# Introduction A common advice to aspiring writer is to *show, don’t tell*. With this it is meant that mental states of characters should not be explicitly described in a narrative, but that the reader should be able to infer the mental state based on a description of the character’s action (or something else). Therefore, instead of saying: >"He read the e-mail and it made him angry", The author should describe: >"He read the e-mail and banged his head to the table". Although this is a common rule in writing, its effectiveness has not been investigated extensively before (but see [Tankard and Hendrickson, 1996](http://journals.sagepub.com/doi/pdf/10.1177/073953299601700105)). In the current behavioral study we aim to investigate the effect of **show don’t tell** on appreciation of a short narrative. Hypothesis: **Show don’t tell will lead to higher appreciation and more 'transportation' than tell don’t show.** # Methods ## Sampling plan Sequential bayesian independent samples t-test will be conducted, which provides a numerical value that quantifies how well alternative hypothesis predicts the empirical data relative to null hypothesis. The sequential hypothesis testing with bayes factors procedure can be outlined as four steps, according to [Schonbrodt and colleagues](https://www.researchgate.net/profile/Felix_Schoenbrodt/publication/286971067_Sequential_Hypothesis_Testing_With_Bayes_Factors_Efficiently_Testing_Mean_Differences/links/5673d52408aee7a427458f83.pdf): 1. Define a prior threshold referring to as "H0 boundary" and "H1 boundary". In this study, we choose **6** as boundary for BF10 (1/6 for BF01), according to the recommended parameters setting of sequential bayesian independent samples t-test; 2. Choose a prior distribution for the effect size under H1. **r=1** is recommended. After the boundary is reached, a sensitivity analysis will be conducted over reasonable r scale parameters to show the results are robust to the choice of the prior; 3. Pre-register this study with the predefined threshold and prior effect size distribution; 4. Run a minimal number of participants. In this case, it will be **20 participants for each condition**. We will calculate the BF and increase sample size after the minimal amount of participants has been collected. A new BF will be calculated after each participant. If the BF >= 6 or the BF <= 1/6, we will stop the data collection. The maximum size sample we are willing to collect is **128** (64 for each condition). If BF does not reach the critical value after 128 participants in total, data collection will be terminated. The following results will be reported: * Bayes factor; * The distribution of the prior and estimated posterior effect size; * The median and 95% CI of the effect size; * The trajectory of bayes factor in sequential analysis The analysis will be conducted in [JASP](https://jasp-stats.org/). ## Procedure The study will be conducted as a pen and paper task in Dutch language. First participants will read a story. Appreciation of the story will be measured via a multi-variable questionnaire (see materials). Transportation will be measured using the story world absorption scale (SWAS) developed by Kuipers and colleagues. After appreciation and transportation questionnaires, participants will answer three comprehension questions about the story. General reading habits will be assessed, which consists of general questions (e.g. How much do you read fiction?) and the Author Recognition Test (ART). In the end, demographic information will be asked. ## Participants We will recruit participants from the general student population at Radboud University. ## Exclusion criterion Three comprehension questions will be asked. Participants will be included only if they answer at least two out of three questions correctly. ## Conditions The materials consist of one short narrative written by the writer Rob van Essen especially for this occasion. The story is called "het geluid" ("the noise"), and it consists in two versions: * show don’t tell (SDT) * tell don’t show (TDS) In the materials that we will use this difference pertains mainly to the more or less explicit description of emotion and cognitive states of characters. In SDT these need to be inferred by the reader, and in TDS these are explicitly described. ## Manipulation check We will match the stories in terms of the following psycholinguistic variables, which are computed at the word level * word length * lexical frequency, [see link](http://crr.ugent.be/programs-data/subtitle-frequencies/subtlex-nl) * orthogra:phic neighbourhood size, [see link](http://clearpond.northwestern.edu/dutchpond.php) * phonological neighbourhood size, [see link](http://clearpond.northwestern.edu/dutchpond.php) * Age of acquisition, [see link](http://crr.ugent.be/archives/1602) * Concrecreteness, [see link](http://crr.ugent.be/archives/1602) Check the manipulation results in Material. No difference between these psycholinguistic variables in two stories. ## Design Between subjects: every participant only reads one version of the narrative. Dependent variables are: * appreciation of the narrative (multi-facetted) * transportation (via SWAS) ### Main analysis We will compare the two conditions on appreciation and subscale of the SWAS (according to Kuipers and colleagues) using independent t-tests (details of statistical procedure see above). Appreciation will be quantified as 1) general appreciation (How much did you like this story?) and 2) the subquestions of the multi-variable appreciation scale and subscale of the SWAS. As stated above, the hypothesis is that SDT will lead to higher appreciation and transportation scores as compared to TDS. ### Exploratory analysis We will first conduct Bayesian ANCOVA by putting general reading habits as covariate, conditions as the fixed factor, and appreciation and transportation as dependent variables. If the BF(inclusion) of interaction between conditions and general reading habits in the analysis of effects is larger than 5, which means the homogeneity of regression slopes assumption of ANCOVA is violated. Then we will conduct the analysis in linear mixed model in R.