The Cambrian Period, primarily known for animal life diversifying, experienced global extinctions. Pulses of extinction in Cambrian Series 2 are exemplified by the disappearance of archaeocyath sponges and olenelline and redlichiid trilobites. However, the effect of such extinctions on outer shelf organisms, as typify Burgess Shale-type (BST) deposits, remains relatively unknown. The phylogeny of naraoiid arthropods, represented in BST deposits globally, has consequently been reconstructed from either side of the Series 2–Miaolingian extinction event to evaluate the response of offshore marine organisms to Cambrian environmental perturbation. As soft anatomy is known for only a subset of naraoiid species, exoskeletal morphology has proven important. Misszhouia and Naraoia (Naraoia) are distinguished morphometrically by posterior shield length/width and anterior shield length/posterior shield length. Morphometry has also been used to strengthen the identification of some cryptic naraoiid species and revise stratigraphic ranges. A revised phylogeny for naraoiids reveals Misszhouia as a monophyletic subgenus, the former genus Pseudonaraoia nests within Naraoia and is placed in synonymy, and the systematic position and status of the Subfamily Liwiinae are sensitive to character weighting. Ten species of Naraoiidae range through the Series 2–Miaolingian boundary, all naraoiid lineages originating during the main BST window. The persistence of outer shelf naraoiids through the Series 2 extinctions suggests that deeper offshore marine environments were resilient to extinction during periods of environmental stress. This study therefore provides novel empirical support for the asylum of BST communities, which may contribute to the taxonomic longevity and widespread geographic distribution of taxa in these biotas.

Detailed descriptions were made of 259 naraoiid specimens from the United States National Museum of Natural History (USNM), Yunnan Key Laboratory for Paleobiology (YKLP) and Royal Tyrell Museum (TMP) collections. Multiple specimens sharing the same diagnostic features were grouped into single species, then compared to previous diagnoses and descriptions in the literature to determine species names and incorporate additional features not preserved in the specimens studied.

Photography

For the USNM and TMP specimens from the Burgess Shale cross polarised light was used for photography (Bengtson 2000).

Equipment used: Canon 600D DSLR, Canon MP-E 65mm f2.8 1-5x Macro Manual Focus Lens, Canon EF-S 60mm f/2.8 Macro USM Lens, Hoya PL CIR Filter, Copy Stand, Canon 550D DSLR, Hoya 52mm Pro-1 MC PL-C Filter.

The specimens were levelled using modelling clay. Low angle, plane polarised light was then provided from the top left. A polarising film was placed in front of the light source, and a polarising filter placed on the camera. The specimen was submerged in alcohol/water, and the white balance was altered slightly to darken the image and better define soft anatomy. Each specimen was photographed at several levels of focus. Alcohol was initially used to submerge the specimens, but it left staining on the specimens upon drying so water was therefore subsequently used. The YKLP Chengjiang specimens were studied by plane polarised light microscopy and fluorescence light microscopy.

Equipment used: Leica M205 FA Fluorescence Stereo Microscope, Leica DFC 7000 T Camera, Leica KL 3000 LED, Keyence VHX 5000 digital microscope, Keyence VHZ20R/W/T Lens.

Plane polarised light microscopy allowed for the detailed study of the specimens through the higher magnification and photography of certain body parts. For more detailed images or more complex structures fluorescence microscopy was used. Specimens were photographed in air with a low-angle, plane polarised light source directed from the top left. The specimen was brought into focus and the white balance and exposure settings of the microscope were edited to allow for detailed, well exposed images of various parts of the organism.

For fluorescence microscopy the lights in the room were dimmed in order for the fluorescent light to show more clearly any fluorescing organic material (e.g. the digestive tract). Fluorescent light with a wavelength of 561nm was used to "stimulate" the specimens and a filter with wavelength of 600-650nm was used to visualise the signal.

Fluorescence microscope settings: exposure- 11 seconds, gain- 3.1x, gamma- 0.83, image type- greyscale.

Quantitative methods

Equipment used: Tacklife DC02 Advanced Vernier Calliper.

USNM specimens that preserved their original dorsal morphology were measured using a digital vernier caliper. The specimens with the least visible deformation and entire shield margins were then selected to characterise naraoiid morphotypes. However, no attempt was made during this study to evaluate morphometry in relation to bedding orientation, or the position of the splitting-plane on the specimen. The calliper provided precise measurements (±0.03mm) of the sagittal length (from the anterior margin of the anterior shield to the posterior margin of the posterior shield) of the specimen and the width and length of both the anterior and posterior shields. These measurements were then used to calculate the proportions of exoskeletal shape.

The YKLP specimens that were studied exhibited less well-defined dorsal morphologies and therefore the proportions of the Chengjiang species (M. longicaudata and N. spinosa) were extracted from specimens in published literature.

Phylogenetic analyses

Interrelationships of all named species of naraoiids and other nektaspids were inferred based on a matrix of 27 characters (Bond & Edgecombe 2020), using parsimony and maximum likelihood as optimality criteria. Parsimony analysis explored the sensitivity of groupings to equal and implied character weights. The trees were built using PAUP* 4.0a build 167 (Swofford 2002) and TNT version 1.5 (Goloboff & Catalano 2016). Synonymous taxa were manually removed in PAUP* and the trilobite Olenoides was defined as the outgroup to Nektaspida. All characters were initially given equal weights. Exact searches using branch and bound or implicit enumeration were performed for parsimony analysis, whereas likelihood analysis under the Mk model would not allow for a branch and bound search so a heuristic search was performed (HSEARCH addseq=random nreps=1000).The characters were then re-weighted using basic settings for implied weights in TNT, and implicit enumeration searches for parsimony were repeated under all integer values for concavity constants from k=2 to k=10. Branch support was quantified by jackknife frequencies for parsimony analysis under equal weights (heuristic search, 36% character deletion, 1000 replicates), symmetric resampling for parsimony analysis under implied weights (heuristic search, 33% change probability, 1000 replicates), and bootstrap values for likelihood analysis (nreps=100). Characters were optimised on trees in PAUP* through the code: DESCRIBETREES 1 / plot=cladogram opt=deltran apolist=yes;.