Typically, in an inflected Indo-European language such as Latin or Ancient Greek, transitive verbs take a direct object in the accusative case. But these languages will often also possess some verbs which take dative or (more rarely) genitive objects. The common understanding is that this is caused by case syncretism.
Proto-Indo-European had at least eight noun cases. Among these were an ablative case (often used to express motion away from something), an instrumental case (used to express an action taken by or with an object), and a locative case (used to express that the subject is in a certain location). Indo-European languages merged these cases to varying degrees. Latin, for instance, retained the use of the locative and ablative. But the instrumental case was merged with the ablative, leading to such constructions as liber a discipulo aperitur [‘the book by the student is being opened’]. This merging of cases is known as ‘case syncretism.’
Proto-Indo-European also had a dative case, originally used to denote the indirect object of a verb. Proto-Germanic merged the ablative and locative cases into the dative case. Old Norse further merged the instrumental case into the dative case. There are several explanations for this sort of case syncretism, including historical phonology and a loss of verbal prefixes Proto-Germanic used to indicate the appearance of a dative, instrumental, or ablative object (Holland 21).
By this argument, the dative form beinum in the phrase ‘hann kastar beinum smám um þvert gólfit’ [‘he throws small bones across the floor’] is a remnant of an instrumental sense in which throwing was conceived (compare the Modern English phrase ‘he splashed the ground with water’). But there are many examples that implied instrumentality does not satisfactorily explain:
bregða augum sundr [‘to open the eyes’]
skjóta hesti undir einhvern [‘to put a horse under someone, to mount someone’]
róa báti [‘row a boat’]
hestrinn varp honum af baki [‘the horse threw him from (its) back’]
verpa eggjum [‘to lay eggs’]
…the hunt for surviving traces of earlier case values in such examples has obscured the basic fact that these Old Norse verbs with dative objects are functionally equivalent to ordinary transitive verbs with accusative objects. (Holland 24)
Holland’s proposed solution is that such uses are constructions, and optional ones at that, used to indicate a causal relationship between subject and object. “The unifying semantic feature of this construction is that all of these verbs impart motion to, or control the motion of, their dative objects.” (Holland 26) A key feature of Holland’s argument is that this relationship was indicated by a suffix in Indo-European and Proto-Germanic which was lost in the Northern Germanic dialects. Adapting the dative case to indicate this relationship may have been a strategy adopted by the speakers of Old Norse because it was “synchronically opaque… it was simply the dative…” (Holland 29).
Barnes, Michael P. A New Introduction to Old Norse. I, Viking Soc. for Northern Research, 2008.
Holland, Gary. "Transitivity, Causativity, and Surface Case in Old Norse". Arkiv för Nordisk Filologi 108 (1993) 19-37