Determining what extinct dinosaurs ate is difficult, but we can infer some aspects of their dietary preferences. Traditionally, this information has been derived from direct evidence, such as stomach contents, and indirect evidence, such as establishing a correlation between particular body characteristics and diets of living animals and then inferring habits for dinosaurs.

Animals such as house cats and dogs have large, stabbing canine teeth at the front of the mouth and smaller, equally sharp teeth farther back in their jaws. Many of these animals are also armed with sharp claws. The advantage of teeth and claws as predatory tools is obvious. Now consider animals like cows, horses, rabbits, and mice. These animals have flat teeth at the back of the jaw that are analogous to and have the same function as grindstones. Unlike the meat-slicing and stabbing teeth of carnivores, the teeth of these animals grind and shred plant material before digestion.

More clues exist in other parts of the skull. The jaw joint of carnivores such as dogs and cats has the mechanical advantage of being at the same level as the tooth row, allowing the jaws to close with tremendous speed and forcing the upper teeth to occlude against the lower teeth with great precision. In herbivorous animals, rapid jaw closure is less important. Because the flat teeth of herbivores work like grindstones, however, the jaws mush move both side to side and front to back. The jaw joints of many advanced herbivores, such as cows, lie at a different level than the tooth row, allowing transverse tearing, shredding, and compression of plant material. If we extend such observations to extinct dinosaurs, we can infer dietary preferences (such as carnivory and herbivory), even though we cannot determine the exact diet. The duck-billed dinosaurs known as hadrosaurs are a good example of a group whose jaw joint is below the level of the tooth row, which probably helped them grind up tough, fibrous vegetation.

Paleontologists would like to be much more specific about a dinosaur's diet than simply differentiating carnivore from herbivore. This finer level of resolution requires direct fossil evidence of dinosaur meals. Stomach contents are only rarely preserved, but when present, allow us to determine exactly what these animals were eating.

In the stomach contents of specimens of Coelophysis (a small, long-necked dinosaur) are bones from juvenile animals of the same species. At one time, these were thought to represent embryonic animals, suggesting that this small dinosaur gave birth to live young rather than laying eggs. Further research indicated that the small dinosaurs were too large and too well developed to be prehatchling young. In addition, the juveniles inside the body cavity were of different sizes. All the evidence points to the conclusion that these are the remains of prey items and that, as an adult, Coelophysis was at least in part a cannibal.

Fossilized stomach contents are not restricted to carnivorous dinosaurs. In a few rare cases, most of them “mummies” (unusually well preserved specimens), fossilized plant remains have been found inside the body cavity of hadrosaurs. Some paleontologists have argued that these represent stream accumulations rather than final meals. The best known of these cases is the second Edmontosaurus mummy collected by the Sternbergs. In the chest cavity of this specimen, which is housed in the Senckenberg Museum in Germany, are the fossil remains of conifer needles, twigs, seeds, and fruits. Similar finds in Corythosaurus specimens from Alberta, Canada, have also been reported, indicating that at least two kinds of Late Cretaceous hadrosaurs fed on the sorts of tress that are common in today's boreal woodlands.

A second form of direct evidence comes from coprolites (fossilized bodily waste). Several dinosaur fossil localities preserve coprolites. Coprolites yield unequivocal evidence about the dietary habits of dinosaurs. Many parts of plants and animals are extremely resistant to the digestive systems of animals and pass completely through the body with little or no alteration. Study of coprolites has indicated that the diets of some herbivorous dinosaurs were relatively diverse, while other dinosaurs appear to have been specialists, feeding on particular types of plants. The problem with inferring diets from coprolites is the difficulty in accurately associating a particular coprolite with a specific dinosaur.

要确定已经灭绝的恐龙吃什么是一件很困难的事情，但是我们能推测出它们在饮食上的一些偏好。传统上，这些信息来自直接证据，如胃内食物，以及间接证据，如通过在恐龙的特定身体特征和现今存活的动物的饮食习惯之间建立关系来推断恐龙的饮食习惯。

像家养的猫和狗这类动物，嘴前部的牙齿又大又尖，后部的牙齿要小一些但是同样锋利。这些动物中有许多都有着锋利的爪子。牙齿和爪子作为捕食工具的优势非常明显。现在再看牛、马、兔子和老鼠这些动物。它们后部的牙齿比较扁平，功能类似于研磨的石头。不像食肉动物的牙齿是刺进肉里并把肉撕开，这些动物的牙齿会将植物磨成碎块再消化。

头骨的其他部分存在更多线索。狗和猫这类食肉动物的颌关节和牙齿在同一高度上，这种机械上的优势能使上下颌快速闭合，使得上下牙齿咬合严密。对草食动物来说，快速颌闭合并不重要。由于食草动物的扁平牙齿像磨刀石一般，上下颌必须前后左右移动。许多高等草食动物，如牛，颌关节和牙齿的高度并不一致，这样它们就能横向撕裂、切碎和挤压植物。如果我们把这种观察方法应用到灭绝的恐龙身上，我们可以推断出恐龙的饮食偏好（如肉食还是草食），虽然我们无法确定准确的食物。鸭嘴龙便是一个颌关节比牙齿低的例子，这种高度不一致可能会帮助它们把坚韧的纤维植物磨碎。

古生物学家不只是想要区分恐龙是食草还是吃肉，他们还想知道恐龙的具体饮食。这种更精细的研究决心需要直接的恐龙食物的化石证据。胃里的食物很少能保存，但如果有，就能让我们确定恐龙到底吃什么。

腔骨龙（一种体型较小、脖子较长的恐龙）的胃内食物样本是同类恐龙的幼崽骨头。科学家一度认为这些幼崽骨头属于腔骨龙的胚胎，认为这种小恐龙是胎生而非卵生。进一步的研究发现，这些幼崽体积太大，而且发育良好，不可能是孵化前的幼崽。另外，这些胃内的幼崽残骸大小也不一。所有的证据都表明，这些都是猎物的残骸，成年腔骨龙中至少有一部分是吃自己的同类的。

并非只有食肉恐龙才有胃内食物的化石。在一些罕见的情况下，鸭嘴龙的体内也发现有植物遗骸的化石或“木乃伊”（罕见的保存完好的标本）。一些古生物学家认为，这些食物是积累下来的食物而不是最后一餐。最著名的案例是施特恩伯格一家收集的第二具埃德蒙顿木乃伊。标本存放在德国森肯伯格博物馆内，在其胸腔内，是针叶、细枝、种子和果实的化石。加拿大阿尔伯塔的冠龙标本也报导了类似的发现，这表明至少有两种晚白垩纪的鸭嘴龙是以现在北半球森林中的常见树木为食的。

另一个直接的证据形式来自粪化石（身体排泄物的化石）。几个恐龙化石地点保存有粪化石。粪化石能明确说明恐龙的饮食习惯。很多植物和动物不容易被动物的消化系统消化，没有改变或者只有些许改变就排出体内了。对粪化石的研究表明，有些食草恐龙的饮食比较多样化，而另一些恐龙似乎专门吃某些特定的食物。从粪化石推断饮食偏好，困难在于如何准确知道哪种粪化石属于哪种特定的恐龙。