The double object construction, also known as the ditransitive construction, has been the focus of much analysis in syntax. In English, this construction can be expressed in two ways: direct object + complement, and indirect object + direct object. This is exemplified by the following two sentences:
Peter gave the ball to Andrew.
Peter gave Andrew the ball.
Some syntacticians argue that the sentence "Peter threw Andrew the ball" is underlying and thus "Peter threw the ball to Andrew" is derived from "Peter threw Andrew the ball". However, an analysis of double object constructions in other languages makes it difficult to make this determination. Also, many syntacticians who argue that "Peter threw Andrew the ball" is underlying do not refer to historical linguistics nor to child language acquisition to support their claim.
I prefer to argue that both structures are common and serve different purposes. For example, the sentence "Peter threw the ball to Andrew" has a different semantic meaning from "Peter threw Andrew the ball". In the first sentence, it is not clear whether or not Andrew has caught the ball. The completeness of the action is not expressed. In the second sentence, the action has been completed. It is clear that Andrew has caught the ball.
With certain verbs such as "lift", English does not allow both double object constructions. We can use an asterisk to mark that one sentence is ungrammatical:
Peter lifted the box to Andrew.
*Peter lifted Andrew the box.
If forms such as "Peter gave Andrew the ball" are underlying, it is difficult to explain why "Peter lifted Andrew the box" is ungrammatical. We know that in Romance languages, the double object construction is expressed with the preposition "to" as in "Peter gave the box to Andrew". For example, in French this is "Pierre a donne la boite a Andres". In Hungarian, "to Andrew" is expressed with a postposition suffixed to "Andrew". Double object constructions can vary greatly from one language to another.
The construction that a person uses tends to be highly influenced by the form that the first speaker uses in a conversation. For example, if speaker A asks: "Did Mark give Linda flowers?", it is likely that speaker B will answer, "Yes, he gave Linda flowers" or if pronominalization is used, "Yes, he gave her flowers." Likewise, if speaker A asks: "Did Mark give flowers to Linda?", it is likely that speaker B will answer, "Yes, he gave flowers to Linda" or with pronominalization, "Yes, he gave flowers to her". This is because of the desire to establish solidarity with the first speaker.
In English, though, it is interesting that the question "Did Mark give flowers to Linda?" is never replaced with *"Did Mark give her them?" English does not like to place two pronouns next to one another. For this reason, it uses the structure "Did Mark give them to her?" If forms such as "Mark gave Linda flowers" are underlying, it is difficult to explain why this structure does not result when the direct and indirect objects are pronominalized.
In questions with the interrogative pronoun "who", only one construction is used. The question *"Who did he give them?" is ungrammatical. Only "Who did he give them to?" or "To who did he give them?" is possible. This also counters the theory that forms such as "Mark gave Linda flowers" are underlying in English.
I do not agree with syntactic analyses which claim that the English double object construction has one form which is underlying and that it is this form from which the other is derived. Rather, I prefer to state that the two forms serve different purposes and are both very common. One may be more common than the other but they are both frequently used in English. Many examples can be provided to counter the claim that forms such as "Peter gave Andrew the key" are underlying.