I don't know how to get at metadata directly, but you can export the metadata to xml (I believe this tool only works FROM INSIDE PYTHON in SP1 and beyond) and then read through the xml with an ElementTree. I can't remember if the translator doesn't export all FGDC tags or doesn't import all FGDC tags correctly, but it's one of the two, so beware.
import arcpy, sys
from xml.etree.ElementTree import ElementTree
from xml.etree.ElementTree import Element, SubElement
from arcpy import env
env.overwriteOutput = True
fcpath = r"X:\working\metadata\BlankGDB.gdb\FullMetaFC" # the input feature class
translatorpath = r"D:\Program Files (x86)\ArcGIS\Desktop10.0\Metadata\Translator\ARCGIS2FGDC.xml" # the translator file to use - it should be in an installed folder like this
xmlfile = r"X:\working\metadata\FullMetaFC.xml" # the output xml file
arcpy.ExportMetadata_conversion(fcpath, translatorpath, xmlfile) # export the metadata to xml
tree = ElementTree() # make an ElementTree object
tree.parse(xmlfile) # read the xml into the ElementTree
spot = tree.find("idinfo/citation/citeinfo/title") # find whatever tag you want
print spot.text # print the text between the tags
edit: updated paths as Curtis advises below.