Assessing pain in non-communicative patients remains challenging in anaesthesia and intensive care. When self-report is unavailable, clinicians infer nociception from behaviour and physiology. Behavioural scales such as the Behavioral Pain Scale and the Critical-Care Pain Observation Tool are simple and reproducible, supporting consistent practice; however, performance declines with deep sedation, neuromuscular blockade, or severe neurological injury. Where behavioural cues are absent or unreliable, physiological and neurophysiological signals provide partial information. Autonomic indicators, including heart rate variability, the Surgical Pleth Index, pupillometry, and skin conductance, capture sympathetic responses to noxious stimuli rather than perceived pain and are sensitive to drugs, haemodynamic instability, shivering, and agitation. Electroencephalography and functional near-infrared spectroscopy identify cortical responses to nociceptive input, yet clinically useful thresholds remain context dependent, and most applications are research-based. Emerging machine-learning systems that integrate behaviour and physiology show promise, but models validated in the operating room are not automatically applicable in the intensive care unit and require new external validation with potential recalibration. Evidence is generally stronger intraoperatively than in intensive care, and paediatric data are limited. No instrument directly measures subjective pain when self-report is absent. Available tools index nociception through behavioural and physiological correlates and must be interpreted within the clinical context.