Temporal Extent and Support

Temporal Extent and Support are fundamental concepts used to describe how far a temporal entity spans, which portions of time are actually represented or scientifically usable, and how those portions are bounded, fragmented, combined, and compared. Temporal extent concerns the temporal reach or duration of an entity, whereas temporal support identifies the set of temporal positions over which a signal, observation, validity condition, state, or analytical claim applies. Equal start and end times do not imply equal usable support, as temporal coverage can contain gaps, and a zero-valued observation can remain validly supported despite having no nonzero values.

Meaning of Temporal Extent and Support

Temporal extent is defined as the amount or reach of time associated with an entity, observation, interval, record, or phenomenon. Temporal support is the set of times at which a declared signal, observation, validity condition, behavioral unit, or analytical quantity is defined, available, applicable, or scientifically usable. The semantics of support must always specify what is being supported: whether it is a mathematical signal, a recorded observation, a valid measurement, an inferred state, a reconstructed value, or a scientific analysis.

Mathematical signal support differs from evidential support. In signal theory, support commonly denotes the set of time points where a function is nonzero. In behavioral evidence, valid temporal support can include zero-valued observations and exclude nonzero observations judged invalid, artifactual, missing, censored, or unusable for a particular purpose. The intended meaning of "support" should always be explicitly stated rather than inferred from notation alone.

Temporal Extent, Span, and Duration

Temporal span is the interval from the earliest supported or declared time to the latest supported or declared time for a nonempty bounded temporal set. Span describes the outer temporal reach of an entity and can include internal periods where there is no observation, no valid evidence, or no target behavior.

It is important to distinguish span, elapsed duration, active duration, valid-observation duration, and cumulative supported duration. For example, a ten-minute recording can span ten minutes while containing eight minutes of valid evidence, six minutes of target-state occupancy, and several gaps. Each quantity answers different scientific questions and should not be substituted for one another.

The span of a nonempty bounded temporal support is rendered as:

Here, S is the temporal support, inf(S) and sup(S) are its earliest and latest temporal limits, and D_span is the outer span. This quantity ignores internal gaps and therefore need not equal the cumulative duration of supported evidence.

Temporal extents can be bounded, unbounded, open-ended, left-censored, right-censored, or truncated. A phenomenon may begin before observation starts, continue after observation ends, or be observed only inside a restricted acquisition interval. The observed extent should never be presented automatically as the full extent of the underlying behavior or process.

Temporal Support and Evidence Status

Observation support refers to the times for which a recording system produced evidence. Valid support is the subset of observation support usable for a declared scientific purpose. Recorded values may fall outside valid support due to artifacts, saturation, timing uncertainty, protocol violations, insufficient quality, or other exclusion criteria.

Temporal statuses such as observed, valid, invalid, missing, rejected, interpolated, reconstructed, and unknown can coexist within a temporal representation. These statuses must remain distinguishable because directly observed support and model-derived support carry different evidential meanings.

Support is purpose-dependent: evidence valid for coarse activity estimation may be invalid for precise event timing, spectral analysis, morphology, cross-source lag estimation, or other demanding uses. Therefore, temporal support is not always a single universal mask shared by every analysis.

Zero values, quiescent periods, and absence claims require careful interpretation. A valid zero-valued sample can indicate a measured zero or inactive state and still belong to evidential support, while a missing interval provides no direct evidence that the underlying phenomenon was zero or absent. Behavioral absence must never be inferred from unsupported time.

Temporal masks and support labels explicitly represent which time positions belong to declared evidence classes. These masks can be binary, categorical, probabilistic, or confidence-weighted depending on scientific needs. Masks should preserve their semantics, source, and resolution rather than functioning as unexplained preprocessing flags.

Continuous, Discontinuous, and Composite Support

Connected or continuous support is evidence available across one uninterrupted temporal interval under the declared validity criterion. This differs from a continuous-looking signal generated through interpolation or reconstruction, which can contain derived values over times that were not directly observed.

Fragmented or discontinuous support is a temporal set composed of multiple separated intervals. Fragmentation can arise from sensor dropout, artifact rejection, intermittent observation, event-triggered recording, nonwear, protocol design, privacy filtering, participant absence, or other causes. Fragmentation itself does not identify the cause or scientific severity of the gaps.

Cumulative duration for a support represented by K mutually nonoverlapping intervals is rendered as:

Here, S is the union of the nonoverlapping intervals [a_k, b_k), μ(S) is their total temporal measure, and overlapping intervals must first be unioned or otherwise accounted for to avoid double counting. This cumulative measure contrasts with the outer span.

Support components and gaps matter independently of total supported time. Two records with eight valid minutes can differ substantially if one has one continuous eight-minute interval and the other has dozens of short fragments.

Composite support is a scientifically defined temporal set built from several intervals, repeated occurrences, or conditional periods treated jointly without pretending they are temporally continuous. Composite support can represent recurring behavior, distributed observation opportunities, or repeated task phases, but the separation among components should remain visible when relevant.

Coverage, Occupancy, Exposure, and Effective Observation Time

Temporal coverage is the proportion or amount of a declared reference support for which the required evidence is available or valid. The denominator or reference interval must be stated because "90% coverage" is uninterpretable without knowing whether it refers to a session, day, event, participant exposure period, or another support.

Absolute supported duration differs from proportional coverage. Eight valid hours can represent high coverage of an eight-hour protocol or low coverage of a twenty-four-hour observation target. Coverage should therefore be interpreted jointly with its reference extent and gap distribution.

Temporal coverage relative to a declared reference support is rendered as:

Here, R is the declared reference support, S_valid is valid evidence support, and μ is temporal measure. Coverage summarizes the amount but not fragmentation, gap placement, event-specific loss, or quality variation within the supported portion.

Occupancy is the amount or proportion of reference time assigned to a state, activity, location, or other condition. Exposure time is the amount of time during which an entity was under the relevant opportunity, environment, treatment, sensor configuration, or observational condition. Occupancy and exposure can use related duration calculations but have different scientific semantics.

Effective observation time is the support over which the evidence required for a specific estimate is both available and sufficiently valid. Effective observation time can be smaller than recording time, wear time, protocol time, or nominal session duration and can differ across scientific quantities derived from the same record.

Overlap, Intersection, Union, and Joint Support

Intersection, union, difference, and complement of temporal supports are set operations that retain evidential semantics.

• Intersection identifies time shared by declared supports.
• Union identifies time present in at least one support.
• Difference identifies support remaining after exclusion of another support.
• Complement identifies unsupported or excluded time only relative to a declared reference universe.

In multichannel and multistream evidence, different channels or streams can have different valid intervals due to dropout, artifacts, occlusion, participant-specific conditions, source latency, device state, or quality requirements. A common recording span should not be treated as common valid support.

Joint temporal support required from all N sources is rendered as:

Here, S_i is the valid support of source i and S_joint is the support on which all required sources contribute simultaneously. Other analyses can legitimately require unions, quorum rules, or source-specific combinations, but the combination rule must be stated and should not be called joint support ambiguously.

Pairwise overlap and higher-order overlap differ: two streams can share substantial support even when three-way or all-source overlap is small, and aggregate coverage statistics can hide which source combinations are actually available. Joint analyses should use the support appropriate to the exact set of required inputs.

Boundary, Extent, and Support Uncertainty

Uncertainty in temporal boundaries and extent arises from limited sampling resolution, delayed detection, gradual transitions, timestamp uncertainty, uncertain artifact boundaries, ambiguous participant presence, or disagreement among sources or observers. A precise numeric boundary must not imply precision absent from the underlying evidence.

Support membership can also be uncertain. Some times may have graded confidence of validity rather than a binary valid-or-invalid status, especially near event boundaries, artifact transitions, occlusions, uncertain wear periods, or probabilistically reconstructed intervals. Confidence-weighted support should remain distinguishable from directly observed certainty.

Sensitivity to gap-bridging and boundary-tolerance rules affects temporal evidence. Merging support components separated by short gaps can increase apparent continuity and cumulative duration, while strict separation can fragment behaviorally coherent observation periods. Any bridge, tolerance, dilation, trimming, or erosion rule should be justified by the scientific semantics of the gap rather than convenience alone.

Support uncertainty propagates into derived quantities such as event counts, occupancy, durations, overlap, cross-source relations, averages, descriptors, and inference. Derived quantities should not silently convert uncertain support into exact exposure or exact observation time.

Support Under Transformation and Across Sources

Preprocessing transformations can alter temporal support. Filtering may require boundary trimming, artifact rejection can remove intervals, reconstruction can supply derived values over gaps, resampling can create new representation times, and windowing can create supports that overlap or extend beyond directly observed samples. The output support should preserve its relationship to the evidence that originally constrained it.

Representational support differs from observational support after interpolation, reconstruction, smoothing, or resampling. A value can numerically exist at a time coordinate even when no direct observation existed there. Dense output support should not be equated automatically with dense observational support.

Support correspondence across sources with different sampling rates, latencies, event structures, and uncertainties is complex. Two supports can refer to the same nominal interval while representing different observation opportunities or effective temporal resolution. Overlap in represented coordinates does not by itself establish synchronization or equivalent evidential precision.

temporal span, valid support, coverage, and joint support describe different properties of the same observation period

Evaluating and Documenting Temporal Support

Evaluating temporal support requires considering supported duration, span, coverage, number and distribution of gaps, connected-component lengths, event-specific loss, source-specific support, joint support, boundary uncertainty, and sensitivity to validity criteria. A high coverage percentage alone is insufficient when gaps systematically remove rare events, specific conditions, participant states, or critical transitions.

Temporal-support provenance includes all information needed to reproduce and interpret extent and availability. When relevant, this includes the reference interval, start and end semantics, mathematical or evidential meaning of support, validity criteria, support masks or labels, gap causes when known, direct-versus-derived status, component boundaries, gap-bridging rules, uncertainty or confidence, source-specific supports, combination rule for joint analyses, temporal measurement unit, timestamp reference, transformations that changed support, and software or implementation version.

Temporal Extent and Support matter in Behavioral Signal Processing because they determine how much evidence actually constrains durations, occupancy, event rates, cross-source relations, descriptors, and inference. Defensible temporal support makes explicit not only how long a record spans, but exactly where evidence exists, where it is valid, where it is derived, and which temporal regions truly support the scientific claim.