Can literary refugee narratives decrease ingroup bias? An experimental reader-response study using implicit methods

:speech_balloon: Speaker: R. L. Victoria Pöhls

:classical_building: Affiliation: Max-Planck-Institute for Empirical Aesthetics

Title: Can literary refugee narratives decrease ingroup bias? An experimental reader-response study using implicit methods

Abstract (long version below): Can narrative empathy of engaged readers carry over to members of the represented group in real life? In this experimental study, it was examined whether readers of refugee short stories show less ingroup bias on an implicit measurement task when compared to a control group. Contrary to expectations, no general decrease was found. The narratives had opposite effects on different readers: While some became less biased and showed this effect even after being retested two weeks later, it had the opposite effect on some, making them more biased towards refugees.


:newspaper: Long abstract


Fiction reading may facilitate the understanding of characters that are very different to the reader and/or belong to a different social group. But does the narrative empathy engaged readers may develop for characters carry over to members of the represented group in real life? Will they be perceived as less different to one’s own ingroup after contact through fiction? To find out whether this is the case, I examined stereotypical expectancies regarding refugees by readers who either read two fictional short stories written from a refugee perspective (test condition) or non-refugee short stories (control) and then answered items on linguistic intergroup bias (LIB).

Description of methods

Participants (N=286, German as only mother tongue) read different short stories based on their random assignment to either test group 1 (TG1), test group 2 (TG2) or control (C): Both TG1 and TG2 read stories about refugees’ troubled settling-in in their ‘adopted home country’ , while stories in the control were of similar length, but did not touch upon the topic of refugees. TG1 read the original stories, focalized through a 1st-person-refugee narrator, whereas TG2 read the same stories but stylistically manipulated, narrated by a 3rd-person narrator, to determine whether focalization matters.
To assess whether stereotypical expectancies regarding refugees changed and ingroup bias decreased through reading stories about this outgroup, an implicit test for Linguistic Intergroup Bias (LIB) was developed based on the theory and findings of Maass et al. (1989) to avoid a social-desirability bias, and participants were not fully informed about the overall study objective.
According to the pertinent findings of the LIB, the same behaviour is linguistically encoded differently when performed by either ingroup or outgroup members.
The developed LIB test consists of 45 images, showing characters introduced as refugees (outgroup), foreigners (outgroup), or native Germans (ingroup), who (inter)act in various social situations either seen as undesirable or desirable behaviours as pretests (N=15) confirmed.
After each image, participants were presented with four slightly different verbal descriptions describing the situation seen. Descriptions vary along dimensions of abstractness vs. concreteness, thereby allowing to implicitly test participants’ stereotypical expectations by having them rate the sentences for image fit.
When viewing a positively valenced action by ingroup members, participants more often describe it with a stability-indicating adjective (e.g., “The person is friendly”). In contrast, the same action performed by outgroup members tends to be labelled with concrete action verbs, suggesting no cue to an underlying positive character trait (e.g., “The person lends the hammer to his neighbour”).
Participants were tested right before, right after and two weeks after reading, allowing possible time-sensitive effects to appear.


Data is analysed using the software R. LIB scores are calculated for each individual at all measuring time points, allowing for within-subject and between-groups comparison.
I hypothesized that LIB scores for control group members will not change significantly across time and baseline measures before reading will not differ significantly between groups.
It is further expected that TG1 and TG2 measures after reading will show a decreased LIB compared to C – as they have engaged with a refugee’s perspective, the felt distance between their own ingroup and the refugee outgroup should have decreased.
As this is, to my knowledge, the first study attempting to measure LIB across time, it was difficult to foresee whether these effects increase or decrease between the second and third measuring time point: Attributing the changed attitude to fiction reading might enlarge the effect two weeks later once participants do not remember this connection (cf. Appel 2007), but, being a largely unconscious effect, it might also wear off once the fictional experience that prompted the change lies further in the past.
Story liking, narrative engagement, and identification with characters was measured by self-report measures. I hypothesized that in TG1 and TG2 higher ratings on these measures correlate with greater decrease in LIB measurements, compared to less impacted readers.


A mixed ANOVA shows that LIB scores are normally distributed and, as expected, not significantly different across groups before reading. As hypothesized, there are also no significant differences within the control group across measuring time points, while a different trend can be observed in test conditions: Contrary to expectations, no overall reduction of bias is observed, but a flattening of the curve to both sides. Thus, some participants become more, whereas others become less biased. Data analysis is still ongoing with regard to the question whether this can be attributed to personal characteristics (age, gender, education, political leaning) or aesthetical experience of the narratives (liking, personal relevance, felt emotions, identification with characters).

Hello Victoria, this is an interesting project along the lines of what I would call applied literariness (applying literary forms to solve real world social issues). I applaud you for the attempt. As to the results, the flattening you describe suggests a entrenchment in a’priori positions. There is some recent data from Markus Appel (not so far away from you in Wurzburg) that when readers feel that they are being influenced (manipulated?) in certain ways it backfires and leads to a strengthening in the opposite direction. I dont know you data set well enough to know if this is the case here. But still a route worth considering.

A different issue is one you referred to in passing. To change an LIB score you might need a very stringent intervention. This might be too high a bench mark for the exposure you include. Having said that perhaps there are shifts going on the are worth considering but far more minor and subtle.

In any case I wish you good luck with this project. I apologize that I cannot be at the session itself. But perhaps we can discuss this at one of the gather sessions.

Cheers David Hanauer

Hello David, thanks for your encouraging words and for sharing your thoughts!
The implicitness of the measurement (i. e. trying to make sure that participants are not aware what they are being tested on and do not feel manipulated either) was indeed a very important consideration when designing the study. I chose the LIB test, because - from all the literature that is out there - participants a) report that they do not realize what they are tested on, b) it still “works” even when they are explicitly told beforehand that this measurement will measure their in-group bias, suggesting that they can’t figure out which response to the picture would cast a positive light on them and make them look unbiased, and c) a person’s results on the LIB correlate with other implicit measures of prejudice (while explicit ones not necessarily do). Additionally, the whole experiment was presented as a study on stylistic preferences and creative writing. When asking a random 15% of our sample after the experiment what the experiment was about, all of them were answering along these lines and nobody made a connection between the refugee texts and the comics. So all in all, I’m quite certain that they did not feel manipulated - the surest sign that this is not the case might actually be that we do not see a similar trend on the other measure (the outgroup portrayal task): if people felt manipulated and show an adverse reaction to the outgroup because of that, this should also show on the complimentary measure, I think.

I therefore believe that, as you say, to change deeply ingrained, spontaneous reactions to outgroups, more is needed than the reading of 2 short stories. And I’m not too surprised that this is the case. I’m actually more surprised that we see all these results on the outgroup portrayal task!

I will chase you in Gathertown later to talk more about this :slight_smile:

Hi @victoria_poehls Thank you for your very interesting presentation. I appreciate the thoroughness of your design and conclusions. I was wondering along the same lines as David about the length of exposure and perhaps also the timing of the task after exposure. I can imagine, for example, that reading a book, rather than a short story, is able to lead to less ingroup bias. Similarly, giving participants more time to reflect on what they have read, maybe using think aloud protocols or diaries beforeusing the lib test could influence the results on the lib test? Are you going to do any follow-up studies? And could you tell us a little more about how you would change your design if you are?

Hi Victoria, thanks so much for the really interesting presentation!
First I have a question: maybe I miss it, but the participants of the creative writing task are the same of the LIB?
If not, the different results you found may be due to a problem of generalization. It is just speculation, but maybe in the creative writing you are pushing the reader to give an identity to the outgroup member, and indeed through reflection, he/she is able to develop a non stereotypical story. Instead, this is not the case in implicit thinking, as the LIB requires a fast judgement, and the reader still base it on his/her stereotypical thinking. It would be very interesting to see if maybe a longer exposure to non-stereotypical stories would influence also this fast judgment, not mediated by reflection

Hi @moniek.kuijpers ! Yes, definitely, length of exposure and engagement tasks after reading might influence results on the LIB. My mindset when designing this study was to find a set-up where effects are already possible, but also - in my mind - not highly likely. So that if I would find any effects of reading under these conditions, it would be strong evidence that reading in more naturalistic settings - you read more than two stories about a topic, you engage in private conversations or classroom activities afterwards, etc. - would be even more likely to have these or stronger effects.
Of course, I would like to follow up on this, and have multiple ideas of what to change - and I hope we can discuss them (and yours) in the session! (If I start rambling about this here and now, I won’t manage to have lunch before the sessions start… :wink: )

Thanks for your presentation Victoria, I loved it. Good to see how your research has progressed since the previous presentation I attended. I really like the clever approach to implicit measurement, very important in this field. In addition, your work is of obvious societal relevance - which I also appreciate highly, thinking of the influx of refugees at the moment in Europe, but also of the differences between how Ukrainian and Syrian refugees are welcomed.

I agree that one possible explanation of your present results is that the exposure is simply not strong enough, although there are studies that suggest it might. In my 2000 book I review some older studies, often conducted in educational settings. But also see for instance Johnson, D. R., Jasper, D. M., Griffin, S., & Huffman, B. L. (2013). Reading narrative fiction reduces Arab-Muslim prejudice and offers a safe haven from intergroup anxiety. Social cognition , 31 (5), 578., which you probably know. Some of the studies also suggest what other factors (other than narrator) may play a role, like transportation or post-reading activities. Are these variables that you consider in your experiments?

Hi @Giulia_Scapin !
Sorry if that was not clear, yes, they are the same participants!
Exactly, that is what I believe and @moniek.kuijpers and @DHanauer also suggest, I think: the reflection period is key for a change in readers portrayal, while quick perception of a situation still leads to biased encoding.

Hi Frank @Hakemulder , thanks for your suggestions!
Yes, Johnson (2013) and also Johnson (2011) are very interesting in comparison, because the effects they are looking at are similar and some findings might be related. For example, I think it is very interesting that they find a bias towards fearful expressions, which suggests that people become more and even overly perceptive that someone might be in need of help, while I also find that participants entertain wishful thinking in terms of finding happy ends for victims and aggressors, but especially victims, when confronted with a situation of conflict. I think that this might be similarly based on the wish to help (even if they only have fictive means to do so in my experimental set-up).
Regarding other factors: I measure transportation, which does not seem to have an effect, but there were no post-reading activities. If I were to design a follow-up, I would definitely include a condition that increases reflection on the literary material by different kinds of post-reading activities, but here I thought it would counteract my idea of “implicitness” of the tasks.

Victoria, thanks for your talk. I enjoyed it!

I think the paper Shapiro talk about could be one of those:

  • Fong, G. T. et al., 2004 - Fong, G. T., Zanna, M. P., & Dal Cin, S. (2004). Narrative persuasion and overcoming resistance. In Resistance and persuasion (Vol. 4, pp. 175–191).
  • Slater & Rouner, 2002 - Slater, M. D., & Rouner, D. (2002). Entertainment—Education and Elaboration Likelihood: Understanding the Processing of Narrative Persuasion. Communication Theory, 12(2), 173–191.
  • Cohen, Tal-or & Mazor-Tregerman, 2015