## data
> * we prepared datasets using the preparation scripts (see preparation scripts in [analyses](https://osf.io/yxju4/) component)
> * all "clean" data are deidentified (i.e., we do not share raw data, but we share the scripts used to generate cleaned data from raw data)
> * each data folder includes study-specific data from Phase 1 (participants generate representations) and Phase 2 (independent participants rate those representations)
> 1. **codebook.** variable names, Qualtrics labels, and some category labels for stimuli (bare bones)
> 2. **questionnaire.** questionnaire responses from MediaLab (Phase 1 data)
> 3. **rcict.** reverse correlation image classification task responses from Qualtrics (Phase 1 data collected on MTurk)
> 4. **directrt.** reverse correlation image classification task responses from DirectRT (Phase 1 data collected in lab)
> 5. **ratings.** independent ratings from Qualtrics (Phase 2 data collected on MTurk)
> 6. **turkprime.** turkprime HIT data
> 7. **cis.rds.** classification image(s) pixel data from the [rcicr package](https://cran.r-project.org/web/packages/rcicr/index.html) in R (.rds is a file extension that refers to R data, which can be loaded into R as an object)
> * for Study 1 and 3, we conducted multiple ratings surveys (i.e., Phase 2), but we only report in the manuscript data from rating survey 5 of Study 1 and from rating survey 2 of Study 3 (for a catalog of all ratings studies, see "representation feature ratings summary" in the [materials](https://osf.io/9fm2q/) component)