X-ray crystallography is the predominant method for macromolecular structure determination. In this method, crystals are illuminated with intense X-ray radiation that damages the crystal. If crystals are cooled to cryogenic temperatures, the effects of radiation damage are greatly reduced. Since protein crystals are aqueous, cryocooling creates the possibility of crystalline ice formation. X-ray diffraction from ice is difficult to disentangle from the protein diffraction and if the ice is internal to the protein crystal, the expansion of solvent into ice degrades the crystal's order and subsequent diffraction quality. While there has been significant research into how to prevent ice formation, there has been very little research into the basic science of ice formation in macromolecular crystallography. The solvent within protein crystals is nanoconfined so the ice formation process should differ significantly from bulk-like solvent. Many open questions remain, such as what is a protein crystal's freezing point? Once the crystal has been cooled below this temperature, how long does it take for ice to form? When ice does form, what is its structure and how much solvent at the protein's surface is restricted from crystallizing? We performed experiments studying the ice formation process inside of protein crystals. The results from these experiments showed that freezing points are depressed, the structure of ice is stacking disordered and two monolayers of water remain uncrystallized at the protein's surface. Completely new results from our experiments show protein crystals can remain supercooled for 10's of seconds suggesting a dramatic suppression of ice nucleation rates unobserved in any other material. A novel observation was made while performing these experiments, following an expected initial contraction on cooling to temperatures as low as 220 K, apoferritin crystals expand to volumes as large or larger than their volumes at room-temperature. This serves as further evidence that the solvent inside of the protein crystals is in liquid form at temperatures above the protein-solvent glass transition. Using data deposited to the Protein Data Bank and the Integrated Resource for Reproducibility in Macromolecular Crystallography, we show that crystals with larger solvent cavities and percent solvent content are more susceptible to ice formation and the prevalence of hexagonal like ice has dramatically increased in the last 10 years.