Reproducing experiments on early verb understanding in infants

Ayer V, Witte C, Cimiano P, Rohlfing K, Nomikou I. Reproducing experiments on early verb understanding in infants. In: Cimiano P, Pietsch C, Wiljes C, eds. Studies in Analytical Reproducibility: the Conquaire Project. Bielefeld: Universität Bielefeld; 2021: 87-96.In this chapter, we describe an effort to reproduce the main result of the paper “Evidence for early comprehension of action verbs” by Nomikou et al. [1]. The study aimed at investigating the ability of 9-and-10-month old infants in understanding verbs. The study in question followed a so called preferential looking paradigm consisting in investigating the ability to understand the meaning of words by testing whether infants look longer at a related stimulus compared to an unrelated stimulus. As a method to track looking time proportions to the target stimulus, an eye tracker was used. Data were collected from 9- to 10-month old infants who were presented with paired-picture trials while listening to corresponding verbs. The infants saw two images on the screen side by side, each one from a different context category (CARE or PLAY). One of the pictures was related to the verb in question, while the other image was a confounder. The percentage of time that infants looked at the matching picture, before and after having heard the verb, was recorded across participants, computing a difference score. In case the difference was positive, this was taken as evidence of understanding the meaning of the verbs. The study could only find a positive difference for 10-month olds but not for 9-month olds, showing that the ability to understand the verb in question emerges between 9 and 10 months. In close interaction with the authors of the original paper we rewrote the analysis scripts which were used by the authors to refine the results during a second iteration of reviewing in response to requests by reviewers. Overall, we could reproduce the central results of the study. This case represents a case of full analytical reproducibility. The data and scripts for the paper described above can be found at https://gitlab.ub.uni-bielefeld.de/conquaire/psycholinguistics