Secondary aquatic adaptations independently evolved more than thirty times from terrestrial vertebrate ancestors 1,2 . For decades, non-avian dinosaurs were believed to be an exception to this pattern. Only a few species have been hypothesized as partly or predominantly aquatic 3,4,5,6,7,8,9,10,11 . However, these hypotheses remain controversial 12,13 largely due to the difficulty of identifying unambiguous anatomical adaptations for aquatic habits in extinct animals. In this study, we demonstrate that the relationship between bone density and aquatic ecologies across extant amniotes provides a reliable inference of aquatic habits in extinct species.We use this approach to evaluate the distribution of aquatic adaptations among non-avian dinosaurs. We find strong support for aquatic habits in spinosaurids, associated with a remarkable increase in bone density, which precedes the evolution of more conspicuous anatomical modifications, a pattern also observed in other aquatic reptiles and mammals 14,15,16 .Spinosaurids are revealed to be aquatic specialists with surprising ecological disparity, including subaqueous foraging behavior in Spinosaurus and Baryonyx, and non-diving habits in Suchomimus.

Intensive research on non-avian dinosaurs in recent decades strongly suggests that these animals were restricted to terrestrial environments 1. Historical views proposing that some groups, such as sauropods and hadrosaurs, lived in aquatic environments 2,3 were abandoned decades ago 4,5,6. Recently, however, it has been argued that at least some spinosaurids, an unusual group of large-bodied Cretaceous theropods, were semi-aquatic 7,8 , but this idea has been challenged on anatomical, biomechanical, and taphonomic grounds and remains controversial 9,10,11. Here we present the first unambiguous evidence for an aquatic propulsive structure in a dinosaur, the giant theropod Spinosaurus aegyptiacus 7, 12. This dinosaur has a tail with an unexpected and unique shape consisting of extremely tall neural spines and elongate chevrons forming a large, flexible, fin-like organ capable of extensive lateral excursion. Using a robotic flapping apparatus to measure undulatory forces in physical tail models, we show that the tail shape of Spinosaurus produces greater thrust and efficiency in water than the tail shapes of terrestrial dinosaurs, comparable to that of extant aquatic vertebrates that use vertically expanded tails to generate forward propulsion while swimming. This conclusion is consistent with a suite of adaptations for an aquatic lifestyle and a piscivorous diet in Spinosaurus 7,13,14. Although developed to a lesser degree, aquatic adaptations are also found in other spinosaurids 15,16 , a clade with a near global distribution and a stratigraphic range of more than 50 million years 14 , documenting a significant invasion of aquatic environments by dinosaurs.

Besanosaurus leptorhynchus Dal Sasso & Pinna, 1996 was described on the basis of a single fossil excavated near Besano (Italy) nearly three decades ago. Here, we re-examine its cranial osteology and assign five additional specimens to B. leptorhynchus, four of which were so far undescribed. All of the referred specimens were collected from the Middle Triassic outcrops of the Monte San Giorgio area (Italy/Switzerland) and are housed in various museum collections in Europe. The revised diagnosis of the taxon includes the following combination of cranial characters: extreme longirostry; an elongate frontal not participating in the supratemporal fenestra; a prominent ‘triangular process’ of the quadrate; a caudoventral exposure of the postorbital on the skull roof; a prominent coronoid (preglenoid) process of the surangular; tiny conical teeth with coarsely-striated crown surfaces and deeply-grooved roots; mesial maxillary teeth set in sockets; distal maxillary teeth set in a short groove. All these characters are shared with the holotype of Mikadocephalus gracilirostris Maisch & Matzke, 1997, which we consider as a junior synonym of B. leptorhynchus. An updated phylogenetic analysis, which includes revised scores for B. leptorhynchus and several other shastasaurids, recovers B. leptorhynchus as a basal merriamosaurian, but it is unclear if Shastasauridae form a clade, or represent a paraphyletic group. The inferred body length of the examined specimens ranges from 1 m to about 8 m. The extreme longirostry suggests that B. leptorhynchus primarily fed on small and elusive prey, feeding lower in the food web than an apex predator: a novel ecological specialisation never reported before the Anisian in a large diapsid. This specialization might have triggered an increase of body size and helped to maintain low competition among the diverse ichthyosaur fauna of the Besano Formation.

Myhrvold et al. suggest that our inference of subaqueous foraging among spinosaurids is undermined by selective bone sampling, inadequate statistical procedures, and use of inaccurate ecological categorizations. Myhrvold et al. ignore major details of our analyses and results, and instead choose to portray our inferences as if they were based on qualitative interpretations of our plots, without providing additional analyses to support their claims. In this manuscript, we thoroughly discuss all the concerns exposed by Myhrvold et al.. Additional analyses based on our original datasets and novel data presented by Myhrvold et al. do not change our original interpretations: while the spinosaurid dinosaurs Spinosaurus and Baryonyx are recovered as subaqueous foragers, Suchomimus is inferred as a non-diving animal.

Here we describe 31 fossil teeth, deposited in the palaeontological collections of the Museo di Storia Naturale di Milano (MSNM), that come from the inland portion of the Mahajanga Basin, NW Madagascar, namely from the Sakahara Formation (classically known as Isalo IIIb subunit), which is dated to the Bathonian stage of the Middle Jurassic. Based on detailed morphological characters, the eight morphotypes recognized herein are tentatively referred to four sauropod taxa: Archaeodontosaurus descouensi, ‘Bothriospondylus madagascariensis’, Lapparentosaurus madagascariensis and an indeterminate specialized eusauropod, which may represent a new species and provides the first evidence of a Bathonian diplodocoid in Madagascar. The identification of the teeth is corroborated by comparative examination of morphometric data. We provide evidence that Titanosauriformes were present in the Bathonian, on the basis of seven specimens referable to this clade. We also discuss in detail some dental characters that support the existence of a clear niche partitioning between the abovementioned taxa that co‐existed in the Malagasy Middle Jurassic terrestrial ecosystem. We hypothesize, for the first time, a direct correlation between the pattern drawn on the tooth crown by the enamel wrinkles and the feeding ecology of sauropod dinosaurs. The enamel wrinkles probably played a structural function: coarse wrinkles were associated with a diet composed mainly of hard foodstuff, whereas fainter wrinkles, which appeared in more derived morphologies, were associated with a diet composed of softer foodstuff.