It’s amazing how many topics and research areas largely just boil down to attitude valence.

My personal hope is that we can refine our measures to do a better job of tapping into different constructs as intended, in part by paying more attention to process evidence. How are participants actually interpreting and responding to these items? And also by statistically parsing out biases/glop

Definitely- there are disadvantages to trying to measure so many things at once. In another project, led by Victoria Pringle, we do have more evidence for the phenomenon in samples with fewer relationship constructs (and collected from dyads too)

Yeah, it's a great question. I definitely think this is a measurement issue. Theoretically, these concepts are distinct, absolutely. But by and large, our measures are failing to sufficiently detect those differences.

Kind of a wild finding, that in hindsight nonetheless makes sense. How many other “multidimensional” constructs boil done to a positive versus a negative evaluation?

I was STUNNED that this model ended up being the winner. No Big 5, or Big 3, or Big Anything. Relationships measures capture positivity, layered on top of item-specific content. That's it.

It feels like we need to start developing new measures from square one.

I think it's a problem in a lot of areas. In general, social psychology just hasn't seriously grappled with shared method biases.

Totally. I'm convinced that it feels disloyal to a lot of participants to report on negative aspects of their otherwise good relationship - as if they're giving their partners a "bad grade"

Thanks Mark!

All the credit goes to James (who's not on bsky)

Big shoutout to our coauthors Ariana González, Brett Murphy (@brettmurpsych.bsky.social), Jacqueline Pérez, Victor Kaufman, Tom Bradbury, Paul Eastwick (@pauleastwick.bsky.social), and Ben Karney. And of course, the brilliant lead author, James Kim.

Results highlight the validity threat that sentiment override poses to relationship measures, and emphasize the need for careful measurement validation work in our field, as in other fields of psychology. (11/12)

There was little to no evidence of any dimensionality above and beyond this global factor, which appears to capture, essentially, relationship vibes. (10/12)

In sum: across two studies, using preregistered bifactor models examining more than 30 prominent relationship constructs, a single, global factor explained about 70% of the common variance across all of the items. (9/12)

Top-loading items for the general factor included, "all things considered, I'm very happy in my relationship with my partner", "my relationship with my partner is enjoyable", and "my relationship with my partner is rewarding" (8/12)

Across both studies, it was this null model that we found consistent evidence for. All the measures primarily loaded onto a single, global factor, which appears to represent general relationship evaluations. (7/12)

A structural equation model in which many different measures load onto a single, general factor

And then finally, it was possible we would find that all of the different relationship measures primarily tap into a single, global construct, with no distinct, substantive factors remaining. (6/12)

A bifactor model in which many different measures cluster into several different, specific latent factors, as well as an overarching, global factor.

A related possibility was that we would find support for a bifactor. With this model, you might see support for several distinct domains, as well as a general factor that taps into something more global, like people's overall satisfaction with their relationships. (5/12)

A structural equation model in which many different measures cluster into a handful of different latent constructs

Another possibility - what we expected/hoped to find - was that the items would cluster into smaller sets of distinct domains. Maybe we'd see a Big 5, or a Big 4 for relationship facets. (4/12)

A structural equation model in which each measure corresponds to a different latent construct.

We preregistered several competing hypotheses for these models. One possibility was that each measure would correspond to a single, different latent construct. This is the model implicitly assumed in the literature. (3/12)

James analyzed two large, US census-matched samples of people in relationships, recruited with online panels. Each sample was given a battery of close relationships measures.

Study 1 (N = 2000 on Dynata): 206 items, 29 measures

Study 2 (N = 1439 on CloudResearch): 408 items, 34 measures (2/12)

The original goal of this project, led by James Kim, was to organize the field's measures and constructs into a coherent framework. Can we use bifactor modelling to distill our many, many measures into a manageable number of dimensions? (1/12)

Predicting relationship quality with itself? A single general factor captures most of the variance across 34 common relationship measures In relationship science, researchers have generated a wide array of constructs and corresponding self-report measures to characterize, explain, and predict relationship quality – the foremost studied ...

New paper, out this week in PLOS One, suggests that most close relationship self-report measures are primarily capturing relationship quality 🧵
journals.plos.org/plosone/arti...

Really bears the lie to the whole "the NDP party is irrelevant" narrative.

And the winner is… - Avi Lewis wins NDP leadership race in decisive first-ballot victory Avi Lewis has won the leadership of the federal NDP. The crowd broke out into a raucous cheer when his vote total was announced. He won with more than 39,000 votes.

Exciting! #cdnpoli

It was a privilege to have an ML expert double-check all our models with a fine-toothed comb. It was also embarrassing to force him to sift through my many, manually generated excel files, all because postdoc me had never heard of dplyr. Such is the cost of transparency!

Given how machine learning practices have evolved in the last decade, I was very curious if these findings would hold up. Many thanks to Florian for his nuanced and thorough review.

Romantic desire: not predictable in 2017, not predictable in 2026.

Study promotion for a great PhD student!

If you and/or your romantic partner are experiencing depressive symptoms, consider participating in this online study. Your participation can help researchers better understand romantic partner support. Open to couples of any gender or sexual orientation 🌈

'No cuts, no fees, no corporate universities': Students stage mass protest over Ford government’s OSAP cuts | CBC News Students across the province are protesting recent changes to the Ontario Student Assistance Program (OSAP) — the post-secondary financial aid system will go from a majority grant structure to a major...

"Students across the province are protesting recent changes to the Ontario Student Assistance Program (OSAP) — the post-secondary financial aid system will go from a majority grant structure to a majority loan structure in the new school year."
www.cbc.ca/news/canada/... #OnPoli

Assistant Professor Tenure Track for Social Psychology with a Focus on Environmental Psychology - Universität Bern Universität Bern is looking for Assistant Professor Tenure Track for Social Psychology with a Focus on Environmental Psychology

Our institute is hiring
1. an assistant professor (with TT) for Social Psychology (focus: environmental psychology)
ohws.prospective.ch/public/v1/jo...

2. an assistant lecturer (with TT) for Experimental Personality Psychology
ohws.prospective.ch/public/v1/jo...

One hundred accounts are behind the majority of conspiracy theory content in Canada Researchers analyzed more than 14 million social media posts from accounts in Canada and found that 87 per cent of conspiratorial claims come from influencers, with just 100 users responsible for near...

One hundred accounts are behind the majority of conspiracy theory content in Canada. "In 2025, Canada’s National Observer reported that Meta accepted $300K from The Epoch Times in ads targeting Canadians under a series of false names..." #cdnpoli
www.nationalobserver.com/2026/02/25/n...