CTT Parental Care Species Dataset: Supplementary Data for "Counterpart Tracking: The Architecture of Predictable Inter-Organism Interaction"

This repository contains the supplementary materials for Masensi (2026), "Counterpart Tracking: The Architecture of Predictable Inter-Organism Interaction," which derives and tests structural impossibility predictions for parental care from a counterpart-tracking framework using a three-arm empirical protocol. The dataset codes 210 species across 6 phyla (Chordata, Echinodermata, Mollusca, Cnidaria, Annelida, Arthropoda) for reproductive mode, offspring characteristics, and parental care presence. It supports a counterexample search testing whether state-dependent parental care can occur when transmissibility between parent and offspring is structurally absent. Across 125 species meeting the impossibility conditions (95 verified broadcast spawners in large environments with parental departure; 30 precocial dispersers with no offspring signaling), zero counterexamples were found. An additional 39 positive-control species, precocial offspring with signaling or travel limitation, all exhibit parental care as predicted. A confinement test across 71 of the 95 broadcast spawner species found no emergent parental care when proximity was artificially restored via spatial confinement. A matched-pair comparison of 10 non-caring species against 16 closely related caring comparators across 23 pair comparisons found that non-caring species scored equal to or higher on capacity in all cases; the differentiator is whether the lineage evolved a parental care template. A phylogenetic robustness check on a corrected 26-species backbone chronogram confirmed all results (phylogenetic logistic regression p < 0.001; tip-shuffle permutation p = 0.0036; sign test p < 0.001). The workbook (CTT_Parental_Care_Dataset_v4_cleaned.xlsx) contains five sheets: Species Data (194 species), Formation Condition Scores, Matched Pairs (23 pair comparisons across 16 caring comparators), Audit Log (documenting 30 reclassifications from v3 to v4 with sources), and Phylogenetic Regression. The codebook (CTT_Codebook.docx) documents all column definitions, the scoring rubric, and the confinement and matched-pair methodologies. The full prompts document (CTT_Full_Prompts_Document.pdf) records all 28 data-generation, cross-validation, and analysis prompts across four phases. Data were collected using a four-phase AI-assisted methodology designed to prevent confirmation bias. Phase 1 used theory-blind prompts (no mention of the predictions being tested) to generate species lists via DeepSeek V3 and Grok 2. Phase 2 used ChatGPT (GPT-4o) and Gemini 1.5 Pro for independent cross-validation and adversarial counterexample searches. Phase 3 used Claude (Sonnet 3.5) for final analysis after data collection was complete. Phase 4 used theory-blind prompts for the confinement test (4A, Claude Opus 4.6) and matched-pair comparators (4B, DeepSeek V3). A 20% subset was independently double-coded (Cohen's κ = 0.91). All 28 prompts are provided in full.