Abstract: Duration perception is a fundamental cognitive ability, yet its mechanisms remain contested. A central question is whether durations of different lengths are processed by a single timing system or by qualitatively distinct mechanisms. Drawing on a series of studies conducted by our research group over the past two decades, together with key findings from independent laboratories, this review integrates evidence across four themes—segmentation, attention and working memory, emotional modulation, and special populations—and proposes three theoretical hypotheses that move the field from description toward mechanistic explanation.
The segmentation issue has a long history. Behavioral, electrophysiological, and neurostimulation evidence converges on the view that duration processing is not homogeneous. Rather than sharp boundaries at 300-500 ms and 2-3 s, the evidence supports functional transition zones. These zones delineate three processing levels: a sensory-automatic level below roughly 300-500 ms, a perceptual level between 300-500 ms and 2-3 s that is sensitive to attentional resources, and a memory-dependent level beyond 2-3 s that increasingly relies on working memory. We emphasize that these zones are probabilistic and task-dependent, not fixed critical points. The proposed representational forms—iconic, pre-semantic, and semantic codes—are offered as theoretical propositions that require direct empirical test.
Closely related to segmentation is the distinction between attention and working memory. We propose the attention-working memory duration-dependency hypothesis. Attention functions as a general resource that modulates pulse accumulation across a wide range of durations, from milliseconds to seconds. Working memory capacity, by contrast, is selectively recruited for durations beyond 2-3 s, when temporal representations require active maintenance and reconstruction. This dissociation helps explain why some studies report working memory effects on timing while others do not. It also yields testable predictions: manipulations of attentional load should affect both short and long durations, whereas individual differences in working memory capacity should primarily affect long durations. We further note boundary conditions, as task paradigms may shift the transition zone.
Emotion adds another layer of complexity. We propose the segmentation-integrated hypothesis of emotional effects on duration perception. Existing accounts often assume that emotion either speeds up or slows down an internal clock or consumes attentional resources. We argue that the direction and magnitude of emotional effects depend on the target duration range being judged. In the short-duration range, attentional capture by salient emotional stimuli dominates, often lengthening perceived duration. In the medium range, arousal-driven clock acceleration becomes primary. In the long-duration range, arousal returns to baseline, and cognitive factors such as memory and decision-making dominate. This framework integrates disembodied (arousal/attention) and embodied (bodily state simulation) pathways and explains why the same emotion can produce opposite effects across studies. It also generates testable predictions: early attentional capture should affect all salient stimuli regardless of valence; high-arousal emotions should produce the strongest distortions in the medium window; and emotional effects should diminish for very long durations.
Clinical and special populations provide a complementary perspective. We propose a dual-pathway model of abnormalities in duration perception. The cognitive resource deficit pathway centers on working memory impairment and selectively affects long-duration processing beyond 2-3 s, as observed in depression and alcohol use disorder. The perceptual encoding deficit pathway centers on reduced temporal resolution and selectively affects short-duration processing below 1 s, as observed in dyslexia and deafness, sometimes with sensory compensation at longer durations. State factors such as anxiety and pain constitute a parallel modulatory mechanism that dynamically shifts the relative weight of arousal and attention without altering stable cognitive architecture. This model offers a unified account of seemingly disparate timing abnormalities and provides converging evidence for the segmentation framework.
The theoretical contributions are threefold. We advance segmentation theory from a descriptive taxonomy to a mechanistic framework with explicit resource-dependency predictions. We dissociate attention from working memory in timing, moving beyond the generic notion of cognitive resources. We integrate emotional effects and clinical abnormalities into the same segmentation framework, demonstrating cross-domain consistency. These hypotheses are falsifiable and point to specific future studies. Clinically, the model offers a theoretical basis for understanding timing disturbances in clinical populations, although direct application to screening or treatment evaluation requires further validation of reliability and incremental validity.
In sum, this work synthesizes two decades of research to propose an integrated, testable account of duration perception. It highlights that time processing is not a single uniform system but a dynamic interplay of sensory, attentional, working memory, and emotional mechanisms that unfold differentially across time scales.

Key words: duration perception, segmentation, attention, working memory, emotion