In the images you're analyzing, I don't think it will matter too much which version you use - connectivity4 or connectivity8. With connectivity4 you are telling the algorithm to only consider items to be together if their pixels are adjacent - touching on one of four sides as in the following:
X
XOX
X
On connectivity8, the two items are together if thier pixels are adjacent or 'kitty-corner' - any of the 8 surronding pixels as shown below-
XXX
XOX
XXX
For more in depth discussion I would look at the vision concepts manual, that ships with the vision toolkit. Go to Start >> Programs >> National Instruments >> Vision >> Documentation.