Darwin attributed the absence of species transitions in the fossil record to his hypothesis that speciation occurs within isolated habitat patches too geographically restricted to be captured by fossil sequences. Mayr’s peripatric speciation model added that such speciation would be rapid, further explaining missing evidence of diversification. Indeed, Eldredge and Gould’s original punctuated equilibrium model combined Darwin’s conjecture, Mayr’s model, and 124 years of unsuccessfully sampling the fossil record for transitions. Observing such divergence, however, could illustrate the tempo and mode of evolution during early speciation. Here, we investigate peripatric divergence in a Miocene stickleback fish, Gasterosteus doryssus. This lineage appeared and, over ~8,000 generations, evolved significant reduction of twelve of sixteen traits related to armor, swimming, and diet, relative to its ancestral population. This was greater morphological divergence than we observed between reproductively isolated, benthic-limnetic ecotypes of extant Gasterosteus aculeatus. Therefore, we infer that reproductive isolation was evolving. However, local extinction of low-armoured G. doryssus lineages shows how young isolate populations often disappear, supporting Darwin’s explanation for missing evidence and revealing a mechanism behind morphological stasis. Exctinction may also account for limited sustained divergence within the stickleback species complex and help reconcile speciation rate variation observed across time scales.

(a) Fossil specimen data

To quantify morphological divergence in the fossils, we compiled published data from the D (Bell et al. 1985), L (Bell et al. 2006), and K series (Stuart et al. 2020, Voje et al. 2022). See Figure 1A for stratigraphic correlations among series. Series D consisted of 26 samples spaced at ~5,000-year intervals over an estimated 108,275 years (Figure 1A). Six traits were measured from D: standard length, pelvic score, and the number of pre-dorsal pterygiophores, dorsal spines, anal-fin rays, and dorsal-fin rays (Bell et al. 1985). Series L comprised a section starting ~4,500 years before the replacement event until ~16,500 years after and was sampled continuously. Three armour traits were measured for series L: pelvic score, number of dorsal spines, and number of touching pre-dorsal pterygiophores. This series confirmed replacement of lineage I by lineage II within ~125 years (Bell et al. 2006) and that subsequent evolution of lineage II was probably caused by directional natural selection (Hunt et al. 2008). Our analysis focuses on series K (Stuart et al. 2020; Voje et al. 2022) because 16 traits were measured for K (Table S1), allowing estimates of multivariate divergence of traits that should reflect swimming, feeding, and defense. These traits are also divergent between benthic and limnetic ecotypes in the species-pair lakes as well as among allopatric generalist populations of G. aculeatus (Walker 1997; Spoljaric and Reimchen 2007; Willacker et al. 2010). Series K consisted of 18 samples taken at ~1000-year intervals, and mean sample times span ~16,363 years. Series K starts at the replacement, when lineage I and lineage II specimens occurred in a single sample. We removed lineage I fish from this sample for morphological analysis. Traits measured and their sample sizes are provided in Table S1 and Table S2, respectively. Finally, we note that a parapatric, highly armored form with a benthic diet (Bell 2009) was collected from Quarry E (of Brown 1986), approximately 1.7km from the depositional environment studied here. This site was dated to roughly the same time as the K series (Brown 1986) and appears to have been a nearshore sample, based on abundant terrestrial plant fossils and thick clastic layers embedded within the diatomite (Bell 2009). This contrasts the open water habitat that characterized the depositional environment of series D, L, and K (Bell 2009). 

(b) Extant specimen data

For comparison to fossil divergence, we measured benthic and limnetic ecotypes from five species-pair lakes (Table 2) for the same 16 traits that were scored in the K-series fossils (Table S1). Specimens were loaned by D. Schluter’s lab (University of British Columbia). They collected from Enos Lake in 1988 and from Emily, Little Quarry, Paxton, and Priest Lakes in 2018. The Enos specimens were fixed in formalin and stored in 40% isopropanol. The other specimens were initially preserved in 95% ethanol in the field before being transferred to water then formalin in the lab and stored in 40% isopropanol. In 2019, we stained these specimens for bone using Alizarin Red. Standard length as well as pelvic-spine lengths were measured with calipers. We used a dissection microscope to count dorsal spines, pelvic spines, dorsal-fin rays, and anal-fin rays. Right and left-side pelvic girdle and ectocoracoid lengths were measured from ventral photographs (Canon EOS Rebel T7, Tamron 16-300 mm MACRO lens, Kaiser RS1 copy stand). Lateral X-rays were used to measure dorsal spine length, number of pterygiophores anterior to the pterygiophore holding the third spine, length of the pterygiophore just anterior to the third spine, cleithrum length, and pre-maxilla ascending branch length. We also counted vertebrae from X-rays: abdominal vertebrae were counted anterior to the first vertebra with a haemal spine contacting an anal fin pterygiophore. Caudal vertebrae were posterior, including the first vertebra with the haemal spine contacting the anal fin pterygiophore (Aguirre et al. 2014). X-rays were taken with an AXR Hot Shot X-ray Machine (Associated X-ray Corporation) at the Field Museum of Natural History. Specimens were exposed at 35kV and 4mA for 7 to 10 seconds. We developed the film and scanned individual fish images using the B&W Negatives setting on an Epson Perfection 4990 Photo flatbed at 2400 dpi. Measurements from photographs and X-rays were taken with FIJI (Schindelin et al. 2012) and its plugin ObjectJ (https://sils.fnwi.uva.nl/bcb/objectj/). Table S3 reports sample sizes by population, ecotype, and trait.

Microsoft Excel is helpful to open the data files.

R and R Studio are necessary to run the analysis scripts.