SiGN The Signal for Good News Model R Package

Overview

Designed for researchers in behavioural science, SiGN is an R package that applies the Signal for Good News (SiGN) model to predict choice behaviour in operant conditioning settings.

Installation

You can install SiGN’s development version from GitHub with:

install.packages("pak")
pak::pak("SiGN-R/SiGN")

Then load the package with:

library(SiGN)

Usage

To get started with the package, begin by reading the Get Started article, which introduces the core functionality of the SiGN package.

Then, explore the following articles to see other uses of the package:

• Classic Concurrent-Chains
  • Re-analyzes classic concurrent-chain experiments using SiGN predictions, and shows how to generate predictions across a range of parameter values.
• Predictions From Data Frames
  • Demonstrates how to generate model predictions in batch from a data frame of input parameters.

For a more advanced use case involving model evaluation, see the following:

• Model Evaluation: Descriptive Statistics
  • Introduces metrics such as RMSE, MAE, and Lin’s Concordance Coefficient for assessing model performance.
• Model Evaluation: Information-Theoretic Statistics
  • Describes how to compare choice models generally (not just the SiGN model) using likelihood-based metrics such as AIC, BIC, and Bayes Factors.

Ready-to-Use Suboptimal Choice Data and Profiles

The package includes:

• Built-in datasets — Two curated datasets from Dunn et al. (2024):

  • subopt_full: the complete dataset of suboptimal choice studies.

  • subopt_avian: a filtered subset focused on pigeons and starlings.

• Built-in parameter profiles — Predefined setups for well-known procedures from the literature, including:

  • "zentall" (Stagner and Zentall, 2010)

  • "kendall" (Kendall, 1985)

  • "fantino" (Fantino, 1969)

• Fully vectorised parameter input — All model parameters can be customised and passed as vectors, enabling efficient simulation of multiple conditions.

Try the Package Without Installing R

You can generate predictions with the SiGN model in your browser using the free, interactive SiGN Calculator — no installation or coding required.

Set up custom choice scenarios and instantly view predicted preferences and model diagnostics. It’s a great way to explore how delay reduction and reinforcement schedules influence decision-making.

📎 This is ideal for students, collaborators, or researchers who want to explore the model without any writing code.

Planned Features

The SiGN package is under active development. Upcoming updates will expand its capabilities to support a generalised version of the SiGN model that incorporates free parameters for model fitting and comparison.

Additional tools for model evaluation, simulation, and visualisation are also being considered.

Stay tuned for updates, and feel free to suggest features or contribute!

References

Dunn, R. M., Pisklak, J. M., McDevitt, M. A., & Spetch, M. L. (2024). Suboptimal choice: A review and quantification of the signal for good news (SiGN) model. Psychological Review. 131(1), 58-78. https://doi.org/10.1037/rev0000416

Fantino, E. (1969). Choice and rate of reinforcement. Journal of the Experimental Analysis of Behavior, 12(5), 723–730. https://doi.org/10.1901/jeab.1969.12-723

Kendall, S. B. (1985). A further study of choice and percentage reinforcement. Behavioural Processes, 10(4), 399–413. https://doi.org/10.1016/0376-6357(85)90040-3

Stagner, J. P., & Zentall, T. R. (2010). Suboptimal choice behavior by pigeons. Psychonomic Bulletin & Review, 17(3), 412–416. https://doi.org/10.3758/PBR.17.3.412