Overview

This Dryad deposit contains:

1) Processed dyad-alignment datasets needed to reproduce the main θ₁–θ₂ alignment heatmaps and analysis reported in the associated manuscript accepted at Science Advances, and

2) Raw per-individual time series (2D position and absolute heading) provided for transparency and for additional/alternative analyses beyond the published workflow.

The repository is designed to be usable without reading the article. All variables, units, coordinate conventions, and geometric definitions are documented below.

Contents at a glance

Folders

1. Datasets_dyad_alignment___r_theta1_theta2/

Primary reproducibility dataset. Dyad-level samples with three columns per row: r, theta1, theta2.

2. Datasets_Raw___Time_X_Y_Alpha/

Raw per-individual trajectories and absolute orientations, organized by condition, year, and observation session.

3. Codes_for_Datasets_dyad_alignment___r_theta1_theta2/ (optional)

Scripts used to create radial windows and generate θ₁–θ₂ heatmaps from the provided dyad files.

Not required to reproduce the manuscript if you only need the finalized dyad .dat files as data products.

Zipped mirrors (if included)

- Datasets_dyad_alignment___r_theta1_theta2.zip

- Datasets_Raw___Time_X_Y_Alpha.zip

- Codes_for_Datasets_dyad_alignment___r_theta1_theta2.zip

If present, the .zip files contain the same structure and files as the corresponding folders.

1) Primary reproducibility dataset: dyad alignment samples (r, θ₁, θ₂)

Folder: Datasets_dyad_alignment___r_theta1_theta2/

These files contain dyad-level alignment samples used for the manuscript’s θ₁–θ₂ heatmaps and analysis.

Reproducibility note: The files in this folder are the primary dataset required to reproduce the manuscript’s dyad-alignment results.

Files

- LC1_adults_dyad_alignments.dat

- LC1_preschoolers_dyad_alignments.dat

- LC2_adults_dyad_alignments.dat

- LC2_preschoolers_dyad_alignments.dat

Each file is whitespace-delimited (spaces/tabs) with one dyad sample per row and three numeric columns.

.dat column definitions (order is exactly as stored)

Each row has three columns in the following order:

1. r

Inter-individual distance between the two individuals (units: meters).

2. theta1

Relative alignment angle for individual 1, defined with respect to the dyad connecting direction from individual 1 to individual 2

(units: degrees, wrapped to [-180, 180]).

3. theta2

Relative alignment angle for individual 2, defined with respect to the dyad connecting direction from individual 2 to individual 1

(units: degrees, wrapped to [-180, 180]).

These angles are dyad-centric (not global). They depend only on the two headings and the vector connecting the pair.

Exact geometric definition of θ₁ and θ₂ (standalone)

For a pair of individuals A and B with positions (x1, y1) and (x2, y2) and absolute headings α1 and α2:

- Define the dyad vectors:

- AB = (x2 − x1, y2 − y1) (from A to B)

- BA = (x1 − x2, y1 − y2) (from B to A)

- Define heading unit vectors from the absolute headings:

- u1 = (cos α1, sin α1)

- u2 = (cos α2, sin α2)

θ₁ is the signed angle from AB to u1 (pivot at A), computed as:

θ₁ = atan2( AB_x * u1_y − AB_y * u1_x , AB_x * u1_x + AB_y * u1_y )

θ₂ is the signed angle from BA to u2 (pivot at B), computed as:

θ₂ = atan2( BA_x * u2_y − BA_y * u2_x , BA_x * u2_x + BA_y * u2_y )

Sign convention: counterclockwise rotations are positive. Angles are wrapped to the principal interval.

Unit convention in this repository:

- Raw headings α in the CSV files are stored in degrees.

- theta1 and theta2 in the .dat files are stored in degrees and wrapped to [-180, 180].

(Implementations may compute angles internally in radians for trigonometric functions and then convert to degrees.)

Interpretation examples

- Face-to-face dyads: θ₁ ≈ 0 and θ₂ ≈ 0.

- Side-by-side (parallel relative to the connecting line): θ₁ and θ₂ near ±90°.

2) Raw data: per-individual time series (Time, x, y, α)

Folder: Datasets_Raw___Time_X_Y_Alpha/

This folder contains raw tracking data for each individual as separate CSV files.

Reuse note: These raw files are provided for transparency and to support alternative analyses. They are not required to reproduce the manuscript’s dyad heatmap results if you use the provided dyad .dat files above.

Nested folder structure (how the raw files are organized)

- Top level: Datasets_Raw___Time_X_Y_Alpha/

- Condition and year folders:

- LC1_Year#1/, LC1_Year#2/, LC1_Year#3/, LC1_Year#4/

- LC2_Year#1/, LC2_Year#2/, LC2_Year#3/

- Within each year folder are observation-session folders named:

- LC*_Year#??_Observation#??/

Each Observation#?? folder corresponds to one observation session. All individuals tracked in that session appear as separate CSV files inside that folder, for example:

- Adult_1.csv, Adult_2.csv, ... (one file per adult when present)

- Child_1.csv, Child_2.csv, ... (one file per child when present)

Each CSV is a time series for one tracked individual during that observation session.

CSV column definitions

Each raw CSV contains the following columns:

- Time

Integer time index for the observation. Rows with the same Time across individuals in the same observation are synchronized.

`Time` is an index with a constant sampling interval within each observation.

- Center_x

x-coordinate of the individual’s estimated 2D position in the classroom (units: meters). Positions are computed from dual-tag localization (each individual wears two tags, and the reported center position is derived from the two tag locations).

- Center_y

y-coordinate of the individual’s estimated 2D position in the classroom (units: meters). Positions are computed from dual-tag localization as described above.

Coordinate system note:

For each classroom, the (0,0) origin of the coordinate system is defined separately and reset, so Center_x and Center_y are only comparable within the same classroom, not across different classrooms.

- Alpha (Degrees)

Absolute heading angle α of the individual measured relative to the +x axis (units: degrees).

α is in the range [0, 360). Angles increase counterclockwise.

Notes:

- These raw CSVs provide absolute quantities in a classroom-specific coordinate system.

- The dyad-level quantities r, theta1, and theta2 can be computed deterministically from pairs of individuals at matched Time using the geometric definition above.

3) Code (optional)

Folder: Codes_for_Datasets_dyad_alignment___r_theta1_theta2/

This folder (if included) contains scripts used to generate radial windows and θ₁–θ₂ heatmaps from the finalized dyad .dat files. These scripts are provided for convenience and transparency, but the processed dyad .dat files in Datasets_dyad_alignment___r_theta1_theta2/ are the primary data product for reproducing the manuscript’s dyad heatmap results.

Files

- Code1_splitting_dyad_alignment_data_across_radial_windows.py

Splits a dyad alignment .dat file into separate radial windows based on r.

Default radial window size is 0.6 m.

- Code2_dyad_alignment_data-to-heatmap.py

For a given radial-window file, computes the θ₁–θ₂ alignment distribution and produces a heatmap normalized by a uniform distribution. It also writes the normalized heatmap matrix to disk.

How to reproduce the θ₁–θ₂ heatmaps (from the provided dyad .dat files)

Step 1: split a dyad .dat file into radial windows

Example:

`python3 Code1_splitting_dyad_alignment_data_across_radial_windows.py LC1_preschoolers_dyad_alignments.dat`

This generates multiple .txt files, one per radial range, for example:

- LC1_preschoolers_dyad_alignments_0.0-0.6.txt

- LC1_preschoolers_dyad_alignments_0.6-1.2.txt

- LC1_preschoolers_dyad_alignments_1.2-1.8.txt

(and so on)

Each radial-window .txt file contains the same three columns in the same order: r, theta1, theta2, restricted to the corresponding r interval.

Step 2: generate the normalized heatmap for a given radial window

Example:

`python3 Code2_dyad_alignment_data-to-heatmap.py LC1_preschoolers_dyad_alignments_0.0-0.6.txt`

This will:

1) Save a normalized heatmap matrix, for example:

`LC1_preschoolers_dyad_alignments_0.0-0.6_normalized_heatmaps.txt`

2) Display a publication-style heatmap plot using the settings used in the manuscript.

Repeat Step 2 for each radial-window file created in Step 1.

Notes on naming conventions

- LC1 and LC2 denote two classroom conditions used in the study.

- preschoolers and adults denote the group membership used when forming dyads for the derived .dat files.

- Year#k denotes the data collection year index within a condition.

- Observation#k denotes a specific observation session folder that contains the per-individual CSV files for that session.

Contact: For questions about the dataset structure, conventions, or variable definitions:

Debasish Sarker (deb.sker@gmail.com)

Repository DOI: 10.5061/dryad.2rbnzs82v

Human subjects data

All study procedures were reviewed and approved by the University of Miami Institutional Review Board (IRB Protocol 20160509). Written informed consent was obtained from all participating teachers, and parental/guardian consent was obtained for all participating children. The study involved preschool-aged children in a familiar classroom environment, and no interventions were introduced by the research team during data collection. All data were anonymized prior to analysis.