In a relational database tables and queries are UNORDERED sets. That means that the database engine might return the results as 1,2, 3 one time and 3,1,2 the next. The ONLY way you can ensure that a recordset is returned in a specific, consistent order is to use a query with an Order By clause.
Access tends to fool people because normally when you open a table in ds view, it "appears" to be sorted by PK order but believe me when I tell you that, you cannot rely on this. The appearance is exacerbated if you have saved a sort order for either the table or a query. When you compact a database, one of the tasks performed is to sort ALL tables into PK order and write them back to disk in that sequence. As long as you don't update any of the early records, they will appear to be in PK order. However, if you update record #1 and make it too large to fit back in its original physical position on the disk, Access will add the changed record to the first available free space, which in the case of Access will be at the end of the physical file. More sophisticated database engines spread freespace throughout the file space so records will generally be written back some place closer to where they were before they were updated.
Anyway, you can prove all of this to yourself if you are so inclined. You just need a large enough set of data and you need to make a sufficient field size increase to force Access to move the record to see the effect. The reason you need a large number of records is because data I/O is done at the physical record level. This is dictated by the device. For example, on one disk, the size of a physical record might be 2K but on another it might be 4k and if your average record size is 100 bytes, there will be multiple logical records in one physical record. Access (like most other relational databases) will always rewrite the entire physical record when one logical record is updated, so you actually need to force the physical record size to be exceeded for Access to put the updated record back at the end of the file space and into the last physical record of the table.
That is a very high level explanation of how records are read and written in a database.