TL;DR summary: This claim seems overstated in its current form. In short, the subjects in the study cited were primed to hallucinate; they were played a song, then told that the song might be embedded in white noise, and then were played white noise for three minutes and told to push a button whenever they heard fragments of the song.
Subjects who self-identified as having high stress and as intaking high amounts of caffeine were found to be more prone to reporting hearing fragments of the song.
Thus, yes, those in this group did "hallucinate," but on a practical level, the amounts of publicity this has received (just note how the floodgates open HERE) seem to overstate the conclusion and blow things out of proportion.
I personally question whether a primed response even counts as "hallucination" in the same way that some are implying it, such as being connected with psychosis.
The claim in whole, "that caffeine causes auditory hallucinations," seems to be false.
Initial results didn't turn up much; I searched "caffiene hallucination" on google scholar, which revealed THIS list. There was one result of interest, a blind study HERE, that found that caffeine given intravenously led to olfactory (smell) hallucinations.
In any case, the actual study, published in the journal, Personality and Individual Differences, is found HERE. Note the following characteristics:
- Sample size: 92
- Divided into four groups: high stress/low caffeine, high stress/high caffeine, low stress/low caffeine, low stress/high caffeine
- All subjects were played Bing Crosby's White Christmas and then told the following:
The White Christmas song you just heard might be embedded in the white noise, below the auditory threshold. If you think or believe that you hear the song, or a fragment of the song clearly, please indicate so by pressing the hand counter. Of course, you may press the counter several times if you think that you heard several fragments of the song (p. 140).
- The high stress/high caffeine group pressed the button more than the other groups
Now, there's some interesting features in the discussion section:
- They mention a similar study from 2009 by Jones and Fernyhough in the same journal, found HERE which asked 219 participants to report their caffeine intake, and then measured the following items via questionnaires:
Hallucination-proneness: This was assessed using the modified Launay-Slade Hallucination Scale (LSHS-M: Laroi & van der Linden, 2005), a 16-item instrument designed to measure predisposition to hallucination-like experiences. Each item is scored on a five-point Likert’s scale ranging from ‘‘Certainly applies to me” (4) to ‘‘Certainly does not apply to me” (0).
Proneness to persecutory delusion-like beliefs: The Persecutory Ideation Questionnaire (PIQ; McKay, Langdon, & Coltheart, 2006) is a 10-item questionnaire designed to measure persecutory ideation in both clinical and non-clinical samples. Items are rated on a five-point Likert’s scale ranging from ‘‘Very True” (4) to ‘‘Very Untrue” (0).
Stress: This was determined by the Perceived Stress Questionnaire (Levenstein et al., 1993), a 30-item instrument which assesses the levels of stress an individual perceives themselves to have been under during the last year. Examples of items include ‘‘You feel tense” and ‘‘You have many worries”. Items are scored on a four-point Likert’s scale ranging from ‘‘Almost Never” (1) to ‘‘Usually” (4). A total score is derived using the formula ((raw score – 30)/90).
They found that caffeine intake was correlated with both stress levels and hallucination proneness.
- Crowe says the following about the above Jones study, however:
Jones and Fernyhough (2009) found that caffeine consumption predicted hallucination proneness. Yet, contrary to our hypothesis, we found no correlation between hallucination proneness and the actual occurrence of false alarms as measured by our procedure.
Interesting. So, if caffeine is correlated to questionnaire determined "hallucination proneness," (via Jones, 2009), but "hallucination proneness" isn't correlated to actual false alarms (hallucinations) (Crowe, 2011)... then hallucination proneness may not be a very reliable quantity of measure with respect to anything usefully predictive.
- Finally, this statement was perhaps the most informative:
Other potential influences that may explain the occurrence of hallucinatory experiences (including creative experiences, imagery ability and social desirability seeking) were also ruled out in the current investigation, which is a strength of the study.
There are however a number of limitations that can be identified with the study. Concerning the effects of caffeine, it is important to note that we did not take into consideration all possibly influential factors that could affect an individual’s ambient level of caffeine. There are a number of factors that are known to mediate the effects of caffeine. This includes factors such as age, gender, body mass, habitual user status, caffeine source, liver function, cigarette smoking, use of oral contraceptives and pregnancy (James, 1997).
That's a lot of other factors to control for in a future study.
- The conclusion of the study was this:
The results from this study demonstrate that increased caffeine consumption, combined with high levels of stress, has the ability to increase the experience of psychotic symptoms, specifically auditory hallucinations.
While that conclusion certainly seems to provide an affirmative source to support the claim, I'd be hesitant to accept it in full, as I stated at the beginning of this answer:
- There are a lot of other variables to control for in a further experiment
- I consider priming to be a huuuge design flaw in the experiment. If one were not going to see something, but saw it due to priming, is that really in the hallucination category? For example, consider this fantastic example from Michael Shermer (LINK). If one hears the words in Stairway to Heaven in reverse after being primed, is this hallucination?
So, high stress and high caffeine have been shown to lead to a higher incidence of auditory hallucination after being directly primed for the "hallucination" -- hearing fragments of a song "embedded" in what was really white noise.
As far as those concerned with such results, my suggestion would be to wait until replication that controls for other factors as well as another test method which does not prime. In other words, I assume that anyone wishing to translate this study into action would do so on the basis of safety or health concerns. I don't personally consider hearing song fragments after being primed to hear them in that category. Thus, if another study comes along and finds that subjects hear things on their own, without priming, and that such occurrences could be unsafe or unhealthy, then I'd act on such studies.