A Review of the Literature on the Cognitive Effects of Alcohol Hangover

Richard Stephens; Jonathan Ling; Thomas M. Heffernan; Nick Heather; Kate Jones


Alcohol Alcohol. 2008;43(2):163-170. 

In This Article

Laboratory Studies

We identified 27 English language peer-review studies that have investigated some aspect of psychological performance during alcohol hangover following controlled alcohol ingestion. Typically, in a between-subjects design, alcohol is given to one group and placebo is given to another group. After the passage of sufficient time for the acute intoxication effects to wear off (typically 11 hours), one or more cognitive/performance tests are applied. Decrements in the alcohol group relative to the placebo group are interpreted as hangover effects. A similar procedure may be applied in a related design where an analogous comparison is made but the same participants are tested twice, once following alcohol and once following placebo. However, close reading reveals that these relatively straightforward designs have not been well implemented in practice in this literature, with the majority of studies having basic methodological shortcomings.

Several early studies either present no data at all or do not present inferential statistical analyses of the data, and it is not possible to estimate statistics, such as effect sizes, based on the information presented (Carroll et al., 1964; Ekman et al., 1964; Ideström and Cadenius, 1968; Dowd et al., 1973). Future studies clearly were able to learn from these pioneering efforts. Nevertheless, without data or a rigorous data analysis it is not possible to draw meaningful conclusions from these studies. In an even earlier ground-breaking study Takala et al. (1958) compared hangover performance in medical students with non-hangover performance in a group of psychology and technical students. Unfortunately, in comparing intact groups of participants it is not possible to distinguish hangover effects from pre-existing participant differences.

Some other studies used a no-alcohol control condition rather than a placebo (Yesavage and Leirer, 1986; Taylor et al., 1996; Kruisselbrink et al., 2006). In the laboratory, placebo controlled trials have become the gold standard for research into effects of imbibed substances such as alcohol. Results from laboratory studies that do not implement a placebo control have little credibility. Without a placebo, participants would have known when they were consuming alcohol and so any effects shown could be expectancy effects, rather than genuine effects arising from the experimental treatment. In the case of alcohol, which is renowned amongst the general public for producing performance decrements, participants may unwittingly have put in less effort when completing the performance tests following alcohol ingestion. Morrow et al. (1990) did employ a placebo but they used a repeated measures design without properly controlling for condition order effects. Participants becoming fatigued during the experiment could explain their findings.

Another common problem is not verifying that BALs have returned to zero at the time the performance testing is carried out. As already mentioned, a defining characteristic of alcohol hangover effects is their presence at zero BAL. Ensuring BAL genuinely is zero (or so low as to be at the limits of detection) is necessary for research purposes in order to distinguish between hangover and acute alcohol intoxication. Lemon et al., (1993) purport to assess hangover effects without verifying that BAL is zero. In addition several supposed hangover studies report carrying out performance testing at elevated BALs. Such studies are likely to be picking up acute alcohol intoxication effects and cannot be interpreted as showing genuine hangover effects (Kelly et al., 1970; Seppela et al., 1976; Collins, 1980; Myrsten et al., 1980; Kim et al., 2003). Several other studies that are sometimes cited as assessing hangover effects were actually concerned with acute alcohol effects on the descending limb of the blood-alcohol curve. As these studies conducted testing at raised BALs, they do not elucidate alcohol hangover effects (Ekman et al., 1963; Jones and Vega, 1972; Peeke et al., 1980; McCaul et al., 1991; Millar et al., 1999).

Seven laboratory hangover studies are sufficiently rigorous to warrant serious attention (Collins et al., 1971; Collins and Chiles, 1980;. Roehrs et al., 1991; Chait and Perry, 1994; Streufert et al., 1995; Finnigan et al., 1998; Verster et al., 2003). One further laboratory study that we have criticized for not verifying zero BAL at testing will also be reviewed further. Lemon et al., (1993) found no effects and so this study is immune from the criticism that, in not verifying zero BAL, hangover results are contaminated by acute alcohol intoxication effects. Key aspects of the design, procedure and results of these eight studies are summarized in Table 1 and Table 2 .

Each study involved the administration of between 0.7 and 1.6 g/kg of alcohol (median 1 g/kg, roughly equivalent to 9 units, that is 3 pints of typical 5% ABV lager, or one bottle of 13% ABV wine), as pure ethanol or vodka mixed with orange juice, tonic, lime or lemonade. The placebo was usually orange juice or another mixer with a small quantity of alcohol to provide an appropriate olfactory sensation. The majority of studies employed all-male samples, the consumption-to-test interval ranged from 7.5-12 hours (median 10 hours), and appropriate steps were taken to control condition order effects. Each study verified that BAL was zero at testing (except Lemon et al., 1993, but see above). A range of cognitive and other performance tests were applied across the memory, attention, processing speed, executive function and psychomotor domains, although attention was tested in more studies than any other function.

Only two of the studies showed hangover effects. In a between-subjects design, Verster et al. (2003) showed poorer delayed recall of items from a 15-word list the morning after 1.4 g/kg of alcohol consumption (mean recall 9.4 items, SD 3.4) compared with placebo (mean recall 11.5 items, SD 3.5). Four of the eight rigorous laboratory studies assessed divided attention (or dual task performance which amounts to the same thing) but only Roehrs et al. (1991) showed a significant hangover decrement, with null effects in the three other studies (Lemon et al., 1993; Chait and Perry, 1994; Finnigan et al., 1998). This low 'hit rate' of effects across studies questions the reliability of the finding by Roehrs et al. and this reliability is further questioned by the small sample size those investigators employed. Therefore, there are only limited data to support the hypothesis that hangover produces decrements in divided attention. Only one of the eight rigorous laboratory studies showed a significant hangover decrement in delayed recall, although this was the sole study to have assessed long-term memory (Verster et al., 2003). At this juncture it is worth considering the extent to which long-term memory decrements have been shown in the less rigorous studies. Takala et al. (1958) noted a poorer rate of improvement due to practice (later trial performance was compared with earlier trial performance) on a visual search task after 1.3 g/kg beer, although, as already stated, this result was based on an intact groups comparison and it is possible that the technical students in the control group had generally superior visual search ability relative to the medical students in the hangover group. Ekman et al. (1964) assessed immediate and 7-minute delayed recognition of letter pairs but neither presented nor analysed the test scores. A small effect is claimed but it is not clear whether this is an acute intoxication effect. In a related design, Kim et al. (2003) showed long-term memory 'decrements' following 1.5 mg/kg alcohol consumption. However, again it is not clear whether this is an acute intoxication effect.

We find these mainly null findings surprising and strongly suspect their explanation lies with study insensitivity rather than a genuine absence of hangover effects. There are three reasons for this. First, five of the above studies employed crossover placebo designs, yet such designs have been argued to be unsuitable for ingestion studies. A combination of proprioceptive changes due to alcohol ingestion and inferential reasoning on the part of participants allows some participants to determine which is the placebo condition and diminishes study sensitivity (Finnigan and Hammersley, 1992). Second, all the above studies employ the pharmacological model of drug action, i.e. the alcohol is administered as a single large dose under as close to possible double-blinded conditions. However, this model may not be applicable to study the social phenomenon of drinking alcohol (Finnigan and Hammersley, 1992). For example, in real life people may control their drinking rate so that they can continue to function socially and will often take food with the alcohol. Third, it is possible that the alcohol doses employed in laboratory studies are insufficient for a hangover to occur; insufficient; due to research-ethics-imposed consumption limits. Nevertheless, five of the identified eight rigorous laboratory studies assessed subjective hangover symptoms and in all cases hangover symptoms were detected (Collins and Chiles, 1980; Roehrs et al., 1991; Streufert et al., 1995; Finnigan et al., 1998; Verster et al., 2003). It is possible that the alcohol dose required to produce cognitive hangover effects is larger than that for somatic hangover symptoms.

While some aspects of the design of laboratory studies can easily be addressed to improve sensitivity, research ethics limitations on alcohol dose and the limitations of the pharmacological model of drug action applied to alcohol ingestion studies have led some investigators to look beyond the laboratory.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: