Data Collection.

Data were either collected by the authors or extracted from references (16, 28, 33–36, 39, 42–84).

Arthropod specimens were collected around Cambridge, United Kingdom or Brisbane, Australia or obtained from the Cambridge Phasmid Study Group. All arthropods were identified (85–87), and their live weight was recorded (ME5, resolution 1 μg, max 5 g or 1,202 MP, resolution 0.01 g, max 300 g; both Sartorius AG). Attachment pads were photographed either with a Canon EOS camera mounted on a stereo microscope (MZ16; Leica Microsystems Ltd.) or by using SEM for large and small specimens, respectively. Some pads were imaged while in contact with glass, visualized using the stereomicroscope with coaxial illumination. For SEM imaging, individual legs were dried, mounted on stubs, sputter-coated at 65 mA for 10–20 s (K575X turbo-pump sputter; Quorum Technologies) and examined with a field emission gun SEM at a beam voltage of 5 kV (Leo Gemini 1530VP; Carl-Zeiss NTS GmbH).

Data on toepad-bearing gecko species were collected from live animals kept in the D.J.I. laboratory [under an Institutional Animal Care and Use (IACUC) protocol 2012–0064 from the University of Massachusetts, Amherst to D.J.I.] and preserved specimens from the American Museum of Natural History and the Museum of Comparative Zoology at Harvard University. For each specimen, photos of one forefoot were obtained by pressing it tightly against the glass plate of an Epson Perfection V500 Photo Scanner (Seiko Epson Corp.) next to a ruler and taking a digital scan. The total toepad area across all digits was measured using ImageJ, version 1.49r (88). We also measured snout-vent length (±1 mm) from each individual using a clear plastic ruler. Where possible, we measured multiple conspecific individuals and used the mean as the species value.

Literature data were taken from the papers’ text or tables or were extracted from figures using WebPlotDigitizer 3.3 (WPD, developed by Ankit Rohatgi; arohatgi.info/WebPlotDigitizer) or ImageJ, version 1.49m. We tested the performance of WPD with an x–y plot of 20 random numbers between 1 and 1,000 (x) and 0.01 and 10,000 (y) on a log-log scale and found an accuracy of ≈0.6% for the raw data.

We used live body weights where available and interpolated live weight from body length where necessary, using established scaling relationships (38). A list of the species included can be found in the Supporting Information.