Benner, Knecht, and Engel have replied to my critique of their interpretation of a Carboniferous trace fossil produced by an insect at the edge of water. Here I respond by pointing out that their reiterated scenario still requires mutually exclusive paths of motion and I show that their assertions of methodological shortcomings are tangential and lack merit. Overall, this discussion provides an opportunity to examine in greater detail competing hypothesis about behaviors and taxonomic identity of the trace maker, and relevance thereof to competing theories regarding early events in the evolution of pterygote insects.

The reply of Benner, Knecht, and Engel (hereafter BKE) to my critique (Marden 2013) of their paper (Knecht et al. 2011; hereafter KEB) describing the earliest trace fossil of a winged insect consists primarily of accusations of methodological failings on my part, but our differences have nothing to do with methods. Most fundamentally, my critique pointed out three likely errors in their interpretation (KEB) of the trace fossil. First, they identified the maker as a mayfly, but the trace contains thorax and wing characters indicative of a neopteran insect. Second, they proposed a series of motions requiring three approximately orthogonal and self-contradictory paths: downward impact by a flying insect, forward motion while skimming the surface and leaving leg-derived marks alongside the body, and lateral motion indicated by marks beyond the reach of the legs on one side of the body. Third, they overlooked nearly all of the literature on skimming locomotion, evidence for shared aquatic habits in basal pterygotes, and the possible relevance of these features to the origin of insect flight. Nowhere in their reply (BKE) do they defend their identification of the trace maker as a mayfly or their original hypothesis about vertical impact causing the body imprint, and they still do not acknowledge the self-contradictory paths of motion inherent in their interpretation. In their concluding remarks, they elaborate what I maintain is an unsupported interpretation about the evolutionary history of key insect traits.

Primarily, BKE attempt to discredit my methods. Foremost among their concerns is that I worked only from their excellent photographs of the imprint and they object to the way I tested and illustrated hypotheses using live insects. Working from published figures has ample precedent. For example, Carpenter (1979) criticized a published account of a fossil mayfly stating, “I have not seen that specimen …… but if the costal brace is formed as shown … the insect could not even be assigned to ….” As in any field, one can judge fundamental errors of logic and interpretation without observing the original material. My critique stated explicitly that it was “based on published images and text in the original report” and that “no features of the fossil were considered beyond what was presented in the original report.” Given that I worked from their own evidence, protestations about methodological shortcomings ring hollow.

In the sections below, I respond individually to their specific concerns.


BKE reproduce (their Fig. 1) images from KEB of marks alongside the body and repeat their hypothesis that flexible spines (striae) on the legs produced these marks during forward motion across the surface of wet sediment. Their primary contention is that marks left by folded wings should number only two and should not show subtle curvature or variation in spacing. The expectation that wings would leave only two edge traces applies to a dead or still insect, but the trace was left by a live insect and the leg impressions indicate struggling. Even minor lateral rocking motions could result in multiple and slightly displaced wing edge contacts on the sediment. Figure 3A in my critique shows gentle curvature and irregularities in spacing between the wing edges of a live insect. Hence, the new remarks in BKE do not refute the experimentally supported hypothesis that such marks can be produced by folded wings of a neopteran insect. Furthermore, these marks in the fossil are not arced or obscured in the areas immediately around the leg bases (BKE Fig. 1A), as would result during struggling if made by leg features.

KEB's and BKE's interpretation that these marks were made while some portion of the legs remained in steady contact with the sediment during wing-propelled forward motion is, by definition, surface-skimming locomotion. If made by an ancestor of modern mayflies, this would serve the surface-skimming theory for insect flight origins (Marden and Kramer 1994; Marden et al. 2000; Marden 2008) much better than the scenario I proposed in my reanalysis of the trace fossil, as it would provide direct evidence for skimming much nearer to the base of the pterygote phylogeny. This contradicts BKE's claim that my interpretation reflects bias in favor of the skimming hypothesis.


The trace fossil contains numerous clear imprints of the legs, sufficient to discern the approximate reach and arc of each leg, but my disagreement with KEB and BKE concerns only the subset of marks beyond the apparent reach of the legs. All of the marks around the foreleg are within its reach without hypothesizing any lateral movement of the body. The same is not true for the other thoracic segments, for which there are many shallow and unjointed impressions located well beyond the maximum lateral reach of the middle and hind legs. As pointed out above, lateral motion of the body is contradictory to the original and now reiterated (BKE objection 1) interpretation of posterior-to-anterior motion claimed to have produced marks alongside the body. The conundrum is solved if the distal lateral marks were made by the wings of a surface skimming insect attempting to continue its wing-flapping motion after coming ashore onto wet sediment, a scenario consistent with the water's edge location of the trace fossil and video I produced of stoneflies continuing to skim on wet mud after exiting water (a behavior previously undocumented, contrary to BKE's statement that this observation “proved nothing new”).

KEB and BKE claim that these distal marks show “clearly discernible tarsal impressions.” They cite an “elongate shape” with the “same width” and “bulging impression” as the undisputed tarsi of the more proximal leg imprints. I was unimpressed with this identification in their original paper and remain so after examination of the new figure included in their reply. These marks do not show joints or a claw and are much more uniform in depth (all shallow) and orientation (nearly parallel) than proximal marks clearly made by the legs. Their new photo includes only the marks distal to the middle leg, whereas those distal to the hind leg (corresponding to the typically lighter and weaker hindwing) appear much fainter with even less resemblance to legs or tarsi.

BKE criticize the way I overlaid an image of a surface skimming stonefly onto their drawing to show the correspondence of flapping wings and the distal lateral marks. In making this figure, I placed the stonefly at an angle that did not obscure relevant features of the fossil. Rotating and scaling the stonefly image so that the body matches exactly the body length and angle of the trace fossil yields the new image shown in Supplementary Figure S1. The fit with all features of the trace fossil is remarkable. BKE state that removal of their drawn prothoracic leg marks from part of my figure (done to focus attention on the distal marks in question) was a misrepresentation, but they do not mention that the same figure included their full interpretation as a separate panel.

BKE include a new analysis of the convergence of lines distal to the hind leg and argue that this indicates a basal articulation not consistent with a wing, but the lines they draw do not fit closely the marks in their own figure, and they do not consider that wings are not rigid structures. My photo of a skimmer in action (included in their figure just below the convergence analysis) reveals a large bend in the left forewing. It is difficult to predict how the dynamic shape of beating wings would affect the layout of wing marks on a soil surface left by a forward-moving insect, but suffice to say that the apparent convergence shown in BKE does not effectively refute that these may be wing marks.

Finally, if these distal marks in the trace fossil were made by legs as the insect moved laterally, where are the corresponding marks made by the forelegs and the body as they too moved laterally? The remarkable feature of this trace fossil is its detail, so lateral movement would likely have been recorded for more than just two of the legs.


BKE claim that I misrepresented and obscured sternal plates present in the imprint fossil by using an overlay, but they do not mention that another figure in my critique shows the same figure without the overlay. They further claim that these sternal elements are not sufficiently discernible to contribute to a taxonomic interpretation, but as detailed in my critique, a subset of the sternal plates are clearly discernible: poststernum 1 (which KEB describe in detail, even giving its dimensions), basisternum 3, and furcasternum 3. The latter two are fused in mayflies and very different from the morphology evident in the trace fossil. BKE's new photograph provides more detail and reveals lateral margins somewhat different than my overlay, but the fundamental correspondence with the size, number, and arrangement of sternal plates of Plecoptera is striking. Readers can make their own conclusions.


BKE claim that the imprint is not deep enough to reveal the full width of the thorax, but the argument addresses the relative size of the thorax and abdomen, each equally impressed in this fossil. Ironically, they used relative thorax width in their own analysis (KEB: “the prothoracic segment is narrower than the head”). More importantly, they overlook the fact that ventral thoracic plates are attachment points indicative of the cross-sectional area of dorsoventral muscles that power the wing stroke (e.g., Fig. 1A in Marden et al., 2001). Muscle cross-sectional area limits maximum contractile force (Marden and Allen 2002; Marden 2005), and so regardless of the other dimensions of the thorax (including its height, which varies according to the cross-sectional area of dorsal longitudinal muscles; e.g., Fig. 1 in Marden 2005), the strikingly small ventral plates revealed by the imprint indicate small muscles and probably a marginally flight-capable or nonflying insect. They criticize my use of fossil reconstructions to derive a distribution of relative thorax width in fossil insects and call the deceased Tillyard fanciful, yet, a photo of a complete specimen of a Permian mayfly (Misthodotes obtusus Sellards 1907) confirms a thorax substantially broader than its abdomen (Tillyard 1932). Here, I need to acknowledge and correct an error in my critique: I cited Chai and Srygley (1990) as showing that relative thorax width relates strongly to flight ability, but the correct citation is Srygley and Chai (1990). BKE argue that relative thorax size in the butterflies used in those species correlates with flight performance rather than flight ability, but the lower limit of vertical acceleration in those (Marden and Chai 1991) and other insects (Marden 1987) with small thoraces is just over 1G (9.8 m/s2), the lower limit of flight ability per se.


BKE complain about my method of pushing an insect into the substrate to make an experimental imprint. My objective in using that method was only to observe the ventral sternal impressions of a stonefly, but in doing so I discovered, unexpectedly, that the wing edges left marks alongside the body with an appearance and spacing closely matching those of the trace fossil. I subsequently observed live stoneflies walking freely across wet sediment (a critical experiment not mentioned in BKE), which confirmed that folded neopteran wings can leave marks similar to the trace fossil. BKE further criticize my choice of sediment, but I am skeptical that changing the substate in the manner they suggest would critically affect the outcome of these experiments.

Regarding their accusation of bias in my choice of experimental animal, I did not subsequently test mayflies because all paleopteran insects rest with their wings above the body and hence could not possibly leave marks on the sediment alongside the body. Tellingly, the reply of BKE includes no defense of their mayfly identification.


As pointed out in 1994 (Marden and Kramer), the wings of mayfly subimagos and stoneflies are covered with the same wet-resistant (Watson et al. 2010) hairs (microtrichia), implying an ancestral need for water repellency by pterygote wings. Subsequent research has revealed that surface skimming locomotion maps to a basal position in stoneflies (Thomas et al. 2000) and occurs also in an odonate (Samways 1996) and certain mayflies (Marden et al. 2000), including a species in which the female does not fly, but rather mates and completes her life cycle without leaving the water surface (Ruffieux et al. 1998). This trace fossil provides evidence that skimming over water was used by insects in the Carboniferous, a finding that adds weight to the notion that skimming has ancient rather than recent origins. In their reply, BKE again cite Will (1995) to argue that skimming is not a plesiomorphic trait of stoneflies, a contention that my critique pointed out was rejected by a study (Thomas et al. 2000) that surveyed surface skimming behavior across stonefly families, mapped the behavior on a rooted molecular phylogeny, and concluded that skimming is a shared ancestral trait of Plecoptera. Skimming is so common in Plecoptera (e.g., Marden and Thomas 2003) that its status as a plesiomorphic trait is robust to alternative phylogenetic hypotheses such as those placing Antarctoperlaria families at the base of the tree, as in Will's phylogeny. The appropriate question then is “how deep in time and pterygote phylogeny does this behavior extend and to what extent are its multiple appearances convergent vs. homologous?” This trace fossil indicating skimming locomotion by a stonefly-like neopteran (or a mayfly if one prefers KEB's interpretation) in the Carboniferous contributes a precious data point by showing that the behavior was present fairly early in the fossil history of pterygotes.

These issues are embedded in a long-running debate about the anatomical origin of insect wings and whether pterygote insects arose from terrestrial or aquatic ancestors (Toms 2007). BKE cite an in-press review as showing that wings are not derived from gills, but I see no such clarity in the developmental genetics literature. The most recent and taxonomically appropriate developmental genetic study (Niwa et al. 2010) concluded that mayfly wings show evidence for a mixture of appendage (i.e., possibly gill) and body wall origins. My critique (Marden 2013) included a video (EVO_1743_sm_video_S2.dv) showing the first evidence that the abdominal segmental gills present in certain Odonata (Polythoridae) are flappable. Another recent study found exquisitely preserved and apparently flappable segmental gills in a newly discovered sister taxon of mayflies, the Coxoplectoptera (Staniczek et al. 2011). As pointed out in my critique, Grimaldi and Engel's (2005) contention that gills are not homologous across aquatic insect orders refers to derived, nonflappable gills that occur in various body locations, irrelevant to the homology of segmental flapping gills thought to have given rise to wings. Similarly, I find little or no basis for the sweeping statements in BKE, Grimalidi and Engel (2005) and elsewhere that terrestriality is the ancestral state for pterygote insects. In contrast, the presence in multiple orders (Odonata, Ephemeroptera, Coxoplectopera, Plecoptera) of flappable segmental gills is compelling evidence that pterygote insects arose from aquatic ancestors. Such complex structures are unlikely to have evolved independently in all of these groups, but would be routinely lost in taxa radiating onto land.