Olfactory research almost exclusively measures ability in controlled lab settings, yet humans encounter odors in complex, multi-sensory environments. This project revisits a published field study testing odor identification and verbal description across three London locations with distinct ambient odor profiles -- a neutral art museum, a polluted urban street, and a food market -- paired with a controlled lab replication. The reanalysis applies binary accuracy scoring, correspondence analysis, Cain four-level coding, and direction tracking to assess whether alternative analytical approaches change the conclusions. The dominant finding is a design problem, not a null result: repeated exposure within a single session confounds learning with environment, masking any true location effect.
Hoover, K.C. (2020). Field-testing olfactory ability to understand human olfactory ecology. American Journal of Human Biology, 32(5), e23411. DOI: 10.1002/ajhb.23411
Note: data-field methods.xlsx and data-cain.xlsx were updated in 2026 to remove actual ages and, in the case of field methods, replace with ranges to prevent any possibility of reidentification of individuals.
This repo contains two sets of files. All files prefixed revised- are revised scripts and figures from the original publication. Revisions offer a fully reproducible end-to-end pipeline from raw data through ETL to results — including data cleaning and wrangling scripts (revised from originals or created where none existed), code modernization, and the addition of environment management to future-proof the revised scripts. All other files are associated with the original publication. Occasionally, an original script may be updated with annotations for clarity; the commit history will note any such changes.
The portfolio page includes a full project narrative, key findings, and figures.
Languages: R
Packages: readxl | dplyr | tidyr | stringr | broom | ggplot2 | scales | irr | skimr | FactoMineR | factoextra | nnet | car | tibble
Designing and reanalyzing field studies that test real-world behavior; applying multiple analytical frameworks to assess robustness of findings; building reproducible workflows that extend and modernize published research.
- Code and scripts are licensed under the MIT License.
- Data, figures, and written content © Kara C. Hoover, licensed under CC BY-NC-ND 4.0.