The Role of Animacy in the Acquisition of Sluiced Questions and Relative Clauses


Victoria Mateu, Nina Hyams



(424) 653-8327, (323) 497-5472


UCLA Department of Linguistics


Lecturer (V. Mateu), Full Professor (N. Hyams)




Mackenzie Lighterink - Undergraduate Research Assistant


Studies show that children under the age of 6 have greater difficulty comprehending sentences that question objects (e.g., Who did Mary call __?) than those that question subjects (e.g., Who __ called Mary?). In this study we investigate the acquisition of two constructions that involve subject/object gaps –wh-questions that contain elided material, a.k.a. sluiced questions (e.g., Someone called Mary, do you know who?), as well as relative clauses (e.g., I know the man that __ called Mary). We aim to investigate two key questions: (i) will children show a correlation in performance between these two constructions? – evidence for theoretical analyses that relate the two constructions in linguistic/mental representations, and (ii) does children’s better performance with subject-related constructions arise from the shorter distance between the moved element and where it was originated (indicated with underscore above), or due to the connection children make between subjects and animate agents? (i.e., if who-question choose the animate subject, if what-question choose the inanimate object).


In this study, we investigate children’s understanding of ellipsis and relative clauses. Ellipsis refers to the omission of part of a sentence that are understood through the linguistic context, e.g. previous sentence. For example, the elided portion of sentence (1a) is interpreted as the string in angle brackets in (1b).

(1) a. Someone called Mary, do you know who?
b. Someone called Mary, do you know who <__ called Mary>?

Sentence (1a) exemplifies a phenomenon called ‘’sluicing’’ in the theoretical linguistics literature, that is, ellipsis of part of a wh-question.

Relative clauses, on the other hand, refer to a type of subordinate clause, such as the one in the bracketed portion of (2).

(2) I know the person [that _ called Mary].

Both (sluiced) questions and relative clauses can be classified as involving the subject of the embedded clause, as in (1) and (2), or the object of the embedded clause, as in (3) and (4)

(3) a. Mary called someone, do you know who?
b. Mary called someone, do you know who <Mary called __>?

(4) I know the person [that Mary called __].

Results from a previous study on sluicing (conducted at the Lab School) show that children aged 3-5 have an easier time interpreting (sluiced) subject wh-questions, e.g., (1a), as opposed to (sluiced) object wh-questions, e.g., (3a). However, it is yet to be determined whether their performance is better because of the shorter distance between the wh-element and the position in which it was originated – indicated with an underscore in (1) and (3), or because of the connection between animacy and subjects, i.e., subjects tend to be animate, therefore ‘who’ questions are expected to ask about the subject/agent of the sentence.


In this study we aim to investigate whether sluiced subject wh-questions and subject relative clauses are easier to comprehend than sluiced object wh-questions and object relative clauses across the board, or only when the subject is animate. We will manipulate the animacy feature of both the subject and object:

(5) a. [+ animate] subject, [+animate] object
b. [+ animate] subject, [-animate] object
c. [- animate] subject, [+animate] object
d. [- animate] subject, [-animate] object

If children do better with subject constructions regardless of animacy, we may deduce that limited verbal working memory is a source of children’s difficulties. If children show sensitivity to animacy, it may suggest that animacy is a relevant feature in children’s linguistic systems.

Additionally, we will compare and contrast children’s performance on sluiced wh-questions and relative clauses. This will provide new empirical evidence relevant to theoretical analyses that relate the underlying, mental derivation of these two structures.


There are no direct benefits to the UCLA Lab School. However, this research addresses important issues in the study of language acquisition that is deemed to be of great interest in the linguistic community. More generally, investigating the developmental profile of constructions that are difficult to acquire can tell us about the underlying grammar and human language, in general. More specifically, sentences that leave recoverable information unsaid are ubiquitous in both the speech of children and adults. However, there has been very few previous studies on how English-speaking children interpret sentences with sluicing (i.e., Wood 2009, and our own from 2017). A better understanding of this topic could provide insight into children’s cognitive ability to recover information from a context and integrate that information into a sentence structure. Moreover, knowing which types of errors children produce and comparing their performance to other similar constructions can provide insight into the source of difficulty, whether grammar-based or due to limits on working memory. An answer to this question could also help speech pathologists to develop more efficient therapies. For example, children with Specific Language Impairment (SLI) or adult aphasics who experience problems with these constructions could find improvement by doing certain memory exercises.


Results from our last experiment conducted at the Lab School have been presented in 3 different conferences: Generative Approaches to Language Acquisition 13 (Mallorca, Spain, Sept. 7-9); Boston University Conference on Language Development 42 (Boston, MA, Nov. 3-5); California Meeting on Psycholinguistics 2017 (Los Angeles, CA, Dec. 2-3). A proceedings paper is also going to be published in the next few months. Our work received excellent reviews from other researchers and theoreticians in the field and we hope to continue to investigate this question. Upon completing our data collection, we expect to submit two new abstracts to two different conferences and write a journal paper with the results from both experiments.




Monolingual English-speaking children aged 3-6 with no diagnosed speech impairment. We need at least 10 children per year interval.


We will show the child images on a laptop. An interactive puppet will ask the child a question about the image and he or she will have to point at the character that the puppet is referring to. The study will be conducted individually. The study will consist of two sessions of 15-20 minutes each. The first session will test sluiced wh-questions. The second session will test relative clauses.





In this study the child listens to a story that is accompanied by child-friendly images on a computer screen. An interactive puppet describes the picture and is asked a yes/no question or asked to point to a character on the screen. The child's response is audio recorded and transcribed in writing. The experiment session has six training items (practice with how the experiment will work) and 24 test items (in each session). Children will be divided in two groups.


If they are more likely to make more errors with object extracted constructions (e.g. I can see that the boy is poking someone, can you see who [the boy is poking __]?) than subject extracted nouns (e.g. I can see that someone is poking the boy, can you see who [__ is poking the boy]?) it will suggest that children have difficulties with longer extractions due to memory interference. If we find they make more errors with object extracted constructions only when the subject is animate, but not when the subject is inanimate (e.g., I can see that the tree is poking someone, can you see who [the tree is poking __]?) it will suggest that animacy is a relevant feature in children’s linguistic systems.


Yes. The IRB-approved consent form is attached at the end.


There are no potential risks associated with this study. If the child seems tired or distracted, we take breaks or can stop the experiment.


There is no deception or debriefing.


The data is associated with a subject number, rather than the child's name. The data and consent forms are stored in a locked cabinet (and room) in the UCLA Language Acquisition Lab.



Ms. Sandra Smith (in 2016-2017)






Any quiet corner will suffice.


Linguistic Development (see Summary Info for more details).


We will only need the information that is requested from the parents in the consent form, i.e., the child’s date of birth and the child’s language background (monolingual English, mostly exposed to English but also exposed to another language, or mostly exposed to a language other than English).


We will bring our own equipment (laptop and microphone). It would be preferred to conduct the study in a quiet corner or room so the child doesn't get distracted.










see attached stimuli