What peer-reviewed research says about taste, personality, values, and shared living. The short version: taste can supply social context, not destiny, and daily routines still need a direct conversation.
The honest thesis
We size each other up by our tastes constantly. The playlist on someone's profile, the books on their shelf, the shows they leave running in the background: we read all of it as a clue to who a person is, and whether we could actually live with them. That instinct is not wrong, just incomplete. Tastes, routines, values, and personality-linked preferences are meaningful social signals. They help people infer identity, decide who feels familiar, coordinate daily life, and avoid friction. What the evidence does not support is the tidy version of the story: that a shared favorite genre makes a great housemate.
The strongest, evidence-based claim: shared tastes matter most when they point beyond taste, to openness, values, lifestyle rhythms, and how two people will actually share a kitchen.
Executive summary
Each is drawn straight from the peer-reviewed record. Read these in thirty seconds, then go as deep as you like below.
01 Taste as a signal
The modern case starts with Rentfrow and Gosling's "The Do Re Mi's of Everyday Life." Across the cited studies, four music-preference dimensions emerged, and Openness correlated r = .44 and .41 with Reflective & Complex music. Extraversion tracked Upbeat & Conventional and Energetic & Rhythmic music.
The honest caveat matters. A 2017 meta-analysis found most music-personality correlations are small (average r ≈ .06). So taste is a signal, not a personality X-ray. It is most useful as one layer among many, which is exactly how a good matching model should treat it.
Evidence: robust for music-personality links, especially Openness02 Similarity
Similarity structures social life. Homophily shapes friendship, marriage, work, and support networks, and dissimilar ties dissolve faster. But the sharpest finding for matching is this: Montoya and colleagues' meta-analysis of 313 studies found actual similarity predicts attraction early (r = .47), yet in established relationships it is perceived similarity that keeps predicting quality, not raw trait overlap.
That is a design instruction, not just a fact. Objective similarity helps two people select in and feel safe. Ongoing harmony depends on whether they experience each other as understandable, aligned, and fair.
Evidence: robust for homophily and similarity-attraction03 Personality
"Opposites attract" is weak in the data, but so is "match the clones." In three large national samples, each person's own personality explained about 6% of relationship satisfaction, while couple similarity explained less than 0.5% after controls. A 2023 study of 1,294 couples reached the same verdict: trait similarity is not a robust predictor.
Personality still matters, just not as similarity. Higher agreeableness and conscientiousness, and lower negative emotionality, predict better outcomes. For housemates that means matching for the consequences of personality: reliability with chores and bills, conflict repair, and stress regulation.
Evidence: robust that trait similarity is a weak incremental predictor04 Values & routines
Values and life goals often show more couple similarity than personality, and they matter more once people share a space. The most housemate-relevant finding comes from situation perception: in dyadic diary studies, similarity in how two people interpret the same everyday moment predicted satisfaction (β ≈ .17 to .19) and its change over two weeks.
Rentzsch and colleagues use the perfect example: if one person reads an untidy kitchen as a minor inconvenience and the other reads it as a major violation, the conflict is not about dishes. It is about mismatched meaning. That is where shared homes are won or lost: sleep schedules, guests, cleaning standards, noise, bills, and conflict repair.
Shared taste starts the conversation. Shared values and routines keep the home from becoming a tiny constitutional crisis over whose turn it was.
05 The home is active
Shared homes create interdependence: people share space, routines, resources, and stress. The cited student survey reports roommate conflict, while other work examines roommate support after move-in. These findings describe groups, not a prediction for one pair.
Random-assignment studies are the eye-opener. Being assigned a roommate with higher cognitive vulnerability predicted increases in one's own rumination months later, and roommates converged emotionally over a year. The home is a contagion environment. Compatibility is not just selection; it is exposure.
Evidence: strong that roommate dynamics matter; moderate on adult generalization06 First impressions
Thin-slice research shows short observations carry real information (overall r = .39 for judgments from under five minutes). Early signals like taste, communication rhythm, and expressive style are not fake. But liking someone quickly and sharing a bathroom with them for a year are different questions.
Machine-learning speed-dating studies drive it home: models predicted who is generally desirable, but could not predict the unique chemistry between two specific people. The lesson for matching is to separate three things: entry compatibility (do I feel safe and interested), living compatibility (can we coordinate chores, noise, guests, money), and dynamic compatibility (do we repair well after friction).
Evidence: robust for thin-slice accuracy; preliminary for long-term prediction07 Why it matters now
Two forces make compatibility consequential. First, cost: a record 49% of U.S. renters are cost-burdened, and house-sharing hit a record 6.8 million households with unrelated housemates in 2023. Sharing is mainstream and rising.
Second, health. The cited meta-analysis links loneliness, social isolation, and living alone with mortality risk. It does not show that a roommate score improves health or prevents conflict.
Evidence: robust for connection-and-health links and rising house-sharingCross-referenced insights
These are the aha moments that only appear when you read the studies together.
Music and film are useful because they encode Openness, values, and identity. Use taste as a signal layer, not the whole engine.
Match on real overlap, then help people build felt alignment through expectations and conversation.
Match for the interaction consequences of personality (reliability, conflict repair), not for trait similarity.
Small routines become repeated negotiations. Situation meaning predicts satisfaction more than abstract similarity.
Roommates shift each other's stress, rumination, and mood over time. Compatibility is exposure, not just selection.
Sleep, noise, guests, bills, and repair attempts are measurable and designable. Chemistry is not.
Fit at day zero is not enough. The best model supports check-ins and conflict repair after move-in.
What this means for finding a roommate
The science does not justify telling anyone "you are a 92% guaranteed fit." It justifies something better: "you align strongly on sleep, guests, cleaning, and conflict style; you differ on music and food, and you both rate those low." That is also the standard for using CoHabby's Synergy Score: treat it as a conversation starter, not a validated outcome prediction. You can try the compatibility quiz in a couple of minutes.
The four layers on the left are ranked by how much they actually govern day-to-day life together. Surface tastes spark the first conversation; household operations and conflict repair are where a lease quietly succeeds or falls apart. A good match weights the layers that carry real weight for each person, instead of scoring everyone on the same surface traits.
See how the platforms compare on real compatibility in our guide to the best roommate apps, and how to screen a potential roommate for the operational fit that matters most.
Methodology & evidence grading
This report is a literature synthesis, not a study of CoHabby's own users. We reviewed 26 peer-reviewed and public sources across psychology, sociology, and housing research, prioritized findings with clear effect sizes, and cross-referenced them to surface patterns individual papers miss.
We grade evidence honestly. Music-personality links, homophily, similarity-attraction, and the connection-health relationship are robust. Values, situation-perception, and college-roommate findings are moderate. Direct prediction of long-term adult co-living outcomes is preliminary, because much roommate research uses college samples. Correlation is not causation, and where studies examine romantic couples we say so rather than silently overreaching. Full detail lives in our research methodology.
The receipts
Writing about roommates, compatibility, co-living, or the loneliness economy? CoHabby's founders are available for interviews and can share data and commentary. Reach us at press@cohabby.com.