**Final Sample** We obtained a sample of 70 participants (female, 52; male, 17; and other, 1). The mean age was 20.91 (SD = 5.31). **Analysis** All analyses were done using R, version 3.5.1, and the lme4 package (Bates et al., 2016). We directly followed the R analysis code provided by the original authors for use in the CREP project where possible, always using multi-level models specifying participant as a random factor (random intercepts). Specifically, we used the lmer function for continuous outcome variables (e.g., RT) and the glmer function for binomial outcome variables (e.g., correct/incorrect answers to comprehension questions, mindwandering Y/N). **Results** The main result of the present study was that mind-wandering differed by text difficulty, where people were more likely to mind-wander during difficult passages compared to during easy passages. Furthermore, the occurrence of mind wandering negatively affected participants reading comprehension scores. These two results both replicated the findings from Feng et al., 2013. However, we did not replicate other findings reported in Feng et al., 2013. For instance, we did not find that the effect of mind-wandering on passage comprehension was moderated by task difficulty (no significant interaction). We also did not find any significant predictors of reading time. For instance, the mind-wandering did not predict time spent reading of text passages within this study. This may have been because reaction time was more highly varied (with more outliers) due to our replication occurring online instead of in a lab environment. All of the results in this study (both direct replications of the Feng analyses and exploratory analyses of our extension questions) are available in "Data and Results", in "Replication Results.pdf". **Unanticipated Changes to Preregistered Procedure** Due to a programing error, participants assigned to condition one did not receive a mind-wandering thought probe for question 4.2. Instead of Windsorizing outliers that were faster than the 1st percentile or slower than the 99th percentile in our own data, we used the same cut-off values that were used to determine outliers in Feng et al., (2013). Reaction times were considered outliers if they were faster than 395ms or slower than 24,640ms. Our treatment of outliers had not been previously stated in the preregistration. Although we had preregistered a sample size of 80 participants, only 70 participants completed the study over the course of the semester. The CREP Feng replication is finishing in December 2020, so we will not be continuing data collection into the next semester.