Figures on American religious attendance vary widely between sources. The variation comes from differences in how the question is asked and who is doing the counting.
Surveys ask people to describe themselves
The most common method asks respondents how often they attend religious services, with answer categories from never to weekly or more.
Attendance is socially approved in much of the United States, which introduces pressure toward answers that present the respondent favorably.
Researchers have found that asking about a specific recent period, rather than about general frequency, produces noticeably lower figures from the same populations.
Head counts measure something narrower
Congregations count people in the building, or count giving envelopes, or estimate from parking. These methods observe behavior rather than relying on recall.
They also miss things. A person attending a different congregation, watching a broadcast or participating online is present religiously and absent from the count.
Aggregating congregational counts nationally is difficult because reporting practices differ by denomination and many congregations report nothing at all. Independent and nondenominational churches, a large and growing share of American congregations, have no central body collecting figures from them.
Question wording moves the result
Asking whether someone attended in the last seven days produces a different figure than asking how often they usually attend, and both are asked in American research.
Including or excluding weddings, funerals and holiday services changes the number, as does whether online participation is counted as attendance.
These are definitional choices rather than errors, which is why two reputable sources can report substantially different levels of participation in the same year.
Affiliation and attendance diverge
Identifying with a tradition and attending its services are separate measures, and in the United States they have moved apart over recent decades.
A large share of Americans name a religious identity while attending rarely, which means affiliation figures overstate participation considerably.
The category of people reporting no religious affiliation has grown, and researchers distinguish within it between the actively nonreligious and the merely unaffiliated.
Streaming redefined the boundary
Broadcast and online services expanded rapidly, and many American congregations retained hybrid participation afterward.
Counting a viewer is genuinely ambiguous. Whether attendance means presence, attention or a completed stream is a question the older methods never had to answer.
Any figure quoted about American religious practice therefore carries an implicit definition, and comparing across sources requires knowing which one was used.