Hank recently wrote a piece that dealt with the problem of a "ghost train" and a "ghost hunter" that was killed by a real train while waiting for the apparition.

Since these were amateur ghost hunters, it would be a bit much to presume that there was any actual scientific query going on, but nevertheless there seems to be a persistence of belief that ghosts are the subject of active research.  However, the question to ask is whether there's any scientific basis for thinking there's something to investigate.

Because research cannot start by assuming what it will find, here is a basic working definition of a ghost:

    * a ghost (or apparition) is a human (sometimes animal) figure, witnessed by someone, which cannot be physically present
http://www.assap.org/newsite/articles/Ghosts.html

A ghost - or spirit or apparition - is the energy, soul or personality of a person who has died and has somehow gotten stuck between this plane of existence and the next. Most researchers believe that these spirits do not know they are dead. Very often they have died under traumatic, unusual or highly emotional circumstances.
http://paranormal.about.com/cs/ghosthunting/f/blfaqs_ghosts01.htm

According to local legend, an apparition of the 1891 Bostian's Bridge crash appears every year on the anniversary of the accident, complete with screaming passengers and grinding metal.
http://www.aolnews.com/nation/article/man-looking-for-ghost-train-killed-by-real-locomotive/19612028

One problem we have immediately is questioning what the basis would be for inanimate objects to be made into "ghosts".  Even if we accepted the idea of ghosts as being spirits, what is the basis for assuming a train could become a ghost?  For that matter, why would ghosts be clothed?  Their clothing certainly has no spiritual dimension to it, so there's no basis for believing that the afterlife requires any such modesty.

Another issue is why a ghost should still follow the physical path of objects it is interacting with.  Why should a ghost train confine itself to real tracks?  Why should a ghost walking down stairs follow such a path?

Of course another fundamental problem in all of this is simple physics.  In order to see an object, light must be reflected from some surface.  To hear something, requires the movement of air.  All of these require some physical interaction with the "real" world, which is problematic for the ghost hunter.

After all, Newton's laws expressly state that for every action there is an equal and opposite reaction.  Therefore if a ghost can interact with the "real" world to create such effects, then the "real" world can also interact with the ghost.

Therefore such interactions would result in real measurements.  Not mysterious electro-magnetic fields (the fallback for any paranormal investigator), but real physical phenomenon.

So how should one investigate ghosts?

Luckily, there is an alternative to assumption-led investigation of ghosts. It was widely used before the TV ghost hunting shows and deserves to be more widely adopted once again. The approach is a more scientific, neutral, evidence-based method. Evidence-based investigations start with no assumptions as to what a ghost is or what may be causing a haunting. So, how does it work?
http://www.assap.org/newsite/articles/Scientific%20ghost%20research.html

Despite the claim that one needs to gather evidence, this doesn't make ghost-hunting science.  Gathering evidence makes no sense in the absence of any hypothesis to test.  Even though it looks like paranormal researchers are loaded up with all kinds of high-tech equipment and measuring devices, there is no ability to establish a control for measurement since no one knows what it is that is supposed to be significant.  In effect, it's like leaving a tape recorder or video camera sitting outside all night and then attempting to guess at what sounds and images are captured that seem unusual.

So the problem isn't gathering data, but formulating an initial hypothesis about what ghosts are.  Once such a claim is made, then it can be logically checked to see what requirements or predictions it must make, and what the criteria for a valid test would be.

More importantly, such a hypothesis wouldn't necessarily be confined just to ghosts.  After all, if there's something like a "ghost train", then clearly there must be some scientific principle that would explain how an inanimate object can make an "appearance" (especially to keep an anniversary schedule).

It is precisely when trying to articular such a hypothesis, that all the inconsistencies and absurdities surface.  Therefore if you're going to do paranormal investigation, it's much easier to come up with the gear and conduct "measurements" than it is to explain what you're trying to measure.