The current study used the stimulant model, caffeine, to assess the effects of stimulant deprivation on impulsivity in humans. Previous research has found substance abusers express greater preference for immediate outcomes and this effect is increased with deprivation from the substance. A delay discounting method was used to examine the effects of stimulant deprivation on caffeine-dependents and non-dependents. Experiment 1 looked at the effects of 12 hour's abstinence from caffeine on High and Low caffeine consumers' decision making, using a pen-and-paper delay discounting method. Experiment 1 produced area under the curve values which showed the largest effect in the Low caffeine consumers, Low caffeine consumers increased discounting when under the influence of caffeine. Experiment 2 used a well supported computer-based, hypothetical monetary, delay discounting task in hope of generating a magnitude effect. Experiment 2 was used to determine the task mode in Experiment 1 was not responsible for the unexpected pattern of results. A magnitude effect was found between the $1000 and $10,000 tasks in the delay discounting procedure. Experiment 3 used the established computer-based, hypothetical monetary amounts, delay discounting task from Experiment 2 to assess the effects of 24 hours abstinence from caffeine in High and Low caffeine consumers. The k values revealed a significant increase in discounting when participants were caffeine deprived for all groups except the High caffeine group in the $10,000 condition. These mixed results were discussed in terms of weaknesses revealed about the common mathematical measures used to analyse delay discounting data. Most importantly, the practice regularly employed by delay discounting researchers to exclude data which does not conform to discounting functions. The exponential, hyperbolic, extended hyperbolic models and the area under the curve measure were compared. The area under the curve was found to be the only measure to not exclude data, and sub-sequentially was preferred. Further research is required to establish if the area under the curve behaved in similar ways to other measures of discounting data as correlations in this study are not clear.