And one of hardest (but most intellectually stimulating) diagnoses to make is that of "spells." That's what I call episodes that come and go, that have a beginning and an end—and something unusual in between. The basic process of diagnosing spells should be familiar to anyone who has taken a squeaking car to a mechanic. The one time that car doesn't squeak is when mechanic is inspecting it. So mechanic has to make an analysis based on what you describe.

The same thing occurs in diagnosing people with spells. When an attack occurs in front of a doctor, it's usually easy to diagnose. But that almost never happens. Usually, all we have to go on is description, or, hopefully, two descriptions—one from person who had spell and a second from someone else who was there to witness it.

Methodically, each of two accounts is broken down into three parts—the events leading up to attack, attack itself, and what happened afterwards. Each account, taken one at a time, is based on what that person actually saw, heard and could remember, reported in a way particular to that person's abilities to observe and articulate. To make matters more challenging, patient who had attack often has significant gaps in their memory.

The list of potential underlying causes—what I think of as differential diagnosis of things that come and go—spans multiple medical disciplines and is almost as broad as medicine itself. For example, let's assemble just a short list of conditions that can occur as episodic symptoms: seizures, pseudoseizures (seizure-like attacks of psychological origin), fainting spells, hypoglycemia, panic attacks, irregular heartbeats, dissociations, transient ischemic attacks (TIAs), migraine and vertigo.