attraction.lab

Micro-Cheating: What Has Actually Been Measured, and the Table of 27 Behaviours Nobody Quotes

PubMed returns zero results for the word. The behaviours it names were measured in 456 people, and the ambiguous ones landed at a coin flip, 30 points apart.

Research-backed writing on attraction, dating and relationships. Drafted by AI against a named bibliography, published only when every claim links to its source.
31 min read
Editorial illustration for: Micro-Cheating: What Has Actually Been Measured, and the Table of 27 Behaviours Nobody Quotes

Before the article. This subject is written about almost entirely as a list of things to watch for in another person, and there is a point at which watching stops being vigilance and becomes control. If what is happening in your relationship includes demands for passwords or device access, having to account for where you are and who you are with, someone deciding who you may see, control of money, pressure about sex, or punishment alternating with affection until you manage your own behaviour to avoid it, that is not a disagreement about boundaries and nothing here addresses it. In the United States the National Domestic Violence Hotline is 1-800-799-7233, or text START to 88788, 24 hours a day. Find A Helpline, in the sources below, lists free and confidential services by country. The first section of the article returns to this.

With that said. Micro-cheating has a Psychology Today topic entry, a Wikipedia article and ten pages of confident explanation on the first screen of search results. It also has, as of 1 November 2026, zero records in PubMed and one in Europe PMC — and that one is a 2026 measurement paper about something else, which mentions the word once in its introduction as an example.

So the term has no definition anyone agreed, no instrument, no prevalence figure and no study of consequences. What it does have is a question underneath it — where is the line, and who decides — and that question was measured carefully in 2013, five years before the word was coined, in a study that almost nobody writing about micro-cheating cites. Four hundred and fifty-six people rated 27 behaviours, and the behaviours this term is really about landed in the middle of the table with standard deviations of 30 to 35 points. The same study found that attachment anxiety raised the ratings on 18 of the 27.

What follows is thirteen sections: where the word came from, the literature counted with the queries written out, what the ranking pages actually cite, the 2013 table and the stretch of it that matters, what moves your own line, how common the behaviours are in a married sample, why the escalation claim has never been tested, the one attempt at an instrument and the half of it that collapsed, what watching has been measured to do, why professional counsellors disagree too, and what the evidence will support you doing. Our editorial policy explains why every empirical claim here carries a linked source.

#1

Before the rest of this page: when looking becomes the problem

This article is about what has been measured, and the measurements point somewhere uncomfortable early on, so it belongs at the top rather than buried at the foot.

Almost every page about micro-cheating is organised as a list of behaviours to watch for in another person. The research that exists on watching does not support that as a plan. In the one study that tested whether time on a partner's social media predicted jealousy about it, the variable that did nearly all the work was the reader's own trait jealousy, not anything the partner did — and the authors describe the result as a loop in which looking produces ambiguous information, and ambiguous information produces more looking. A later section sets that study out in full.

There is also a line that the vocabulary of this subject blurs, and it is worth naming before anything else. Wanting to know what a partner is doing is ordinary. Checking a phone once because something felt off is ordinary. What is not ordinary, and is not a dispute about boundaries, is a pattern: demanding passwords or access to devices, requiring an account of where you are and who you are with, deciding who you are allowed to see, controlling money, pressure about sex, or alternating punishment with affection until you start managing your own behaviour to avoid the punishment. Those are the warning signs the US National Domestic Violence Hotline lists, and a relationship that has them in it is not having a disagreement about whether liking a photo counts.

In the United States the National Domestic Violence Hotline is 1-800-799-7233, or text START to 88788, 24 hours a day. Find A Helpline, in the sources at the foot of this page, lists free and confidential services by country. If you are the one being monitored, nothing on this page is about you being insufficiently transparent.

With that said, the rest of this is about the evidence, and the evidence is in an unusual state: the word has almost nothing behind it, while the question underneath it has a 2013 data table that answers it better than anything written since.

#2

Where the word came from, and it was not a study

Micro-cheating entered circulation through the media, not the literature, and the trail is short enough to walk.

The term is attributed to Melanie Schilling, an Australian psychologist and television personality, in interviews given around the end of 2017; it reached mass circulation in the first days of 2018, when outlets on both sides of the Atlantic ran the same explainer within days of each other. The definition that travelled with it has stayed stable ever since: a series of small actions suggesting that someone is emotionally or physically focused on a person outside their relationship.

Notice what kind of claim that is. It is a category, offered complete, with examples attached. It did not arrive as a hypothesis with a measurement behind it, and nothing has been done since to turn it into one. Wikipedia's entry, which is more careful about this than most of the pages that rank for the term, calls it a neologism, says discussion has mostly happened in mass media, and notes limited scholarship as of the mid-2020s.

This site has now written about enough of these to recognise the shape. Breadcrumbing came out of a newspaper column and later acquired a real, if narrow, body of research. Cushioning has no research under its own name but sits on top of a behaviour that has been studied since 2014 under a different one. Pocketing overlays a term on the literature about relationship concealment. The 3-6-9 rule and the three-day rule have nothing underneath them at all.

Micro-cheating is a fourth case, and it is the most interesting of the lot: the word has nearly nothing, and the question it points at has been measured carefully, with a sample of 456 people, five years before the word was coined. Almost nobody writing about the term cites it.

#3

The literature, counted on 1 November 2026

Here is the base rate for the subject itself, with the queries written out so you can re-run them rather than take my word for it.

PubMed returns zero results for `"micro-cheating" OR "microcheating"`. Not a small number — none. The same query without the hyphen also returns nothing. For comparison, the same database returns results for breadcrumbing, and Europe PMC, which indexes more widely, returns 19 records for that word and one for this one.

That one record is worth describing precisely, because it is not what it sounds like. It is a 2026 paper in Frontiers in Psychology by Ahmet Aldemir on something called digital fidelity, and the word micro-cheating appears in it exactly once — in the introduction, as one example of an online action that makes it hard to tell what counts as betrayal. It is not a study of micro-cheating. A later section covers it anyway, because what happened when its authors tried to measure the neighbouring idea is the most instructive result in this whole area.

Searching Crossref by title for micro-cheating returns one hit about academic integrity — micro-outsourcing of coursework — and nothing about relationships.

What else exists, after looking: a conference paper and a review article. The conference paper is Margaret Foster, Aspen Omapang and Marina Johnson-Zafiris at Cornell, presented at the Association of Internet Researchers in 2023, and it describes itself as an in-progress mixed-methods pilot. It does not survey anyone; it applies topic modelling and discourse analysis to Reddit posts about possible relationship transgressions, and calls microcheating an ambiguous new term. Its one finding matters here and comes back later: the discussion they analysed rests on a belief that proof of fidelity or its absence can be found on a partner's phone, which functions as a licence for surveillance. The review article is by Ali Karpuz and Gizem Çelik Özkan, published in Turkish in the Turkish Academic Research Review in June 2024, and it is explicitly a review rather than a study — it summarises what is known about infidelity generally and discusses where micro-cheating might fit.

So: no definition agreed by anyone, no instrument, no prevalence estimate, no sample, no study of consequences. That is the entire research base for a term that has a Psychology Today topic entry, a Wikipedia article, and a first page of search results in which eight separate publications explain it to you with confidence.

#4

What the pages that rank for it actually cite

It seemed worth checking rather than assuming, so on 1 November 2026 the first page of results for micro cheating in a relationship was pulled and each page opened and read for citations.

The pages were marriage.com, the Psychology Today topic entry, ChoosingTherapy, The Modest Man, a Psychology Today blog post, a therapist's own site, Real Simple, TherapyTips, Verywell Mind and Simply Psychology.

Across all of them there is not one sample size. No `n =`, no number of participants, no survey, in any of the ten. The only percentages anywhere were a pair on Verywell Mind — 20% of married couples and 70% of unmarried couples engaging in infidelity — and those are about infidelity in general, not about this. The survival figures that circulate for infidelity in general turn out to have the same problem one level up, and we trace each of them to its denominator in whether couples stay together after infidelity.

Four of the pages do carry references, and two of them are good ones. ChoosingTherapy cites Moller and Vossler's 2015 study of how couple counsellors define infidelity, which is a genuinely relevant piece of work and appears in a later section here. Real Simple says in its own text that there is no universal definition of what counts. Verywell Mind lists three journal articles. The Psychology Today topic entry cites nothing at all: it says research indicates that women may be more emotionally affected by online micro-cheating behaviours than men, and names no study, no author and no year.

The point is not that these pages are dishonest. Most of them are written by licensed therapists and several say in plain words that the concept is subjective. The point is structural, and it is the gap this page is trying to fill: the subject is covered entirely in the register of definition and example, and not once in the register of measurement. Every one of the ten tells you what micro-cheating is and gives you a list. None tells you what happened when somebody asked 456 people where the line was and wrote down the answers.

#5

27 behaviours, 456 people, and no agreement

The study that answers the question the term is asking was published in Evolutionary Psychology in 2013 by Daniel Kruger, Maryanne Fisher, Robin Edelstein, William Chopik, Carey Fitzgerald and Sarah Strout, and it is called, straightforwardly, Was that Cheating?

They built an inventory of 27 interpersonal behaviours spanning several dimensions: sexual acts, erotic ones, behaviours implying relationship status, emotional bonding, financial support, extensive socialising, and deliberately mundane comparison items. Then they asked 456 heterosexual undergraduates — 67% women, mean age 19.31 — at two public Midwestern universities to imagine a person in a long-term relationship doing each thing with someone other than their partner, and to rate on a 0% to 100% scale how far that behaviour indicated cheating. The items were presented in random order. Just under half the sample, 48%, were in an exclusive relationship at the time.

The averages ran from 8.1% at the bottom (giving the other person $5) to 97.7% at the top (intercourse). Oral sex came in at 96.8, showering together 96.2, kissing on the lips 88.7, emailing nude pictures 88.2, texting erotic messages 82.6.

So far this is unsurprising, and that is the part the genre quotes when it quotes anything. What matters is what the gradient looks like, because the authors found that intensity mattered within categories as well as across them. Kissing on the lips scored far above kissing on the cheek (d = 0.97). Giving $500 scored far above giving $5 (d = 0.89). A hug of more than ten seconds scored far above a brief one (d = 0.88). Sharing a bed scored above sharing a hotel room (d = 0.59). And calling a third party when upset about something that happened with your partner scored above calling them when upset about something that happened at work (d = 0.62).

Their own summary of the pattern is the honest one: some behaviours are clearly identified as cheating, some are clearly not, and some are ambiguous — and even among the ambiguous ones, some are read as more suspicious than others. The whole of micro-cheating lives in that third group.

#6

The middle of the table is where the word lives

This is the section nobody writes, so here is the relevant stretch of that table, with the standard deviations, because the standard deviations are the finding.

  • Sleeping in the same bed — 68.4 (SD 30.5)
  • Holding hands — 63.2 (SD 32.0)
  • Staying in the same hotel room — 52.7 (SD 35.3)
  • Forming a deep emotional bond — 52.4 (SD 34.6)
  • Spending lots of time together — 52.2 (SD 30.8)
  • Sitting in someone's lap — 52.2 (SD 32.8)
  • Accompanying them to a formal event — 43.3 (SD 33.7)
  • Going out to dinner — 41.4 (SD 32.9)
  • Talking on the phone several times a week — 40.1 (SD 33.2)
  • Kissing on the cheek — 36.9 (SD 32.7)
  • Sharing secrets — 36.5 (SD 32.1)
  • Hugging for more than ten seconds — 34.5 (SD 30.5)
  • Calling them when upset about your relationship — 33.0 (SD 31.4)
  • Taking a road trip out of state — 32.6 (SD 32.6)
  • Telling dirty jokes — 25.9 (SD 29.9)
  • Calling them when upset about work — 19.2 (SD 26.9)
  • Hugging briefly — 12.2 (SD 21.5)

Read the list of behaviours that pages about micro-cheating tell you to watch for — a close friendship, a long hug, confiding in someone else, dinner with a colleague, frequent messaging, keeping a conversation going with an ex — and you are reading this part of the table. The behaviours in question average somewhere between a third and two thirds of the way to cheating, which means that on average they are not classified either way.

And the standard deviations, which sit around 30 to 35 points on a 100-point scale, say something stronger than the averages do. A mean of 52 with an SD of 35 does not describe a population that mildly agrees the behaviour is halfway suspicious. It describes a population split, with a large group near zero and a large group near one hundred. Two people can look at the same hug and be at opposite ends of the scale, and both will be perfectly representative of how people actually answer.

This is also why the question comes up as often from the accused side as from the accusing one. Warmth that one person reads as ordinary friendliness is read by another as something else, and the research on that mismatch — including the finding that the problem usually sits with the person reading rather than the person smiling — is in what to do when you keep coming across as flirty and the difference between flirting and friendliness.

When the authors grouped the items statistically, the middle group — the eighteen items they called behaviours indicating close relationships — averaged 41.90 with an SD of 21.98. The sexual and erotic group averaged 93.08 with an SD of only 13.96, and the casual group averaged 10.50. In other words: the two ends of the subject are settled, and the disagreement is concentrated precisely in the zone where the word micro-cheating was invented to operate.

This is why the arguments do not resolve. They are not arguments about the facts of what happened. They are arguments conducted between two people who placed the same behaviour 60 points apart, each believing the other is being obtuse.

#7

Your own anxiety moves the line, and that has a number

The same study measured attachment, using the first five items of each dimension of the Experiences in Close Relationships inventory, and the result is the single most useful thing in this literature for anyone actually in the situation.

People higher in attachment anxiety rated 18 of the 27 behaviours as more indicative of cheating. Not one or two. Eighteen. And the effect was concentrated in exactly the ambiguous middle: taking a road trip (β = .192), sitting in someone's lap (.186), going out to dinner (.185), telling dirty jokes (.184), accompanying them to a formal event (.166), staying in the same hotel room (.162), spending lots of time together (.159), holding hands (.149), talking on the phone several times a week (.143), a long hug (.134). On the grouped dimensions, anxiety predicted the close-relationship cluster at β = .180 and the sexual cluster not significantly at all.

Attachment avoidance did something close to the mirror image, and smaller: five items rated lower, all of them at the sexual and erotic end, plus one low-rated item higher.

These are modest coefficients from a student sample and one questionnaire sitting, and nobody has followed anyone over time. But the direction is unambiguous and it reframes the whole subject. Where you draw the micro-cheating line is partly a fact about you, and specifically about how much you fear being abandoned. It is not a neutral reading of your partner's conduct, and the pages that hand you a checklist of behaviours to monitor are handing the most alarming checklist to the people whose own measurements are most inflated by worry.

That is not a reason to dismiss your own reaction, and it is emphatically not a reason for anyone to tell a partner that their discomfort is just their attachment style — which is a way of winning an argument, not a finding. It is a reason to treat the line as something the two of you have to agree rather than something one of you can detect. Our attachment style test will show you your own two scores on these dimensions, with the usual warning on its own page that it is an educational tool and not a diagnosis, and what is known about whether those patterns change covers the rest. The difference between a feeling of threat and a behaviour of control, which is the distinction this whole area turns on, is set out at length in the piece on jealousy.

Further illustration for: Micro-Cheating: What Has Actually Been Measured, and the Table of 27 Behaviours Nobody Quotes
#8

How common the behaviours actually are: between 5% and 12%

There is no prevalence figure for micro-cheating, because there is no instrument and no study. There is a prevalence figure for a set of behaviours that overlaps it almost exactly, and it is much lower than the genre implies.

Brandon McDaniel, Michelle Drouin and Jaclyn Cravens published it in Computers in Human Behavior in 2017. Their sample was 338 individuals — 173 wives and 165 husbands — from 176 families in the United States, all living together with a child aged five or younger, recruited from a family database, parenting websites and local announcements. They were 96% married, 92% white, 73% university-educated, with an average relationship length of 10.02 years. The seven items they wrote are, in substance, the micro-cheating list:

Feeling uncomfortable if a partner read your messages. Wondering whether a partner would be upset by them. Having messages you would want to hide. Sharing deep or intimate information with someone online instead of your partner. Liking to message old romantic partners. Getting defensive when interrupted online. Hiding what you say to others online.

Counting everyone who agreed at all, the figures were: 12% uncomfortable if a partner read their messages, 7% sometimes confiding in someone else online instead of their partner, 6% with messages they would want to hide, 6% messaging old romantic partners, 6% getting defensive when interrupted, 5% wondering whether their partner would mind, and 5% hiding what they say online. Men and women did not differ significantly on any item.

Those are not epidemic numbers. The single most common item — a feeling of discomfort, not an act — reached one person in eight, and the behavioural items clustered around one in sixteen.

One of those items deserves a note, because it is the one the genre treats as most damning: confiding in somebody else instead of your partner, reported by 7%. There is a literature on confiding, and it does not describe a transgression — it describes the mechanism by which closeness gets built at all, in every relationship a person has. What the research on how people open up actually shows sets it out. Whether a particular instance of it is a betrayal is not a question that research can answer, which is the whole difficulty of this subject. More pieces like this under Dating.

Where the study gets sharper is the association. Among women, more of these behaviours went with lower relationship satisfaction (r = −.37), higher relationship ambivalence (r = .49), and higher attachment avoidance (.27) and anxiety (.29). Among men the pattern held and ran smaller on satisfaction (−.32) and larger on ambivalence (.50) and avoidance (.34).

Read carefully, that is a correlation between reporting these behaviours and reporting an unhappy relationship, measured at the same moment, in a sample of married American parents of toddlers. It cannot tell you which came first, and the authors do not claim it can. What it does suggest is that these behaviours are more plausibly a symptom than a cause — which is roughly the opposite of the gateway framing, and is the next section's subject.

#9

The gateway claim has never been tested in time order

The engine of almost every page about micro-cheating is an escalation story: small things lead to bigger things, so catch the small things. It is intuitive, it is why the subject feels urgent, and it has not been tested.

What exists is cross-sectional. Surveys find that people who report more social-media-based infidelity behaviours also report lower satisfaction, more ambivalence, and more attachment insecurity, all at the same sitting. Other surveys find that heavier or more compulsive social media use goes with more of these behaviours. None of that establishes an order of events, and in a single snapshot it cannot: an unhappy relationship producing secretive messaging looks identical in the data to secretive messaging producing an unhappy relationship.

What would settle it is a study that measured ambiguous online behaviour at one point and infidelity later, in the same people. Looking for one turns up nothing. The nearest longitudinal work in the area studies what happens after someone discovers an affair, and one of its findings runs the wrong way for the gateway model: discovery was associated with less subsequent monitoring of social media, not more.

So the honest statement is this. Nobody has shown that liking a photo, keeping a warm conversation going, or confiding in a friend predicts anything at all about later infidelity. It might. The claim is not absurd. It is simply undemonstrated, and the confidence with which it is asserted across this subject has no empirical basis whatsoever. This site has had to write that sentence about a long list of things — about what distance does, about the rules with numbers in them, about intermittent reinforcement in dating — and it is worth saying each time, because an untested mechanism presented as a known one is how a reader ends up acting on nothing.

#10

The one attempt to build a measure, and the part of it that failed

In 2026 somebody tried to turn the neighbouring idea into an instrument, and what happened is more informative than a successful validation would have been.

Ahmet Aldemir's paper in Frontiers in Psychology evaluates a model of digital fidelity built on two axes: an individual one, which the paper calls Digital Fidelity Ethics, and a couple-level one, Boundary Management. The data come from an online convenience sample collected in September and October 2025 — 762 people in total, 285 for the exploratory analysis and 477 for the confirmatory and Rasch work — with a mean age of 28.15, mostly women, mostly graduates, mostly in a relationship. The authors are at Tokat Gaziosmanpaşa University in Türkiye; the paper does not state where participants lived.

The conventional analyses looked fine. The two-factor structure held up, and it held up identically for men and women, which is not nothing in an area where gender differences are assumed everywhere.

Then they ran a Rasch analysis, which asks a harder question: can this scale actually tell two people apart? For the seven-item ethics dimension, the answer was no. Person-separation reliability came out at 0.432, against a conventional threshold of 0.70, with a targeting mismatch of +2.053 logits and what the authors describe as a severe ceiling effect. In plain terms, the items were far too easy: almost everybody agreed they would behave ethically online, so the scale could not distinguish among them. The authors' own explanation is the obvious one — people do not volunteer that they are unethical on a questionnaire.

There is a second result in the same paper that deserves attention. The dimension that did not collapse was the other one: Boundary Management, the three items about what a couple has agreed is private and what is shared. Its reliability is borderline by the paper's own account, at CR = 0.688, and three items is thin. But it is the half that survived contact with a serious psychometric test, and it is the half that is about an agreement between two people rather than a judgement about one person's character.

That is the shape of the whole subject in miniature. Asked as a question about whether someone is the kind of person who does this, it does not measure. Asked as a question about what the two of you have agreed, it starts to.

#11

What watching does, and this is the measured part

The thing a micro-cheating article usually recommends — closer attention — is also, as it happens, one of the few things in this area with real data behind it. The data are not encouraging.

Start with Amy Muise, Emily Christofides and Serge Desmarais in 2009, which is a small, old, cross-sectional study whose numbers nevertheless say something that has not been improved on. Three hundred and eight undergraduates, aged 17 to 24, completed a survey about Facebook use and jealousy. In a hierarchical regression, trait jealousy and gender alone accounted for 46% of the variance in Facebook-related jealousy. Personality and relationship factors added 2%. Time spent on Facebook added another 2%, significant but small, for a final model explaining 48%.

That 46% is the figure to keep. The overwhelming majority of what predicted how jealous someone felt about what they saw was a disposition they brought with them. The platform contributed something real and marginal on top.

Their interpretation is the loop, and it is worth quoting the mechanism accurately because it is exactly what micro-cheating describes: using the site exposes people to often ambiguous information about their partner that they would not otherwise have access to, and the new information incites further use. Their own coding of participants' written answers found a theme they labelled lack of context in 7.4% of responses, and another they labelled addiction — difficulty limiting time spent on a partner's profile — in 10.3%. The authors are careful about what they have shown: further research is needed, they write, because the nature of their data could not fully explain the process. The loop is their reading, not a tested causal model.

Two further pieces fit around it. Robert Tokunaga published the scale that measures this behaviour in 2011, under a title that does the summarising for us: Social networking site or social surveillance site? And in 2022 Janneke Schokkenbroek and colleagues traced a path in 346 people from a partner's phone use, through lower perceived responsiveness, through anxiety, to electronic surveillance of that partner — with the authors noting that their model accounts for around 5% of the variance. That study, and the broader research on phones between two people, is set out in what phone behaviour on a date actually signals.

And here the Cornell conference paper earns its place. The belief its authors found running through the online discussion of relationship transgressions was that proof of fidelity or its absence can be located on a partner's phone — and they argue that this belief functions as a licence, turning an invasion of privacy into an entitlement. That is the thing to watch for in yourself, and it is the reason the safety section is at the top of this page rather than the bottom.

#12

The counsellors do not agree either, which tells you what kind of question this is

If the disagreement were only among undergraduates, you might expect expertise to settle it. It does not.

Naomi Moller and Andreas Vossler interviewed seven experienced couple counsellors about how they define infidelity and analysed the transcripts thematically. What they found was a tangle of conflicting definitions — among the counsellors themselves, and in what the counsellors reported hearing from clients. Their conclusion is that infidelity is socially constructed, and they note the practical consequence for research: definitions used in studies range from intercourse only to much broader categories of extradyadic activity, and the broader the definition, the higher the prevalence figure that comes out of it. Which is to say that every statistic about how common cheating is depends on a definitional choice made before anyone was surveyed.

That is worth holding next to the Kruger table. The 456 undergraduates disagreed about where the line was. The seven professional counsellors disagreed about where the line was. Nobody has produced the authority that would settle it, and the structure of the thing suggests nobody can, because it is not that kind of question. There is no fact about hugging for eleven seconds waiting to be discovered.

What follows from that is not relativism. Two things follow, and they are both practical.

The first is that a partner who draws the line somewhere other than you do is not thereby lying, gaslighting or minimising. They are, statistically, a normal respondent. The genre's framing — that there is a real answer, that your partner knows it, and that their disagreement is therefore evidence of something — is the part of this subject that does the actual damage.

The second is that the boundary has to be stated to exist. This is where micro-cheating as a concept has it backwards: it describes a line that has been crossed without ever having been drawn, and then treats the crossing as discoverable. The research says the line is not out there to be found. Which leaves one place it can come from, and it is the subject of the last section.

#13

What the evidence will support you doing

The defensible list is short, and worth separating from the much longer list of things this subject asserts.

Name the specific behaviour, not the category. The word micro-cheating imports a verdict — it has cheating in it — into a conversation about a behaviour that, in the one study that measured it, half the population does not classify as cheating at all. Saying "when you message her late at night I feel shut out" opens a conversation. Saying "you are micro-cheating" announces a finding that the evidence does not contain. This is the same problem as the rest of the vocabulary: a label that sounds like a diagnosis, built on nothing that could diagnose.

Expect to be 30 points apart, and treat that as normal. The standard deviations in the Kruger table are the most useful thing on this page for a real conversation. Your partner is not pretending not to understand. On this class of behaviour, a large minority of people genuinely sit at the far end from you.

Agree the boundary rather than auditing for it. It is the only version of this that survived a psychometric test — the boundary-management half of the 2026 instrument held while the character-judgement half collapsed — and it is the only version that does not require one person to be the authority on an answer that does not exist. The research on what questions actually reveal about compatibility is relevant here, and its central finding applies: what predicts how a relationship goes is overwhelmingly what the two people report about the relationship, not traits either of them has.

Treat an increase in your own watching as information about you. 46% of the variance, in the one study that partitioned it, was the watcher's own disposition. If the checking has gone up, the most likely thing that changed is not the evidence.

Know what is actually a symptom. In the one sample of married couples, these behaviours travelled with relationship ambivalence (r = .49 in women, .50 in men) more strongly than with anything else measured. If secretive messaging has appeared, the evidence points at the state of the relationship rather than at the messages, and nothing has shown that stopping the messages addresses the state.

And the limits, in the same type size. There is no prevalence estimate for this. There is no instrument. There is no longitudinal study. The behaviour table is from 456 American undergraduates in 2013 and has not been replicated in an older sample or outside the United States. The prevalence figures are from 338 white, educated, married American parents of young children. The attachment findings are correlations from one sitting. Everything in this section is an inference offered as one.

The one move with support behind it is the same one that ends most pages here, and the one the whole apparatus of watching is built to avoid: ask. Ambiguity resolved by conversation is resolved. Ambiguity resolved by surveillance, in the only study that looked at the loop, produces more ambiguity and more surveillance.

Pulling it together

The short version is that this is a real question wearing a fake answer.

The question — when does ordinary warmth towards somebody else become a betrayal — is genuine, common and worth taking seriously. The answer the term supplies, that there is a category of small betrayals which can be identified from outside by watching, is not supported by anything. There is no measurement of it, and the measurement of the thing next to it says the opposite: the behaviours in question sit where people disagree most, the disagreement is large enough to put two reasonable people 60 points apart, and how high you rate them is partly a fact about your own attachment anxiety rather than about your partner.

Three things are worth carrying away. The first is that the line is agreed, not discovered. The only half of the one serious attempt at an instrument that survived a hard psychometric test was the half about what a couple has decided together; the half about whether someone is the kind of person who does this could not tell two people apart at all.

The second is that the escalation story — small things lead to big things, so catch the small things — has never been tested in time order by anyone. In the one married sample where these behaviours were counted, they ran between 5% and 12% and travelled with relationship ambivalence more than with anything else, which makes them a more plausible symptom than a cause.

The third is about the method the genre recommends. In the one study that partitioned the variance, 46% of how jealous people felt about what they saw on a partner's profile was their own trait jealousy, and the platform added two points. Looking produces ambiguous information, and ambiguous information produces more looking. The way out of that loop has never been a better checklist.

Frequently asked questions

Is micro-cheating a real thing?

As an experience, yes — people really do feel betrayed by things that fall short of an affair, and that feeling is not irrational. As a researched category, no. Searched on 1 November 2026, PubMed returns zero records for "micro-cheating" or "microcheating", and Europe PMC returns one — a 2026 Frontiers in Psychology paper about digital fidelity that mentions the word once, in its introduction, as an example. Beyond that there is a 2023 Association of Internet Researchers conference paper that describes itself as an in-progress pilot analysing Reddit discussions, and a 2024 Turkish-language review article. There is no agreed definition, no validated instrument, no prevalence estimate and no study of consequences. The term was popularised by an Australian psychologist and television personality in media interviews around the end of 2017, not by a study.

What counts as micro-cheating?

Nobody can tell you, and the closest thing to an answer is a measurement of how much people disagree. Kruger and colleagues asked 456 undergraduates in 2013 to rate 27 behaviours on a 0–100% scale for how far each indicated cheating. The behaviours this term usually covers landed in the middle: forming a deep emotional bond 52.4, spending lots of time together 52.2, going out to dinner 41.4, talking on the phone several times a week 40.1, sharing secrets 36.5, a hug of more than ten seconds 34.5, telling dirty jokes 25.9. The standard deviations were around 30 to 35 points, which means the population is split rather than mildly undecided. Seven experienced couple counsellors interviewed by Moller and Vossler disagreed with each other too, and concluded infidelity is socially constructed. So the honest answer is that it counts if the two of you have agreed it counts, and it has to be stated to exist.

Does micro-cheating lead to actual cheating?

Nobody has tested it. The escalation story is the engine of almost every page on the subject and it rests on cross-sectional surveys, which cannot establish an order of events: an unhappy relationship producing secretive messaging looks identical in a single snapshot to secretive messaging producing an unhappy relationship. There is no study that measured ambiguous online behaviour at one point and infidelity in the same people later. The nearest longitudinal work looks at what happens after an affair is discovered, and found that discovery was associated with less subsequent social media monitoring, not more — which runs against the gateway model rather than for it. The claim may turn out to be true. It is currently undemonstrated.

How common is micro-cheating?

There is no prevalence figure for the term, because there is no instrument. There is one for a nearly identical set of behaviours. McDaniel, Drouin and Cravens surveyed 338 married or cohabiting individuals from 176 US families in 2017 using seven items — feeling uncomfortable if a partner read your messages, having messages you would hide, confiding in someone else online instead of your partner, messaging old romantic partners, getting defensive when interrupted, hiding what you say online. Counting everyone who agreed at all: 12%, 6%, 7%, 6%, 6% and 5% respectively, with 5% wondering whether their partner would mind. Men and women did not differ significantly on any item. The sample was 92% white, 73% university-educated parents of children under five, so it does not generalise far — but it is the only real base rate in the area, and it is nothing like an epidemic.

Why do my partner and I disagree about what counts as cheating?

Because that is the normal result, and because your own attachment pattern shifts your answer. In the Kruger data the ambiguous behaviours carried standard deviations of 30 to 35 points on a 100-point scale, so two people placing the same behaviour 60 points apart are both ordinary respondents. On top of that, participants higher in attachment anxiety rated 18 of the 27 behaviours as more indicative of cheating, with the effect concentrated in exactly that ambiguous middle — a road trip (β = .192), sitting in someone's lap (.186), dinner (.185), a formal event (.166), holding hands (.149). Attachment avoidance went the other way on five sexual items. Those are modest coefficients from one student sample, and they are not a licence to tell a partner their discomfort is just their attachment style. They are a reason to treat the line as something to negotiate rather than something one person can detect.

Should I check my partner's phone?

The research on checking is some of the only measured material in this area, and it does not recommend it. Muise, Christofides and Desmarais surveyed 308 undergraduates about Facebook use and jealousy: trait jealousy and gender alone accounted for 46% of the variance in how jealous people felt about what they saw, personality and relationship factors added 2%, and time on the site added 2% more. The authors describe a loop in which use exposes people to often ambiguous information about a partner, and that information drives further use — and they say explicitly that further research is needed because their data could not fully explain the process. Tokunaga published the scale for measuring this behaviour in 2011 under the title "Social networking site or social surveillance site?". Schokkenbroek and colleagues traced a path in 346 people from a partner's phone use through lower perceived responsiveness and anxiety to electronic surveillance. And if checking has become demanded rather than chosen — passwords, location, an account of your day — that is in the territory the warning signs at the top of this page describe.

How is micro-cheating different from cushioning, breadcrumbing or pocketing?

They describe different arrangements and they have very different amounts of evidence. Micro-cheating is ambiguous warmth towards somebody outside the relationship, and has the least research of the four: zero PubMed records. Cushioning is keeping other people in reserve while in a relationship, and the behaviour underneath it has been studied since 2014 as back burners, where the striking result was how ordinary it turned out to be. Breadcrumbing is sporadic non-committal contact that keeps a connection alive without progressing it, and has the most direct research of the four under its own name — 19 records in Europe PMC, against one for micro-cheating. Pocketing is being kept out of someone's visible life, and sits on the literature about relationship concealment. The common pattern is worth noticing: in each case the term arrived from media or social media, and the research either came later or belongs to something that was already being measured under another name.

If there is no research, is the feeling of being micro-cheated on invalid?

No, and that is a different question from whether the category holds up. The absence of a literature means nobody can tell you where the line is or that your partner crossed it — not that your reaction is wrong. Two measured things support taking the feeling seriously. In the married sample, more of these behaviours went with higher relationship ambivalence (r = .49 in women, .50 in men) and lower satisfaction (−.37 and −.32), so a sense that something is off often accompanies a relationship that genuinely is not going well, even if the direction of cause is unknown. And the one dimension of the 2026 digital fidelity instrument that held up psychometrically was boundary management — what a couple has agreed is private and what is shared — which is the part a conversation can actually resolve. What the evidence does not support is the step from feeling to verdict, delivered as a category with the word cheating inside it.

Sources

What this article draws on. For the broader bibliography see our sources page.

Keep reading