So, here are some of the implications of choosing secundative alignment. They aren't really strict implications so much as Stuff You Might Consider Thinking About™. I think the effects of choosing secundative alignment are pretty local overall and won't have cascading effects on your grammar. In my opinion, the thorniest issue here is dealing with valence-changing operations on verbs.
- What do secundative languages tend to look like in the real world?
- How does secundative alignment manifest itself in morphology or syntax?
- Valence-changing operations
- Relative Clauses
Real World Secundative Languages.
If you want your language to be naturalistic, I suggest taking a look at WALS to see what kinds of combinations of features are typologically uncommon. If you are not trying to make a naturalistic language, knowing what natural languages tend to do is a good way to spend your "weirdness budget" wisely.
This map marks the strategy used for the verb
to give. This analysis could be misleading because it's possible for the verb
give to use a different strategy from other verbs, but I think this feature is the best fit for what you're after of the available WALS features.
Looking at the map, my assessment of the broad areal trends is that Eurasia is dominated by the indirect object construction. The secondary object construction seems to be common in North America and in New Guinea.
The WALS Chapter for this feature provides some examples of the different types of constructions as well as a note about what other features commonly co-occur with secondary object marking, excerpt below:
...head-marking languages with verb indexing of two core arguments tend
to show the secondary-object pattern...
How does secundative alignment manifest itself?
Basically, there are four ways your language can manifest secundative alignment.
- word order
- presence of dedicated adpositions
- case markers
- verbal agreement
I think most of the real world secundative languages are secundative because the recipient in the verb
to give or other similar verbs gets indexed in the verb the same way that a prototypical direct object would.
This can occur in a language with no case marking. So, I give you the book might be something like
In this example, book bears no case marking but is clearly the theme.
The case marking strategy is covered in your example.
For word order, we can take an example from English.
I give you the book
you is the recipient, but occupies the position directly after the verb. In English you have a choice between using a preposition like
for instead. If no such option existed, this would be a pristine example of syntactic secundativeness.
Note that the language can be partially secundative.
For instance, the following is possible.
you.sg.DAT book.sg.ACC give.1sg-subj.2sg-obj
In this example, the two non-agent arguments
give receive indirect-object-style case marking, but the verb shows secundative alignment.
Valence changing operations
It is important to think about how the passive voice, if it exists, should behave.
Suppose we have a language with secundative alignment in its cases.
I.NOM you.OBJ book.INST give.PST
I gave you a book.
If you were to passivize the verb, which argument would be promoted to the agent? Would (101), (102), or neither be acceptable?
(101) you.NOM give.PSV.PST
(102) book.NOM give.PSV.PST
Do you have an anti-applicative voice that deletes the recipient and promotes the theme to primary object? As far as I know, anti-applicative voices do not exist in the real world, but applicative voice does exist. It's possible to allow this promotion without an explicit voice for it.
I.NOM book.OBJ give.ANTIAPP.PST
This situation gets murkier if the language is both ergative and secundative.
I.ERG you.ABS book.INST give.PST
I gave the book.
In this case, the antipassive voice would delete the absolutive argument and promote the ergative argument.
I.ABS book.INST give.ANTIPSV.PST
I gave the book (to someone).
Some ergative languages mark instruments and agents the same way.
I.ERG book.ERG you.ABS give.PST
In this case, the antipassive voice could potentially target either argument.
Deciding how relative clauses in secundative languages should work might present some subtle issues.
Languages frequently have a limit on what kinds of arguments can be targeted by relative clauses.
You may want to choose to make only the agent and the primary object extractable, but not secondary objects.