The stunning process is an important component of slaughter with implications for animal welfare due to the potential distress and pain in the case of a sub-effective or lengthy stun. This study examined the factors correlated with variation in responses to carbon dioxide (CO2) stunning of pigs in five Australian commercial abattoirs. A total of 1 769 pigs (199–492 focal pigs per abattoir) were individually followed from lairage to post-stunning. A standardised observation protocol was used based on a literature review of the pre-slaughter factors that may influence the reaction to CO2 stunning, such as animal background, lairage conditions, handling, stunning system and conditions. Pigs lost posture 22.5 ± 0.2 s after commencement of descent of the gondola into the CO2 chamber. Latency to loss of posture was associated with farm of origin and time of day, which could be linked to various factors. Pigs that crawled or attempted to escape while in the gondola within the CO2 chamber took longer to lose posture. Crawl and escape attempts differed between abattoirs (0.6–46.2% of the pigs observed) as well as mounting other pigs (1.0–24.3%). Greater amounts of forceful contacts during handling in the race were related to more mounting in the gondola, but to less pigs crawling or attempting to escape. Mounting in the gondola was more frequent for pigs from lairage pens of mixed sexes, followed by pens of entire males and finally pens of females. Males were also twice as likely to show crawl and escape attempts than females. Gasping in the gondola was relatively frequent (63.1–81.8%) and was associated with higher activity in the lairage pen and higher skin injuries. Convulsions (60.1–69.6%) were generally observed after loss of posture. The type of CO2 system (group-wise vs single-file loading) had no significant effect on behaviour in the gondola. Nevertheless, pigs slaughtered in abattoirs with group-wise loading systems and automatic gates had lower cortisol concentrations post-stunning, which may be linked to minimal handling by stockpeople, other factors related to the systems, or differences in timing of when blood samples were taken. In conclusion, substantial variation in the reaction of pigs to CO2 stunning was observed between and within abattoirs using a uniform protocol for data collection. This variation in outcomes between abattoirs and stunning systems and the relationships between handling and behavioural outcomes indicates that improvements can be made to reduce aversive responses to CO2 stunning. In particular, avoiding mixing pigs of different sexes in lairage and aversive handling in the race may reduce aversive response to CO2 stunning.