Cognitive Daily

I like rock music, but my father-in-law doesn’t. My son Jim likes horror movies, but his mom doesn’t. While some of our preferences can be explained easily—for example, we usually don’t like things that cause us pain—others are more difficult to understand. When there’s not an obvious reason for a preference, mere exposure to an item can lead to preference. Studies have found this “mere exposure effect” for words, photos, objects—nearly anything, really.

What’s less certain is what causes the mere exposure effect: two competing explanations have been proposed. The first is the uncertainty reduction hypothesis—the idea that we simply prefer things which are more familiar. The second explanation, the misattribution hypothesis, is more complex. It begins with the same concept, that we’re sometimes more familiar with one item than another, but suggests that whether we actually develop a preference depends on whether or not we remember why we’re familiar with it. If an item was only presented subliminally, meaning we’re not aware of the instance when it was presented, then we’re more likely to misattribute that recollection to actual preference, compared to items presented liminally (overtly). So preference occurs more readily when it’s subtly suggested to us, compared to when we’re hit over the head with it.

Some research supports both hypotheses, however: both overt and subliminal presentations of items can lead to preference. But what if some items are presented overtly and some are presented subliminally? Which will we prefer then? Man-Ying Wang and Hsio-Chuan Chang of Soochow University believe they have devised an experiment to determine which of the two views better explains how we form preferences (“The Mere Exposure Effect and Recognition Memory,” Cognition and Emotion, 2004). To develop their experiment, Wang and Chang needed to make one more distinction: between knowing and remembering. When we “know” something, we’re completely aware of its existence, but we don’t recall the specific instance when we learned of it. When we “remember,” we’re recalling a particular occasion. If remembering leads to preference, then that supports the uncertainty reduction hypothesis. If knowing does, that supports the misattribution hypothesis.

In their experiment, Wang and Chang played music excerpts of classical music for listeners. Next they played the same excerpts in a random order, along with some new excerpts that the listeners hadn’t yet heard. Listeners rated the items for how much they liked them on a scale of 1 to 5 and indicate whether or not they were new. Here is a summary of the results.

Actually Old
Actually New
Judged Old
Judged New

When listeners thought the excerpt was old, they liked it better—whether or not the excerpt actually was old. So listeners prefer the items they remember, rather than those they know—when they believe they recall hearing a specific excerpt, they like it better. Wang and Chang argue that this result supports the uncertainty reduction hypothesis. They suggest that we prefer things we’ve seen or heard before because these things are less likely to be dangerous: after all, if it didn’t kill us the first time we saw it, it’s probably safe. So it might be that the reason I like rock music while my father in law doesn’t is because I had heard it as a child, while he didn’t. Since I had more exposure to rock, I like it and he doesn’t. Now apparently we just need to find out who showed all those horror movies to Jim when he was younger!


  1. #1 Garrett
    June 28, 2005

    I’m not a statistician, but couldn’t the results simply indicate a higher probability of a preference for music the subjects have had a second exposure to? Meaning, if I hear a piece of music and then hear it again an hour later, I’ve heard the piece twice in a short time, and am more apt to be able to make a decision.

    Maybe it just indicates that people’s first impression of classical music isn’t always accurate. 🙂

  2. #2 Dave Munger
    June 28, 2005

    “if I hear a piece of music and then hear it again an hour later, I’ve heard the piece twice in a short time, and am more apt to be able to make a decision.”

    Actually, what the study is saying is the opposite of what you’ve suggested here. Whether you’ve heard the piece twice or not isn’t what matters; it’s whether you think you’ve heard it before. When you believe it to be familiar, you prefer it over a piece you believe to be unfamiliar, whether or not that’s actually the case.

  3. […] Psychology, Research ::

    From Elearningpost, June 27, 2005: This is a wonderful explanation for why we prefer some items over others. This can have some learning design implications. The findin […]

  4. #4 Garrett
    June 29, 2005

    Ah. That makes sense. Thanks for the very interesting piece; I enjoy your articles.

    Oh, ok, I re-read it and see the part about ‘misattribution’ now. I get it.

    Now I’d love to see someone create entirely new music, perhaps at random, and play it for a large population, trying to isolate the melodies that people think they’ve heard, but never possibly could have.

  5. #5 J.K.
    July 5, 2005

    Interesting. How did my post get displayed as a comment here? Is this an automatic process in WordPress?

    Wrote my comments based on a cursory reading of your article here. Now that I’ve read it more carefully, like Garrett, I understand the ‘misattribution’ and ‘uncertainty reduction’ hypotheses better and what the researchers did to test the hypotheses.

    I have a few questions though:
    1. What was Wang & Chang’s sample size?
    2. Couldn’t it be that one simply remembers things that one likes and forgets things that one dislikes? There’s a far greater difference (around 16 percent) between Actually Old, Judged Old (3.33) and Actually Old, Judged New (2.79) than the difference (around 10.4 percent) between Actually New, Judged Old (3.07) and Actually New, Judged New (2.75).

  6. #6 J.K.
    July 5, 2005

    Sorry. To re-phrase:

    Could the researchers be inverting the symptom (remembering) and the cause (preference)?

    The difference (around 16 percent) between Actually Old, Judged Old (3.33) and Actually Old, Judged New (2.79) and the difference (around 10.4 percent) between Actually New, Judged Old (3.07) and Actually New, Judged New (2.75) are far greater than the difference (4.9 percent) between Actually Old (3.06) and Actually New (2.91).

  7. […] ferences

    Enregistr¨¦ dans : G¨¦n¨¦ral, Esth¨¦tique ¡ª Pierre @ 8:03 pm

    Cognitive Daily — Some insight into how we develop preferences: Some insight into how we develop preferences Fil […]

  8. #8 Dave Munger
    July 5, 2005

    J.K., I think you’re saying the same thing Wang and Chang were saying–that the perception of remembering is more important than actually remembering. It doesn’t matter, in fact, whether the item was subliminally presented or overtly presented. So they’re discounting another theory that supported subliminal presentation as more effective in causing preference than overt presentation.

    I also don’t think it’s appropriate to average the figures together the way you have. You can’t combine all the “actually old” numbers with a simple average, because they correspond to different numbers of individuals. To take a simple example, it might be that in the “actually old” condition, two people judged it old, and one person judged it new. So you can’t just average the two preference ratings together–you’d need to weight the “judged old” portion twice as much as the “judged new” portion.

    Regarding sample size, I don’t have the article in front of me, but I think it was about 140. But this shouldn’t matter—as long as the results are statistically significant, the sample size could be 10 or 10,000, unless there’s some reason to believe there is sampling bias. But sampling bias could be present even in a sample of 10,000.

  9. #9 J.K.
    July 5, 2005

    Dave, I’m not saying the same thing Wang and Chang were saying. If I understand your article correctly, they claim that the symptom is preference and the cause is remembering while I think that the symptom is remembering and the cause is preference.

    But you are right that I cannot simply combine all the ¡°actually old¡± numbers with a simple average, because they correspond to different numbers of individuals. If the figures are available, weighted averages would be more correct.

  10. #10 Mary Hall
    July 5, 2005

    Hi, Dave
    JK’s comments ring a bell with me – here’s an example to illustrate. My kids & I listen to popular music (not classical). Every now and then we hear a song for the first time, and it sounds familiar! These ones almost inevitably seem to go on to become hugely successful on the charts.

    My assumption is that the song appeals at a level where I assume that I’ve heard it before because I like it so much – the causality being in the opposite direction towhat your research seems to be suggesting. The observation that songs that fall into this category usually become huge hits suggests that the same thing happens with other listeners as well.


  11. #11 J.K.
    July 5, 2005

    My assumption is: some people judge certain things that are actually old as something old (correctly), and other things that are actually old as something new (incorrectly). If so, I think the reason they do so is simply because they remember certain things when the impression is greater (often because they like it).

    NOT the other way round: that people like certain things because they remember them. From what I’ve learnt (though quite limitedly) through a psychology course earlier, a correlation of figures does not always translate into a specific cause-effect relationship.

  12. #12 J.K.
    July 5, 2005

    Sorry, not “some people” but “most people” (if not all).

  13. #13 Dave Munger
    July 6, 2005


    I think if you take Wang and Chang’s results in isolation, yours is a valid interpretation. Where I may not have made myself clear enough in the post is that Wang and Chang’s goal was not to determine if mere exposure led to preference, but instead, to explain the dozens of studies which show that mere exposure does indeed lead to preference. To do so, they needed to come up with a way to distinguish between the two leading hypotheses as to why this is true. So in their study, they exposed participants to much less music than typically leads to a preference, in order to be able to distinguish between preference based on specific recollection of an item and general knowledge of an item—to see which of these two was more likely to lead to preference. If you take their study in isolation, it’s difficult to see which causes which. However, in the context of many other studies which show that exposure does indeed lead to preference, it offers additional insight as to what the mechanism for this effect is.

New comments have been disabled.