the Crimes Act 1900 (NSW) now defines consent as free and voluntary agreement (s 61HE), and affirmative consent models have been adopted in several jurisdictions. These reforms represent genuine improvements. They also illustrate, with unusual clarity, the constructed character of legal definitions: the meaning of ’consent’ for the purposes of criminal law is whatever the legislature decides it is, and that decision is a political one. The point is not that consent is meaningless or that its presence or absence is always contested. The point is that ’consent’ in criminal law is a legal term of art — a definition that operates within a specific institutional context, interpreted according to specific evidentiary rules, applied by decision-makers with specific cognitive and cultural frameworks — and that the gap between the legal concept and the lived experience it purports to address is substantial. Research on juror decision-making in sexual assault cases has documented persistent myth-consistent reasoning — the application of false beliefs about how ’genuine’ victims behave, how perpetrators present, and what circumstances are consistent with non-consent — that operates in the gap between the legal definition and jurors’ interpretive frameworks (Larcombe, 2002; Temkin & Krah?, 2008). The legal definition of consent, however well-drafted, is applied by human beings whose understanding of what consent looks like in practice is shaped by cultural narratives that the law cannot simply override through definitional revision.
The theoretical framework established in Chapter 2 introduced the observation that identity categories in law are fictions — not in the pejorative sense of falsehoods, but in the technical sense of constructed categories that produce real effects while describing nothing in the natural world. The category ’Australian’ — as a legal designation — illustrates this with particular clarity. From a zoological perspective, a person described as ’Australian’ is a member of the species Homo sapiens, currently residing on the landmass that was named ’Australia’ by European colonisers in the nineteenth century. Nothing in the biological, geological, or cosmological description of this person and their location requires the category ’Australian.’ The category is a legal and political construction — one that was assembled through colonial dispossession, federation, immigration legislation, citizenship law, and passport administration. It has no natural referent. Yet the legal consequences of being ’Australian’ versus ’not Australian’ are profound: they determine which criminal jurisdiction has authority over one’s body, which rights one can assert, which protections one can claim, and which obligations one bears. The construction does real, material work — coercive work — in the world. Criminal law operates entirely within this constructed landscape. Every category deployed in criminal proceedings — ’defendant,’ ’victim,’ ’witness,’ ’offender,’ ’juvenile,’ ’recidivist’ — is a legal fiction of the same type. Each category carries a definitional network that activates specific institutional procedures, rights, obligations, and presumptions. The person who is classified as a ’juvenile offender’ is not merely described by this term; they are constituted as a particular kind of legal subject who will be processed through a different institutional system, subject to different dispositions, and described in different records than the person classified as an ’adult offender.’ The word does not follow the person. The word precedes them, and the person is fitted to it.
When a person is charged with a criminal offence, the charge document is not merely an administrative record. It is a narrative — one that constructs a specific account of events, attributes specific intentions and actions to a specific person, and frames that account within the definitional structure of a specific statutory provision. The words of the charge are not neutral. They are selected from a range of possible characterisations, each of which would produce different legal consequences, and the selection is made by a prosecution service that has an institutional interest in conviction. The prosecutorial discretion to charge — to choose among available offences, to decide whether to proceed, to select which counts to include and which to leave out — is a site of substantial constructed power that receives insufficient critical attention. The person charged with aggravated assault rather than common assault, with murder rather than manslaughter, with supply rather than possession, has had a narrative imposed on their conduct that is not uniquely determined by the facts. It is one of several possible narrativisations, chosen by an institutional actor, and that choice will shape everything that follows: the available defences, the maximum penalty, the stigma, and the institutional pathway through which the case proceeds. The charge is the first act of courtroom narrative construction — and it takes place before the court convenes.
This chapter has argued that criminal legislation does not describe a pre-existing moral landscape but constructs the categories through which guilt becomes possible. The performative character of legislative language — the fact that statutes create crimes rather than discovering them — means that every criminal category is contingent, politically constructed, and historically specific. The key terms through which guilt is determined — 'intent,' 'consent,' 'reasonable,' 'recklessness' — are semantically unstable in ways that the legal system persistently conceals. The reasonable person standard encodes a normatively loaded fiction as an objective reference point. Mental state requirements demand inferences that are shaped by the same cognitive biases documented in the preceding chapters. Consent definitions operate in a gap between legal language and juror interpretation that legislative drafting cannot close. Identity categories construct the legal subject before the legal subject enters the courtroom. And the charge document — the prosecution's first narrative act — frames the entire proceeding before a word of evidence is heard.
# Chapter 6: The Courtroom as Construction Site
The courtroom is, in the popular imagination, the site where truth is established. Two competing accounts are presented, evidence is tested, witnesses are examined, and a neutral decision-maker — judge or jury — determines which account is true. This model of the adversarial trial as truth-finding mechanism is foundational to common law legal culture; it is the premise of the presumption of innocence, the basis of the right to cross-examine, and the legitimating narrative of the verdict. This chapter argues that the model is false — not in the sense that trials always produce incorrect verdicts, but in the deeper sense that the adversarial courtroom is not designed as a truth-finding mechanism and does not function as one. It is designed as a narrative contest, and what it produces is not truth but the official version: the account that has been most successfully constructed within the specific rhetorical and institutional constraints of the trial process.
The adversarial trial proceeds on the assumption that truth is best approximated through the clash of competing partisan accounts, each subjected to rigorous challenge by the opposing party, with a neutral decision-maker determining the outcome. This is a coherent epistemological hypothesis. It is also one for which there is remarkably little empirical support. Research comparing adversarial and inquisitorial trial systems — the two dominant models across common law and civil law jurisdictions respectively — has not produced consistent evidence that adversarial systems generate more accurate verdicts (Damaska, 1997; Findley & Scott, 2006). What the adversarial system does demonstrably produce is a particular kind of narrative contest in which rhetorical skill, the strategic management of evidence, and the exploitation of juror psychology are systematically rewarded. Walter Fisher’s (1984) narrative paradigm provides a useful analytical framework here. Fisher argued that human beings are fundamentally narrative creatures — that we evaluate claims not primarily through logic and evidence but through narrative rationality: the assessment of whether a story hangs together coherently (narrative probability) and whether it resonates with our experience of how the world works (narrative fidelity). In Fisher’s framework, the most compelling argument is not the most logically sound but the most narratively coherent and resonant. The adversarial trial, in this analysis, is not a logical contest but a narrative competition — and it will be won by the party whose story best satisfies the jurors’ narrative rationality, regardless of its correspondence to what actually happened. This analysis has direct implications for the construction of guilt. The prosecution, which opens first, has the structural advantage of establishing the primary narrative — the framework within which all subsequent evidence will be interpreted. Research on primacy effects in juror decision-making (Penrod & Hastie, 1979; Furnham, 1986) has consistently established that information presented first has disproportionate influence on final judgments. The prosecution’s opening is not merely a preview of evidence; it is the installation of a narrative framework that will operate as a cognitive filter on everything the jury subsequently hears. Evidence that is consistent with the prosecution’s narrative will be remembered and weighted; evidence that is inconsistent will be reinterpreted, minimised, or forgotten. The jury that reaches the end of a criminal trial has not evaluated two competing accounts from a neutral baseline. They have evaluated the defence account from within the cognitive structure that the prosecution’s opening constructed.
Cross-examination is the adversarial system’s primary mechanism for testing evidence. A witness who gives evidence in chief is then subjected to questioning by the opposing party, whose goal is to undermine the credibility, accuracy, or completeness of that evidence. In principle, this is a truth-seeking operation: the cross-examiner challenges the witness’s account and exposes its weaknesses, leaving the jury better positioned to evaluate its accuracy. In practice, cross-examination as routinely conducted is not a truth-seeking operation but a narrative-construction operation — one whose principal tools are linguistic, and whose effects operate directly on the witness’s memory. The foundational empirical research here is Elizabeth Loftus’s work on the misinformation effect, introduced in Chapter 2. Loftus and Palmer’s (1974) demonstration that the verb used in a question about a car accident affected both speed estimates and subsequent memory of physical details established that post-event language does not merely elicit memory — it alters it. Subsequent decades of research have established the robustness and generality of this effect across a wide range of events, populations, and types of misinformation (Loftus, 2005). The relevance to cross-examination is direct and devastating. The cross-examining attorney who asks ’When the defendant ran out of the building, what did he do next?’ has embedded the presupposition that the defendant ran — and research suggests that witnesses who answer this question, even those who initially reported that the defendant walked, will subsequently be more likely to remember the defendant as having run. The cross-examination has not revealed the truth. It has edited the memory of the witness.
The scope of the misinformation effect has been documented across more than four decades of research. Key studies include:
Loftus & Palmer (1974): Participants watched a film of a car accident and were asked "How fast were the cars going when they [smashed / collided / bumped / hit / contacted] each other?" The verb used in the question produced significantly different speed estimates: 'smashed' produced an average estimate of 40.5 mph; 'contacted' produced 31.8 mph. One week later, participants who had received the 'smashed' condition were more than twice as likely to report seeing broken glass (there was none in the film). The question did not merely influence the answer. It altered the memory.
Loftus, Miller & Burns (1978): Participants who were exposed to misleading post-event information about a traffic sign (yield sign vs. stop sign) subsequently 'remembered' seeing the sign described in the misleading question, even when they had originally correctly identified the actual sign. The misinformation did not merely coexist with the original memory. It replaced it.
Loftus (1993): In a review of the field, Loftus documented that approximately 22% of participants across studies accepted false information as their own memory after exposure to misleading post-event information. The effect was robust across populations, types of events, and forms of misinformation.
Hyman, Husband & Billings (1995): Participants who were told by family members that a false childhood event had occurred (e.g., being lost in a shopping centre) came to 'remember' the event with vivid detail — including sensory details, emotional responses, and narrative structure — after as few as three interviews. Approximately 25% of participants created full false memories of events that never occurred.
Application to cross-examination: The cross-examining attorney has available every technique documented in this literature: presuppositional questions ("When you saw the defendant running..."), repetition with variation (repeating a question with slightly altered framing until the witness's memory conforms), confirmation bias exploitation ("So you would agree that..."), and authority-based suggestibility (the social authority of the courtroom, the robed judge, and the assertive questioner all increase the witness's susceptibility to suggestion).
The witness who leaves the stand after sustained cross-examination has had their memory systematically altered by a process that the legal system presents as truth-seeking. The cross-examination was not a test of the witness's memory. It was an intervention in it.
The scope of the misinformation effect has been documented across more than four decades of research. Key studies include:
Loftus & Palmer (1974): Participants watched a film of a car accident and were asked "How fast were the cars going when they [smashed / collided / bumped / hit / contacted] each other?" The verb used in the question produced significantly different speed estimates: 'smashed' produced an average estimate of 40.5 mph; 'contacted' produced 31.8 mph. One week later, participants who had received the 'smashed' condition were more than twice as likely to report seeing broken glass (there was none in the film). The question did not merely influence the answer. It altered the memory.
Loftus, Miller & Burns (1978): Participants who were exposed to misleading post-event information about a traffic sign (yield sign vs. stop sign) subsequently 'remembered' seeing the sign described in the misleading question, even when they had originally correctly identified the actual sign. The misinformation did not merely coexist with the original memory. It replaced it.
Loftus (1993): In a review of the field, Loftus documented that approximately 22% of participants across studies accepted false information as their own memory after exposure to misleading post-event information. The effect was robust across populations, types of events, and forms of misinformation.
Hyman, Husband & Billings (1995): Participants who were told by family members that a false childhood event had occurred (e.g., being lost in a shopping centre) came to 'remember' the event with vivid detail — including sensory details, emotional responses, and narrative structure — after as few as three interviews. Approximately 25% of participants created full false memories of events that never occurred.
Application to cross-examination: The cross-examining attorney has available every technique documented in this literature:
The witness who leaves the stand after sustained cross-examination has had their memory systematically altered by a process that the legal system presents as truth-seeking. The crossexamination was not a test of the witness's memory. It was an intervention in it.
This effect is not limited to single questions. A sustained cross-examination — conducted over minutes or hours, comprising dozens of questions, many of them leading, many of them embedding contested presuppositions — is a systematic intervention in the witness’s episodic memory. The witness who leaves the stand at the end of cross-examination does not have the same memory of the events in question as they had when they took the stand. The cross-examination has been a memory surgery — conducted without anaesthetic, in public, before a jury who will evaluate the post-surgical account as though it were the witness’s authentic recollection.
Presupposition Trap The leading question is the primary instrument of cross-examination. A leading question is one that suggests its own answer — typically by embedding a specific claim as a presupposition. ’Isn’t it true that you were angry with the defendant on the day in question?’ does not neutrally inquire into the witness’s emotional state. It proposes anger as a presupposition and invites the witness to confirm or deny it. The cognitive and conversational dynamics of this structure systematically favour confirmation: the presupposition is already established as the background against which the witness must formulate their response, and the social dynamics of the courtroom — the authority of the examiner, the formality of the setting, the time pressure of real-time response — all reduce the witness’s capacity to identify and challenge the embedded assumption. Loftus (1975) demonstrated experimentally that the form of a question affects not only how witnesses answer but what they subsequently remember. Witnesses who were asked ’Did you see the broken headlight?’ were significantly more likely to subsequently report having seen a broken headlight — which had not been present — than witnesses asked ’Did you see a broken headlight?’ The definite article ’the’ presupposes the existence of the headlight; the indefinite ’a’ does not. This is a difference of a single word. In a complex cross-examination comprising hundreds of carefully crafted questions, the cumulative effect of embedded presuppositions on witness memory and testimony is potentially profound. The rules of evidence governing leading questions in Australian courts (Evidence Act 1995 (Cth), ss 37, 42) permit leading questions in cross-examination as a matter of right — a recognition that the adversarial purpose of cross-examination is to challenge evidence rather than to elicit it fresh. What this permission does not acknowledge is that leading questions in cross-examination do not merely challenge prior evidence. They produce new evidence — new memories, new accounts — that is then presented to the jury as the witness’s authentic recollection. The evidentiary rules that regulate leading questions were not designed with the Loftus research in mind. They reflect a model of memory as stable retrieval that the cognitive science of the past five decades has comprehensively dismantled.
Prior Convictions, and the Mythology of the Bad Person Among the most powerful narrative mechanisms available to the prosecution — and among the most extensively regulated — is evidence of the defendant’s prior criminal history. The general rule in Australian evidence law, embodied in the tendency evidence provisions of the Uniform Evidence Acts (Evidence Act 1995 (Cth), s 97), is that evidence of a person’s prior conduct is not admissible to prove a tendency to act in a certain way unless the probative value substantially outweighs the prejudicial effect. This rule reflects an awareness, at the level of doctrinal principle, that prior conviction evidence is highly prejudicial — that juries who learn of a defendant’s previous offending will be systematically inclined to convict on the current charge regardless of the independent evidence. The doctrinal awareness does not resolve the problem. Research on the effects of prior conviction evidence on juror decision-making has consistently found that even when juries are instructed to disregard such evidence for improper purposes, the evidence affects verdicts (Wissler & Saks, 1985; Lloyd-Bostock, 2000; Dempsey & Beauregard, 2014). This is not a failure of jury discipline. It is a predictable consequence of the way human cognition processes narrative information. Once the story of the defendant as ’a person who has done this sort of thing before’ has been installed in the juror’s cognitive framework, it cannot be surgically removed by a judicial direction. The instruction to ’disregard the prior conviction except for the limited purpose of assessing credibility’ asks jurors to perform a cognitive operation — the selective use of information — that is inconsistent with the integrated, schema-driven processes through which human beings actually make judgments about other people (Kahneman, 2011). In Barthes’s (1957/2009) terms, the prior conviction is a myth in the technical sense: a second-order signification that transforms contingent legal history into natural character. The defendant who has a prior conviction for violence is not merely someone who was previously found guilty of a violent act. Within the narrative architecture of the trial, they become ’a violent person’ — a subject type whose current conduct is to be read through the lens of their established character. The legal rules that attempt to prevent this reading are, in the face of the cognitive dynamics that produce it, largely ceremonial.
: The Authorisation of Interpretation Expert witnesses occupy a distinctive position in the adversarial trial. Unlike lay witnesses, who are in principle limited to reporting their direct observations, expert witnesses are permitted to offer opinions — to interpret facts, draw inferences, and present conclusions. The basis of this permission is the assumption that the expert possesses specialised knowledge that equips them to make inferences beyond the capacity of the ordinary juror. In practice, expert witnesses are retained by the parties, and their evidence — however framed in the language of objective analysis — tends to support the case of the party who retained them. The adversarial system’s approach to expert evidence — each side retaining their own experts, whose conflicting opinions are then presented to a lay jury — has been extensively criticised as epistemologically incoherent (Edmond, 2000; Freckelton & Selby, 2019). The jury is asked to choose between two competing expert analyses of, for instance, DNA evidence, forensic accounting, or psychiatric diagnosis, without possessing the technical knowledge required to evaluate the methodological validity of either. In practice, juries evaluate expert evidence using the same narrative criteria they apply to all evidence: credibility, coherence, and resonance with existing schemas. The expert witness who presents most confidently, most clearly, and in a way most consistent with the jury’s prior expectations will tend to prevail — regardless of the technical quality of their analysis. From the perspective of this thesis, the most significant function of expert witnesses is their role in authorising interpretations of ambiguous conduct. The forensic psychologist who testifies that the defendant exhibits personality characteristics consistent with the offence type, the behavioural analyst who interprets post-offence conduct as indicative of consciousness of guilt, the police investigator who offers expert evidence on the significance of specific behavioural indicators — each of these witnesses is performing a Foucauldian power/knowledge operation: deploying institutional authority to constitute an authoritative account of what the defendant’s behaviour means. The jury, who lack the framework to challenge this account on its own terms, is left to evaluate it narratively. A confident, credentialled expert presenting a coherent account of what the defendant’s behaviour means is a powerful narrative actor, regardless of the empirical validity of the interpretive framework they are deploying.
Language Nobody Understands At the conclusion of a criminal trial, the judge directs the jury on the law applicable to their deliberations. These directions — explanations of the elements of the offence, the standard of proof, the application of specific evidentiary rules — are the primary mechanism through which the legal system attempts to ensure that jury decisions are grounded in legally correct reasoning. Research on juror comprehension of judicial directions has produced consistently dispiriting results. Ogloff and Rose (2005) conducted a comprehensive review of Australian research on jury comprehension and concluded that jurors routinely misunderstand critical legal directions, including the standard of proof (beyond reasonable doubt), the presumption of innocence, and specific evidentiary directions regarding prior conviction evidence. Studies using simulated trials in Australian and comparable jurisdictions have found that comprehension of standard judicial directions rarely exceeds 50% on specific content questions, and that revised, plain-English directions produce only modest improvements (Lieberman & Sales, 1997; Trimboli, 2008). The jury that deliberates on a criminal charge without having understood the legal framework it has been instructed to apply is not engaging in legally constrained reasoning. It is engaging in unadulterated lay judgment — narrative assessment, folk psychology, and the application of cultural schemas — dressed in the institutional authority of a legal verdict. The language of judicial directions is itself a site of constructed meaning. The direction that the jury must be satisfied ’beyond reasonable doubt’ communicates nothing determinate to a person without legal training. Research on lay understanding of the beyond reasonable doubt standard has found wildly varying interpretations: some jurors interpret it as requiring virtual certainty, others as requiring only that guilt be more likely than not (Horowitz & Kirkpatrick, 1996; Young et al., 1999). The standard’s function as a safeguard against wrongful conviction depends on its being interpreted as a high threshold. Its systematic misinterpretation as a lower threshold — combined with the prosecution-favouring cognitive architecture described throughout this chapter — means that the standard operates in practice at a level significantly below its doctrinal intention. The words say ’beyond reasonable doubt.’ The understanding hears something considerably more permissive.
Rhetoric of Closing Addresses The closing address is the moment at which the partisan nature of the adversarial trial is most openly acknowledged. Counsel is explicitly permitted to be persuasive — to advocate for their client’s case, to emphasise favourable evidence, to challenge the credibility of witnesses, and to construct a narrative that makes sense of the trial as a whole. It is also the moment at which the gap between the rhetorical and epistemic functions of the trial is most clearly visible. Research on the persuasive strategies employed in closing addresses has identified systematic use of narrative framing, emotional appeals, and strategic use of evidence ordering (Spiecker & Worthington, 2003). Prosecutors who invoke the suffering of victims, who construct emotionally resonant narratives of the defendant’s culpability, and who end their closing on a note of moral certainty are not engaging in truth-telling; they are engaging in advocacy. The rules governing closing addresses — the prohibition on misleading the jury, the requirement of fairness to the accused — are enforced through judicial supervision and appellate review, but they address explicit misconduct, not the pervasive and entirely permissible rhetorical architecture of persuasive advocacy. The defendant who is convicted following a skillful prosecution closing address and an ineffective defence closing has not been found guilty by a process that weighted the evidence neutrally. They have been found guilty by a process that was won by better storytelling. In a system that presents its verdicts as objective findings of fact, this is a more significant admission than it is usually acknowledged to be.
6.9
Additional References (Chapters 5 & 6) (To be combined with full reference list in final document.) Crimes Act 1900 (NSW). Damaska, M. R. (1997). Evidence law adrift. Yale University Press. Dempsey, J., & Beauregard, E. (2014). Profiling tendency evidence and prior convictions: A case study analysis. Journal of Criminal Law, 78(4), 296–312. Edmond, G. (2000). Whigs in court: Historiographical problems with expert evidence. University of New South Wales Law Journal, 23(3), 1–31. Evidence Act 1995 (Cth). Findley, K. A., & Scott, M. S. (2006). The multiple dimensions of tunnel vision in criminal cases. Wisconsin Law Review, 2006(2), 291–397. Freckelton, I., & Selby, H. (2019). Expert evidence: Law, practice, procedure and advocacy (6th ed.). Thomson Reuters. Furnham, A. (1986). The robustness of the recency effect: Studies using legal evidence. Journal of General Psychology, 113(4), 351–357. https://doi.org/10.1080/00221309.1986.9710569 Goff, P. A., Jackson, M. C., Di Leone, B. A. L., Culotta, C. M., & DiTomasso, N. A. (2014). The essence of innocence: Consequences of dehumanizing Black children. Journal of Personality and Social Psychology, 106(4), 526–545. https://doi.org/10.1037/a0035663 Graham, S., & Lowery, B. S. (2004). Priming unconscious racial stereotypes about adolescent offenders. Law and Human Behavior, 28(5), 483–504. https://doi.org/10.1023/B:LAHU.0000046430.65485.1f Horowitz, I. A., & Kirkpatrick, L. C. (1996). A concept in search of a definition: The effects of reasonable doubt instructions on certainty of guilt standards and jury verdicts. Law and Human Behavior, 20(6), 655–670. https://doi.org/10.1007/BF01499236 Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus and Giroux. Lange v Australian Broadcasting Corporation (1997) 189 CLR 520. Larcombe, W. (2002). The ’ideal’ victim v successful rape complainants: Not what you might expect. Feminist Legal Studies, 10(2), 131–148. https://doi.org/10.1023/A:1016539403573 Lieberman, J. D., & Sales, B. D. (1997). What social science teaches us about the jury instruction process. Psychology, Public Policy, and Law, 3(4), 589–644. https://doi.org/10.1037/10768971.3.4.589 Lloyd-Bostock, S. (2000). The effects on juries of hearing about the defendant’s previous criminal record: A simulation study. Criminal Law Review, 2000, 734–755. Loftus, E. F. (1975). Leading questions and the eyewitness report. Cognitive Psychology, 7(4), 560–572. https://doi.org/10.1016/0010-0285(75)90023-7 Model Criminal Code — Model Criminal Code Officers Committee. (2009). Model Criminal Code. Attorney-General’s Department. Ogloff, J. R. P., & Rose, V. G. (2005). The comprehension of judicial instructions. In N. Brewer & K. Williams (Eds.), Psychology and law: An empirical perspective (pp. 407–444). Guilford Press. Penrod, S., & Hastie, R. (1979). Models of jury decision making: A critical review. Psychological Bulletin, 86(3), 462–492. https://doi.org/10.1037/0033-2909.86.3.462 Police Powers and Responsibilities Act 2000 (Qld). Rachlinski, J. J., Johnson, S. L., Wistrich, A. J., & Guthrie, C. (2009). Does unconscious racial bias affect trial judges? Notre Dame Law Review, 84(3), 1195–1246. Richardson, L. S., & Goff, P. A. (2012). Implicit racial bias in public defender triage. Yale Law Journal, 122(1), 2626–2649. Schneider, E. M. (1992). Particularity and generality: Challenges of feminist theory and practice in work on woman-abuse. New York University Law Review, 67(3), 520–568. Spiecker, S. C., & Worthington, D. L. (2003). The influence of opening statement/closing argument organizational strategy on juror verdict and damage awards. Law and Human Behavior, 27(4), 437–456. https://doi.org/10.1023/A:1024041505879 Stubbs, J., & Tolmie, J. (1999). Falling short of the challenge? A comparative assessment of the Australian use of expert evidence on the battered woman syndrome. Melbourne University Law Review, 23(3), 709–748. Temkin, J., & Krah?, B. (2008). Sexual assault and the justice gap: A question of attitude. Hart Publishing. Trimboli, L. (2008). Juror understanding of judicial instructions in criminal trials (Crime and Justice Bulletin No. 119). NSW Bureau of Crime Statistics and Research. Wissler, R. L., & Saks, M. J. (1985). On the inefficacy of limiting instructions: When jurors use prior conviction evidence to decide on guilt. Law and Human Behavior, 9(1), 37–48. https://doi.org/10.1007/BF01044288 Young, W., Cameron, N., & Tinsley, Y. (1999). Juries in criminal trials (NZLC PP37). New Zealand Law Commission. CONSTRUCTED GUILT: LANGUAGE, POWER, AND THE CRIMINAL JUSTICE SYSTEM 1 Figure 3.1: Cross-examination and memory. Chapter 4 Media, Jury, and Synthesis
# Chapter 7: The Pre-Trial Verdict
By the time a defendant stands in the dock of a criminal court, they have frequently already been tried and convicted in a jurisdiction with no rules of evidence, no presumption of innocence, no right of cross-examination, and an audience of millions. The media trial — the construction of a public narrative of guilt through news reporting, social media commentary, and true crime content — precedes the legal trial and shapes its conditions in ways that are systematically underdiscussed within criminal law scholarship. This chapter argues that media framing of criminal defendants is a site of guilt construction operating through the same semiotic and narrative mechanisms identified in previous chapters, but at a scale and with a reach that the institutional protections of the courtroom cannot contain. The jury that deliberates on a high-profile criminal matter is not a blank slate. It is a panel of people who have lived, in many cases for months or years, inside a media environment that has already told them who did it.
Criminal Identity Framing theory, as developed by Goffman (1974) and applied to media analysis by Entman (1993), argues that the way information is presented — the frame through which events are described — determines not merely how audiences understand specific facts but what questions they ask, what causes they infer, what moral evaluations they make, and what remedies they consider appropriate. A frame is not a bias in the pejorative sense; it is an unavoidable feature of all communication. Every account of events selects some details and omits others, foregrounds some actors and backgrounds others, and invokes some causal narratives and ignores alternatives. The question is not whether media coverage of criminal cases frames events — it does, necessarily — but whose interests particular frames serve and what effects they produce on the audiences who receive them. Research on media framing of criminal defendants has identified consistent patterns. Greer and McLaughlin (2012) documented the emergence of what they term ’trial by media’ as a distinctive institutional phenomenon: the construction of a public verdict on individual culpability through sustained, intensive, and frequently prejudicial media coverage that precedes, accompanies, and sometimes determines the outcome of formal legal proceedings. The hallmarks of this process include the early designation of a suspect as perpetrator in news framing, the selective reporting of incriminating details while omitting exculpatory context, the use of loaded language (’monster,’ ’predator,’ ’killer’) that preemptively characterises the defendant in terms that presuppose guilt, and the mobilisation of victim-centred narratives that position the defendant as the appropriate object of collective condemnation. In Australian contexts, the relationship between media coverage and criminal proceedings has been the subject of sustained legal and scholarly attention, particularly following high-profile cases in which intensive pre-trial publicity was alleged to have contaminated jury pools (Crofts, 2007; Keyzer et al., 2010). The legal response — through suppression orders, change of venue applications, and directions to juries to ignore prior publicity — has been consistently demonstrated to be inadequate. Steblay et al.’s (1999) meta-analysis of 44 studies examining pre-trial publicity effects on juror judgments found a consistent and significant relationship between exposure to prejudicial publicity and conviction-favourable judgments, and found that judicial instructions to disregard prior publicity were largely ineffective in eliminating this effect. The law’s primary tool for managing media contamination of jury decision-making is a judicial instruction. The instruction does not work.
Language of Guilt: Presupposition in News Reporting News reports of criminal allegations deploy a linguistic architecture that systematically presupposes guilt while maintaining formal deniability through the mechanism of attribution. The headline ’Police charge man over brutal attack’ does not formally assert guilt — the charge, not a conviction, is the subject of the report. But the use of ’brutal’ as a descriptor of the attack (rather than ’alleged attack’), the selection of ’over’ rather than ’in connection with,’ and the passive construction that renders the victim’s experience as a fixed fact while the defendant’s role remains formally attributed — these are micro-linguistic choices that cumulatively construct a framing of guilt. Van Dijk’s (1991) critical discourse analysis of news language identified the systematic use of what he termed ’semantic macrostructures’ — the thematic frames that organise the meaning of news texts at a level above individual sentences. In crime reporting, the dominant semantic macrostructure is typically the detection narrative: a wrongful act has been committed, it has been attributed to a specific person, and the justice system is in the process of confirming that attribution and administering appropriate consequences. This narrative structure positions the defendant as the already-established wrongdoer before any evidence has been evaluated. The legal process appears within this macrostructure not as a mechanism of inquiry but as a mechanism of confirmation. The word ’alleged’ — the standard journalistic hedge against defamation liability — is frequently cited as evidence that news coverage respects the presumption of innocence. Its function is, however, more complex. Research on the linguistic processing of negation and hedging has established that qualifications embedded in otherwise assertive framings tend to be cognitively subordinated to the primary assertion (Mayo et al., 2004). The reader of ’alleged killer John Smith’ does not process the ’alleged’ as equivalent in cognitive weight to ’killer.’ The hedge is formal; the impact is nominal. The designation ’alleged killer’ operates, in practice, more similarly to ’killer’ than to ’person accused of killing’ — a distinction that matters considerably to the cognitive framework within which subsequent information about the defendant is processed.
, Viral Conviction, and the Collapse of Sequencing The traditional media trial operated within a temporal sequence that, while imperfect, preserved some structural separation between pre-trial publicity and trial process: reports appeared in print or broadcast media, with some lead time between publication and jury assembly. The digitalisation of news and the emergence of social media have collapsed this sequencing entirely. A criminal allegation can now be reported, amplified, commented upon, and subjected to crowdsourced investigation within hours of an arrest, reaching a potential jury pool of the entire population of any connected device before any formal legal proceedings have commenced. The epistemological consequences of this collapse are severe. Social media platforms reward emotional engagement over accuracy; the content that spreads most rapidly is typically the most emotionally provocative, which in crime contexts means the most guilt-presumptive, victim-centred, and perpetrator-condemning material (Bail et al., 2018). True crime content — podcasts, documentary series, and investigative social media accounts that reconstruct criminal cases with varying degrees of accuracy and almost universal guilt-presumptive framing — has become a major media genre with audience sizes that dwarf traditional news coverage. A potential juror in a high-profile criminal case in the contemporary media environment has typically not merely read some news reports. They have listened to multiple podcast episodes, watched documentary reconstructions, read Reddit threads containing names, photographs, and amateur ’analysis’ of the defendant’s behaviour, and formed a settled view of guilt months or years before the trial commences. The legal system’s response to this environment — the voir dire process for identifying and excluding jurors with prejudicial prior knowledge, suppression orders of disputed efficacy, and judicial directions — was designed for a media environment that no longer exists. The doctrinal mechanisms for protecting the presumption of innocence in jury selection and trial management have not kept pace with the transformation of the information environment within which potential jurors are embedded. This is not a regulatory failure that could be corrected by better suppression orders. It is a structural condition of the contemporary information environment, and it means that the presumption of innocence, for high-profile defendants, is practically inoperative before the trial begins.
Class, and the Differential Application of the Suspect Frame The framing of criminal defendants in media is not applied uniformly. Extensive research across multiple jurisdictions has documented systematic differences in how defendants of different racial and socioeconomic backgrounds are framed in news coverage. Dixon and Linz (2000) analysed local television news coverage of crime in Los Angeles and found that Black and Latino suspects were significantly overrepresented relative to their actual proportion of those arrested, and that they were more likely to be shown in police custody, less likely to be identified by name, and less likely to be described in terms that contextualised their conduct or humanised their circumstances. The media suspect frame — the visual and linguistic grammar of crime reporting — is racialised in ways that consistently position defendants of colour as more threatening, more culpable, and less deserving of the benefit of the doubt. In Australia, this dynamic is most starkly visible in the media treatment of Aboriginal and Torres Strait Islander defendants and in the framing of crime in communities with high First Nations populations. Research on Australian media coverage of Indigenous Australians in criminal justice contexts has documented consistent patterns of deficit framing — the presentation of Indigenous defendants within a narrative of cultural dysfunction and pathological criminality — that positions Aboriginal defendants as a category of person for whom guilt is presumed rather than established (Harding et al., 1998; Dreher, 2010). This framing operates at the level of the semantic macrostructure: the individual defendant is positioned within a pre-existing narrative about who commits crime in Australia, and that narrative is racialised in ways that the formal language of legal neutrality cannot obscure.
The analysis of this chapter points to a conclusion that is uncomfortable but empirically well-supported: for defendants in cases that attract significant media attention, the presumption of innocence is, in any practical sense, not operative. It exists as legal doctrine. It is formally asserted in jury directions. And it is systematically overridden by the cumulative effect of pre-trial publicity that installs guilt-presumptive narratives in the cognitive frameworks of potential jurors, who then enter the courtroom not as blank slates but as people who have already decided, in an informal but cognitively real sense, that the defendant is guilty. The trial is then the process by which that pre-existing judgment is either confirmed — the most common outcome — or overturned by evidence sufficiently powerful to dislodge a settled cognitive commitment. This is not merely a critique of media practice, though it is that. It is a critique of a legal system that invokes the presumption of innocence as a foundational principle while making no serious institutional provision for its protection in the contemporary information environment. The gap between the doctrine and the reality is not an incidental feature of modern criminal procedure. It is a structural condition that serves the system’s operational interest in conviction rates by ensuring that the most publicly significant cases — the cases that most require the presumption of innocence to function — are precisely the cases in which it has been most thoroughly dismantled before the jury is sworn.
7.7
# Chapter 8: Twelve People Who Weren't There
Weren’t There The Jury, Narrative Rationality, and the Construction of the Verdict
The jury is the institution through which the criminal justice system performs its most significant act of legitimation. The verdict of a jury of peers is presented as the closest approximation to truth that a fallible human system can achieve — the judgment of ordinary citizens, free from institutional bias, evaluating evidence according to the standard of proof and the judge’s directions, and reaching a collective conclusion about what happened. This legitimating narrative is foundational to the common law criminal trial. It is also, this chapter argues, substantially false. The jury that deliberates on a criminal charge is not a neutral truth-finding panel. It is a group of people who have been subjected to the full rhetorical and institutional architecture described in the preceding chapters, who bring to their deliberations the cognitive schemas, cultural assumptions, and narrative frameworks of their social formation, and who reach a verdict through processes that are far more narrative and far less evidential than the system’s self-presentation acknowledges.
Fisher’s Narrative Paradigm and Jury Decision-Making Walter Fisher’s (1984) narrative paradigm proposes that human beings are fundamentally storytelling creatures — that the primary mode of human cognition and judgment is narrative rather than logical, and that we evaluate claims according to narrative rationality rather than formal logic. Narrative rationality has two components: narrative probability (does the story hang together coherently? are its elements consistent with each other and with the established narrative framework?) and narrative fidelity (does the story resonate with our experience of how the world works? does it cohere with the stories we already believe?). Fisher’s paradigm has been extensively applied to jury decision-making, and the research consistently supports the view that jury verdicts are predominantly narrative judgments rather than logical deductions from evidence (Bennett & Feldman, 1981; Pennington & Hastie, 1992). Pennington and Hastie’s (1992) story model of jury decision-making provides the most empirically developed account of this process. Their research established that jurors do not weigh evidence item by item in a cumulative logical process. They construct a story — a narrative account of what happened — and then match that story to the available verdict categories to determine which verdict best fits the story they have constructed. The verdict is not the output of evidence evaluation. It is a narrative classification. And the story that jurors construct is not built exclusively from the evidence presented at trial. It is built from the evidence, plus the jurors’ prior knowledge, cultural schemas, and narrative expectations about how the world works and about what kind of people do what kind of things. The implications are significant. A prosecution case that produces a coherent, emotionally resonant, and culturally familiar narrative will tend to produce conviction, regardless of whether its specific factual claims have been established beyond reasonable doubt in the logical sense. A defence case that challenges the prosecution’s narrative without offering a coherent alternative will tend to fail, regardless of the formal adequacy of the reasonable doubt it has raised. The question for the jury is not ’has guilt been proven beyond reasonable doubt?’ It is, in practice, ’which story makes more sense?’ — and that question is answered by narrative rationality, not by the standard of proof.
The cognitive psychological concept of the schema — an organised mental framework for interpreting and assimilating new information, developed through experience and cultural learning — is central to understanding how jurors process evidence (Bartlett, 1932; Neisser, 1976). Schemas operate as cognitive templates: they guide what information is attended to, how ambiguous information is interpreted, what is remembered, and how gaps in available information are filled. In the context of jury decision-making, schemas about crime, about defendants, about victims, and about how criminal events typically unfold are the cognitive infrastructure within which trial evidence is processed. The operation of schemas in jury decision-making produces systematic departures from the evidence-based neutrality that the system presupposes. Information that is consistent with an activated schema is processed more fluently, remembered more accurately, and weighted more heavily than schema-inconsistent information (Taylor & Crocker, 1981). Information that is inconsistent with the activated schema is more likely to be reinterpreted to fit it — what Bartlett (1932) called ’rationalisation’ — or to be forgotten. A juror whose schema for ’sexual assault perpetrator’ includes a specific profile of behaviour and presentation will interpret the defendant’s conduct in the dock through that schema, regardless of whether the schema is empirically valid. Research on rape myth acceptance — the endorsement of false beliefs about how sexual assaults occur and how genuine victims behave — has documented that jurors with high levels of rape myth acceptance acquit at significantly higher rates in sexual assault cases, not because they evaluate the evidence differently but because they apply a schema that designates the complainant’s behaviour as inconsistent with ’real’ victimhood (Larcombe, 2002; Schuller & Hastings, 1996). The schema also operates at the level of the defendant’s social identity. Research on racial schemas and juror decision-making (Sommers & Ellsworth, 2001; Levinson & Young, 2010) has established that jurors apply different schemas to defendants of different racial identities — schemas that encode different presumptions about culpability, dangerousness, and credibility. These schemas operate at an implicit level: jurors who endorse racial equality explicitly, and who would deny that race affected their judgment, nonetheless show systematic racial disparities in verdict outcomes when race is manipulated in simulated trial research. The schema is not a consciously held belief. It is a cognitive structure that operates below the level of reflective awareness and shapes judgment in ways the juror cannot introspect or report.
Jury deliberation is frequently presented as the mechanism that corrects individual error — the process through which the idiosyncratic biases of individual jurors are cancelled out through collective discussion and the requirement of consensus. This is a plausible hypothesis. The empirical evidence does not support it. Research on group decision-making has consistently found that group deliberation does not average individual judgments toward a neutral centre; it tends to amplify the pre-deliberation majority position through a process called group polarisation (Myers & Lamm, 1976; Sunstein, 2002). A jury that enters deliberation with a majority leaning toward conviction will, through deliberation, tend to move further toward conviction rather than toward the median of individual pre-deliberation views. The social dynamics of the deliberation room also produce conformity pressures that systematically disadvantage minority viewpoints. Asch’s (1951) classic research on conformity to group pressure established that individuals will publicly state judgments they privately believe to be incorrect when faced with unanimous group disagreement. In the deliberation room, the juror who holds a minority view — who believes, for instance, that a reasonable doubt exists where the majority do not — faces social pressure to conform that operates independently of and often in opposition to the epistemic merits of their position. Research on hung juries and minority influence in deliberation has established that lone holdout jurors face intense social pressure to change their votes, and that the system’s unanimous verdict requirement (in Australian superior court criminal trials) tends to resolve this pressure through capitulation rather than persuasion (Sandys & Dillehay, 1995; Hastie et al., 1983). The practical consequence is that the unanimous verdict — the doctrinal guarantee that every juror has been persuaded of guilt beyond reasonable doubt — frequently reflects the social dynamics of group pressure rather than the independent epistemic judgment of twelve individuals. The juror who votes guilty under social pressure from eleven peers who are confident, assertive, and impatient has not been persuaded that guilt has been established. They have been worn down. The verdict records both outcomes identically.
Chapter 6 addressed the research on juror comprehension of judicial directions in the context of courtroom procedure. It is worth returning to this issue in the context of the deliberation process, because the gap between what the law instructs and what jurors understand becomes operationally significant at the deliberation stage. When jurors retire to deliberate, they carry with them their comprehension — or, as the research consistently establishes, their miscomprehension — of the legal framework they are supposed to apply. The deliberation is then not an application of the legal standard to the evidence; it is an application of each juror’s idiosyncratic interpretation of the legal standard to their narrative account of the evidence. Research by Finkel (1995) on ’commonsense justice’ — the moral intuitions that jurors actually apply in reaching verdicts, as distinct from the legal standards they are instructed to apply — found that jurors systematically deviate from legal standards in the direction of their own moral intuitions, and that these intuitions frequently diverge from the law in ways that advantage the prosecution. Jurors who believe that a person would not confess to a crime they did not commit will discount evidence of false confession without reference to the Kassin literature or to the Reid Technique. Jurors who believe that innocent people cooperate readily with police investigations will interpret the exercise of the right to silence as indicative of guilt, despite explicit judicial direction to the contrary. The jury applies commonsense justice. Commonsense justice, as a general matter, presumes guilt more readily than the law requires.
The preceding analysis generates a prediction: a system characterised by pre-interrogation degradation that produces suggestible suspects, interrogation methodology designed to produce confessions regardless of guilt, legislative language that encodes normatively loaded terms as objective standards, courtroom processes that rewrite memory and install guilt-presumptive narratives, media coverage that contaminates jury pools before trial begins, and jury deliberation processes driven by narrative rationality and group polarisation rather than evidence-based reasoning — such a system should produce wrongful convictions at a predictable and significant rate. The empirical record is consistent with this prediction. The Innocence Project, operating in the United States since 1992, had secured exonerations for more than 375 wrongfully convicted individuals by 2023, the vast majority through post-conviction DNA testing (Innocence Project, 2023). The National Registry of Exonerations, which employs a broader evidentiary threshold, documented more than 3,300 exonerations in the United States between 1989 and 2023 (National Registry of Exonerations, 2023). In Australia, formal exonerations are less systematically documented, but cases including those of Lindy Chamberlain-Creighton, Henry Keogh, and Derek Bromley — each of whom served extended periods of imprisonment for offences they did not commit, and each of whose convictions was produced by a recognisable combination of the mechanisms identified in this thesis — establish that the Australian criminal justice system produces wrongful convictions through the same structural processes. The wrongful conviction is not an anomaly to be explained by individual failure. It is the predictable output of a system that is architecturally designed to produce conviction. The false confession is the predictable output of a pre-interrogation sequence that degrades cognitive capacity and an interrogation methodology that exploits suggestibility. The wrongful eyewitness identification is the predictable output of a memory system that is reconstructive and susceptible to post-event suggestion, combined with identification procedures that embed leading cues. The wrongful verdict is the predictable output of a narrative competition conducted before jurors who have been contaminated by pre-trial publicity, who apply schema-driven reasoning rather than evidence-based evaluation, and who are subject to group polarisation toward the majority view. These are not failures. They are features.
8.7
# PART III: SYNTHESIS
# Chapter 9: The System Is Not Broken
This thesis has advanced a single, sustained argument across nine substantive chapters: that guilt, as produced by the criminal justice system, is not an objective finding of fact but a performative linguistic construction—one that can be assembled from the behaviour of any individual, regardless of actual culpability, through a sequential institutional process in which the presumption of innocence is practically negated at every stage at which it should operate.
The argument has proceeded in three interlocking registers:
Theoretical: Drawing on Saussurean semiotics, Wittgensteinian philosophy of language, Foucauldian discourse theory, speech act theory, and critical legal studies, Chapter 2 established that legal language does not describe a pre-existing reality but constructs the institutional realities it purports to merely reflect. Guilt is constituted by the verdict, not discovered by it.
Empirical: The psychological and criminological literature on pre-interrogation detention, interrogation methodology, memory, jury decision-making, and wrongful conviction provides extensive evidence that the guilt-construction mechanisms identified theoretically operate with measurable, documented, real-world effects.
Structural: The various mechanisms of guilt construction identified across individual chapters are not independent dysfunctions but components of a coherent system. At every stage, the structural advantage is with the prosecution. At every stage, the formal apparatus of procedural protection exists as doctrine while being practically overridden by institutional design.
The thesis's most confronting claim—and the one most directly supported by the empirical record—is that innocence provides no reliable protection against conviction within this system.
The innocent suspect who is:
—that person's innocence was never structurally relevant to the system's operation. The system did not fail them. It processed them.
The argument is not that everyone who is convicted is innocent, nor that the justice system produces only wrongful convictions. The argument is more specific: that within this system, the distinction between guilt and innocence is not reliably operative as a determinant of outcome.
A common response to critiques of criminal justice systems in liberal democratic societies is the language of systemic failure: the system is broken; it is not living up to its ideals; reform is required to bring it into conformity with its stated principles.
This language preserves the foundational legitimacy of the system's stated principles—neutrality, equal treatment, the presumption of innocence—while attributing their non-realisation to correctable malfunction rather than to design.
This thesis consistently rejects this framing.
The design includes the pre-interrogation detention regime because a physiologically compromised suspect is more likely to confess.
The design includes the Reid Technique because confession rates are institutionally valued over accuracy rates.
The design includes the voluntariness doctrine's inattention to the neurobiological consequences of detention because a doctrine that attended to those consequences would exclude a substantial proportion of confession evidence.
The design includes juror miscomprehension of the standard of proof because accurate comprehension of a genuinely demanding standard would produce more acquittals.
None of these features are accidents. They are the accumulated institutional choices of a system whose operational success is measured in conviction rates, not in truth production.
The analysis developed in this thesis has specific implications for criminal procedure reform:
This thesis has argued that the criminal justice system constructs guilt through language. It is appropriate, in conclusion, to reflect on what it means to name that construction.
There is a politics to naming the coercive machinery of justice as machinery, to calling the presumption of innocence a performative contradiction, to identifying the confession of an innocent person as the predictable output of a designed system rather than an unfortunate accident.
That naming is not comfortable for institutions that exercise substantial power and legitimate themselves through the language of neutral justice.
The contribution is not a new politics of criminal justice but a more precise account of the mechanisms through which the existing politics operates—an account grounded in the theoretical and empirical literatures that, taken together, make the case that language does not merely describe criminal justice.
Language is criminal justice.
The empirical evidence reviewed in this thesis supports the following conclusions:
An individual may be:
The findings documented in this thesis support the following specific conclusions regarding evidentiary weight:
Investigator credibility assessments:
Evidence that an investigator assessed a person as "deceptive" or "evasive" based on behavioural observation carries no reliable probative value. The meta-analytic evidence establishes that such assessments operate at chance levels. Courts should not admit investigator testimony regarding assessments of credibility based on behavioural cues, or should instruct juries that such assessments are no more reliable than random assignment.
Confession evidence following extended detention:
Confession evidence obtained following detention periods exceeding four hours should be viewed with substantial scepticism. The compound effects of sleep disruption, stress, isolation, and environmental degradation on suggestibility are well-documented. The longer the pre-interrogation detention, the lower the probative value of any subsequent confession.
Testimony from neurodivergent witnesses:
Credibility assessments of neurodivergent individuals based on presentation features—eye contact, affect, communication style, narrative structure—are systematically unreliable. The features that reduce credibility in neurotypical assessment frameworks are features of neurodivergent presentation, not indicators of deception. Expert evidence on neurodivergent communication should be standard in cases involving neurodivergent defendants, complainants, or witnesses.
FND symptom variability:
Symptom variability in functional neurological disorder is a diagnostic feature, not evidence of fabrication. Evidence that symptoms varied across time, context, or stress level is consistent with genuine FND and should not be treated as evidence of malingering without specific expert evidence to the contrary.
Trauma memory characteristics:
Fragmented narrative, non-linear recall, delayed disclosure, and evolving accounts are consistent with genuine traumatic memory. These features are not indicators of fabrication. Expert evidence on trauma memory should be standard in cases involving traumatic events.
Accounts that become more detailed over time:
Memory for traumatic events may become more detailed as avoidance decreases and as therapeutic or other processing occurs. An account that becomes more detailed over time is not thereby less credible; this pattern is consistent with the documented phenomenology of trauma memory.
Direct communication style:
Direct communication, factual correction of errors, literal response to questions, and absence of performed deference are features of autistic communication style, not indicators of hostility, contempt, or consciousness of guilt.
Absence of expected emotional display:
Flat affect, absence of performed distress, and calm presentation during accusation may reflect autistic presentation, FND, PTSD-related dissociation, or other conditions. Absence of expected emotional display is not evidence of lack of genuine experience.
Presence of emotional display:
Visible distress, crying, shaking, or emotional dysregulation may reflect genuine distress, trauma response, autistic overwhelm, or FND symptoms. Presence of emotional display is not evidence of performance or manipulation.
The critical point:
Both presence and absence of any behavioural indicator are compatible with both guilt and innocence. Behavioural heuristics cannot distinguish between them. The appropriate evidentiary weight for behavioural presentation evidence is zero.
This thesis has documented numerous ways that specific behaviours are misinterpreted. A reader seeking to weaponise this analysis might argue: "The defendant has read this thesis and is now performing the behaviours described as 'innocent.'"
This argument fails for the following reasons:
The only valid response to this thesis is the abandonment of behavioural heuristics as an assessment method—not the sophisticated reapplication of them to assess whether someone has read the thesis.
The evidence presented in this thesis does not suggest. It proves.
Deception detection does not work. Meta-analysis of 24,483 judgments establishes accuracy at 54.1%—four percentage points above chance. Police officers, judges, psychiatrists, and trained investigators perform at the same level as untrained civilians. Training increases confidence without increasing accuracy. This is not a limitation to be acknowledged. It is a fundamental invalidity.
Confession evidence is unreliable. DNA exonerations prove—not suggest, prove—that innocent people confess to crimes they did not commit. 12–30% of exonerations involve false confessions. These are not edge cases or anomalies. They are the documented output of a system designed to produce confessions regardless of guilt.
Behavioural heuristics are invalid. 91% of the cues people use to assess credibility are either empirically unrelated to deception or directionally inverted. The behaviours observers interpret as deceptive are more strongly associated with truth-telling. The behaviours interpreted as honest are more strongly associated with deception. The instrument is not merely imprecise. It is backwards.
The reasonable conclusion:
The criminal justice system relies on:
Chamberlain was exonerated in 1988 after a jacket was discovered near a dingo lair at Uluru. She had served more than three years in prison. A 2012 coroner's inquest finally determined that Azaria Chamberlain was killed by a dingo.
Every mechanism documented in this thesis was operative in her conviction. Her innocence was structurally illegible to the system that processed her.
Henry Keogh was convicted in 1995 of the murder of his fiancée Anna-Jane Cheney, who was found drowned in a bathtub. He was sentenced to 25 years. The conviction relied principally on the forensic pathology evidence of Dr Colin Manock, who testified that bruises on the body were consistent with the deceased being held under the water.
Mechanisms operative:
Keogh served 20 years before his conviction was quashed in 2014. A review of Manock's broader forensic career revealed systemic concerns about the quality of his evidence across multiple cases.
Derek Bromley was convicted in 1984 of the murder of Stephen Docoza, who was found dead in a pond in Adelaide. Bromley, an Aboriginal man, was convicted primarily on the basis of forensic evidence and witness testimony. He has maintained his innocence for over 40 years.
Mechanisms operative:
As of 2026, Bromley remains in prison — one of the longest-serving prisoners in South Australia. His case is the subject of ongoing legal challenge.
The National Registry of Exonerations (US) documents over 3,300 exonerations since 1989. Analysis of contributing factors reveals systematic patterns consistent with the mechanisms documented in this thesis:
| Contributing Factor | % of Exonerations | Thesis Mechanism |
|---|---|---|
| Perjury/false accusation | 57% | Courtroom narrative construction (Ch. 6) |
| Official misconduct | 54% | Institutional architecture (Ch. 9) |
| Mistaken witness identification | 29% | Memory distortion (Ch. 2, Ch. 6) |
| False confession | 12% | Reid Technique + detention (Ch. 3, Ch. 4) |
| False/misleading forensic evidence | 24% | Expert evidence hierarchy (Ch. 6) |
| Inadequate legal defence | 23% | Structural inequality |
Note: Categories are not mutually exclusive; most exonerations involve multiple contributing factors.
The pattern is consistent with the thesis: wrongful convictions are not produced by single-point failures (one bad cop, one lying witness, one incompetent lawyer). They are produced by the convergence of multiple mechanisms within an architecture designed to produce conviction. The architecture ensures that each mechanism's output feeds into the next, creating a cumulative construction of guilt that becomes increasingly resistant to challenge at each successive stage.
# Chapter 10: Cultural and Neurodivergent Structural Bias
If truthful speech patterns differ significantly by cultural background, then any deception-detection instrument calibrated on a dominant cultural baseline will produce structurally higher false-positive rates for minority speakers—not because they lie more, but because their truth is linguistically illegible to the instrument.
| Feature | Kruskal-Wallis H | p (cultures differ) | Significant |
|---|---|---|---|
| Hedging Rate | 383.64 | .05), indicating a consistent effect. |
Good research designs specify disconfirmation criteria. The Signal Inversion Effect thesis would be weakened or killed if:
Any of these outcomes would require substantial revision of the thesis. None would eliminate the Phase 1 disfluency finding, which stands on its own regardless of the broader framework.
The final output is a single correlation matrix or forest plot showing the Signal Inversion Effect across domains: a consistent pattern where the cues humans use to assess credibility are negatively correlated with actual credibility across criminal justice, education, and cross-cultural belief data. One visualisation. One devastating finding. Implausible to dismiss.
# APPENDIX C: Drug Policy and the Construction of Criminality
The criminalisation of drug use provides the clearest illustration of the thesis's core argument: that criminal categories are politically constructed rather than naturally discovered.
Between 1906 and 1914, Bayer marketed heroin as a cough suppressant. Cocaine was an ingredient in Coca-Cola until 1903. Cannabis was prescribed by physicians until the 1930s. Opium was legal in Britain throughout the Industrial Revolution. None of these substances changed their pharmacological properties. What changed was the legislative category applied to them — and the populations that were using them.
The Harrison Narcotics Tax Act of 1914 in the United States was explicitly linked to racial anxieties about Chinese opium use and Black cocaine use (Musto, 1999). Australia's opium prohibition was directly connected to anti-Chinese sentiment in the goldfields (Manderson, 1993). The construction of 'drug crime' is inseparable from the construction of racialised criminal subjects.
In 2001, Portugal decriminalised personal possession and use of all drugs. This was not legalisation — supply remained criminal — but the removal of criminal penalties for personal use and the redirection of resources from criminal justice to health.
Twenty-five-year outcomes (2001–2026):
| Measure | Before (2001) | After (2020s) | Change |
|---|---|---|---|
| Drug-induced deaths | ~80/year | ~16/year | -80% |
| HIV among PWUD (new diagnoses) | 52% | 7% | -87% |
| Drug-related incarceration | — | -75% | — |
| Overall drug use | Baseline | No significant increase | Stable |
| Drug use among 15–24 year olds | Above EU average | Below EU average | Reversed |
These outcomes are not explained by reduced drug use — overall use rates did not change significantly. They are explained by the removal of the criminal justice system from the response pathway. The system that was constructing 'drug criminals' and processing them through arrest, charge, trial, and incarceration was replaced by a system that treated drug use as a health issue.
The mechanism is precisely what this thesis describes: when you remove the guilt-construction architecture, you remove the guilt it constructs. The drugs remain. The crime disappears — because the crime was never in the substance. It was in the statute.
Prohibition creates the conditions it claims to prevent. The safety paradox of drug criminalisation operates through three mechanisms:
The deaths are not caused by the drugs. They are caused by the policy.
The chemical compound sodium oxybate (GHB) provides a particularly stark illustration of the constructed character of drug criminalisation.
In Schedule 8 pharmaceutical form, sodium oxybate is marketed as Xyrem and prescribed for narcolepsy. It is a legal, FDA-approved, PBS-listed medication. Patients receive it through pharmacies, take it under medical supervision, and are protected by the full apparatus of medical regulation.
In Schedule 9 form, the identical molecule is classified as a prohibited substance. Possession is a criminal offence carrying penalties of up to 15 years imprisonment in some Australian jurisdictions.
The molecule is the same. The pharmacological effects are the same. The risk profile is the same. The difference is entirely legislative — a political construction that determines whether possessing this substance makes a person a patient or a criminal.
The person convicted of GHB possession has not been found to have done something harmful. They have been found to be in possession of a molecule that the legislature has classified as criminal when obtained outside the pharmaceutical supply chain. The harm — if any — is identical regardless of the source. The criminality is entirely constructed by the definitional category.
This is Chapter 5's argument in miniature: the criminal justice system does not respond to harm. It responds to the violation of political categories. The categories determine who is a criminal. The categories are contingent. The system processes people through an architecture that constructs guilt from the satisfaction of definitional conditions, not from the presence of harm.
The chemical compound sodium oxybate (GHB) provides a particularly stark illustration of the constructed character of drug criminalisation.
In Schedule 8 pharmaceutical form, sodium oxybate is marketed as Xyrem and prescribed for narcolepsy. It is a legal, FDA-approved, PBS-listed medication. Patients receive it through pharmacies, take it under medical supervision, and are protected by the full apparatus of medical regulation.
In Schedule 9 form, the identical molecule is classified as a prohibited substance. Possession is a criminal offence carrying penalties of up to 15 years imprisonment in some Australian jurisdictions.
The molecule is the same. The pharmacological effects are the same. The risk profile is the same. The difference is entirely legislative — a political construction that determines whether possessing this substance makes a person a patient or a criminal.
The person convicted of GHB possession has not been found to have done something harmful. They have been found to be in possession of a molecule that the legislature has classified as criminal when obtained outside the pharmaceutical supply chain. The harm — if any — is identical regardless of the source. The criminality is entirely constructed by the definitional category.
This is Chapter 5's argument in miniature: the criminal justice system does not respond to harm. It responds to the violation of political categories. The categories determine who is a criminal. The categories are contingent. The system processes people through an architecture that constructs guilt from the satisfaction of definitional conditions, not from the presence of harm.
Drug criminalisation constructs guilt from a health behaviour. The person who uses drugs and is arrested has not harmed another person. They have been defined into criminality by a legislative act — the same performative speech act described in Chapter 5.
The system then processes them through the same architecture documented in Chapters 3–8: pre-interrogation detention, interrogation, charge, trial, verdict. At every stage, the same credibility-assessment heuristics apply. The drug user who presents with anxiety, flat affect, or inconsistent recall — common features of substance use disorders — triggers the same Signal Inversion effects documented throughout this thesis.
The person convicted of drug possession was not found to have done something harmful. They were found to have satisfied the conditions of a definitional category constructed by legislators with specific political interests, interpreted by police using invalid behavioural heuristics, and adjudicated by juries applying narrative schemas about 'the kind of person who uses drugs.'
Goal 7 (legalise and regulate drugs) (legalise and regulate drugs) of the prevention framework is not a policy proposal. It is the logical consequence of the evidence reviewed in this thesis.
# APPENDIX D: The Economic Architecture of Incarceration
The criminal justice system that this thesis has demonstrated constructs guilt rather than discovers it costs Australia approximately $32 billion per year (Productivity Commission, Report on Government Services, 2023).
| Component | Annual Cost (AUD) |
|---|---|
| Police services | $14.2 billion |
| Corrective services | $6.1 billion |
| Courts and legal aid | $1.9 billion |
| Juvenile justice | $1.0 billion |
| Emergency services (overlap) | ~$9.0 billion |
| Total | ~$32 billion |
The cost per prisoner in Australia is approximately $110,000 per year — making Australian imprisonment among the most expensive in the world. This cost purchases a recidivism rate of 45–70%, meaning that nearly half of the people the system processes at $110,000 per year will return to the system.
For comparison: Norway spends approximately $93,000 per year per prisoner on a rehabilitation-focused model and achieves a recidivism rate of 20%.
The Australian system costs more per person and produces worse outcomes on every measured dimension.
The observation that the system's operational success is measured in conviction rates rather than truth production (Chapter 9) has an economic correlate. The system employs approximately 300,000 people in Australia — police officers, prosecutors, magistrates, judges, prison officers, parole officers, court administrators, lawyers.
These 300,000 livelihoods depend on the continued operation of the system that this thesis has demonstrated does not reliably distinguish guilt from innocence. This is not a conspiracy. It is an institutional incentive structure. No individual within the system needs to intend the construction of guilt for the system to construct guilt. They need only to do their jobs — jobs whose continued existence depends on the system processing people.
Goal 6 (re-employ displaced workers) (re-employ displaced workers) of the prevention framework addresses this directly: re-employ all displaced staff in functional positions. Nobody loses a livelihood. The skills transfer. The roles change. The prison officer becomes a community support worker. The detective becomes a conflict mediator. The prosecutor becomes an advocate. The institutional knowledge is preserved; the institutional architecture is replaced.
The $32 billion currently spent on the guilt-construction system would fund:
| Prevention Investment | Estimated Annual Cost | Evidence Base |
|---|---|---|
| Universal mental health access | $8 billion | Medicare-funded, OECD models |
| CAHOOTS-model community response | $2 billion | 35 years of evidence (Eugene, OR) |
| Housing First (chronic homelessness) | $3 billion | Finland, US VA studies |
| Drug treatment at Portuguese scale | $1.5 billion | 25-year national data |
| Early childhood intervention | $4 billion | Perry Preschool, Abecedarian studies |
| Community emergency response | $0.5 billion | Hatzolah, surf lifesaving |
| Total prevention investment | $19 billion | — |
| Remainder (returned to citizens) | $13 billion | — |
The prevention model costs $13 billion less than the system it replaces and produces:
The Perry Preschool Program (Schweinhart et al., 2005) provides the longest longitudinal evidence for the economic returns of prevention over punishment.
Beginning in 1962, 123 low-income African American 3- and 4-year-olds in Ypsilanti, Michigan, were randomly assigned to receive high-quality preschool or no preschool. Participants have been followed for over 50 years.
By age 40:
| Outcome | Programme Group | No-Programme Group |
|---|---|---|
| Arrested 5+ times | 12% | 49% |
| Earning $20,000+/year | 60% | 40% |
| Graduated high school | 77% | 60% |
| Homeowner | 37% | 28% |
| Monthly earnings | $1,856 | $1,308 |
Return on investment: Every $1 invested returned $7.16 to society — primarily through reduced criminal justice costs and increased tax revenue.
The critical insight: the programme did not provide criminal justice intervention. It provided high-quality early childhood education — play, curiosity, social interaction, responsive adult attention. It prevented crime by creating the conditions in which crime does not occur.
This is the prevention architecture that the criminal justice system's resources could fund. The system currently spends $110,000 per year per prisoner to warehouse people in conditions that increase recidivism. The same resources, redirected to prevention, would produce fewer crimes, fewer victims, and fewer people processed through a system that cannot reliably determine whether they are guilty.
# APPENDIX E: Companion Research and Cross-References
This thesis draws on a body of companion research that extends the analysis into domains not fully addressed in the main text. The following summaries indicate the scope and key findings of each companion study.
Location: Constructed Guilt Signal Inversion (2026). Preprint.
The Signal Inversion Effect paper provides the detailed statistical foundation for the claims made in Chapter 2, Section 2.8. The paper analyses 23 behavioural cues used in credibility assessment and finds that 21 (91.3%) are either empirically unrelated to deception or directionally inverted — meaning the behaviour that observers interpret as deceptive is more strongly associated with truthful communication.
The paper includes a forest plot comparing believed vs. actual deception cues, a belief-reality scatter plot, and a weighted inversion index showing that the cues most strongly endorsed as deception indicators are the most inverted.
Key statistic: 91.3% inversion rate (p