A sensory investment syndrome hypothesis: Personality and predictability are linked to sensory capacity in the hermit crab Pagurus bernhardus
Data files
May 15, 2025 version files 56.26 MB
-
fit_mass_file.rds
18.74 MB
-
fit_MJC_file.rds
18.66 MB
-
fit_MNC_file.rds
18.73 MB
-
README.md
4.83 KB
-
SensoryStartle.csv
112.06 KB
-
SensoryStartle.R
17.72 KB
Abstract
Correlated phenotypic traits (i.e., syndromes) may manifest as associations between different behavioural types or between behavioural and non-behavioural phenotypes. While research on syndromes is extensive, correlations involving behavioural type and sensory morphology have yet to be investigated. Sensation is essential in decision-making and should be correlated with behavioural phenotypes involved in risk response, including boldness. We investigated correlations between boldness and sensory capacity in Pagurus bernhardus hermit crabs, taking repeat measures of startle response durations to assess hermit crab personality and predictability. The correlation between startle response and the sensillar density (i.e., number of sensilla per unit surface area) of both chelipeds was assessed using Bayesian-fitted double hierarchical general linear models (DHGLMs). Negative correlations between these two traits support the existence of a syndrome linking sensory capacity and behavioural type, hereafter distinguished as a ‘sensory investment syndrome’. Increasing sensillar density on the major claw also corresponded with reduced within-individual variation, or predictability, in startle response duration. By demonstrating a correlation between a sensory and a behavioural phenotype, our results demonstrate the importance of considering sensory morphology and performance in behavioural ecology and show how sensory investment syndromes may be linked to support behavioural strategies that help to maximise fitness.
Dataset DOI: 10.5061/dryad.ksn02v7gm
Description of the data and file structure
The following data files were used to investigate the potential for a correlation between a behavioural phenotype (boldness) and sensory morphological phenotype (sensillar density or the number of sensilla per unit chelar surface area) in the common hermit crab Pagurus bernhardus. Morphological data were collected by scanning electron microscopy and image analysis. Behavioural phenotype (animal personality and predictability) data were collected by collecting repeat measures (n=11) of the startle response durations for individual hermit crabs. Data were analysed using Bayesian-fitted DHGLMs.
Files and variables
File: SensoryStartle.R
Description: R code used to analyse data from the attached .csv file
File: SensoryStartle.csv
Description: Datasheet with all variables analysed in this research study
File: fit_MJC_file.rds
Description: The output for the brms() model examining the effect of major chelar (MJC) sensillar density on hermit crab startle response duration (SRD)
File: fit_MNC_file.rds
Description: The output for the brms() model examining the effect of minor chelar (MNC) sensillar density on hermit crab startle response duration (SRD)
File: fit_mass_file.rds
Description: The output for the brms() model examining the effect of crab mass on hermit crab startle response duration (SRD)
Variables
- crab_ID: Unique ID assigned to each focal P. bernhardus hermit crab
- site: Field collection site (Mount Batten, Plymouth, Devon, UK, and Hananfore, Looe, Cornwall, UK)
- hcsx: Hermit crab sex (female vs. male)
- hcmss: Hermit crab mass (g)
- shmss: Shell mass (g) provided to each focal crab
- SRD_no: The startle response number (0 = initial through 10 = final) for each hermit crab
- SRD: The startle response duration (s) for each hermit crab
- includeRPT: Include (1) or exclude (0) from the analysis; reasons for exclusion are elaborated in the Methods section of the paper, but include death and autotomy.
- MJC_SEM: Was a scanning electron micrograph of the major cheliped generated for this individual hermit crab (1 = Yes, 0 = No)
- single sensilla: The number of single sensilla identified on the major cheliped’s dorsal surface.
- dublet sensilla: The number of double sensilla identified on the major cheliped’s dorsal surface.
- triplet sensilla: The number of triple sensilla identified on the major cheliped’s dorsal surface.
- sensillar bundle: The number of bundled sensilla identified on the major cheliped’s dorsal surface.
- TOTAL_SENSILLA: The total number of sensilla identified on the major cheliped’s dorsal surface.
- MJC chelar area: The surface area of the major cheliped
- density total: The total sensillar density of the major cheliped (total sensilla / chelar area)
- MNC_SEM: Was a scanning electron micrograph of the minor cheliped generated for this individual hermit crab (1 = Yes, 0 = No)
- MNC_SINGLE: The number of single sensilla identified on the minor cheliped’s dorsal surface.
- MNC_DOUBLET: The number of double sensilla identified on the minor cheliped’s dorsal surface.
- MNC_TRIPLET: The number of triple sensilla identified on the minor cheliped’s dorsal surface.
- MNC_BUNDLED: The number of bundled sensilla identified on the minor cheliped’s dorsal surface.
- MNC_TOTAL: The total number of sensilla identified on the minor cheliped’s dorsal surface.
- MNC chelar area: The surface area of the minor cheliped
- MNC_SENS_DENSITY: The total sensillar density of the minor cheliped (total sensilla / chelar area)
- premoult_death: Did the hermit crab die before moulting (necessary for SEM methodology) but after all 11 startle responses had been collected? (1 = Yes, 0 = No)
- fit_MJC_file: The output for the brms() model examining the effect of major chelar (MJC) sensillar density on hermit crab startle response duration (SRD)
- fit_MNC_file: The output for the brms() model examining the effect of minor chelar (MNC) sensillar density on hermit crab startle response duration (SRD)
- fit_mass_file: The output for the brms() model examining the effect of crab mass on hermit crab startle response duration (SRD)
Code/software
All data were analysed in R (version 4.4.1) with RStudio (version 2024.09.0+375). Figures were generated with ggplot2 (version 3.5.1). Linear mixed effects models were analysed using lme4 (version 1.1-35.5) and lmerTest (version 3.1-3). The DHGLM was fitted using a Bayesian approach with the ‘brms’ package (version 2.22.0).
Access information
NA