In the landscape of modern cognitive psychology, few theoretical frameworks have reshaped our understanding of executive functioning, transient information maintenance, and attentional allocation as fundamentally as the Time-Based Resource Sharing (TBRS) model. Formulated at the turn of the twenty-first century by French cognitive psychologists Pierre Barrouillet and Valérie Camos, the TBRS framework emerged in response to persistent empirical anomalies that plagued classical modular architectures of human memory. For decades, structural accounts treated short-term memory as a collection of specialized, passive storage repositories overseen by an ill-defined central executive. However, these paradigms frequently failed to explain how human beings seamlessly coordinate immediate, transient storage with concurrent, intensive computational processing in real time.
The core insight of the TBRS model lies in its radical re-conceptualization of mental capacity: working memory is neither a static container nor a set of structurally insulated buffers. Instead, it is a highly dynamic, time-constrained processing-storage system governed by the rapid, sequential deployment of a central attentional resource. At its core, the model posits that while conscious processing and active memory maintenance rely on the same single-capacity central bottleneck, the human brain achieves the illusion of parallel multi-tasking through high-frequency attentional switching occurring over fractions of a second. Within this micro-temporal landscape, memory traces are inherently fragile, undergoing spontaneous biological decay the moment executive attention is diverted toward ongoing processing routines.
Consequently, the survival of ephemeral representations in working memory depends on an active, covert mechanism termed attentional refreshing. Through this process, degraded memory items are repeatedly restored to peak activation levels during momentary pauses or micro-gaps within the primary processing stream. By introducing a mathematically rigorous, chronometric index of Cognitive Load—defined as the proportion of total time during which an intervening activity captures central attention—Barrouillet and Camos established an elegant, predictive framework that links the micro-temporal dynamics of cognitive tasks directly to macro-level memory capacity. The following treatise provides an exhaustive, multi-dimensional exposition of the TBRS architecture, detailing its theoretical origins, operational mechanics, computational formalizations, neurobiological substrates, lifespan developmental trajectories, and ongoing debates within contemporary cognitive science.
1. Introduction to the Time-Based Resource Sharing (TBRS) Model
1.1 Historical Context and Epistemological Origins
The emergence of the Time-Based Resource Sharing (TBRS) framework during the late 1990s and early 2000s marked a pivotal paradigm shift within cognitive psychology, driven by increasing dissatisfaction with the explanatory limits of classic structural architectures. For nearly three decades, the study of human short-term storage had been dominated by the multi-component working memory model proposed by Alan Baddeley and Graham Hitch (1974). In this classic tripartite architecture, memory maintenance was attributed to domain-specific, structurally independent “slave systems”—namely, the phonological loop for acoustic-verbal information and the visuospatial sketchpad for visual and spatial configurations. Both systems operated under the supervision of a domain-general “central executive.” While this architecture successfully accounted for simple span phenomena and modality-specific dual-task dissociations, it struggled to explain complex span performance and cross-domain interference patterns.
In parallel, researchers investigating human performance through complex span tasks (such as the reading span and operation span tasks developed by Daneman and Carpenter) observed that engaging in concurrent computational processing severely impaired the retention of both verbal and non-verbal items, regardless of whether the processing task matched the storage modality. The modular independence assumed by Baddeley’s framework could not easily accommodate these cross-domain processing-storage trade-offs without attributing an ever-expanding, vaguely defined set of maintenance responsibilities to the central executive. Furthermore, competing capacity-sharing models, such as those inspired by Just and Carpenter’s capacity hypothesis, conceptualized working memory capacity as an expandable pool of activation resources that could be continuously subdivided between concurrent processing and storage operations. However, these resource-sharing accounts lacked temporal precision, failing to articulate the micro-level chronometric mechanisms through which this division was executed.
Confronted with this theoretical impasse, Pierre Barrouillet and Valérie Camos initiated a systematic experimental program aimed at addressing the fundamental trade-off between maintenance and processing. Rather than treating storage capacity as a spatial container or a fluid energetic pool, they turned to the temporal dimension of cognition. Influenced by chronometric approaches in experimental psychology, the psychological refractory period paradigms formalized by Harold Pashler, and the embedded-processes view of Nelson Cowan, Barrouillet and Camos hypothesized that the limiting factor in working memory performance is time itself. Their initial formulations radically departed from static storage views, positing that human cognition operates under strict temporal constraints where attentional engagement must alternate rapidly between ongoing computational demands and the restorative maintenance of fading memory representations.
1.2 Core Premises and Fundamental Axioms
The Time-Based Resource Sharing model is anchored by four foundational axioms that collectively describe how the human cognitive apparatus balances transient storage with continuous computation. The first axiom establishes the conceptualization of working memory as a unified, dual-function system that simultaneously executes processing and storage operations. Rather than locating these operations within structurally discrete anatomical containers, the TBRS model asserts that processing tasks (such as evaluating an arithmetic equation or verifying a logical proposition) and maintenance operations (retaining a sequence of letters, digits, or visuospatial coordinates) compete directly for a shared, limited cognitive resource.
The second axiom posits the existence of a single-capacity central bottleneck. Drawing heavily from attentional theories, the TBRS model asserts that the executive resource required to construct new cognitive relations, make decisions, and manipulate task-relevant representations is fundamentally indivisible. Attention cannot be simultaneously split between conscious processing and active storage maintenance at any single, indivisible millisecond of cognitive execution. Instead, the central focus of attention behaves as a single-channel processing stream. Whenever a primary task demands conscious executive control, the focus of attention is captured entirely, leaving unattended memory traces vulnerable to degradation.
The third axiom defines the temporal mechanism that enables dual-task execution: rapid attentional switching. Because the focus of attention cannot be structurally divided, multi-tasking is achieved through high-frequency, sequential alternation between processing and maintenance routines. When human subjects perform complex working memory tasks, they do not hold items in mind while simultaneously executing cognitive operations; rather, they rapidly switch their attentional focus back and forth between the computational steps of the processing task and the stored memory representations. These switches occur across brief inter-stimulus intervals and micro-pauses that emerge naturally within the execution of complex cognitive behaviors.
The fourth axiom defines the relationship governing memory retention and temporal duration: the inverse relationship between cognitive load and memory trace survival. In the absence of attentional focus, memory traces are subject to spontaneous, autonomous time-based decay. As soon as attention is diverted to process an external stimulus, activation levels of the stored traces begin to decline continuously. To counteract this decay, the cognitive system must reactivate degraded traces via a covert mechanism known as attentional refreshing. The survival probability of a given memory item is therefore an exact, quantitative function of the temporal ratio between the duration of attentional diversion (decay) and the availability of free micro-intervals dedicated to attentional refreshing (restoration).
1.3 Significance Within Modern Cognitive Science
The formulation of the TBRS model has had profound epistemological and methodological ramifications across cognitive science, neuropsychology, and educational measurement. Foremost among its achievements is the elegant resolution of long-standing, conflicting empirical findings regarding processing complexity versus processing duration. Prior to the TBRS model, researchers engaged in fierce debates regarding whether the difficulty of an intervening task (e.g., performing complex mental arithmetic versus simple visual parity judgments) or the pure duration of the retention interval was the primary driver of forgetting in working memory. Barrouillet, Camos, and their colleagues demonstrated that this dichotomy was largely an artifact of imprecise temporal measurement. Complex tasks induce greater forgetting not because of inherent qualitative complexity, but because they capture the central bottleneck for a longer proportion of the available time, thereby starving stored items of essential refreshing opportunities.
By formalizing this dynamic through quantifiable temporal metrics, the TBRS model introduced rigorous mathematical modeling to executive function paradigms. Through the derivation of the Cognitive Load (CL) index—the ratio of total time of attentional capture to total time available—the model enabled researchers to predict immediate recall accuracy with striking linear precision across a multitude of experimental settings. This chronometric approach moved the field beyond subjective or structural categorizations of task difficulty toward standardized, objective temporal parameters that can be manipulated parametrically.
Furthermore, the TBRS framework bridged the theoretical divide between modular structural accounts and process-oriented emergent accounts of human cognition. It preserved the intuitive utility of executive control mechanisms while rooting them firmly in dynamic, temporal neurocognitive processes. In doing so, it transformed working memory assessment protocols. Standard complex span paradigms—historically criticized for relying on unconstrained, self-paced intervals that permitted idiosyncratic rehearsal and grouping strategies—were systematically redesigned into computer-paced, chronometrically controlled laboratory instruments. This methodological refinement has had far-reaching implications, extending into developmental psychology, aging research, and the assessment of atypical cognitive trajectories, where temporal dynamics serve as sensitive biomarkers of cognitive capacity.
2. Theoretical Foundations: Evolution Beyond Classical Working Memory Frameworks
2.1 Shortcomings of Structural Multi-Component Accounts
To fully grasp the theoretical leap represented by the TBRS model, one must examine the empirical and architectural shortcomings that undermined Baddeley and Hitch’s classic multi-component model during the late 1990s. The central thesis of the classic model posited that working memory maintenance was preserved by peripheral, modality-dedicated slave systems: the phonological loop operating via subvocal rehearsal, and the visuospatial sketchpad maintaining mental imagery. However, accumulating empirical data challenged the operational independence of these modules. Foremost among these challenges was the pervasive finding of cross-domain interference. If the phonological loop functioned as an autonomous, self-contained buffer insulated from domain-general processing, concurrent visuospatial, motoric, or abstract logical processing tasks should have exerted minimal impact on verbal recall, provided that phonological articulatory suppression was not introduced. Instead, studies revealed that demanding non-verbal tasks consistently and severely depressed verbal span, and vice versa.
A second major structural limitation involved the ambiguous supervisory mechanism known as the central executive. In Baddeley’s early formulations, the central executive functioned as an omniscient homunculus responsible for strategy selection, dual-task coordination, selective attention, and resource allocation. However, the precise operational mechanisms governing how this executive directed resources or interacted with the slave systems remained largely unformalized. When empirical evidence demonstrated that verbal items could be maintained under complete articulatory suppression, the multi-component model was forced to posit alternative storage vehicles, culminating in the post-hoc addition of the episodic buffer. Yet, the episodic buffer primarily acted as a descriptive repository rather than an explanatory mechanism for the dynamic interactions observed between processing tasks and storage items.
Perhaps the most decisive failure of the structural accounts was their inability to account for the precise temporal pacing of intervening tasks within complex span paradigms. If working memory capacity were determined strictly by the structural capacity of static buffers, then variations in the temporal pacing of an intervening processing task—holding the number and nature of operations constant—should have produced negligible variations in span performance. Yet, as Barrouillet and Camos systematically documented, altering the presentation rate of intervening stimuli while holding cognitive operations constant produced dramatic, predictable fluctuations in memory survival. Traditional models completely underestimated the profound degree to which verbal maintenance relies continuously on domain-general attentional resources, revealing the need for a framework that unified maintenance and executive processing under a shared, temporally constrained architecture.
2.2 Integration of Cowan’s Embedded-Processes Perspective
In developing their architecture, Barrouillet and Camos drew heavily upon the embedded-processes model advanced by Nelson Cowan (1999, 2005). Cowan proposed an alternative, non-modular taxonomy that conceptualized working memory not as an assemblage of specialized anatomical boxes, but as a hierarchical arrangement of cognitive activations: working memory consists of the entire pool of activated long-term memory (aLTM) representations, a subset of which falls within the direct focus of attention (FoA). Cowan argued that while activated long-term memory has an expansive capacity subject to decay and interference, the focus of attention itself is capacity-limited, typically accommodating approximately four distinct, unchunked informational items or ensembles at any given moment.
The TBRS model adopted Cowan’s core premise that the focus of attention is the central vehicle through which mental representations are highlighted and maintained against background cognitive noise. However, Barrouillet and Camos introduced a crucial, radical divergence regarding the structural capacity and operational dynamics of this attentional focus. While Cowan conceptualized the FoA as a multi-item spotlight capable of simultaneously holding three to four discrete chunks in an activated state, the TBRS model aligns with strict single-channel bottleneck theory. Barrouillet and Camos argued that the FoA is fundamentally limited to a single item or cognitive operation at any precise millisecond of time. What appears to be a multi-item attentional capacity in macroscopic observations is, in reality, the product of rapid, serial cycling of this singular focus across candidate items.
Furthermore, the TBRS model transformed Cowan’s capacity-based constraints into time-based constraints. In Cowan’s embedded-processes view, capacity limits are typically characterized in terms of the number of items that can simultaneously occupy the FoA. In contrast, the TBRS model posits that working memory limitations emerge from the rapid temporal dynamics governing the entry, exit, and refreshing of representations within this bottleneck. Representations outside the focus of attention do not merely linger passively in activated long-term memory; they suffer immediate, autonomous temporal degradation. The fundamental constraint on human working memory is therefore not a spatial or numerical limit on how many items can be held at once, but a temporal race between the rate of spontaneous trace decay in the absence of attention and the speed with which the single-item focus of attention can be redeployed to refresh those traces before they drop below an irreversible retrieval threshold.
2.3 The Evolution from Static Capacities to Dynamic Resource Allocation
The theoretical transition inaugurated by the TBRS model fundamentally altered how cognitive psychology interprets historical landmarks in memory research, beginning with George Miller’s seminal “magical number seven, plus or minus two.” For decades, Miller’s chunking metric had reinforced the conceptualization of short-term memory as a set of static structural slots or storage bins. Even as subsequent research revised Miller’s estimate downward to four items, the underlying paradigm remained structurally static: working memory capacity was viewed as a physical limit on the number of discrete entities that could reside simultaneously in a central storage buffer.
The TBRS framework challenged this spatial metaphor by demonstrating that static span metrics are macroscopic emergent properties of microscopic temporal trade-offs. Working memory capacity is not an immutable structural container with four or seven fixed slots; it is a dynamic equilibrium established through continuous temporal scheduling. By analyzing performance across complex span tasks, Barrouillet and Camos illustrated that the number of items a participant can reliably recall is directly determined by the micro-intervals of time available to shift attention to those items during task execution. If an experimental manipulation systematically compresses these micro-intervals—shortening the pauses between processing steps—the observed “capacity” of the individual collapses, even if the total number of items presented remains completely unchanged.
This dynamic perspective replaced static structural containers with the concept of continuous temporal scheduling. In this paradigm, cognitive performance is governed by the interplay of two competing rates: the rate of informational degradation (decay rate) and the rate of executive restoration (refreshing rate). The human mind is viewed as a high-speed scheduler that must continuously negotiate the conflicting demands of processing novel perceptual inputs, executing central transformations, and dispatching refreshing pulses to fading memory traces. This evolution from static spatial metrics to continuous temporal dynamics allowed cognitive scientists to construct formal chronometric models that simultaneously account for individual differences, task difficulty, processing speed, and forgetting curves within a unified mathematical framework.
3. The Core Architecture: Attention as a Shared, Limited Resource
3.1 The Central Bottleneck Hypothesis
At the architectural core of the TBRS model stands the central bottleneck hypothesis, a construct historically rooted in the chronometric study of human multi-tasking and the psychological refractory period (PRP) formalisms established by Harold Pashler. The PRP effect demonstrates that when two separate stimuli requiring distinct, non-automatic responses are presented in rapid succession, the response time to the second stimulus is systematically delayed as the interval between the stimuli—the stimulus-onset asynchrony (SOA)—is reduced. This empirical delay occurs because while peripheral sensory perceptual processing and motoric output execution can proceed in parallel, central executive operations—such as response selection, decision-making, and rule-based transformations—are constrained by an absolute, single-channel structural bottleneck.
Barrouillet, Bernardin, and Camos (2004) extended Pashler’s central bottleneck hypothesis from the domain of immediate perceptual-motor coordination directly into the internal architecture of working memory maintenance. They argued that the covert maintenance of internal representations is not a passive, background operation delegated to isolated neurological buffers; rather, it is a central operation that relies on the identical executive mechanism required for computational processing. Consequently, the brain faces a strict biological serialization constraint: it cannot simultaneously engage in a conscious, response-selection processing operation and execute the attentional refreshing of a stored memory trace at the exact same millisecond.
This serialization constraint renders simultaneous conscious processing and active memory maintenance physically impossible at the micro-temporal scale. If a human subject is tasked with reading a word, judging whether an arithmetic equation is correct, or making a spatial discrimination, the single-channel central executive is captured entirely by that cognitive step. During this period of executive occupation, the pathway to memory maintenance is structurally occluded. Experimental verifications using chronometric dual-task paradigms have repeatedly substantiated this axiom, demonstrating that whenever an intervening task requires central decision-making—even when sensory and motor requirements are controlled—memory maintenance ceases entirely, triggering immediate representational decay.
3.2 Rapid Attentional Switching Dynamics
Because the central executive bottleneck enforces a strict regime of sequential processing, the human cognitive architecture must rely on rapid attentional switching to accomplish concurrent processing and storage. Rather than dividing its processing capacity into parallel streams, the cognitive system engages in what Barrouillet and Camos term micro-time-sharing. In this mode, the focus of attention rapidly shifts back and forth between executing steps of the processing task and refreshing stored memory items, taking advantage of brief micro-intervals, post-processing pauses, and inter-stimulus gaps that occur in the course of task execution.
These covert reallocations of attention occur across fractions of a second, operating within temporal windows ranging from 50 to 500 milliseconds. When an external stimulus appears (for instance, an arithmetic operation within an operation span task), the central executive directs attention toward resolving that problem. Once the computational step is completed or reaches an automatic plateau, attention is liberated. In the brief pause before the appearance of the next stimulus or processing requirement, the executive disengages from the processing task, targets the most degraded memory trace stored in activated memory, redirects the focus of attention onto it, and elevates its activation level. It then disengages to either refresh subsequent items in the retention set or re-engage with the next computational requirement of the primary task.
Crucially, this rapid switching is not cognitively free; it incurs quantifiable switching latencies and energetic costs. Disengaging the focus of attention from an external task set, redirecting it toward an internal memory representation, and subsequently retrieving and executing the original task set involves measurable cognitive overhead. Highly practiced, automatic processes impose minimal demands on this mechanism because they bypass the central bottleneck entirely, allowing memory maintenance to proceed unimpeded. Conversely, novel, non-automatic, or rule-shifting operations mandate intensive executive involvement, lengthening switching latencies, locking the central bottleneck, and severely restricting the cognitive system’s ability to execute restorative maintenance cycles.
3.3 The Functional Role of the Central Executive in TBRS
Within the TBRS framework, the central executive is neither a descriptive homunculus nor a passive storehouse; it is an active, dynamic supervisory engine that operates as a temporal scheduler and attentional coordinator. Its primary functional mandate is to continuously calibrate the allocation of central resources over millisecond timescales, balancing the competing demands of ongoing environmental processing against the internal imperative to preserve fragile memory traces from permanent decay.
One of the vital responsibilities of the central executive in this model is the precise temporal calibration and sequencing of task-switching routines. The executive must continuously monitor the degradation levels of items held in working memory while simultaneously tracking the arrival rate and processing requirements of external stimuli. If the executive miscalculates the duration of a micro-interval and directs attention to a memory trace just as an external stimulus requires urgent response selection, performance on the processing task deteriorates. Conversely, if it prioritizes external processing to the total exclusion of memory maintenance, the stored representations rapidly decay past the threshold of retrieval, causing catastrophic forgetting.
Furthermore, the central executive is responsible for the active suppression of task-irrelevant intrusions and internal interference during maintenance intervals. When attention is redirected toward internal memory traces, the executive must inhibit interference generated by recently processed stimuli, prevent proactive interference from prior experimental trials, and protect the focus of attention from task-irrelevant environmental distractions. Through this coordinated orchestration—scheduling switches, mitigating capture latencies, and insulating fragile traces—the central executive maintains the stability of the entire cognitive architecture.
4. The Mechanics of Attentional Refreshing and Trace Decay
4.1 Spontaneous Time-Based Trace Decay
A defining and hotly debated pillar of the TBRS model is the concept of spontaneous time-based trace decay. Barrouillet and Camos argue that working memory representations are inherently transient neurobiological patterns. The moment a perceptual item or semantic concept is disengaged from the conscious focus of attention, its corresponding memory trace begins to suffer continuous, time-dependent degradation. This decay occurs autonomously, driven by the thermodynamic and biological dissipation of transient synaptic activations, regardless of whether direct representational interference or feature overwriting is present.
From a computational perspective, the degradation of an unattended memory trace can be modeled as a continuous decay function wherein the activation level $A$ of an item $i$ decreases as a function of the elapsed unattended time $t$:
$$A_i(t) = A_{\max} – d \times t$$
where $A_{\max}$ represents the initial peak activation achieved when the item was actively processed within the focus of attention, and $d$ represents the parameter governing the baseline rate of autonomous temporal decay. If the unattended duration $t$ extends beyond a critical duration without restorative intervention, the trace’s activation level drops below a basal retrieval threshold ($A_{crit}$), rendering the representation inaccessible for subsequent conscious recall and redintegration.
The biological plausibility of this autonomous decay is supported by neurophysiological models of short-term synaptic plasticity and recurrent neural network dynamics. When an ensemble of neurons in prefrontal and posterior associative cortices fires to sustain a mental representation, this sustained firing is maintained via recurrent collateral excitation and transient intracellular calcium kinetics. In the absence of sustained, top-down re-excitation driven by executive attentional networks, these recurrent loops naturally attenuate due to synaptic adaptation, neurotransmitter depletion, and intrinsic membrane leak currents. Consequently, unattended traces inevitably slide down an entropic gradient toward resting baseline states—a biological reality that the TBRS model captures through its formalization of time-based decay.
4.2 The Mechanism of Attentional Refreshing
To counteract this relentless entropic decline, the TBRS architecture incorporates an active maintenance mechanism: attentional refreshing. Refreshing is defined as a non-articulatory, domain-general cognitive process whereby degraded memory traces are covertly reactivated by directing the central focus of attention onto them. Unlike domain-specific rehearsal systems that depend on acoustic or motoric codes, refreshing is an executive act of attentional refocusing that can be applied to any cognitive representation, whether it consists of a phonological word, an abstract geometric shape, a spatial location, or an emotional valence.
When the central focus of attention is directed onto a fading memory trace during an inter-stimulus pause or micro-gap, the trace undergoes immediate restoration. The model posits that during refreshing, the activation level of the memory item is restored toward its asymptotic maximum ($A_{\max}$) at a specific refreshing rate ($r$):
$$A_i(t_{refresh}) = A_i(t_0) + r \times t_{refresh}$$
Through this pulse of attentional energy, the decay process is momentarily reversed, lifting the item’s activation back into a secure zone well above the critical retrieval threshold. Because refreshing operates through the central bottleneck, it can only target one item (or chunk) at a time, requiring the executive to cycle sequentially through all active items within the memory set.
Attentional refreshing must be clearly distinguished from classical articulatory rehearsal. Whereas rehearsal relies on the motor planning and subvocal execution systems of speech production (anchored in Broca’s area and premotor pathways), refreshing is an endogenous, top-down attentional modulation that operates independently of the speech apparatus. Developmental and chronometric research demonstrates that attentional refreshing emerges reliably around the age of seven in human children, coinciding with the maturation of the prefrontal cortex and fronto-parietal control networks. Its operational speed and efficiency vary across individuals and development, serving as a primary determinant of individual differences in general fluid intelligence.
4.3 The Maintenance Dichotomy: Refreshing vs. Phonological Rehearsal
A major achievement of the TBRS framework has been the clarification of the human memory maintenance dichotomy, cleanly disentangling the unique, independent contributions of attentional refreshing and phonological rehearsal. While both mechanisms can maintain verbal information in working memory, they operate via entirely distinct neurocognitive pathways, rely on different computational rules, and exhibit differential vulnerability to task demands.
The structural autonomy of these two mechanisms is demonstrated through classical articulatory suppression paradigms. In standard complex span tasks where participants are required to continuously articulate an irrelevant verbal utterance (such as repeating “the, the, the” or “baba, baba”), the phonological loop is entirely engaged by motor-speech routines, preventing subvocal rehearsal. Traditional models predicted that this suppression should collapse verbal span down to baseline levels, particularly if the intervening processing task was non-verbal. However, TBRS experiments reveal that even under complete articulatory suppression, participants can maintain verbal items, provided that the cognitive load of the intervening task is sufficiently low to permit attentional refreshing micro-gaps. The residual memory performance observed under articulatory suppression is directly modulated by the cognitive load of the concurrent task, proving that attentional refreshing operates independently of speech production networks.
Conversely, when an intervening task imposes a massive cognitive load that completely captures central attention—leaving zero micro-gaps for attentional refreshing—participants can still maintain verbal traces if their phonological loop remains free to engage in rapid subvocal rehearsal. Dual-task dissociations have confirmed this reciprocal independence across multiple stimulus modalities. Non-verbal materials, such as abstract visual textures, tonal sequences, or spatial trajectories that cannot be readily translated into verbal phonemes, rely exclusively on attentional refreshing for their transient preservation. Human subjects strategically toggle between these two maintenance systems based on the nature of the materials and the structural constraints of the concurrent processing task, utilizing the phonological loop as an autonomous verbal offloader whenever domain-general attentional refreshing is monopolized by intensive executive processing.
5. Quantifying Cognitive Load: The TBRS Mathematical Metric
5.1 Derivation and Mathematical Formulation of Cognitive Load
Prior to the work of Barrouillet and Camos, the concept of “cognitive load” in experimental psychology was frequently criticized as a vague, circular construct: tasks were deemed to impose high cognitive load if they resulted in poor performance, yet poor performance was explained by appeal to high cognitive load. The TBRS framework resolved this tautology by deriving a mathematically rigorous, chronometrically objective definition of Cognitive Load (CL) grounded strictly in physical time parameters.
In the TBRS architecture, Cognitive Load is formally defined as the proportion of total time during which an intervening processing activity captures the central focus of attention, thereby denying access to restorative attentional refreshing. Mathematically, the cognitive load ($CL$) imposed by an intervening task is expressed through the seminal equation:
$$CL = \frac{\sum_{i=1}^{N} T_{capture, i}}{T_{total}} = \frac{T_p}{T_{total}}$$
where $T_{capture}$ (or $T_p$) represents the aggregate processing duration during which the central bottleneck is fully occupied by stimulus evaluation, response selection, and decision execution, and $T_{total}$ represents the total available time allotted for that processing phase (including all intervening pauses, blank screens, and inter-stimulus intervals).
Because cognitive load is expressed as a ratio, it is bounded between 0 and 1. A cognitive load approaching 0 represents an experimental condition where processing requirements are exceptionally brief or entirely absent, leaving nearly 100% of the retention interval available for attentional refreshing. Conversely, a cognitive load of 1.0 represents a saturated condition where the processing task continuously and completely monopolizes the central bottleneck from the onset to the termination of the interval, entirely eliminating any temporal micro-gaps for trace restoration. Across dozens of parametric experiments, Barrouillet, Camos, and their colleagues established that immediate memory recall ($M$) is a precise, inverse linear function of this cognitive load metric:
$$M = k – s \times CL$$
where $k$ represents the baseline structural or contextual capacity of the individual, and $s$ represents the linear slope of degradation across increasing cognitive loads.
5.2 Duration vs. Difficulty: Disentangling Cognitive Demands
The predictive power of the TBRS cognitive load metric was empirically demonstrated through a series of experiments designed to disentangle the traditional confound between task difficulty (or computational complexity) and task duration. In classical paradigms, researchers routinely assumed that complex cognitive operations—such as calculating multi-digit sums or executing complex logical transformations—depressed working memory performance because they consumed a specialized, non-temporal mental “fuel” or structural computational capacity. Barrouillet and Camos hypothesized instead that computational complexity is merely a proxy for processing time: complex tasks harm memory primarily because they require longer durations of uninterrupted attentional capture.
To test this hypothesis, Barrouillet, Bernardin, and Camos (2004) devised an experimental paradigm that systematically crossed task difficulty with task duration while measuring real-time response times. Participants were presented with memory items (consonants) interleaved with intervening processing tasks. In one condition, the intervening task was intrinsically simple (e.g., visual parity judgments: determining whether a single digit is odd or even), but it was presented at a rapid pace, forcing participants to execute judgments continuously with minimal intervening pauses. In another condition, the intervening task was cognitively more complex (e.g., solving multi-step mental arithmetic equations), but the presentation rate was relaxed, providing extended blank intervals after each response.
The empirical results challenged traditional modular assumptions: brief, simple tasks presented under tightly compressed temporal schedules imposed a higher cognitive load—and produced significantly greater memory loss—than intrinsically complex tasks accompanied by generous temporal micro-intervals. By precisely manipulating reaction times ($T_p$) and total interval durations ($T_{total}$), the researchers demonstrated that memory degradation is indifferent to the qualitative nature or perceived subjective difficulty of the cognitive operation. Rather, memory trace survival is dictated by the chronological proportion of time that the central bottleneck remains closed to attentional refreshing. Processing duration, not computational complexity per se, is the decisive causal determinant of forgetting in working memory span tasks.
5.3 The Cognitive Load Effect Across Diverse Domains
A critical test for any unified theory of cognitive architecture is its generalizability across different sensory modalities and cognitive domains. If the TBRS cognitive load metric reflected only the specific dynamics of verbal memory or linguistic processing, its value as a universal model of working memory would be limited. Consequently, extensive research programs were conducted to establish whether the linear decay function formalized by the TBRS model holds true across visual, spatial, and cross-modal complex span paradigms.
The empirical evidence confirmed broad cross-domain invariance. In spatial complex span experiments, participants maintained sequences of visuospatial coordinates (such as highlighted cells on a matrix or sequential dot locations) while performing intervening non-spatial tasks, such as reading aloud words or evaluating parity judgments. When cognitive load was parametrically varied by altering the presentation rate and attentional capture times of the verbal task, the spatial memory recall curves mapped onto the identical linear degradation slope observed in purely verbal tasks. Symmetrically, when verbal memory targets were interleaved with continuous spatial processing tasks (e.g., tracking a moving dot or judging the alignment of geometric figures), memory survival decayed as a direct function of the temporal load imposed by the spatial tasks.
This cross-domain invariance demonstrated that the linear relationship between cognitive load and trace survival is not an artifact of domain-specific interference, such as acoustic confusion or visual masking. Instead, it reveals the footprint of a domain-general central resource. The linear decay slope ($s$) remains consistent whether the stored targets are digits, abstract consonants, phonologically complex non-words, or visuospatial locations. These findings established that the cognitive load metric derived from the TBRS framework is a robust, universal chronometric index capable of predicting working memory performance across diverse cognitive tasks.
6. Experimental Paradigms and Empirical Validation in Complex Span Tasks
6.1 Standard Working Memory Span vs. TBRS Complex Span Protocols
The methodological revolution introduced by the TBRS framework is directly visible in the structural redesign of working memory assessment protocols. For decades, the standard experimental instruments used to quantify working memory capacity were complex span tasks based on the Daneman and Carpenter paradigm, including the Reading Span Task, the Operation Span Task, and the Counting Span Task. In these traditional paradigms, participants read sentences or solved equations, maintained final target words or letters in memory, and proceeded through the task at their own self-paced rhythm. However, from a chronometric perspective, these traditional self-paced protocols possessed a critical methodological flaw: they left temporal scheduling entirely unconstrained.
Under self-paced conditions, participants routinely adopt idiosyncratic cognitive strategies. High-performing individuals often deliberately pause for hundreds of milliseconds after completing an arithmetic problem or reading a sentence before triggering the appearance of the next stimulus. During these unmonitored free-time pauses, they deploy attentional refreshing, rehearse phonological codes, or organize items into strategic semantic chunks. Conversely, other participants rush through the processing steps, inadvertently creating high cognitive load conditions for themselves. Consequently, traditional complex span tasks often measured an unstandardized mixture of individual differences in processing speed, strategy selection, and chronometric time management, rather than pure working memory capacity.
To eliminate these confounding artifacts, Barrouillet and Camos developed TBRS-specific computer-paced complex span protocols. In these paradigms (such as the computer-paced Parity Span or Reading Span tasks), every discrete processing step is rigorously controlled by the experimental software. Visual stimuli appear on the computer screen for precise durations (e.g., 800 ms), followed by fixed, standardized intervals. Participants are required to vocalize or key in their responses within strict response windows. By standardizing the duration of stimulus exposure, recording exact reaction times for each discrete processing step, and regulating the remaining duration of post-processing intervals down to the millisecond, TBRS protocols eliminate unmonitored free time. This level of chronometric control successfully prevents idiosyncratic chunking and strategic rehearsal, isolating the raw time-based resource-sharing dynamics of the cognitive system.
6.2 Empirical Verification of Micro-Intervals and Attentional Pauses
The central operational claim of the TBRS model—that memory traces are actively restored during micro-gaps occurring between processing steps—has been subjected to direct empirical verification using chronometric, ocular, and psychophysiological techniques. If attentional refreshing occurs covertly within post-processing pauses, then systematically expanding these pauses while keeping the physical processing requirements strictly identical should produce immediate, measurable recoveries in memory retention.
Chronometric studies systematically manipulating the stimulus-onset asynchrony (SOA) provided early causal validation of this principle. In these experiments, participants were required to perform an identical number of computational operations (e.g., judging whether $4 + 3 = 7$). However, the inter-stimulus interval following each calculation was systematically varied from 100 milliseconds to 1500 milliseconds. Across thousands of experimental trials, memory recall increased as a monotonic function of the cumulative free time provided by these micro-intervals. Even when an intervening processing operation was exceptionally demanding, inserting brief, recurring pauses of 200 to 400 milliseconds between successive operations allowed the central executive to switch attention back to the decaying traces, elevating their activation levels and preventing memory collapse.
Further empirical verification has been obtained through high-speed eye-tracking and pupillometry methodologies. When participants engage in computer-paced complex span tasks, their gaze patterns and pupillary responses provide objective, real-time markers of attentional reallocation. During the execution of an intervening arithmetic task, the pupil dilates significantly, reflecting intensive mental effort and the capture of the central bottleneck. However, during subsequent micro-pauses, pupil diameter exhibits a characteristic, partial constriction coupled with micro-saccadic suppression—a physiological signature corresponding to the disengagement of attention from external sensory inputs and its internal redirection toward endogenous memory representations. These ocular dynamics confirmed that the cognitive system actively utilizes micro-pauses to deploy restorative attentional refreshing.
6.3 Cross-Modality Empirical Findings
To confirm that the operational mechanics of the TBRS model are mediated by a shared central resource rather than modality-specific structural channels, Barrouillet and Camos conducted extensive cross-modality experiments. These studies paired storage items from one sensory-representational domain with intervening processing tasks from a completely disparate domain, systematically tracking whether interference patterns followed temporal cognitive load equations rather than structural similarity metrics.
In one series of cross-modal experiments, participants were instructed to maintain sequences of phonologically balanced verbal consonants while the intervening processing task consisted of a continuous spatial choice-reaction task (e.g., determining whether an arrow flashed on the screen pointed upward or downward, or tracking a moving dot on a grid). Standard modular working memory theory predicted that because verbal storage relies on the phonological loop and spatial choice reactions rely on visuospatial mechanisms, the two streams should proceed in near-total autonomy, resulting in minimal interference. In reality, the results confirmed the TBRS predictions: as the temporal presentation rate of the spatial arrows was accelerated—increasing the spatial task’s cognitive load by consuming larger proportions of available time—verbal recall performance declined along the identical linear decay function observed when verbal processing tasks were used.
Reciprocally, when participants maintained sequences of abstract visual designs or spatial matrix locations while performing purely auditory or verbal verification tasks (e.g., verifying spoken mathematical facts), the degradation of spatial memory was strictly determined by the cognitive load of the auditory processing stream. The cross-domain interference was found to be statistically robust across an extensive variety of experimental stimuli, including:
- Printed and auditory digits
- Phonologically similar versus dissimilar consonants
- Non-sense pseudo-words
- Abstract geometrical figures lacking verbal labels
- Unpredictable visuospatial path configurations
The cross-modal invariance of these findings provided empirical proof that the central interference observed in complex span tasks does not arise from structural representational overlap within sensory buffers. Rather, it is mediated by the continuous competition for a singular, domain-general pool of central attention, the allocation of which is governed strictly by the temporal parameters of the task environment.
7. TBRS vs. Competing Working Memory Models: A Comparative Analysis
7.1 Comparison with Baddeley and Hitch’s Multi-Component Model
The structural divergence between the Time-Based Resource Sharing framework and Baddeley and Hitch’s multi-component model centers on their contrasting views of cognitive modularity, the nature of storage buffers, and the mechanism of central control. Baddeley’s architecture is fundamentally modular and structural: it posits that working memory consists of physically or functionally segregated storage buffers (the phonological loop, the visuospatial sketchpad, and the episodic buffer) coordinated by a central executive. In contrast, the TBRS model is fundamentally process-oriented, chronometric, and unitary with respect to its central attentional resource, viewing maintenance not as passive residence within dedicated structural bins, but as the active, temporal scheduling of attention across decaying neurocognitive states.
This theoretical divergence is sharpest when evaluating the autonomy of the phonological loop. In Baddeley’s model, verbal items can be maintained indefinitely within the phonological store via the automated, circular mechanism of the phonological loop, operating without continuous supervision from the central executive. The TBRS model fundamentally disputes this supposed autonomy. While acknowledging that subvocal articulatory rehearsal can run as an auxiliary motor loop when permitted, TBRS demonstrates that verbal items remain continuously dependent on domain-general central attention. If the cognitive load of an intervening non-verbal task is elevated to 1.0, verbal traces decay and are lost even if the articulatory apparatus is theoretically unoccupied, because participants cannot access the central executive resources required to initiate and coordinate retrieval strategies.
Furthermore, the models differ fundamentally in how they explain the integration of cross-modal information. To account for how verbal, visual, and long-term memory representations are combined into cohesive mental episodes, Baddeley introduced the episodic buffer—a passive storage unit with limited structural capacity. TBRS rejects the necessity of positing an additional structural buffer. Instead, it conceptualizes cross-modal integration as a dynamic, emergent temporal process: disparate features residing in activated long-term memory are temporarily bound and maintained through the rapid, cyclical deployment of the single-channel focus of attention. Where Baddeley posits structural containers to house mental representations, Barrouillet and Camos demonstrate that dynamic temporal scheduling across a central bottleneck is sufficient to explain the data.
7.2 Comparison with Cowan’s Embedded-Processes and Oberauer’s Concentric Models
The relationship between the TBRS model and other attention-based architectures—notably Nelson Cowan’s embedded-processes model and Klaus Oberauer’s concentric working memory model—is one of shared epistemological lineage coupled with critical mechanistic divergences. All three frameworks reject the proliferation of isolated peripheral buffers, conceptualizing working memory as hierarchical layers of attentional activation operating over the broad expanse of long-term memory representations. However, they disagree fundamentally regarding the operational bandwidth of the focus of attention and the primary mechanisms driving memory degradation.
The primary point of contention with Cowan’s model centers on the capacity of the focus of attention (FoA). Cowan maintains that the human focus of attention is a multi-item spotlight capable of simultaneously encompassing approximately three to four distinct information chunks. The TBRS model, alongside Oberauer’s concentric architecture, firmly rejects this multi-item view. Barrouillet and Camos argue that the central bottleneck identified in psychological refractory period experiments limits the conscious focus of attention to an indivisible, single-item capacity at any precise millisecond. TBRS demonstrates that experimental findings suggesting a multi-item focus of attention are macroscopic illusions produced by the rapid, high-frequency time-sharing of a single-channel attentional beam cycling sequentially across candidate items.
In contrast to Oberauer’s concentric model—which posits a broad region of direct access (holding approximately three to four items bound to context) containing a singular, one-item focus of attention—the TBRS model diverges primarily regarding the causal mechanism of forgetting. Oberauer’s framework dismisses time-based decay entirely, attributing all working memory loss to interference, specifically feature overwriting and cue overload within the direct-access region. The TBRS model, while acknowledging the reality of interference, insists that spontaneous, autonomous time-based decay is a physical and biological reality that operates continuously whenever attention is withdrawn. While Oberauer views interference as the sole driver of cognitive breakdown, TBRS maintains that temporal decay and attentional refreshing cycles are the primary engines governing working memory capacity.
7.3 Comparison with Resource-Sharing Models of Just and Carpenter
During the early 1990s, Marcel Just and Patricia Carpenter introduced a highly influential capacity-sharing framework of working memory, designed primarily to explain computational and storage dynamics in language comprehension. Just and Carpenter posited that working memory consists of a single, continuous pool of activation resources—often conceptualized as a finite energetic or metabolic capacity—that can be continuously divided and shared simultaneously between concurrent processing demands and concurrent storage needs. If a task requires heavy linguistic computation, the pool of activation is diverted toward processing, leaving less remaining activation for the passive maintenance of storage representations.
The TBRS model fundamentally rejects the concept of simultaneous, fractional resource sharing. Barrouillet and Camos argue that the cognitive architecture is physically incapable of dividing its central executive capacity into parallel, simultaneous fractions at the millisecond level. The brain cannot dedicate, for instance, 60% of its central attention to solving an equation while simultaneously allocating 40% of that same attention to maintaining three consonants. Drawing upon chronometric paradigms and neurophysiological evidence of the central bottleneck, TBRS demonstrates that resource sharing is strictly temporal rather than quantitative.
In the TBRS framework, the system is always allocated 100% to a single cognitive target at any given millisecond: it is either 100% focused on processing the external stimulus, or 100% focused on refreshing a memory trace. What appears to be simultaneous resource splitting in Just and Carpenter’s macro-level linguistic experiments is revealed at the micro-level to be rapid, sequential time-sharing. This distinction is of vital theoretical importance: while Just and Carpenter’s capacity-sharing model lacks predictive precision regarding temporal pacing and response chronometry, the TBRS model leverages its temporal serialization axiom to derive exact mathematical predictions of memory performance based directly on physical time parameters.
8. The Great Working Memory Debate: Decay vs. Interference
8.1 The Anti-Decay Critique: Lewandowsky and Oberauer’s Arguments
The inclusion of autonomous, time-based trace decay as a foundational pillar of the TBRS model ignited one of the most intense and sustained theoretical controversies in contemporary cognitive science: the great Decay versus Interference debate. Led primarily by Stephan Lewandowsky and Klaus Oberauer, critics of decay theory argued that spontaneous, autonomous temporal decay is an unfalsifiable, scientifically unproductive, or empirically non-existent construct. Drawing on a century of research descending from McGeoch (1932), the anti-decay camp asserted that time itself cannot act as a causal physical agent of forgetting, just as time alone does not cause iron to rust (oxidation being the true causal agent).
Lewandowsky, Oberauer, and their colleagues launched a series of empirical investigations designed to demonstrate that memory retention over unfilled retention intervals is remarkably invariant, provided that interference is rigorously controlled. In a famous series of complex span experiments, Lewandowsky and colleagues manipulated the duration of intervening tasks by varying the display duration or articulation rate of intervening materials, claiming to show that prolonging blank delays did not induce memory loss. They posited that working memory failure is caused entirely by interference mechanisms, which they categorized into three primary forms:
- Feature Overwriting: The perceptual or semantic features of intervening processing items overwrite or displace the features belonging to the stored memory representations within a shared feature space.
- Cue Overload: As additional stimuli are processed, the retrieval cues associated with the target items become associated with multiple competing targets, reducing the probability of successful target access at retrieval.
- Retroactive and Proactive Interference: Direct conflict emerging from prior experimental trials or subsequently presented information that blurs the distinctiveness of positional context bindings.
Based on these findings, the anti-decay camp argued that apparent decay effects in earlier studies were methodological artifacts resulting from failures to control for feature similarity, structural interference, or covert rehearsal strategies during supposedly unfilled retention intervals.
8.2 The Counter-Arguments: How Barrouillet and Camos Defend Temporal Decay
Pierre Barrouillet, Valérie Camos, and their collaborators responded to the anti-decay critique with theoretical refinements and a comprehensive series of chronometrically controlled counter-experiments. They demonstrated that the experiments marshaled by Lewandowsky and Oberauer suffered from a fundamental methodological flaw: the failure to fully occupy central attention during the manipulated temporal intervals.
Barrouillet and Camos pointed out that when an experimenter simply inserts a “prolonged blank delay” into a memory task without engaging the participant in an attention-demanding secondary task, the participant does not sit idly in a state of cognitive arrest. Instead, the participant immediately and covertly exploits this free, unmonitored time to execute attentional refreshing and phonological rehearsal. Thus, the finding that memory performance does not decline across long, unfilled retention intervals does not prove the absence of decay; rather, it demonstrates that participants successfully use the available free time to refresh and restore decaying traces back to peak activation. To observe pure decay, central attention must be continuously captured by a non-interfering task, thereby preventing attentional refreshing without introducing feature overlap.
To provide definitive empirical proof, Barrouillet, Portrat, and Camos (2011) designed experiments that held the number and qualitative nature of intervening stimuli entirely constant—thereby keeping feature overwriting and cue overload perfectly fixed—while parametrically varying only the duration of the capture intervals. The empirical results were decisive: even when feature overlap was kept completely identical, increasing the duration of attentional capture produced severe, monotonic reductions in recall accuracy. Furthermore, Barrouillet and Camos established that temporal decay is a biological and computational necessity for the human cognitive apparatus. Without an autonomous decay mechanism, transient information would linger permanently in working memory, producing catastrophic cognitive saturation and overwhelming the system with outdated, proactive interference from prior tasks.
8.3 Current State of the Consensus: Hybrid Perspectives
Following decades of vigorous debate, contemporary cognitive science has largely converged toward a nuanced, hybrid consensus that synthesizes the valid empirical insights of both the decay and interference camps. The scientific community increasingly recognizes that debating whether working memory failure is caused exclusively by decay or exclusively by interference represents a false dichotomy. In realistic cognitive environments, decay and interference operate as inextricably linked, complementary mechanisms of forgetting.
Modern revisions of the TBRS architecture, alongside updated computational implementations of competing models, acknowledge that memory representations consist of both item features (phonological, visual, semantic codes) and context-binding signals (associations linking an item to its specific temporal or serial position). The contemporary consensus suggests that while feature representations may indeed suffer from interference, feature overwriting, and cross-talk when similar stimuli are encountered, the *bindings* that anchor those features to their specific temporal and serial coordinates undergo rapid, continuous degradation in the absence of top-down attentional maintenance.
Consequently, the TBRS model has evolved to accommodate both forces within a unified framework: autonomous temporal decay constantly lowers the baseline activation of unattended representations, increasing their vulnerability to ambient noise and representational interference. If a trace’s activation drops significantly due to prolonged attentional diversion (decay), its distinctiveness is diminished, making it far more susceptible to being overwritten or displaced by competing features (interference). Attentional refreshing serves as the vital counter-force, elevating degraded representations back into high-activation states where their structural integrity and contextual bindings are shielded against both temporal entropic decay and retroactive interference.
9. Computational Modeling: Simulating TBRS Architecture and TBRS*
9.1 The Formal Computational Architecture: TBRS*
To transition the TBRS framework from a conceptual, verbal-mathematical model into an explicit, algorithmic system capable of generating quantitative, trial-by-trial simulations of human cognition, Pierre Barrouillet, Valérie Camos, and computational modelers formulated TBRS*. Published in seminal papers by Oberauer and Lewandowsky (2011) and further refined by Barrouillet and colleagues, TBRS* provides a fully implemented computational architecture that simulates the exact micro-temporal mechanics of attentional capture, trace decay, task switching, and refreshing cycles.
In TBRS*, memory representations are formalized as vectors of activation residing within an activated memory space. Each memory item $i$ possesses a continuous activation value $A_i$ bounded between 0 and 1. When an item is presented visually or aurally, its activation is initialized to $A_{\max} = 1.0$. The temporal dimension is operationalized through discrete, millisecond-level time steps. During each millisecond that the central executive bottleneck is occupied by an external processing task (such as evaluating an equation), the activation of every unattended memory item decays according to a continuous exponential or linear decay function governed by a decay parameter $d$:
$$\Delta A_i = -d \times \Delta t$$
Whenever the processing task releases the central bottleneck, an algorithmic scheduling module checks for the presence of available free time. If the duration of the micro-gap exceeds the executive switching latency parameter ($S_{cost}$), the architecture initiates an attentional refreshing cycle. The algorithmic core of TBRS* schedules refreshing by directing the focus of attention onto the item currently possessing the lowest activation level above the critical threshold, restoring its activation at a standardized refreshing rate parameter $r$:
$$\Delta A_i = +r \times \Delta t$$
Through this explicit computational loop, TBRS* simulates the precise, real-time dynamics of human cognitive performance, tracking parameter-dependent activation curves as they fluctuate across the experimental timeline.
9.2 Simulating Serial Order and Retrieval Mechanisms
A critical requirement for any complete computational model of working memory is the ability to account for serial order retention and the generation of realistic error typologies. In human working memory experiments, participants do not merely forget items; they commit systematic, predictable errors, including *transpositions* (recalling the correct item in the wrong serial position), *omissions* (failing to recall an item entirely), and *intrusions* (recalling an item from an earlier list or an intervening processing step). TBRS* accounts for these phenomena through an integrated positional-activation mechanism.
In the TBRS* architecture, serial order is preserved through dynamic position markers that bind each memory item to an abstract positional gradient (e.g., Position 1, Position 2, Position 3). The strength of the binding between an item vector and its corresponding position marker is directly proportional to the item’s instantaneous activation level ($A_i$). When the computational system enters a free micro-interval and must decide which item to refresh first, it executes a priority algorithm. TBRS* implements this priority selection through two primary competing rules:
- First-In, First-Refreshed (Cyclical Queue): The system iterates sequentially through the items in their original chronological order of presentation, mimicking the rhythmic cadence of natural human rehearsal.
- Urgency-Based Refreshing (Lowest-Activation-First): The system directs the focus of attention onto the item whose activation level is closest to the critical decay threshold, preventing immediate catastrophic loss.
At the retrieval phase, the model executes a noisy, competitive retrieval process formalized via a Luce choice or softmax selection rule. The probability $P$ of successfully retrieving item $i$ at its correct serial position marker $j$ is calculated as a function of its relative activation compared to all competing items and background noise:
$$P(i | j) = \frac{\exp(A_{i, j} / \tau)}{\sum_{k} \exp(A_{k, j} / \tau) + \exp(Threshold / \tau)}$$
where $tau$ represents a stochastic neural noise parameter. If an item’s activation has decayed below the threshold, an omission error occurs. If temporal decay has weakened positional bindings, a neighboring item may win the competitive retrieval selection, generating a classic transposition error. Finally, TBRS* incorporates a redintegration process: if a retrieved representation has suffered partial decay of its individual feature nodes, the cognitive system queries long-term memory to reconstruct the degraded trace, successfully modeling human phonological and semantic reconstruction effects.
9.3 Validation and Predictive Power of Computational Simulators
The scientific validity of the TBRS* computational architecture has been established through extensive simulation studies that pit synthetic performance data directly against large-scale empirical human datasets. By optimizing a minimal set of free parameters—namely, decay rate ($d$), refreshing rate ($r$), switching latency ($S_{cost}$), and retrieval noise ($tau$)—the TBRS* computational simulator has demonstrated extraordinary predictive accuracy across a wide range of complex span conditions.
A benchmark validation of TBRS* was its ability to reproduce human complex span performance curves across varying processing loads. When the simulator was subjected to synthetic complex span experiments matching the exact chronometric protocols of Barrouillet, Bernardin, and Camos (2004), the resulting synthetic recall scores mirrored human empirical data with correlation coefficients frequently exceeding $R^2 = 0.95$. The simulator accurately reproduced the linear degradation slope relating cognitive load to recall accuracy, confirming that the simple mathematical interaction between time-based decay, single-channel bottlenecking, and attentional refreshing is sufficient to generate human-like working memory limits.
Furthermore, TBRS* demonstrated high predictive power when simulating individual differences across the human population. By systematically modulating baseline parameters within the model—such as slowing processing speed (which artificially increases attentional capture time $T_p$) or lengthening the switching latency ($S_{cost}$)—the simulation reproduced the empirical performance profiles characteristic of developmental children, older adults, and clinical cohorts. Crucially, the model successfully simulated catastrophic performance collapse as the cognitive load metric approached 1.0, capturing the physical breakdown of the human cognitive system when temporal micro-gaps are entirely eliminated.
10. Developmental Trajectories: Working Memory Maturation Across the Lifespan
10.1 Ontogeny of Attentional Refreshing in Childhood
The Time-Based Resource Sharing framework has provided profound insights into the developmental psychology of executive functions and transient memory. A central developmental question within cognitive science has long been: why does working memory capacity expand so dramatically between early childhood and late adolescence? Traditional accounts attributed this growth to the mysterious expansion of static storage buffers or the spontaneous emergence of the phonological loop around age seven. The TBRS model revolutionizes this developmental narrative by mapping memory maturation directly onto the ontogeny of chronometric processing speed and the spontaneous emergence of attentional refreshing.
Through pioneering developmental investigations, Valérie Camos and Pierre Barrouillet documented a crucial chronological milestone: attentional refreshing does not operate as an innate, fully operational mechanism from infancy, but emerges spontaneously around the age of seven. In experiments with children aged four to six, Barrouillet and colleagues observed that manipulating the cognitive load of an intervening task by inserting micro-pauses produced negligible variations in memory recall. Young children were unable to capitalize on available free time; when an intervening task ceased, their focus of attention remained passive, leaving unattended memory traces to decay rapidly. Consequently, young children rely almost exclusively on passive storage, resulting in low working memory spans.
Around age seven, however, a fundamental neuro-developmental transition occurs. Coinciding with the functional maturation of the prefrontal cortex and fronto-parietal control networks, children spontaneously develop the executive capability for endogenous attentional reallocation. At this age, children begin to actively exploit micro-gaps between processing operations, deploying their central focus of attention to refresh fading representations. From this developmental junction onward, children exhibit the classical TBRS cognitive load effect: their recall performance becomes a direct linear function of the temporal ratio between processing duration and available free time. The developmental expansion of working memory capacity across primary education is thus primarily driven by the emergence, optimization, and automation of this active, time-sharing refreshing routine.
10.2 The Developmental Mechanics: Speed of Processing vs. Executive Control
As children mature from age seven into adolescence, their working memory capacity continues to exhibit substantial, linear increases. Within the TBRS framework, this developmental trajectory is unpacked into two interacting chronometric mechanisms: the developmental growth of pure processing speed and the refinement of executive task-switching efficiency.
The primary driver of working memory expansion across childhood is the age-related acceleration of pure cognitive processing speed. As the central nervous system matures through extensive axonal myelination, synaptic pruning, and the establishment of high-velocity white matter tracts, children can execute cognitive operations at much faster rates. In an operation span or reading span task, an older child executes the identical arithmetic calculation or word verification in a fraction of the time required by a younger child. Within the TBRS cognitive load equation:
$$CL = \frac{T_p}{T_{total}}$$
this acceleration in processing speed dramatically reduces the duration of attentional capture ($T_p$). Because the processing task captures the central bottleneck for a significantly smaller window of time, the duration of the remaining micro-intervals within $T_{total}$ is expanded. Consequently, older children liberate larger reserves of unattended time, allowing their central executive to execute multiple, deep attentional refreshing cycles that protect stored traces from dropping below the retrieval threshold.
Simultaneously, the developmental trajectory is shaped by the maturation of executive control parameters, specifically the reduction of switching latencies ($S_{cost}$). Younger children exhibit considerable cognitive inertia: disengaging attention from an external perceptual task and redirecting it toward an internal memory representation is a slow, effortful process. As executive control matures throughout late childhood, switching latencies drop from several hundred milliseconds down to optimized adult baselines. This enables older children to initiate refreshing routines within extremely narrow micro-pauses that younger children cannot exploit. This combined developmental optimization—faster processing freeing up time, coupled with faster switching utilizing that time—directly underpins the maturation of academic learning, complex reading comprehension, and advanced mathematical proficiency.
10.3 Cognitive Senescence and the Decline of Time-Sharing Capabilities
At the opposite end of the human lifespan, the TBRS framework provides an explanatory architecture for understanding the decline of working memory capacity during healthy cognitive aging and neurodegenerative senescence. Historically, age-related deficits in complex span tasks were attributed to vague, global reductions in “mental energy” or generalized frontal lobe degradation. The TBRS model provides a chronometrically precise account of these senescence trajectories.
The primary engine driving working memory decline in older adults is age-related cognitive slowing, a neurobiological reality that triggers an artificial inflation of cognitive load. As individuals age, reductions in white matter integrity, neurotransmitter receptor density (particularly dopaminergic systems), and neural conduction velocity result in the systematic lengthening of processing times. When an older adult performs an intervening operation within a complex span task, the duration of attentional capture ($T_p$) is prolonged. Consequently, even when an experimental protocol presents stimuli at a pace that provides ample free time for young adults, that identical physical timeframe becomes saturated for an older adult. The cognitive load metric ($CL$) inflates toward 1.0, effectively eliminating the temporal micro-gaps required to execute attentional refreshing.
Furthermore, cognitive senescence is accompanied by a severe degradation in attentional switching dynamics. Neuroimaging and chronometric data reveal that older adults suffer from heightened switching costs and lengthened executive latencies when transitioning between external perceptual processing and internal trace maintenance. In many instances, the micro-pauses available between processing steps fall below the elevated switching threshold of the senescent brain, rendering older adults structurally incapable of deploying refreshing pulses in rapidly paced environments. To mitigate these catastrophic storage deficits, older individuals frequently adopt compensatory strategies, such as relying more heavily on passive, domain-specific phonological loops or shifting toward semantic gist representations, to offset their declining ability to execute rapid, dynamic time-sharing.
11. Neurocognitive Correlates and Functional Substrates of TBRS
11.1 Neural Substrates of the Central Attentional Bottleneck
The behavioral axioms of the TBRS model—particularly the central bottleneck and the rapid switching of executive attention—map onto specific, large-scale neuroanatomical networks within the human brain. Functional neuroimaging (fMRI) and lesion studies have firmly established that the single-capacity central bottleneck is grounded within the fronto-parietal control network (FPN), with critical epicenters located in the dorsolateral prefrontal cortex (dlPFC) and the posterior parietal cortex.
The dorsolateral prefrontal cortex, particularly within the middle frontal gyrus, serves as the neurobiological engine of executive task-switching and temporal scheduling. When participants transition from active stimulus processing to internal memory maintenance within complex span paradigms, the dlPFC exhibits intense, transient metabolic activation corresponding to the disengagement and redirection of the focus of attention. Concurrently, the anterior cingulate cortex (ACC), situated on the medial surface of the frontal lobes, acts as a critical conflict-monitoring and capacity-allocation hub. When the cognitive load of a task approaches high saturation levels—generating intense competition between incoming perceptual inputs and fading memory traces—the ACC fires robustly, signaling the necessity for top-down cognitive control to resolve bottleneck interference.
The posterior anchor of this bottleneck architecture resides in the intraparietal sulcus (IPS) and superior parietal lobule. The IPS is heavily implicated in the top-down spatial and non-spatial selection of mental representations. Neuroimaging experiments isolating the central processing bottleneck during dual-task execution reveal that when two tasks demand central decision-making within overlapping temporal windows, the fronto-parietal control network exhibits an invariant, serialized activation profile: the neural circuitry processing the second task is held in an enforced state of metabolic latency until the fronto-parietal network completes the response-selection computations for the first task. This provides direct neurobiological confirmation of the single-channel central processing bottleneck postulated by Barrouillet and Camos.
11.2 Electrophysiological Markers of Decay and Attentional Refreshing
While hemodynamic neuroimaging (fMRI) provides precise anatomical localization, its poor temporal resolution (on the order of seconds) is insufficient to capture the rapid, sub-second dynamics of the TBRS model. To track the temporal signature of attentional redirection, decay, and refreshing occurring within micro-intervals of 100 to 500 milliseconds, cognitive neuroscientists turn to high-density electroencephalography (EEG) and event-related potentials (ERPs).
A premier electrophysiological index of working memory maintenance is the Contralateral Delay Activity (CDA)—a sustained, negative-polarity event-related potential observed over posterior parietal electrodes during the retention interval of unilateral visual arrays. The amplitude of the CDA tracks the number of individual representations currently held in an active state. In chronometric studies designed around TBRS protocols, researchers have documented that when the focus of attention is captured by an intervening processing task, the CDA amplitude undergoes a continuous, time-dependent attenuation. This electrophysiological decay directly reflects the autonomous, neurobiological degradation of unattended memory traces predicted by the model.
Crucially, high-density EEG has isolated the electrophysiological signature of attentional refreshing occurring during post-processing micro-intervals. Time-frequency spectral analyses reveal that the execution of an attentional refresh is signaled by a burst of frontal midline theta oscillations (4–8 Hz) coupled with transient posterior alpha desynchronization (8–12 Hz). The frontal theta burst corresponds to the top-down executive command emitted by the prefrontal cortex to retrieve a fading representation, while the posterior alpha desynchronization reflects the visual cortex being released from inhibition to render the refreshed memory trace active once more. These electrophysiological dynamics occur within micro-pauses as brief as 200 milliseconds, confirming that the brain exploits fleeting temporal pauses to reactivate fading neurochemical traces.
11.3 Dissociating Neural Networks for Refreshing and Articulatory Rehearsal
The theoretical claim that attentional refreshing and phonological rehearsal represent structurally distinct maintenance mechanisms has received strong empirical validation from functional neuroimaging and cognitive neuropsychology. By contrasting brain activation patterns during complex span tasks with and without articulatory suppression, cognitive neuroscientists have dissociated the neural networks underlying these two cognitive functions.
Functional neuroimaging studies conducted by Raye, Johnson, Mitchell, Greene, and Camos (2007) established this double dissociation. When human participants maintain verbal items via phonological rehearsal, the blood-oxygen-level-dependent (BOLD) signal is localized within the classical speech production and phonological processing loop: Broca’s area (left inferior frontal gyrus, Brodmann areas 44/45), the left premotor cortex (Brodmann area 6), and the left posterior superior temporal gyrus (Wernicke’s area). In sharp contrast, when participants maintain items via attentional refreshing—holding identical verbal stimuli under conditions of complete articulatory suppression—the speech-production network falls completely silent. Instead, neural metabolic activity shifts entirely to a domain-general fronto-parietal network consisting of the left middle frontal gyrus (dorsolateral prefrontal cortex, BA 9/46) and the bilateral superior parietal lobules.
This functional dissociation is further corroborated by targeted repetitive Transcranial Magnetic Stimulation (rTMS) and clinical neuropsychology. Applying inhibitory rTMS over Broca’s area severely disrupts articulatory rehearsal speeds but leaves the linear cognitive load effect mediated by attentional refreshing completely intact. Reciprocally, applying inhibitory rTMS over the left dorsolateral prefrontal cortex leaves speech articulation and simple phonological span unaffected, but severely impairs the participant’s ability to refresh items across temporal micro-intervals during complex span tasks. In clinical cohorts, patients with focal strokes localized to the left inferior frontal cortex exhibit profound articulatory rehearsal deficits while retaining normal attentional refreshing dynamics, whereas patients with focal lesions in the dorsolateral prefrontal cortex show the inverse deficit. These findings confirm that refreshing and rehearsal are mediated by neurobiologically autonomous brain systems.
12. Contemporary Challenges, Applications, and Future Directions for TBRS
12.1 Translational Applications in Educational and Occupational Psychology
The mathematical and operational insights of the Time-Based Resource Sharing model extend far beyond the laboratory, offering powerful translational applications across educational design, classroom pedagogy, and high-stress occupational ergonomics. Foremost among these applications is the radical optimization of instructional pacing in educational curricula.
Traditional instructional design frameworks, such as Sweller’s Cognitive Load Theory, historically categorized cognitive load into intrinsic, extraneous, and germane forms, often lacking precise temporal metrics for real-time classroom delivery. The TBRS model provides educators and software designers with an objective, chronometric engineering formula. In modern digital learning environments, presentation pacing can be dynamically modulated to prevent student working memory collapse. By breaking complex pedagogical explanations into discrete instructional modules interleaved with mandatory, computer-enforced micro-pauses (300–600 ms), educational software allows the student’s central executive to refresh newly formed concepts. This prevents immediate temporal decay, drastically improving reading comprehension and mathematical learning outcomes, particularly for children with developmental working memory deficits and Attention-Deficit/Hyperactivity Disorder (ADHD).
In occupational and human factors psychology, the TBRS framework has reshaped the design of high-stress human-machine interfaces, such as those found in aviation cockpits, air traffic control centers, and nuclear power plant operations. In these critical environments, human operators are routinely overwhelmed not by the sheer intellectual complexity of individual decisions, but by continuous, high-density perceptual alerts that completely saturate available processing time ($CL to 1.0$). By applying the TBRS cognitive load equation to interface telemetry, human factors engineers can design automated pacing algorithms that dynamically reschedule secondary alerts during critical operational windows. This preserves the essential micro-gaps human operators require to maintain situational awareness representations in working memory, preventing catastrophic human error.
12.2 Open Theoretical Questions and Unresolved Controversies
Despite its extensive empirical successes and mathematical elegance, the Time-Based Resource Sharing framework faces several open theoretical questions, unresolved controversies, and boundary conditions that occupy contemporary cognitive science. Primary among these questions is the exact physical and representational nature of the decaying trace.
Within current formulations of the model, trace decay is frequently conceptualized as a continuous, analog decline in general activation or signal-to-noise ratio. However, competing cognitive architectures argue that memory representations do not fade like dimming lightbulbs; rather, they suffer from discrete, stochastic feature loss, wherein individual phonological, visual, or contextual feature nodes drop out abruptly due to cellular noise or quantum metabolic variations. Determining whether the decaying trace undergoes continuous analog decay or discrete all-or-none feature disintegration remains an active, intensely pursued empirical frontier requiring ultra-high-resolution electrophysiological recording.
A second major theoretical challenge involves the scalability of the TBRS model to highly complex, naturalistic, multi-modal environments. While the model functions with near-flawless predictive accuracy in computer-paced laboratory tasks using simple, discrete stimuli (digits, consonants, parity judgments), real-world cognition frequently involves continuous, fluid sensory streams—such as navigating a vehicle through heavy traffic while participating in a conversation. Modeling how the central bottleneck and attentional refreshing cycles operate over continuous, analog perceptual inputs that do not possess clear-cut inter-stimulus micro-gaps represents a complex theoretical endeavor.
Finally, researchers are actively investigating how affective, motivational, and emotional factors integrate into the TBRS cognitive load equation. Intrusive emotional states, clinical anxiety, and acute physiological stress appear to consume a baseline fraction of central attentional processing time ($T_{capture}$), artificially elevating the cognitive load of even mundane environments. Integrating these dynamic emotional vectors into the chronometric architecture represents an essential step toward developing a comprehensive, ecologically valid model of human mental capacity.
12.3 The Next Horizon: Methodological Innovations and Theoretical Extensions
As cognitive science accelerates into the twenty-first century, the paradigm established by Pierre Barrouillet and Valérie Camos is expanding into new methodological horizons driven by mobile neuroimaging, computational neuroscience, and artificial intelligence. The next frontier of TBRS research is marked by the deployment of high-resolution mobile eye-tracking and wireless functional near-infrared spectroscopy (fNIRS) in fully naturalistic, immersive settings.
These advanced mobile recording modalities allow researchers to measure attentional capture times ($T_p$) and micro-pauses in real time as individuals navigate real-world environments, such as driving vehicles, performing surgical procedures, or interacting within virtual reality educational environments. By tracking pupillary dilations, micro-saccades, and prefrontal oxygenation fluctuations down to the millisecond during unconstrained human behavior, these technologies are validating the core axioms of the TBRS model outside the confines of the computer-controlled laboratory.
Simultaneously, machine-learning algorithms are being integrated with the TBRS* computational architecture to develop predictive models of real-time human cognitive failure. By feeding real-time physiological telemetry, stimulus presentation pacing, and individual processing speed baselines into artificial neural network implementations of TBRS*, computational systems can anticipate when an operator’s cumulative cognitive load is approaching the tipping point of working memory collapse. This predictive capability enables adaptive artificial intelligence systems to intervene dynamically, shedding non-critical cognitive tasks and preserving the integrity of human executive performance.
Through these continuous methodological and computational innovations, the Time-Based Resource Sharing model continues to demonstrate profound explanatory power. By grounding human working memory in the dynamic, temporal scheduling of an indivisible attentional resource, Barrouillet and Camos provided cognitive science with a transformative framework that unifies mind, time, and computational biology—cementing the TBRS architecture as a foundational pillar of modern cognitive psychology.
Conclusion
The Time-Based Resource Sharing (TBRS) model formulated by Pierre Barrouillet and Valérie Camos represents a landmark paradigm shift in our understanding of human working memory. By deconstructing the long-standing, static spatial container metaphors that dominated twentieth-century cognitive architectures, TBRS reframed working memory as an inherently dynamic, time-constrained cognitive balancing act. The framework demonstrates that the human mind does not multi-task through parallel processing channels; rather, it achieves complex cognitive coordination through high-frequency, sequential time-sharing across a single-channel central attentional bottleneck.
Within this chronometric architecture, the survival of fragile memory traces is governed by an unyielding race against time: representations outside the conscious focus of attention suffer rapid, spontaneous temporal decay, their preservation contingent upon the executive system’s ability to exploit fleeting micro-intervals to deploy restorative attentional refreshing pulses. By formalizing this dynamic through an objective, mathematically derived Cognitive Load equation, Barrouillet and Camos provided experimental psychology with a predictive tool of exceptional elegance, resolving historical debates regarding processing difficulty versus duration and establishing cross-domain invariance across verbal, visual, and spatial realms.
From its rigorous computational formalization in the TBRS* architecture to its neurobiological grounding within the fronto-parietal control network, and from its explanatory power across developmental ontogeny to its translational impact on educational design and occupational safety, the TBRS model stands as an enduring theoretical triumph. It reminds cognitive science that the definitive currency of human thought is not merely the volume of information we can hold, but the precision with which we schedule our limited attention across the relentless, flowing canvas of physical time.
References
- Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H. Bower (Ed.), The Psychology of Learning and Motivation (Vol. 8, pp. 47–89). Academic Press. https://doi.org/10.1016/S0079-7421(08)60452-1
- Baddeley, A. (2000). The episodic buffer: A new component of working memory? Trends in Cognitive Sciences, 4(11), 417–423. https://doi.org/10.1016/S1364-6613(00)01538-2
- Barrouillet, P., Bernardin, S., & Camos, V. (2004). Time to refresh: Cognitive load in working memory. Journal of Experimental Psychology: General, 133(1), 83–100. https://doi.org/10.1037/0096-3445.133.1.83
- Barrouillet, P., & Camos, V. (2001). Developmental increase in working memory span: Resource sharing or temporal decay? Journal of Memory and Language, 45(1), 1–20. https://doi.org/10.1006/jmla.2001.2767
- Barrouillet, P., & Camos, V. (2007). The time-based resource-sharing model of working memory. In N. Osaka, R. H. Logie, & M. D’Esposito (Eds.), The Cognitive Neuroscience of Working Memory (pp. 59–80). Oxford University Press. https://doi.org/10.1093/acprof:oso/9780198570394.003.0004
- Barrouillet, P., & Camos, V. (2012). As time goes by: Temporal constraints in working memory. Current Directions in Psychological Science, 21(6), 413–419. https://doi.org/10.1177/0963721412459513
- Barrouillet, P., & Camos, V. (2015). Working Memory: Loss and Reconstruction. Psychology Press. https://doi.org/10.4324/9781315754598
- Barrouillet, P., De Paepe, A., & Gorin, S. (2012). Does refreshing in working memory require executive control? Journal of Experimental Psychology: General, 141(4), 629–635. https://doi.org/10.1037/a0026857
- Barrouillet, P., Portrat, E., & Camos, V. (2011). On the law relating cognitive load to forgetting in working memory: A response to Oberauer and Lewandowsky (2011). Journal of Experimental Psychology: General, 140(1), 130–135. https://doi.org/10.1037/a0022320
- Camos, V., & Barrouillet, P. (2014). Attentional refreshing and rehearsal in working memory: Two maintenance mechanisms. Annals of the New York Academy of Sciences, 1316(1), 58–68. https://doi.org/10.1111/nyas.12461
- Camos, V., Lagner, P., & Barrouillet, P. (2009). Two maintenance mechanisms of verbal information in working memory. Journal of Memory and Language, 61(3), 457–469. https://doi.org/10.1016/j.jml.2009.06.002
- Cowan, N. (1999). An embedded-processes approach to working memory. In A. Miyake & P. Shah (Eds.), Models of Working Memory: Mechanisms of Active Maintenance and Executive Control (pp. 62–101). Cambridge University Press. https://doi.org/10.1017/CBO9781139174909.005
- Cowan, N. (2005). Working Memory Capacity. Psychology Press. https://doi.org/10.4324/9780203342398
- Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19(4), 450–466. https://doi.org/10.1016/S0022-5371(80)90312-6
- Just, M. A., & Carpenter, P. A. (1992). A capacity theory of comprehension: Individual differences in working memory. Psychological Review, 99(1), 122–149. https://doi.org/10.1037/0033-295X.99.1.122
- Lewandowsky, S., & Oberauer, K. (2009). No evidence for temporal decay in working memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(6), 1545–1551. https://doi.org/10.1037/a0017010
- Oberauer, K. (2002). Access to information in working memory: An inspection of the focus of attention. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28(3), 411–421. https://doi.org/10.1037/0278-7393.28.3.411
- Oberauer, K., & Lewandowsky, S. (2011). Modeling working memory: A computational implementation of the time-based resource-sharing model. Journal of Experimental Psychology: General, 140(1), 117–129. https://doi.org/10.1037/a0024615
- Pashler, H. (1994). Dual-task interference in simple tasks: Data and theory. Psychological Bulletin, 116(2), 220–244. https://doi.org/10.1037/0033-2909.116.2.220
- Portrat, E., Barrouillet, P., & Camos, V. (2008). Time-related decay or interference-based forgetting in working memory? Journal of Experimental Psychology: Learning, Memory, and Cognition, 34(6), 1561–1564. https://doi.org/10.1037/a0013356
- Raye, C. L., Johnson, M. K., Mitchell, K. J., Greene, E. J., & Camos, V. (2007). Refreshing: A minimal executive function. Cerebral Cortex, 17(10), 2243–2252. https://doi.org/10.1093/cercor/bhm018
- Vergauwe, E., Barrouillet, P., & Camos, V. (2010). Do mental processes share a domain-general resource? Psychological Science, 21(3), 384–390. https://doi.org/10.1177/0956797610361340