A dissertation, derived from the Latin verb disserto, dissetare, dissertavi, dissertatus (meaning "to discuss") is the presentation of the results of one's endeavors as candidate for a research doctorate. Dissertations often run from 50 pages (in the engineering sciences) to over 300 (in the humanities and social sciences). Most dissertations form the basis of a later book.
A dissertation in the humanities and social sciences often takes a standardized format. The first two chapters sketch out the problem the author is attempting to address. It describes the question, reasons why previous scholarship is inadequate, and presents an outline of the author's approach. In most cases, it also includes a careful review of previous literature (that the author has read and is familiar with) along with comments on why they, too, are inadequate. As the first two chapters contain these elements, they are usually submitted as a dissertation proposal.
Science and engineering dissertations are often much briefer and to the point.
Dissertation writing can commence as soon as the candidate completes his oral and written exams, usually after the second or third year in the PhD program. In most cases, the research work involved in the dissertation far outweighs the time actually spent writing the dissertation. During this time, the candidate is known as an ABD ("All But Dissertation") and if he opts out, may be awarded the M.Phil (Master of Philosophy) degree.
Reearch work associated therewith and the writing of the dissertation can take as little as two or three years (some PhDs are completed in four or five years) to as long as seven or eight years (on the other end of the spectrum, a PhD can take a decade or more). GSAS has set in place an upper-bound limit of seven years in how long a PhD candidate may take to complete his degree. However, this is not a hard and fast rule. Candidates can, and often do, apply for extensions, deferrals, leaves of absence, et. al.
Once the dissertation has been successfully completed, the candidate may schedule a viva voce defense of the dissertation. In this defense, he must successfully present his work to his examiners and answer their questions on his procedure, methodology, and results. In other words, it's a final examination.
Typically, a defense committee is made up 3 members, at least 2 of whom must be from the dissertation writer's department (including his advisor). The third may be from an associated department (for example, Political Science for a History dissertation) or even another university (if the candidate is specializing in a particularly obscure topic, or if there is a notable faculty member at another university that the candidate would like to be examined by), but this is not a necessity. In engineering and applied sciences, however, a representative from the industry may be invited to sit as the fourth member of the defense committee.
Each committee member will have received a copy of the dissertation beforehand and will be expected to have read and understood it, and come prepared with questions. However, the candidate will still usually know far more about the subject than his examiners. Humanities and social sciences defenses normally take two hours; science and engineering defenses usually take one.
The candidate begins the defense with an overview of the results of his work. This usually lasts for 1/2 or 2/3 of the alloted time. For the remaining time, the committee members pose questions and may request further explanation or clarification. At some universities, the general public is invited to pose questions after the committee is finished. This is not the case at Columbia, as dissertation defenses are closed to the public.
Following the session, the committee may render one of four decisions:
- Accept. The dissertation is complete and in acceptable condition for permanent deposit and publication. Rarely granted.
- Accept with minor revisions. The dissertation has minor revisions, in wording, grammar, typos, etc. that must be completed before it can be accepted. But otherwise, the conclusions, methodology, argument, results, etc. are all fine. But maybe the candidate just forgot to dot an i or cross a t or spelled someone's name wrong. Or something. The most common decision of the defense committees.
- Provisionally accept with major revisions. The dissertation needs a major rewrite, as the candidate's process and methodology is flawed, or perhaps the conclusions were too much of a stretch for the presented evidence, or an argument is unconvincing. In some cases, the candidate is required to schedule a second defense. Rarely granted.
- Not accepted. There are major things wrong with the dissertation. Like the author claimed the moon is made of goat cheese or something. This dissertation is rejected, in its entirety. This usually leads the candidate being required to withdraw from the program, and as such, is almost never imposed.
The candidate is informed of the decision on the spot, and champagne is usually broken out if the results are as expected.
The reason why few, if any, dissertations are outright rejected (a history professor has estimated one or two per decade) is that the dissertation advisor will usually not permit a candidate to even attempt to schedule a defense unless the work satisfies the advisor. Therefore, the dissertation defense is usually a ceremonial affair -- the actual examination has already been done beforehand by the advisor, and usually with the input and advice of other members of the committee. In addition, it reflects very badly on the advisor if he permits an unready candidate to proceed to the defense. Thus, in almost all cases, the dissertation is accepted.
It is also usually accepted that the American dissertation defense is tame and boring. In the Netherlands, for example, due to dissertation defenses in the 17th and 18th centuries sometimes escalating to physical altercations, a PhD candidate is required to bring a pair of armed bodyguards to protect himself from the mad professors. However, their role is largely ceremonial and they are usually chosen from the candidate's family. A toga-wearing mace-bearer announces the commencement and end (hora est, or "It is the hour!") of the proceedings.
- Max Horlick '54 received his PhD in 2007, 53 years after he successfully defended it. The university retroactively granted him his PhD. 
- Wm. Theodore de Bary CC '41, MA '48, PhD '53 successfully defended his dissertation, which was accepted with minor revisions in 1953. De Bary promised to make them as soon as possible. Unfortunately, he then started working on Sources of Chinese Tradition and designing the major texts program in East Asian humanities. Then 1968 happened. Then he had to be Provost. Then he started working on what became the Major Cultures requirement. Somewhere in there, he found time to be Chair of EALAC. Then he retired and began wrapping up the dissertations he sponsored. In 1994, de Bary finally found time to make revisions to his dissertation and turned it in. After that, he was awarded a Litt. D. honoris causa. Then he moved onto newer and better things, like the Nobility and Civility program.