A recent description and analysis of an imprint fossil from the Carboniferous concluded that it was made by a mayfly landing in sediment at the edge of water. Here, I reanalyze that trace fossil and supply experimental evidence regarding wing traces and behavior. The thorax of the trace maker lacked structures characteristic of mayflies, but closely matches a modern neopteran insect family (Taeniopterygidae, Plecoptera) little changed from Early Permian fossils. Edges of the folded wings of live Taeniopteryx leave marks on sediment closely matching marks in the trace fossil. Faint marks lateral to and beyond the reach of meso- and metathoracic legs match the location where wings of surface-skimming Taeniopteryx stoneflies lightly touch the sediment when these insects skim onto wet ground at shorelines. Dimensions of the thorax of the trace indicate relatively weak flight ability compared to fossils from the Early Permian, making doubtful the hypothesis that the trace maker was flight capable. Ultimately, this fossil best fits a scenario in which a neopteran insect skimmed across the surface of water, then folded its wings. Surface skimming as a precursor to the evolution of flight in insects is supported by this fossil evidence of skimming behavior in a Carboniferous insect.

Winged insects (Pterygota) diverged and radiated during the Carboniferous (360–300 mya), yet insect fossils from that period are sparse and consist primarily of wings or wing fragments (Grimaldi and Engel 2005). A remarkably detailed full-body imprint fossil of a pterygote insect was recently discovered from the Late Carboniferous (Knecht et al. 2011) and provides rare information about morphology, habitat, and behavior, but correct interpretation of the imprint features and its maker's taxonomic identity are critical. Knecht et al. concluded that the imprint was most likely produced by a lineage that gave rise to mayflies (Ephemeropterida). Here, I critique and reinterpret their data and reconstruction, and provide experimental evidence that the trace fossil contains marks left by the wings of a neopteran insect. The results change conclusions not only about the taxonomy, but also the behavior of the trace maker.

This fossil imprint (University of Kansas Natural History Museum, SEMC-F97) was made in saturated, water-rippled alluvial sediment at the edge of fresh water. Ultimately, these sediments became fine-grained sandstone of the Wamsutta Formation (Late Carboniferous, Westphalian B-C; 308–314 mya) of southeastern USA, MA. None of this critique challenges the geological analysis or interpretation of the habitat. A full description of the fossil and its setting is contained in the original publication (Knecht et al. 2011).

Materials and Methods

This reinterpretation is based on published images and text in the original report (Knecht et al. 2011), and a high-resolution unpublished photograph of the fossil. No features of the fossil were considered beyond what was presented in the original report. New data were provided by experiments performed using live Taeniopteryx sp. (possibly T. burksi, nebulosa, or another of a group of highly similar species) collected as adults from Bald Eagle Creek, Centre County, PA in February and March of 2011 and 2012. These stoneflies were used to create shallow impressions in fine-grained kaolinite sediment, and to film their behavior skimming across a water surface onto rippled, saturated sediment. Impressions were made by allowing live stoneflies to walk freely over wet kaolinite, or by gluing a pin to the dorsal thorax of a stonefly and pressing it lightly into kaolinite to the approximate depth of the trace fossil. Photos of stoneflies and of experimental impressions were taken with a digital camera attached to a stereomicroscope. Video sequences were filmed at normal speed using a Sony TRV9 camera and at 240 frames per second using a Casio Exilim EX-ZR100 camera.


A first step toward taxonomic identification of the SEMC-F97 trace maker requires consideration of the head and terminus of the abdomen, locations of key features for distinguishing orders of insects. Mayflies and dragonflies (Odonata) have tiny antennae unlikely to leave a discernable trace, whereas early Neoptera typically had long, segmented antennae. In this trace fossil, only the ventral rear of the head left an impression, and anterior to the head is a disturbance obscuring the sediment surface. Regardless, live insects tend to hold their antennae above the ground in a position unlikely to leave marks. Knecht et al. concluded that long antennae were absent, but the fossil provides no clear evidence for either presence or absence of long antennae.

Mayflies typically have three filaments at the terminus of the abdomen whereas insects such as stoneflies (Plecoptera) have only two. In the trace fossil, Knecht et al. concluded that “two narrow impressions lie in curved projection from the terminus of the final tapered segment, representing the drag marks of some of the terminal structures, either both cerci or one cercus and the median caudal filament.” This unambiguous description of two impressions extending from the abdomen terminus contrasts with their Figure 2C legend, which claimed that “basal articulations of the three structures are discernible along the margin of the last segment.” How two impressions in the primary description became three basal articulations in the figure legend is not clear, but lacking definitive evidence of three filaments, the abdomen terminus provides no reason to identify the trace maker as a mayfly.

Figure 2.

Side view of an Allocapnia stonefly using its wings to sail across water. Note the weakly developed thorax, not wider than the abdomen. This species has small flight muscles and cannot fly (Marden and Kramer 1995).

Much of the remaining taxonomic interpretation in Knecht et al. was based on a process of elimination. Paleodictyoptera were eliminated by the absence of a ventrally folded rostrum, which would have been clearly visible in this well-resolved impression. Other eliminations of taxa were less well justified, particularly their consideration of Plecoptera, which they incorrectly described as “not known or predicted to extend deep into the Carboniferous.” The earliest fossil stonefly is from the Carboniferous (Bethoux et al. 2011) and a diverse group of families are represented in Early Permian fossils (reviewed in Illies 1965), hence indicating radiation prior to the Permian and disproving the claim (Grimaldi and Engel 2005) that Plecoptera is not a particularly old order of insects. Knecht et al. referred to “basal” and “primitive” Plecoptera without citing a source, and used assumptions about basal plecopteran morphology not traceable to a rooted phylogeny or to fossil Plecoptera. By this method, they excluded Plecoptera based on the trace's contiguous rather than broadly separated leg bases (coxae), a prothorax not wider than other body segments, and absence of relatively short forelegs. As I show below, these exclusions of Plecoptera are baseless.

Plecoptera fossils from the Early Permian (Illies 1965) include Paleotaeniopteryx (Fig. 1D). Modern members of this family (Taeniopteryx; Taeniopterigidae) have ventral morphology strikingly similar to the trace fossil (Fig. 1A, B), closely matching the overall body shape, posture, position of the leg bases, relative breadth of the prothorax, and ventral sclerites of all thorax segments. Notably, the trace fossil does not possess the ventral thorax morphology of mayflies (Fig. 1C, showing Siphlonurus, thought to have the most primitive morphology of modern Ephemeroptera; Matsuda 1956), which have a mesothoracic furcasternum (F2) that is large and posterior to the mesothoracic leg bases, and a fusion of basisternum 3 and furcasternum 3 (B3+F3). All modern Ephemeroptera have an enlarged and strongly convex F2 sclerite, the ventral attachment point of the subalar–sternal muscles, direct wing depressors absent in Neoptera (Kluge 1994). Lack of ephemeropteran F2 and B3+F3 morphology in the imprint fossil indicates that its maker was not a mayfly, or that the sternal morphology and arrangement of flight muscles in modern mayflies diverged radically subsequent to the time of this imprint fossil.

Figure 1.

Comparison of thorax ventral morphology in stoneflies and mayflies with the SEMC-F97 trace fossil (Knecht et al. 2011). (A, B) Ventral morphology of an extant neopteran insect (Plecoptera, Taeniopteryx) and the trace fossil (scale bars: A, 5 mm; B, 10 mm). Highlighted are sternal sclerites of Taeniopteryx, which, with minimal changes in size (primarily width) and relative position, match the fossil. Poststernum 1, basisternum 3, and furcasternum 3 are clearly defined in the fossil. (C) Diagram (adapted from Matsuda 1956) of the ventral pterothorax of the mayfly Siphlonurus (Ephemeroptera). Note how the mayfly furcasternum 2 (F2) is enlarged and posterior to the mesothoracic leg bases (L2), and that basisternum 3 and furcasternum 3 are fused (B3+F3). This ventral morphology is a shared derived feature in Ephemeroptera, in which F2 provides the ventral attachment point of the subalar–sternal muscles, direct wing depressors absent in Neoptera (Kluge 1994). (D, E) Forewing of Paleotaeniopteryx (top, scale bar 1 mm; adapted from Illies 1965) and modern Taeniopteryx (lower, scale bar 3 mm; adapted from Béthoux 2005). This lineage has undergone little change in wing design over 260 million years.

Flight ability of all types of insects is a function of the mass of the flight muscles relative to total body weight (Marden 1987). Flight muscles occupy most of the interior volume of the meso and metathorax, and in situations in which it is not possible to dissect and weigh the body segments, relative dimensions of these thoracic segments and the abdomen provide a metric strongly correlated with flight ability (Srygley and Chai 1990). In the SEMC-F97 trace fossil, the width of the meso and metathorax does not exceed that of the abdomen (both 3.8 mm; Knecht et al. 2011). Even in Taeniopteryx, a marginally flight-capable insect (Marden and Kramer 1994), the thorax is broader than the abdomen (Fig. 1A; ratio = 1.26). Across a taxonomically broad sample of early fossil pterygotes (Early Permian; Table S1, N= 14), the highest ratio of thorax to abdomen breadth (1.59) occurs in a mayfly, Protereisma, whereas only one species has a ratio less than one. Modern insects in which the thorax is not wider than the abdomen tend to be small and wind-blown (e.g., Thysanoptera), have abdomens swollen with eggs (e.g., temporarily flightless female mantids), or have an abdomen mass reduced by dorsoventral flattening (e.g., Hemiptera, many Coleoptera), large abdominal air sacs, or a narrow waist and terminus (many Hymenoptera). Among insects sharing the generalized cylindrical body design and habitat of the SEMC-F97 trace fossil, a thorax not wider than the abdomen occurs in flightless Allocapnia stoneflies (Plecoptera) that sail across the surface of water until they encounter shorelines (Fig. 2; Marden and Kramer 1995).

The canonical difference between paleopteran (including Ephemeroptera) and neopteran insects is the ability of Neoptera to fold their wings parallel to and just above the long axis of their body (a trait independently evolved in the extinct Diaphanopterodea; Kukalova-Peck and Brauckmann 1990). In many Neoptera, the folded wings curve around (e.g., Plecoptera: Leuctra and Taeniopteryx;Fig. 3A) or slant downward (e.g., Megaloptera: Sialis) along the sides of the body, with the leading edges lightly touching the ground surface when the insect is at rest. Marks likely to have been left by the edges of wings held in this fashion are present in the SEMC-F97 trace fossil but were not interpreted as such by Knecht et al. The marks in question comprise pairs of closely spaced lines running parallel to and equidistant from the sides of the body (Fig. 3B, E). Knecht et al. attributed these marks to scratches left by spines projecting from the posterior femora that dragged as the animal came to rest in wet mud while moving in a posterior-to-anterior direction. However, this path of motion is contradicted by their conclusion that the insect moved laterally, as indicated by faint marks beyond the farthest reach of the fully extended legs (the linear green arrows in Knecht et al.'s Fig. 3). Clearly the animal could not have moved simultaneously in two different directions, nor could these marks have been made by leg motions after the insect had settled, as such traces would follow arcs rather than lines parallel to the body axis.

Figure 3.

Evidence for marks made by the edges of folded neopteran wings in the SEMC-F97 trace fossil. (A) Ventral view of a live Taeniopteryx stonefly. Arrows show the fore and hind wing leading edges, which sit flush against the ground when the wings are folded. Inset shows close-up of the paired wing edges. (B) Wing edge marks (at arrowheads) beside the abdomen of the SEMC-F97 trace fossil (Knecht et al. 2011). (C, D) Traces of wing edges of Taeniopteryx stoneflies in experimental impressions in kaolinite. (E) Close-up of the two fine parallel marks (at arrowheads) beside the abdomen of the SEMC-F97 trace fossil. Scale bars: A, 5 mm (inset 0.1 mm); B, 10 mm; C, 5 mm; D, 0.2 mm; E, 0.4 mm.

To test the hypothesis that folded neopteran wings can leave marks on the sediment surface, I experimented with live insects. When Taeniopteryx stoneflies walked across a fine wet kaolinite sediment, they left wing edge traces adjacent and parallel to the abdomen (Fig. 4). Similarly, resting Taeniopteryx stoneflies with folded wings (Fig. 3A), pressed lightly into kaolinite to a depth similar to the SEMC-F97 trace fossil, left parallel lines where the edges of the overlapping fore and hind wings contacted the sediment surface, strikingly similar to the marks present in the trace fossil (Fig. 3C, D). Distance between these paired lines was 0.1 mm in Taeniopteryx (Fig. 3D) compared to 0.18 mm in the trace fossil (Fig. 3E), consistent with the twofold difference in body length. Across experimental replicates, marks left by edges of folded Taeniopteryx wings were variable in their presence, absence, and number at all locations along the body, including between the meso- and metathoracic legs, caused by variations in the lateral tilt of the body, irregularities of the sediment surface, and the movement and wing posture of individual insects.

Figure 4.

Wing trace made by a Taeniopteryx stonefly walking on wet kaolinite. Walking direction was from bottom to top. Angled lateral marks were made by the legs; a broad band was left by the abdomen (right edge at arrow). The line parallel to the abdomen was left by the edges of the folded wings.

Additional faint marks on the sediment surface lateral to the right legs of the trace fossil provide additional evidence regarding the behavior of the trace maker. As described by Knecht et al., these marks are outside the reach of the limb impressions of all three right legs. Their description however conflicts with their reconstruction (reproduced here in Fig. 5), which shows marks beyond the reach of only the meso and metathoracic legs, but not the forelegs. These distal marks are considerably shallower and less distinct than the deeper and more proximal impressions clearly made by legs. Given the setting of the trace fossil, these distal lateral marks could have been made by beating wings of an insect like Taeniopteryx as it skimmed across the surface of water and came ashore onto saturated sediment. A photo of a skimming Taeniopteryx superimposed onto Knecht et al.'s reconstruction (Fig. 5) shows that the tips of the fore and hind wings, at maximum downstroke, align with these distal lateral marks. Furthermore, it is easy to recreate this scenario experimentally with Taeniopteryx skimming across water onto mud. Upon reaching the shore, the stoneflies often continued wing-powered skimming across a short distance of saturated sediment, and continued flapping their wings until they were unable to proceed farther, only then folding their wings and beginning to move their legs (see Supporting Information Video S1 online). This behavior is a good fit for the setting and events evidenced in the trace fossil. Also clear in this video is that the stoneflies folded their wings in a raised position above their dorsum, with no portion of the wings touching the ground during the wing folding maneuver. Only when the wings were fully folded and parallel to the body axis were they lowered to a point where the wing edges contacted the ground. Hence, no arced traces made by fold wings, rotating from the wing bases, are expected or observed in the trace fossil. Considered at a population level, surface-skimming insects repeatedly arrive at shorelines onto soft sediment, consistently recreating the conditions necessary to create the SEMC-F97 trace fossil.

Figure 5.

Comparison of striations distal to the legs in the SEMC-F97 trace fossil with location of wingtips of a surface skimming stonefly. Left: Interpretation of Knecht et al. (2011) requiring lateral movement of the entire body to leave striations beyond the reach of the meso- and metathoracic legs while also moving forward to leave the linear marks parallel to the abdomen (i.e., two contradictory paths). Right: Dorsal view of a Taeniopteryx stonefly skimming on the surface of water or wet sediment (see Supporting Information Video S1 online). Tips of the fore and hind wings contact the surface at locations consistent with the distal lateral marks in the trace fossil.


Knecht et al. concluded that the maker of the trace fossil flew and landed at a shoreline location and rejected the hypothesis that the insect skimmed across the water surface. The only evidence to support the flying hypothesis is absence of a walking track leading to the body impression, but insects commonly disembark from floating objects, jump from rocks or stems above the ground, or skim across water without leaving marks. The body impression in the sediment was not likely to have been caused by impact after flight, because the lightweight bodies of small insects do not displace by impact a substantial volume of heavy wet sediment held together by surface tension. A more physically realistic scenario is that the maker of the trace skimmed across the wet sediment surface, leaving minimal marks, until it encountered a patch of sticky, possibly organic-laden mud in which it became temporarily stuck. Struggling and adhesion, as opposed to impact, were most likely the forces responsible for creating the impression. How the insect extracted itself and departed without leaving an exit track is unclear, but insects commonly respond to obstacles or impedance by reaching upward and lifting themselves onto overhanging objects.

Knecht et al.'s rejection of surface-skimming as a plausible behavior for creation of the imprint was based on a set of false arguments rather than evidence from the fossil. Surface-skimming locomotion by adult aquatic insects was first described in Plecoptera (Marden and Kramer 1994), accompanied by mechanical, physiological, and morphological data indicating that skimming could have been an evolutionary forerunner to powered flight. Almost immediately, the flight-from-skimming hypothesis was challenged on the grounds that the phylogenetic placement of skimming indicated evolutionary novelty (Will 1995). However, no other taxa or fossils had yet been examined for presence/absence of the trait, and hence a phylogenetic analysis was premature at that time. Subsequent studies revealed that surface-skimming locomotion is widespread across modern Plecoptera families (Marden et al. 2000; Marden and Thomas 2003), appears to be an ancestral trait within the order (Thomas et al. 2000), and occurs sporadically in modern mayflies (Ruffieux et al. 1998; Marden et al. 2000) and damselflies (Samways 1996). Wing structures important for aerial flight became more elaborate and diverse during the radiation of modern stoneflies (Thomas et al. 2000), consistent with weaker flight in early Plecoptera and the weakly developed thorax of the SEMC-F97 trace fossil. Similarly, wings and thoraces of fossil mayflies (Carboniferous and Permian) indicate reduced flight ability compared to modern species (Wootton and Kukalova-Peck 2000). These results agree with a model in which flight originated on water, a setting in which the body weight is supported and only a small amount of thrust is required for locomotion, followed by a transition to three-dimensional flight requiring a more elaborate flight motor and wings. Developmental studies have found support for the wings-from-gills hypothesis for wing origins (Cohen et al. 1993; Averof and Cohen 1997; Franch-Marro et al. 2006), and molecular phylogenetics indicate that hexapods evolved from aquatic, crustacean ancestors (Regier et al. 2010). Although the proximal antecedents of pterygote insects were fully terrestrial (e.g., Protura, Archaeognatha, Zygentoma), each of these groups may have lost gills early in their divergence from an aquatic stem group that retained gills and an aquatic habit all the way through to the emergence of pterygotes. Clearly, evidence supporting a possible aquatic and surface-skimming origin for insect flight has grown considerably since the original publication of the hypothesis, yet inexplicably Knecht et al. cited the premature and unrooted 1995 phylogenetic analysis by Will as authoritative and concluded that “support for the skimming hypothesis is entirely ad hoc.”

Knecht et al. extended their arguments against the skimming hypothesis by stating there is no evidence for aquatic forms in any life stage of early Odonata, Ephemeroptera, and Plecoptera. This is a curious stance because flappable abdominal gills, commonly thought to be serial homologs of wings (Kukalova-Peck 1978, 2008; Kluge 2004) are present in Ephemeroptera (many taxa), Plecoptera (Eustheniidae, Diamphipnoidae; Marden and Thomas 2003), the extinct fossil order Coxoplectoptera, a sister group of mayflies (Staniczek et al. 2011) and the aquatic immatures of fossil (Kukalova-Peck 2009), and some modern Odonata (Polythoridae; Cora marina; (Pritchard 1996) here shown in video that is the first report of segmental gill flapping in Odonates: see Supporting Information Video S2 online). Flappable gills are complex structures possessing muscles, nerves, articulation, and fluid dynamic function, unlikely to have arisen independently and would be rapidly and independently lost in terrestrial species. An argument that gills are not homologous across aquatic insect orders (Grimaldi and Engel 2005) refers to filamentous and other types of nonflappable gills that occur in various body locations, but these derived structures are irrelevant to the homology of the segmental flapping gills thought to have given rise to wings.

In conclusion, the original description of the SEMC-F97 trace fossil and interpretations in that report do not withstand a critical analysis. Rather than providing evidence of the landing position of a flying mayfly, this imprint fossil was more likely made by a weakly flying or flight-incapable neopteran insect. The setting in wet rippled shoreline sediment, along with a weakly developed thorax and marks where wingtips of a skimmer would have repeatedly struck the ground, indicate that the trace maker may have skimmed across water onto wet ground before becoming stuck and folding its wings into a resting position. Such behaviors are readily observable in certain extant insects like Taeniopteryx and are a predicted transitional stage during the evolution of three-dimensional flight in pterygote insects (Marden and Kramer 1994; Marden 2008). Ultimately, this trace fossil provides evidence that surface-skimming locomotion has a deep history in neopteran insects and perhaps insects in general, thereby strengthening the likelihood that skimming across water was an intermediate stage in the evolution of flight.

Associate Editor: D. Carrier


C. Marone generously donated the kaolinite used for making imprints. J. Benner shared a high-resolution photograph of the fossil. R. Schilder helped with videos of surface-skimming and he, T. Carlo, and two anonymous reviewers provided helpful comments on the manuscript. This work was supported by NSF IOS-0950416.