To facilitate theory generation and replication, we include a Statement of Limits of Generality (SLOG). For more informmation on the SLOG, see [here][1]. **Materials/Stimuli** - *Will the effect generalize to other materials? What are the critical features of the materials that must be maintained to obtain the same result? What manipulation checks are necessary (if any) to verify that they are comparable to the originals?* It is clear from our paper that our effects pertain to three items of the ECR. We thus expect that at the very least these items will replicate in the future. That being said, we expect that the underlying mechanism relates to *social outsourcing*. Thus, better items can and should be created in future research. Manipulation checks could be included (e.g., to see whether higher temperatures were noticed by participants) but we fear that such manipulation checks elicit associative networks as they pertain to warmth and sociability. --------- **Subjects** - *Define the target generalization population. If you are trying to generalize to college undergraduates, then your study might not be expected to generalize beyond undergraduate participants. If you are trying to generalize to all adults, then the fact that your participants were testing of undergraduates should not be essential for your effect. What manipulation checks would someone need to conduct in order to verify that their subjects are drawn from a comparable population?* Our effects were obtained with undergraduates. We expect them to generalize beyond undergraduates. That being said, thermoregulatory abilities start to change with old age. We thus do not yet have clear hypotheses how these effects vary amongst for example the elderly. Finally, as we indicate above, a manipulation check for these populations is harder. What future researchers should check if they want to investigate a main effect is the distribution of people who are more versus less secure in their RCR. --------- **Task procedures** - *What aspects of the task procedures must be followed exactly in order to obtain the effect? What would other researchers need to do in order to follow those procedures (i.e., would they need to test participants in your lab? Would they need your code?). If they would need to follow procedures exactly, have you provided everything they need to do so? For those procedures that need not be followed exactly, what manipulation checks might be required?* We describe our procedure elsewhere. Other researchers should ascertain that participants held the cup for at least 30 seconds (note: so far, this time is arbitrary). If participants quickly put the cup down, they should be excluded from analyses. Researchers should also probe participants for awareness of the hypotheses; more and more psychology students are getting aware of these findings. -------- **Testing context** - *To what extent does the study result depend on duplicating the original testing context? For example, must participants be isolated from each other in cubicles? Will the study work only in a large classroom or only at a large university? Will it work only in a lab or will it also work online? Does the study require an unusual testing setting (e.g., an airport, a shopping mall)? Is it likely that the personality and skills of the task administrator affect the results obtained? Is the effect culturally specific? What aspects of the context are unlikely to matter, and what manipulation checks (if any) would be required to verify the correct context?* We think that these effects may lessen the closer one gets to the equator. Languages around the equator lack metaphors tying affection to warmth (cf. Koptjesvkaja-Tamm, 2015). Although absence of the metaphor does not necessarily imply absence of the effect, it may well be that effects are less, or absent. We did run both tests in this case in busier environments, and the effects held up. We do recommend the task administrator to keep interaction with the participant the least as possible, because we think that the interaction with the participant will matter. This is also the reason most instructions were provided via qualtrics, and why our second experimenter took care of the temperature condition. ------- **Historical/temporal specificity** - *Does the effect depend on timing relative to significant events in society (i.e., right before or after an election)? Does it depend on cultural norms that might change over time? What aspects of the temporal or historical context need to be stable to observe the effect? For example, attitudes about same-sex marriage in a study conducted in the 1990s might differ from those in a study conducted in 2016. Can you anticipate any such changes that might affect whether or not you would observe your effect?* In part, we are unsure about this. The underlying construct, we suspect, is "outsourcing" to others. But how people understand outsourcing may be culturally variant. Ideally, future researchers develop questionnaries that deal with "outsourcing" and examine how culturally variant this variable is. If people conduct follow up research with the current materials, we recommend the following: 1. Ask the full 36-item ECR. 2. Run regressions with main effects (temperature condition, RCR) and interaction term with the 3-item RCR. 3. In case 2 does not replicate *in another cultural context*, repeat our exploratory analyses, and seek to detect a comparable interaction as we detected (greater accessibility for warm for insecures, and for cold for secures). 4. Replicate the analyses, and seek whether cultural differences can explain differences with our study. The differing cultural/temporal context is a prerequisite. If the effects do not hold in the same context, they simply do not hold. We realize that this is a posthoc approach. However, we think that our "outsourcing idea" received strong support in the first report, but also that our outsourcing idea still requires considerable investigation. [1]: https://docs.google.com/document/d/17KS4zU7--Gi5HqiOn-tIoSFMFTGaBDgO6PraSXFSTAM/edit