The Journal
How to talk to your partner about chores without a row
Most advice on this conversation is about tone. This is a procedure for one conversation with a fixed ending: a named job, a named person and a date, written down. It also sets out what the research on openings and on the demand/withdraw pattern can support, and where a related finding from the same laboratory has failed to replicate.
Why the first minute decides the hour
An opening that makes a claim about character invites a defence of character, which is not a negotiation. "You never do anything in this house" can only be answered by disputing "never". "The bins didn't go out on Tuesday" can be answered by doing something about the bins.
Sybil Carrère and John Gottman videotaped 124 newlywed couples discussing a disagreement for fifteen minutes, coded the emotion each partner showed in three-minute blocks, and followed them for six years, by which time 17 had divorced. Their 1999 paper in Family Process reports that the first three minutes alone separated the couples who later divorced from those who did not.
That is one sample, and this kind of model is fitted to and reported on the same couples. Richard Heyman and Amy Smith Slep, in the Journal of Marriage and Family in 2001, counted fifteen published divorce-prediction studies, six from Gottman's laboratory, and found that none using general-population couples had tested its equation on a fresh sample. Such an equation is, in their words, "tailor-made for that sample", and accuracy, they showed with survey data, drops sharply on a fresh one.
The three-minute result itself has not, in the sources read here, been tested on a fresh sample. A neighbouring claim from the same laboratory has: its 1998 newlywed paper reported that a wife's "negative start up" predicted separation. Kim, Capaldi and Crosby (same journal, 2007) tested that paper's models on 85 young couples from the Oregon Youth Study, coding start up as negative emotion following a partner's neutral emotion. Neither partner's start up predicted whether the couple were still together two and a half years later.
So "the first minute decides the hour" is a working rule, not a finding. It earns its place because the opening is the one part you control completely. The procedure has six steps: pick one job; open with the instance; make one ask; return to it when you get a defence; watch for pursuit and retreat; end with a written line and a date.
Open with this week's bins, not with the person
The University of Washington's release on the newlywed study drew the distinction that matters, setting a criticism, which it called "a global attack", against "a specific complaint". Its example of a complaint: the rubbish not taken out the night before.
The sentence itself has been tested. Rogers, Howieson and Neame (PeerJ, 2018), working at universities in Perth, Western Australia, gave 253 university students (mean age 28, 77% women) six hypothetical conflicts, one of them a couple who both work full time and disagree about the cleaning, with eight possible opening lines each. The opener rated least likely to provoke a defensive reaction had three parts: it acknowledged the other person's position, stated the speaker's own view as their own, and made the request. A bare demand, the authors conclude, is more likely to meet defensiveness whether it begins with "I" or "you". They state the main limit themselves: participants rated written statements rather than talking, so the findings, they say, generalise more safely to text messages and email.
Applied to the bins: "I know Tuesday is your late night. The bins didn't go out this week and I did them at eleven, which I minded. I'd like to settle who does them." Relate, the relationship-support service run by the charity Family Action, says the same more plainly: "Don't go in with all guns firing", and speak in 'I' statements. One caution: "I feel that you are lazy" is a 'you' statement wearing a disguise.
One instance, one ask, one decision
Choose one job before you start. Not the whole division of the house, a longer exercise this site covers in its article on splitting chores. One job that recurs, that you can name, and that went wrong recently enough for both of you to remember.
Make one ask about it, specific enough to be refused: who does it, on which day, and what counts as done. "Help more" cannot be agreed to or declined. "Bins and recycling out on Tuesday night, new bags in" can.
Then get one decision: yes, no, or a counter-offer. Leave out the history, the other jobs, and other people's partners. Relate's page on arguing uses a row about household chores as its example of an argument really about something deeper, such as feeling unsupported. That may be true of yours; it is a reason to keep this one small. What accumulated over the months before is a separate conversation.
Relate suggests a time limit: after ten to fifteen minutes an argument has probably stopped being productive. One job fits inside that.
When the answer is a defence and not an answer
Defences come in a few standard forms. The counter-complaint: you never deal with the car. The explanation: work has been impossible. The standards argument: they did not need doing yet. Each answers an accusation. None answers the ask.
Do not rebut. Acknowledge whatever is true in it, then return to the ask. "Work has been bad, I can see that. I still want to settle who does the bins." Relate's advice is to say out loud that the other person's feeling makes sense to you. A counter-complaint gets queued: "That's fair, and I'll take the car next. Bins first."
If two returns produce no decision, stop. Relate's guidance is to come back when you have both calmed down. Name when: "Saturday after breakfast" is a pause; "forget it" is a withdrawal.
Demand and withdraw: the pattern to watch for in yourselves
Researchers call one shape of this argument demand/withdraw. One partner presses with requests, criticism and complaints; the other retreats into defensiveness and passive inaction. The sturdiest evidence is a meta-analysis by Schrodt, Witt and Shimkowski in Communication Monographs in 2014, pooling 74 studies (N = 14,255). The pattern correlated with outcomes overall at r = .360, more strongly with relational (.423) and communicative (.418) outcomes than with individual wellbeing (.249), and about equally whether the woman demanded and the man withdrew (.380) or the reverse (.392). Ross, Karney, Nguyen and Bradbury (Journal of Personality and Social Psychology, 2019), citing it, give the direction: demand/withdraw interactions are "stronger among unhappy couples than happy couples". These are pooled correlations. They show what travels together, not which causes which. Nor is it universal: in Ross and colleagues' first study, of 515 couples, it harmed wives' satisfaction among relatively affluent couples and helped it among relatively disadvantaged ones.
The roles are positions, not personalities. Christensen and Heavey, in a 1990 study of 31 married couples, found each partner more demanding when the change discussed was one they wanted, and more withdrawing when it was their partner's. Kluwer, Heesink and Van de Vliert (Personal Relationships, 2000) found the same structure in scenario studies of housework: one partner discontented, the other content as things are. In those studies the defender of the status quo, not the complainant, was expected to reach their goal. If nothing is decided, nothing changes.
Pursuit is repeating the ask louder, adding instances, following someone into another room. Retreat is "fine, whatever", the phone, or leaving with no time to resume. Relate advises against starting just before someone leaves for work or with the television on. Asking in advance for ten minutes on Saturday gives notice, not an ambush.
Close it with a rota line and a date
The conversation ends when one line is written down where you can both see it: the job, the name, the day, what done means, the start date and the date you will look again. "Bins and recycling: Sam. Tuesday night, at the kerb, new bags in. From 22 September. Look again 20 October."
A trade is a legitimate ending: one of you takes the bins, the other the Thursday shop. "I'll try to do more" is not an ending, and nor is a better mood. If fifteen minutes pass with no line, write down when you will resume.
The written line is what stops you asking again next Tuesday. When one has held for a month, a full rota is the next step.
All of this assumes a partner you are not afraid of. If you are, the opening sentence is not the problem. GOV.UK names a helpline for each nation: Refuge's National Domestic Abuse Helpline, 0808 2000 247 (England); Live Fear Free, 0808 80 10 100 (Wales); the Domestic Abuse and Forced Marriage Helpline, 0800 027 1234 (Scotland); the Domestic and Sexual Abuse Helpline, 0808 802 1414 (Northern Ireland).
Sources
- Carries every figure in the first paragraph of section 5: the abstract states that the meta-analysis reviews "74 studies (N = 14,255)", gives the overall effect as r = .360, wife demand/husband withdraw .380, husband demand/wife withdraw .392, relational outcomes .423, communicative outcomes .418, well-being outcomes .249 (and demographic .239, distressed/clinical samples .413 against non-distressed .345, which the piece does not use). The abstract nowhere says what N counts, so the piece prints it as the abstract does and calls it neither participants nor couples. The brief named Human Communication Research as the journal; the Crossref deposit (api.crossref.org/works/10.1080/03637751.2013.813632, re-read 20 September 2026, which carries no abstract) gives Communication Monographs 81(1), 28-58, and the piece follows Crossref. The publisher page at tandfonline.com, where doi.org resolves, returned 403 to automated requests again on 20 September 2026 and Semantic Scholar reports the abstract as withheld by the publisher, so this URL is the OpenAlex record, which serves the full abstract; read 20 September 2026. The abstract speaks only of associations. The full text is paywalled and no open repository copy was found on 20 September 2026, so the authors' own limitations section has never been opened: the two caveat sentences in the piece ("These are pooled correlations. They show what travels together, not which causes which.") are the piece's own, attributed to nobody, and the boundary that follows them is taken from Ross and colleagues instead. — Schrodt, P., Witt, P. L., & Shimkowski, J. R., "A Meta-Analytical Review of the Demand/Withdraw Pattern of Interaction and its Associations with Individual, Relational, and Communicative Outcomes", Communication Monographs, 81(1), 28-58, doi 10.1080/03637751.2013.813632 (abstract via OpenAlex), 2014
- Used for the direction of the pooled effect in the first paragraph of section 5, in place of a 2014 press release. The introduction says: "demand/withdraw interactions are stronger among unhappy couples than happy couples (meta-analytic r = .36; Schrodt, Witt, & Shimkowski, 2014)"; the piece quotes the words before the bracket. Also the source for the closing sentence of that paragraph, which is the piece's stated boundary on the pooled effect: the abstract says Study 1 (N = 515 couples; 18-month follow-up) replicates the harmful effects of observed demand/withdraw communication on changes in wives' satisfaction among relatively affluent couples, while documenting benefits among relatively disadvantaged couples. Study 2 (431 couples) is not used. The publisher page is paywalled; this is the copy of the published article hosted by the authors' laboratory at UCLA, whose text was extracted and read on 20 September 2026. — Ross, J. M., Karney, B. R., Nguyen, T. P., & Bradbury, T. N., "Communication That Is Maladaptive for Middle-Class Couples Is Adaptive for Socioeconomically Disadvantaged Couples", Journal of Personality and Social Psychology, 116(4), 582-597, doi 10.1037/pspi0000158, 2019
- Source for the definition of demand/withdraw in section 5 ("the demander, usually the woman, pressures the other through emotional requests, criticism, and complaints, and the withdrawer, usually the man, retreats through defensiveness and passive inaction") and for the finding that roles follow whose issue it is: "31 couples were assessed in 2 conflict situations: 1 in which husband wanted a change in wife and 1 in which wife wanted a change in husband", and "both husband and wife were more likely to be demanding when discussing a change they wanted and more likely to be withdrawing when discussing a change their partner wanted". The abstract says "31 couples" rather than "31 married couples"; the piece writes "married" on the strength of the abstract's own "marital conflict" and its husband-and-wife design, and states no country or setting, because the abstract names neither (it does not say where the couples were recruited, nor that the assessment was in a laboratory). The abstract also reports that men were overall more withdrawn than women, which the piece does not rely on. The PubMed record page serves a cookie notice to automated requests, so this URL is the NCBI E-utilities endpoint serving the abstract for PMID 2213491; re-read 20 September 2026. — Christensen, A., & Heavey, C. L., "Gender and social structure in the demand/withdraw pattern of marital conflict", Journal of Personality and Social Psychology, 59(1), 73-81, 1990
- Source for the housework-specific sentences in the second paragraph of section 5. The abstract reports two experiments in which spouses responded to scenarios; spouses reported more conflict over housework than over paid work and child care; wife-demand/husband-withdraw was reported more than the reverse only when the wife desired a change in her spouse's contribution, a situation the authors describe as asymmetrically structured, with one partner discontent and the other wanting to maintain the status quo; and "defenders of the status quo were expected to reach their goal" more than complainants. That clause is passive in the abstract and names nobody as doing the expecting, so the piece keeps it passive and does not attribute the expectation to the study's participants; the OpenAlex inverted index also drops the comparative adverb from the clause, which is why the piece paraphrases the direction rather than quoting it. These are reports about hypothetical conflicts, which is why the piece says "scenario studies" and "expected". Sample sizes are not in the abstract and are not stated in the piece. Abstract read via the OpenAlex record on 20 September 2026; the Wiley page, where doi.org resolves, returned 403 to automated requests, so this URL remains the record that was read; it is JSON, with the abstract stored as a word index. — Kluwer, E. S., Heesink, J. A. M., & Van de Vliert, E., "The division of labor in close relationships: An asymmetrical conflict issue", Personal Relationships, 7(3), 263-282, doi 10.1111/j.1475-6811.2000.tb00016.x (abstract via OpenAlex), 2000
- Source for the design and headline result in the second paragraph of section 1: the marital conflict discussion of 124 newlywed couples coded with the Specific Affect Coding System, divided into five 3-minute intervals, and the statement that marital outcome over a 6-year period could be predicted using just the first 3 minutes of data for both husbands and wives. The piece reports the result as separating the groups and quotes no prediction percentage. NCBI E-utilities abstract for PMID 10526767, read 20 September 2026. — Carrère, S., & Gottman, J. M., "Predicting divorce among newlyweds from the first three minutes of a marital conflict discussion", Family Process, 38(3), 293-301, 1999
- Source for the details of the newlywed study that the abstract omits (couples married less than nine months, mirroring the racial and ethnic makeup of the greater Seattle area, fifteen-minute videotaped discussions, 17 couples divorced by the end of the study) and for the criticism/complaint distinction at the top of section 2: the release contrasts a criticism, described as "a global attack", with "a specific complaint", and its example of the complaint is the partner not taking out the trash the night before. The release frames this as advice to wives in particular; the piece does not repeat that framing. University press release, not a peer-reviewed page. Read 20 September 2026. — University of Washington News, "First three minutes of discussion about on-going area of marital conflict are predictive of divorce for newlyweds", 27 September 1999, 1999
- Source for the methodological criticism stated in the third paragraph of section 1. The paper says only 15 published studies have predicted who will divorce, that the six studies of Gottman and colleagues have received extraordinary attention, that predictive equations must be crossvalidated for their true value to be known and that no published study predicting divorce with general population couples had done this; it explains that a fitted equation is "tailor-made for that sample" and may be overly influenced by idiosyncrasies of the data set (overfitting). Its own demonstration uses the 1985 National Family Violence Survey, split into development and crossvalidation samples, and the abstract says accuracy and predictive value drop precipitously during crossvalidation - the basis for the piece's "accuracy, they showed with survey data, drops sharply on a fresh one". The paper's list of the fifteen names Gottman et al. (1998), Gottman and Levenson (1999) and Carrère et al. (2000) among others; it does not list the 1999 first-three-minutes paper by name, so the piece attributes the criticism to the laboratory's prediction studies as a group, not to that paper specifically. Author manuscript on PubMed Central, read 20 September 2026. — Heyman, R. E., & Smith Slep, A. M., "The Hazards of Predicting Divorce Without Crossvalidation", Journal of Marriage and Family, 63(2), 473-479, 2001
- Source for the fourth paragraph of section 1. The paper tests Gottman, Coan, Carrère and Swanson (1998), the laboratory's newlywed study; it does not test, and nowhere mentions, the 1999 first-three-minutes measure, and the piece says so. Abstract: a community-based sample of 85 married or cohabiting couples with at-risk backgrounds (young men from the Oregon Youth Study and their partners), affective processes assessed at about age 21 and relationship status and satisfaction approximately 2.5 years later; "The major findings of Gottman et al. failed to replicate", and in particular women's negative start up was not predictive of relationship status. The introduction summarises the 1998 finding as the wife's negative start up having predicted marital separation. The paper defines the code sequentially ("Either low- or high-intensity negative affect following partner's neutral affect"), so it is not a measure of how a conversation opened, and the Results say "neither men nor women's negative start up was predictive of separation at T2". The authors note differences in sample and procedure from the original (married and cohabiting couples from higher-risk backgrounds against educated newlyweds), and the piece does not claim the result settles the question. Read on PubMed Central, 20 September 2026. — Kim, H. K., Capaldi, D. M., & Crosby, L., "Generalizability of Gottman and Colleagues' Affective Process Models of Couples' Relationship Outcomes", Journal of Marriage and Family, 69(1), 55-72, 2007
- Source for the second paragraph of section 2. The Participants section reads in full, on the point of population: "Participants were 253 university students (Mean = 28 years; SD = 8.75; 77% female)", followed by "Prior to commencement of the research ethical approval was obtained from Edith Cowan University ethics committee (Ref: 15257)". It never says which university the students attended or in which country they were recruited, so the piece gives the age and sex split as the population and places the geography where the paper does place it, on the authors: the bylines read "Psychology Department, Edith Cowan University, Perth, WA, Australia" and "Law School, University of Western Australia, Perth, WA, Australia". The piece therefore says the authors were working at universities in Perth, Western Australia, and does not call the students Australian. Six hypothetical scenarios, one of them a couple both working full time in which one partner feels it unfair to be responsible for all the cleaning; eight opening statements per scenario varying I/you language and communicated perspective; participants rated the likelihood of a defensive reaction. The statement combining the other person's perspective, the speaker's own perspective and I-language was rated least likely to produce a defensive response, and the discussion says a defensive reaction is more likely when no perspective is communicated, regardless of I- or you-language. The Limitations section says "the non-interactive rating statements paradigm used in the present study limits the generalisability of findings beyond non-interactive communication contexts" and that "our findings reported here are more appropriately generalised to less interactive forms of communication such as text messaging and email"; that is the only limit the piece attributes to the authors. Open-access full text read on PubMed Central, 20 September 2026; the PeerJ page itself (peerj.com/articles/4831/) returned 403 to automated requests that day. — Rogers, S. L., Howieson, J., & Neame, C., "I understand you feel that way, but I feel this way: the benefits of I-language and communicating perspective during conflict", PeerJ, 6, e4831, 2018
- The non-academic anchor, and the source for every piece of Relate advice in sections 2 to 5. Under "Talking it over" the page advises choosing an appropriate time ("coming back to the topic when you've both calmed down"; "not immediately before someone must go to work or with the TV on in the background"), making the discussion time-limited ("if you're still arguing after 10 to 15 minutes you're getting so worked up that it's not going to be productive"), starting amicably ("Don't go in with all guns firing, or with a sarcastic or critical comment"), using 'I' statements in place of 'you' statements, and validating a partner's feelings ("It makes sense to me that you feel like that"). Its "onion" passage uses an argument about a fair share of household chores as the example of a surface issue that taps wider feelings of being supported. On the descriptor: the page's own footer reads "Relate is part of Family Action. Charity number: 264713, Company number: 01068186 (Registered in England and Wales)", and 264713 is Family Action's registration, so the piece introduces Relate as a relationship-support service run by the charity Family Action rather than as a charity in its own right, and uses the bare name thereafter. No publication or review date is shown on the page; read 20 September 2026. — Relate, "We can't stop arguing", 2026
- Source for the four helplines named in the last paragraph of section 6, so that the safety signpost can be acted on rather than gestured at. The page gives, by nation: "Refuge's National Domestic Abuse Helpline" on 0808 2000 247 (England); "Live Fear Free" on 0808 80 10 100 (Wales); the "Domestic Abuse and Forced Marriage Helpline" on 0800 027 1234 (Scotland); and the "Domestic and Sexual Abuse Helpline" on 0808 802 1414 (Northern Ireland). Names and numbers are taken verbatim. Page published 5 October 2018, last updated 29 April 2025; read 20 September 2026. Nothing else in the piece rests on this page, and the piece makes no claim about what any of these services will do. — GOV.UK, "Domestic abuse: how to get help", 2025
Barriers is a household planner for tasks, money and receipts. A job carries someone's name, and whoever it lands on can accept it or say no with a reason. The task list is free.
Get Barriers