Message cards separated into what happened and what it might mean

Research brief

Dating Communication Signals: What the Research Can and Cannot Show

Message counts, silence and inconsistent contact can describe an experience. They cannot reveal one person's motive.

In this brief
The short answer: record the behavior before explaining it. “Three messages arrived after I said no” is observable. “That person cannot handle rejection” is an interpretation the messages alone cannot prove.

Findings at a glance

6,034U.S. adults in Pew's 2022 survey
54%of recent women users felt overwhelmed by message volume
64%of recent men users felt insecure about too few messages

Pew's 2022 U.S. survey shows how differently people can experience the same platform. Among recent online-dating users, women were more likely than men to say the number of messages felt overwhelming. Men were more likely than women to say too few messages made them feel insecure. These are population-level reports. They do not tell us whether one sender is interested, avoidant, busy, unsafe or trying to manipulate someone.

The same caution applies to unwanted contact. In Pew's 2019 survey, 37% of online-dating users said someone continued contacting them after they said they were not interested, and 35% reported an unsolicited sexually explicit message or image. Those are reported experiences, not explanations of motive.

Silence has to be defined before it can be studied

Research does not use one universal test for ghosting. A 2023 paper developed a ten-item questionnaire for reported experiences of sudden communication exclusion. Its items cover several different behaviors: abrupt silence, blocking, delayed or vague replies, inconsistent engagement and avoiding a meeting.

That breadth matters. A measurement scale can group answers for research, but it does not make every delayed reply a case of ghosting. Nor does a score explain the other person's reason. The study tested an instrument's structure. It did not estimate prevalence for all daters or build a motive detector.

One study's percentage is not everybody's percentage

In a 2020 online snowball sample of 409 Italian adults, 67.4% reported stopping message replies, 58.5% reported punitive silence and 18.6% reported ending a relationship by disappearing. The behaviors are stated plainly, which makes the study useful.

The limits belong beside the numbers. The sample was not designed to represent all adults, and the three-item ghosting measure had low internal consistency. “18.6% of this sample” is accurate. “18.6% of people ghost” is not supported by the study.

Association is not diagnosis

A 2023 study of young adults in India and Spain found positive correlations between breadcrumbing-engagement scores and both anxious and avoidant attachment scores. The associations differed between the two country samples. That is evidence that measured variables moved together in those groups.

It is not evidence that a sporadic message diagnoses an attachment style. The study was cross-sectional, used non-probability samples and compared groups with different demographic profiles. It cannot establish which factor came first or whether either caused the other.

The evidence has blind spots

A 2025 scoping review of 125 qualitative online-dating studies found repeated sample imbalances. Research often centered young, educated, ethnic-majority participants in Western societies, while other groups were underrepresented. It also tended to emphasize risk and negative communication experiences.

That does not make the research useless. It changes the question a careful reader asks: who was actually studied, and what did the measure ask? If the answer does not resemble the person or situation at hand, the study should not be used as a personalized explanation.

How this brief was built

Freshness: Pew figures will be reviewed when a newer comparable wave appears. Scholarly sources are checked at least annually for corrections or retractions. To report a correction, use the contact form.

Sources, samples and limits

SourcePopulationWhat it supportsMain limit
Pew, 20236,034 U.S. adults; fielded July 2022Platform use, message-volume experiences and reasons for useCross-sectional self-report; subgroup bases vary
Pew, 20204,860 U.S. adults; fielded October 2019Reported unwanted contact and perceptions of deceptionPerception is not a verified deception rate
Jahrami et al., 2023811 online respondentsStructure of a ten-item ghosting-experience measureInstrument study, not prevalence; crowdsourced sample
Biolcati et al., 2022409 Italian adults; online snowball sampleThree specifically worded communication behaviorsNon-probability sample; three-item alpha 0.56
Khattar et al., 2023304 adults in India and 334 in SpainAssociations between breadcrumbing and attachment scoresCross-sectional association; no diagnosis or causation
Diesen et al., 2025125 qualitative studies from 2014–2023Coverage and sample bias in the research literatureScoping review, not a pooled prevalence estimate

Put the distinction to work

Use the observation worksheet for one moment, or the pattern timeline when you need to compare several dated events.

Open the observation worksheet Build a pattern timeline

Keep reading