A range of gamblers, from low-frequency social gamblers through to problem gamblers in treatment, participated in focus groups discussing the characteristics of Electronic Gaming Machines (EGMs) that they found attractive. Analyses of the resulting transcripts resulted in two groups of EGM characteristics being identified as important, one group associated with winning and one with betting. Overall, free spin features were identified in all groups as the most attractive characteristic of EGMS. Beyond that it was smaller win-related characteristics, and low-denomination machines with multiple playable lines that were associated with increased duration and intensity of gambling behaviour. The important characteristics were consistent across different levels of gamblers, with the key behavioural difference being a self-reported 'expertise', and 'strategic' approach to gambling amongst higher-frequency gamblers and problem gamblers in treatment. The key characteristics all occur frequently and result in more wins and extended gambling sessions. The patterns identified resonated with established behavioural principles, and with models describing the development of problem gambling and addictions more generally.Keywords: gambling, problem gambling, electronic gaming machines, game characteristics, qualitative Electronic gaming machine characteristics: It's the little things that countThe core characteristic underlying Electronic Gaming Machines (EGMs) is a variation on a simple schedule of reinforcement, a variable-or random-ratio schedule (Skinner, 1953;Ferster & Skinner, 1957). In their simplest form, variable-ratio schedules specify that a reinforcer (in the case of EGMs, a win or related characteristic) is delivered after an average number of responses (plays). These schedules reliably create a very high and consistent rate of responding. This pattern of behaviour is very robust, and most importantly resistant to extinction (i.e., the behaviour persists for very long periods in the absence of any payout) or other independent variables/interventions. However, the schedules underlying EGMs are more complex than those used in basic research (e.g., Griffiths, 1993). EGMs have programmed variations to the payout magnitude, with large jackpots, much more frequent very-small wins, and characteristics such as freespins. Multiple playable lines add yet another dimension that can affect EGM outcomes and gambler behaviour (e.g., Harrigan, Dixon, & Brown, 2015;Jensen, Dixon, Harrigan, Sheepy, Fugelsang, & Jarick, 2013;Templeton, Dixon, Harrigan, & Fugelsang, 2015). These more sophisticated variants of simple schedules add additional levels of complexity to the behaviour of both EGMs and gamblers, likely meaning very high rates of gambler behaviour that is very resistant to intervention. In the gambling context, this means gamblers are likely to spend more time and money in any given gambling session than they initially intended. Delfabbro and Winefield (1999) used a similar methodology to observe a sample of regular a...