Skip to main content

Swida amomum seed germination

Cite this dataset

Padgett, Donald; Gouthro, Alexis; Adams, Michelle; Surasinghe, Thilina (2022). Swida amomum seed germination [Dataset]. Dryad.


Premise: Effective seed dispersal is essential in population dynamics of plant species. Swida amomum (Silky dogwood) exhibits a dispersal syndrome characteristic of autumn-ripening shrubs with fleshy fruits, where attached fruits are ingested and defecated by birds while fallen fruits are consumed by ground-foraging birds and mammals. 

Methods: We documented that fallen fruits of this shrub were consumed by two aquatic turtle species (Eastern painted turtle, Chrysemys picta and Red-eared slider, Trachemys scripta) and that their seeds were defecated.  We compared germination success (percentage of seeds germinated) of defecated seeds, seeds collected from pond surface, and seeds removed from shrubs.

Results: While four seed taxa were identified in fecal samples, seeds of S. amomum were the most frequent (93%) among samples and the most numerous (106 seeds) in any sample.  Average proportion of fecal seeds germinated (85.99%) exceeded that of seeds from pond surface (82.76%) and from shrubs (60.24%), albeit the difference in germination success was insignificant. When analyzed using fecal samples from Painted turtles only, the difference in germination success between fecal seeds and those collected from pond or shrub became significant.

Conclusions: Our findings represent the first report of S. amomum seeds being dispersed by turtle gut passage and suggest aquatic turtles could be an important part of a secondary seed dispersal process influencing woody plant community composition in temperate wetland ecosystems.


Study site:

We conducted this study at a small (surface area 0.03 km2) freshwater impoundment (Sturtevant Pond; UTM Easting: 433438.26, UTM Northing: 46170104684.94, UTM zone: 24Z) of a semi-perennial stream located in Bridgewater, Massachusetts, USA. Pine-mixed eastern temperate deciduous forests and woodlands comprised the dominant vegetation of the upland habitats. The watershed is largely forested with a substantial extent of suburban and residential development.   

Field sampling and seed collection­:

We conducted this study for three consecutive years (2017–2019), from late August to early October in each year. This timeframe coincided with both the turtle activity, autumn bird migration, and fruit production in silky dogwood.  For capturing turtles, we targeted two species (Chrysemys picta, Eastern painted turtles, and Trachemys scripta, Red-eared sliders) and deployed a single hoop-net trap (1 m diam., 2.5 cm mesh; Memphis Net & Twine Co., Memphis, Tennessee, USA) baited with canned, oil-emersed sardines.  For a given year, we deployed the trap overnight for 24 hours, 1–2 times per week. We recovered the trap after each trap night to check for captured turtles. Our methods are standard field techniques for surveying turtles and have been employed for capturing a representative sample of freshwater turtles elsewhere (Dodd, 2016).

We individually housed turtles in a 38 L aquarium filled with tap water to a depth of 2–3 cm for five consecutive days without feeding to collect fecal matter.  Afterward, we returned all animals to the site of capture.  For each turtle captured, seeds were manually isolated from fecal matter, rinsed, sorted, and counted.  Two sources of Swida amomum control seeds were collected from the study site in each year we trapped for turtles: 1) fruits floating on the pond surface (a random sample of 11–26 fruits) and 2) ripe fruits harvested from nearby shrubs (donor shrubs were 1–2 m apart, 11–15 fruits collected per plant, one fruit per inflorescence).  Prior to storage, all pulp was removed by agitation after soaking in water for 24–48 h.  In the first year of study (2017), control fruits were limited to only those floating on pond surface.  All seeds were surface sterilized with a 30 s 0.05% hypochlorite solution wash, followed by two sterile water rinses, and stored in sterile water for 4 months at 5°C to break dormancy (Allen and Farmer, 1972). 

To measure germination, we placed entire seed samples (1–64 seeds per fecal sample; 12–26 seeds per “pond” control; 12–15 seeds per “shrub” control) on moist filter paper, with individual seeds physically apart from other seeds, in petri dishes.  Dishes were positioned in a growth chamber (Percival CU30L2C8; Percival Scientific, Perry, Iowa, USA) under a 12h light/dark, 25/15°C cycle for 50 consecutive days. Seeds were observed once every two days for germination and filter paper wetted as needed with sterile water. Seeds were removed from the dishes as they germinated. For each seed sample, both controls and fecal, percentage seed germination was calculated ((number germinated/total seeds in sample)*100).

Statistical analyses:

To test for significant differences in percent seed germination between fecal, pond, and shrub samples, we ran a one-way Analyses of Variance (1-ANOVA) permutational (20,000 permutations) test (Legendre and Legendre 2012). For post hoc pairwise comparisons, we performed unpaired, Yuen’s trimmed means robust test for unequal variance (Welch approximation), and applied Benjamini and Hochberg (1995) method to correct for false discovery rate and adjust probability estimations. The graphics were generated following Patil (2021).

We opted for permutation tests since our sample sizes were low, non-normal, and unbalanced. In permutation tests, instead of comparing test statistics to a standard statistical distribution, the reference distribution is generated by randomly permuting the dataset under the study (Legendre and Legendre 2012). Likewise, the robust test we applied for pairwise comparisons is distribution independent and performs well despite heteroscedasticity and small sample sizes. Since just one fecal sample was collected from Trachemys scripta, we ran tests both including and excluding the fecal samples collected from this species. Statistical inferences were made at 95% confident intervals. We used R Statistical Programming Environment (R Core Team 2022) within RStudio integrated Developmental Environment for all the analyses (RStudio 2022.02.3 Build 492).  


Bridgewater State University