The current research examined whether caffeine interacts to modulate ethanol-induced changes in anxiety, locomotion, and learning in the plus-maze discriminative avoidance task (PMDAT). Developed by Silva and colleagues (1997), the PMDAT uses an elevated plus-maze consisting of two opposing, open arms and two opposing, enclosed arms. During training, animals are free to explore all four arms but are conditioned to avoid one enclosed arm (the aversive arm) by the presentation of both light and white noise stimuli when they enter that arm. Time spent in the open arms, compared to time in all arms, is used as a measure of anxiety. Time in the aversive enclosed arm compared to time in the non-aversive enclosed arm is used as a measure of learning. Total entries into all arms are used as a measure of locomotion. On testing day, animals are again free to explore all four arms, with no presentation of the conditioning stimuli, and the same measures observed at training are recorded. Thus, the PMDAT allows measurement of learning, anxiety, and locomotion at both training and testing within each subject.

The effects of ethanol and caffeine on time in the aversive arm most likely represent changes in the processes underlying learning rather than associated changes in arm preference or anxiety. We previously demonstrated that animals presented with neither light or tone aversive stimulus at training spend a similar amount of time in the (to be) aversive and non-aversive arms at testing, suggesting that avoidance of the aversive arm in trained animals reflects a learned association between the aversive stimuli and the arm in which they were presented (Gulick & Gould, 2009b). In addition, of the doses of caffeine that increased anxiety, only the 40 mg/kg dose also impaired learning, suggesting that the learning deficits associated with caffeine were not due to changes in anxiety. Thus, the effects of caffeine and ethanol on learning may not be mediated by changes in anxiety or arm preference, but rather by changes in the substrates of learning. Furthermore, ethanol and caffeine may yield selective effects on learning. Both ethanol and caffeine produced deficits in learning to associate the aversive arm with the cues presented in it, but neither drug blocked habituation to the plus-maze, as demonstrated by a decrease in arm entries per minute of the maze at testing compared to training. Studies from another paradigm which models habituation learning, the novel object recognition task, support our finding that the doses of ethanol and caffeine used in the current study do not impair habituation. In mice, higher doses of ethanol than those used in the current study are required to impair novel object recognition (Ryabinin, Miller, & Durrant, 2002), and one study suggests that caffeine may produce beneficial or null effects on novel object recognition depending on the administration paradigm and time of testing (Costa et al., 2008).

In the current study, caffeine failed to reverse ethanol-induced learning deficits. On the surface, this finding contradicts work by Silva and Frussa-Filho (2000), who found that caffeine reversed memory deficits in mice when co-administered with a memory-impairing dose of the benzodiazepine chlordiazepoxide; they interpreted their findings as evidence that the antagonistic actions of the anxiolytic chlordiazepoxide and anxiogenic caffeine produced a normalized level of anxiety and arousal that allowed animals to learn to avoid the aversive arm. In the current study, however, the moderate doses of caffeine (10-20 mg/kg) increased anxiety without altering learning, and although caffeine and ethanol interacted to alter anxiety, this interaction was not associated with a rescue of learning, suggesting that the learning deficit does not depend on changes in anxiety. Thus, some mechanism other than neutralization of changes in anxiety may underlie the interactive effects of caffeine with ethanol and benzodiazepines on learning.

Both ethanol and chlordiazepoxide have amnesic and anxiolytic actions via GABA receptors (Grobin, Matthews, Devaud, & Morrow, 1998; Hevers & Luddens, 1998; Stackman & Walsh, 1992; White, Simson, & Best, 1997), but the differential effects of caffeine on the learning-impairing effects of each drug may be due to changes in other processes. Benzodiazepines primarily work in the central nervous system to alter the function of GABA receptors (Richter, 1981), whereas ethanol alters receptor function for a variety of neurotransmitter systems, including glutamate, GABA, and acetylcholine (Nevo & Hamon, 1995). Caffeine may not be able to antagonize all of the functional changes that ethanol produces at these diverse receptors. This is in contrast with the effects of nicotine on ethanol-induced learning deficits in the PMDAT and fear conditioning, as nicotine reversed learning deficits associated with ethanol administration (Gulick & Gould, 2008, 2009a). Thus, it appears that the reversal of ethanol-induced learning deficits is due to actions at specific receptors targeted by both ethanol and nicotine, rather than some general property of drugs classified as stimulants; in the current study, there was no interactive effect of caffeine and ethanol on learning that could explain the frequent co-use of these drugs.

Although caffeine did not block the learning deficits associated with ethanol treatment, ethanol did block the anxiogenic effect of caffeine. Caffeine also reverses ethanol-induced sedation and decreases self-reports of intoxication in humans (Drake et al., 2003; Liguori & Robinson, 2001; Marczinski & Fillmore, 2003, 2006). Thus, one reason for the popularity of drinks combining alcohol and caffeine may be the interactive effects of these drugs on sedation and anxiety.