Fiction feelings in Harry Potter

Cognitive neuroscience and neuropsychology
Fiction feelings in Harry Potter: haemodynamic response in
the mid-cingulate cortex correlates with immersive reading
Chun-Ting Hsua,b, Markus Conrada,b,d and Arthur M. Jacobsa,b,c
Immersion in reading, described as a feeling of ‘getting lost
in a book’, is a ubiquitous phenomenon widely appreciated
by readers. However, it has been largely ignored in cognitive
neuroscience. According to the fiction feeling hypothesis,
narratives with emotional contents invite readers more
to be empathic with the protagonists and thus engage the
affective empathy network of the brain, the anterior insula
and mid-cingulate cortex, than do stories with neutral
contents. To test the hypothesis, we presented participants
with text passages from the Harry Potter series in
a functional MRI experiment and collected post-hoc
immersion ratings, comparing the neural correlates
of passage mean immersion ratings when reading
fear-inducing versus neutral contents. Results for the
conjunction contrast of baseline brain activity of reading
irrespective of emotional content against baseline were in
line with previous studies on text comprehension. In line
with the fiction feeling hypothesis, immersion ratings were
significantly higher for fear-inducing than for neutral
passages, and activity in the mid-cingulate cortex correlated
more strongly with immersion ratings of fear-inducing than
of neutral passages. Descriptions of protagonists’ pain or
personal distress featured in the fear-inducing passages
The phenomenon called immersion has been described
as a feeling of ‘getting lost in a book’ [1]. It has been
studied at subjective-experiential and behavioural levels
(e.g. Green [2]), but largely ignored in cognitive neuroscience, and its neuronal correlates during reading have
been so far unexplored [3]. In a recent neurocognitive
model of literary reading, neural, affective and cognitive
mechanisms likely to be involved with immersive processes have been proposed to involve the activation of
familiar situation models, sensory-motor grounded
affective and bodily resonance, and autobiographical
emotional memories of similar real situations [4,5]. These
are assumed to result in a feeling of familiarity, or the
experiencing of nonaesthetic, narrative or fiction emotions [6] such as empathy, sympathy and identification
[7], or vicarious anger or fear. According to the fiction
feeling hypothesis of the model, narratives with emotional contents, especially negative, arousing and suspenseful ones, facilitate immersion by inviting readers to
be more empathic with protagonists compared with
neutral texts, thus engaging the affective empathy network, described by Walter [8] to include the anterior
0959-4965 © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins
apparently caused increasing involvement of the core
structure of pain and affective empathy the more readers
immersed in the text. The predominant locus of effects
in the mid-cingulate cortex seems to reflect that the
immersive experience was particularly facilitated by the
motor component of affective empathy for our stimuli from
the Harry Potter series featuring particularly vivid
descriptions of the behavioural aspects of
emotion. NeuroReport 25:1356–1361 © 2014
Wolters Kluwer Health | Lippincott Williams & Wilkins.
NeuroReport 2014, 25:1356–1361
Keywords: absorption, affective empathy, fiction feeling hypothesis,
mid-cingulate cortex, pain empathy, reading, transportation
Department of Education and Psychology, bLanguages of Emotion, cDahlem
Institute for Neuroimaging of Emotion (DINE), Freie Universität Berlin, Berlin,
Germany and dDepartment of Social, Cognitive and Organizational Psychology,
Universidad de La Laguna, San Cristóbal de La Laguna, Spain
Correspondence to Chun-Ting Hsu, MD, MSc, Department of Education and
Psychology, Freie Universität Berlin, Habelschwerdter Allee 45, D-14195 Berlin,
Tel: + 49 30 838 50388; e-mail: [email protected]
Received 22 August 2014 accepted 2 September 2014
insula (AI), the mid-cingulate cortex (mCC), the amygdala, the secondary somatosensory cortex and the inferior
frontal gyrus. Among them, AI and mCC were identified
in a meta-analysis [9] as the shared neural network
between pain empathy and pain experience, also associated with empathy of personal distress [10]. Using the
labelling by Vogt [11], Lamm et al. [9] defined mCC as
‘the border of anterior medial cingulate cortex and posterior anterior cingulate cortex’, which is analogous to
the ‘(dorsal) anterior cingulate cortex’ used by other
researchers (e.g. Craig [12]), who proposed that interconnected von Economo neurons in both regions enable
fast and highly integrated representations of emotional
moments and behaviours and generate metarepresentations of ‘global emotional moments’.
Here we investigated a specific version of the fiction
feeling hypothesis that assumes that readers immerse in
texts reaching the sensory-motor grounded affective
resonance and autobiographical emotional memories
through metarepresentations of ‘global emotional
moments’ generated in the AI and mCC. Indirect evidence supporting this hypothesis comes from Altmann
DOI: 10.1097/WNR.0000000000000272
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Testing the fiction feeling hypothesis Hsu et al. 1357
et al. [13], who showed that stories with negative, arousing, but – at the same time – liked contents engaged the
medial prefrontal cortex, which was functionally coupled
with the left AI.
To test our hypothesis, we presented entire text passages
from the Harry Potter series comparing the neural correlates of immersion ratings when reading fear-inducing
versus neutral contents. Pain, fear and personal distress
are the focus of affective empathy that is essential for
altruistic, prosocial behaviour in the social context [14].
Furthermore, self–other matching for autonomic/emotional
states of pain and fear has been evident across species (e.g.
rodents and monkeys [15,16]). Therefore, we expected
(i) higher immersion ratings for fear-inducing passages,
which often describe pain or personal distress, as compared
with neutral passages, and (ii) significant correlations
of immersion ratings with activity in the affective empathy
network, particularly AI and mCC, associated with
pain empathy for fear-inducing, but not for neutral,
(iii) respective emotional contents were unambiguous
and consistent throughout the passage. To control for
other factors potentially influencing reading performance, we matched numbers of letters, words, sentences
and subordinate sentence per passage, the number of
persons or characters (as the narrative element), the type
of intercharacter interaction and the incidence of supranatural events (i.e. magic) involved in text passages
across the emotional categories.
The 120 text passages were divided into two subsets of
60, each containing 20 passages per emotional condition,
maintaining control for all relevant variables mentioned
above. During the experiment, each participant read one
subset in German and the other one in English. Each
subset was presented to 12 participants in German and
English each. Only the data for reading in German and for
the ‘Fear’ and ‘Neutral’ conditions were used for analyses
in this study (see Hsu et al. [17] for analyses of the complete data following a bilingualism research question).
Participants and methods
Twenty-four right-handed native German speakers (16
women) gave written consent to take part in the experiment, which was approved by the ethics committee of
the Freie Universität Berlin and conducted in compliance with the Code of Ethics of the World Medical
Association (Declaration of Helsinki). Their ages ranged
from 18 to 31 years (mean ± SD = 23.71 ± 3.67). All participants had read at least one Harry Potter book and
were therefore familiar with its context enough to
understand the novel-specific contents. They all had
normal or corrected-to-normal vision, and reported no
neurological or psychiatric disorders. Participants were
adequately compensated monetarily or with course
credits for their participation.
The experiment consisted of four runs, each containing
15 German passages and 15 English passages. The order
of presentation was pseudorandomized so that the distribution of the language switch positions was balanced in
each run and across all participants. Similar to the design
of a previously successful functional MRI (fMRI)
experiment on text-reading [13], each passage was presented for 14 s in the MR scanner, distributed along four
lines (shown consecutively for 3.5 s each), and then followed by 14 s of fixation cross. The visual input was
presented on a computer screen and was reflected onto
the participants’ eyes by means of a mirror.
To prepare our stimulus material we screened all seven
Harry Potter novels (German translations by Klaus Fritz,
Carlsen Verlag, Hamburg) for text passages representing
either strongly fear-inducing or particularly neutral
moments or events. Eighty text passages were selected as
stimuli: each of 40 passages entered the ‘Fear’ or
‘Neutral’ categories and were presented together with 40
fillers used to balance the emotional contents. The
selection of passages was based on a pilot study of
emotional ratings on the dimensions of valence, arousal,
fear and happiness (see Hsu et al. [17] for further details).
Each passage was about four lines in length. Passage
selection further ensured the following: (i) understanding
the passage did not require a high level of familiarity with
Harry Potter novels; (ii) emotional connotations of the
passage clearly evolved at the very beginning; and
To keep participants attentive, four randomly selected
passages in each run were immediately followed by an
emotion-unrelated, context-specific, yes/no question
(e.g. ‘Was Harry in a train station?’ ‘Was the alarm clock
broken again?’), to which participants responded by
button press.
Functional magnetic resonance imaging data acquisition
Functional data were acquired on a Siemens Tim Trio
3 T MR scanner (Erlangen, Germany). Four runs of 440
volumes were measured using a T2Ã -weighted echo-planar
sequence [slice thickness: 3 mm, no gap, 37 slices,
repetition time (TR): 2 s, echo time (TE): 30 ms, flip
angle: 70°, matrix: 64 × 64, field of view (FOV): 192 mm,
voxel size: 3.0 mm × 3.0 mm × 3.0 mm] and individual
high-resolution T1-weighted anatomical data (MPRAGE
sequence) were acquired (TR: 1.9, TE: 2.52, FOV: 256,
matrix: 256 × 256, sagittal plane, slice thickness: 1 mm,
176 slices, resolution: 1.0 mm × 1.0 mm × 1.0 mm).
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
NeuroReport 2014, Vol 25 No 17
Functional magnetic resonance imaging preprocessing
The fMRI data were preprocessed and analysed using the
software package SPM8 (http//
Preprocessing consisted of slice-timing correction, realignment for motion correction, and sequential coregistration. Structural images were segmented into grey
matter, white matter, cerebrospinal fluid, bone, soft tissue
and air/background with the ‘New Segment’ module [18].
A group anatomical template was created with the
DARTEL (Diffeomorphic Anatomical Registration using
Exponentiated Lie algebra [19]) toolbox from the segmented grey and white matter images. Transformation
parameters for structural images were then applied to
functional images to normalize them to the brain template of
the Montreal Neurological Institute (MNI) supplied with
SPM. For the univariate analysis, functional images were
resampled to a resolution of 1.5 mm × 1.5 mm × 1.5 mm and
spatially smoothed with a kernel of 6 mm full-width at halfmaximum during normalization.
Post-hoc immersion ratings
For the current study, we collected post-hoc ratings from
20 native German speakers (15 women, age 18–34 years,
mean age ± SD = 22.95 ± 4.5) who were rewarded with
course credits. They all liked the Harry Potter novel
series and had read at least one book of it. The participants were asked to read each passage in the authorized
German translation and to provide ratings on immersion
by matching their subjective experience with the
description ‘I forgot the world around me while reading’
on a scale from 1 (totally untrue) to 7 (totally true) on an
online questionnaire. The rating was based on the reading experience and narrative engagement scales by Appel
et al. [20]. We calculated the mean immersion values of
each passage for further analyses.
Fear > fixation] and [German-Neutral > fixation] conditions was made in the analysis of variance analysis
to cover the neural correlates of reading in this study.
Contrasts representing the parametric modulatory effects
of passage immersion ratings on haemodynamic responses in both the [German-Fear] and the [German-Neutral]
conditions were used at the group level to model a
random-effect paired t-test analysis to explore potential
differential correlation between immersion ratings and
BOLD responses between two experimental conditions.
The fMRI analyses were conducted at the whole brain
level; we used an initial voxel-level threshold of uncorrected P value less than 0.005, and then a cluster-level
threshold of family-wise error corrected P value less than
0.05 for the entire image volume, as suggested by
Lieberman and Cunningham [21], for studies in cognitive, social, and affective neuroscience. The labels
reported were taken from the ‘TD Labels’ [22] in the
WFU Pickatlas Tool. The Brodmann areas (BAs) were
further checked with the Talairach Client using nearest
grey matter search after coordinate transformation with
the WFU Pickatlas Tool.
Behavioural performance
All participants correctly responded to context-specific
questions for German passages in the scanner above
chance (≥62.5%) with an overall mean accuracy of
81.47 ± 13.16%.
Conditional comparison for passage immersion ratings
In the Student t-test, immersion ratings of fear-inducing
passages (mean ± SD = 3.75 ± 0.29) were significantly higher
than those of neutral passages (mean ± SD = 3.18 ± 0.26;
Student’s t-test: P < 0.001).
Functional magnetic resonance imaging data analysis
Statistical parametric maps were obtained by multiple
regressions of the data onto a model of the haemodynamic response. At the subject level, we constructed one
model of parametric analyses. This model contained
regressors for [German-Fear], [German-Neutral] and
[Fillers] conditions, and each passage lasted 14 s. The
context-specific questions were modelled as the fourth
condition, and each question was presented for 4 s. The
six realignment parameters were modelled as six additional regressors. For the [German-Fear] and [GermanNeutral] conditions, we entered the mean values of
immersion ratings for each passage in the condition as
parametric modulators for linear regression.
Regressors were convolved with the canonical haemodynamic response function in SPM8. For each participant, contrasts of the [German-Fear > fixation] and the
[German-Neutral > fixation] conditions were used at the
group level to model a random-effect one-way analysis of
variance. A conjunction contrast of the [German-
Functional magnetic resonance imaging results
Conjunction of fear-inducing and neutral narratives
Results of the conjunction contrast of [GermanFear > fixation] and [German-Neutral > fixation] conditions are summarized in Table 1 and displayed in Fig. 1a,
showing the neural substrates associated with the general
reading process irrespective of emotional contents in the
current study.
Differential parametric effects of immersion
With respect to our hypothesis, we found one cluster in
the middle cingulate gyrus (BA 32, family-wise errorcorrected cluster-level P = 0.037, cluster size: 414 voxels,
containing three peaks: MNI coordinates [x y z] = [8 14 39],
[12 23 42], and [3 23 31]) (Fig. 1b), in which the correlation
between immersion ratings and BOLD responses was
significantly more positive when reading fear-inducing
passages than when reading neutral passages, as shown in
the contrast estimates in the coordinate [8 14 39] (Fig. 1c).
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Testing the fiction feeling hypothesis Hsu et al. 1359
Table 1
Results of the conjunction analysis for [German-fear > fixation] and [German-neutral > fixation] reading conditions
[x, y, z]
Occipital pole (lingual gyrus)
Medial SMA (SFG)
STS (STG, MTG, including aTL)
LPFC (MFG and IFG) and precentral gyrus
[ − 12 − 97 − 8]
[18 − 87 − 3]
[22 − 87 − 11]
[ − 48 − 3 49]
[ − 54 − 39 4]
[ − 58 − 33 1]
[ − 3 2 66]
[8 11 54]
[50 − 28 − 0]
[54 − 12 − 9]
[53 12 − 20]
[54 6 45]
[57 − 3 43]
[56 30 7]
[2 − 36 − 39]
[ − 2 − 36 − 32]
[3 − 33 − 23]
Precentral gyrus
Pons and midbrain
aTL, anterior temporal lobe; BA, Brodmann area; H, hemisphere; IFG, inferior frontal gyrus; L, left hemisphere; LPFC, lateral prefrontal cortex; MFG, middle frontal gyrus;
MTG, middle temporal gyrus; R, right hemisphere; SFG, superior frontal gyrus; SMA, supplementary motor area; STG, superior temporal gyrus; STS, superior temporal
sulcus; T, t values.
This study investigated the fiction feeling hypothesis of
the neurocognitive model of literary reading [5], which
states that emotional contents, especially negative,
arousing and suspenseful ones, activate the affective
empathy network and facilitate immersive reading
For the conjunctional contrast focussing on brain activation during narrative reading, results are in line with a
previous study on short story reading [13] and a metaanalysis on neural correlates of text comprehension [23].
The bilateral medial superior frontal gyrus (BA 6) has
been associated with encoding written language [23].
The bilateral lateral prefrontal and temporal cortices are
part of the Extended Language Network [23].
At the behavioural level, although participants obviously
immerse in both fear-inducing and neutral passages
(mean rating values of immersion in both conditions were
between three and four out of seven), they felt significantly more immersed in fear-inducing than in neutral
passages. This suggests that, although emotional content
such as descriptions of vicarious fear is not necessary for
immersive reading experience, arousing, negatively
valenced texts do indeed facilitate immersive processes.
At the neuronal level, we expected immersion ratings to
correlate more strongly with haemodynamic responses in
AI and mCC, the core substrates of affective empathy [8,9],
when reading fear-inducing versus neutral passages.
Although being first associated with pain empathy [9], AI
and mCC were also reported to be responsive to pain of
others in patients with congenital insensitivity to pain [24].
Thus, it has been proposed that these regions might also
be associated with empathy of personal distress [10], or
understanding and sharing of the same emotion perceived
in others [9]. As expected, mCC, but not AI (discussed in
the next paragraph), showed higher correlation with
immersion experience for passages of the Fear condition
featuring many descriptions of protagonists’ pain or personal distress as compared with the Neutral condition.
With regard to both structures, Craig [12] considered
mCC to be the limbic motor cortex and the site of
emotional behavioural initiation, whereas AI is the sensory counterpart. With respect to our stimuli from Harry
Potter series, in which behavioural aspects of emotion are
particularly vividly described, the motor component of
affective empathy (i.e. mCC) might predominate during
emotional involvement, and facilitate immersive experience. This assumption is in line with the claim of Gygax
et al. [25] that descriptions of behavioural components of
emotions, as opposed to the labelling of emotions per se,
are better markers of readers’ mental representations of
the protagonists’ affective status – which might explain
why our effects were restricted to mCC and did not
extend to AI.
Our study is the first attempt to understand the neural
mechanisms of immersive reading experience.
Consistent with the assumption that pain empathy is
crucial for the cohesion of social groups [14], correlations
between immersion and BOLD responses were most
pronounced for fear-inducing stimuli. Although the Harry
Potter series offer exceptionally beloved literary materials, we remain of course cautious not to generalize the
present results to all daily immersive reading experiences. Further studies are necessary to elucidate the
relationship between the reading process, affective
empathy, and immersion at the behavioural and neuronal
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
NeuroReport 2014, Vol 25 No 17
Fig. 1
fMRI results. (a) Results of the conjunction contrast of [German-Fear > fixation] and [German-Neutral > fixation] conditions. The grey scale indicates t
values. (b) The mid-cingulate gyrus showing a significant correlation difference between passage immersion ratings and BOLD response in the Fear
versus Neutral conditions, cross-hair highlighting the peak voxel (Montreal Neurological Institute coordinate [x y z] = [8 14 39]). (c) The estimated
response strength in the peak [8 14 39] for both experimental conditions. The error bars represent 90% confidence intervals.
This research was supported by two grants from
Deutsche Forschungsgemeinschaft (DFG) to Markus
Conrad: ‘Bilingualism and Affectivity in Reading’,
Project 201, and ‘Sound-physiognomy in Language
Organization, Processing and Production’, Project 410,
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Testing the fiction feeling hypothesis Hsu et al. 1361
from the Research Excellence Cluster ‘Languages of
Emotion’ at the Freie Universität Berlin.
Conflicts of interest
There are no conflicts of interest.
1 Nell V. Lost in a book: the psychology of reading for pleasure. New Haven,
London: Yale University Press; 1988.
2 Green MC. Transportation into narrative worlds: the role of prior knowledge
and perceived realism. Discourse Process 2004; 38:247–266.
3 Schrott R, Jacobs AM. Brain and poetry: how we construct our realities.
München, Germany: Carl Hanser Verlag; 2011.
4 Jacobs AM. Neurocognitive poetics: elements of a neurocognitive model of
literary reading. In: Schrott R, Jacobs AM, editors. Brain and poetry: how we
construct our realities. Munich, Germany: Carl Hanser Verlag; 2011.
pp. 492–524.
5 Jacobs AM. Towards a neurocognitive poetics model of literary reading. In:
Willems RM, editor. Towards a cognitive neuroscience of natural language
use. Cambridge: Cambridge University Press; 2014.
6 Lüdtke J, Meyer-Sickendiek B, Jacobs AM. Immersing in the stillness of an
early morning: testing the mood empathy hypothesis in poems. Psychol
Aesthet Creativity Arts 2014; 8:363–377.
7 Oatley K. A taxonomy of the emotions of literary response and a theory of
identification in fictional narrative. Poetics 1995; 23:53–74.
8 Walter H. Social cognitive neuroscience of empathy: concepts, circuits,
and genes. Emotion Rev 2012; 4:9–17.
9 Lamm C, Decety J, Singer T. Meta-analytic evidence for common and distinct
neural networks associated with directly experienced pain and empathy
for pain. NeuroImage 2011; 54:2492–2502.
10 Yamada M, Decety J. Unconscious affective processing and empathy: an
investigation of subliminal priming on the detection of painful facial
expressions. Pain 2009; 143:71–75.
11 Vogt BA. Pain and emotion interactions in subregions of the cingulate gyrus.
Nat Rev Neurosci 2005; 6:533–544.
Craig AD. How do you feel – now? The anterior insula and human
awareness. Nat Rev Neurosci 2009; 10:59–70.
Altmann U, Bohrn IC, Lubrich O, Menninghaus W, Jacobs AM. The power of
emotional valence-from cognitive to affective processes in reading. Front
Hum Neurosci 2012; 6:192.
De Waal FB. Putting the altruism back into altruism: the evolution of empathy.
Annu Rev Psychol 2008; 59:279–300.
Hecht EE, Patterson R, Barbey AK. What can other animals tell us about
human social cognition? An evolutionary perspective on reflective and
reflexive processing. Front Hum Neurosci 2012; 6:224.
Panksepp J, Panksepp JB. Toward a cross-species understanding of
empathy. Trends Neurosci 2013; 36:489–496.
Hsu CT, Jacobs AM, Conrad M. Can Harry Potter still put a spell on us in a
second language? An fMRI study on reading emotion-laden literature in Late
Bilinguals. Cortex (in press).
Ashburner J, Friston KJ. Unified segmentation. NeuroImage 2005;
Ashburner J. A fast diffeomorphic image registration algorithm. Neuroimage
2007; 38:95–113.
Appel M, Koch E, Schreier M, Groeben N. Assessing experiential states
during reading: scale development. Z Medienpsychol 2002; 14:149–154.
Lieberman MD, Cunningham WA. Type I and type II error concerns in fMRI
research: re-balancing the scale. Soc Cogn Affect Neurosci 2009;
Lancaster JL, Woldorff MG, Parsons LM, Liotti M, Freitas CS, Rainey L, et al.
Automated Talairach atlas labels for functional brain mapping. Hum Brain
Mapp 2000; 10:120–131.
Ferstl EC, Neumann J, Bogler C, von Cramon DY. The extended language
network: a meta-analysis of neuroimaging studies on text comprehension.
Hum Brain Mapp 2008; 29:581–593.
Danziger N, Faillenot I, Peyron R. Can we share a pain we never felt? Neural
correlates of empathy in patients with congenital insensitivity to pain. Neuron
2009; 61:203–212.
Gygax P, Tapiero I, Carruzzo E. Emotion inferences during reading
comprehension: what evidence can the self-pace reading paradigm provide?
Discourse Process 2007; 44:33–50.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.