Artefaktum initiated this research following its recent involvement in the resolution of a complex international fraud case. During the investigation, our experts encountered patterns of deception that raised a broader question extending well beyond the individual case: what happens when a false narrative is repeated, refined and defended over an extended period of time? The case provided practical reason to examine whether prolonged and repeatedly rehearsed deception can alter not only the way a person communicates a falsehood, but also the cognitive and physiological processes associated with producing it. Of particular interest was the possibility that familiarity, repetition, memory reconstruction and gradual psychological adaptation may reduce some of the responses traditionally associated with deceptive behaviour. These observations formed the starting point for the present research into adaptive deception, memory, neuroplasticity and the potential consequences for physiological and ocular deception detection.
Deception detection has traditionally rested on a simple and intuitively convincing assumption: lying creates a disturbance. The deceptive individual knows one thing and communicates another, and somewhere between knowledge, intention, inhibition and expression this conflict is expected to leave a measurable trace. Such traces may include autonomic arousal, changes in respiration, electrodermal activity, cardiovascular response, delayed reaction time, pupil dilation, altered fixation behaviour or changes in cortical activation. For more than a century, technologies have attempted to capture these signals. Yet the underlying model contains an assumption that deserves far greater attention: that the physiological and cognitive disturbance associated with deception remains sufficiently stable over time to remain detectable. The question, therefore, is what happens if it does not.
Research into dishonesty, memory, habituation and cognition increasingly suggests that lying is not a fixed cognitive operation. Deception can change with repetition. A lie can be rehearsed, a response can become faster, emotional reactions may diminish, and fabricated information can interfere with memory for the original event. Repeated retrieval can strengthen some representations while weakening or distorting others. False denials can affect source memory, and fabricated narratives can become increasingly familiar. Under certain conditions, individuals may even become uncertain whether particular information originated in actual experience, imagination or earlier fabrication. The implication is important. The relevant question for credibility assessment may no longer be simply whether a person is lying, but what stage of its cognitive development the lie has reached. A spontaneous lie told seconds after an event may be neurologically, cognitively and physiologically very different from a narrative that has been repeated for years, refined under scrutiny, embedded in autobiographical memory and eventually treated by the speaker as substantially true. If that proposition is correct, deception detection faces a problem more fundamental than the search for a better sensor because the target itself may be changing.
Research already provides important evidence for this. Hu and colleagues showed in reaction-time experiments that deceptive responses could become substantially more efficient after training. Following specific practice, the reaction-time difference between deceptive and truthful responses disappeared. Their findings suggested that processes associated with deception are malleable and that behaviour initially requiring greater cognitive control can become more automatic. Garrett and colleagues approached the question from another angle. Their work on repeated self-serving dishonesty showed that dishonest behaviour could escalate while amygdala responses associated with repeated dishonesty decreased. This does not mean that the brain simply stops responding to lies. The evidence supports a more careful conclusion: neural responses associated with repeated dishonest behaviour can change as the behaviour becomes familiar.
Memory research introduces another layer. A growing body of work suggests that deception can itself alter subsequent recollection. False denial, fabrication and feigned amnesia have been associated with source-monitoring errors, denial-induced forgetting, false recognition and fabrication-related distortion. Different forms of deception appear to alter memory in different ways. Eye-tracking research adds further complexity. Pupil dilation, fixation behaviour and other ocular measures have been associated with deceptive responding or concealed recognition under controlled experimental conditions. Yet pupil size does not measure lying directly. It reflects processes such as cognitive effort, attention, autonomic activation and arousal, all of which can change through habituation and repetition. A reduced physiological response does not therefore necessarily mean deception has disappeared. It may indicate that the cognitive process producing the response has changed.
A recent Scientific Reports study using eye tracking in a Concealed Information Test reported very high classification accuracy in a small mock-crime sample and compared favourably with polygraph assessment under the specific study conditions. At the same time, the researchers emphasised the limitations of the findings: small sample size, absence of genuine real-world consequences, methodological differences between approaches and the need for further work on external validity, countermeasures and specificity. The overall picture is therefore neither a validation nor a rejection of physiological deception detection. It is more complicated. Deception may produce measurable effects, but those effects arise from several psychological mechanisms, and those mechanisms do not necessarily adapt at the same rate. The detectable signature of deception may therefore migrate over time.
This has particular relevance in intelligence, counterintelligence, insider-threat investigation, security screening, criminal interviewing and forensic credibility assessment. Subjects in these environments are not always naive individuals improvising a first falsehood. They may have rehearsed cover stories, repeated denials, compartmentalised memories, rationalised misconduct or maintained false identities over long periods. The experienced deceiver may therefore not simply be a better liar. He or she may represent a different psychophysiological problem.
One reason is that there is no universally accepted physiological signature that appears every time a person lies. Polygraphs measure respiration, cardiovascular activity and electrodermal response. Eye-tracking systems measure pupil diameter, gaze behaviour and fixation patterns. EEG-based approaches may examine P300 responses, while functional neuroimaging investigates patterns of brain activation. None of these technologies measures falsehood itself. They measure processes that can accompany deception, and that distinction defines their limitations. A frightened innocent person may show strong autonomic responses. A highly motivated liar may do the same, while a calm liar may show relatively little. A truthful person recalling traumatic material may appear more physiologically activated than someone calmly repeating a rehearsed cover story. A person who believes a reconstructed false memory may experience little of the cognitive conflict normally associated with deliberate deception.
Most candidate indicators are therefore not deception-specific. They reflect conflict, inhibition, emotional salience, fear of detection, cognitive effort, recognition, surprise, orienting and social evaluation. A more useful model is to understand deception as a temporary configuration of these processes rather than a single neurological event. When a person lies deliberately, the truthful representation may become available first. It may then need to be inhibited, replaced or modified. The alternative account must be retrieved or constructed, checked for consistency and delivered while the speaker monitors the situation and assesses possible consequences. None of these processes, however, necessarily remains equally demanding after repetition.
This is where rehearsal becomes important. One of the foundations of cognitive lie detection is the observation that lying is often more demanding than truth telling. Truth generally has a processing advantage because a remembered event may be retrieved relatively automatically. Lying can require suppression of that response and substitution of another, creating additional cognitive load. But the statement that lying is often cognitively demanding should never be converted into the claim that lying is always cognitively demanding. Practice matters, rehearsal matters, preparation matters, and the nature of the lie matters. Repeated cognitive operations are subject to learning. Tasks initially requiring substantial executive control can become more efficient as stimulus-response associations strengthen. Retrieval becomes easier, competing responses are inhibited more quickly, and behaviour can develop into a routine.
A rehearsed deception can therefore move from construction toward retrieval. Consider a false alibi created under pressure. At first, the person must invent locations, times and activities while avoiding contradictions. After extensive rehearsal, however, the task may change. The person may no longer be constructing the answer but simply retrieving it. The lie remains false, but the cognitive pathway used to produce it has changed. This matters because many deception-detection approaches depend on the assumption that deceptive responses impose a stable cognitive cost. Research comparing spontaneous and rehearsed lies has in fact found different neural activation patterns, reinforcing the conclusion that deception is not neurologically uniform.
A useful operational concept is therefore deception automaticity. A lie may exist somewhere along a continuum between improvised and automated. At one end, the individual constructs and monitors a novel falsehood. At the other, a highly rehearsed response may be triggered almost immediately by a familiar question. The more a detection method depends on reaction time, effort or conflict, the more important this distinction becomes.
Emotional adaptation creates a similar problem. Repeated dishonesty may not carry the same emotional meaning indefinitely. Research into self-serving dishonesty suggests that emotional responses associated with transgression can adapt. The first serious act may feel highly consequential, while repetition can reduce novelty, uncertainty and internal resistance. This does not mean habitual liars become emotionless or physiologically invisible. A more defensible interpretation is that one source of deception-related signal, the emotional discomfort associated with the act itself, may weaken.
This leads to an important distinction: fear of lying is not the same as fear of being caught. A habitual deceiver may experience little internal discomfort when repeating the lie but still react strongly to the possibility of exposure. A professional operative may experience minimal moral conflict while remaining highly alert to operational risk. Another person may feel guilt but little fear, while another may have rationalised the behaviour completely. The spoken falsehood may be identical while the underlying autonomic state is different. The signal may therefore not disappear; it may move.
An unrehearsed lie may initially generate emotional discomfort, executive inhibition and narrative construction. With repetition, construction may disappear because the answer has been memorised, inhibition may become more efficient and moral discomfort may decline through habituation or rationalisation. A detector optimised for those processes may become less sensitive. But recognition of incriminating information may remain. Working memory may still be activated when unexpected details appear. Conflict may return when new evidence contradicts the rehearsed narrative, and anticipatory arousal may increase when a dangerous topic is introduced. The deception signal has not necessarily vanished. It may simply be located elsewhere.
This has direct implications for interviewing. Repeated questions can produce trained answers. Static questioning is therefore especially vulnerable to adaptation, while dynamic questioning changes the task. A rehearsed answer to “Where were you at 22:00?” may require almost no construction. Asking what the person could see immediately to the left on entering the building, or why one entrance was chosen instead of another, may force the subject back into active cognitive processing. The objective is not aggression but diagnostic novelty. The investigative opportunity shifts away from repeatedly testing the lie itself and toward testing the limits of the cognitive model behind it.
Memory makes the problem even more complex because human memory is reconstructive. It is not a fixed digital recording. Retrieval is active and can alter what is remembered. Repeated recall can strengthen memory, but it can also transform it. The conventional image of a liar assumes two permanently separate representations: the true event and the false account. In reality, those representations may begin to interact. Fabricated details become familiar, sources may be confused, and the person may remember having said something without remembering clearly whether it was experienced or later constructed.
Research has shown that false denial, fabrication and feigned amnesia affect memory differently. False denial may reduce rehearsal of denied material. Fabrication can create source-monitoring problems because invented information later becomes familiar. Feigned amnesia can change later retrieval because the person repeatedly practises withholding certain information. For investigators, the distinction is important. A contradiction discovered months later may reflect deliberate continuing deception, but it may also reflect genuine memory distortion caused partly by earlier deception. The historical account remains false, yet the individual’s present psychological relationship with it may have changed.
Repeated fabrication may also increase familiarity with invented autobiographical information and, under some conditions, influence belief in that information. A deceptive narrative can therefore evolve. At first, the person knows one proposition to be true and deliberately reports another. With repetition, the fabricated version becomes increasingly accessible. As time passes, memory for the original event may weaken while the narrative structure supporting the false account strengthens. Eventually, the rehearsed representation may become the most easily available version of what happened. At that point, a detector designed to identify deliberate deception encounters a difficult problem: what exactly should it be detecting?
Self-deception and rationalisation make this even more significant. Human beings have strong incentives to maintain a coherent and morally acceptable image of themselves. Responsibility can be minimised, harm reframed, intentions reconstructed and conduct justified as necessary. The most difficult liar to detect may therefore not be someone with extraordinary voluntary control over physiological responses. It may be someone who has gradually changed the psychological meaning of the act. If the individual no longer experiences the statement as a serious moral violation, one potential source of physiological differentiation becomes weaker. If the false account has partly entered genuine belief, cognitive conflict may also decline. Historical reality has not changed, but the current cognitive state has.
That creates a fundamental problem because credibility technologies examine present cognitive state, while investigators are interested in historical reality. A useful conceptual model is therefore to think of deception as developing through stages: initial construction, stabilisation through repetition, increasing automation, integration into broader memory, rationalisation and, in some cases, partial assimilation into subjective belief. This model is theoretical. Existing evidence does not show that every lie develops in this way or follows a fixed sequence, but it is nevertheless useful for explaining why detectability may change over time.
Ocular measures illustrate the point particularly well. The pupil responds not only to light but also to cognitive effort, attention and autonomic activation. Deceptive responses can therefore produce increased pupil dilation when additional processing or emotional engagement is required. Yet pupil dilation is not itself a lie response. It reflects processes that may be present during lying. This makes pupillometry attractive and difficult at the same time. It gives access to relatively involuntary processes, but those processes are not unique to deception. If rehearsal reduces cognitive effort, one source of pupil dilation may decline. If emotional habituation reduces arousal, another may decline. Yet concealed recognition may remain salient even when the verbal response itself has become automatic.
This is where the distinction between deception detection and recognition detection becomes critical. A person may become extremely efficient at saying something false without significant measurable effort. That does not necessarily mean genuinely familiar information can be made unfamiliar at will. The Concealed Information Test is built around this distinction. Rather than attempting directly to identify lying, it tries to determine whether a person recognises information they should not know. A person may have repeated “I have never been inside that room” hundreds of times, making the response almost automatic. But if the person is shown the actual stolen object among several alternatives, recognition may still trigger an orienting response. The lie has become practised, while the memory may remain.
Recognition-based approaches therefore offer an important alternative, but they have limitations of their own. Memory traces can weaken, relevant information may be learned innocently, investigators can contaminate a subject by revealing details, and media reporting can make supposedly concealed facts widely known. The usefulness of recognition-based assessment therefore depends on both memory integrity and the secrecy of the information being tested.
The same caution applies to pupil habituation. Evidence that pupil responses diminish after repeated presentation does not automatically mean ocular detection becomes useless. Absolute response magnitude and diagnostic contrast are different. If both relevant and irrelevant responses decline while the difference between them remains, useful information may survive. If deception-related responses decline faster, sensitivity may deteriorate. The key issue is differential habituation. Existing research still lacks sufficient longitudinal evidence over the periods most relevant to real-world deception. Many experiments last minutes or hours, while operational deception may continue for years.
This gap becomes especially important with professional deceivers. Much experimental research necessarily relies on volunteers participating in mock crimes or instructed lies. Such work is valuable but has obvious ecological limits. Participants usually face no genuine consequences and have not spent years building a cover identity, maintaining supporting documentation or incorporating real experiences into fabricated biography. A trained operative under cover may live within an invented identity for so long that parts of that identity become genuine experience. A cover workplace may become the place the person actually visits every day. The answer to “Where do you work?” may no longer feel internally like a fabricated response even though the broader identity remains false. The boundary between fabrication and lived reality becomes much more complex than laboratory paradigms suggest.
Insider-threat cases demonstrate a similar progression. The first unauthorised act may produce substantial anxiety. Repeated success can reduce uncertainty, increase confidence and encourage rationalisation until the conduct becomes psychologically routine. This is why nervousness or visible behavioural disruption should never be treated as reliable indicators of continuing misconduct. An established insider may appear less anomalous than a first-time offender. Habituation also does not mean that stakes cease to matter. A person may repeat the same false story calmly for years and still react strongly when confronted with evidence capable of destroying it. Narrative familiarity is not the same as situational familiarity. The same rehearsed sentence may therefore produce different physiological responses when spoken to a friend, an employer, an investigator or a court. A subject can become habituated to the lie while remaining highly sensitive to exposure.
For this reason, familiarity with the narrative, familiarity with the stimulus and familiarity with the test environment should be distinguished. They can interact, but they are not the same. Practice can also create new vulnerabilities. A rehearsed liar may become faster and more consistent, but every repeated statement creates another record that must remain compatible with external reality. Every fabricated detail increases the number of dependencies within the narrative. Long-lived lies eventually collide with metadata, access logs, financial records, travel data or third-party testimony. Cognitive automation may therefore make a liar harder to detect physiologically while making the story easier to challenge evidentially.
That is a crucial operational shift. The answer to adaptive deception may not be a more sensitive sensor. It may be to move from state detection to model testing. Instead of asking whether the person appears deceptive, investigators can test whether the account makes correct predictions about independently verifiable reality. A truthful autobiographical representation normally contains sensory, spatial, contextual and causal relationships derived from experience. A fabricated narrative can reproduce many of these features, especially after rehearsal, but it may remain weaker around peripheral facts that were never anticipated during construction. Investigators should therefore often probe the edges of the narrative rather than repeatedly attacking its polished centre, because the polished centre is exactly where the experienced liar is strongest.
Repeated deception can also produce what might be called retrieval inversion. At first, the truth may be easier to retrieve than the lie. After extensive rehearsal, however, the fabricated answer can become the dominant response to a familiar cue. A person repeatedly asked whether documents were transferred may eventually answer “no” almost automatically. Change the question, however, and the automatic mapping may fail. Ask what folder was opened immediately before the transfer or what appeared on the screen, and the subject may again have to assess what the interviewer knows, evaluate the premise and construct a new response. Cognitive load returns. The practical consequence is straightforward: investigators should not continue testing the same answer simply because it appears important. A well-rehearsed response may become increasingly stable through repetition, while the surrounding factual structure remains much harder to control. The more useful approach is therefore to examine whether the account is consistent with the wider reality it describes.
The same principle can be understood through prediction error. A rehearsed deceiver develops expectations about questioning, and as long as the interaction remains predictable, processing is efficient. Unexpected evidence changes this. A photograph that should not exist, a timestamp that contradicts the alibi or a question revealing knowledge the subject did not expect the interviewer to possess can create sudden informational significance. From this perspective, the future of credibility assessment may depend less on finding a universal signature of lying and more on identifying moments when the subject’s internal model encounters something it did not predict. This is also why stimulus design matters as much as sensor quality. The most advanced eye tracker cannot compensate for poor questions.
Artificial intelligence does not remove this underlying problem. Machine learning can combine multiple features, but if none of them uniquely measures deception, combining them does not create a direct measurement of falsehood. The system identifies statistical relationships within its training data. If those relationships change with experience, rehearsal or memory reconstruction, the model faces a form of concept drift. A classifier trained mainly on first-time deception may encounter experienced subjects whose physiological profile differs significantly from the training population, yet it may still produce a confident classification. Confidence is not proof of validity.
For security professionals, the deeper problem is familiar: the person being assessed is an adaptive adversary. Spam filters face spammers who change behaviour, malware detection faces attackers who change code, and fraud controls face offenders who learn procedures. Deception detection operates in a similarly adaptive environment. Future systems should therefore be evaluated against informed and experienced subjects, not only naive participants. It is equally important to distinguish intentional countermeasures from natural adaptation. A countermeasure is a deliberate attempt to interfere with detection. Habituation, learning, memory reconstruction and self-deception can occur without any intention to beat the test. The practical outcome may look similar because diagnostic differentiation decreases, but the mechanisms are different. A subject using a deliberate countermeasure may produce secondary signs of manipulation. Someone who has simply rehearsed the lie until it requires less effort may not. The person has not necessarily defeated the machine; the cognitive process the machine expected may simply no longer exist in the same form.
No single technology therefore answers every relevant question. A professional assessment should distinguish between whether a person shows unusual processing while answering, whether they recognise information they claim not to know and whether their narrative corresponds with independently verifiable evidence. These are different questions, and the strongest investigations combine them. This is why the phrase “lie detection” may itself be limiting. A more useful concept is cognitive forensics. Cognitive forensics would examine recognition, memory accessibility, cognitive effort, orienting, prediction error, narrative structure and response consistency. Physiological measurement would become one source of evidence rather than a verdict. Eye tracking may reveal disproportionate attention to supposedly unfamiliar information, reaction time may show unusual processing, a P300 response may indicate recognition, strategic interviewing may expose the limits of a rehearsed account, and digital evidence may contradict the narrative. None proves deception independently, but together they can form a stronger inferential structure.
The relevance to HUMINT, counterintelligence and insider-threat investigations is clear. Professional deceivers may mix truthful and false elements, maintain elaborate legends and embed genuine lived experience inside fabricated biography. Such narratives cannot always be separated neatly into truth and lie. The objective should therefore be to understand how the account is built. Which elements appear grounded in genuine episodic memory? Which have been rehearsed? Which can be verified independently? Where is knowledge unusually precise? Where does specificity suddenly disappear? Which parts become unstable when perspective or chronology changes? The aim is not to identify the instant when someone “looks like a liar.” It is to understand the architecture of the narrative.
For intelligence services, police, prosecutors and corporate or governmental security officers, these findings have consequences that go considerably beyond the question of whether a particular detection technology works. They affect how investigations are structured, how sources are assessed and managed, and how interviews or interrogations should be interpreted. Apparent calmness should never be confused with credibility. An experienced fraudster, intelligence source, insider or suspect may have repeated the same account so often that its delivery no longer produces the hesitation, cognitive effort or emotional disturbance an investigator might expect from deception. Conversely, a truthful person confronted with serious accusations may show considerable stress. Behavioural confidence, physiological stability and factual accuracy are therefore different variables and should be treated as such.
For police and criminal investigators this means that the centre of gravity should gradually move away from asking whether a person “looks deceptive” and toward testing the explanatory model contained in the statement. A statement should create consequences that can be checked. Locations should correspond with records, timelines should fit digital and physical evidence, claimed actions should leave expected traces, and relationships and communications should be capable of independent confirmation. The more established and rehearsed a deceptive narrative has become, the more valuable independent corroboration becomes. This is equally important for prosecutors. A witness or suspect who repeats an inaccurate account consistently over several interviews is not necessarily becoming more reliable. Consistency can result from genuine memory, but it can also result from rehearsal. Repetition itself can stabilise a narrative. Investigators and prosecutors should therefore be cautious about treating consistency alone as evidence of truthfulness, particularly when the same narrative has been told repeatedly before formal examination.
At the same time, inconsistencies should not automatically be interpreted as evidence of lying. Human memory is reconstructive, and genuine recollection can vary, particularly concerning peripheral details. The investigative value lies not simply in identifying contradiction but in understanding what kind of contradiction has occurred and whether it affects a central, independently verifiable element of the account.
For intelligence services the implications may be even more significant because source relationships can continue for months or years. A human source may repeatedly provide the same biographical explanation, motivation, access story or account of past events. Over time, both the source and the handler can become familiar with that narrative. Familiarity creates its own risk because what initially required verification can gradually become accepted background knowledge simply because it has been heard repeatedly. Source validation should therefore remain a continuous process rather than an event conducted only at recruitment. The relevant questions are not limited to whether the source appears credible or whether reporting has previously been accurate. Analysts and handlers should continue to examine whether claimed access remains plausible, whether reporting can be independently corroborated, whether the source’s explanations evolve when new facts emerge, and whether apparent consistency reflects genuine access or simply a well-established narrative.
This becomes particularly important when a source mixes truth and deception. Professional deception rarely requires an entire story to be false. A source can provide large amounts of accurate information while concealing one relationship, intention, affiliation or access limitation. Indeed, truthful material can make the critical false element substantially more convincing. Reliability should therefore be assessed at the level of individual claims and access pathways, not only at the level of the person.
The same principle applies to counterintelligence. A long-established cover story may be cognitively very different from an improvised lie. A person who has lived within a cover identity for years may respond to routine biographical questions without measurable hesitation because much of the surrounding life has become genuinely experienced. Attempting to detect deception solely within the rehearsed centre of such a narrative may therefore be less productive than examining its boundaries: the parts where claimed biography intersects with records, other people, travel, communications, chronology and events that were not anticipated when the original narrative was constructed.
For source handling and HUMINT interviewing, this also argues against repeatedly asking essentially the same verification questions. Repetition may strengthen the very response pattern the interviewer hopes to examine. More information can often be gained by approaching the same subject through different retrieval routes, returning to relevant events from another temporal or spatial perspective, or examining unexpected but independently verifiable peripheral details. This does not mean creating artificial cognitive stress. Excessive pressure can damage truthful recall and contaminate the information being collected. The objective is not to make the subject uncomfortable enough to reveal deception. It is to create opportunities to distinguish a rehearsed verbal representation from a richer underlying memory and to generate information that can subsequently be tested.
For police interviews and interrogations, the practical consequence follows the same logic. The primary objective should be accurate information gathering rather than obtaining behavioural confirmation of an investigator’s existing suspicion. Rapport-based, non-coercive and evidence-led interviewing is particularly relevant because it allows the account to develop before its internal and external structure is tested. Unexpected evidence can then be assessed against what the interviewee has already said rather than immediately becoming part of the narrative the person is trying to explain. The timing of evidence disclosure can therefore matter. If all available evidence is revealed at the beginning, a deceptive subject may simply incorporate it into an adapted account. If the investigator first obtains a sufficiently detailed account, later comparison with independently held evidence can reveal where the narrative corresponds with reality and where it requires revision.
For security officers and those responsible for internal investigations, the research carries another warning. Long-running insider behaviour may become increasingly normal to the person performing it. An employee who initially experiences substantial anxiety when violating a security rule may display far fewer behavioural indicators after performing the same action repeatedly without consequence. The absence of visible nervousness should therefore carry very little evidential weight. Internal investigations should instead combine interviews with access records, communication data, document histories, physical-security information, system logs, authorisation records and other independently available evidence. Behavioural and physiological observations may contribute context, but they should not replace evidential reconstruction.
There is also an important lesson for the sequencing of investigations. When possible, investigators should preserve information that only the person responsible is likely to know. Once such information has been disclosed through interviews, media reporting, internal circulation or poorly controlled questioning, its later value for recognition-based assessment is reduced. Information discipline therefore matters not only for operational secrecy but potentially for the evidential value of subsequent interviewing.
Across intelligence, policing, prosecution and security investigations, the resulting principle is remarkably similar: the professional objective should not be to identify a behavioural moment that supposedly proves deception. It should be to understand the relationship between the person’s current cognitive representation and independently observable reality. That requires disciplined interviewing, evidence preservation, corroboration, source evaluation and continuous hypothesis testing. A physiological reaction may be interesting, an ocular response may indicate recognition, a hesitation may suggest increased processing and a contradiction may deserve investigation, but none of these, by itself, establishes the historical truth.
For investigators and intelligence professionals, the more useful question may therefore not be how to make the lie visible, but what would have to be true in the real world if the person’s account were correct. Once that question becomes the organising principle of the investigation, adaptive deception becomes less a contest between an interviewer and a liar and more a systematic comparison between narrative and reality.
This also leads to an important paradox. A lie that has been repeated successfully over a long period may gradually lose some of the psychological characteristics normally associated with deception. Familiarity can increase fluency, repeated justification can reduce moral discomfort, and memory reconstruction may make the original event less accessible while the fabricated account becomes easier to retrieve. Social reinforcement can strengthen this process further, particularly when the narrative has been accepted or repeated by others. As a result, a false account may eventually be delivered with considerable confidence and emotional stability. For a detection method that relies heavily on conflict, hesitation or physiological arousal, this can reduce the strength of the observable signal. It does not, however, change the underlying historical facts. A statement can become easier to tell, easier to remember and even easier to believe without becoming more accurate.
The proposition that repeated lying eventually suppresses physiological deception indicators remains plausible but has not been conclusively proven. Establishing it properly would require longitudinal research that compares rehearsed and unrehearsed lies, repeated truthful memories and multiple physiological, ocular and behavioural measures across extended periods. Researchers would also need to distinguish repetition of the lie from repetition of the stimulus and repetition of the test itself. Memory and subjective belief would need to be measured in parallel. Otherwise a reduced physiological response could reflect habituation, memory transformation or both. Novel questioning would then be necessary to determine whether apparent habituation survives once familiar response patterns are disrupted.
These studies would be difficult, but they are essential if deception technologies are to be evaluated seriously against experienced, high-stakes subjects. The research question should therefore change. Instead of asking only which physiological indicators distinguish lying from truth telling, researchers should ask how those indicators change with repetition, which survive rehearsal, whether recognition persists after responses become automatic, how fabrication alters memory, whether novel questioning restores cognitive differentiation, and whether algorithms trained on spontaneous deception generalise to long-term deception.
The Artefaktum assessment is therefore not that physiological deception detection does not work, nor that experienced liars can reliably defeat such systems. The evidence does not support either categorical conclusion. The stronger and more defensible proposition is that deception should be understood as an adaptive process rather than a fixed state. The person’s history of lying matters, the narrative’s rehearsal history matters, memory retrieval matters, emotional adaptation matters, testing history matters, perceived consequences matter and the degree of self-belief matters. The detector is not confronting a static object. It is confronting a cognitive system shaped by previous behaviour.
The final distinction is perhaps the most important: detecting deception is not the same as discovering truth. A person can lie about a true memory. A person can truthfully report a false memory. Someone can conceal information while making factually accurate statements. Incorrect information can be reported without intent to deceive. A deliberate lie can later become partly believed, while recognition of incriminating information can have innocent explanations. These states cannot be reduced to a simple binary variable. Credibility assessment should therefore remain part of investigative reasoning rather than being treated as an autonomous truth machine. A physiological response can narrow uncertainty, an interview can generate propositions for testing, and recognition measures can identify information that deserves further investigation. Historical truth still has to be established through evidence.
Research does indicate that deception is malleable. Training can make deceptive responses faster and more automatic. Repeated dishonesty can alter neural response. Different forms of lying can affect memory, belief and recall. Ocular measures can distinguish deceptive or concealed-information conditions in controlled experiments, while pupil responses themselves can habituate. What has not been demonstrated is a universal rule according to which repeated liars eventually cease producing all physiological indicators of deception. That would go beyond the evidence.
The more defensible conclusion is that the processes generating deception-related cues are plastic and do not necessarily adapt together. Emotional arousal may decline while recognition persists. Cognitive effort may decrease while memory conflict remains. A rehearsed answer may become automatic while novel questioning restores processing demand. Pupil-response amplitude may habituate while differences between relevant and irrelevant stimuli remain. Eventually, the person’s relationship with the lie may change enough that the detector is no longer confronting the same cognitive phenomenon that existed when the deception began.
The future challenge is therefore not simply to build a more sensitive lie detector. It is to understand the evolving architecture of deceptive cognition. For intelligence, security and forensic professionals, the operational consequence is significant. The experienced deceiver should not be regarded merely as an inexperienced liar who has learned to remain calm. The individual may have learned a different cognitive task. The narrative may have moved from fabrication to retrieval. Its emotional meaning may have changed. Memory may have been altered. The lie may have become embedded within autobiographical identity. The physiological signal may then migrate away from the act of lying itself and toward the moments when the constructed reality encounters something it cannot predict.
The most difficult deception may ultimately be neither the spontaneous falsehood nor the expertly controlled lie. It may be the lie that has survived long enough to become cognitively ordinary. Once that happens, the central question changes. It is no longer simply whether the person is lying, but what the person now believes, what they still recognise, how that representation was formed, and where their internal model ceases to match reality. That is a harder question, but it is also a much better one.
Garrett, N., Lazzaro, S. C., Ariely, D., & Sharot, T. “The Brain Adapts to Dishonesty.” Nature Neuroscience, 19, 1727-1732, 2016.
Hu, X., Chen, H., & Fu, G. “A Repeated Lie Becomes a Truth? The Effect of Intentional Control and Training on Deception.” Frontiers in Psychology, 2012.
Ganis, G., Kosslyn, S. M., Stose, S., Thompson, W. L., & Yurgelun-Todd, D. A. “Neural Correlates of Different Types of Deception: An fMRI Investigation.” Cerebral Cortex, 2003.
Battista, F., & Otgaar, H. “Research on the Effects of Lying on Memory: A Scientometric Analysis and a Call for New Studies.” Frontiers in Psychology, 2022.
Li, Y., et al. “More Lies Lead to More Memory Impairments in Daily Life.” 2022.
Polage, D. “The Effect of Telling Lies on Belief in the Truth.”
Mei, D., et al. “Self-deception: Distorted Metacognitive Process in Ambiguous Moral Decision-Making.” Human Brain Mapping, 2022/2023.
Meijer, E. H., Klein Selle, N., Elber, L., & Ben-Shakhar, G. “Memory Detection with the Concealed Information Test: A Meta-Analysis of Skin Conductance, Respiration, Heart Rate, and P300 Data.” Psychophysiology, 2014.
Cook, A. E., Hacker, D. J., Webb, A. K., Osher, D., Kristjansson, S. D., Woltz, D. J., & Kircher, J. C. “Lyin’ Eyes: Ocular-Motor Measures of Reading Reveal Deception.” 2012.
Kuhlman, B. B., Webb, A. K., Patnaik, P., Cook, A. E., Woltz, D. J., Hacker, D. J., & Kircher, J. C. “Evoked Pupil Responses Habituate During an Oculomotor Test for Deception.” Society for Psychophysiological Research, 2011.
Fang, X., et al. “Assessing Deception in Questionnaire Surveys With Eye-Tracking.” Frontiers in Psychology, 2021.
Celniak, W., Slapczynska, D., Jarszak, B., & Augustyniak, P. “Evaluating Eye-Tracking as a Method for Deception Detection Against the Polygraph as Deployed in Practice.” Scientific Reports, 16, 25352, 2026.
National Research Council. The Polygraph and Lie Detection. National Academies Press, 2003.
The overall evidential picture supports the central proposition only in its qualified form: deceptive cognition is demonstrably malleable, several physiological and neural processes relevant to deception are capable of adaptation or habituation, and deception can alter subsequent memory. Whether long-term habituation systematically enables experienced individuals to evade particular physiological or ocular deception-detection systems remains an open empirical question rather than an established scientific fact.
