Issuu on Google+

Earworms from three angles Victoria Williamson & Daniel M端llensiefen A British Academy funded project run by the Music, Mind and Brain Group at Goldsmiths in collaboration with BBC 6Music

Points of contact for our studies

BBC 6Music site (Short reports, emails and texts)

What is left unanswered‌ 1.

What triggers earworms in everyday life? Do they have a purpose?


Are some personalities more vulnerable than others?


What makes a tune sticky?

Project 1: Everyday triggers What triggers earworms? 

Method: Qualitative analysis (grounded theory) of earworm episodes

Result: Identification of high-risk situations

Do they have a use?

Williamson et al., (2012) Psychology of Music 

Earworm reports coded using grounded theory analysis techniques (2 independent raters)

6 Music corpus: 333 reports = 942 codes  Survey (.com) corpus: 271 reports = 657 codes 

Two models of codes show everyday earworm triggers and their relations

Emphasise importance of musical exposure but also memory function, and cognitive and affective state.

Some examples of memorable reports... 

Stress - My ear worm is ‘Nathan Jones' by Bananarama. I first caught it in 1989 during my GCSE chemistry exam and have been plagued by it in moments of extreme stress since, e.g. wedding, childbirth etc” (6Music Text).

Person Association- My earworm today is ‘This Charming Man' by The Smiths because every time I see David Cameron, that song just appears in my head, for some particular reason” (6Music Emails)

Musical media        

1. Live Music (e.g. concerts or gigs) 1.Video Media (e.g. TV, film, internet site) 3. Radio 4. Private Music (e.g. in the home or the car) 5. Contagion (e.g. another individual singing or humming) 6. Learning (e.g. practising for performance or a lesson) 7. Public Music (e.g. restaurant, shop or gym) 8. Ringtones


Musical exposure – ubiquity (Sacks, 2007;

Beaman & Williams,

2010; Liikkanen, 2012)

But also non musical association triggers in (involuntary) memory

Heightened emotional states (including Media): Levels of encoding = ‘resurfacing’ potential?

Project 2: Individual differences Are some people more vulnerable than others? 

Method: Statistical analysis of personality inventory (OCI-R) and factors of musical behaviour questionnaire (MuBQ) in relation to INMI factors (

Müllensiefen et al. (in review) 

Why are we interested in OC trait?

“people with obsessive compulsive disorder are more likely to report being troubled by earworms – in some cases medications for OCD can minimise the effects” (Levitin, 2006, p.151)

Let’s find out …


Individuals who measure highly on sub-clinical OC will experience more INMI that is more disturbing (Garcia-Soriano, Belloch, Morillo, & Clark, 2011)

People who are more ‘musical’ will experience more frequent earworms (INMI) that are longer and more troubling (Beaman & Williams, 2011; Liikkanen, 2012)


1536 participants (58.1% women).

M Age = 34.2, SD = 12.6, range: 12-75

Exploratory analysis (n=512): â—Ś Factor analysis of musical behaviour and INMI questionnaire

Confirmatory analysis (n=1024): â—Ś Structural equation modelling to test hypotheses between OC, musical behaviour, and INMI.

Testing Hypotheses 

Structural Equation Modelling: â—Ś Only some hypotheses confirmed â—Ś Good fit of final model:  adjusted goodness-of-fit = 0.929  RMSEA index = 0.06

Results: Only Singing is linked (positively) to INMI But: Singing makes INMI more pleasant

OC traits = INMI Frequency and Disturbance Mediated evaluative response between OC & INMI Length: High OC => INMI disturbing => longer INMIs Similar paradoxical relationships found in OCD (Wegner et al., 1987)

To follow up 

Should we be medicating earworms with OCD drugs?...

Should we prescribe singing to OCD patients? ‌

Posters on individual differences 

G. A. Floridou, V. J. Williamson, D. Müllensiefen “Contracting Earworms: The Roles of Personality and Musicality” (Friday 3.30pm)

M. Wammes, D. Müllensiefen,V.J. Williamson: “Schizotypal Influences on Musical Imagery Experience” (Wednesday 11am)

Project 3: Stickiness of tunes What is it that makes a tune sticky? 

Method: Computational analysis of tunes from frequently reported earworms

Tools: FANTASTIC software package i.abs.std =

Result: Classification model predicting stickiness

∑ (∆p i



− ∆p

N −1

) = 2.83

Step 1: Gathering earworms •

~2000 participants (.com survey)

1960 different earworm tunes (Artist, song title, exact part)

Top earworm list: 5.5% of songs identifiable and named at least 3 times

Method 1.

Control for popularity and recency and find ‘sticky tunes’: => tunes with a positive residual after poisson regression (using popularity data as predictors)

2. 3.

Find tunes most similar to INMI tunes (match by genre and chart success etc.) Use melodic features (Müllensiefen, 2009) of tunes to predict INMI vs non-INMI tunes (logistic regression)

Data Most frequent earworm tunes: artist lady gaga lady gaga journey katy perry queen

song bad romance alejandro don't stop believing california gurls bohemian rhapsody

incs 13 11

hi.entry 1 7

weeks 38 10 281 253 15 183

genre pop pop

11 10

6 1

47 6

477 43

149 1

rock pop







Similarly successful but never mentioned as earworms: artist gorillaz jessica simpson stereophonics nelly elvis presley

song feel good inc. these boots are made for walkin' handbags and gladrags my place way down

incs 0

hi.entry 2

weeks 39 1940 1667

genre pop







0 0 0

4 1 1

15 11 13

3164 2164 12054

3059 2087 11963

rock pop rock

Earworm classification model p (earworm = 1) =

1 1+ e

−(1.079+ 0.064 ⋅ d.median -0.723 ⋅ i.leaps)

= Longer durations and smaller intervals make tunes sticky (maybe because they are easier to sing?) BUT results only preliminary, because: • • • • 

Melody only one aspect of INMI Small sample (58 songs) Interactions of features Different types of earworms => different structural models?

Latest analysis on 214 tunes: Sebastian Finkel (Friday 3.30pm poster session)

FINAL conclusions 

Musical exposure important (Sacks, 2007; Bailes, 2012) that is recent and repeated (Beaman & Williams, 2010); but so is the activity of nonmusical, involuntary memories

State of mental arousal (wakefulness, excitement and stress) and ‘mind wandering’ – a possible function? (Leverhulme Grant)

Singing behaviour predicts features of INMI plus ease of singing may predict stickiness: activity of brain areas?

Melodic structure alone is a powerful predictor of inherent stickiness

Multi-method approach for generating future hypotheses

Special thanks to Sagar Jilka, Sebastian Finkel, Josh Fry, Alex Handler, Mandi Goldberg, Andre Lira & all at the BBC


QUESTIONS?? This project was kindly supported by: