Are they losing interest?

They might be, but the signals worth trusting are slow and boring: fewer plans they start, less follow-through, less curiosity about your life, less talk about a future that includes you, holding for weeks. What anxious minds tend to over-weight is the single moment, the off mood, the night that felt cooler than the last.

You can be anxious and right at the same time. The point here is to know which evidence carries weight.

What does research say actually predicts where a relationship is going?

Commitment, mostly, and how committed each person believes the other is. In a 2020 project led by Joel and colleagues, researchers ran machine learning across self-reports from 11,196 couples in 43 separate long-term studies, looking for whatever best predicted relationship quality.1 The top relationship predictor was perceived partner commitment, followed by appreciation, sexual satisfaction, perceived partner satisfaction and conflict.1

So your question, “are they still in this?”, is the right one. Research suggests it matters more than almost anything else you could ask.

The older framework behind it is the investment model. In a 2003 meta-analysis of 52 studies and more than 11,000 people in five countries, Le and Agnew found that commitment tracked three things: how satisfied someone is, how much they have put in (time, shared friends, plans, identity), and how appealing their alternatives look, with more appealing alternatives going with less commitment.2 More committed relationships were more stable, and the pattern held across genders, cultures and relationship types.2

Notice what that list leaves out. Nothing on it is a single gesture. It’s all accumulation.

Which behaviors are worth watching?

The ones people use to keep a relationship going. Researchers call them maintenance behaviors, and they’ve been studied enough to pool. A 2013 meta-analysis by Ogolsky and Bowers combined 35 studies and 12,273 individual reports on five of them: positivity, openness, assurances, shared social networks and sharing tasks.3 People who used these were more satisfied and more committed, and the strongest predictors were positivity and assurances.3

Translated into what you can actually see:

• Assurances: they talk about the two of you continuing. Next month, next summer, meeting their sister.
• Positivity: they’re generally glad to be around you, not just polite.
• Openness: they still tell you what’s going on with them, including the hard parts.
• Social networks: they bring you into their life, and they show up for yours.
• Sharing tasks: they pull their weight on the logistics of being together.

These are the signals that move with investment. If several have thinned and stayed thin, that’s information. If one has dipped during a brutal month at work, that’s a month.

What do anxious readers tend to over-weight?

Their own sense that the other person cares less than they do. A 2018 set of three studies with 333 couples compared what each partner reported feeling with what the other partner believed they felt.4 People higher in attachment anxiety showed a pessimistic bias: they saw their partners as less satisfied and less committed than those partners actually said they were.4 People higher in avoidance, and people with avoidant partners, did too.4

That doesn’t mean your read is wrong. It means an anxious read comes with a known tilt, and it’s worth correcting for before you act on it.

The other things that get too much weight: one short evening, a distracted dinner, a mood you can’t explain. (How fast they reply is not the best signal either.) None of those show up in the research as strong predictors. Patterns in effort do.

Does losing interest happen suddenly?

Usually not, even when it feels that way. A 2024 study drew on four national long-term surveys from the UK, Australia, Germany and the Netherlands, following satisfaction in people whose relationships later ended.5 It found two stages: a long, small decline, then a sharp drop that began somewhere between about seven months and a little over two years before the split.5 People on the receiving end of the breakup entered that steep phase later than the people who initiated it, then fell faster.5

Two things follow from that. One: a single bad week is not what a fading relationship looks like on the inside. Two: if someone is pulling back, the change in their effort may be visible before they’ve said anything, which is why behavior over time is worth more than any one conversation.

How do you tell a dip from a decline?

Meet Priya. Eight months with Tom. He used to plan Saturdays; for the last five weeks she has. He still says goodnight, still kisses her when he walks in. He stopped mentioning the concert in March.

• What happened: She’s made every plan for five weeks, and the March concert hasn’t come up.
• The reading: He’s checking out.
• The fear: That she’ll be the last to know, again.
• The next step: Name the plan shift to him and ask about March directly.

Her facts are real and specific, and they’ve lasted more than a week or two. That’s enough to ask about. It’s not enough to know what’s going on inside him. Tom could be losing interest. He could also be swamped, low, or assuming she likes planning. The only way to find out is to ask.

What to do this week

• Log effort, not feelings. For seven days, note who started plans, who followed through, who asked about whose day. Read it on Sunday.
• Check the five. Assurances, positivity, openness, shared people, shared tasks. Which ones actually changed, and since when?
• Correct for your tilt. If you run anxious, assume your read is somewhat more pessimistic than reality, and look for evidence against it as hard as evidence for it.
• Ask about one specific thing. “I’ve noticed I’ve been making most of our plans lately. What’s going on for you?” beats “Are you losing interest?”
• Get support if the worry won’t loosen. If this fear follows you through the day, wrecks your sleep or shows up in every relationship, a licensed mental-health professional is the right help.

What doesn’t the research show?

• The prediction study measured self-reports in established couples. It shows what tends to predict quality, not what one person feels right now.1
• The investment model meta-analysis pooled studies up to 1999 and shows associations, not a formula for any one relationship.2
• The maintenance meta-analysis is correlational. Fewer maintenance behaviors could follow a decline as much as cause one.3
• The perception-bias studies show an average tilt, not that every anxious worry is mistaken.4
• The decline study followed couples in national surveys over years, mostly longer relationships than a few months of dating.5
• No study can tell you what this person intends. Only their words and what they do next can.