<h2>ManyBabies 1: Infant-Directed Speech Preference</h2>
<p>This is the project page for the first study of the ManyBabies collaborative research project.</p>
<p>Data collection began on May 1, 2017 and continues until April 30, 2018. We are still soliciting contributions from new labs joining the project! If you are interested, please fill in this <a href="https://docs.google.com/forms/d/e/1FAIpQLScitS5enkjbdYa3OBLN5hu8vNUVGkrsZB6C3eyhJaD6sVHysg/viewform?ts=577e76d4" rel="nofollow">expression of interest form</a>.</p>
<p>You can also join the manybabies listserv (<a href="" rel="nofollow">email@example.com</a>).</p>
<p>The recent focus on power, replication, and replicability has had important consequences for many branches of psychology. Beliefs in influential theories and confidence in textbook experiments have been shaken by repeated demonstrations that much of the experimental literature is underpowered, that surprisingly few empirical claims are subject to direct replication, and that, when direct replication does occur, those empirical claims are often found wanting. Still, these pessimistic conclusions have had important positive consequences, as scientific organizations, journals and researchers have worked to improve the transparency and replicability of psychological science.</p>
<p>In developmental psychology these issues have been much less discussed. The typical power associated with most developmental psychology experiments is unclear, and the reproducibility of many important findings is uncertain. These issues arise in part due to a set of quite particular challenges that are faced when collecting and interpreting developmental data. For example, it is very costly to replicate experiments conducted on infants, and it is very hard to interpret contradictory findings given how children’s behavior and development can vary across ages, cultures, and socioeconomic groups. But while these challenges may mean that discussions about issues like power and replicability are difficult, they do not mean that such discussions are unimportant. </p>
<p>On the contrary: in recent a handful of recent meta-analyses of phenomenon in infant development, statistical power has varied widely, sometimes being as low as 30% (see the <a href="http://metalab.stanford.edu" rel="nofollow">MetaLab project</a>).</p>
<p>Inspired by the ManyLabs projects (Klein et al., 2014), we aim to provide an empirical basis for understanding how power and replicability affect developmental psychology, and the study of infant cognition in particular. Our current study examines consistency and variability in data collected from multiple different labs. </p>
<p>But ManyBabies differs from ManyLabs in that we focus on what developmental psychology can learn by assessing the replicability of a single important phenomenon, infants’ interest in infant-directed speech over adult-directed speech, which we will study across a range of ages (from 0 months to 15 months), cultures (among others North America and Europe), and linguistic backgrounds.</p>
<p>We have taken this “single study” approach for two reasons. First, because the costs and difficulties of replicating multiple infant studies in multiple groups are simply too high for an initial investigation. Second, because focusing on a single, well-studied phenomenon allows us to address a series of important issues about how statistical power interacts with developmental methodologies, particularly looking time experiments. In particular:
a) What should we expect the bounds of replicability to be in infant cognition? Infant-directed speech preferences are known to be replicable and robust. By analyzing the strengths of these preferences across ages, cultures, and laboratories, we can better estimate the conditions under which we should, and should not, expect other results in developmental psychology to replicate.
b) How should we think about the relationship between statistical significance and developmental change? For example, when a reliable effect is found in 7-month-old infants but not 6-month-old infants, that finding is often taken to reflect developmental change. However, it could also reflect a failure to replicate the phenomenon in the younger group or, in the worst case, a false positive result in the older group. By more precisely measuring how effect sizes change over age (and also across cultures), the data from ManyBabies can clarify how future work should approach this issue.
c) In the same vein, we can ask how standard looking time measures (such as familiarity preferences) change over development? For example, if infants of different ages show differently strong looking preferences, then it would be important to account for this when estimating power and computing sample sizes in future work.
d) How should we estimate statistical power for infant looking time experiments? Publication bias, the tendency to only report significant results, may have led the field to overestimate the typical effect size in infant research. While the current pre-registered project will provide an unbiased estimate of effect size for only one important phenomenon, it will indicate how we should expect effect sizes to vary across age and culture, which is important information for scientists planning future work.
e) What are the most appropriate statistical analyses for looking time data? Are they best analyzed with parametric or non-parametric statistics, and what types of data transformation are most appropriate?</p>
<p>This broad replication of the Infant-Directed Speech preference will therefore help to answer basic questions about the replicability of developmental psychology, and will also provide useful positive advice for how to design infant cognition studies going forward. Just as projects such as ManyLabs have led to important positive improvements to research practice in cognitive and social psychology, we hope that ManyBabies will play a similar role for developmental science.</p>
<p><strong>Note on authorship</strong>: Final authorship on the project will be determined based on final participation in the project via a policy that will allow authorship based on contribution of data, analytic effort, or other organizational contributions. </p>
<p><strong>Citation for the original article</strong>: Cooper, R. P. & Aslin, R. N. (1990). Preference for Infant-Directed Speech in the First Month after Birth. <em>Child Development, 61</em>, 1584-1595.</p>