I am somewhat following what you are saying, however are the three links any particular way? I really need a better picture of your database structure to give you a viable solution.
You haven't told us too much about either the database structure or your process, so I'll make some assumptions and base a suggestion on those assumptions.
First, I assume that a print is made from a negative. I also assume that you somehow mark records that represent negatives, so that you know they're negatives. Let's say you have a field "type" that is either "N" (negative) or "P" (print).
You should also have a field on each record that carries the ID of a negative (let's call it the "negative_id" field). It will have a different meaning for P (print) and N (negative) records.
In the case of a print, that field would contain the ID of the negative from which the print was made. In the case of a negative, it would contain the ID of the negative itself.
With those assumptions, it would seem that whenever the user adds a record for a print, the user should be required to select the negative from which the print was made. The ID of the selected negative would be carried in the negative_id field. In the case of a negative, the field would simply repeat its own ID.
Now all records contain a negative ID that can be used to drive a relationship to all other records that carry the same negative ID.