This week you might have seen some coverage resulting from a press release out of the University of Utah about a new inattentional blindness study by Seegmiller, Watson, & Strayer. Here are some examples of headlines resulting from the press release:
Why we don’t see what’s right in front of our eyes
(I’m not quite sure I understand the last one. Do Brits really have a new dance move that involves chest thumping? If so, please do tell.)
The implication of these headlines are: (1) that some people typically experience inattentional blindness and others don’t, and (2) that the new study entirely explains such individual differences. Both implications are false. The first is entirely unsubstantiated and the second is a massive overextension of what actually is an interesting result. The actual study, the press release, and the subsequent media coverage make a nice case study to explore how media coverage of science can create a false understanding among non-scientists about the nature of scientific inquiry.
The media is reacting to the finding that, under some conditions, differences in working memory capacity predict noticing of an unexpected gorilla. They over-generalize the finding to suggest that people who are high in working memory capacity are immune to inattentional blindness.
A look at the findings and their scope
Let’s take a step back and look further at those constraints. First, the study used a difficult task — keeping track separately of bounce passes and aerial passes — that likely taxes both working memory and attention. It’s possible that working memory capacity only predicts noticing when the task is particularly taxing and not in the more common case in which the primary task is relatively easy. Given that inattentional blindness has been documented across groups that differ widely in intelligence and presumably in working memory capacity, the predictive value of individual differences in working memory capacity might be limited to a fairly small subset of cases for which attention and memory are particularly strained. From the result, we just don’t know whether the demands of the task actually matter. My guess is that they do.
Second, when including all of the subjects in the analysis, individual differences in working memory did not predict noticing. You read that right. Overall, working memory capacity was unrelated to noticing. In order to reveal an effect of working memory on noticing, the authors first excluded anyone who counted the passes with less than 80% accuracy (that percentage is somewhat arbitrary, but fine). The logic for the restriction is reasonable—people might give up counting in the middle of the task if they happen to notice the gorilla, and people who don’t bother counting might also notice the gorilla for uninteresting reasons. The authors justifiably argue that such contaminating factors might obscure a relationship between working memory and noticing. But, the choice of cutoffs could well influence the resulting relationship, and the lack of an overall relationship means that you can’t predict who will notice and who won’t just by measuring working memory capacity.
I think there might be a more theoretically interesting reason why the authors found an effect of working memory capacity on noticing in those subjects who accurately counted passes, one that further limits the scope of the relationship: People were counting passes by the players wearing black, and the unexpected object was also black.
Typically, more people notice the gorilla when counting passes by the players wearing black than when counting passes by the players wearing white. The authors suggest in a footnote that with a video like this one—although not the exact one they used—the difference in rates of noticing isn’t huge. My concern is not about the difference in noticing rates (although they typically are about 20% more likely to spot the gorilla when counting players by the black team). My concern is about how attending to the black players might produce a correlation between noticing and working memory.
When focusing on the players wearing black, people have an attention set for black, and the gorilla falls into that attention set (i.e., it is black as well). It’s possible that people with high working memory capacity are more likely to notice other items matching their attention set than are people low in working memory capacity. That is, their spare capacity helps them incorporate other aspects of a scene that are consistent with their focus of attention. In other words, they only had an advantage in noticing because the gorilla shared features with the people they were attending. If true, there should be no effect of working memory if the unexpected object were more similar to the ignored items. The relationship between working memory and inattentional blindness revealed by the experiment might only apply in that special case in which the unexpected object happens to share the attended feature (black color) with the attended items (players wearing black). I would actually predict that there will be no relationship whatsoever between working memory and noticing if the unexpected object is different from the attended items on the critical attended dimension.
Implications for science and the media
The paper is among the first to carefully explore whether individual differences in working memory contribute to inattentional blindness. I think this is an interesting and intriguing result, one that shows how individual differences in cognition affect inattentional blindness, at least under some conditions. The paper carefully acknowledges prior work and does hedge its claims (although a little more hedging might have been merited in places). It is laudable for its rigor, including large enough samples to look for individual differences and controlling for factors that might have masked evidence for a link between working memory and noticing in the past.
So how did the media coverage take this interesting, but potentially limited-in-scope result and infer that the study somehow solved the case of the invisible gorilla? Any scientist reading the journal article would recognize that the correlation between working memory and noticing is imperfect and would separate speculative conclusions from definitive results. Unless the press release makes those limitations explicit, the media will not either. Unless the press release explicitly identifies the limited scope and imperfect correlation and flags speculation as such, an untrained reader (or headline writer) will naturally infer that the result and the speculation are one and the same. In this case, they will infer that working memory differences explain inattentional blindness in its entirety. By not reining in the speculation, the release suggests that the working memory is the primary (if not the only) reason that some people notice and some people miss unexpected objects.
This mistaken inference is dangerous. Everyone is subject to inattentional blindness because we all have limits on our attention capacity. Nobody is immune to the effects of inattentional blindness, and many people who were high in working memory capacity missed the gorilla, even under the conditions tested in this study. This is an important new finding, but it does not justify broad claims that working memory explains inattentional blindness in general. Perhaps these working memory effects will generalize to other tasks and contexts, but we don’t know that yet (and I have reasons to doubt they will generalize to all inattentional blindness tasks).
Scientists understand the value of a strong claim. Speculation based on an initial finding can drive new research designed to test or refute strong claims. Over time, and across many studies, the field settles on a more nuanced understanding. The problem comes when an initial strong claim and the associated speculations are presented to the public as definitive conclusions. If later studies reveal the limits and scope of the relationship between working memory and noticing, and press releases tout the relative lack of individual differences, the public will perceive the combined results as contradictory. That hurts the perception of science in general. If one day chocolate is bad for you and the next it is good for you, people assume that they can just choose what to believe because scientists really don’t have a clue. In some cases, scientists might well be clueless. But, in most cases, strong claims from preliminary evidence are not a sign of cluelessness but of a desire to make interesting findings known to the scientific community so that they can be refined. I’m fine with speculation in journal articles. Most media outlets don’t read the original scientific paper anyway. My concern is with overly strong claims pushed as definitive evidence in press releases that are designed to grab media coverage. To the extent that they can control the publicity process, scientists must rein in the speculative conclusions in their press releases, noting the limits of their evidence. If we’re lucky, press releases might even convey more basic information about such limits, thereby giving the public a more nuanced view of the scientific process.
Update 4-21-11: Fixed typos
Source Cited:
Seegmiller JK, Watson JM, & Strayer DL (2011). Individual differences in susceptibility to inattentional blindness. Journal of experimental psychology. Learning, memory, and cognition PMID: 21299325
[…] Invisible Gorillas, Working Memory & Social Media- via Invisible Gorilla Blog- So what did the new paper actually find? If you give people a highly demanding version of the basketball-gorilla task (separately count the aerial and bounce passes by the players wearing black) and exclude subjects who were unable to do the counting task well, those subjects remaining who scored highest on a measure of working memory were more likely to notice the gorilla than were those subjects scoring low on the working memory measure. Even with those constraints, working memory differences only help a little bit in predicting who will notice. […]