The study area is situated in the southeastern part of Jianhe County, Guizhou Province, China, at 26°45.014'N, 108°24.982'E (Fig. 1). The Kaili Formation in the Miaobanpo section is in unconformable contact with the dolostone of the Tsinghsutung Formation (Cambrian Stage 4) and is overlain by the argillaceous limestone of the Jialao Formation (Miaolingian Series, Wuliuan Stage) (Yuan et al. 2002). Detailed measurements show that the Kaili Formation of the Miaobanpo section is divided into 27 beds based on lithological characteristics, with a total thickness of 156.05 m, and lithologically alternates among silty mudstone, calcareous mudstone, and a set of terrigenous clastic rocks dominated by shale, with limestone at the top and bottom (Zhou et al. 1979; Yin 1987; Zhao et al. 2019). The Kaili Formation is generally considered to have been deposited in a slope environment with a high sedimentation rate, below the storm-wave base and near the redox interface, with two microfacies depositional models, calcareous claystone facies and carbonate mudstone facies (Zhang et al. 1996; Zhu et al. 1999; Gaines et al. 2011). The abundance of macroalgae with holdfasts (Wu et al. 2016), echinoderms attached to biological debris (Lin et al. 2008; Liu et al. 2023), trace fossils (Wang et al. 2009), and trilobite exuviae (Chen et al. 2023) further indicates that the Kaili Formation was deposited under low-energy conditions below the storm-wave base and was dominated by multiphase sedimentation (Gaines et al. 2011). Within the Kaili Formation, the following trilobite zones are represented from the bottom upward: 1) Bathynotus kueichouensis–Ovatoryctocara sinensis Assemblage-Zone; 2) Oryctocephalus indicus Zone; and 3) Peronopsis taijiangensis Zone (Zhao et al. 2012, 2019).

The specimens of the trilobite Kaotaia globosa were collected from the Kaili Formation of the Miaobanpo section, mainly from the 9^th bed (indicated by specimen number GTBM9) and the 17^th bed (indicated by specimen number GTBM17). The authors selected 136 internal molds and nine external molds, with minimal evidence of postmortem distortion, as the objects of study in this paper. A total of 145 specimens are deposited at the Guizhou Provincial Key Laboratory for Paleontology and Paleoenvironment. Detailed sample numbers and measurements are provided in Suppl. materials 1, 2, respectively.

To eliminate color differences between the fossils and highlight subtle structural features, the fossils were coated with magnesium oxide before being photographed with a Canon 5D Mark IV camera and a VHX100K microscope. The approach of landmark-based geometric morphometrics aims to summarize the pattern of change in biomorphology via extracting the morphological information contained in landmarks (Bookstein 1991; Rohlf and Marcus 1993; Zelditch et al. 2004). Compared with the complex trunk homology arising from segmental exchange between the meraspid pygidium and the thorax, the structures associated with the cranidial furrows and the facial suture of trilobites are readily discernible and can be considered homologous. Consequently, research into the ontogenetic and evolutionary biology of trilobites has primarily centered on the characteristics of their cephalon (Jacobs and Carlucci 2019). The tpsUtil and tpsDig v.216 software (Rohlf 2010) were used to measure specific parts of the Kaotaia globosa cephalon and obtain coordinates for 19 landmark points (Fig. 2). The specimens examined in this paper are well preserved and minimally affected by deformation, which enabled the acquisition of data from 19 landmarks across all specimens. The collected raw morphological coordinate data (Suppl. material 1) were imported into PAST software (Hammer et al. 2001) to generate a variance-covariance matrix, and partial Procrustes superimposition was performed on the morphological coordinate data of each sample to eliminate the influence of non-shape differences.

Through regression transformation, principal component analysis (PCA) simplifies multidimensional data into fewer composite variables, with new axes accounting for maximum original variability (Quinn and Keough 2002). Applying this method, a PCA was conducted based on landmark data from the detailed ontogenetic sequence of Kaotaia globosa. Accurate stratigraphic occurrence data afford the opportunity to examine the partitioning of variation among collections with modest environmental or temporal differences (Hughes et al. 2021). In this study, the inclusion of specimens from both bed 9 and bed 17 introduces a potential confounding factor, as inter-bed differences may constrain interpretation (Hughes et al. 2021). Although these two stratigraphic beds have similar petrological characteristics, combining specimens from different beds within the dataset is an unfortunate and unsatisfactory phenomenon. To assess this potential issue, a PCA was conducted on the holaspid specimens of K. globosa from different beds, concluding that their morphological characteristics exhibit minimal variation (see Suppl. material 3: fig. S1).

Fusco et al. (2012, 2014) proposed formulas for the average per-molt growth rate (AGR; absolute growth rate) and the allometry coefficient (relative growth rate) of trilobite segments. They defined the AGR as the antilogarithm of the average per-molt growth increment (AGI). The AGI of stage i (AGIi) is equal to the mean natural logarithm of developmental stage i minus the mean natural logarithm of developmental stage 1 in the ontogenetic sequence and divided by the value of stage i minus 1. The allometry coefficient is defined as the slope of a least-squares linear regression of ln (LTS_i,d) on ln (TRL_d) (Fusco et al. 2012, 2014). Based on observed values of the thoracic segment relative to trunk length (RLS) in Aulacopleura koninckii, Fusco et al. (2014) further proposed the Segmental Gradient Hypothesis (SGH) and Trunk Gradient Hypothesis (TGH) to explain the control of segmental growth. To quantify these hypotheses, they established two growth models for SGH (geometric series function, SGgmp; power series function, SGpwp) and two for TGH (exponential function, TGexp; power law function, TGpwl). Among these, the SGH postulates that each segment released into the thorax has a constant growth rate, with its final size determined by its relative position in the sequence, and that each segment exhibits distinct autonomous growth progression (Fusco et al. 2014). Alternatively, the TGH assumes a continuous and stable trunk growth gradient, which is determined by the differing growth rates at each position along the trunk (Fusco et al. 2014).

Taphonomic compaction affects fossil morphology in both the horizontal and vertical planes, which can exaggerate the estimated magnitude of shape changes and modify the observed patterns of shape change (Webster 2007b). Although the trilobite specimens here experienced minimal taphonomic distortion, the axial ring of Kaotaia globosa is not exactly at the same level as the pleurae, resulting in a poor match between the axial ring length and pleural length (sag.). Therefore, a transverse line connecting the fulcra on both sides of each segment was drawn and used to measure the length of the thoracic segments (Fig. 2B). To eliminate the effects of measurement error, each segment was measured twice, with the mean value used for analysis. The generated images were digitized using the ImageJ software package (Schneider et al. 2012) to obtain the length of a given structure (sag.) by measuring the distance between different line segments and to generate Suppl. material 2, a section of measurements that will be used to study the growth of the thoracic segments of K. globosa. In this approach, the slight overlap between the posterior edge of the occipital ring and the anterior end of the first thoracic segment (TS1) results in the measured cephalon length (sag.) being slightly smaller than the actual value, which in this paper is approximated as equal to the body length minus the trunk length. The number of thoracic and pygidial segments was determined by placing specimens under a stereomicroscope and examining them several times to ensure maximum accuracy. The calculations and subsequent statistical tests pertaining to trunk segment growth were conducted using Microsoft Excel 2010 and Statgraphics Centurion, Version XVI. Additionally, CorelDRAW 2017 (a free trial version) and Origin 2020 were employed in the graphical representation of the findings.

The morphological terms and abbreviations used in the description generally follow Whittington and Kelly (1997). The classification criteria are principally derived from the scheme proposed by Fortey (1997). Abbreviations used in the descriptions include sag. = sagittal, exsag. = exsagittal, tr. = transverse, and S1–S3 = glabellar furrows 1 to 3.

Order Ptychopariida Swinnerton, 1915

Family Alokistocaridae Resser, 1939

The ontogenetic sequence of Kaotaia globosa Chang et Zhou in Lu et al., 1974, described herein ranges from meraspid degree 3 to the holaspid period (Figs 3, 4, 5); the number and size range of specimens at each stage are shown in Table 1. Ontogenetic stages of trilobites have traditionally been recognized on the basis of the development of exoskeletal segmentation, and the division of each phase relates specifically to the development of articulated joints between exoskeletal elements (Hughes et al. 2006). Therefore, in this paper, the ontogenetic sequence of meraspid degrees is based on an ascending number of thoracic segments, whereas the holaspid period includes all individuals that have reached the stable and mature number of thoracic segments, which in this species is 14. The main morphological variations are as follow:

Represented by three articulated specimens, with two specimens illustrated (Fig. 3A, B), shields of this phase range from 1.53 to 1.60 mm in length, with the thorax having three segments and the pygidial axis having five segments (morph; m3₅, morphological conditions are indicated as mx_y, where x is the number of thoracic segments and y is the number of pygidial segments, without the terminal piece). Cranidium semicircular, length (sag.) less than width between the midpoint of the two palpebral lobes (tr.); glabella convex, long columnar, glabellar furrows extremely indistinct; occipital furrow shallow and straight in the middle, slightly deeper and curved forward on both sides; occipital ring wider in the middle than on both sides (sag.); librigenal spine long and slender, extending to the posterior part of the pygidium; preglabellar field broad (sag.), with a small spherical bulge in front of the glabella; border furrow indistinct; fixigena wider (tr.), being approximately twice the width of the glabella between the palpebral lobes; eye ridge obliquely extending outward and slightly posteriorly from the frontal lobe of the glabella, parallel to the border furrow; palpebral lobe moderately sized, situated in a median anterior position relative to the glabella; trunk subequal to cephalon (sag.), end of pleura rounded, no pleural spine seen; pygidium relatively large, width (tr.) slightly smaller than width of cranidium (tr.).

Represented by two articulated specimens, with one specimen illustrated (Fig. 3C), shields of this phase range from 1.86 to 2.27 mm in length, with the thorax having four segments and the pygidial axis having five segments (m4₅). Cranidium semicircular, length (sag.) less than maximum width between the two palpebral lobes (tr.); glabella convex, long columnar; occipital ring convex, wider in the middle than on both sides (sag.), with a faint occipital node; librigenal spine long and slender, with the ends extending to the level of the posterior end of the pygidium; preglabellar field broad, with a spherical bulge; anterior border narrow, ca. 1/2 as long as preglabellar field (sag.); preglabellar field is broad (sag.), having a small spherical bulge; fixigena wider (tr.), being approximately twice the width of the glabella between the palpebral lobes; other characteristics are similar to those of the previous stage.

Represented by three articulated specimens, with one specimen illustrated (Fig. 3D), shields of this phase range from 2.09 to 2.32 mm in length, with the thorax having five segments and the pygidial axis having four segments (m5₄). Cranidium semicircular, length (sag.) less than width between the two palpebral lobes (tr.); glabella convex, long columnar, glabellar furrows extremely indistinct; occipital furrow shallow and straight in the middle, slightly deeper and curved forward on both sides; occipital ring wider in the middle than on both sides (sag.), with a faint occipital node; librigenal spine long and slender, extending to the level of the first pygidial segment; preglabellar field broad (sag.), with a spherical bulge; border furrow deep; fixigena wider, being approximately 1.5 times the width of the glabella between the palpebral lobes; eye ridge obliquely extending outward and slightly posteriorly from the frontal lobe of the glabella, parallel to the border furrow; palpebral lobe moderately sized, situated in a median anterior position relative to the glabella; trunk slightly longer than the cephalon (sag.), end of pleura rounded, no pleural spine seen; pygidium relatively large, width (tr.) approximately 3/4 of the width of the cranidium (tr.).

Represented by four articulated specimens, with one specimen illustrated (Fig. 3E), shields of this phase range from 2.50 to 2.63 mm in length, with the thorax having six segments and the pygidial axis having four segments (m6₄). Cranidium semicircular, length (sag.) less than width between the two palpebral lobes (tr.); glabella convex, long columnar; glabellar furrows extremely indistinct; librigenal spine long and slender, extending horizontally to the opposite end of the 6^th thoracic segment; other characteristics are similar to those of the previous stage.

Represented by five articulated specimens, with one specimen illustrated (Fig. 3F), shields of this phase range from 2.47 to 3.06 mm in length, with the thorax having seven segments and the pygidial axis having four segments (m7₄). Cranidium semicircular; glabella convex, long columnar, with three pairs of fuzzy lateral glabellar furrows; occipital ring crescent-shaped, with sides curved forward and with a faint occipital node; librigenal spine long and slender; preglabellar field broad (sag.), with a spherical bulge; anterior border narrow and gently convex, ca. 1/2 as long as the preglabellar field (sag.); fixigena wider (tr.), being approximately 1.5 times the width of the glabella between the palpebral lobes; eye ridge obliquely extending outward and slightly posteriorly from the frontal lobe of the glabella, parallel to the border furrow; palpebral lobe moderately sized, situated in a median anterior position relative to the glabella; trunk slightly longer than the cephalon (sag.), pleural spine curved backward at the end of the pleura; pygidium relatively large, width (tr.) approximately 1/2 the width of the cranidium (tr.).

Represented by five articulated specimens, with three specimens illustrated (Fig. 3G–I), shields of this phase range from 2.67 to 3.50 mm in length, with the thorax having eight segments and the pygidial axis having four segments (m8₄). Cranidium semicircular, length (sag.) less than width between the two palpebral lobes (tr.); glabella convex, truncate-conical, with three pairs of lateral glabellar furrows; occipital ring crescent-shaped, with sides curved forward and with a faint occipital node; beneath the glabella lies an ovoid and natant hypostome; librigenal spine long and slender, with the ends extending to the level of the posterior end of the pygidium; preglabellar field broad (sag.), with a spherical bulge; anterior border narrow (sag.) and gently convex, ca. 1/2 as long as the preglabellar field (sag.); border furrow deep; fixigena wider (tr.), being approximately 1.5 times the width of the glabella between the palpebral lobes; palpebral lobe moderately sized, situated in a median position relative to the glabella; trunk subequal to the cephalon (sag.), pleural spine curved backward at the end of the pleura; pygidium relatively large, width (tr.) approximately 2/5 the width of the cranidium (tr.).

Represented by five articulated specimens, with one specimen illustrated (Fig. 3J), shields of this phase range from 2.92 to 3.91 mm in length, with the thorax having nine segments and the pygidial axis having three segments (m9₃). Cranidium trapezoidal, length (sag.) less than width between the two palpebral lobes (tr.); glabella convex, truncate-conical, with three pairs of lateral glabellar furrows, S3 extending diagonally near the occipital furrow; occipital furrow shallow and straight in the middle, slightly deeper and curved forward on both sides; occipital ring crescent-shaped, with sides curved forward and a faint occipital node; pygidium relatively small, transverse width equal to or slightly greater than the width of the base of the glabella; other characteristics are similar to those of the previous stage.

Represented by nine articulated specimens, with two specimens illustrated (Fig. 3K, L), shields of this phase range from 3.55 to 4.31 mm in length, with the thorax having 10 segments and the pygidial axis having three segments (m10₃). Cranidium trapezoidal, length (sag.) less than width between the two palpebral lobes (tr.); glabella convex, truncate-conical, with three pairs of lateral glabellar furrows; occipital ring crescent-shaped, with sides curved forward and with a faint occipital node; librigenal spine long and slender, extending horizontally to the opposite end of the 9^th thoracic segment; preglabellar field broad (sag.), with a spherical bulge; pygidium relatively small, transverse width equal to the width of the base of the glabella.

Represented by 13 articulated specimens, with two specimens illustrated (Fig. 3M, N), shields of this phase range from 3.61 to 5.70 mm in length, with the thorax having 11 segments and the pygidial axis having three segments (m11₃). Cranidium trapezoidal, length (sag.) less than width between the two palpebral lobes (tr.); glabella convex, truncate-conical, with three pairs of lateral glabellar furrows; librigenal spine long and slender, with the ends extending to the level of the posterior end of the pygidium; fixigena wider (tr.), being approximately 1.5 times the width of the glabella between the palpebral lobes; palpebral lobe moderately sized, situated in a median position relative to the glabella; occipital ring crescent-shaped, with sides curved forward and with a faint occipital node; opisthoparian suture; trunk subequal in length to the cephalon (sag.), pleural spine curved backward at the end of the pleura; pygidium relatively small, width (tr.) slightly smaller than the width of the base of the glabella.

Represented by 11 articulated specimens, with three specimens illustrated (Fig. 5A–C), shields of this phase range from 4.24 to 6.02 mm in length, with the thorax having 12 segments and the pygidial axis having three segments (m12₃). Cranidium trapezoidal, length (sag.) less than width between the two palpebral lobes (tr.); glabella convex, truncate-conical, with three pairs of lateral glabellar furrows; librigenal spine long and slender, extending horizontally to the opposite end of the 10^th thoracic segment; palpebral lobe moderately sized, situated in a median position relative to the glabella; opisthoparian suture; trunk slightly longer than the cephalon (sag.); pygidium relatively small, width (tr.) slightly smaller than the width of the base of the glabella.

Represented by 15 articulated specimens, with three specimens illustrated (Fig. 5D–F), shields of this phase range from 3.57 to 9.39 mm in length, with the thorax having 13 segments and the pygidial axis having three segments (m13₃). Cranidium trapezoidal, length (sag.) less than width between the two palpebral lobes (tr.); glabella convex, truncate-conical, with three pairs of lateral glabellar furrows; occipital lobe wider in the middle than on both sides (sag.), with a faint occipital node; genal spines long and slender, extending horizontally to the opposite end of the 9^th thoracic segment; preglabellar field broad (sag.), with a spherical bulge; anterior border narrow (sag.) and gently convex, ca. 1/2 as long as the preglabellar field (sag.); palpebral lobe moderately sized, situated in a median position relative to the glabella; opisthoparian suture; pygidium relatively small, width (tr.) smaller than the width of the base of the glabella.

Represented by 70 articulated specimens, with six specimens illustrated (Fig. 5G–L), shields of this phase range from 4.882 to 18.480 mm in length, with the thorax having 14 segments and the pygidial axis having two segments (m14₂). Detailed morphological descriptions are given in the previous section on systematic paleontology.

Yuan and Zhao (1994) systematically described Kaotaia and pointed out that, in the ontogenetic sequence of Kaotaia transversa, the main morphological variations were concentrated in the glabella, palpebral lobe, anterior border, and pygidium. This paper describes the successive ontogenetic sequence of Kaotaia globosa from meraspid degree 3 to the holaspid period, and the main trends in exoskeletal changes during ontogeny are summarized below.

Based on the morphological descriptions of the developmental stages provided in the preceding section, the ontogenetic sequence of Kaotaia globosa is divided into three stages: meraspid degrees 3–8, meraspid degrees 9–13, and the holaspid period. To facilitate a visual analysis of the morphological changes occurring in the cranidium of K. globosa during ontogeny, 19 landmark points were extracted, and principal component analysis (PCA) was performed (Fig. 6). Thin-plate spline analysis, canonical variate analysis (CVA), analysis of similarities (ANOSIM), and allometric growth curves were also employed to identify the morphological differences between the three developmental stages and test their significance (see Suppl. material 3).

Because the landmark data of different developmental stages of the same species were statistically analyzed, the total proportion of the first three principal components was only 53.73%. Among them, PC1 accounts for 29.81% of the variance and is primarily associated with variation in the anterior branch of the facial suture and the anterior border (Fig. 6A). Positive scores along PC1 are associated with the anterior branch of the facial suture converging toward the sagittal axis and the anterior border protruding anteriorly. Conversely, negative scores are associated with a posteriorly and laterally extending anterior branch of the facial suture and a retracted anterior border. PC2 explains 13.02% of the variance and is primarily related to the glabella and the occipital ring (Fig. 6A). Positive scores along PC2 correspond to a narrower and more elongated glabella (sag.) combined with a shorter occipital ring (tr.). Conversely, negative scores are associated with a shorter and thicker glabella (sag.) and an elongated occipital ring (tr.). The vast majority of samples from meraspid degrees 3–8 were concentrated on the positive half of the PC2 axis. PC3 accounts for 10.90% of the variance; however, the relative warps visualized through vector and thin-plate spline transformation grids exhibit a disordered state. This may suggest that such shape deformation is not of biological origin but rather stems from either mild tectonic or compression-related deformation.

As with the preceding traditional description of ontogeny, the findings based on geometric morphometrics indicate that alterations in the cranidium are predominantly attributable to morphological modifications in the facial suture, palpebral lobe, and glabella across the various stages of ontogeny in Kaotaia globosa. From the overlapping areas in the scatter plots of the first and second principal components, the meraspid degrees 3–8 and the holaspid period have almost no overlap, whereas the meraspid degrees 9–13 are almost completely covered by the other two stages (Fig. 6A). In the scatter plot of the first and third principal components, the stages overlap almost completely without any distinct boundaries (Fig. 6B). Setting the constraint value to 4, as required for correctly calculating the p-value of Procrustes-fitted landmark data, resulted in a multivariate analysis of variance (MANOVA) showing that the Wilks’ lambda value was 0.07359 (df1 = 68, df2 = 216, F = 8.533, p-value = 3.487E-34) and the Pillai trace value was 1.42 (df1 = 68, df2 = 218, F = 7.856, p-value = 7.174E-32). These values indicate significant differences between groups (p < 0.05).

Based on the detailed description of the ontogenetic sequence of Kaotaia globosa in the previous section, a trunk segmentation schedule of K. globosa was drawn (Fig. 7). In general, the number of meraspid pygidial segments in K. globosa showed a decreasing trend with a gradual increase in the number of thoracic segments during ontogeny, with a total of three stages of decrease. In meraspid degrees 3–4, the number of pygidial segments was five; in meraspid degrees 5–8, the number of pygidial segments was four; in meraspid degrees 9–13, the number of pygidial segments was three; and in the holaspid period, only two pygidial segments were identified. During the holaspid period, K. globosa exhibited a consistent total number of trunk segments with the last meraspid degree (meraspid degree 13), except that the exoskeletal length increased from 4.88 to 18.48 mm. Therefore, the onset of the epimorphic stage of K. globosa is earlier than that of the holaspid period, and its mode of development is protomeric (hypoprotomeric; Hughes et al. 2006).

In this study, the absolute and relative growth rates of the thoracic segments of Kaotaia globosa were calculated using the computational method proposed by Fusco et al. (2012, 2014). The average per-molt growth rate showed a gradually increasing growth gradient from the first to the 13^th thoracic segment, which was abnormally high at the 13^th thoracic segment (Fig. 8A); the allometry coefficient in relation to segment length also showed an increasing trend but slightly decreased at the 3^rd and 12^th thoracic segments (Fig. 8B). In addition, in the allometry coefficient of K. globosa, thoracic segments 1 to 7 showed a negative allometry coefficient (all. c. < 1), and thoracic segments 8 to 13 showed a positive allometry coefficient (all. c. > 1).

In two distinct mechanisms governing segmental growth, the segmental gradient hypothesis (SGH) proposes that individual body segments grow autonomously at constant rates, whereas the trunk gradient hypothesis (TGH) suggests that growth is controlled by a continuous gradient along the entire trunk (Fusco et al. 2012, 2014). To determine what controlled the growth of the segments of Kaotaia globosa and how well it fits with the two hypotheses (SGH or TGH), the growth trends of the segments of K. globosa were plotted using the RLS values for each developmental stage and the calculated values obtained by introducing TGexp, SGgmp, and SGpwp (Fig. 9).

The trend plot of segment growth in Kaotaia globosa (Fig. 9A), based on the estimated value of segment length relative to trunk length (RLS), shows that the relative segment length (RLS) gradually decreases during ontogeny, and the RLS also shows a decreasing trend from anterior to posterior during the same developmental stage. The mechanism of segment growth control is consistent with the trunk gradient hypothesis (TGH), and this specific posterior-to-anterior decaying trajectory of the growth gradient is detected in K. globosa, suggesting graded signal control in its development. To further test whether the segmental growth of K. globosa is consistent with the control mechanism of the TGH and the fit to the SGH, two growth models of the SGH (SGgmp and SGpwp) and two growth models of the TGH (TGexp and TGpwl) were fitted to K. globosa (Fig. 9B–D). Based on the predictions of the different growth models and the actual measurements of the segment data, the fits of the SGgmp, SGpwp, and TGexp models were 0.915632, 0.916207, and 0.975137, respectively, while the TGpwl model failed, and the growth curves fit the exponential function equation with the higher r-value: TGexp(x) = 0.274915*e^{-3.57564*(1-x)}.

In the hemianamorphic development exhibited by trilobites (Minelli et al. 2003), the developmental mode of an individual can be classified as hypoprotomeric, synarthromeric, or protarthrous, based on whether it enters the epimorphic phase during the meraspid period, at the transition to the holaspid period, or after the onset of the holaspid period, respectively (Hughes et al. 2006). Based on a summary of the developmental modes of 39 trilobites and an assessment of environmental context, the three aforementioned developmental modes occur with similar frequency in deeper waters; however, the protomeric (hypoprotomeric) developmental mode overwhelmingly predominates in shallow waters, and this predominance may be attributable to abiotic factors (Fusco et al. 2012). As highlighted by Hughes et al. (2006), developmental mode varies widely among trilobites. Even within the same taxonomic group, different species exhibit variation. For example, Ctenopyge (Eoctenopyge) angusta and C. gracilis follow a protomeric developmental mode, whereas C. (C.) ceciliae follows a protarthrous developmental mode. Early Cambrian corynexochid trilobites exhibit distinct developmental modes (Du et al. 2019). Similarly, among trilobites traditionally assigned to the “ptychopariid” taxa, the developmental mode of the solenopleurid Sao hirsuta has been interpreted as hypoprotomeric (Laibl et al. 2014), whereas the alokistocarid Elrathia kingii exhibits a synarthromeric developmental mode (Hopkins 2020). As the onset of the epimorphic stage in Kaotaia globosa precedes the holaspid period, its developmental mode is protomeric (hypoprotomeric). Trilobite developmental mode exhibits a significant phylogenetic signal (Fusco et al. 2012), and this diversity may, to some extent, reflect the phylogenetic polyphyly observed in trilobites.

In the ontogenetic sequence of the redlichiid Bathynotus kueichouensis, the number of pygidial segments during meraspid degrees 1–11 is three (excluding the terminal piece; the same applies below), but subsequently decreases in the late meraspid degrees (M12, 13) to two, the number maintained in the holaspid period (Zhang et al. 2022; Dai et al. 2023). In contrast, the number of pygidial segments in Zhangshania typica was two from meraspid degrees 2–13, with a gradual increase in the number of pygidial segments observed after entering the holaspid period; it was thus protarthrous in its development (Hou et al. 2017). The reduction in the number of meraspid pygidial segments observed in Kaotaia globosa differs from that seen in B. kueichouensis. In K. globosa, a three-stage decrease was evident throughout the ontogenetic sequence, whereas in B. kueichouensis, this decrease was concentrated in the late meraspid degrees. In Z. typica, the number of pygidial segments continued to increase even after the thoracic segments stabilized. Conversely, in K. globosa, the number of pygidial segments remains unchanged following attainment of the holaspid period.

The body proportions of trilobites vary among different species. Based on a growth model generated from extensive body size data for Aulacopleura koninckii, Hopkins (2020) suggested that depressing the growth gradient during the holaspid period yields body proportions that result in micropygous trilobites. In all or most redlichiid taxa with a micropygous body pattern, the relative size of the meraspid pygidium is maintained by producing and releasing segments in a 1:1 ratio (Dai et al. 2023). The results presented above indicate that the formation and addition of K. globosa segments do not occur on a one-to-one basis. During the transition from meraspid degree 4 to 5, the number of trunk segments does not increase; instead, one segment is released from the pygidium into the thorax. Similarly, the number of trunk segments remains unchanged during the transition from meraspid degree 8 to 9 and from meraspid degree 13 to the holaspid period (Fig. 7). This ultimately reduces the number of pygidial segments of K. globosa from five in the early meraspid period to two in the holaspid period. As the number of meraspid pygidial segments decreases, the pygidial relative width of K. globosa gradually decreases from slightly less than the width of the cephalon (tr.) in meraspid degree 3 to a width that is equal to or less than the width of the base of the glabella (tr.) in the holaspid period.

Growth gradients are reflected in differences in growth rates between organs or parts along the axis of an organism (Huxley 1932; Fusco et al. 2014), and studying the growth gradient of trilobite segments can help reveal patterns of segmental growth and development, as well as potential control mechanisms. In this study, the growth gradient of Kaotaia globosa segments was analyzed through the ontogenetic measurement data from meraspid degree 3 to the holaspid period. The calculated absolute and relative growth rates indicate that K. globosa also exhibits an increasing growth gradient toward the posterior part of the trunk (Fig. 8). The absolute growth rate of the trunk segment of K. globosa demonstrates a pronounced increase in the 13^th segment (Fig. 8A), and this phenomenon is also observed in the 16^th segment of Aulacopleura koninckii and Changaspis elongata (Fusco et al. 2014; Du et al. 2019).

In this paper, the absolute growth rate of trunk segments in Kaotaia globosa is expressed as the antilogarithm of the calculated value of the average incremental growth rate per molt. Notably, holaspid specimens constitute a disproportionately large fraction of the sample (70 of 145), far exceeding any meraspid degree. This prompted the question of whether the apparent increase in absolute growth rate could be correlated with sample-size disparity across stages. To investigate this, the holaspid period data were randomly subsampled, retaining only 20 specimens. Recalculation showed that the absolute growth rate of the 13^th thoracic segment remained significantly elevated. During the holaspid period of K. globosa, the smallest specimen had an exoskeletal length of 4.88 mm, whereas the largest specimen reached 18.48 mm, an almost fourfold difference. Taking this into account, the data were reanalyzed after removing the measurements of larger holaspid period specimens, finding that the absolute growth rate curve tended to rise steadily. Meanwhile, the absolute growth rate of the 13^th thoracic segment declined from 1.69 to 1.19 after adjustment, comparable to that of the 12^th segment. To test the generality of this relationship between absolute growth rate and body length range within the holaspid period, the segmental data for Aulacopleura koninckii published by Fusco et al. (2014) were reanalyzed. After excluding holaspid period specimens exceeding 7 mm in length, the absolute growth rate of the 16^th segment decreased from 1.21 to 1.05 (see Suppl. material 2 for details). This result is consistent with the pattern observed in K. globosa. Therefore, it is hypothesized that the abrupt increase in the absolute growth rate of the 13^th thoracic segment in K. globosa is attributable to the substantially larger range of exoskeleton length observed during the holaspid period than in any of the meraspid stages.

Although collection biases are recognized, the number of Kaotaia globosa specimens from the early meraspid degree analyzed in this paper is significantly lower than that from the late meraspid degree and the holaspids—a pattern also observed in other trilobite ontogenetic studies (e.g., Fusco et al. 2014; Hughes et al. 2017; Du et al. 2019). Fortunately, Dai et al. (2021a) provided a wealth of body length data for early meraspid degree specimens of Oryctocarella duyunensis, which made it possible to assess the impact of morphometric analysis when the number of specimens is limited. After artificially and randomly reducing the early meraspid degree data of O. duyunensis, changes in the growth gradients and the estimated values of segmental growth trends under different model fits were only minimally affected. High-quality, abundant ontogenetic series can minimize the impact on descriptive statistics and hypothesis testing (Fusco et al. 2012). However, given the scope of the current work, this requirement has not yet been met, and further work in this area is necessary.

A recent study suggested that changes in the body size of Cambrian–Ordovician trilobites were closely related to marine redox fluctuations and that larger-bodied animals may have been more susceptible to impact during biological crises (Sun et al. 2025). Trilobites achieved growth through molting behavior (Daley and Drage 2016; Drage et al. 2018); the presence of larger individuals in the holaspid period suggests that molting behavior continues throughout their life cycle (Henningsmoen 1975). Unfortunately, complete ontogenies have only been reported for a few species. Most of these show a significant difference in exoskeletal length between the smallest and largest holaspid stages (Fig. 10). Some individuals of Elrathia kingii experienced a 7.5-fold increase in exoskeletal length during the holaspid period, and assuming that the cephalic growth rate remains constant, E. kingii would need to undergo at least 28 molts to achieve its maximum body length (Hopkins 2021). Since the absolute growth rate of the 13^th thoracic segment is greatly influenced by the size range of the holaspid period (see above), and assuming that the rate remains constant throughout subsequent ontogeny, the absolute growth rate of the 12^th segment (1.34) was used to estimate the minimum number of molts required for K. globosa to achieve its maximum exoskeletal size. The results show that K. globosa would have experienced at least five molts from the smallest holaspid specimen (4.88 mm) to the largest holaspid specimen (18.48 mm) when growing at this rate. Although growth rates vary among different species, trilobites with greater ranges in exoskeletal length will, in general, require more molting cycles to attain their maximum size.

Here, the holaspid exoskeletal lengths of articulated trilobites across five orders were analyzed using previously published data (Fig. 10). This result cannot fully represent the exact growth of a particular trilobite because of factors such as the uncertainty of fossil specimen preservation, differences in the degree of research, and survivor bias. Nevertheless, this compilation shows that most holaspid trilobites encompass a wide range of sizes. Based on substantial differences observed in size distribution between the pre-holaspid and holaspid periods, Dai et al. (2023) proposed that the holaspid period constitutes the majority of the life cycle of certain trilobite species. The gradually increasing anterior-to-posterior growth gradient observed in Kaotaia globosa and several other species appears to confirm the hypothesis proposed by Dai et al. (2023). In other words, the increasing growth gradient enables the number of trunk segments to reach a stable state within a relatively short period (i.e., shorter anamorphic development), after which prolonged epimorphic development continues to increase the body size of trilobites.

Arthropod phylogeny suggests that sexual dimorphism was probably widespread among Trilobita (Hughes and Fortey 1995). However, the specific features that effectively distinguish dimorphism vary to some extent among different trilobites. Several studies have proposed various potential indicators, including numerical ratio, relative size, and exoskeletal ornamentation (Hu 1971); the number and morphology of macropleural segments (Gozalo et al. 2003); the spherical enrolled shape (Cederström et al. 2009); and the presence or absence of pygidial margin spines (Lei et al. 2023). Specifically, Fortey and Hughes (1998) proposed that a spherical bulge on the preglabellar field in several trilobites functioned as a brood pouch. Interestingly, features reminiscent of some proposed dimorphic traits are observed in two closely related trilobites from the Kaili Formation. Kaotaia globosa and K. magna are highly similar, with one of the primary distinctions being a more pronounced preglabellar bulge in K. globosa (Zhang et al. 1980). Noting that K. globosa possesses a preglabellar bulge almost as prominent as that of Parabolinella bolbifrons, Fortey and Hughes (1998) proposed that K. globosa and K. magna may represent manifestations of sexual dimorphism. Although they share the same stratigraphic range in the Miaobanpo section, their recorded frequencies differ considerably (approximately 7:1) based on existing collections. Moreover, there is no evidence that trunk segment stabilization in trilobites is linked to sexual maturity (Hughes et al. 2006). Critically, the preglabellar bulge in K. globosa is clearly present from as early as meraspid degree 8—five instars before the holaspid phase. Taken together, these observations suggest that sexual maturity at this early developmental stage is highly unlikely; instead, the preglabellar bulge in K. globosa is more parsimoniously interpreted as a species-specific ontogenetic feature rather than a reproductive adaptation.