RESEARCH DOSSIER
Nonconscious Information Processing
Abstract. This fundamental research monograph systematically examines how perceptual, semantic, memory-related, and action-relevant information can influence measurable behavior without corresponding conscious report. The study reconstructs the history of research, defines nonconscious processing in relation to awareness, attention, automaticity, and implicit learning, and critically evaluates masking, priming, blindsight, continuous flash suppression, confidence and metacognitive measures, and neuroscience. Primary experimental investigations and methodological reviews are examined alongside competing global-access, recurrent-processing, and high-level unconscious cognition accounts. The evidence establishes task-dependent nonconscious effects while delimiting their duration, representational scope, generalization, and measurement uncertainty. Special attention is given to direct-versus-indirect dissociations, false negatives in awareness testing, statistical power, alternative mechanisms, and the disputed breadth of unconscious deliberation. The synthesis identifies priorities for causal and reproducible research and considers carefully bounded cross-system comparison in the Artificial Era. This is a critical narrative research synthesis; it reports no new empirical experiments or independent psychological theory.
Keywords: nonconscious information processing; unconscious cognition; subliminal perception; masked priming; visual masking; implicit memory; implicit learning; blindsight; continuous flash suppression; conscious awareness; signal detection theory; metacognition; semantic priming; global neuronal workspace; recurrent processing; unconscious decision making; experimental psychology; Artificial Era
Introduction: The Scientific Object of Nonconscious Processing
Nonconscious information processing denotes transformations of sensory, cognitive, affective, or action-relevant information that influence measurable activity without the relevant contents being consciously accessible at the time of processing. Its scientific identification requires a specified process, a specified criterion of conscious access, and a measurable consequence. A participant can categorize an input, become biased toward a response, or learn a regularity while reporting little or no corresponding experience. The appropriate inference depends on the quality of that report and on alternative sources of information. The field therefore studies a family of experimentally separable processes rather than one hidden faculty with unlimited powers Kihlstrom (1987) Merikle et al. (2001).
The research questions concern how far stimulus properties propagate when conscious identification is absent, how their effects are maintained, which forms of integration and selection remain possible, and what differs between conscious and nonconscious versions of an otherwise comparable task. These questions demand behavioral observations, subjective ratings, neural recordings, and explicit models that link an intervention to a result. A masked-prime effect indicates that some stimulus-related information influenced the target response. By itself it leaves open whether activation was semantic, motoric, perceptual, decisional, or carried through residual awareness. The explanatory ambition grows only as these alternatives are experimentally separated Kouider & Dehaene (2007) Van den Bussche et al. (2009).
This monograph critically synthesizes historical experiments, contemporary consciousness research, measurement theory, and disputed claims about unconscious reasoning, learning, and decision making. Its source hierarchy gives particular weight to experimental designs with appropriately measured awareness, independent comparisons, and reviews addressing contradictory evidence. The expression nonconscious is used when a process can be established without asserting a permanent inaccessibility of its representations. Unconscious is retained when a source uses the term or proposes a stronger theoretical claim. The synthesis distinguishes empirical effects from interpretations of what consciousness contributes, and treats clinical observations as mechanistic case evidence rather than universal descriptions of human perception Dehaene et al. (2006) Newell & Shanks (2014).
Historical Origins and Experimental Transformation
Studies of processing outside awareness have repeatedly changed with the methods used to define awareness. Earlier traditions emphasized introspection, divided attention, subliminal stimulation, and dissociations between reported experience and action. Historical claims cannot be judged by vocabulary alone because the practical meaning of detection, identification, and voluntary report differed across laboratories. Holender's 1986 assessment of dichotic listening, parafoveal perception, and visual masking questioned whether purportedly unconscious semantic activation had truly excluded conscious identification. Its enduring methodological lesson concerns the dependence of strong claims on sensitive and appropriate awareness tests, not the dismissal of every later experimental dissociation Holender (1986).
A modern cognitive account was consolidated through experiments that separated performance on indirect tasks from participants' direct access to a stimulus or regularity. Reber's artificial-grammar experiments examined how exposure could improve classification of grammatical sequences and stimulated decades of work on learning without articulate rules. The original demonstrations established learning under an instructional arrangement; they did not establish that every acquired regularity was inaccessible under every possible test. Kihlstrom subsequently distinguished automatisms developed through practice from perceptual events and memories unavailable to immediate report, providing a framework in which the cognitive unconscious comprised multiple mechanisms and levels of availability Reber (1967) Kihlstrom (1987).
The late twentieth-century focus on visual masking made awareness manipulable on a trial-by-trial basis while electrophysiology and functional imaging allowed investigators to trace processing beyond overt judgments. Greenwald and colleagues demonstrated transient semantic priming from masked words with tight temporal constraints. Subsequent discussions separated access to report from the mere arrival of sensory information in cortical networks. Contemporary studies combine stimulus-level interventions with quantitative awareness scales, offering more discriminating evidence than an impressionistic statement that a person saw nothing. Historical development is therefore best reconstructed as the progressive refinement of operational contrasts between residual information and reportable awareness Greenwald et al. (1996) Dehaene et al. (2006) Del Cul et al. (2007).
Core Distinctions: Stimulus, Representation, Access, and Report
A stimulus may be physically present, encoded by a sensory system, represented in a local neural population, influence an action, and become consciously reportable. These events have related but nonidentical operational signatures. Signal transmission does not automatically imply identification of meaning, and an accurate response does not always imply awareness of why the response occurred. Conversely, a failure of verbal description does not establish absence of experience: subtle phenomenology may remain while a participant lacks a precise word, confidence, or memory for it. A credible design states which stage is manipulated and which alternative explanations its awareness measure can exclude Merikle et al. (2001) Sandberg et al. (2010).
Dehaene and colleagues distinguish subliminal information whose strength is insufficient for broad access, preconscious information capable of access but currently unattended, and conscious processing characterized by access under favorable conditions. This taxonomy is an explanatory proposal associated with global neuronal workspace theory. It provides predictions about attention, recurrent amplification, report and cerebral dynamics, but it does not make subjective experience directly observable. The categories are experimentally valuable when investigators independently measure sensory strength, attentional allocation, and report. Confusion arises when a failure to report is treated as equivalent to stimulus subliminality even though distraction, selection, response demands, or memory failure may be responsible Dehaene et al. (2006).
Automaticity refers to properties of an operation such as speed, learned efficiency, limited intentional control, or reduced demand on attention. It is analytically distinct from awareness of the operation's content. A practiced movement can be guided automatically while its goal remains conscious; a masked stimulus may affect performance without being processed with the effortless stability often associated with habit. Implicit denotes measurement by indirect consequences or learning expressed without explicit articulation; it also describes task format rather than necessarily proving lack of experience. Keeping automaticity, implicitness, unconsciousness, and reportability separate allows results from memory, learning, and motor control to be compared without definitional shortcuts Kihlstrom (1987) Reber (1967).
What Counts as Evidence: Direct and Indirect Measures
The central experiment measures an indirect effect of information while independently examining whether the information was accessible through an appropriately matched direct task. In a masked-prime study, an indirect index could be the difference in response latency or error probability between congruent and incongruent prime–target trials. The direct measure asks which prime was presented, perhaps through forced-choice identification under comparable timing, attention, and stimulus conditions. Demonstrating a prime-related consequence is often straightforward; establishing that the consequence occurred without task-relevant awareness requires careful sampling, sensitive instruments, and an inference that respects uncertainty in the awareness estimate Merikle et al. (2001) Van den Bussche et al. (2009).
A dissociation compares two dependent variables, often an implicit performance effect and an explicit awareness score. Such a dissociation can be informative when their measurement scales are calibrated and competing response pathways are considered. However, measuring an indirect effect with high reliability and a direct effect with a small or noisy sample invites a misleading contrast of significance against nonsignificance. Failure to reject chance-level discrimination cannot be treated as proof that discrimination is absent. A more informative analysis estimates plausible bounds on residual awareness, compares effect magnitudes under matched precision, and tests how much visible information could explain the indirect response Vadillo et al. (2016) Sandberg et al. (2010).
Qualitative dissociations can provide complementary evidence. Conscious and nonconscious stimuli might yield different patterns of persistence, generalization, temporal integration, or interference even when each produces a measurable response. Greenwald's brief time window for masked semantic activation exemplifies a difference in functional profile rather than merely a positive prime effect. Such contrasts sharpen theoretical conclusions because they constrain the kind of process that plausibly operated. Their interpretation still depends on whether the conscious and masked conditions are comparable in sensory energy and task demands; otherwise, apparent differences may follow stimulus strength or attentional allocation instead of awareness itself Greenwald et al. (1996) Kouider & Dehaene (2007).
Awareness Measurement and the Problem of Exhaustiveness
Objective awareness tests ask participants to detect, discriminate, or identify stimuli; subjective measures ask what was experienced or how clearly it was experienced. The same trial may reveal different capacities under these procedures. Forced-choice accuracy supplies an observable criterion, but the selected alternatives and decision policy can omit weak phenomenology. Subjective ratings come closer to the participant's experience while varying with confidence thresholds, language, motivation, and instructions. The Perceptual Awareness Scale and related approaches allow finer categories than a binary seen–unseen judgment. Comparing measurements within the same paradigm exposes their different sensitivity and their distinct response assumptions Sandberg et al. (2010) Merikle et al. (2001).
Signal detection theory distinguishes discriminative sensitivity from willingness to use a particular response. A tendency to claim visibility, select an option, or assign high confidence should not be confused with increased stimulus information. When a claim depends on direct-task performance being effectively at chance, estimation must include sampling variability and potential response bias. Metacognitive sensitivity concerns the extent to which confidence distinguishes correct from incorrect choices; it is different from the accuracy of the primary task. Fleming and Lau show why metacognitive measures need explicit calibration rather than an unexamined correlation between confidence and success Fleming & Lau (2014).
An exhaustive awareness assessment would detect every type of consciously usable information relevant to the observed indirect effect. No single test automatically meets that standard. An identification test may miss rough category knowledge, affective valence, familiarity, or a fragment of form; a confidence test may miss weak but consistent discrimination. Experiments should include several direct questions when theoretically relevant, maintain matched stimulus timing, and ideally estimate how indirect effects vary across the graded range of awareness. Such analyses turn awareness from a post hoc label into an experimentally characterized variable with measurement error, individual variation, and a defined inferential scope Sandberg et al. (2010) Vadillo et al. (2016).
Perceptual Masking and Temporal Control
Visual masking arranges stimuli so that a target is followed, surrounded, or otherwise interrupted by another visual event. Changing target duration and target–mask timing can reduce identification while leaving evidence of partial processing. Backward masking is particularly useful because it permits a brief signal to initiate processing before its reportable visibility is disrupted. The resulting dynamics depend on stimulus properties, relative contrast, spatial arrangement, and the task. A masked stimulus is thus a physical experimental condition; whether it qualifies as nonconsciously processed must be established through the awareness measures applied to the particular participants and trials Kouider & Dehaene (2007) Dehaene et al. (2006).
The masked-priming paradigm typically pairs a prime with a consciously presented target and evaluates how the prime changes target performance. A lexical decision, category judgment, or response compatibility task can expose information that would otherwise be difficult to measure directly. Researchers must control for visible fragments, repeated stimulus–response mappings, and post-target information that might make the prime retrospectively inferable. Congruency effects are sometimes carried by learned motor preparation rather than a broadly conceptual representation. The interpretive question is therefore which properties of the prime are encoded and transferred to the target, and under what awareness and task constraints Greenwald et al. (1996) Van den Bussche et al. (2009).
Masking provides a tractable window into processing time. Del Cul and collaborators associated conscious access to masked targets with a nonlinear threshold and later distributed electrical activity in their experimental paradigm. Earlier target-related responses remained measurable under conditions in which participants failed to report the stimulus. This distinction supports time-resolved investigation of partial information propagation but leaves open whether the late activity is necessary for experience, report, task execution, or some combination. Strong evidence uses manipulations that independently vary visibility and report demands while observing neural signals and behavior. Such converging methods can delimit the reach of nonconscious processing more precisely than behavioral masking alone Del Cul et al. (2007) Lamme (2003).
Masked Priming: Mechanisms and Meta-Analytic Evidence
Masked priming remains one of the best investigated families of nonconscious effects because its defining comparisons can be repeated across materials and laboratories. Van den Bussche, Van den Noortgate, and Reynvoet synthesized masked-priming studies and reported evidence that masked information affects subsequent behavior, while examining moderators of the effect. A meta-analytic average does not determine the mechanism of any individual experiment. The meaning of a priming difference depends on stimulus overlap, visibility control, target categories, response requirements, and the distribution of trial-level outcomes. A reliable mean latency difference demonstrates sensitivity to preceding information, not a general capacity for independent unconscious reasoning Van den Bussche et al. (2009).
Priming can occur through perceptual repetition, phonological or semantic activation, response preparation, and expectation. A prime and target that require the same response may speed a prepared motor program without demonstrating a rich representation of the prime's meaning. Conversely, a prime–target relation involving distinct words and transferred category knowledge can support a semantic interpretation if perceptual and response confounds are controlled. The experimental design should therefore separate identity repetition, semantic relatedness, response congruence, stimulus familiarity, and changes in task set. These factors matter because different paths can converge on the same reaction-time outcome while implying markedly different cognitive operations Greenwald et al. (1996) Kiefer & Martens (2010).
A further distinction concerns an effect's size, duration, and generalizability. Greenwald and colleagues described masked semantic activation as fleeting and sensitive to the interval before the target, providing a functional boundary on the phenomenon they observed. Meta-analysis demonstrates that a class of effects survives aggregation, but pooled results inherit variation in awareness tests, prime durations, publication practices, and task designs. Robust interpretation requires preregistered contrasts, effect estimates with uncertainty, direct-task checks, and tests of whether the effect persists when response mappings change. The scientific conclusion is strongest when a specific mechanism predicts a distinctive pattern across several controlled tasks Greenwald et al. (1996) Van den Bussche et al. (2009).
Semantic Activation and the Depth-of-Processing Question
Semantic processing refers to transformations sensitive to meaning or category relations rather than only visual or acoustic identity. A masked word's influence on a later category decision can indicate activation of lexical or semantic information. Yet semantic classification is not a unitary operation: a learned association between a stimulus and a response can imitate meaning-sensitive behavior, and partial perceptual cues may correlate with the category. The history of this controversy includes Holender's critique of awareness measurement and later demonstrations that carefully masked primes produce reliable short-lived semantic effects under some conditions Holender (1986) Greenwald et al. (1996).
Research should ask how far semantic representations generalize beyond previously practiced items. Transfer to new exemplars, arbitrary changes in response mapping, and cross-modal relations provide different tests of abstraction. Such tests require careful matching because a new item may also vary in frequency, visibility, familiarity, or task difficulty. Theoretical discussions of levels of processing emphasize that early visual features, lexical identity, category membership, and complex relational integration need separate evidence. A simple congruency effect supports only the representational distinction established by its controls; claiming propositional reasoning or contextual understanding demands more complex discriminating conditions Kouider & Dehaene (2007) Mudrik et al. (2014).
The most defensible account treats nonconscious semantic processing as a conditional capacity whose observable expression depends on prime visibility, prior associations, current task, and temporal proximity. Automatic category activation may be highly efficient without maintaining a representation over long delays or flexibly recombining new concepts. Experiments that vary both the duration of suppression and the required semantic relation can identify these constraints. This approach resists extrapolating from a brief lexical bias to extended thought. It also invites direct comparison of conscious and nonconscious versions of the same semantic operation to determine which computations are shared and which require sustained access Greenwald et al. (1996) Mudrik et al. (2014).
Attention, Task Sets, and Preparatory Control
Attention changes the probability that information will be selected, enhanced, maintained, and used for a goal. Conscious report is often supported by attention, yet the constructs are separated experimentally and theoretically. An unattended visible stimulus and a deeply masked attended stimulus need not undergo equivalent processing. Lamme argued that selection and phenomenal awareness have different biological and psychological determinants, while the global-workspace account predicts important interactions between attention, signal strength, and widespread access. The point of disagreement can be examined by manipulating spatial attention, temporal expectation, and report demands independently rather than assigning a single unobserved state to every missed event Lamme (2003) Dehaene et al. (2006).
Kiefer and Martens experimentally examined how current task sets modulate subsequent masked semantic priming. Their evidence undermines the assumption that an unconscious stimulus automatically activates every possible aspect of its meaning. Task preparation can tune what is processed even when the participant does not consciously identify the stimulus. An apparently effortless effect may therefore depend on conscious instructions or goals established earlier. Disentangling preparation from stimulus processing requires measuring the current task set, maintaining comparable prime characteristics, and varying whether a relevant semantic relation is anticipated. Conditional unconscious processing is compatible with meaningful top-down control Kiefer & Martens (2010).
Attention also limits the inference from missed stimuli during demanding tasks. A failure to notice information could reflect weak encoding, competition, rapid forgetting, failure to form a stable report, or a genuinely absent perceptual experience. These possibilities can be separated through immediate discrimination, graded confidence, eye movements, time-resolved neural recording, and controlled cueing. Studies should specify whether their target is experience, report, memory, or behavior. A theory predicting unconscious information use under attention withdrawal must describe how residual representations survive selection and how those representations differ from the effects elicited by masking under otherwise attended conditions Lamme (2003) Merikle et al. (2001).
Blindsight and Dissociations After Visual Cortical Damage
Blindsight research provides historically important evidence that visual capacities can persist in parts of the visual field that a patient describes as unseen. In 1974, Sanders, Warrington, Marshall, and Weiskrantz reported discriminations in a field affected by occipital damage despite an absence of ordinary visual awareness. A related case report by Weiskrantz and colleagues investigated residual capacity following restricted occipital ablation. Such findings are striking because the indirect outcome is an intentional forced-choice response rather than merely a speeded reaction to an unseen prime. Their interpretation depends on lesion location, residual vision, the wording of subjective reports, and the precise task being performed Sanders et al. (1974) Weiskrantz et al. (1974).
Blindsight is a family of dissociations rather than proof of one uniform alternative visual pathway. Residual discrimination can depend on stimulus motion, contrast, location, affective significance, and the patient's available neural circuits. Clinical case designs often provide unusually strong within-person contrasts but limited population generalization. Researchers also must distinguish a denial of ordinary visual seeing from a total absence of any experience such as vague movement, feeling, or uncertain presence. Reports should be elicited sensitively and compared with objective performance across stimulus strengths. A clinical deficit can reveal component organization without implying that all normal nonconscious perception has the same mechanism Sanders et al. (1974) Merikle et al. (2001).
Lau and Passingham used masking in healthy observers to construct conditions with comparable objective discrimination but differing reported awareness, a paradigm described as relative blindsight. They associated the awareness contrast with particular prefrontal activity in their imaging experiment. This does not make a lesion-induced syndrome and a laboratory masking state physiologically equivalent. Rather, the comparison illustrates a general strategy: match objective task capacity while varying subjective access. Such designs constrain accounts that identify conscious experience solely with correct discrimination, although interpretation still depends on how well confidence and visibility ratings capture experience and how task requirements engage the relevant brain networks Lau & Passingham (2006) Sandberg et al. (2010).
Continuous Flash Suppression and Invisible Stimulus Competition
Continuous flash suppression manipulates perception by delivering rapidly changing high-contrast images to one eye while presenting a target to the other. Tsuchiya and Koch introduced a powerful implementation and showed that a stimulus can remain perceptually suppressed for extended periods compared with conventional rivalry arrangements. Their work also revealed changes in negative afterimage strength under suppression, illustrating that information can affect sensory adaptation in ways not transparent to subjective visibility. Unlike brief backward masking, prolonged interocular competition permits tests of temporal persistence and content sensitivity, but its specific binocular mechanisms make simple generalizations across paradigms precarious Tsuchiya & Koch (2005).
The interpretation of a suppressed stimulus depends on more than its objective presence. Visibility can break through unpredictably, fluctuate across trials, and vary with spatial position or adaptation. Detection and discrimination tasks administered under the same display conditions are necessary to establish what was available to the observer. Experiments that use time to breakthrough as an index of high-level nonconscious processing require caution: early visual differences can change suppression duration without implying semantic analysis. Matching spatial frequency, contrast, edge structure, and other low-level features is a central experimental requirement when testing faces, words, categories, or emotionally salient images Tsuchiya & Koch (2005) Kouider & Dehaene (2007).
Extended suppression makes information integration an especially relevant question. A masked or suppressed event may be processed in a limited local channel yet fail to support stable, flexible combination across space and time. Mudrik, Faivre, and Koch reviewed evidence that some forms of integration can occur without reportable awareness while identifying important limits, particularly for novel or extended combinations. Their synthesis cautions against using the presence of any integration as proof that consciousness makes no functional difference. The research program instead compares dimensions of integration, prior familiarity, stimulus strength, temporal span, and the effort required to incorporate information into current goals Mudrik et al. (2014).
Integration Across Features, Modalities, and Time
Information integration can mean binding visual features into objects, combining signals from multiple senses, applying semantic context, or maintaining relations among events. These operations differ in complexity and in the temporal windows they require. A demonstration that two suppressed visual features jointly affect behavior cannot establish that arbitrarily complex narratives or instructions are integrated without awareness. Mudrik, Faivre, and Koch surveyed the evidence across perceptual, multisensory, semantic, and novel integration, emphasizing that some combinations occur under limited awareness conditions while other forms become fragile. Research must identify the elements integrated, their familiarity, the outcome measure, and the mechanism proposed to combine them Mudrik et al. (2014).
Experimental integration studies confront the problem of partial awareness. An observer may fail to identify a compound event yet consciously register a feature or relationship sufficient to guide behavior. The correct comparison tests awareness for each candidate cue alongside awareness of the combined representation. In sequential tasks, retention of the first event may support later integration even when its original identity was not accurately reported. Strong designs vary temporal separation, attention, and the need for novel recombination while estimating both direct sensitivity and integrated-task performance. The resulting evidence supports a map of capacities across conditions rather than a single threshold between simple and complex cognition Sandberg et al. (2010) Mudrik et al. (2014).
A mechanistic account should predict how representations are maintained between events. Short-lived feedforward activation could facilitate immediate target decisions, whereas integration over delays may need recurrent maintenance, working-memory support, or prior learned associations. Conscious access might enhance flexible routing and temporal stability without being the exclusive source of every binding operation. That interpretation remains a theoretical hypothesis whose tests should distinguish integration from sequential priming, unconscious storage from fleeting activation, and novel inference from retrieval of practiced combinations. The same vocabulary applies to sensory experiments and higher-level decisions only after the underlying operational criteria are held constant Dehaene et al. (2006) Greenwald et al. (1996).
Implicit Memory, Familiarity, and Retrieval Without Recollection
Implicit memory refers to influences of past experience on present performance that are measured without requiring deliberate recollection of a study episode. Familiarity, perceptual fluency, skill acquisition, and priming can be expressed when someone cannot retrieve the source of an earlier encounter. A failed source report therefore reveals one dimension of episodic access rather than proving that the earlier information was never consciously encoded. Kihlstrom's account of the cognitive unconscious placed implicit memory alongside automatic skills and subliminal perception while acknowledging their different mechanisms and evidential histories Kihlstrom (1987).
Researchers usually compare a direct memory task, such as recognition or recall, with an indirect measure such as completion of a fragment, classification speed, or changes in performance. Interpretation requires care because an ostensibly indirect test can be intentionally controlled by a participant who recognizes the materials; direct recognition can also be undermined by task difficulty or inappropriate cues. The relation between the two outcomes depends on what is represented, the time interval, and the retrieval demands. Separating effects of encoding, storage, and retrieval is essential before concluding that a memory trace is inaccessible to consciousness in an enduring sense Merikle et al. (2001) Kihlstrom (1987).
Implicit memory offers a bridge from the tightly controlled masking experiment to learning across longer timescales. The key questions concern how prior exposure changes subsequent processing, which elements of the exposure are retained, and how flexible the consequences are across settings. A speed advantage for repeated perceptual material may result from stimulus-specific tuning; improved performance on new combinations suggests a more general representation. Neither result alone determines whether the learner possesses articulate knowledge about the change. Comparing indirect transfer with multiple explicit tests and manipulating retrieval cues provides a stronger account of the learned information's structure Reber (1967) Vadillo et al. (2016).
Implicit Learning and Sensitivity to Statistical Regularities
Implicit-learning experiments investigate how exposure to structured environments changes choices and predictions when participants cannot fully verbalize the governing regularities. Reber's 1967 artificial-grammar experiments used structured letter strings to examine learning beyond the explicit discovery of rules. Such tasks helped launch a research tradition involving sequence learning, statistical patterns, and classification transfer. Yet acquisition of probabilistic knowledge and absence of conscious access are separate propositions. Participants may possess fragmentary rules, familiarity for common combinations, or degrees of confidence that a binary explanation task fails to capture Reber (1967).
A strong design evaluates learning with novel sequences that differ from training stimuli, while testing awareness using questions sensitive to the same predictive information that supports performance. The experimenter must separate memory for particular exemplars from abstraction of structural regularities and from simple frequency sensitivity. Learning may remain distributed across many weak associations that participants can exploit before they can express a concise rule. Conversely, poor verbal explanation may reflect linguistic demands rather than genuine unavailability of useful knowledge. Graded judgments, forced choice, recognition probes, and postexperimental interviews each constrain different possible interpretations Reber (1967) Newell & Shanks (2014).
Statistical claims about unconscious learning face an asymmetric challenge. A study seeks a detectable indirect learning effect while treating awareness as absent; the second assertion is often supported by a nonsignificant test. Vadillo, Konstantinidis, and Shanks showed through a review of contextual-cueing work why insufficient power can conceal real explicit knowledge and inflate apparent dissociations. Their analysis emphasizes precision of the direct test, not simply collecting more learning trials. Future work benefits from equivalence approaches, preregistered awareness thresholds, hierarchical models of individual differences, and trial-matched tests that make weak conscious knowledge easier to detect Vadillo et al. (2016).
Motor Preparation, Action Selection, and Agency
Behavior can be shaped by information that is not incorporated into a person's conscious account of why an action occurred. Masked response primes may influence the speed or likelihood of selecting a motor response, especially when stimulus–response associations are already established by task instructions. This can reveal rapid coupling among sensory signals, decision thresholds, and motor execution. Such an effect is narrower than evidence that the full goals of an action were formed without consciousness. Experiments should vary response mapping, prime–response contingency, target visibility, and initiation requirements to determine how much processing is automatic preparation rather than goal-directed evaluation Greenwald et al. (1996) Kiefer & Martens (2010).
Nisbett and Wilson examined the limits of verbal reports about higher-order mental processes and argued that explanations of actions are often constructed from plausible causal theories rather than direct observation of the mediating operations. Wegner later analyzed conscious will as an experience of apparent mental causation that can diverge from the processes actually producing behavior. These analyses address the reliability of introspective causal attribution, which is distinct from whether people experience their goals, intentions, or movements. A person may be consciously aware of choosing an action yet misunderstand the influences that weighted the choice Nisbett & Wilson (1977) Wegner (2003).
Soon and colleagues demonstrated that patterns of cerebral activity could contain information about subsequent choices before participants reported conscious awareness of deciding in a constrained voluntary-choice paradigm. Prediction at an above-chance level is informative about preparatory activity but does not establish that a choice was determined with certainty or that complex life decisions follow the same course. The task structure, restricted response alternatives, temporal report method, and statistical classification all delimit the inference. Motor preparation research is thus most useful when it links neural precursors, trial-level decisions, and subjective timing while distinguishing predictive information from experimentally demonstrated causation Soon et al. (2008).
Decision Making, Intuition, and Claims of Unconscious Deliberation
A broad scientific literature examines whether decisions reflect evidence accumulated without deliberate attention or access to the specific reasons for choice. Dijksterhuis and Nordgren formulated unconscious-thought theory, which proposes circumstances under which distraction following information presentation could support decisions, especially on complex options. This account is a testable theory, not a general prescription that distraction improves judgment. Experimental findings depend on how options are presented, whether preference criteria are known, what participants remember, and whether conscious reflection and distraction conditions receive equivalent time and information Dijksterhuis & Nordgren (2006).
Newell and Shanks critically reviewed evidence on multiple-cue judgment, deliberation without attention, and decisions under uncertainty. Their evaluation emphasizes how strong unconscious-influence claims may depend on inadequate awareness tests or alternative strategies that remain consciously available. A participant can know some cue values, apply simple rules, or retain salient features without being able to reconstruct every intermediate step. A superior outcome after distraction, where observed, also admits explanations involving interference, response strategy, fatigue, and the distribution of attention. Mechanistic conclusions require tests that distinguish these explanations directly Newell & Shanks (2014).
A scientifically usable distinction separates the existence of unconscious influences on choices from the claim that unconscious thought can perform comprehensive multi-attribute optimization without awareness. Evidence for the former includes robust effects of prior stimuli and accessible or inaccessible associations; the latter requires demonstrations of integration, abstraction, and flexible transfer under exhaustive awareness assessment. Choosing a good option can also result from partial conscious knowledge or familiar heuristic policies. The unresolved questions concern when nonconscious influences help, when they bias outcomes, and whether explicit review can correct or reinforce those influences in realistic settings Hassin (2013) Newell & Shanks (2014).
Neural Signatures and the Architecture of Access
Neural recordings permit investigators to compare stimulus-related processing that survives masking with activity associated with conscious report. Del Cul, Baillet, and Dehaene reported a nonlinear transition in their masking paradigm: early brain responses occurred below the visibility threshold while later, broadly distributed activity accompanied reportable targets. The result supplies evidence for temporally differentiated processing. It does not settle the interpretation of every component because report preparation, confidence, attentional selection, and the subjective event can overlap in time. A neural correlate must be characterized through manipulations, temporal specificity, and comparison to alternative task demands Del Cul et al. (2007).
The global neuronal workspace account proposes that local stimulus processing can proceed without broad availability, while sufficiently strong and amplified representations gain access through large-scale recurrent interactions. The framework offers an operational contrast between subliminal, preconscious, and conscious states and predicts consequences for report, flexible use, and sustained propagation. Other accounts emphasize the importance of recurrent processing within sensory cortex or challenge the equation between globally reportable information and phenomenal experience. Experiments comparing these models must specify predicted signals and distinguish perceptual experience from the additional operations required to communicate it Dehaene et al. (2006) Lamme (2003).
Anatomical location is not a synonym for psychological function. A measured response in prefrontal cortex may support a report or decision rather than constitute consciousness itself; early activity in sensory regions can carry discriminable information without establishing subjective seeing. Lau and Passingham's relative-blindsight study offers one approach to matching performance while varying report, and its conclusions remain conditioned by the design's measurement assumptions. Strong causal inference would require interventions that change a specified circuit while holding relevant sensory and behavioral variables as constant as possible. Across methods, convergence between behavioral dissociations and neural timing is more persuasive than activation maps interpreted in isolation Lau & Passingham (2006) Del Cul et al. (2007).
Temporal Dynamics, Persistence, and Limits of Unseen Information
A nonconscious effect can be robust in amplitude and restricted in duration. Greenwald and colleagues reported that their masked semantic priming effect operated in a short temporal interval, with little evidence for persistence in a subsequent prime–target sequence. This boundary has theoretical significance: rapid activation sufficient to facilitate an immediate response is distinct from information that can be retained and deliberately manipulated. Temporal generalization tests should therefore vary the interval to the target, the number of intervening events, and the need for integration across those events. These procedures reveal what a processing route maintains and what it rapidly loses Greenwald et al. (1996).
Time also interacts with stimulus strength and awareness measurement. A prime initially below the identification threshold can become partly available through repeated exposure, changed attention, or predictive context. Conversely, an easily recognized stimulus may produce a weak indirect effect if its relationship to the current response is irrelevant. This creates a multidimensional space in which reportability, the duration of neural activation, memory retention, and behavioral influence can vary independently. A well-controlled study models trajectories across conditions rather than labeling a trial permanently unconscious from an average awareness score measured elsewhere Sandberg et al. (2010) Kiefer & Martens (2010).
Evidence from continuous flash suppression and masked neural recordings adds temporal resolution but also introduces distinctive physiological mechanisms. The extended invisibility achieved under interocular suppression does not mean semantic or motor representations persist for the same duration as the physical stimulus. Likewise, late cortical dynamics observed during report do not establish that all nonconscious computation is limited to early feedforward processing. Investigators must measure the temporal profile of each representational outcome separately, ideally using multiple dependent variables. Functional limits emerge from the joint pattern of activation, transfer, recovery, and task sensitivity rather than from stimulus duration alone Tsuchiya & Koch (2005) Del Cul et al. (2007).
Competing Theoretical Accounts and Discriminating Predictions
Accounts of nonconscious processing can be organized around representational depth, global availability, recurrent interaction, and functional capacity. The global-workspace taxonomy specifies when locally available representations become broadly accessible. Recurrent-processing approaches identify local feedback dynamics as relevant to awareness. The Yes It Can argument advanced by Hassin emphasizes that functions conventionally associated with consciousness may sometimes occur nonconsciously. These perspectives differ in their emphasis and in what must be demonstrated. The decisive test is whether their predictions diverge under matched sensory strength, attentional setting, cognitive demands, and measurements of experience Dehaene et al. (2006) Lamme (2003) Hassin (2013).
A claim that unconscious processes can implement a class of computation requires care at the algorithmic level. A similar final response can arise from different mechanisms, including cached associations, perceptual generalization, sequential priming, or flexible integration. Two conditions may be functionally similar in one task yet differ in transfer, stability, and susceptibility to interference. Hassin's thesis offers an agenda for testing broad functional capacity, while reviews of information integration identify conditions that constrain such generality. Theories gain explanatory resolution by predicting which components remain available when memory load, novelty, and temporal distance increase Hassin (2013) Mudrik et al. (2014).
The field also contains theories directed at decisions and metacognitive access rather than basic perception. Unconscious-thought theory proposes possible benefits of offline processing after information presentation, whereas critical analyses examine whether awareness and alternative choice strategies were properly controlled. A rigorous model comparison should include competing predictions about behavior, reports, confidence, and error patterns, not simply one significant interaction selected after the fact. The failure of a model on a suitably sensitive test has evidential value, and a successful result must generalize beyond the exact materials used to construct it. Models remain hypotheses whose precision improves with disconfirming as well as supporting evidence Dijksterhuis & Nordgren (2006) Newell & Shanks (2014).
Introspection, Confidence, and the Experience of Agency
Introspective evidence must be parsed into distinct questions: what a person experienced, what they believe happened, and how they infer that an internal event caused an action. Nisbett and Wilson reviewed evidence that people can be unaware of influences on their responses and may rely on causal theories when explaining their behavior. This does not imply that all introspective reports are false or scientifically useless. A report of visual clarity can measure phenomenology; a story explaining why an option was chosen addresses a different construct. Treating these questions as interchangeable distorts both the empirical design and the interpretation Nisbett & Wilson (1977).
Metacognitive studies can quantify how confidence covaries with actual task success, and those relations can be examined separately from bias toward high or low confidence. Fleming and Lau formalized distinctions among metacognitive sensitivity, efficiency, and response bias. Awareness experiments benefit from applying these distinctions because a participant can be good at discriminating targets but cautious in reporting visibility, or confident for reasons unrelated to discrimination. Calibration can be studied over repeated trials rather than inferred from one global questionnaire. Careful design also allows subjective reports to constrain neural explanations instead of merely decorating an objective performance measure Fleming & Lau (2014) Sandberg et al. (2010).
The experience of acting voluntarily creates another inferential challenge. Wegner argued that conscious will often involves an attribution of apparent mental causation, while Soon and colleagues examined neural information predictive of simple choices before reported conscious decision time. The two research programs focus on different phenomena: causal introspection and predictive brain activity. Neither result warrants a wholesale denial of agency in ordinary social or legal contexts. The relevant scientific investigation specifies timing, task freedom, decision alternatives, prior preparation, and the accuracy of subjective judgment. General theories of agency must account for both actions consciously controlled and influences that enter behavior without explicit awareness Wegner (2003) Soon et al. (2008).
Individual Differences, Adaptation, and Boundary Conditions
Nonconscious effects vary across observers partly because sensory thresholds, attention, prior experience, and response policies vary. A physical mask that makes a prime undetectable for one person may leave partial visibility for another. Aggregating across participants can therefore combine qualitatively different states and create apparent dissociations that disappear under individual-level scrutiny. Proper calibration establishes participant-specific visibility functions and models uncertainty instead of assuming a universal stimulus onset or contrast threshold. This principle is particularly important for clinical populations or studies using unusual perceptual tasks, where equivalent stimuli need not yield equivalent processing Sandberg et al. (2010) Lau & Passingham (2006).
Learning history can change what an unseen cue predicts. A familiar association may automatically bias a response that a novel relation would fail to influence, and task instructions can make specific semantic dimensions selectively relevant. This dependence on context suggests that the observed capacity of a nonconscious route must be expressed as a conditional relation among experience, current task, and stimulus quality. Disentangling learning from awareness requires longitudinal comparisons, transfer to untrained exemplars, and changes in stimulus–response mapping. Such tests can distinguish a robust learned predictive pathway from the illusion of general conceptual processing Kiefer & Martens (2010) Reber (1967).
Population transfer requires further discipline. Evidence from healthy young adults completing rapid visual judgments cannot automatically establish equivalent processing in children, older adults, patients with visual lesions, or people working in naturalistic environments. Blindsight provides a case where altered neural anatomy reshapes behavior, while contextual-cueing studies demonstrate how uncertainty in direct awareness measurement complicates broad interpretation. Researchers should describe sensory function, neurological status, familiarity with stimuli, cultural and linguistic demands, and task incentives where they are relevant. Generalization is a scientific question to be measured, not a property conferred by a large number of citations Weiskrantz et al. (1974) Vadillo et al. (2016).
Replication, Publication Bias, and Statistical Power
Replication in nonconscious-processing research concerns both the direction of an indirect behavioral effect and the absence or small size of direct awareness effects. These claims require different statistical treatment. A laboratory may repeatedly obtain masked priming under carefully specified conditions and still lack adequate evidence that every participant was unaware of all relevant prime information. Meta-analysis by Van den Bussche and colleagues supports the existence of masked-priming effects across studies, but reproducibility must be evaluated at the level of tasks, operational definitions, and the exact inferential claim. Selective reporting and analytic flexibility can affect estimates of effect magnitude and consistency Van den Bussche et al. (2009).
Tests of absence deserve special attention. Vadillo and collaborators identified the risks of low-powered awareness analyses in contextual-cueing research, where nonsignificant explicit tests can be used to assert unconscious learning. Increasing power only for the indirect measure creates systematic asymmetry: evidence of learning is precise, while evidence about knowledge remains diffuse. The problem is addressed by designing both measurements for adequate precision, specifying the smallest relevant awareness effect, reporting uncertainty bounds, and analyzing individual observers. Researchers should avoid selecting only trials described as unseen without considering how noisy trial classification can bias the indirect comparison Vadillo et al. (2016).
Credible replication includes variation that tests the mechanism rather than merely reusing favorable conditions. Independent laboratories can preregister their stimuli, exclusion rules, awareness calibration, response mappings, and planned analyses. Replication across semantically distinct targets or newly trained associations tests generalization; failures identify meaningful boundary conditions when sensitivity is sufficient. Transparent reporting also documents null and opposite-direction findings, which allow synthesis to distinguish stable phenomena from publication-shaped distributions. The goal is a reproducible causal explanation specifying when the effect occurs, not an unqualified claim that all unseen information influences behavior in the same way Newell & Shanks (2014) Van den Bussche et al. (2009).
Causal Identification and Experimental Design
The strongest causal inference begins with an intervention on a candidate source of information. Prime visibility can be manipulated through contrast, timing, masking, or attention; target response mappings can be changed independently; and concurrent neural measures can identify the temporal sequence of effects. Each intervention carries additional consequences that may themselves change behavior. A stronger design anticipates these pathways and uses matched comparisons or factorial manipulation to isolate them. The mere correlation between a neural response and an unseen target is insufficient to establish whether the response carries the relevant information or causes the eventual choice Del Cul et al. (2007) Kouider & Dehaene (2007).
Control conditions must be chosen according to the proposed mechanism. A perceptual explanation demands visually balanced stimuli; a semantic explanation demands meaningful contrasts that survive response remapping; a learning explanation demands transfer beyond trained exemplars; and a metacognitive explanation demands calibrated reports. A blind field, masked event, or suppressed image is not a universal control condition for consciousness because each manipulation changes sensory processing differently. Clear preregistered predictions should indicate which comparison would support the model and which observation would require revision. An experimental null can be informative when its sensitivity is demonstrated Weiskrantz et al. (1974) Mudrik et al. (2014).
Longitudinal and naturalistic research can complement highly controlled experiments, especially for habits and adaptive routines. Yet complexity increases the number of potential conscious cues and uncontrolled confounds. Investigators can combine laboratory manipulations with field measurements, trial-wise confidence, and computationally specified predictions, while recording when a participant acquires explicit knowledge of a regularity. Source of information, process of transformation, availability of representations, and eventual behavior should be measured separately. Maintaining these distinctions prevents a descriptive label such as intuitive choice from substituting for a causal mechanism and allows meaningful comparisons across psychological subfields Reber (1967) Newell & Shanks (2014).
Applications and Ecological Interpretation
Nonconscious processing research is relevant to skilled action, education, perceptual learning, communication, and judgment because people often behave efficiently without articulating every mediating operation. Scientific translation nevertheless requires close correspondence between an experimental mechanism and the proposed application. A millisecond-scale masked-priming effect is evidence of transient influence under structured laboratory conditions; it does not establish durable persuasion or reliable control of consumer behavior. Claims that hidden messages can produce complex and lasting behavioral change require direct field evidence, appropriate control groups, and follow-up measurements rather than analogy with lexical priming Greenwald et al. (1996) Newell & Shanks (2014).
Implicit learning may inform the study of accumulated expertise because repeated environmental structure can alter expectations and performance before observers can explain the regularities fluently. The relevant question is when indirect learning translates to robust transfer and when explicit instruction aids correction or generalization. Similarly, the limits of introspective causal explanation matter when professionals review decisions, but fallible explanation does not make evidence-based self-reflection worthless. Researchers can examine how feedback, deliberation, and different forms of reporting improve calibration across repeated decisions. Practical utility arises from specifying which procedure improves which measurable outcome Reber (1967) Nisbett & Wilson (1977).
Ethical interpretation includes the epistemic limits of using nonconscious processes as an explanation of another person's behavior. Behavioral effects produced by subtle cues do not license mind-reading, confident attribution of hidden motives, or reduction of personal responsibility to a laboratory association. Clinical and educational uses require population-specific evidence, informed consent, and careful communication about uncertainty. Attention to the conditions under which people become aware of cues can guide transparent interfaces and better decision environments. A useful application should identify benefits, foreseeable harms, alternative interventions, and a method for assessing generalization beyond the original experiment Merikle et al. (2001) Newell & Shanks (2014).
Artificial Systems and Cross-System Comparison
The Artificial Era makes nonconscious processing a productive object of comparison between human psychological research and computational systems. An artificial system may transform information internally without producing a user-facing explanation of each operation. Such opacity resembles human limits of introspective reporting only at a superficial functional level unless the comparison specifies what counts as representation, internal access, report, task demand, and causal influence in each system. Human consciousness measures were designed around subjective reports and biological perceptual capacities, and cannot be transferred unchanged to computational architectures. Cross-system work therefore requires its own operational vocabulary and appropriately chosen interventions Dehaene et al. (2006) Fleming & Lau (2014).
Computational models can nevertheless assist this research. A model may implement competing pathways that differ in recurrent maintenance, attention-like selection, response mapping, and access to an output channel. Investigators can then test whether controlled changes reproduce patterns such as brief priming, transfer under new mappings, or the relation between confidence and accuracy. Matching behavior across architectures demonstrates functional similarity on the measured tasks; identifying a mechanism requires intervention on the model's internal pathways and independent validation against human data. A simulation that reproduces a dissociation does not establish shared phenomenology or biological equivalence Mudrik et al. (2014) Lau & Passingham (2006).
Human–artificial collaboration also raises questions about information presented below a user's focal attention, recommendation cues, and explanations generated after an automated decision. Here the relevant scientific constructs include cognitive offloading, transparency, confidence calibration, and distribution of information across participants and tools. Evidence from human masked priming supplies methodological warnings about confusing an observed outcome with an account of the process producing it. Comparative analysis should examine causal pathways, report accuracy, opportunities for correction, and changing task goals. The Artificial Era extends the range of systems that can be compared while preserving the evidential discipline developed in experimental psychology Nisbett & Wilson (1977) Newell & Shanks (2014).
Open Problems and Research Priorities
The central unresolved problem is how to identify the amount and structure of information available without reportable awareness while controlling differences in sensory strength and measurement precision. The field needs designs in which direct and indirect measures are calibrated across participants and tasks, with confidence intervals or posterior uncertainty expressing what residual awareness remains plausible. Simultaneous measurement of several awareness dimensions could reveal whether effects attributed to unconscious processing depend on fleeting experiences, coarse category information, or metacognitive thresholds. These improvements would also strengthen comparisons between global-access and recurrent-processing accounts by specifying what each theory must explain Sandberg et al. (2010) Dehaene et al. (2006).
A second problem concerns processing depth and temporal generalization. Researchers should investigate when unseen features influence immediate responses, when associations persist across delays, and when genuinely novel relations can be integrated. Such studies require factorial manipulations of novelty, task set, interval, and response mapping combined with independent tests of partial knowledge. Stronger models would predict not merely whether an indirect effect occurs, but its shape across increasing demands and its change under deliberate attention. Cross-paradigm replication can identify which limits belong to the cognitive operation and which follow from a particular masking or suppression procedure Greenwald et al. (1996) Mudrik et al. (2014).
A third priority is accountable scientific synthesis. Coordinated studies should publish complete task specifications, stimulus code, exclusion decisions, direct awareness measures, and data supporting both positive and null results. Evidence quality can then be evaluated separately for perceptual priming, implicit learning, clinical dissociation, action preparation, and complex decisions. Research should include populations and contexts beyond narrowly sampled visual laboratories while maintaining explicit criteria for transfer. This agenda makes possible progressively precise claims about where nonconscious processing contributes to adaptive behavior and where conscious access supports stability, flexibility, and communication of information Vadillo et al. (2016) Newell & Shanks (2014).
Conclusion
Nonconscious information processing is an established research domain comprising multiple demonstrable forms of stimulus sensitivity, memory influence, learning, and action preparation outside immediate conscious report. Masked priming, visual dissociations, implicit-memory tasks, and clinical observations supply different evidence for this broad proposition. Each phenomenon has its own representational content, temporal constraints, measurement procedures, and inferential vulnerabilities. The most reliable synthesis recognizes an empirical hierarchy: detecting an indirect behavioral influence, identifying its mechanism, establishing absence of relevant awareness, and generalizing the result to broader cognition are distinct achievements requiring distinct tests Greenwald et al. (1996) Van den Bussche et al. (2009) Merikle et al. (2001).
The major controversies concern the extent and flexibility of processing when awareness is rigorously assessed. Theoretical claims about global availability, recurrent dynamics, high-level unconscious functions, and superior unconscious deliberation generate different predictions. The literature supports the importance of prior learning and task settings while showing that weak awareness measurement and inappropriate comparisons can exaggerate the apparent independence of unconscious cognition. Strong explanatory progress comes from designs that compare conscious and nonconscious conditions at matched levels of performance, model uncertainty in awareness, and measure how effects change across time, novelty, and transfer Dehaene et al. (2006) Hassin (2013) Newell & Shanks (2014).
The enduring scientific result is a differentiated account of psychological information processing. Human activity reflects many interacting pathways, some accessible to awareness and some detectable chiefly through their consequences. Conscious access, implicit influence, subjective confidence, and causal explanation can be analyzed together without collapsing their definitions. For the Psychology by Bogdanova research architecture, this dossier establishes the evidence-centered account of nonconscious processing and its relations to consciousness, perception, learning, and decision making. Future experiments and comparative work in the Artificial Era should extend this account through precise operational definitions, transparent reporting, causal interventions, and independently reproducible tests Fleming & Lau (2014) Vadillo et al. (2016).
References
- Dehaene, S., Changeux, J.-P., Naccache, L., Sackur, J., & Sergent, C. (2006). Conscious, preconscious, and subliminal processing: A testable taxonomy. Trends in Cognitive Sciences, 10(5), 204–211. https://doi.org/10.1016/j.tics.2006.03.007
- Del Cul, A., Baillet, S., & Dehaene, S. (2007). Brain dynamics underlying the nonlinear threshold for access to consciousness. PLoS Biology, 5(10), e260. https://doi.org/10.1371/journal.pbio.0050260
- Dijksterhuis, A., & Nordgren, L. F. (2006). A theory of unconscious thought. Perspectives on Psychological Science, 1(2), 95–109. https://doi.org/10.1111/j.1745-6916.2006.00007.x
- Fleming, S. M., & Lau, H. C. (2014). How to measure metacognition. Frontiers in Human Neuroscience, 8, Article 443. https://doi.org/10.3389/fnhum.2014.00443
- Greenwald, A. G., Draine, S. C., & Abrams, R. L. (1996). Three cognitive markers of unconscious semantic activation. Science, 273(5282), 1699–1702. https://doi.org/10.1126/science.273.5282.1699
- Hassin, R. R. (2013). Yes it can: On the functional abilities of the human unconscious. Perspectives on Psychological Science, 8(2), 195–207. https://doi.org/10.1177/1745691612460684
- Holender, D. (1986). Semantic activation without conscious identification in dichotic listening, parafoveal vision, and visual masking: A survey and appraisal. Behavioral and Brain Sciences, 9(1), 1–23. https://doi.org/10.1017/S0140525X00021269
- Kiefer, M., & Martens, U. (2010). Attentional sensitization of unconscious cognition: Task sets modulate subsequent masked semantic priming. Journal of Experimental Psychology: General, 139(3), 464–489. https://doi.org/10.1037/a0019561
- Kihlstrom, J. F. (1987). The cognitive unconscious. Science, 237(4821), 1445–1452. https://doi.org/10.1126/science.3629249
- Kouider, S., & Dehaene, S. (2007). Levels of processing during non-conscious perception: A critical review of visual masking. Philosophical Transactions of the Royal Society B: Biological Sciences, 362(1481), 857–875. https://doi.org/10.1098/rstb.2007.2093
- Lamme, V. A. F. (2003). Why visual attention and awareness are different. Trends in Cognitive Sciences, 7(1), 12–18. https://doi.org/10.1016/S1364-6613(02)00013-X
- Lau, H. C., & Passingham, R. E. (2006). Relative blindsight in normal observers and the neural correlate of visual consciousness. Proceedings of the National Academy of Sciences, 103(49), 18763–18768. https://doi.org/10.1073/pnas.0607716103
- Merikle, P. M., Smilek, D., & Eastwood, J. D. (2001). Perception without awareness: Perspectives from cognitive psychology. Cognition, 79(1–2), 115–134. https://doi.org/10.1016/S0010-0277(00)00126-8
- Mudrik, L., Faivre, N., & Koch, C. (2014). Information integration without awareness. Trends in Cognitive Sciences, 18(9), 488–496. https://doi.org/10.1016/j.tics.2014.04.009
- Newell, B. R., & Shanks, D. R. (2014). Unconscious influences on decision making: A critical review. Behavioral and Brain Sciences, 37(1), 1–19. https://doi.org/10.1017/S0140525X12003214
- Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84(3), 231–259. https://doi.org/10.1037/0033-295X.84.3.231
- Reber, A. S. (1967). Implicit learning of artificial grammars. Journal of Verbal Learning and Verbal Behavior, 6(6), 855–863. https://doi.org/10.1016/S0022-5371(67)80149-X
- Sandberg, K., Timmermans, B., Overgaard, M., & Cleeremans, A. (2010). Measuring consciousness: Is one measure better than the other?. Consciousness and Cognition, 19(4), 1069–1078. https://doi.org/10.1016/j.concog.2009.12.013
- Sanders, M. D., Warrington, E. K., Marshall, J., & Wieskrantz, L. (1974). “Blindsight”: Vision in a field defect. The Lancet, 1(7860), 707–708. https://doi.org/10.1016/S0140-6736(74)92907-9
- Soon, C. S., Brass, M., Heinze, H.-J., & Haynes, J.-D. (2008). Unconscious determinants of free decisions in the human brain. Nature Neuroscience, 11, 543–545. https://doi.org/10.1038/nn.2112
- Tsuchiya, N., & Koch, C. (2005). Continuous flash suppression reduces negative afterimages. Nature Neuroscience, 8, 1096–1101. https://doi.org/10.1038/nn1500
- Vadillo, M. A., Konstantinidis, E., & Shanks, D. R. (2016). Underpowered samples, false negatives, and unconscious learning. Psychonomic Bulletin & Review, 23, 87–102. https://doi.org/10.3758/s13423-015-0892-6
- Van den Bussche, E., Van den Noortgate, W., & Reynvoet, B. (2009). Mechanisms of masked priming: A meta-analysis. Psychological Bulletin, 135(3), 452–477. https://doi.org/10.1037/a0015329
- Wegner, D. M. (2003). The mind's best trick: How we experience conscious will. Trends in Cognitive Sciences, 7(2), 65–69. https://doi.org/10.1016/S1364-6613(03)00002-0
- Weiskrantz, L., Warrington, E. K., Sanders, M. D., & Marshall, J. (1974). Visual capacity in the hemianopic field following a restricted occipital ablation. Brain, 97(4), 709–728. https://doi.org/10.1093/brain/97.1.709