--- title: "iobj: indirect object — definition and key papers" description: "The indirect object (iobj) is the Universal Dependencies (UD) relation linking a predicate to a nominal core argument that is neither its subject…" canonical: "https://research.elamur.ai/free/en/concepts/ud-iobj/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/concepts/ud-iobj/) [home](https://md.research.elamur.ai/free/en/) / [concepts](https://md.research.elamur.ai/free/en/concepts/) / iobj: indirect object # 🌐 iobj: indirect object Universal Dependencies · 6 key papers · [all concepts](https://md.research.elamur.ai/free/en/concepts/) Part of: [🗨️ Natural language processing](https://md.research.elamur.ai/free/en/concepts/natural-language-processing/) · [🌐 Universal Dependencies](https://md.research.elamur.ai/free/en/concepts/universal-dependencies/) The indirect object (iobj) is the Universal Dependencies (UD) relation linking a predicate to a nominal core argument that is neither its subject nor its direct object. Typically this is the recipient, addressee or beneficiary in a ditransitive clause, such as *me* in *she gave me a book*. Also: dative object, IO ## What it is In [Universal Dependencies](https://md.research.elamur.ai/free/en/concepts/universal-dependencies/), every word attaches to exactly one head through a typed relation, and the main predicate attaches to an artificial root node. Content words serve as heads, and function words depend on them ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). `iobj` belongs to the core arguments. These are the small set of dependents that a predicate selects and that a language codes as central rather than oblique. Like `nsubj` and `obj`, it is kept apart from `obl`, which covers both oblique arguments and adjuncts ([Nivre et al. 2016](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/); Zeman 2017). Which coding counts as core is decided language by language. A bare nominal or a dative case can be core in some languages, while adpositions and other oblique cases mark an `obl` dependent (Zeman 2017; [Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). The head is usually a verb, and the dependent is a nominal: a noun, a proper noun or a pronoun. UD makes the content word the head and lets the copula depend on it ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). An adjectival predicate can therefore host the relation, as in French *Il **lui** est fidèle* "He is faithful to him/her", iobj(fidèle, lui). The relation is defined against its neighbours: - `obj` marks the direct object. When a clause has only one object, that object is normally `obj`, even if it denotes a recipient. - `obl` marks a nominal introduced by an adposition or an oblique case. English *to Mary* in *gave a book to Mary* is therefore `obl`, not `iobj` ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). - `nsubj` marks the subject. `expl` marks expletives, which are semantically empty fillers such as *it* in *it is raining*. It also marks certain clitics, for example inherent reflexives, that occupy an argument position without being a semantic argument. - `ccomp` and `xcomp` mark clausal complements. The indirect object relation is reserved for nominals. The decision rests on how an argument is coded, not on its semantic role. The recipient role links the relation to [semantic role labeling](https://md.research.elamur.ai/free/en/concepts/semantic-role-labeling/), but the label itself is syntactic. Subtypes are permitted in UD v2 only as language-specific refinements of the universal set ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). Refinements of `iobj` itself are rare and treebank-specific, and none has universal status. ## Examples - *She gave **me** a book.* — iobj(gave, me), obj(gave, book) - *They sent **the committee** a letter.* — iobj(sent, committee), obj(sent, letter) - French *Je **lui** donne le livre* "I give him/her the book" — iobj(donne, lui), obj(donne, livre) Compare *She gave a book **to me***. Here *me* carries a preposition, so it is `obl` ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). ## Across languages Languages code the indirect object differently. Some use a dative case, some use word order, and some use clitic pronouns. Others use only prepositions, and in those languages the relation barely occurs. UD's typological framing compares these strategies within a single inventory of relations ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/); Nivre 2025). In [English](https://md.research.elamur.ai/free/en/topics/english/), the relation corresponds to the first object of the double-object construction (*gave **me** a book*). The prepositional variant goes to `obl` ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). In [French](https://md.research.elamur.ai/free/en/topics/french/), the indirect object is realised mainly by dative clitics: *lui*, *leur*, and *me* and *te* when dative. Full nominals with *à* go to `obl`. Some French treebanks give them the language-specific subtype `obl:arg`, which refines `obl` rather than `iobj` ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). For Portuguese, the boundary between indirect objects and adverbial adjuncts has been contested. The borderline between `iobj` and adverbial obliques has been re-examined directly (Souza & Freitas 2022). Some predicates are not prototypical verbs, as in the French adjectival case above. There it becomes a question which word heads the clause and therefore hosts the indirect object (Nivre et al. 2026). ## In treebanks and research ### Origins `iobj` was inherited from the universal Stanford dependencies ([Marneffe et al. 2014](https://md.research.elamur.ai/free/en/papers/marneffe-2014-universal-stanford-dependencies-cross-linguistic/)). It belonged to the first UD inventory, where the direct object was still called `dobj` ([Nivre et al. 2016](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/)). UD v2 revised the guidelines and widened the use of subtypes ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). Among the changes, `dobj` became `obj`, and the old `nmod` was split. Nominals attached to predicates became `obl`, while `nmod` was kept for dependents of nouns. In v1, *to me* in *gave a book to me* was therefore `nmod`; since v2 it is `obl` ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). This split is what now separates prepositional recipients from the indirect object. ### Parsing and consistency In [dependency parsing](https://md.research.elamur.ai/free/en/concepts/dependency-parsing/), the relation must be told apart from `obj` and `obl` on the basis of coding that differs from language to language. Cross-treebank checks have assessed how consistently the UD corpora are annotated ([Marneffe et al. 2017](https://md.research.elamur.ai/free/en/papers/marneffe-2017-assessing-annotation-consistency-universal-dependencies/)). Enhanced UD is an optional graph layer built on top of the basic tree ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). It has been specified in detail for English only, keeping the basic tree while adding edges and refining labels ([Schuster & Manning 2016](https://md.research.elamur.ai/free/en/papers/schuster-2016-enhanced-english-universal-dependencies-improved/)). For indirect objects this has two effects. A dependent shared by coordinated verbs, as in *She bought and gave **me** a book*, is propagated to each conjunct. A prepositional recipient carries its preposition in the label, which was `nmod:to` in that English specification ([Schuster & Manning 2016](https://md.research.elamur.ai/free/en/papers/schuster-2016-enhanced-english-universal-dependencies-improved/)). ### Indirect objects in neural models Outside treebanks, "indirect object" names a benchmark task, indirect object identification (IOI), used in mechanistic [interpretability](https://md.research.elamur.ai/free/en/topics/interpretability/) research. In IOI, a model completes sentences such as *After John and Mary visited the shop, John handed a drink to …*. A circuit of attention heads that carries out IOI has been reverse-engineered in GPT-2 small (Wang et al. 2022). The target there is a prepositional recipient, which UD itself would label `obl`. How often the relation iobj occurs: its share of all tokens Counted by code over the Universal Dependencies 2.18 treebanks of English and French (every token once, subtypes with their base relation). **The numbers** | Treebanks | Share | | --- | --- | | English treebanks | 0.34% of tokens (3,781 tokens) | | French treebanks | 0.62% of tokens (6,742 tokens) | What iobj joins: the most frequent head tag → dependent tag The five most frequent in each language's treebanks, by tokens. **The numbers** | Pair | Tokens | | --- | --- | | English: VERB → PRON | 3,131 | | English: VERB → NOUN | 346 | | English: VERB → PROPN | 289 | | English: VERB → DET | 4 | | English: NOUN → PRON | 3 | | French: VERB → PRON | 6,483 | | French: ADJ → PRON | 90 | | French: NOUN → PRON | 87 | | French: ADV → PRON | 37 | | French: VERB → ADV | 22 | ## Key papers The works this page cites, the most cited first; citations as counted by OpenAlex plus the papers of this library that cite the work, read 2026-10-24. - [Universal Dependencies v1: A Multilingual Treebank Collection](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/) — Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter et al. · 2016 · 1,741 citations - [Universal Stanford dependencies: A cross-linguistic typology](https://md.research.elamur.ai/free/en/papers/marneffe-2014-universal-stanford-dependencies-cross-linguistic/) — Marie-Catherine de Marneffe, Timothy Dozat, Natalia Silveira et al. · 2014 · 627 citations - [Universal Dependencies](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/) — Marie-Catherine de Marneffe, Christopher D. Manning, Joakim Nivre et al. · 2021 · 480 citations - [Enhanced English Universal Dependencies: An Improved Representation for Natural Language Understanding Tasks](https://md.research.elamur.ai/free/en/papers/schuster-2016-enhanced-english-universal-dependencies-improved/) — Sebastian Schuster, Christopher D. Manning · 2016 · 336 citations - [Universal Dependencies v2: An Evergrowing Multilingual Treebank Collection](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/) — Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter et al. · 2020 · 325 citations - [Assessing the Annotation Consistency of the Universal Dependencies Corpora](https://md.research.elamur.ai/free/en/papers/marneffe-2017-assessing-annotation-consistency-universal-dependencies/) — Marie-Catherine de Marneffe, Matias Grioni, Jenna Kanerva et al. · 2017 · 30 citations ## Related concepts - [Universal Dependencies](https://md.research.elamur.ai/free/en/concepts/universal-dependencies/) — Universal Dependencies is a framework and an open collection of treebanks that annotates the grammar of many languages with one shared… - [Dependency parsing](https://md.research.elamur.ai/free/en/concepts/dependency-parsing/) — Dependency parsing is the task of recovering a sentence's syntactic structure as directed, labelled links between its words. Each word is… ## More papers Topics: [🌐 Universal Dependencies](https://md.research.elamur.ai/free/en/topics/universal-dependencies/) · [🌳 Dependency parsing](https://md.research.elamur.ai/free/en/topics/dependency-parsing/) Shelves: [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) - [← previous concept Inverted index](https://md.research.elamur.ai/free/en/concepts/inverted-index/) - [next concept → Knowledge graph](https://md.research.elamur.ai/free/en/concepts/knowledge-graph/) - [↑ back to concepts](https://md.research.elamur.ai/free/en/concepts/)