The Meanings of "Think" and "Believe"

Today's post is by Neil Van Leeuwen who talks about his recent research with Larisa Heiphetz on the differences in meanings between "think" and "believe". For related research by Heiphetz and Van Leeuwen, see herehere, and here

Neil Van Leeuwen

Do “think” and “believe” mean the same thing? Consider two sentences:
  • Jill believes that God exists.
  • Jill thinks that a lake bigger than Lake Michigan exists.

Both sentences attribute mental states to Jill. And each breaks down into an attitude (thinks/ believes) and a content (that God… / that a lake…). So we can sharpen our question: if we set the contents aside, do the words “thinks” and “believes” convey the same attitude type (or manner of processing)?

Many philosophers and cognitive scientists talk and write as if the answer were yes—as if the words “think” and “believe” were interchangeable, at least in propositional attitude reports (i.e., as if “thinks that p” and “believes that p” referred to the same attitude).

But do lay speakers use “think” and “believe” to convey the same thing? In addition to its intrinsic interest, answering this question is important for philosophers and psychologists: if the answer is no, then we theorists should be cautious about treating them as interchangeable, unless we clarify their meanings as terms of art (which we rarely do).

Larisa Heiphetz

Larisa HeiphetzCasey Landers, and I address this very question in our new paper “Does ‘think’ mean the same thing as ‘believe’? Linguistic Insights into Religious Cognition” (forthcoming, Psychology of Religion and Spirituality). We have two main findings:

  • Lay speakers of American English are more likely to use “think” for everyday, matter-of-fact attitudes (factual beliefs) and “believe” for religious and political/ideological attitudes (religious or ideological credences).
  • This difference remains even when contents are held fixed: a factual belief that pis more likely to be described as “thinks”, but a religious credence that pis more likely to be described as “believes”.

These findings emerged from both corpus linguistics and behavioral studies. (Important caveat: we are not saying that these are the only uses of these two verbs; other uses exist, such as hedging. We are saying, rather, that an important pattern of differential usage exists that tracks differences in attitude type.)

Our first study explored the Corpus of Contemporary American English (COCA), a body of text with over 520 million words going back to 1990. Over several searches, we found that “believe that” had a multiple religious collocates (words or phrases associated at a higher rate than would be expected by chance), words like “miracles” and “Allah,” while “think that” had no religious collocates whatsoever. 

Casey Landers

Furthermore, we searched various religious propositional complements, like “that God is,” to see whether “think” or “believe” would precede them with significant frequency: “believe” occurred as a collocate prior to such phrases often; “think” was never a collocate. So “believe” collocates with religious complements in both directions; “think” has no specifically religious associations at all.

We then did four behavioral studies to see what attitude verbs participants would choose for different attitude report contexts (religious vs. factual for Studies 2-4; religious vs. factual vs. political for Study 5). Consider the following fill-in-the-blanks, which are representative of our stimuli in Studies 2 and 3. 

  • Zane _______ that Jesus turned water into wine.
  • Fred and Yuriana _______ that George Washington was the first U.S. President.
  • Nick _______ that cassiterite is the chief source of tin.
  • Sharon _______ that she will meet her mother at the grocery store today. 

The first context is of course religious. The other three are matters of fact: well-known fact, esoteric fact, and everyday life fact. We found participants were more likely to choose “believe” for religious attitude reports and “think” for factual attitude reports. This was true in both forced-choice (Study 2) and free-response paradigms (Study 3).

In Study 2, for example, participants chose grammatically appropriate forms of “believe” in 89% of the religious contexts (11% for “think”) and only 18% of the time in factual contexts (82% for “think”). Study 2 data look like this:





Study 3 used the same stimuli as Study 2, just with a free-response paradigm (participants could fill in any words they wanted). We found that participants still used “believe” for religious contexts 51% of the time; they used “believe” for matter-of-fact contexts only 8% of the time. Another result from Study 3 was that participants were more likely to use “know” for factual than religious items. (We weren’t expecting this, since our overall sample had a reasonable balance of religious and non-religious participants.) 

A skeptic might respond to all this with a question I’ve heard before in response to my research that differentiates religious credence and factual beliefCan’t these differences be accounted for by differences in content (rather than attitude type)? I’ve always maintained the answer to this is no, because in principle anything can be sacralized, which means you can’t capture the difference between religious and non-religious “beliefs” just by appealing to contents. 

Our study 4 gave empirical traction to this point. Consider these two contrasting vignettes:
All last year, Terry would get splitting headaches in the afternoons. Sometimes her friends would offer her aspirin. But Terry belonged to the Church of Christ Scientist, which teaches that prayer, not medicine, is the way to cure medical ills. So Terry always refused when her friends offered, because she _______ that aspirin was not a cure.
All last year, Kerry would get splitting headaches in the afternoons. Sometimes her friends would offer her aspirin. But Kerry had tried using aspirin many times in the past, and the headaches just kept happening, whether she took aspirin or not. So Kerry always refused the aspirin her friends offered, because she _____ that aspirin was not a cure. 

Our participants saw these and other matched pairs and chose between forms of “think” or “believe.”

The key feature of Study 4 was that we matched the contents of the attributed attitudes in each vignette pair. In the pair above it’s that aspirin is not a cure. We reasoned that if the difference between use of “think” and “believe” persisted when contents were held fixed, that would best be explained by the view that the two verbs were being used to report different attitudes (“believe” for religious credence and “think” for factual belief).

Sure enough, the difference held up. Participants chose “believe” for 74% of the religious vignettes and for 38% of the matter-of-fact vignettes.

Finally, Study 5, which only appears in supplemental materials, had the result that participants treated political convictions intermediately: they use “believe” for political attitudes less than for religious credences but more than for matter-of-fact contexts. (Example stimulus: Rob used to _______ that immigration helps the economy.)

What’s the upshot of this research? Three points are significant.

First, the pattern of lay speaker differentiation between “think” and “believe” supports earlier theoretical work distinguishing religious credence and factual belief: a good explanation for why people use those verbs differently is that there are different attitudes to which to refer.

Second, many philosophers and cognitive scientists use “think” and “believe” in less discriminating ways than lay speakers of American English. Many theorists write as if there is just one propositional attitude reported by these verbs, and they implicitly assume the “folk” assume this too. If so, such theorists are wrong on two accounts: (1) in thinking there is only one cognitive attitude for both verbs and (2) also in thinking it is an assumption of folk psychology that there is only one.

Third, a range of intellectuals concerned with religious “belief” should rethink their positions. Various scholars in religious studies and anthropology of religion write as if religious “believers” regard the existence of their respective gods in the same matter-of-fact way they regard the existence of their next-door neighbor. Such writing suggests what my co-authors and I call The Single Belief View, which holds that all “beliefs” (as scholars widely use the term) are of one cognitive attitude type. 

Furthermore, New Atheists and religious apologists—including some philosophers of religion—are strange bedfellows in clinging to The Single Belief View, the former for the sake of attacking religious “beliefs” and the latter for the sake of justifying them (or pretending to).

But our experimental results support The Varieties of Belief View, which posits more than one attitude (possibly many). I find it interesting that the otherwise heterogeneous scholarly groups just mentioned share a view that is mistaken, as many lay speakers of English implicitly know. 

Comments