A family of Microsoft relational database management systems designed for ease of use.
In my case, I want to make sure that no two locations point to the same Parish.
In that case you'd simply need to include the Parish, District, and County columns in Locations_NatKey in a single unique index. It would still be possible to have more than one row in Locations_NatKey with the same value in the Parish column, but the parishes would then be distinct entities, which happen to have the same names, but be in different districts. This would be perfectly valid as some common parish names occur in multiple places throughout England.
As regards your final paragraph, the inclusion of a 'natural key' model in the demo is simply to show that it can be done, and has two advantages: (a) correlated combo box controls can be used in continuous forms view without having to use 'hybrid' controls, as is the case with surrogate keys, as illustrated elsewhere in the demo; (b) it allows simpler queries to be used as these can be based on one table only, rather than having to join all four tables. In an operational database I would be more likely to use surrogate keys, however, and go the extra mile when building the interface.
One thing I didn't include in the demo was the use of the NotInList event procedure when using correlated combo boxes. However, you will find an example of this in DatabaseBasics.zip in my same OneDrive folder at:
https://onedrive.live.com/?cid=44CC60D7FEA42912&id=44CC60D7FEA42912!169
In this little demo file the section on 'entering data via a form/subforms' includes a simple contacts form in which correlated combo boxes are used to select a country, region, then city, the first two being unbound so as to keep the table in Third Normal Form. The NotInList event procedure of each combo box allows new values to be inserted into the referenced tables, by typing the name of the new country, region, or city into the combo box. The code in each case is only slightly more complex than would normally be the case.