Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Human Motor Performance, Target Acquisition, and Control Compatibility

Abstract. Human motor performance in engineered systems emerges from the relation between action intentions, target geometry, motor control, response selection, device mechanics, and the consequences of activating a control. This fundamental research monograph examines the historical development and experimental foundations of aimed movement and stimulus–response compatibility, with particular attention to Fitts-type speed–accuracy relations, motor-output variability, optimized submovement models, visual and proprioceptive feedback, internal prediction, sensorimotor adaptation, and spatial response coding. It integrates evidence from reciprocal tapping, discrete pointing, Simon-effect and mapping experiments, trajectory steering, two- and three-dimensional target acquisition, accessible goal crossing, ergonomic assessment, and contemporary virtual-reality device comparisons. A critical narrative synthesis distinguishes reaction time from movement time, nominal target dimensions from effective accuracy, task selection from action confirmation, and laboratory model fit from transfer to operational settings. It evaluates measurement conventions, individual differences, motor impairment, physical exertion, safety-critical errors, automation of target selection, and the implications of human–Artificial cooperation. The scientific synthesis identifies a layered architecture of intended action, response choice, movement regulation, result confirmation, and recovery, followed by an agenda for predictive validation across populations, modalities, and tasks. The review reports no new empirical observations, independently conducted systematic review, or meta-analysis.

Keywords: human motor performance; Fitts’ law; target acquisition; speed–accuracy tradeoff; stimulus–response compatibility; Simon effect; response mapping; motor variability; motor control; pointing devices; movement time; throughput; goal crossing; steering law; bimanual coordination; assistive interaction; ergonomics; virtual reality pointing; control design; human–Artificial interaction; Artificial Era

Human Motor Performance as a Scientific and Engineering Object

Human motor performance is the measurable organization of voluntary action under biomechanical, perceptual, cognitive, and environmental constraints. In an engineered system, a person must identify an intended outcome, select a permissible action, coordinate the body or an input device, and determine whether the resulting state satisfies the goal. Target acquisition is the directed movement of an effector or controlled pointer into a defined spatial region, often followed by an explicit selection event. Control compatibility is the relation between the information specifying an action and the physical or representational organization of the response required by the system. These phenomena belong to one applied research object because an apparently fast movement can still fail at selection, while a mechanically easy control can impose substantial response-selection demands. The foundational distinction between reaction time and movement time in Fitts & Peterson (1964) and the spatial-mapping analysis in Fitts & Seeger (1953) establish complementary experimental routes into this problem.

The dependent variable called performance must be specified before evidence can be interpreted. A pointing experiment may quantify onset latency, movement duration, endpoint variance, missed targets, corrective submovements, selection errors, or speed–accuracy-adjusted throughput. A workstation assessment may add reach distance, physical load, fatigue, posture, and recovery after an erroneous action. A control-panel task may require the operator to convert a displayed state into the correct control movement; spatial compatibility can determine which action is selected before biomechanical execution begins. These outcomes are related but not interchangeable. A single average completion time conceals whether time was spent searching, choosing, reaching, correcting, confirming, or undoing. Cockburn et al. (2007) demonstrates why models of interface performance must often join visual search, choice latency, movement time, and learning rather than assigning every delay to target acquisition.

This dossier is a critical narrative synthesis of original aimed-movement experiments, theories of motor variability and feedback, stimulus–response correspondence studies, human–computer interaction experiments, accessible input research, and ergonomic assessment standards. It reports no original participant data, preregistered systematic search, or independent meta-analysis. Its guiding questions concern how target geometry governs time and precision, why physical and symbolic mappings affect correct action, how control devices change movement dynamics, how performance generalizes to atypical users and settings, and what evidence is required before translating laboratory coefficients into a deployed design. The research boundary is applied human action in systems; general sensorimotor integration belongs to its neighboring scientific dossier, whereas this work follows its consequences for target selection, device-mediated control, and reliable operation MacKenzie (1992), Proctor & Vu (2006).

From Voluntary Movement to Modern Pointing Research

The modern experimental lineage begins before electronic interfaces. In his 1899 investigation of the accuracy of voluntary movement, Robert S. Woodworth examined the relation between movement speed and end-point precision and proposed a distinction between an initial movement component and later adjustments informed by sensory feedback. The historical argument has remained influential because it separates early motor preparation from information available during execution. Elliott et al. (2001) reexamines Woodworth's program a century later and relates it to contemporary evidence on rapid aiming. The early work established a question that persists in both physical and digital interfaces: how can an action remain accurate when the available time for detecting and correcting deviations is limited?

Midcentury human engineering connected this question to information theory and the design of controls. Fitts & Seeger (1953) investigated spatial properties of stimulus–response arrangements, while Fitts (1954) measured rapid reciprocal movements between targets of different widths and separations. Fitts treated movement time and precision through an information-theoretic analogy, explicitly recognizing visual and proprioceptive feedback as parts of the measured motor system. Fitts & Peterson (1964) subsequently studied discrete responses, separating initiation latency from movement duration and showing that target amplitude and width exert large, systematic effects on movement time. These investigations should be read as distinct experiments: reciprocal tapping, discrete reaching, and control mapping do not estimate the same psychological mechanism merely because they were examined by the same researcher.

Research after the original studies refined the account. Schmidt et al. (1979) developed a motor-output variability theory for rapid action; Meyer et al. (1988) proposed optimized primary and corrective submovements; and Harris & Wolpert (1998) related movement planning to signal-dependent noise. In parallel, Welford et al. (1969) examined departures from a simple speed–accuracy formulation and age-related variation. Computer-interface research then tested keyboards, mice, touchpads, joysticks, and other mechanisms using target-selection tasks, as surveyed by MacKenzie (1992). Soukoreff & MacKenzie (2004) proposed reporting practices that make pointing studies more comparable, while Zhai (2004) clarified serious ambiguities in throughput calculation. This sequence documents both the usefulness of a compact empirical law and the growing requirement to measure the action system rather than one geometry variable in isolation.

Aimed Movement, Target Geometry, and the Speed–Accuracy Relation

A commonly used Shannon-form expression of Fitts-type behavior is MT = a + b log2(A/W + 1), where MT is movement time, A is movement amplitude, W is target width along the direction of approach, and a and b are fitted parameters. The logarithmic quantity is the index of difficulty, expressed conventionally in bits. It represents a design-dependent operational measure, not a literal count of neural symbols. Fitts's original 1954 formulation and later adaptations do not use identical index definitions; MacKenzie (1992) explains why the Shannon form became attractive for device evaluation. A correct application identifies the experimental task, geometrical definitions, whether target center distance or actual path length is used, and whether width is nominal or inferred from observed endpoints. Without those decisions, coefficients across two studies may be numerically precise while referring to different behavioral constructs.

The central empirical tendency is that movements to farther or smaller targets require more time when accuracy is constrained. Both experimental instructions and the physical action determine what counts as an error. A study that rewards maximum speed can produce different endpoint variability from one that penalizes every miss, even with the same A/W ratio. Fitts & Peterson (1964) found a strong relation between a defined difficulty index and movement time under its discrete-response conditions; that reported relation is evidence for a controlled paradigm, not proof that every reach in a real workstation obeys an identical slope. Researchers must distinguish movement onset from initial reaction, physical arrival from electronic click confirmation, and intended target geometry from the effective area the user's pointer can reliably select. The law estimates regularity across conditions, while causal explanation requires examination of planning, feedback, and execution.

An important measurement refinement uses effective target width based on observed endpoint dispersion, rather than assuming that instructed width fully represents performed precision. Soukoreff & MacKenzie (2004) develops standardized recommendations around effective width, amplitude, accuracy, and reporting, building on the history of ISO pointing-device methods. Effective-width calculations typically use the observed standard deviation of endpoints under explicit distributional assumptions, so the analyst must disclose how outliers, misses, and repeated selections are handled. Estimating the regression on participant-level or trial-level data also changes uncertainty. A design comparison should report the range of tested difficulties, regression coefficients, scatter, and failure rates, because the same overall slope may coexist with clinically or operationally important differences in correction behavior.

Reaction, Movement, Confirmation, and Error as Separate Processes

Task completion begins before motion. A visual signal must be detected and interpreted, an intended response selected, and an action plan sufficiently prepared for movement onset. The classical separation of reaction time and movement time in Fitts & Peterson (1964) indicates that target geometry may strongly affect movement duration while having a smaller influence on the latency of response initiation under the studied conditions. This separation is useful in engineering because a cursor that moves efficiently cannot compensate for an ambiguous control label or an overloaded choice set. At the same time, initiation and execution can interact when people anticipate the required precision, delay action to prepare a trajectory, or select a cautious movement policy. The measured stages are analytically separable but should not automatically be treated as strictly serial independent modules.

Completion also extends beyond reaching the geometric target. The user may need to press a button, release a drag, maintain a dwell interval, verify the selected command, and sometimes correct a system state. Wobbrock & Gajos (2008) identifies a particularly consequential difference between entering a target area and activating it without cursor displacement. For people with motor impairments, executing a precise click after successful pointing may introduce a fresh error opportunity. A design can therefore have excellent movement time and poor end-to-end accessibility. Experiments should code target entry, entry-to-selection interval, accidental activation, release accuracy, and successful task completion as separate events. The appropriate aggregate measure depends on whether the operational task prioritizes speed, precision, safety, or independence.

Errors themselves require a taxonomy. An aiming miss may result from trajectory variability, tracking latency, altered gain, occlusion, fatigue, or inaccurate estimation of the target. A wrong-control selection may arise from incompatible mapping, attention to an irrelevant location, or confusion between similar control states. A correct control activated at an unsafe time is a coordination or task-state error. Treating each of these as a single error-rate percentage obscures intervention mechanisms. Kornblum et al. (1990) supplies a principled basis for separating correspondence effects from generic task difficulty, while Meyer et al. (1988) models movement correction rather than response selection. A scientifically informative assessment records both the observed outcome and a plausible stage-specific mechanism, and tests that mechanism through controlled manipulation.

Motor Variability, Feedforward Preparation, and Corrective Submovements

Motor responses vary from trial to trial even when the intended target and instructions remain unchanged. This variability creates a fundamental performance constraint: trajectories that are sufficiently fast may produce more variable endpoints, while movements that allow additional sensing and correction require time. Schmidt et al. (1979) offered an early explanatory account linking motor-output variability to rapid-movement accuracy. The model belongs to the historical evolution of motor-control theory, and its predictions have to be distinguished from later approaches that incorporate explicit optimization and sensory state estimation. It is scientifically useful to characterize variability in amplitude, direction, acceleration, and terminal position rather than assuming that all fluctuations can be summarized by one endpoint standard deviation. The physical properties of a controller, friction, hand tremor, and the scaling of pointer movement can all contribute to observed variability, but their mechanisms differ.

The optimized-submovement formulation of Meyer et al. (1988) explains aimed action through a fast initial movement and, when needed, a secondary correction. It proposes that control parameters can be chosen to minimize expected movement time while maintaining a high probability of reaching the target. This model makes predictions about movement distributions and corrective behavior beyond the usual Fitts regression line. Kinematic decomposition can separate a rapid approach from late deceleration, reversal, or localized adjustment, although algorithms for classifying submovements require explicit thresholds and can mistake measurement noise for a correction. In interface research, submovement counts and movement-path efficiency help explain why two devices produce similar median selection times while differing in subjective controllability or fatigue. Wobbrock & Gajos (2008) applies such path-level analysis to accessible pointing and goal crossing, where the action required for success differs by interaction technique.

Harris & Wolpert (1998) models a movement-planning pressure arising from noise that increases with control-signal magnitude, allowing an optimal trajectory to reduce expected endpoint variance. This is an explanatory proposal about motor planning that connects to the observed regularity of smooth trajectories; it is not equivalent to the empirical claim that every motor task follows a fixed logarithmic law. An experiment can distinguish candidate mechanisms by changing inertial load, introducing visuomotor delay, manipulating endpoint penalty, or comparing trajectories under different accuracy requirements. Time-varying velocity and acceleration, endpoint dispersion, and the occurrence of corrections provide a richer test than mean movement time alone. The resulting model comparison should use out-of-sample predictions when possible because alternative mechanisms can produce similar mean speed–accuracy curves.

Visual, Proprioceptive, and Predictive Contributions to Control

Goal-directed action depends on information about both the target and the moving body. Vision supplies spatial relationships and visible consequences of movement; proprioceptive signals help specify limb position and motion; tactile information reveals contact and activation. These sources arrive with different delays and uncertainties, so effective action requires a representation of what the effector is doing while the movement unfolds. Wolpert et al. (1995) investigated estimates of hand position under altered mechanical forces and interpreted the pattern of errors through internal models and state estimation. This work supports the value of predictive sensorimotor representations while leaving open the exact neural realization of the computational operations. In a digital pointing task, the controlled cursor is an additional object whose displayed behavior may depart from hand motion because of gain, latency, filtering, or software assistance.

Visual feedback can correct a developing trajectory only when the display makes the relevant error detectable and the remaining movement time permits a response. Very small targets require more stringent terminal accuracy; delayed feedback can force a different control policy from immediate feedback even when target geometry remains constant. The two-component historical account reviewed in Elliott et al. (2001) captures the practical distinction between rapid preparation and late feedback control. However, contemporary control theories also describe continuously evolving predictions and corrections, so the operational question concerns the timing and reliability of information, not a fixed division into precisely two movements. Studies that remove visual feedback, perturb cursor position, or hide the limb can identify which cues contribute to stability under a defined task. The same study should measure error consequences because an apparent adaptation in average time can conceal increased misses.

Control interfaces can support prediction by making mapping, consequences, and confirmation states coherent across action. A cursor moving directly with hand displacement offers a different informational relation from a joystick that changes cursor velocity, and a rotary control changes the body's action geometry again. These distinctions determine how easily a user estimates where a pointer will arrive. Shadmehr & Mussa-Ivaldi (1994) showed adjustment of reaching movements to imposed force fields through practice, revealing adaptable representations of task dynamics. The transfer from that laboratory finding to a particular interface design remains an empirical question: users can learn a mapping, but learning rate, retention, performance during unexpected changes, and error recovery must be measured in the device's own context. Predictability and stable feedback are therefore testable interface properties rather than purely aesthetic qualities.

Sensorimotor Adaptation, Learning, and Transfer Between Devices

An individual rarely encounters a control system with an entirely blank motor history. Previous mouse use, touchscreen experience, professional manipulation, or rehabilitation practice shapes expectations about gain, friction, movement direction, and selection. Adaptation occurs when sensorimotor predictions are updated in response to discrepancies between intended and observed action. In the force-field experiments of Shadmehr & Mussa-Ivaldi (1994), trajectories initially deviated under a new mechanical environment and became more similar to baseline trajectories with practice. That pattern demonstrates adaptation to a specific dynamic perturbation; it does not establish a universal rate of interface learning. Experiments on controllers should separately estimate initial acquisition, within-session improvement, retention after delay, transfer to new devices, and savings when a familiar mapping returns.

Theories of predictive internal models provide a basis for describing why transfer can be limited. A person may learn the movement required to select a target under one mapping without obtaining a general representation that transfers unchanged to another display scale or posture. Wolpert et al. (1995) supports the relevance of internal prediction and estimated limb state, while Harris & Wolpert (1998) indicates why altered signal reliability can shift trajectory planning. Training may reduce movement time through earlier anticipation, greater movement efficiency, fewer corrections, or a more permissive interpretation of acceptable error. These mechanisms require different measures. If an evaluator records only task completion time at the end of training, a system with a steep initial learning cost may appear equal to an immediately usable design.

A useful evaluation deliberately includes counterbalanced device orders, both novice and practiced phases, washout or retention intervals when feasible, and realistic task transitions. The mapping and instructions should remain explicit because people often trade accuracy for speed after becoming comfortable. Salthouse (1984) provides a historical example of the limits of simple age-based inference: skilled older typists could preserve overall typing speed while differing on component speed and anticipatory processing. The study concerns typing, not pointing, yet its design illustrates why acquired skill can reorganize performance in ways that a single elementary latency cannot capture. For engineering, transfer is a defined dependent variable measured under specified changes, not an assumption that a familiar movement metaphor will be immediately usable by every population.

Handedness, Bimanual Coordination, Reach, and Embodied Constraints

Motor performance is performed by a body situated relative to the work surface, display, and controls. Reach direction, forearm support, hand posture, seating, grip, handedness, and visual access can alter the feasible set of movements. A distance on the screen is not equivalent to a distance between physical controls or a free-hand movement in three-dimensional space. Changes in target depth may also require different joint combinations and stabilization strategies. Guiard (1987) proposed an asymmetric coordination account of skilled bimanual activity, emphasizing that two hands often have complementary roles within one task. A study that forces a single-hand technique on an otherwise bimanual task can therefore change the task's functional organization, and device comparisons should record whether the nonpreferred hand provides stabilization, navigation, or contextual framing.

Handedness should be treated as a distribution of preferences, roles, and skill histories rather than a one-word binary explanation. The assignment of hands to controls may affect compatibility because an interface can align a displayed location with one hand's expected action while conflicting with the other hand's position or action goal. Similarly, a physically reachable target may be difficult because of awkward wrist extension, occlusion, or repeated contact pressure. Ergonomic standards such as ISO (2007) address input-device design at a level broader than pointing time and direct attention to the relation between physical devices and human capabilities. A performance study confined to short repeated movements cannot by itself establish safe or comfortable sustained use, since fatigue and cumulative exposure have different timescales.

Research across age and ability groups further shows why average motor performance is an inadequate design target. Welford et al. (1969) reported age-related changes in a tapping paradigm while also identifying departures from a simple linear speed–accuracy account. Salthouse (1984) found that task expertise can compensate for differences in elementary processing measures during skilled typing. These findings indicate heterogeneity and support participant stratification, but they do not justify assuming a universal age penalty or a particular adaptive interface effect. A stronger program samples the actual user population, measures reach and endpoint variability under practical posture, and follows performance across realistic sessions. Physical comfort, reachability, and independent control are integral outcomes when the system is designed for repeated use.

Control Compatibility and Stimulus–Response Correspondence

Control compatibility describes how readily an operator can translate an identified stimulus, display, or task intention into the required control action. A compatible arrangement supports reliable response selection through spatial, conceptual, or learned correspondence between a displayed event and an available action. In the foundational experiments of Fitts & Seeger (1953), relationships between stimulus and response codes affected reaction-time performance even when the task required a simple, fully specified response set. This finding was consequential for engineering psychology because it established that physical arrangement is part of the information-processing problem. Position, direction, movement range, and displayed consequence can all participate in a mapping. To diagnose a particular design, the investigator must specify which dimensions are intended to correspond and which are incidental to the required decision.

Stimulus–response compatibility has more than one operational meaning. Dimensional overlap refers to a relation between representational features of stimuli and responses, such as left–right spatial position; mapping compatibility refers to which assignments of stimuli to responses lead to better performance after instructions and practice. Kornblum et al. (1990) developed a dimensional-overlap taxonomy to explain how relevant and irrelevant stimulus dimensions can activate response tendencies. The taxonomy is useful for distinguishing a direct location-based effect from difficulty arising from arbitrary coding or memory load. A panel can be physically symmetrical yet impose a conceptual reversal, and a visually intuitive display can still present an action that is mechanically awkward. Experimental comparisons should vary mapping while holding response number, signal discriminability, target distance, and physical load as constant as feasible.

The applied outcome is both speed and error risk. More compatible stimulus–response relationships often reduce selection cost in a defined paradigm, but the size and even direction of an effect depend on instructions, response mode, practice, reference frames, and the user's task representation. Proctor & Vu (2006) integrates decades of compatibility research into design-relevant principles and cautions against interpreting isolated congruency effects as a universal interface rule. A robust engineering claim must establish which mapping is preferable for the actual operators, which errors matter, and whether a laboratory correspondence advantage persists under fatigue, unexpected events, or concurrent work. Control compatibility becomes a property of the human–system task arrangement that is estimated empirically rather than inferred from visual resemblance alone.

The Simon Effect and Task-Irrelevant Spatial Codes

Spatial features can influence action even when they provide no information needed to choose the correct response. In Simon (1969), participants operated a control in relation to lateralized auditory stimulation and responded faster under a corresponding directional relation than under the reversed relation. This historical observation contributed to what later became known as the Simon effect: response speed and sometimes accuracy are altered when an irrelevant stimulus location corresponds or conflicts with the required response. The important scientific feature is irrelevance. A task can require identifying a nonspatial signal attribute while the signal's spatial origin still influences response selection. Such effects show that a system's unused spatial information can be behaviorally consequential.

One explanation is that spatial stimulation activates a response code that competes with the instructed action when correspondence is absent. Kornblum et al. (1990) treats the resulting interaction as a dimensional-overlap problem. Other accounts emphasize referential coding and the spatial relation among stimuli, responses, and task frames. Hommel (1993) investigated how the Simon effect depends on the frame of reference in which an event and a response are represented, highlighting that correspondence does not belong to physical distance alone. Engineering applications therefore have to ask whose coordinate system is being used: screen orientation, the operator's body, a vehicle's movement, a controller's mechanical axis, or a visually rotated display. The same control action may be compatible in one reference frame and incompatible in another.

Simon-type results must be distinguished from long-term learned mapping costs. A sudden trial-level congruency manipulation can reveal interference under highly constrained laboratory conditions. A fixed panel in which operators develop stable associations may produce different behavior after sustained training. Proctor & Vu (2010) examined mixed mappings and stimulus–response compatibility, illustrating why mapping context and preparation matter when predicting outcomes. The appropriate engineering study compares immediate errors, learning curves, retention, and performance after perturbation, not simply the reaction-time benefit in a single conventional task. In consequential environments, the most relevant measure may be the likelihood of selecting the wrong control when an irrelevant signal becomes unusually salient.

Mapping Consistency, Reference Frames, and Response Coding

A control mapping is an assignment between task states, perceived indications, actions, and resulting system changes. It can be spatial, semantic, numerical, or operational. A slider that increases an indicated value when moved upward supports one conventional correspondence; a vehicle control may use a direction interpreted relative to the operator's seat or the machine's movement; a medical interface may expose different choices depending on mode. These cases share a response-coding problem but differ in the conventions that participants have learned. Fitts & Seeger (1953) establishes the historical experimental foundation for mapping effects, while Proctor & Vu (2006) explains how representational correspondence and acquired associations jointly shape selection. The design question is whether the action implied by a control can be inferred consistently from the information the user receives.

Consistency has several separable components. Within-control consistency means the same physical action produces the same class of result across repetitions. Across-control consistency concerns similar functions sharing similar movement rules. Display–control compatibility concerns how displayed locations or quantities align with physical controls. Cross-modal compatibility concerns visual or auditory information paired with manual, vocal, or other responses. These components can diverge: an arrangement that minimizes hand travel may locate a control far from the associated display, while physical proximity can create accidental selection between neighboring buttons. Kornblum et al. (1990) provides a theoretical language for identifying correspondence dimensions; the applied evaluation must establish which dimension actually drives errors in the tested context.

Reference-frame ambiguity is especially important in devices that alter viewpoints, such as remote manipulators, cameras, drones, mirrored environments, and immersive displays. In each case the operator may need to transform an action relative to the body into a consequence relative to an external scene. Hommel (1993) demonstrates that spatial correspondence depends on how locations are encoded during the task, and Shadmehr & Mussa-Ivaldi (1994) shows that learned movement dynamics can change under a new environment. The two literatures address different mechanisms: choice coding and sensorimotor adaptation. A complete control design can be evaluated by factorially varying visual frame, response frame, movement gain, and practice, allowing the investigator to identify whether a delayed response reflects remapping, unfamiliar dynamics, or endpoint correction.

Information, Choice Uncertainty, and Action Mapping

The number and probabilities of response alternatives affect selection demands. Hick–Hyman relations describe how choice time can vary with uncertainty under constrained tasks. They provide a useful neighboring framework when a person must first choose among several controls, whereas Fitts-type relations concern the time required to execute a movement to a spatial target. Cockburn et al. (2007) explicitly brought search, choice, and pointing contributions into a unified predictive analysis for computer menus. Its importance for target-acquisition science is methodological: a user who is slow to select an option may have experienced difficult visual search, difficult response choice, difficult motor acquisition, or their combination. A complete task measure should identify which processing stage the intervention actually changes.

An information measure used to model response choice is not automatically the same as the logarithmic difficulty measure used in aimed movements. The first concerns uncertainty across alternatives and the conditions of stimulus discrimination and response selection; the second uses geometric amplitude and tolerance for a defined movement task. A menu containing many closely spaced options can increase both demands, but experiments must manipulate them separately before causal contributions are inferred. MacKenzie (1992) reviews pointing-device models while Cockburn et al. (2007) provides a model that treats multiple selection components explicitly. For example, manipulating menu length while holding target width fixed helps distinguish search and choice demands from a pure size effect, while changing width at a fixed menu position tests movement precision more directly.

The combined design problem is to minimize total time and harmful error without assuming all performance can be described by one law. A familiar control vocabulary can speed response choice yet produce difficult reaches; large targets can improve acquisition yet enlarge visual search distance; automation can remove reaching effort while introducing mode uncertainty. Proctor & Vu (2006) provides principles for correspondence, and Soukoreff & MacKenzie (2004) provides consistent measurement recommendations for pointing. Taken together, these literatures support a decomposition strategy: measure choice latency, movement dynamics, endpoint precision, activation reliability, and correction cost, and then estimate which design component contributes to task success. The resulting analysis guides changes to layout, labeling, activation mechanism, and training without conflating their causal pathways.

Two-Dimensional Pointing and Device-Mediated Target Acquisition

The two-dimensional pointing task became a central experimental model because it provides reproducible target locations, adjustable amplitude and width, and measurable cursor trajectories. MacKenzie (1992) reviewed how Fitts's law can be adapted to human–computer interaction and compared formulations relevant to pointing devices. A pointer on a screen is controlled indirectly through a hardware mechanism; the mapping between hand displacement and cursor displacement introduces gain, nonlinear acceleration, clutching, and possible latency. Consequently, the same target in pixels can impose different hand movements across devices. A meaningful evaluation specifies screen geometry, pointer sensitivity, device gain, starting position, instructed speed, and the activation criterion. These details determine whether the measured quantity concerns motor control, response choice, cursor feedback, or a device-specific combination.

Mouse, touchpad, trackball, pen, and joystick tasks can be organized around nominally identical onscreen targets but differ in control transfer functions and required muscle coordination. Direct touch removes the mediated pointer in some tasks but adds finger occlusion and a contact patch whose location differs from the viewed target point. A stylus may permit precision while increasing the need for fine contact control; a joystick may demand integrated velocity regulation rather than a proportional displacement. Comparisons therefore need to separate physical input motion from cursor motion and interpret parameters within the sampled task range. Zhai (2004) showed that throughput calculations can change materially with evaluation conventions, challenging an unqualified ranking of devices on a single bit-per-second score. The broader scientific lesson is that operational equivalence must be demonstrated, not assumed from common target coordinates.

Soukoreff & MacKenzie (2004) synthesized recommendations for controlled pointing evaluation: distinguish nominal and effective target dimensions, document errors, use consistent task designs, and evaluate throughput under transparent conventions. These recommendations strengthen replication because two investigators can otherwise fit linear models to different underlying data and report apparently comparable numbers. They also imply that performance reports should include participant-level variation rather than only aggregated coefficients. A device may improve median movement time while impairing the slowest users, and such differences are directly relevant to accessibility or sustained work. In applied studies, the unit of evaluation can be successful completion with confirmation and recovery, so authors should report error-adjusted outcomes alongside the conventional laboratory measures.

Target Geometry Beyond the Linear Fitts Paradigm

Targets in applied systems are rarely all one-dimensional strips approached along a fixed axis. They may be rectangular, circular, irregular, nested, overlapping, or partly obstructed. A two-dimensional target has width and height, and its effective tolerance depends on approach direction and movement path. If the intended path is oblique, the smaller physical dimension may become the limiting constraint; if the task permits selection from any direction, geometry changes the available movement strategies. MacKenzie (1992) documents the migration of Fitts-type modeling into interface design, while Grossman & Balakrishnan (2005) reports empirical work on target acquisition in a three-dimensional environment. A model that uses one nominal width for every target shape may lose explanatory value when contact tolerance and spatial uncertainty change across conditions.

Large graphical areas are often easier to acquire, but target size interacts with separation, crowding, visual search, and error cost. Increasing the width of one target can reduce spacing to neighboring controls and raise the probability of an unintended choice. Enlarging a button can also alter interface density, navigation distance, or hand occlusion. An experiment addressing layout optimization should manipulate target area and intertarget spacing separately, with identical visual identification demands where possible. Cockburn et al. (2007) illustrates how menu performance requires more than pointing geometry, and Soukoreff & MacKenzie (2004) supplies a measurement framework for isolating endpoint accuracy. The relevant optimization objective is expected successful task completion under a defined arrangement, not the minimum index of difficulty considered in isolation.

The edge of a physical surface creates a special geometrical condition because a pointer can sometimes be stopped by a hard boundary while an unconstrained cursor can overshoot an on-screen object. Similarly, entering a target from multiple directions may create unequal effective tolerance depending on the path. Such features motivate direct testing instead of extrapolation from a stylized tapping paradigm. A researcher may sample radial approach directions, compare aspect ratios, introduce distractor targets, and model directional endpoint distributions. Reporting the distributions permits an applied analyst to identify when a simple Fitts regression remains adequate and when path-dependent or multidimensional models are required. The distinction is a principled boundary of model applicability rather than an argument against the regularity found in controlled movements Zhai (2004).

Trajectory-Constrained Actions and the Steering Law

Some interface tasks require movement along an extended path rather than arrival at a single terminal region. Examples include navigating a menu corridor, moving through a narrow channel, tracing a path, and maneuvering a virtual tool around obstacles. Accot & Zhai (1997) formulated and tested the steering law for trajectory-based interactions, extending the experimental logic of constrained movement to paths where the available corridor changes along the motion. The key variable is not only the endpoint width but the clearance available throughout the trajectory. A narrow segment can force careful steering even if the destination target is large. This makes steering tasks a separate research problem with their own geometry and error conditions.

The distinction between terminal acquisition and movement through a path is reflected in later modeling. Accot & Zhai (2003) refined Fitts-law models for bivariate pointing, in which the target has two planar dimensions, rather than studying a three-dimensional steering law. This result addresses geometry at the endpoint and cannot by itself establish a theory of continuous movement through depth. A handheld controller guiding a pointer along a corridor also requires trajectory regulation, whereas an immersive controller may depend on shoulder stabilization and tracking accuracy. In each case, an ideal geometric path is only part of the task. The investigator should describe allowable deviations, whether contact with a boundary terminates the trial, whether backtracking is permitted, and how the system signals an error. A steering-law fit is informative within a clearly specified task family and should not be substituted automatically for models of discrete target selection.

One methodological challenge is to determine whether movement time reflects path length, corridor width, corner curvature, or changes in movement dynamics at transitions. Straight paths and curved paths of equal integrated difficulty need not elicit equivalent motor strategies. A strong experiment counterbalances starting orientation and direction and records path deviation, pause locations, reversal, and endpoint confirmation. Accot & Zhai (1997) provides the core mathematical and behavioral starting point; Soukoreff & MacKenzie (2004) offers a complementary reminder that errors and effective tolerances require transparent measurement. For practical interface design, a narrow route that allows undo may be acceptable in exploratory work but unsuitable for high-stakes control where a single boundary crossing triggers an irreversible operation.

Three-Dimensional Reaching, Immersive Interfaces, and Spatial Selection

Three-dimensional target acquisition introduces distance, depth, occlusion, viewpoint, and extended joint coordination. A virtual target may be selected by direct hand contact, a ray cast from a controller, gaze-assisted pointing, or a constrained tool. These techniques alter the effector, feedback, and tolerance rules, so two tasks with the same visible target do not necessarily share the same sensorimotor demands. Babu et al. (2022) experimentally compared controller input with two hand-pointing modalities when participants selected spherical targets positioned in personal space. Its investigation provides direct evidence that target distance relative to arm reach and tracking modality can modify selection performance. The scientific significance lies in determining which geometrical and interaction variables actually predict selection under the tested paradigm, not asserting that the two-dimensional law transfers without adaptation.

Immersive systems add tracking noise, optical occlusion, depth perception, arm fatigue, and variable visual–proprioceptive correspondence. Babu et al. (2022) examined target-acquisition techniques and movement performance in immersive contexts, showing why task-specific interaction geometry matters for performance evaluation. Luong et al. (2023) likewise studied device-mediated target selection in immersive interfaces and provides a contemporary example of controlled three-dimensional selection research. These studies broaden the design space beyond the original reciprocal tapping conditions, but they also make the limits of comparison more visible. A virtual cursor and a physical fingertip may occupy similar visual coordinates while imposing very different correction demands and muscular exposure. For this reason, device studies should record tracking accuracy and arm posture in addition to selection time.

Recent work further emphasizes that performance calibration can be task dependent. Rodin et al. (2026) studied Fitts-style standardized tests in virtual-reality interaction and compared task conditions in ways that highlight transfer limitations. The appropriate inference is about the tested instruments, population, and interaction modalities; it is not proof that a single model has universal parameters across all immersive displays. An applied protocol should distinguish one-dimensional tapping, planar pointing, depth-reaching, and trajectory steering. It should also include independent validation tasks rather than assuming that performance on a benchmark predicts complex operational use. These boundaries make classical motor theory more useful by clarifying precisely what an experimental index can and cannot certify.

Accessibility and Individual Differences in Motor Interaction

A target is accessible when an intended user can complete the required action with acceptable reliability, physical effort, and independence under realistic conditions. The evaluation population therefore matters as much as nominal target geometry. People differ in endpoint variability, tremor, range of movement, control speed, fatigue, proprioceptive feedback, sensory function, and previous interaction experience. A study conducted only with young, experienced computer users is informative about that sample but cannot establish usability for people with motor impairments or limited access to conventional devices. Wobbrock & Gajos (2008) compared goal crossing with pointing for people with and without motor impairments and demonstrates the importance of defining the activation gesture as part of the task. Crossing allows successful selection when a trajectory intersects a target region, while conventional pointing may require stopping and clicking; the tasks have different motor demands.

The distinction between target arrival and activation has practical consequences. A user may position a cursor accurately yet inadvertently move it during a physical button press, or may avoid selecting small targets because maintaining a stable position is difficult. Alternative activation methods include dwell, switch input, continuous movement gestures, and combinations of assisted pointing with explicit confirmation. Wobbrock & Gajos (2008) provides direct experimental grounds for treating crossing as a candidate technique for particular users and tasks. However, no technique can be recommended universally from a single comparison because gesture speed, false positives, controllability, learning, and the severity of mistakes differ across settings. Evaluation should stratify by relevant functional ability and report individual-level distributions, enabling designers to detect whether a method helps the users for whom it is intended.

Accessibility is also an environmental relation. A control that is reachable while standing may be inaccessible from a wheelchair; touch input optimized for one hand size may increase occlusion for another; a high-mounted panel may require sustained shoulder elevation. ISO (2007) addresses physical input-device principles and provides a recognized framework for thinking about hardware design. Functional assessment should document work posture, device location, approach angle, available support, and duration of use. The resulting evidence may favor adaptable layouts, adjustable sensitivity, alternative activation techniques, or interoperable input devices. Such adaptations should be evaluated as interventions with measured benefits and costs, including false activations and task interruption, rather than as assumptions about disability categories.

Age, Expertise, Fatigue, and Physical Context

Age-related performance differences have multiple plausible sources: sensory change, biomechanical factors, processing speed, motor variability, and experience with particular tasks. Welford et al. (1969) examined age and speed–accuracy relations under controlled tapping conditions and identified departures from simplified movement-law descriptions. The findings demonstrate that performance models require empirical testing across age groups, but they should not be treated as a universal equation for older users. Older adults with extensive task-specific experience can adopt strategies that partly preserve operational performance. Salthouse (1984) documented such compensatory organization in skilled typing, showing that expertise and component speed should be analyzed together. When evaluating an interface, investigators should record device familiarity and strategy before attributing differences to chronological age.

Fatigue emerges through exposure, repetition, and posture, whereas classic Fitts trials often emphasize short bouts of directed movement. Repeated target selection can be accurate in a brief laboratory session yet uncomfortable during continuous work. The relevant study may measure change in trajectory variation over time, sustained hand position, reported discomfort, intermittent rest, and missed selections. ISO (2007) concerns physical input-device design and complements the psychologically oriented movement studies by placing performance within human capabilities and workstation constraints. Inferences about clinical musculoskeletal outcomes require appropriately designed health research; discomfort ratings or a short-term throughput decline are insufficient to establish injury risk. The proper engineering question is whether the control remains usable and appropriately safe under representative duration and workload.

Expertise is a confounder and a scientific outcome. In a comparison between a standard mouse and an unfamiliar controller, the familiar device may lead to faster initial selections because its mappings are well learned. If training is unequally distributed, a short experiment estimates familiarity effects alongside device mechanics. Shadmehr & Mussa-Ivaldi (1994) provides experimental evidence that motor responses adapt to altered dynamics, while Salthouse (1984) illustrates the value of studying anticipatory skill organization. A credible evaluation may specify an initial-use phase, a criterion-trained phase, and delayed retention, allowing designers to separate onboarding cost from long-term efficiency. It may also stratify by prior expertise while testing whether users can recover when the established mapping unexpectedly changes.

Motor Impairment, Assistive Input, and Alternative Activation

Assistive interaction often changes the object of selection from a small physical button to a configurable virtual target. A switch interface may require serial scanning and an activation moment, while eye gaze can provide spatial selection with different fixation and confirmation demands. Voice input removes some manual acquisition demands yet introduces speech-recognition uncertainty, privacy constraints, and the need for command disambiguation. These methods should be evaluated as distinct interaction systems rather than compared only by nominal motor pointing time. Wobbrock & Gajos (2008) shows why the shape of the action gesture matters in motor-accessible computing. Its findings provide a rationale for evaluating whether crossing or another technique reduces a specific coordination burden, not for assuming that alternative gestures automatically solve attention, vision, or cognitive access problems.

Many assistive techniques depend on a speed–accuracy–error-cost tradeoff. A larger dwell threshold reduces some accidental activations but requires the user to maintain an action longer; a shorter threshold can make navigation feel responsive while increasing false selections. Assisted target expansion or cursor snapping may reduce movement precision demands, but it can also obscure intended fine-grained control. These consequences should be studied with representative tasks that include intentionally close targets, interruptions, and recovery from a selection mistake. Soukoreff & MacKenzie (2004) emphasizes consistent reporting in pointing evaluation, while Zhai (2004) demonstrates how throughput conventions influence the interpretation of device rankings. For assistive access, success rate and independence often deserve equal or greater weight than raw throughput.

The relevant scientific variables include functional range of movement, tremor amplitude and frequency if directly measured, ability to initiate and terminate movement, tolerance of sustained effort, and consistency under repeated sessions. Respectful participant sampling should reflect the intended usage population while avoiding inference from diagnostic labels alone. A person with the same diagnosis as another may use entirely different motor strategies and devices. ISO (2012) offers an evaluation framework for physical input-device usability and can guide task specification, although actual protocol requirements must be checked against the applicable standard and its edition. The strongest accessibility claims result from representative testing, explicit criteria for independence and error recovery, and transparent description of who was studied.

Safety-Critical Controls, Reversibility, and Error Recovery

The design objective changes when the selected action has significant consequences. In a casual drawing application, a missed target may be corrected with a quick gesture; in a transport, industrial, or clinical control system, a mistaken activation can create a hazardous system state. Under these conditions, selection accuracy, mode awareness, feedback, and the ability to reverse a choice may matter more than a small gain in average movement time. Proctor & Vu (2006) demonstrates the theoretical basis for evaluating control mapping, and Kornblum et al. (1990) distinguishes competing response codes that can generate interference. A safety-relevant study should manipulate correspondence, visual prominence, error costs, and time pressure while observing actual wrong-control responses. The design may use physical separation, distinct movements, confirmation, interlocks, or reversible stages, but effectiveness requires task-specific evaluation.

An interface may increase the nominal size of a frequently used target and inadvertently increase the likelihood of selecting it when a neighboring action is intended. Similarly, grouping related controls can reduce search time while complicating fine selection when the controls are densely packed. Cockburn et al. (2007) supplies a framework for studying multiple contributors to selection, and Soukoreff & MacKenzie (2004) emphasizes explicit errors and movement measures. The practical analysis should consider a matrix of possible slips: wrong target, premature activation, duplicated action, incomplete action, and failure to notice system feedback. The consequences of these error types are not proportional to their frequencies. A design assessment can weight outcomes by their operational severity, but such weights must be justified with domain-specific evidence rather than assigned as if they were universal psychological constants.

Recovery is an independent component of performance. A safe system may detect an unintended action, prevent hazardous execution, or provide immediate and intelligible feedback that permits correction. These features can change behavior by altering the perceived cost of error and the willingness to act quickly. Fitts & Peterson (1964) measured target-precision constraints under defined laboratory conditions; those constraints are not equivalent to the expected loss of an irreversible operation. Engineering studies therefore benefit from complete transaction-level measures: successful action, mistaken initiation, detection time, correction effort, final state, and physical strain. Such data support a realistic balance between fast control, clear mapping, and bounded consequences.

Experimental Paradigms for Aimed Movement and Control Mapping

Research on targeted action depends on tasks whose manipulations distinguish plausible causal mechanisms. A reciprocal tapping paradigm varies amplitude and width across repeated movements; a discrete pointing paradigm begins from a defined location and measures initiation, trajectory, and target activation; a spatial compatibility experiment changes the relation between stimulus and response locations; a steering task constrains the entire movement path. Fitts (1954) supplies the classic reciprocal-movement design, Fitts & Peterson (1964) investigates discrete actions, Fitts & Seeger (1953) manipulates response correspondence, and Accot & Zhai (1997) examines path-constrained interaction. These tasks should be understood as a family of complementary instruments. Each one samples only some of the processes involved in a real interaction, and the validity of an inference depends on the particular controlled variables and error criteria.

A well-specified motor experiment should report participants, eligibility, functional capability relevant to the task, prior device experience, hardware, display characteristics, screen resolution and scaling, input sampling rate, target layout, approach direction, trial sequence, practice, feedback, and response instructions. It must state when the clock starts and stops. If selection is defined by a mouse click, the recorded movement time may include the click delay; if it is defined by cursor entry, the result measures a different action. Soukoreff & MacKenzie (2004) proposes comparable reporting conventions, but researchers still need to disclose deviations and device-specific parameters. Repeating a paradigm with identical nominal amplitude and width is insufficient when pointer acceleration, display latency, or activation requirements differ.

Compatibility experiments should independently vary mapping or correspondence while preserving perceptual discriminability and motor difficulty wherever possible. A rightward visual signal and a rightward lever movement may be naturally corresponding, but a reaction-time benefit can also reflect spatial attention or familiarity. Kornblum et al. (1990) suggests experiments that disentangle dimensional overlap from other stimulus and response attributes; Hommel (1993) highlights the reference frame in which locations are encoded. Factorial manipulations of stimulus location, instructed response identity, hand used, and physical direction can reveal whether a compatibility effect is tied to the screen, body, or external task frame. A priori definitions and sufficient counterbalancing are essential because a pooled mean across conflicting reference frames can conceal different individual response strategies.

Kinematic, Temporal, Accuracy, and Throughput Measures

The movement-time distribution is only one view of human action. Initial reaction time, cursor-path length, velocity profile, peak velocity, acceleration, pauses, corrective submovements, endpoint error, false activation, and recovery can each describe an experimentally distinct feature. Trajectory data allow a researcher to test whether a controller produces overshoot, inefficient reversal, or unstable terminal positioning. Meyer et al. (1988) offers an optimized-submovement account that provides specific predictions about the structure of correction, while Harris & Wolpert (1998) offers a prediction based on control-signal-dependent noise. These models cannot be discriminated adequately from a regression coefficient alone. Their evaluation requires compatible kinematic recordings and explicit handling of filtering, temporal sampling, and device noise.

Throughput attempts to summarize speed and precision in one operational metric, but its computation varies with whether nominal or effective index of difficulty is used, how width is estimated, and whether results are pooled across difficulties and participants. Zhai (2004) examines these issues directly, and Soukoreff & MacKenzie (2004) recommends detailed normalization and reporting. A figure in bits per second should therefore be accompanied by its calculation formula and treatment of misses. Statistical analysis should estimate uncertainty around means, slopes, and condition contrasts, rather than ranking two devices by small differences in point estimates. A more transparent report provides both throughput and disaggregated components so readers can see whether the purported improvement arises from faster actions, a change in accepted error, or a different task geometry.

For compatibility, reaction-time and accuracy outcomes should be analyzed jointly. An apparently faster corresponding response can be a speed–accuracy tradeoff if participants adopt a lower decision threshold, while prolonged response latency may indicate cautious correction of an inappropriate initial activation. Error rates may be low in an easy laboratory task, making them statistically difficult to estimate without adequate trial numbers. Simon (1969) illustrates the historical use of latency and movement measures in a spatially coded task; Kornblum et al. (1990) provides mechanisms that motivate distributional analyses. Practical reporting can include median and tail latencies, trial-wise error classes, response distributions by mapping condition, and learning over repeated blocks. The relevant hypothesis should be stated before selecting a preferred summary statistic.

Statistical Design, Within-Participant Dependence, and Model Comparison

Pointing studies often collect many trials from each participant. Those trials share motor skill, device history, and individual speed–accuracy policy and are therefore correlated. Treating every movement as an independent participant overstates the effective sample size and can produce overly narrow uncertainty intervals. A defensible analysis recognizes participant and, when appropriate, target or device effects, with explicit consideration of learning, fatigue, and trial order. Randomization and counterbalancing limit systematic order bias but do not remove within-session trends. The comparison of regression intercepts and slopes described in MacKenzie (1992) provides a useful historical modeling context; contemporary evaluation should add out-of-sample or held-out validation where a predictive model will guide engineering decisions.

The independent variables of interest are often partly collinear. Increasing target distance can also increase physical reach and screen search, while decreasing width can increase the density of neighboring controls. A factorial design with independently manipulated parameters provides stronger causal interpretation than an uncontrolled comparison of two commercial devices. The same principle applies to response mapping: physical control order, displayed labels, and the direction of resulting state change should be varied deliberately when possible. Proctor & Vu (2006) demonstrates why compatibility hypotheses require precise specification, while Cockburn et al. (2007) models multiple selection contributors that would otherwise be attributed to one variable. Explicit task decomposition thus serves both theoretical identification and practical design diagnosis.

A model should be evaluated on more than the proportion of variance it explains in the training data. A high within-sample fit can coexist with systematic bias at high difficulty, poor predictions for motor-impaired users, or failure when the controller's transfer function changes. Researchers can compare linear Fitts regressions, movement-phase models, and interface-level composite models using held-out tasks, participant groups, and error distributions. Zhai (2004) discusses the consequences of evaluating performance with inconsistent throughput conventions, whereas Grossman & Balakrishnan (2005) illustrates how changing dimensionality can expose limits of planar models. The engineering decision should depend on calibrated predictions within the intended usage conditions and transparent costs of being wrong.

Ecological Validity, Replication, and Evidence Synthesis

Laboratory paradigms are valuable because they isolate target geometry, response correspondence, or device dynamics. Their interpretive limits emerge when the real task includes visual clutter, uncertainty about action consequences, interruptions, posture change, fatigue, collaboration, or rarely used emergency controls. A selection technique that excels at repeated targets on a neutral screen may be slower during occasional use in a complex interface because the operator must first locate the control and recall its function. Cockburn et al. (2007) makes this limitation explicit through multicomponent menu modeling, while Proctor & Vu (2006) shows the relevance of response coding and learned mappings. Transfer to field performance should be tested in representative workflows that preserve operational constraints without sacrificing measurement clarity.

Replication requires more than reusing a familiar equation. Studies must reproduce the relevant control mapping, training schedule, feedback, and success definition. In a device comparison, the apparent effect may reverse when participants become familiar with an initially novel controller or when the task emphasizes error recovery. Soukoreff & MacKenzie (2004) addresses comparability through experimental reporting, and Wobbrock & Gajos (2008) reminds researchers that effects can vary by motor ability and interaction method. A replication plan should prespecify primary outcomes, document the original sampling context, and report both converging and divergent findings. The aim is to determine under which conditions a mechanism remains predictive.

The present synthesis combines original controlled studies, theoretical accounts, usability methods, and recent immersive-interface research. It does not imply that this collection was acquired by a reproducible systematic search. Broad historical consistency supports claims about a regular speed–accuracy tendency in controlled aimed movement, whereas precise device rankings and practical error reductions require narrower, task-specific evidence. Fitts (1954), Fitts & Peterson (1964), and MacKenzie (1992) provide the central movement lineage; Fitts & Seeger (1953) and Kornblum et al. (1990) anchor correspondence; Wobbrock & Gajos (2008) contributes accessibility evidence. Disagreement among theories is productive when the differences lead to measurable predictions about initiation, trajectory, correction, or transfer. Claims of universal usability or safety are not warranted by a small number of laboratory demonstrations.

Control Layouts in Workstations, Vehicles, and Clinical Environments

The point of an engineering-psychology model is to support decisions about actual controls. Workstation interfaces combine visual search, target location, posture, and physical activation. Vehicle interfaces additionally impose time-varying external demands, visual attention constraints, and action consequences that may unfold faster than the operator can correct an error. Clinical equipment can present high-stakes combinations of controls, modes, and confirmation states. These domains require the same foundational psychological distinctions but different evaluation criteria. Fitts & Seeger (1953) establishes that correspondence between signals and controls affects response performance, while Proctor & Vu (2006) details how mapping principles apply to design. A control layout should be tested using representative sequences of action, not simply the reaction time required to press an isolated button after a clearly visible cue.

The first design stage identifies the actions the system must support, the information used to choose them, and the consequences of selection. Designers can then map where controls are located, how far a hand must travel, how controls are distinguished, and whether their motion agrees with displayed results. A target can be geometrically easy but operationally ambiguous when several similar controls are adjacent. Likewise, a mechanically convenient knob may be difficult to operate when the label is distant from the display showing its effect. Cockburn et al. (2007) provides a quantitative reminder that search and action costs can coexist. The evaluation must include misidentification, wrong activation, recovery, physical movement, and action confirmation before an intervention's operational effectiveness can be established.

Domain evidence governs the treatment of risk. A wrong selection in a music player and a wrong selection on critical equipment may be equally likely in a pointing test but have dramatically different consequences. The applied experiment should distinguish likelihood from harm and measure whether safeguards reduce both mistakes and their downstream severity. Physical form factors, independent confirmation, movement resistance, and spacing are plausible design interventions, but none is sufficient by definition. The psychological contribution is a causal account of why each intervention might alter selection or activation, with procedures to measure the effect. In this respect, classic compatibility research remains relevant through its experimental logic rather than through a universal prescription for every console Kornblum et al. (1990), Fitts & Peterson (1964).

Touch, Gestures, Crossing, and Multimodal Interaction

A touchscreen joins a visual display to a physical contact surface and thereby removes some indirect cursor-mapping demands. It simultaneously adds occlusion, finger-contact uncertainty, and the need to stabilize the hand when activating small regions. A gestural interface replaces precise dwell with a recognized movement pattern, which may change the balance between motor effort and software classification error. Wobbrock & Gajos (2008) compares crossing with traditional pointing in accessible interaction and illustrates why selection rules determine the movement that the user must perform. In interface research, the task must define whether contact, crossing, sustained presence, release, or a recognized gesture constitutes success. Apparent equivalence in screen location does not imply equivalence in the action required for selection.

Some multimodal systems combine manual input with gaze, speech, or contextual assistance. Such combinations can reduce the need for large hand movements while creating new demands for mode awareness and confirmation. If gaze is used only to indicate a candidate target and a manual action confirms it, performance includes the accuracy of gaze estimation, the interpretation of intent, and the motor selection. If speech can substitute for manual control, the system must handle ambiguous commands and recognition errors. Existing Fitts-type metrics can describe residual motor acquisition but cannot measure the complete human–computer coordination task. Cockburn et al. (2007) offers an instructive component-level framework, while Proctor & Vu (2006) emphasizes response-coding consistency when multiple response modalities are available.

Gesture-based techniques are useful when the system's movement affordance aligns with a user's available motor repertoire. A crossing gesture may avoid the need to stop precisely inside a small target, but it may fail in crowded areas where a path intersects several unintended targets. This reveals a general design tradeoff between relaxed terminal accuracy and increased path-based ambiguity. Accot & Zhai (1997) provides the experimental foundation for path constraints, and Wobbrock & Gajos (2008) supplies a study of crossing under motor impairment. The strongest applied test compares techniques over both densely packed and sparse layouts, uses task-relevant error penalties, and measures accessibility at the participant level. Adaptability is empirically meaningful when measured improvements persist across contexts rather than merely increasing the number of available input options.

Adaptive Target Assistance and Automation of Movement

An adaptive interface may enlarge likely targets, adjust pointer gain, stabilize a cursor near a candidate, or predict an intended control from prior action. These interventions alter the effective relationship between the user's movement and the system's response. Their possible advantages follow from movement variability and geometric tolerance: a wider effective acceptance region can require less terminal correction under some conditions. Soukoreff & MacKenzie (2004) provides a way to evaluate changes in effective width and pointing performance, while Harris & Wolpert (1998) provides a motor-control rationale for why accuracy depends on control signal and endpoint variability. The empirical question is whether the adaptation improves successful task completion across users and tasks and whether assistance changes the operator's own movement strategy.

Assistance also introduces a new source of error when the system predicts the wrong intention. A target snap that helps in a sparse interface may pull the pointer toward the wrong destination in a dense display; an automatic gain change may disrupt learned movements at precisely the moment an unexpected action is needed. Shadmehr & Mussa-Ivaldi (1994) demonstrates that motor behavior adapts to changed dynamics, implying that assistance policies can themselves become learned control environments. Evaluation should therefore include false assistance, policy changes, first use, long-term learning, and recovery after incorrect predictions. A system that improves mean speed under accurate prediction may be unacceptable if rare false predictions produce difficult-to-correct critical actions. The predictor's confidence and the reversibility of assisted actions become part of the behavioral mechanism.

Agency and transparency matter because a user must know when an output resulted from their movement and when a system changed the effective action. A controller that covertly modifies pointer movement may reduce visible error while making a user's internal mapping less reliable. Such effects are hypotheses requiring direct comparison of assisted and unassisted conditions, including explicit tests of transfer after assistance is withdrawn. Wolpert et al. (1995) provides a theoretical foundation for the importance of state estimation in action, while Zhai (2004) warns that single-number performance summaries can conceal crucial methodological differences. An adaptive-control study should publish a policy description, error classifications, user-group coverage, and conditions under which the system overrides or preserves intended action.

Human–Artificial Coordination and the Boundaries of Automated Agency

In the Artificial Era, control interfaces increasingly mediate interaction between human intention and systems that estimate, recommend, or execute actions. A person may guide a physical robot, command a simulated agent, or collaborate with software that proposes a likely target. Such settings make the distribution of control explicit: human movement provides one information source, automated inference another, and the interface determines how they jointly affect external behavior. The experimental core remains recognizable. A target has an operational geometry, response choices have mappings, and action consequences create error costs. Fitts (1954) and Fitts & Seeger (1953) establish the separate movement and correspondence traditions that any contemporary shared-control system must still address.

The psychological effects of assistance should be measured through observable task behavior. Relevant outcomes include the user's movement time, correction rate, dependence on assistance, recognition of a mismatched prediction, time to regain direct control, and accuracy after the system changes its policy. These variables support testable hypotheses about reliance and skill maintenance without requiring claims about unobserved inner states of the artificial component. The system itself can be characterized by policy stability, latency, prediction error, and the constraints under which it acts. Shadmehr & Mussa-Ivaldi (1994) provides a relevant model of how a human operator adapts to changing dynamics; Wolpert et al. (1995) underscores the role of prediction in motor control. Whether these mechanisms generalize to any specific AI-assisted controller is a question for new experiments.

This distinction is significant for Psychology of Human–Artificial Relations and Psychology of Artificial, but it does not make those domains synonymous with human motor performance. The present dossier addresses the human action and interface layer and proposes research protocols for systems that redistribute action selection or pointer control. The adequacy of such systems depends on reliable measurement of correspondence, error recovery, and continuity of skill across autonomous and manual phases. A design can promote effective human–Artificial cooperation by making predictions inspectable, assistance reversible, and manual overrides predictable. These are design propositions with empirical consequences, to be tested in representative tasks rather than treated as already established psychological laws Proctor & Vu (2006), Soukoreff & MacKenzie (2004).

Contemporary Experimental Evidence from Immersive Systems

Recent investigations extend motor-performance research to environments where an individual selects targets using tracked hand motion, controllers, or virtual rays. These studies are valuable because they vary sensing modality and movement geometry while measuring the speed–accuracy relationship. Babu et al. (2022) reported two virtual-reality experiments involving targets in personal space, comparing controller input with lower- and higher-fidelity hand-pointing methods. Its abstract and accessible paper indicate that lower-fidelity pointing performed worse in the examined conditions and that selection near the limits of arm reach was particularly challenging. These observations are informative about the tested tracking systems and reach ranges. They also reinforce a design lesson often absent from a simple Fitts formula: the measured geometry extends beyond the virtual screen to the operator's physical capacity and sensor reliability.

Luong et al. (2023) compared VR controllers with free-hand interaction across selection and trajectory-tracing tasks, examining performance, physical exertion, agency-related reports, and motor behavior. The published abstract reports that controllers yielded favorable speed and accuracy in the raycasting setting, whereas modality preferences varied by interaction technique. The study therefore supports a conditional comparison rather than a universal ranking of controllers and hand tracking. It demonstrates why psychological performance and ergonomic burden must be evaluated together in repeated mid-air action. Device choice affects both perceived controllability and actual movement dynamics, and the task's selection mechanism can moderate the observed preference. Claims about long-term occupational health would require research beyond the duration and outcomes of the reported laboratory study.

Abe et al. (2025) investigated VR pointing methods for onsite exhibitions using a handheld-style head-mounted display, a context in which device configuration and surrounding activity differ from standard seated laboratory demonstrations. Rodin et al. (2026) evaluated modality-specific Fitts-type models in virtual reality and their transfer to applied interaction tasks. These recent works extend the evidence base to evolving hardware and usage contexts, but their models remain conditional on participant sampling, implementation details, and test geometry. The relevant synthesis is methodological: a predictive model trained on one set of standardized movements must be validated on independent tasks before it guides deployed interface decisions. The more a technology changes the operator's posture, sensory feedback, or activation rules, the more carefully the model's original assumptions should be reconsidered.

Assisted Target Selection and the Bubble Cursor

Adaptive target acquisition can be studied through systems that expand a pointer's effective activation region in response to surrounding targets. Grossman & Balakrishnan (2005) introduced the bubble cursor, an area-cursor technique that changes its activation region according to the proximity of selectable objects, with the design goal of allowing only one candidate to be selected at a time. The original CHI paper reported controlled studies in one- and two-dimensional pointing configurations and evaluated performance under varying target layouts. Its significance lies in modifying the action geometry rather than merely asking users to move faster. The candidate selection region changes as part of the interface, so the physical motion necessary to activate a target differs from that with a conventional point cursor.

The bubble-cursor approach illustrates how a useful engineering intervention can emerge from a psychological model without reducing to that model. Fitts-type reasoning predicts benefits when effective target width increases, but the intervention also creates a spatial partition among nearby controls and changes how users perceive the currently selectable object. Grossman & Balakrishnan (2005) therefore invites measurement of selection ambiguity, cursor feedback, correction, and performance at different target densities. A change that benefits discrete pointing may be less suitable when users need to trace a path, cross a narrow boundary, or interact with layered three-dimensional objects. The broader methodological principle is to treat effective target geometry as a controlled property of the human–system system and to test the side effects of changing it.

Adaptive selection also raises a continuity problem. If the activation region changes from one moment to the next, the user must predict the resulting selection boundary. The same algorithm may be beneficial in a sparse toolbar and confusing in a dense scientific visualization. Zhai (2004) underscores why device performance cannot be reduced to an unqualified throughput number, while Soukoreff & MacKenzie (2004) provides compatible reporting guidelines for speed and accuracy. A deployment-oriented evaluation should include target identification, inadvertent activation, repeated selection sequences, and comprehension of the assistance mechanism. It should test conditions where the system's nearest-target choice conflicts with the operator's intention and report the time and steps needed to restore the intended action.

Theoretical Disagreement and Explanatory Limits

The enduring popularity of a logarithmic movement-time relation can conceal substantial disagreement about why the relation arises. An information-channel analogy treats accuracy tolerance as a constraint on the capacity of the motor system; a submovement account emphasizes initial approach and corrective action; an optimal-control formulation emphasizes movement planning under noise; and sensorimotor state-estimation models emphasize prediction from incomplete feedback. Fitts (1954), Meyer et al. (1988), Harris & Wolpert (1998), and Wolpert et al. (1995) represent different components of this explanatory history. These accounts can agree on a broad speed–accuracy regularity while generating distinct predictions about trajectories, response variability, perturbation responses, and movement corrections. Empirical model choice therefore requires observations designed to discriminate among the candidate mechanisms.

Stimulus–response compatibility likewise combines distinguishable mechanisms. A benefit for corresponding spatial positions may reflect response coding, irrelevant stimulus interference, learned associations, or the reference frame chosen by the operator. Fitts & Seeger (1953) and Simon (1969) established important experimental phenomena, Kornblum et al. (1990) modeled dimensional overlap, and Hommel (1993) addressed referential coding. A control that is compatible under one set of instructions may lose its advantage when the user changes viewpoint or interprets a different operational goal. The evidence supports conditional design principles linked to the manipulations that produced them. It does not support a single innate compatibility mapping for all control environments.

A further limitation is the tendency to equate successful prediction of mean time with a complete account of motor behavior. Two models may explain similar averages while diverging markedly at the edges of the speed–accuracy function, for persons with motor impairment, or during learning. Zhai (2004) warns against metrics that change with the sampled task difficulty, and Wobbrock & Gajos (2008) documents how interaction techniques can behave differently across user groups. For research that informs practice, the strongest test is not whether a preferred formula fits the familiar sample but whether it predicts previously unseen users, devices, layouts, and consequences with uncertainty appropriate to the design decision.

Measurement Validity, Bias, and Generalizability

Construct validity begins with the meaning of the outcome. Movement time is not overall task performance; selection throughput is not physical comfort; perceived ease is not proof of accuracy; and a low mean error rate is not evidence that rare critical mistakes are impossible. Each claim requires an appropriate observable measure and a procedure that makes alternative explanations plausible or less plausible. Soukoreff & MacKenzie (2004) supports consistent movement measurement, while Cockburn et al. (2007) demonstrates the need to account for search and decision processes in real interface tasks. A study using a fully cued target cannot establish efficient visual navigation, just as a study of visual recognition cannot establish precise motor activation. The research construct should determine the design rather than be defined retroactively by whatever variable was recorded.

Selection bias is particularly relevant when study recruitment depends on facility with the technology. An experiment comparing VR controllers and free-hand input may overrepresent participants with gaming experience, good upper-limb function, and tolerance of immersive displays. Babu et al. (2022) and Luong et al. (2023) provide useful device-specific evidence; the external validity of any reported ranking depends on their sampled users, tasks, and hardware. Reporting average performance across a heterogeneous sample can conceal the experiences of small but important user groups. Where the design targets an accessibility population, subgroup estimates and individual trajectories should be part of the scientific account, subject to appropriate sample-size and privacy considerations. Practical generalization demands evidence in the population that will use the system.

Publication and measurement biases affect the field as a whole. Novel interactions that significantly improve a familiar benchmark may attract attention, while studies with null or context-dependent outcomes can be harder to find. A critical narrative synthesis cannot quantify such bias without a systematic search and explicit inclusion criteria. The appropriate response is to state the evidence-selection method, emphasize replications and independent settings when available, and distinguish closely verified experimental effects from promising hypotheses. The present dossier's sources provide a representative cross-section of seminal and contemporary research, not an exhaustive census. Its conclusions are calibrated to study design and applicability, preserving space for results that contradict a simple device ranking or an expected geometric advantage.

An Integrated Architecture of Human Action in Control Systems

The evidence reviewed here supports a layered understanding of controlled action. An operator must obtain relevant task information, choose an intended state change, select the corresponding response, initiate and regulate movement, confirm whether the action succeeded, and recover when the result diverges from the goal. The stages are analytically distinguishable while their actual execution can overlap. A system may support fast movement but introduce wrong-control choices; another may present a clear mapping while making precise contact difficult. Fitts & Peterson (1964) documents the distinction between reaction and movement time in discrete actions, Kornblum et al. (1990) explains aspects of response-code conflict, and Cockburn et al. (2007) shows the practical importance of analyzing search, choice, and pointing together. A credible account of human performance therefore identifies the functional units that make a task succeed.

Target geometry constrains movement planning and terminal accuracy; mapping compatibility shapes which movement is selected; physical posture and device dynamics shape the achievable trajectory; feedback shapes correction and confidence; and task consequences shape the operator's preferred balance of speed and caution. These are interacting processes, so a design change can improve one measure while worsening another. Enlarging a control may reduce terminal precision demands while increasing visual clutter. Moving a frequently used control nearer may shorten reach but disrupt spatial mapping learned through practice. Automation may reduce movement effort while making action consequences less transparent. Soukoreff & MacKenzie (2004) provides a foundation for measuring the motor component, and Proctor & Vu (2006) provides the response-compatibility framework. Integration is achieved by keeping their distinct causal questions visible within one empirical protocol.

This architecture suggests evaluating performance at three levels. At the movement level, measure amplitude, accuracy, dynamics, and correction. At the selection level, measure identification, response mapping, and wrong-control activation. At the task level, measure successful consequences, recovery, persistence, and physical sustainability. The levels correspond to different possible interventions and forms of evidence. Claims that an interface supports autonomy or safe operation require task-level validation even when movement-level metrics improve. In a human–Artificial system, the same architecture can be extended by measuring the conditions under which an automated component predicts, modifies, or executes a response and whether the human operator can detect and correct an unwelcome deviation. These are empirically tractable questions anchored in motor and human-factors psychology.

Research Priorities for Accessible, Predictive, and Reliable Control

The first research priority is transportability of performance models across modalities and populations. Studies can train predictive models on a standard pointing task and evaluate them on independent real-world selections, with results stratified by device, experience, functional ability, and age. Rodin et al. (2026) offers a contemporary virtual-reality example of cross-task predictive evaluation, while Babu et al. (2022) shows why reach depth and tracking quality may alter performance. The relevant questions are whether a model's slope and intercept remain stable, whether its uncertainty correctly predicts error, and where the model systematically fails. Researchers should publish task geometry, controller dynamics, and measurement rules so the results can support reproducible model comparison rather than an isolated performance ranking.

A second priority is evaluation of assistance under rare but consequential mispredictions. Adaptive pointers, target expansion, and shared-control methods should be tested under both favorable and deliberately misleading assistance. Grossman & Balakrishnan (2005) provides a concrete target-expansion intervention, and Shadmehr & Mussa-Ivaldi (1994) supports the study of adaptation when control mappings change. Researchers can evaluate the immediate benefit, the cost of false assistance, and the effect of repeated reliance on later unaided performance. A design that improves standard benchmark throughput can still be unsuitable if it makes operator intent difficult to restore after an erroneous prediction. This problem is relevant across rehabilitation, industrial interfaces, accessible computing, and human–Artificial interaction.

A third priority is ecologically valid measurement of fatigue, accessibility, and control reliability over time. Luong et al. (2023) demonstrates the importance of including physical exertion and agency-related reports alongside selection performance in immersive input. Wobbrock & Gajos (2008) shows why heterogeneous motor capabilities and activation gestures should shape experimental design. Longer studies should measure how posture, endurance, interface familiarity, and error recovery evolve across realistic work periods. A fourth priority is explicit experimental comparison of compatibility mappings under changed reference frames and operational consequences, building on Fitts & Seeger (1953), Hommel (1993), and Proctor & Vu (2006). Across these programs, the scientific objective is predictive understanding of successful action under stated constraints, supported by primary observations and independently testable claims.

Scientific Synthesis and Implications for the Artificial Era

Human motor performance is best understood as an organized relation between intention, information, bodily movement, control mechanics, and the system state that results. The research traditions examined here identify repeatable contributions from target geometry and precision demands, response-code correspondence, predictive and feedback-based regulation, practice, posture, and interaction modality. The strongest experimental findings concern specified paradigms: Fitts (1954) and Fitts & Peterson (1964) for aimed movement, Fitts & Seeger (1953) and Simon (1969) for spatially organized response selection, Meyer et al. (1988) and Harris & Wolpert (1998) for explanatory models of action, and Wobbrock & Gajos (2008) for a consequential accessible interaction comparison. They form a coherent scientific foundation when their methods and scope remain explicit.

The modern interface expands the locus of control. Some applications rely on an individual's direct hand motion; others translate movements through nontransparent transfer functions or supplement them with inferred intent. These conditions change the feedback available for motor adaptation and the evidence needed to establish reliable task performance. Luong et al. (2023), Babu et al. (2022), and Rodin et al. (2026) contribute evidence from immersive and modality-dependent tasks, showing that the experimental setting matters to both performance and physical experience. A human–Artificial collaboration model can be built from observable properties: action proposals, corrections, latency, reversibility, learned reliance, and maintenance of human operating skill. The author's Artificial Era framework organizes the historical significance of these questions, while empirical claims remain grounded in studies of measured behavior and system performance.

The practical implication is a research discipline of matched measures. A movement-law analysis can justify investigating smaller movement difficulty; a compatibility experiment can justify investigating clearer response mapping; an ergonomic assessment can justify investigating posture and exertion; and a field simulation can justify investigating recovery from errors. These approaches are strongest when they inform one another without collapsing into a single surrogate outcome. The lasting contribution of human motor research to control engineering lies in making action measurable and its limiting conditions experimentally revisable.

Conclusion

Human motor performance, target acquisition, and control compatibility constitute a unified field of investigation centered on the successful transformation of intended action into reliable system change. Historical experiments established distinct but complementary regularities. Fitts-type studies demonstrated systematic relations among movement amplitude, target tolerance, and time; stimulus–response compatibility studies demonstrated that the mapping between perceived signals and required controls shapes response selection; subsequent work explained movement variability through feedback, correction, prediction, and learned dynamics. The accumulated evidence supports the use of these principles as experimental instruments for analyzing designed action under specified constraints Fitts (1954), Fitts & Seeger (1953), Meyer et al. (1988), Proctor & Vu (2006).

Contemporary research has extended the field into pointer interfaces, constrained trajectories, accessible selection, and virtual environments. It has also clarified the conditions required for useful inference. Movement time must be separated from response choice and activation; device rankings must state how accuracy and throughput were calculated; representation and physical posture matter; accessibility requires population-relevant evidence; and high-stakes interaction must be judged by error consequences and recovery as well as speed. Soukoreff & MacKenzie (2004), Zhai (2004), Wobbrock & Gajos (2008), and Luong et al. (2023) provide complementary foundations for these conclusions. No single pointing benchmark determines the usability of every control system.

The next phase of research concerns systems in which artificial components transform the movement process through prediction, adaptive target acquisition, or shared control. These developments require new experimental measurements of assistance errors, mapping stability, transfer, fatigue, and the human ability to regain intended control. A mature scientific architecture links theory to these observable outcomes, preserves the distinction between explanatory models and demonstrated effects, and tests proposed interventions across representative users and operational contexts. Within the historical horizon of the Artificial Era, human motor control remains a foundational case for understanding how embodied human action is organized in increasingly artificial environments.

References

  • Abe, Y., Kusakabe, K., Choi, M., Sakamoto, D., & Ono, T. (2025). Understanding usability of VR pointing methods with a handheld-style HMD for onsite exhibitions. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (Article 554, pp. 1–21). Association for Computing Machinery. https://doi.org/10.1145/3706598.3713874
  • Accot, J., & Zhai, S. (1997). Beyond Fitts’ law: Models for trajectory-based HCI tasks. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 295–302). ACM. https://doi.org/10.1145/258549.258760
  • Accot, J., & Zhai, S. (2003). Refining Fitts’ law models for bivariate pointing. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 193–200). ACM. https://doi.org/10.1145/642611.642646
  • Babu, S. V., Huang, H.-C., Teather, R. J., & Chuang, J.-H. (2022). Comparing the fidelity of contemporary pointing with controller interactions on performance of personal space target selection. In 2022 IEEE International Symposium on Mixed and Augmented Reality (ISMAR) (pp. 404–413). IEEE. https://doi.org/10.1109/ISMAR55827.2022.00056
  • Cockburn, A., Gutwin, C., & Greenberg, S. (2007). A predictive model of menu performance. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 627–636). ACM. https://doi.org/10.1145/1240624.1240723
  • Elliott, D., Helsen, W. F., & Chua, R. (2001). A century later: Woodworth’s (1899) two-component model of goal-directed aiming. Psychological Bulletin, 127(3), 342–357. https://doi.org/10.1037/0033-2909.127.3.342
  • Fitts, P. M. (1954). The information capacity of the human motor system in controlling the amplitude of movement. Journal of Experimental Psychology, 47(6), 381–391. https://doi.org/10.1037/h0055392
  • Fitts, P. M., & Peterson, J. R. (1964). Information capacity of discrete motor responses. Journal of Experimental Psychology, 67(2), 103–112. https://doi.org/10.1037/h0045689
  • Fitts, P. M., & Seeger, C. M. (1953). S-R compatibility: Spatial characteristics of stimulus and response codes. Journal of Experimental Psychology, 46(3), 199–210. https://doi.org/10.1037/h0062827
  • Grossman, T., & Balakrishnan, R. (2005). The bubble cursor: Enhancing target acquisition by dynamic resizing of the cursor’s activation area. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 281–290). ACM. https://doi.org/10.1145/1054972.1055012
  • Guiard, Y. (1987). Asymmetric division of labor in human skilled bimanual action: The kinematic chain as a model. Journal of Motor Behavior, 19(4), 486–517. https://doi.org/10.1080/00222895.1987.10735426
  • Harris, C. M., & Wolpert, D. M. (1998). Signal-dependent noise determines motor planning. Nature, 394(6695), 780–784. https://doi.org/10.1038/29528
  • Hommel, B. (1993). Inverting the Simon effect by intention: Determinants of direction and extent of effects of irrelevant spatial information. Psychological Research, 55, 270–279. https://doi.org/10.1007/BF00419687
  • International Organization for Standardization. (2007). Ergonomics of human-system interaction—Part 400: Principles and requirements for physical input devices (ISO 9241-400:2007). https://www.iso.org/standard/38896.html
  • International Organization for Standardization. (2012). Ergonomics of human-system interaction—Part 411: Evaluation methods for the design of physical input devices (ISO/TS 9241-411:2012). https://www.iso.org/standard/54106.html
  • Kornblum, S., Hasbroucq, T., & Osman, A. (1990). Dimensional overlap: Cognitive basis for stimulus-response compatibility—a model and taxonomy. Psychological Review, 97(2), 253–270. https://doi.org/10.1037/0033-295X.97.2.253
  • Luong, T., Cheng, Y. F., Möbus, M., Fender, A., & Holz, C. (2023). Controllers or bare hands? A controlled evaluation of input techniques on interaction performance and exertion in virtual reality. IEEE Transactions on Visualization and Computer Graphics, 29(11), 4633–4643. https://doi.org/10.1109/TVCG.2023.3320211
  • MacKenzie, I. S. (1992). Fitts’ law as a research and design tool in human-computer interaction. Human–Computer Interaction, 7(1), 91–139. https://doi.org/10.1207/s15327051hci0701_3
  • Meyer, D. E., Abrams, R. A., Kornblum, S., Wright, C. E., & Smith, J. E. K. (1988). Optimality in human motor performance: Ideal control of rapid aimed movements. Psychological Review, 95(3), 340–370. https://doi.org/10.1037/0033-295X.95.3.340
  • Proctor, R. W., & Vu, K.-P. L. (2006). Stimulus-response compatibility principles: Data, theory, and application. CRC Press. https://books.google.com/books?id=NISHh4ZJV4AC
  • Proctor, R. W., & Vu, K.-P. L. (2010). Stimulus–response compatibility for mixed mappings and tasks with unique responses. Quarterly Journal of Experimental Psychology, 63(2), 320–340. https://doi.org/10.1080/17470210902925270
  • Rodin, N., Ogrizović, D., Batistić, L., & Ljubić, S. (2026). Empirical validation of Fitts’ law in virtual reality: Modeling, prediction, and modality comparison. Multimodal Technologies and Interaction, 10(5), Article 49. https://doi.org/10.3390/mti10050049
  • Salthouse, T. A. (1984). Effects of age and skill in typing. Journal of Experimental Psychology: General, 113(3), 345–371. https://doi.org/10.1037/0096-3445.113.3.345
  • Schmidt, R. A., Zelaznik, H., Hawkins, B., Frank, J. S., & Quinn, J. T., Jr. (1979). Motor-output variability: A theory for the accuracy of rapid motor acts. Psychological Review, 86(5), 415–451. https://doi.org/10.1037/0033-295X.86.5.415
  • Shadmehr, R., & Mussa-Ivaldi, F. A. (1994). Adaptive representation of dynamics during learning of a motor task. Journal of Neuroscience, 14(5), 3208–3224. https://doi.org/10.1523/JNEUROSCI.14-05-03208.1994
  • Simon, J. R. (1969). Reactions toward the source of stimulation. Journal of Experimental Psychology, 81(1), 174–176. https://doi.org/10.1037/h0027448
  • Soukoreff, R. W., & MacKenzie, I. S. (2004). Towards a standard for pointing device evaluation, perspectives on 27 years of Fitts’ law research in HCI. International Journal of Human-Computer Studies, 61(6), 751–789. https://doi.org/10.1016/j.ijhcs.2004.09.001
  • Welford, A. T., Norris, A. H., & Shock, N. W. (1969). Speed and accuracy of movement and their changes with age. Acta Psychologica, 30, 3–15. https://doi.org/10.1016/0001-6918(69)90034-1
  • Wobbrock, J. O., & Gajos, K. Z. (2008). Goal crossing with mice and trackballs for people with motor impairments: Performance, submovements, and design directions. ACM Transactions on Accessible Computing, 1(1), Article 4, 1–37. https://doi.org/10.1145/1361203.1361207
  • Wolpert, D. M., Ghahramani, Z., & Jordan, M. I. (1995). An internal model for sensorimotor integration. Science, 269(5232), 1880–1882. https://doi.org/10.1126/science.7569931
  • Zhai, S. (2004). Characterizing computer input with Fitts’ law parameters—the information and non-information aspects of pointing. International Journal of Human-Computer Studies, 61(6), 791–809. https://doi.org/10.1016/j.ijhcs.2004.09.006