Sentiment Override

One specific threat to the validity of measures in the context of relationship science is a phenomenon called "sentiment override": the tendency for people to project their global, gestalt relationship sentiments onto specific relationship measures they complete. If Rachel is asked how affectionate her partner Sanjay has been over the past week, we would ideally like her to recall her many recent interactions with Sanjay and think about how many hugs, kisses, etc. she received. However, she may instead think to herself, "I love Sanjay a lot. So, he's probably been really affectionate lately."

This poses a problem for the field because it means that measures intended to capture specific constructs (e.g., physical affection) may actually be capturing a more global construct (overall relationship quality). We have several ongoing projects with the goal of detecting and isolating sentiment override from existing relationship measures.

ManyScale: Building Repositories of Self-Report Measures

One promising approach to addressing measurement proliferation is to create open repositories of existing self-report measures (Anvari et al., 2025; Elson et al., 2023). The goal of ManyScale is to help researchers to create these measurement repositories in their own respective subfields. Building repositories by subfield keeps each repository manageable in scope, managed by researchers who are embedded within those communities and who know and use the measures. At the same time, being able to search across a network of repositories will highlight similar constructs and measures used across subfields, hopefully reducing siloing.

As proof of concept, our lab has created RelaScale: a live, working repository of measures for relationship science (relascale.com). Researchers can submit their published measures for inclusion by filling out a form with relevant meta-data (e.g., primary citation, constructs being represented), as well as a document containing the verbatim items, instructions, and scoring. Users can then search within the repository to find relevant measures and constructs.

RelaScale's code and workflow can be easily duplicated for use in other subfields within the same web environment. Thus far, we have teamed up with collaborators who are working on several additional repositories, representing measures pertaining to aggression (AggreScale), political psychology (PolitScale), emotions (EmotoScale), and wellbeing (WellScale). We are actively seeking new collaborators to help expand the network further. By helping researchers find relevant measures, we hope these repositories will reduce the incentives for researchers to create new, ad-hoc measures, while also improving measurement standardization and facilitating meta-science.