Temporal Units

Temporal Units are the formal and scientific entities used to locate, delimit, measure, group, and compare behavioral-signal evidence in time. They provide the foundational framework for representing when behaviors occur, how long they last, and how different temporal aspects relate to each other in behavioral signal processing. It is critical to distinguish temporal measurement units—such as seconds or milliseconds, which quantify elapsed time—from structural temporal units—such as instants, intervals, windows, events, states, episodes, trials, or other temporally bounded evidential supports—that organize and interpret evidence within specific temporal boundaries. A temporal unit is meaningful only when its temporal semantics, reference system, boundaries, duration or extent, precision, and scientific role are sufficiently specified and understood.

Meaning of Temporal Units

A temporal unit can be defined in two complementary senses:

1. As a unit of temporal quantity, it expresses a duration or a position in time. Examples include seconds, milliseconds, or any other fixed or relative measure used to quantify elapsed time or timestamp occurrences.

2. As an evidential unit, it identifies a temporally delimited portion or occurrence within recorded evidence. This includes identifiable segments such as events, states, episodes, or trials that represent meaningful behavioral or signal phenomena.

The use of the single word unit should not obscure the fundamental difference between measuring time (quantifying duration or position) and organizing evidence in time (defining temporal segments with scientific interpretation).

Temporal concepts can be distinguished as follows:

• Temporal position: A temporal coordinate or timestamp that locates an instant relative to a reference system.
• Temporal extent: The duration or elapsed time between two temporal positions.
• Temporal support: The declared interval or set of intervals over which evidence is considered applicable or valid.
• Temporal identity: The scientific or operational characterization of a temporal entity—such as an event, state, episode, or trial—that carries meaning beyond mere temporal coordinates.

Instants, Intervals, and Durations

An instant is a point-like temporal entity with no interior temporal extent in the representation. It marks a specific time position but does not cover any measurable duration. Representing an occurrence as an instant can be an analytical idealization even when the underlying physical or behavioral process had nonzero duration.

An interval is a temporal entity with distinguishable beginning and end points. It represents a continuous or coherent portion of time during which something occurs or persists.

Interval boundaries and their semantics involve specifying the start and end times, which are temporal positions expressed in the same reference system. Intervals can be represented as:

• Closed: Both start and end included.
• Open: Both start and end excluded.
• Half-open: Either start included and end excluded, or vice versa.

Defining these boundary conventions is essential because adjacent intervals must have an unambiguous membership rule for observations exactly at a boundary. This rule must be compatible with timestamp precision, event semantics, and the intended computation or analysis.

The duration ( d ) of an interval is given by the elapsed time between its start and end:

Here, ( t_{start} ) and ( t_{end} ) are temporal positions expressed in a compatible reference system, and ( d ) is the resulting elapsed duration. This subtraction is valid only when the coordinate representation supports meaningful elapsed-time subtraction. Calendar labels, clock adjustments, discontinuous clocks, or differing reference systems may require additional interpretation before subtraction.

Duration must be distinguished from related concepts such as occupancy, exposure, active time, and observation time. For example, an interval may last ten minutes but contain only intermittent valid observations, partial participant presence, or multiple state changes. Therefore, the temporal extent of a support should not be equated automatically with the amount of usable or behaviorally active evidence within it.

Time Measurement Units and Reference Coordinates

Physical time measurement is commonly based on the second, which is the International System of Units (SI) base unit of time. Scaled units such as milliseconds, microseconds, minutes, and hours are contextually useful representations derived by exact conversions.

Changing from seconds to milliseconds changes only the numerical scale, not the underlying temporal quantity, when the conversion is exact.

Fixed-duration units like seconds and their decimal multiples express precise elapsed times. In contrast, calendar-based durations such as days, months, years, local civil times, daylight-saving transitions, leap-related conventions, or calendar periods carry reference-system semantics that cannot be treated as interchangeable fixed numbers of seconds without justification.

Temporal coordinates can be classified as:

• Absolute or reference-system timestamps: Locating observations on a declared time scale.
• Elapsed time: Measuring duration from a known origin.
• Relative time: Expressing position with respect to a chosen anchor such as event onset, session start, or another scientifically defined reference.

Changing the temporal origin shifts the coordinate system but does not inherently improve timestamp accuracy.

An event- or reference-relative temporal coordinate ( \tau ) is given by:

Here, ( t ) is the original temporal coordinate, ( t_{ref} ) is the declared reference instant, and ( \tau ) is the relative time. The value ( \tau = 0 ) identifies the chosen reference in the transformed coordinate system but does not imply that the reference instant itself is known without uncertainty.

It is important to distinguish a temporal coordinate from a sample, frame, packet, or record index. An index identifies order or storage position; converting it to elapsed time requires valid knowledge of sampling behavior or timestamp mapping. Equal increments in index should not be assumed to represent equal increments in physical time when jitter, dropped observations, variable frame rate, asynchronous events, or irregular acquisition are possible.

Structural Units of Behavioral Evidence

Events are temporally identifiable occurrences that can be encoded as point-like anchors or as extended intervals depending on the phenomenon and analytical purpose. It is crucial to distinguish event occurrence time from event duration, event identity, detection time, and recording time.

States are temporally persistent conditions or regimes that occupy intervals, usually characterized by continuity or coherence relative to declared properties. A state can contain many local fluctuations or events and should not be equated with a constant-valued signal.

Episodes are temporally coherent occurrences that can include several events, states, transitions, or actions. Their identity depends on substantive behavioral, physiological, interactional, contextual, or operational meaning. An episode should not be defined solely by having a convenient fixed duration.

Windows and bins are analytical temporal supports created to aggregate, estimate, compare, or represent evidence over declared extents. Their boundaries can be independent of natural behavioral boundaries, and the same event or state can contribute to several overlapping windows.

Trials, bouts, turns, encounters, sessions, and other protocol- or phenomenon-defined temporal units are named evidential supports whose semantics depend on the domain. Their inclusion criteria, start and end semantics, permissible interruptions, and relationship to events or states should be explicit rather than inferred from the label.

Relations Between Temporal Units

Qualitative relations between temporal intervals allow expression of temporal structure independently of exact numerical timestamps. Representative Allen-style interval relations include:

• Before and After
• Meets and Met-by
• Overlaps and Overlapped-by
• Starts and Started-by
• During and Contains
• Finishes and Finished-by
• Equals

These relations describe interval topology and do not by themselves specify metric distance or duration.

Temporal adjacency, overlap, containment, and disjointness are scientifically meaningful relations among behavioral units:

• Adjacent episodes differ from overlapping activities.
• One event can occur during a state.
• One state can contain several events.
• Two temporal supports can be disjoint despite having nearby timestamps.

Metric proximity should not be conflated with overlap or containment.

Granularity, Scale, Resolution, and Precision

Temporal granularity refers to the characteristic level at which time is represented or grouped, such as microseconds, samples, frames, seconds, actions, episodes, or sessions depending on context. Fine granularity means representing time in small units, but this does not necessarily imply high accuracy.

Distinguish:

• Temporal resolution: The smallest temporal distinctions the system or representation can meaningfully resolve.
• Timestamp precision: The repeatability or detail level in representation.
• Timing accuracy: Closeness to the relevant temporal truth or reference.
• Uncertainty: Incomplete knowledge about timing accuracy or precision.

These terms must not be used interchangeably.

Temporal units are scale-dependent. The same evidence may be represented meaningfully as fine-grained events, intermediate episodes, or coarse states or sessions. Each scale preserves different information, and no single temporal scale is universally fundamental for all behavioral questions.

Temporal quantization and rounding occur when timestamps or durations are mapped to coarser units. This process can merge distinct events, move boundaries, change apparent simultaneity, or alter interval relations near quantization thresholds. Displayed precision should not exceed the precision justified by the underlying temporal evidence.

A scale mismatch between the temporal unit and phenomenon can fragment coherent behavior if the unit is too short or merge distinct events or states if the unit is too long. Choosing temporal units requires matching representational scale to the temporal properties needed for the scientific claim.

Temporal Units Across Multiple Sources

Different sources can use source-specific temporal units. For example, streams may represent time through samples, frames, packets, events, states, or intervals with distinct native rates, timestamp semantics, and uncertainties. A unit meaningful for one source should not be assumed to have the same temporal resolution or evidential meaning in another.

Common and source-relative temporal references can express several streams on a shared time scale or relative to a common event. However, a common coordinate system alone does not establish valid synchronization, identical latency, or equivalent timestamp uncertainty.

Temporal correspondence between units from different sources acknowledges that related aspects of a process (e.g., movement episode, speech turn, physiological response) may begin and end at different times. Correspondence can tolerate scientifically meaningful lag rather than require identical boundaries.

Composite and aggregate temporal units are scientifically useful temporal entities composed of several disjoint or adjacent intervals, repeated occurrences, or source-specific supports rather than a single continuous interval. Aggregation should preserve whether members are continuous, recurring, overlapping, or separated by gaps when those relations affect interpretation.

Choosing and Evaluating Temporal Units

Criteria for choosing a temporal unit include:

• Phenomenon duration
• Event frequency
• Expected state persistence
• Required timing precision
• Sampling behavior
• Uncertainty
• Analytical purpose
• Cross-source comparability
• Amount of evidence needed for stable estimation

A conventional unit should not be adopted without verifying its semantics match the intended claim.

Temporal-unit choice impacts invariance and sensitivity:

• Some scientific quantities remain stable under harmless changes of temporal measurement unit (e.g., seconds to milliseconds).
• Others change substantially when structural units, window extents, boundary conventions, or aggregation scales change.

Distinguish numerical re-expression from substantive reorganization.

Validation of temporal-unit semantics can use:

• Known events
• Protocol definitions
• Independent timing references
• Behavioral observations
• Device metadata
• Repeated measurements
• Sensitivity analysis

Validation tests whether the chosen unit faithfully represents the phenomenon and support required by the scientific use, not merely whether timestamps are syntactically valid.

Temporal-unit sensitivity arises from:

• Boundary convention
• Granularity
• Unit conversion
• Anchor choice
• Event representation
• Aggregation scale

When plausible temporal-unit definitions materially change event counts, state durations, overlap relations, temporal descriptors, cross-source comparisons, or scientific conclusions, that dependence should remain visible.

Ambiguity and uncertainty in named temporal units arise because terms such as epoch, bout, episode, session, encounter, turn, or trial can have discipline-specific meanings. They should be operationally defined by temporal and substantive criteria. Familiar labels should not substitute for explicit duration, boundary, membership, and reference semantics.

Temporal-Unit Provenance and Scientific Interpretation

Temporal-unit provenance is the information needed to reproduce and interpret temporal entities and quantities. When relevant, this includes:

• Time measurement unit
• Temporal reference system
• Timestamp semantics
• Origin or anchor
• Start and end conventions
• Duration rule
• Structural-unit definition
• Sampling or index mapping
• Granularity
• Precision
• Uncertainty
• Boundary tolerance
• Membership criteria
• Source identity
• Unit conversion
• Aggregation or composite-unit rules

Temporal units matter in Behavioral Signal Processing because they determine how duration, order, overlap, persistence, event occurrence, state occupancy, cross-source relations, and temporal summaries are interpreted. A defensible temporal unit makes explicit whether it represents a quantity, coordinate, support, occurrence, state, or analytical grouping and preserves the reference, scale, boundaries, uncertainty, and substantive meaning needed for the intended scientific claim.