The critical health effects of formaldehyde exposure include sensory irritation and the potential to induce tumours in the upper respiratory tract. In literature, a concentration as low as 0.24 ppm has been reported to be irritating to the respiratory tract in humans. Nasal tumour-inducing levels in experimental animals seem to be 1-2 orders of magnitude larger. In this paper, the subjectively measured sensory irritation threshold levels in humans are discussed in line with findings obtained in animal experiments. In addition, a Benchmark dose (BMD) analysis of sensory irritation was used to estimate response incidences at different formaldehyde concentrations. Data on respiratory irritation and carcinogenicity of formaldehyde were retrieved from public literature and discussed. BMD analysis was carried out on human volunteer studies using the US-EPA BMD software. Subjective measures of irritation were the major data found in humans to examine sensory (eye and nasal) irritation; only one study reported objectively measured eye irritation. On a normalized scale, mild/slight eye irritation was observed at levels 1 ppm, and mild/slight respiratory tract irritation at levels 2 ppm. With the BMD software, it was estimated that at a level of 1 ppm, only 9.5% of healthy volunteers experience 'moderate' (i.e., annoying) eye irritation (95% upper confidence limit). An important factor modulating the reported levels of irritation and health symptoms most probably includes the perception of odour intensity. In several studies, the 0-ppm control condition was missing. From the results of the long-term inhalation toxicity studies in experimental animals, a level of 1 ppm formaldehyde has been considered a NOAEL for nasal injury. Sensory irritation is first observed at levels of 1 ppm and higher. From both human and animal studies, it was concluded that at airborne levels for which the prevalence of sensory irritation is minimal both in incidence and degree (i.e., <1 ppm), risks of respiratory tract cancer are considered to be negligibly low.