V-IO-DO is nothing new
You asked when we started doing this, putting the indirect object between the verb and the direct object, foregoing any preposition.
There's a slight bit of recency illusion here, because the specific answer
to your question is that both orderings of DO and IO have always been
permitted in English. It was present in Old English as shown below, and so is ancient. It predates English itself. It’s the restrictions which are “new”; although Old English was never as position-independent as Latin was, it was still much more flexible in its syntactic ordering than the Present Day English we use today is.
The loss of a case system in determiner phrases (not just a lone personal pronoun, which has some case markings left) has caused us to mark the
non-Patient NP with a preposition whenever we use any ordering other than IO-DO.
This first started happening in Early Middle English as case inflections of regular noun phrases fell away, and it persists to this day.
Dative alternation, allowed and forbidden
Some types of double-object constructions in Present Day English
allow two different possibilities under dative alternation, both the unmarked IO-DO order with the IO glued to the constituent slot directly right of the verb without interventions, and also the DO-IO where the IO gets marked with a preposition and so can now move around a bit, usually to the end.
- John gave Jane the book.
John gave the book to Jane.
- John made Jane a cake.
John made a cake for Jane.
But the only ones that can experience dative alternation are those where one of the two objects is the
semantic Recipient or Beneficiary. The Wikipedia article on Dative shift phrases that restriction this way:
The double object construction requires a possessor/possessed relationship. This means the indirect object in the oblique dative construction must have the theta-role of beneficiary (PP introduced by [for]) or recipient/goal (PP introduced by [to]) to be a candidate for the dative alternation.
You can’t do that with these next four double-object scenarios because you can’t use a preposition to mark the non-Patient argument (the indirect object), and we no longer distinguish dative from accusative even in pronouns:
- John made Jane president.
- The court denied John a public defender.
- The flat tire cost John the race.
- John bet Jane an ice cream cone.
And here is why you cannot:
In (3) this is more like having two direct objects, if you want to think of
it that way; historically those were both inflected into the accusative, unlike (1) and (2) where one was accusative and the other dative. There’s no dative to alternate via a preposition there in the first place. This is a type of causative where John causes Jane to become the president.
In (4) and (5), you did once have datives there, but the verbs are privative ones and so the indirect
object is the Deprivee not the Recipient. It still took the dative case historically, just like most (but not all) prepositional objects once did back in Old English.
A commenter has taken exception to including deny in verbs that do not alternate, but this is explained by Peter Hallman in Syntactic Neutralization in Double Object Constructions from Linguistic Inquiry volume 46, issue 3, Summer 2015, who mentions the “inter-speaker variation on the grammaticality of these verbs.”
Grammaticality can vary by speaker and by historical epoch. There is disagreement among researchers about whether, or to what extent, that deny-class double-object verbs are able to alternate with prepositional versions, starting with Levin. Other verbs in this class include bar, begrudge, envy, forbid, forgive, guarantee, refuse, spare, and vouchsafe. Jane can begrudge John the time which it cost her to change out his flat tire, but it seems impossible for her to begrudge that time to or for him. Similarly, if Jane envies John his good looks, then it is hard to see how you could rewrite that with a prepositional phrase.
The situation in (6) is that bet is in the bill-class of verbs, which also includes verbs like charge, tax, wager, save. These are not easily alternated, if at all.
There are really all kinds of complex restrictions on double object alternations with prepositional versions. Researchers have been working for a good 50 years on devising theoretical frameworks that explain and predict what is and is not permissible. Shakespeare’s writings include double object constructions like:
- I will bar no honest man my house.
- How can I then return in happy plight, That am debarred the benefit of rest?
- Then the world go's hard When Clifford cannot spare his Friends an oath.
- What villaine boy, barst me my way in Rome?
- I am for you, though it cost me ten nights watchings.
- Go, go into old Titus’ sorrowful house, And hither hale that misbelieving Moor To be adjudged some direful slaught’ring death As punishment for his most wicked life.
- All your Interest in those Territories, Is utterly bereft you.
- This Gloster should be quickly rid the World.
Those are like when Dickens wrote:
- Will you forbid him the house where I know he's safe?
But those may no longer all be completely acceptable to all native speakers of English today. An old example that is almost universally considered unacceptable now is the old Jacobist rhyme:
What is the rhyme for porringer?
The King he had a daughter fair,
And gave the Prince of Orange her.
Today’s speakers don’t like using a personal pronoun for the direct object being given over to the recipient, preferring to write that last line as
And he gave her to an Oranger.
Who gave whom whose what?
In general, without a case inflection system to distinguish the indirect
object from the direct object, those two clause constituents must either be
strictly ordered (such as IO before DO or vice versa) and their order with
respect to the clause's verb fixed such as Objects-Verb or
Verb-Objects, or alternately, a preposition or other particle can mark one
of those but not the other. (BTW, those with a background in Romance will recognize this same thing happening in languages that aren't Germanic ones, but just how it all works out in each varies a bit.)
That's because whenever a clause has multiple noun phrases, there must
always be some way to express the role each noun phrase is playing within
that clause. Available mechanisms include ordering these syntactic
constituents one way or another, morphological inflections of individual
words, or by using function words to connect one to another. Uninflected
languages necessarily have fewer choices here, but even when a language
does use morphological case for its noun phrases, more complex factors can
influence which of several possible strategies the speaker selects.
That's what we see happening in Present Day English in double-object
transitive clauses. Under dative alternation, the constituent that's the
verb's Beneficiary can occur either as a core argument by placing it
immediately after the verb, or else it can be connected via a preposition
such as to or for as an adjunct not an argument.
When you use a preposition to mark the Beneficiary,
ordering is slightly less important because now other constituents can
intervene. But normally we put it after the DO.
Why not direct object first and indirect object second?
In the north of England, there do exist dialects in which the IO-DO order can in
the case of pronouns become DO-IO. This is not grammatical in standard
English, but is normal in those Englishes. This is no longer possible in Modern English because you can’t tell the indirect object from the direct object otherwise. But it used to be possible to use either one without anybody ever getting confused.
Old English marked NP components by grammatical case, but Modern English
does so only when the NP is a personal pronoun, not when it is a determiner
phrase. Old English allowed for either ordering of DO and IO because it
marked each NP via a distinct case inflection.
This example uses dative before accusative:
- hwi God wolde forgifan [þam yfelum mannum]DAT [agenne freodom]ACC
meaning “why God wanted to give evil men their own freedom”. But
this one uses accusative before dative:
- God betæhte [þone wineard]ACC [þam wisum bocerum]DAT.
meaning “God showed the wise scholars the vineyard” but using the other
ordering, so more like “God showed the vineyard to the wise scholars”.
It turns out that the same factors that conditioned which of those two
orderings was selected continue through to this present day. Therefore the
general observation about ordering in systems with or without case
inflection does not take into account extra factors that can influence the
choice a speaker will make when more than one choice is available. Please
see Verb–Object Order in Old
English by S. Pintzuk
for much more about this.
Notes harvested from comments
There was some question about the difference between arguments and adjuncts. I wrote:
- Direct and indirect objects are two elements of a verb's core arguments; the other is its subject. They occupy syntactic slots and are unmarked by other particles. When a noun phrase is the object of the preposition, it's acting as a prepositional object, which is usually an optional adjunct the way it is in John made a cake for Jane, not a required oblique argument the way it is with John put his glasses on the table.
Which one of core argument vs oblique argument vs adjunct doesn't affect actual assignment of the Agent, Patient, and Recipient semantic roles, just the syntax used.
And I provided these two comments to mention some of the historic shifts in the grammars of the Germanic languages:
Primitive Germanic was originally a verb-second (V2) language in the main clause but a verb-final (VF) language otherwise. It was also an object-verb (OV) one, not the verb-object (VO) one that Present Day English has become. Word-order evolution over time was not always connected to loss of inflectional case in that some that lost case kept the old ordering while some that kept the case still lost the old ordering.
Using the original SOV word-order today sounds archaic or poetic: With this ring I thee wed. But OSV order is just a rhetorical focusing device: His name he told them that night; his reasons, never. You can still see V2 ordering under locative inversion, where SV becomes VS because there's still a non-verb constituent before the verb: Down the lane came the firetruck. The many changes to Germanic languages’ grammar—i.e., to their syntax and morphology—is a very broad topic, and much of it is still under active academic investigation by linguists studying Germanic philology.