The current Icelandic dataset contains the replication data for Somers & Barðdal (2022) and Somers, Jenset & Barðdal (2024a). The data have also been used in Somers & Barðdal (2023), Somers, Jenset & Barðdal (2024b) and Elens, Somers & Barðdal (2024). The dataset, which has been compiled using the Icelandic Web 2020 corpus, or isTenTen20, contains 200 observations each for 15 Icelandic verbs, thus amounting to a total of 3,000 observations. The verbs in question belong to one of three syntactic classes: (1) Dat-Nom verbs, (2) Nom-Dat verbs and (3) Dat-Nom/Nom-Dat verbs. All tokens have been annotated for lemma, verb class and the internal order of the dative and the nominative arguments. In addition, each constituent, dative and nominative, has been annotated for (1) case, (2) (pro)nominality, (3) pronoun type (if applicable), (4) referentiality, (5) person, (6) number, (7) definiteness, (8) animacy and (9) length.