We construct a framework for meta-analysis and other multi-level data structures that codifies the sources of heterogeneity between studies or settings in treatment effects and examines their implications for analyses. The key idea is to consider, for each of the treatments under investigation, the subject’s potential outcome in each study or setting were he to receive that treatment. We consider four sources of heterogeneity: (1) response inconsistency, whereby a subject’s response to a given treatment would vary across different studies or settings, (2) the grouping of nonequivalent treatments, where two or more treatments are grouped and treated as a single treatment under the incorrect assumption that a subject’s responses to the different treatments would be identical, (3) nonignorable treatment assignment, and (4) response-related variability in the composition of subjects in different studies or settings. We then examine how these sources affect heterogeneity/homogeneity of conditional and unconditional treatment effects. To illustrate the utility of our approach, we re-analyze individual participant data from 29 randomized placebo-controlled studies on the cardiovascular risk of Vioxx, a Cox-2 selective nonsteroidal anti-inflammatory drug approved by the FDA in 1999 for the management of pain and withdrawn from the market in 2004.

The authors argue that children prefer fictions with imaginary worlds. But evidence from the developmental literature challenges this claim. Children's choices of stories and story events show that they often prefer realism. Further, work on the imagination's relation to counterfactual reasoning suggests that an attraction to unrealistic fiction would undermine the imagination's role in helping children understand reality.

Phillips et al. present a number of arguments for the premise that knowledge is more basic than belief. Although their arguments are coherent and sound, they do not directly address numerous cases in which belief appears to be a developmental precursor to knowledge. I describe several examples, not necessarily as a direct challenge, but rather to better understand their framework.

We investigated longitudinal relations among gaze following and face scanning in infancy and later language development. At 12 months, infants watched videos of a woman describing an object while their passive viewing was measured with an eye-tracker. We examined the relation between infants' face scanning behavior and their tendency to follow the speaker's attentional shift to the object she was describing. We also collected language outcome measures on the same infants at 18 and 24 months. Attention to the mouth and gaze following at 12 months both predicted later productive vocabulary. The results are discussed in terms of social engagement, which may account for both attentional distribution and language onset. We argue that an infant's inherent interest in engaging with others (in addition to creating more opportunities for communication) leads infants to attend to the most relevant information in a social scene and that this information facilitates language learning.

Preschoolers participated in a modified version of the disambiguation task, designed to test whether the pragmatic environment generated by a reliable or unreliable speaker affected how children interpreted novel labels. Two objects were visible to children, while a third was only visible to the speaker (a fact known by the child). Manipulating whether a novel object was visible to both interlocutors or hidden from the child tested the child's understanding of pragmatic expectations of interlocutor competence. When interacting with a speaker with a history of accurately labeling familiar objects, children responded appropriately in both cases. Whn interacting with a speaker who previously generated inaccurate labels for familiar objects, children's behavior and eye-movements reflected their belief that the speaker was not a competent communicator. These data support the hypothesis that children consider the pragmatic environment constructed by an interlocutor when that speaker asks them to make a lexical inference.

L. W. Sumner's excellent book Welfare, Happiness, and Ethics has four central goals. He aims (1) to demonstrate the subjectivity of welfare (chaps. 1 to 3); (2) to show that the traditional subjective accounts of welfare—namely, hedonism and preference theories—are inadequate (chaps. 4 and 5); (3) to construct an alternative subjectivist account of welfare (chapter 6); and (4) to argue for the moral primacy of welfare (chap. 7). I will not comment here on Sumner's treatment of the subjectivity of welfare, as I have had my say on this topic elsewhere. In fact, my space here will only permit me to discuss the second and fourth goals of the book. Even with this narrow focus, I will be forced to leave unexplored a wealth of truly lovely work which passes by us on the road to these central conclusions.

Shelly Kagan and Leonard Katz have offered versions of hedonism that aspire to occupy a middle position between the view that pleasure is a unitary sensation and the view that pleasure is, as Sidgwick put it, desirable consciousness. Thus they hope to offer a hedonistic account of well-being that does not mistakenly suppose that pleasure is a special kind of tingle, yet to offer a sharp alternative to desire-based accounts. I argue that they have not identified a coherent middle position.

An adequate moral theory must take (at least) each person into account in some way. Some think that the appropriate way to take an agent into account morally involves a consequentialist form of promoting something about her. Others suggest instead that morality requires a Kantian form of respecting something about an agent. I am interested here in pursuing the former line. When we pursue the broadly consequentialist line we come to this question: what should we promote on the agent's behalf when we are taking her into account morally?