A brain basis for musical hallucinations



Hello Dear Reader,

I hope your January is going well. We are experiencing an unusually mild winter here in Switzerland so far. Very little snow has reached Luzern, which stands 400m above sea level. But at least I can now see snow on the mountains (including the stunning Mount Pilatus) from my office window, and very lovely it looks too.

Today I have been reading a new article (in press) in Cortex which claims to have identified a brain basis or at least a brain based explanation for musical hallucinations (MH). My interest was peaked – perhaps this might give a first clue about a brain basis for earworms?

Earworms (tunes that get stuck in your head) and MHs (complex musical perceptions with no external source) are not the same thing though they have a number of common features, both being musical and related to mental imagery.

The main difference as far as I am aware is one of conscious inference regarding the likely source of the sound: One is clearly recognizable as a memory (earworm) whereas one could easily be mistaken for the experience of real music listening (an hallucination).

A few weeks ago I write a blog about the difference between MHs and tinnitus. This blog was based on a new paper (Vanneste et al., 2013) that compared the resting brain state activity (using EEG) of people who experienced regular tinnitus (a sensation of ringing in the ears) or MH to that of spontaneous activity.

Neuron_in_tissue_cultureThe result of this paper was a theory that abnormal firing of neurons in some bandwidths (alpha, gamma-theta) in the lower centres of the brain was associated with tinnitus.

By contrast, MHs were associated with abnormal neuronal firings in the higher centres of the brain, those associated with memory and language processing.


This idea of ‘hierarchical levels’ of abnormal neuronal patterns in the musical brain pathway was a nice intuitively sensible conclusion. It was also interesting to see some brain basis for these conditions as opposed to the more common idea of blaming everything on the inner ear.

The new paper by Sukhbinder Kumar and colleagues takes a case study approach instead. The team looked at the experiences and brain activity of one 62 year old keen amateur musician who had absolute pitch. This lady had experienced MH 15 months after the onset of acute hearing loss, approximately a year and a half before she took part in the study.

When she first started experiencing MH the lady assumed that the sounds came from an external source – by my definition that qualifies them as an hallucination rather than earworms, even though she now knows that these musical sounds are not real (they appear to always be the same few bars of recognizable melodies).

NIMH_MEGThe authors used a clever technique to assess her MHs as they happened in an MEG scanner.

Her MHs could be suppressed by playing short excerpts of music by Bach (like a mask) so the authors compared her brain activity across time in a single scanning session as she moved in and out of a state where she experienced MHs.


A beamforming analysis was then performed on the brain data to isolate patterns in oscillatory activity during MH across five frequency bands: 1-4Hz (delta), 5-14Hz (theta/alpha), 14-30Hz (beta), 30-60Hz (gamma) and 70-140Hz (high gamma)

Results: Significant power changes during high periods of MH were observed in the theta/alpha, beta and gamma bands but not in delta or high gamma.

None of these changes were localised to the right hemisphere and all changes referred to increases (rather than decreases) in oscillatory power.

Area 1 of activity was the orbitofrontal cortex (theta/alpha activity). This area has been associated with responses to unpleasant music and imagery. It is perhaps not surprising to see this activity therefore, since the lady was often bothered by her MH. 

Area 2 of activity was the motor cortex (beta activity) which the authors link to the well established activation of motor areas in response to musical imagery, particularly in musicians.

Area 1 of activity was the secondary auditory cortex (aSTG – gamma activity). This area is involved in melody perception.

Traffic_light_greenHow do these results compare to the previous paper? This paper and that of Vanneste et al. (2013) show an increase in gamma in relatively ‘lower’ brain areas (secondary sensory cortices – green) and an increase in alpha and beta power in ‘higher’ brain centres during MH (motor cortex – orange ), which fits with a hierarchical theory of MH.  The present paper takes this hierarchy idea to propose a new model for the brain basis of MH. Their theory presupposes only the presence of hearing loss.

Crucial to this model is the existence of a top down predictor system that we build through a lifetime of musical listening. This system of ‘priors’ sends predictions back through the musical perception pathway in response to sensory stimulation in the level below. Ascending (upward going) information about the music being heard then consists only of any information on prediction error so that higher level expectations can be modified.

It is a Bayesian optimised prediction system for music.

Top down musical prediction from priors – bottom up prediction errors

When someone loses their hearing the brain responds by lowering the sensory precision of the lower sensory centres of the brain (in this case the auditory cortex). That leaves the next level of the hierarchy increasingly sending through prediction error messages to the higher systems, unchecked. And the higher centres reciprocate with backward prediction messages, creating a loop that leaves out the lower level.

In theory this leaves a cycle of communication between the brain areas that drive basic melody perception and imagery (and memory) without the strong input of the lower sensory systems to feedback a prediction error based on what is actually being heard. This leads to a MH.

musicWhy music? As compared to speech or images, music is more predictable and repetitive. It is also rapid and temporal meaning there is more pressure to alleviate strain on the sensory systems, to support them with predictions from higher centres. These combined characteristics mean music is more subject to the activity of priors; music’s own recursive cyclic characteristic is what lends it to be the basis for hallucinations

…and perhaps is why earworms tend to be musical too.

What does this tell us about earworms, and what is missing? If we accept the hierarchical prediction model of the musical pathway then we might presume that any spontaneous activation of the system (for example, in memory) might trigger reciprocal communication within part of this loop. A tune therefore might get stuck in our mind when we have an earworm in the same manner as a MH.

What we don’t know is: a) how this activity is triggered in the first place in either MH or earworms; b) why the loop goes on and on; and c) what part of the model makes the difference between a MH and an earworm

It might explain however, why listening to music often helps people deal with earworms (as you can read about in my upcoming PLOS ONE paper on earworm cures!), as the ascending musical input of prediction errors in this case would break the cycle of internal musical imagery.  

What does it NOT tell us about MH? This paper provides an explanation for MH as a result of hearing loss – it does not provide an explanation for MH that are experienced by people with a psychosis or as a result of a focal brain lesion. Why do people experience MH when there is, on the face of it, nothing wrong with the lower hierarchy (the sensory systems)?


This is a really nice paper that stimulated great conversation in my office about the nature of perception and mental imagery, both in the musical world and beyond!

I think these Bayesian network ideas are here to stay so I advise you to have a go at reading this paper and to think about how the brain’s way of dealing with rapid, temporal sensory input (i.e. relying on top-down predictions) might influence what we experience as part of consciousness.

Article: Kumar, S., et al., A brain basis for musical hallucinations, Cortex (2014)

7 thoughts on “A brain basis for musical hallucinations”

Comments are closed.