Temporal Integrity and Continuity

Temporal Integrity and Continuity refers to the quality of preserved temporal structure in recorded behavioral, physiological, neurophysiological, digital, or contextual evidence. Temporal integrity concerns whether ordering, timing, spacing, duration, timestamp meaning, and temporal relationships remain scientifically interpretable. Continuity concerns whether evidence progresses across the intended temporal support without unexplained breaks, substitutions, duplications, or discontinuities. Temporal integrity is not equivalent to synchronization, time referencing, sampling rate, completeness, or the mere presence of timestamps.

Meaning of Temporal Integrity and Continuity

Temporal integrity is the preservation of the temporal relationships required to interpret the evidence: which observation came first, when it occurred according to the declared time semantics, how long it lasted, what interval separated observations, and whether temporal transitions remained valid. A temporally intact record need not be perfectly regular or continuously sampled if irregularity or intermittency is part of the intended observation scheme.

Temporal continuity is the preservation of an interpretable progression of evidence through time over the support where continuity is expected. Continuity can be broken by missing intervals, source interruption, timestamp discontinuity, sequence reset, duplicated or substituted observations, or uncertainty about how two retained portions connect. A planned observation break does not constitute a continuity failure when its boundaries and semantics are known.

Expected Temporal Structure

Expected temporal structure refers to the ordering, timing rules, sampling behavior, event logic, state progression, continuity requirements, or known observation schedule that should govern a valid record. Temporal-quality assessment requires a declared expectation because regular, irregular, event-driven, burst-based, intermittent, and state-based records can all be valid under different acquisition semantics.

Nominal versus realized timing distinguishes between a system’s configured temporal plan—such as nominal sampling interval, frame rate, polling period, event schedule, or logging cadence—and the actual times at which observations occur. Scientific interpretation must differentiate the intended temporal structure from the timing actually retained in the evidence.

Temporal support is the interval, event sequence, state interval, recording span, or other temporal scope over which an integrity or continuity statement applies. A record can be temporally intact over one support and compromised over another, so global labels should not erase localized timing failures.

Order, Monotonicity, and Sequence Integrity

Temporal order is the preservation of the correct before-and-after relationship among observations, events, states, or frames. Reordered packets, delayed writes, sorting errors, buffering, asynchronous logging, clock resets, or incorrect merge operations can make the retained order differ from the physical or operational order.

Timestamp monotonicity requires nondecreasing or strictly increasing temporal coordinates when the declared clock semantics demand it. Legitimate equal timestamps can occur due to finite timestamp resolution or simultaneous events, but these differ from duplicated records or invalid backward time jumps. Monotonic timestamps alone do not prove correct absolute timing.

Sequence identifiers, frame numbers, sample counters, event counters, or other ordinal evidence help check order and detect gaps, repeats, or resets. However, an intact numeric sequence does not prove valid physical timing, and irregular identifiers do not necessarily imply lost temporal evidence if their semantics differ.

Reordered or out-of-order observations can occur even when all expected values are present. Such reordering can change event precedence, derivative estimates, transition interpretation, temporal descriptors, and apparent causal direction without missing samples, compromising temporal integrity.

Continuity, Gaps, and Discontinuities

Recording gaps are intervals in which expected evidence is unavailable or cannot be connected confidently across time. True gaps must be distinguished from genuine periods of behavioral inactivity, constant physiological states, planned acquisition pauses, or sources legitimately producing no events.

Discontinuity is an abrupt break in the expected temporal relation between adjacent retained observations or recording segments. Causes include restart, reconnection, clock reset, source reassignment, file rotation, stream interruption, skipped data, or configuration change. Discontinuities can exist even when two records are stored contiguously.

Continuity differs from completeness. A deliberately intermittent observation schedule can be incomplete relative to continuous time yet temporally intact because all planned observation windows and boundaries are known. Conversely, a record may contain the expected number of observations but lack temporal integrity if values are duplicated, reordered, mislabeled in time, or associated with the wrong interval.

Uncertain gap boundaries arise when the exact time at which valid observation ceased or resumed differs from the first missing or returning timestamp, due to buffering, delayed writes, stale values, or detection latency. Such boundary uncertainty should be preserved rather than assigning falsely precise continuity limits.

Interval Regularity, Jitter, and Timing Variability

Inter-observation interval integrity evaluates whether realized time differences among observations closely match the temporal behavior expected from the acquisition scheme for the intended use. Nominally periodic streams may exhibit small interval variability, occasional long intervals, compressed intervals, or abrupt timing changes.

Jitter is short-timescale variability in timing around an expected temporal relation. Usage varies across clocking, communication, sensing, and signal-processing contexts. Jitter is distinct from long-term drift, missing samples, variable biological latency, or intentionally irregular sampling.

Irregular sampling can be a legitimate acquisition pattern when observations are intentionally obtained at unequal intervals and actual times are preserved. Irregularity alone does not constitute temporal failure; failure occurs when realized timing is inconsistent with declared semantics or becomes unknown.

Rate changes and timing regime changes occur legitimately when a source modifies sampling or reporting cadence, enters low-power mode, switches from periodic to event-driven reporting, or alters acquisition configuration. Temporal integrity requires these changes be identifiable and their timing known sufficiently for interpretation rather than assuming a single rate for the entire record.

Timestamp and Clock Discontinuities

Timestamp jumps, resets, wraparound, repeated timestamps, epoch changes, clock-source changes, and nonmonotonic time present temporal-integrity hazards. A numerically valid timestamp can still be scientifically incorrect if its clock, epoch, unit, or assignment semantics change without representation.

Clock drift and rate error degrade temporal integrity only when they alter retained temporal relations beyond the tolerance required for the evidence. Clock progression error must be distinguished from genuine changes in event timing or participant behavior.

Clock-source switching and temporal-domain transitions occur when a device changes time source, reboots into a different epoch, resynchronizes, or alters timestamp generation behavior while recording continues. These can create jumps, duplicated time regions, compressed intervals, or apparently missing time unless their semantics are preserved.

Timestamp location and delayed assignment refer to the stage at which timestamps are assigned—sensing, device buffering, packet creation, receipt, persistence, or other. Changes in buffering or assignment location can alter temporal meaning even if timestamp format remains unchanged.

Duplication, Substitution, and Stale Evidence

Duplicate observations are repeated retention of the same underlying observation or event when evidence is intended to represent distinct temporal occurrences. True repeated behavior or identical consecutive values differ from duplicated records; repeated values alone do not prove duplication.

Repeated-last-value, stale-value, or hold behavior occurs when systems preserve the most recent valid value when no new observation arrives, creating an apparently continuous record despite missing new evidence. Held values should not be interpreted as direct evidence that the phenomenon remained unchanged unless justified.

Substitution and synthetic continuity arise from interpolation, frame duplication, resampling, buffering recovery, packet concealment, model-based completion, or software repair that produce temporally continuous-looking output after source evidence was missing or irregular. Such operations can be useful but should remain distinguishable from directly observed temporal continuity.

Event, State, and Duration Integrity

Event-time integrity preserves event occurrence time, order, count, onset, offset, and temporal relation to surrounding evidence. Missed, duplicated, merged, split, delayed, prematurely timestamped, or reordered events can compromise event-based evidence even when continuous signals appear intact.

State-transition integrity preserves state identity, transition order, transition timing, and dwell duration when records represent discrete or categorical states. Missing transitions, duplicated state updates, stale states, or uncertain transition boundaries can alter inferred durations and state sequences.

Duration integrity depends on valid onset, offset, and time semantics of events, episodes, states, or recording intervals. Duration can be biased by missing boundaries, timestamp jumps, delayed state updates, truncation, or use of nominal rather than realized timing.

Assessment and Temporal-Quality Indicators

Temporal integrity assessment uses evidence such as timestamp progression, realized inter-observation intervals, sequence counters, frame or packet identifiers, expected event schedules, clock-status records, restart indicators, buffer or transport logs, source-state information, duplicate detection, gap detection, and known reference events. No single indicator is universally sufficient.

Temporal-quality metrics include valid temporal coverage, gap duration and frequency, duplicate fraction, out-of-order count, interval-deviation statistics, timestamp-jump counts, continuity intervals, and event-timing uncertainty. Each metric requires definition of expected timing model, denominator, support, and tolerance.

Purpose-dependent temporal tolerance means timing deviations negligible for some uses (e.g., daily activity totals) may be unacceptable for others (e.g., reaction-time measurement, waveform morphology, event ordering, phase relationships, cross-signal delay estimation). Temporal integrity judgment must align with the temporal precision required by the scientific claim.

Uncertainty in temporal-quality assessment arises from unknown timestamp location, incomplete logging, uncertain gap boundaries, ambiguous duplicate detection, unrecorded clock changes, and imperfect reference events. Quality assessment should preserve bounded or unresolved uncertainty rather than impose false precision.

Scientific Consequences and Temporal Provenance

Failures in temporal integrity can alter Behavioral Signal Processing by changing event order, apparent duration, rates, transition timing, temporal descriptors, spectral estimates, temporal dependencies, cross-signal relationships, behavioral sequence interpretation, and model inputs. A record can remain numerically complete yet become scientifically misleading if its temporal structure is incorrect.

Temporal provenance is the information needed to reconstruct how time was represented and where continuity could have failed. This includes preserving the time base and clock source, timestamp semantics and assignment location, expected and realized cadence, sequence identifiers, gap and discontinuity intervals, resets, wraparound, clock-source changes, buffering or transport behavior, restart events, rate changes, duplicate or stale-value behavior, synthetic completion, uncertainty, and the validity interval of each temporal regime. Temporal integrity is an evidential property of the retained timeline—not a cosmetic property of evenly spaced data. A defensible temporal claim must establish what timing behavior was expected, what was actually retained, where continuity or ordering changed, whether apparent gaps or repeated values reflect the phenomenon or the recording process, and what uncertainty remains. Regular-looking output must not be confused with directly observed temporal continuity.