Thursday, April 18, 2013

Existential Dread of Absurd Social Psychology Studies

Scene from Rabbits by David Lynch

In a nameless city, deluged by a continuous rain, three rabbits live with a fearful mystery.”

The latest "elegant and breathtaking"1 paper in Psychological Science presents a rather muddled view of film aesthetics, continental philosophy, surrealism, mortality salience, and stigmatizing attitudes towards sex work (Randles et al., 2013). Oh, and how Tylenol® brand acetaminophen can ease the existential dread evoked by all of these modern horrors.

The authors explained the purpose and implications of their study in the APS press release:
According to lead researcher Daniel Randles and colleagues at the University of British Columbia in Canada, the new findings suggest that Tylenol may have more profound psychological effects than previously thought:

“Pain extends beyond tissue damage and hurt feelings, and includes the distress and existential angst we feel when we’re uncertain or have just experienced something surreal. Regardless of the kind of pain, taking Tylenol seems to inhibit the brain signal that says something is wrong.”

Randles and colleagues knew from previous research that when the richness, order, and meaning in life is threatened — with thoughts of death, for instance — people tend to reassert their basic values as a coping mechanism.

The researchers also knew that both physical and social pain — like bumping your head or being ostracized from friends — can be alleviated with acetaminophen. Randles and colleagues speculated that the existentialist suffering we face with thoughts of death might involve similar brain processes. If so, they asked, would it be possible to reduce that suffering with a simple pain medicine?

No!!  I think this is a ridiculous assertion that gets away with using language (and dependent measures) that not only lack precision, but also lack an analogical relation to the real phenomenon under discussion. The leaps of logic were so egregious that I don't know where to begin... let's start with the meaning-maintenance model (MMM) that motivated the work. MMM "posits that any violation of expectations leads to an affective experience that motivates compensatory affirmation" (Randles et al., 2013). Any violation?? So all sorts of psycholinguistics experiments that involve syntactic violations 2 will motivate compensatory affirmation? If that's the case, then David Lynch films will often "motivate compensatory affirmation."

But does a David Lynch film “hurt” you?
...Lynch’s films have the ability to “disturb, offend or mystify” (Rodley, 2005, p. 245). Insofar as it “hurts” to watch some of Lynch’s films, as it arguably hurts whenever one is assaulted by thoughts and experiences that are at odds with one’s expectations and values, the question arises as to how this uncomfortable feeling is represented in the brain.

First, David Lynch is one of my favorite directors, and I have never felt "hurt" by watching one of his films. Second, Randles et al. never, at any point in their experiments, address how Lynch-viewing is represented in the brain.

What did the authors actually do? In brief, they asked ~350 young Vancouverites to participate in one of two experiments. In the first study, 121 subjects wrote about death or about dental pain. In the second study, 228 subjects watched a 4 min clip from Rabbits or from The Simpsons. In each case, half of the participants received acetaminophen, half received placebo. Why? What motivated the choice of acetaminophen, as opposed to aspirin, ibuprofen, or naproxen? This was based on a study by Dewall et al. (2010), another problematic paper3 in Psych Sci. There was no mechanistic reason for the original choice.

Here's the neuro-rationale for the current study (Randles et al., 2013):
The present research is predicated on four key findings in the literature: (a) Both physical and social pain are associated with activation in the dACC [dorsal anterior cingulate cortex]4 (e.g., Eisenberger et al., 2003), (b) the dACC is activated in response to anomalies (e.g., Botvinick et al., 2004), (c) social rejection can produce the same compensatory affirmation as other meaning threats (e.g., Nash et al., 2011), and (d) acetaminophen has been shown to reduce physical and social pain, as well as activation in the dACC (DeWall et al., 2010). These findings led us to predict that acetaminophen may also inhibit compensatory affirmation following meaning threats.

The acetaminophen group in Dewall et al. (dose of 2,000 mg a day for 3 weeks) did show less dACC activity in response to cyberball exclusion, but they did not report lower hurt feelings in that situation. The treatment administered by Randles et al. was quite different: a single acute dose of 1,000 mg Tylenol-brand acetaminophen (Rapid Release formula) or 1,000 mg sugar placebo, given 30 min before the critical manipulation.

In Exp. 1, writing two paragraphs about what will happen to your body after death was designed to trigger mortality salience, or thoughts about the inevitability of death. This in turn would lead to compensatory affirmation of cultural views. How was this measured? By assessing the severity of punitive attitudes towards women who engage in sex work! This is the worst part of the study, in my opinion.
Social judgment survey

Finally, participants read a hypothetical arrest report about a prostitute and were asked to set the amount of the bail (on a scale from $0 to $999). This measure has been used in a number of other meaning-threat studies (Proulx & Heine, 2008; Proulx et al., 2010; Randles et al., 2011; Rosenblatt, Greenberg, Solomon, Pyszczynski, & Lyon, 1989). Participants are expected to increase the bond amount after experiencing a threat, because trading sex for money is both at odds with commonly held cultural views of relationships and against the law. Increasing the bond assessment provides participants n opportunity to affirm their belief that prostitution is wrong.

The study took place in Vancouver, Canada. What are the laws on prostitution?
In Canada, the buying and selling of sexual services are legal, but most surrounding activities, such as public communication for the purpose of prostitution, brothels and procuring are offences under the law.

What are current attitudes towards prostitution in Canada?
The views of Canadians on prostitution vary greatly according to age and gender, with a large proportion of men and older respondents voicing support for some kind of decriminalization, while most women and younger respondents are not as comfortable with the idea...
. . .

As evidenced in surveys conducted by Angus Reid Public Opinion in 2009 and 2010, only about a quarter of Canadians (22%) are aware that exchanging sex for money is legal in Canada, while seven-in-ten (70%) mistakenly believe that the practice is illegal.
. . .

Still, there is no clear consensus on how some of these guidelines are currently applied. While 36 per cent of respondents believe the Criminal Code provisions related to communication and brothels are fair to the purpose of protecting the public good, almost half (47%) think the rules are unfair and force prostitutes into unsafe situations.

Here are my reactions to the Prostitute-Bail dependent measure:

(1) Yay! Let's stigmatize the prostitute, not the johns!

(2) Does the baseline for these bail judgments differ by sex? age? religion? ethnicity? As professional polling can attest, attitudes vary greatly along demographic lines. The participant pool was quite diverse, and we know nothing about age.
We recruited 121 participants (81 women, 40 men). The sample was predominantly of East Asian (45%), European (29%), and South Asian (12%) descent.
(3) Participants were randomly assigned to one of four groups, but we don't know anything about the randomization  - perhaps the most religious and judgmental people ended up in the mortality salience/placebo condition.

(4) To reiterate, we don't know anything about possible demographic differences in the amount of bail set. And that is the only dependent measure!! We don't know how anyone would allocate money or set a price in another situation that is not "morally laden". Let's say you're selling a used car - what would you charge?

At any rate, the authors reported that the mortality-salience/placebo group punished the "norm violator" by a significantly larger amount than the other three groups, t(112) = 2.33, p = .02, d = 0.52.

Fig. 1 (Randles et al., 2013). Results from Study 1: mean bond value set for the prostitute as a function of group (mortality-salience vs. control condition crossed with placebo vs. acetaminophen condition). The scale ranged from $0 to $999. Error bars represent the standard error for each group.

Moving right along to Exp. 2, we discover that the authors decided to use a different dependent measure for no clearly motivated reason. This makes it impossible to compare the outcome of the salience mortality manipulation to the David Lynch manipulation.
We also changed the dependent measure [in Exp. 2]. This study was conducted 3 to 6 months after a well-publicized local riot that followed the Vancouver Canucks’ loss in their bid for the Stanley Cup, and we expected that most students held a negative view of the riot. Thus, we expected that after a threat, participants would affirm this view by calling for stronger punishment for the rioters. Participants were informed that people were debating whether the rioters should be given sentences more lenient than those for comparable individual acts of vandalism, because the rioters had acted impulsively, or should be given stiffer sentences, because they had taken advantage of the city while it was vulnerable. Participants then marked a spot on a line from 0% to 200%. They were told that 0% indicated that rioters should not be fined, that 100% indicated that rioters should receive a normal fine, and that 200% indicated that rioters should receive a doubled fine.

One initial critique is that the Vancouver hockey riot itself provoked MMM. It was a mob event that people could not explain rationally. The subjects were more likely to have been directly affected by this event (in comparison to the hypothetical sex worker bail), by either knowing someone who participated or who was present, or by witnessing the event live or through social media, or by having a favorite business vandalized. In addition, the assigned fines were relative, not absolute. A 150% fine out of... $100 or $1,000 or $10,000?

At any rate, the authors reported that participants in the Lynch/placebo group wanted to punish the rioters by a significantly larger amount than did participants in the other three groups, t(203) = 2.64 p < .01, d = 0.43.

Fig. 2 (Randles et al., 2013). Results from Study 2: mean preference for the penalty to be given individuals convicted of vandalism or theft during the Vancouver hockey riot as a function of group (threat vs. control condition crossed with placebo vs. acetaminophen condition). The rating scale ranged from 0% (no fine for a conviction), through 100% (a normal fine), to 200% (a doubled penalty).

Collectively, the results were taken as evidence that Tylenol can potentially treat chronic anxiety disorders, a conclusion that filled me with existential dread:
The study demonstrates that existentialist dread is not limited to thinking about death, but might generalize to any scenario that is confusing or surprising — such as an unsettling movie.

“We’re still taken aback that we’ve found that a drug used primarily to alleviate headaches can also make people numb to the worry of thinking about their deaths, or to the uneasiness of watching a surrealist film,” says Randles.

The researchers believe that these studies may have implications for clinical interventions down the road.

“For people who suffer from chronic anxiety, or are overly sensitive to uncertainty, this work may shed some light on what is happening and how their symptoms could be reduced,” Randles concludes.

I have a few final questions for the authors, since this violation of my expectations led to an affective experience that motivated my own compensatory affirmation processes:
  • Why wasn't the dose adjusted by weight? A 45 kg woman got the same dose as a 90 kg man. 
  • Was physical pain assessed in the subjects pre/post-treatment? No it was not. 
  • Did anyone have a headache or any other physical pain before treatment? We don't know... which would be important to know, since relieving physical pain will make you less cranky and irritable. 
  • Is there a single neuroimaging study that has administered acetaminophen at the dose and time course used here? No. 
  • What is the evidence that acetaminophen affects the hemodynamic response in the same exact dACC region hypothesized to control physical, existential, and social pain? 
  • Has there been a single fMRI study in which subjects have watched Rabbits and Simpsons, counterbalanced in a single session while their brains were scanned? 
  • What is the Rabbits > Simpsons neural activation pattern? 
  • Why wasn't there a measure that the Lynch clip was actually "disturbing" or that the Simpsons clip was enjoyable? Actually, none of the manipulations induced changes in affective state on the PANAS.

To ease my existential dread, it's time to watch Rabbits in its entirety.


1 Former Psychological Science Editor Robert V. Kial:
At meetings and via email, authors often asked me, “What sort of paper is a good candidate for Psychological Science?” ... And when feeling particularly candid, I might say that the ideal Psychological Science manuscript is difficult to define, but easily recognized — the topic is fundamental to the field, the design is elegant, and the findings are breathtaking.
2 For example: "The metal was for refined by the goldsmith who was honored" (Friederici et al., 1996).

3 For a lengthy exposition on the problematic aspects of the Dewall et al. paper, see Suffering from the pain of social rejection? Feel better with TYLENOL®.

4 At the risk of sounding like a broken record, the dACC has been associated with a wide array of cognitive and emotional control functions (Posner et al., 2007). In the TYLENOL® post, I said:
The "shared neurobiological systems" [for social and physical pain] are thought to be located in the dorsal anterior cingulate cortex (ACC), a brain structure that contains discrete regions responsive to physical pain (Kwan et al., 2000). Interestingly, externally applied vs. self-administered thermal pain activate anatomically distinct areas of the ACC (Mohr et al., 2005). Furthermore, it is not at all clear whether the same regions of ACC represent social pain and the affective components of physical pain. In a study designed to dissociate expectancy violations from social rejection, the dorsal ACC was activated when expectations were violated, while ventral ACC (quite distant from the physical pain regions) was activated by social rejection (Somerville et al., 2006).


Dewall CN, Macdonald G, Webster GD, Masten CL, Baumeister RF, Powell C, Combs D, Schurtz DR, Stillman TF, Tice DM, Eisenberger NI. (2010). Acetaminophen reduces social pain: behavioral and neural evidence. Psychol Sci. 21:931-7.

Randles, D., Heine, S., & Santos, N. (2013). The Common Pain of Surrealism and Death: Acetaminophen Reduces Compensatory Affirmation Following Meaning Threats Psychological Science DOI: 10.1177/0956797612464786

Further Reading on Surrealism, Dread, and Tylenol:

Surrealistic Imaging Experiment #1

Of Mice and Women: Animal Models of Desire, Dread, and Despair

Suffering from the pain of social rejection? Feel better with TYLENOL®

Subscribe to Post Comments [Atom]


At April 19, 2013 1:48 PM, Blogger Neuroskeptic said...

"meaning-maintenance model (MMM)"

My rule of thumb is that if a 'model' can be well-encapsulated in its own name, then it's a description (and a shallow one), not a model.

I call this the Beware Trite Pseudoexplanations Model (BeTriPaMo).

At April 23, 2013 10:51 AM, Anonymous Lew said...

Your comments on the dependent measures can be summed up as "maybe random assignment didn't work". I don't really care for this study either, but those criticisms are weak sauce. Stick to the neurological issues (e.g., they propose a brain-based mechanism but include no physiological measures).

At April 23, 2013 11:34 AM, Anonymous Anonymous said...

Absolutely I agree with Lew.

This post doesn't include a single legitimate methodological critique, other than "well..I didn't like it, so therefore randomization didn't work. Just trust me on that"

At April 23, 2013 2:39 PM, Blogger The Neurocritic said...

Thanks for the comments.

Lew - You said "Stick to the neurological issues (e.g., they propose a brain-based mechanism but include no physiological measures)." You're right, they suggested the effects are based on dACC response but included no neuroimaging.

I expanded on a more fundamental point in my follow-up post, on whether Tylenol Exerts its Analgesic Effects via the Spinal Cord?

"So if it's really true that acetaminophen exerts its pain-relieving effects through synapses in the spinal cord, then what does this say about providing relief from the angst of social exclusion, mortality salience, and existential dread? That it's based on nociceptive spinal cord neurons in laminae I, II, and V?"

As for the "randomization critique" critique shared by you and Anonymous, let's assume that the randomization worked perfectly and there were no discrepancies in baseline attitudes between the groups. I still think the Prostitute Bail Dependent Measure is stigmatizing (and a strange choice for Canadians), but that's irrelevant here. One persistent question is why switch to the Hockey Riot Dependent Measure for the second study? If the authors wanted to compare the effects of mortality salience to puzzling cinema, why not use the same measure? Or is this what is called a conceptual replication in social psychology? I would think the field might be more careful now about replication of "elegant and breathtaking" results in light of the Stapel fiasco (note comments by Nobel Laureate Daniel Kahneman)

And I do think a control measure of money allocation that is not based on moral norms is important here.

Neuroskeptic - Turns out there's another model... the MAID (model of ambivalence-induced discomfort)!

At April 24, 2013 8:15 AM, Anonymous Lew said...

The authors didn't come up with the prostitute measure, it has often used in these types of worldview defense studies. I do agree that it's stigmatizing and wouldn't use it myself. But if the first DV is problematic, wouldn't switching to another for Study 2 be a strength of the paper, not a weakness? ;)

You're correct, this is a conceptual replication. The manipulations are also conceptual replications of one another, with writing about death intended to evoke similar psychological processes as watching Lynch. And you're right that there's some debate about whether conceptual replications are strengths or weaknesses. Though here the DV's are actually pretty similar--they both involve punishing people. Wouldn't a brain paper that used ERP's in one study and fMRI in another be seen as stronger because of the converging evidence from multiple methods? Here's a longer explanation of conceptual replications.

I take some offense to your Stapel comment because conceptual replications had nothing to do with Stapel's fraud. Stapel made up data. If more conventional replications were the norm, he would have made up those too. (A lot of his "findings" actually "replicated" just fine because his predictions happened to be good ones!)

But in the end, I think where you and I can agree is that in the current paper they definitely haven't linked these effects to brain processes. I also think it's risky to wade into the physiological pool when that's usually not our training--it carries the potential for rookie mistakes.

At April 24, 2013 10:05 AM, Blogger The Neurocritic said...

Lew - Thanks again for your comment. I'm not a social psychologist, that must be clear. We cognitive types are more inclined to do direct replications of tasks and/or manipulations. If I did ERPs in Exp. 1 and fMRI in Exp. 2, the whole point would be to use the exact same experimental parameters in both cases.

My question then is why not use more than one dependent measure in the mortality / Lynch manipulations? Why not use prostitute bail (or preferably something better) AND hockey riot? Plus a third control task, e.g., "set the price to sell a used car with these x y z problems."

I brought up the Stapel investigation because of the resulting controversy, not only from the fraud itself but also from questions regarding standard scientific practices in social psychology. "Nobel laureate challenges psychologists to clean up their act" (and this includes replications).

PsychFileDrawer.Org is one place that replication attempts are being documented.

I do agree that the manuscript sections on proposed neural underpinnings and mechanisms of action are rather weak and speculative. It took me about an hour to look up and write about Tylenol's proposed mechanisms of action, and that was from a paper cited by Randles et al. themselves.

At April 25, 2013 8:13 AM, Anonymous lew said...

Here's one I saw today that I'm sure will blow your mind: Clenching Right Fist May Give Better Grip On Memory

"My question then is why not use more than one dependent measure in the mortality / Lynch manipulations? Why not use prostitute bail (or preferably something better) AND hockey riot? Plus a third control task, e.g., "set the price to sell a used car with these x y z problems."

I think those are great suggestions actually.

At April 27, 2013 5:29 AM, Blogger B. Fiend said...

I'm with some of the rest of the commentators: if randomization worked (and the p-values are there precisely for this purpose), then something happened, and it happened fairly similarly between the two studies. The implicit criticisms suggesting that there was no actual treatment effect are pretty weak in that light. In addition, I'm also of the opinion that the second DV being technically different from, but conceptually similar to the first can be seen as adding stronger evidence to the study's assertions than if it had been identical.

Much of your criticism seems to suggest that, if a researcher doesn't have a clear, plausible explanation for all findings at the physical neurological level of analysis, the findings should be ignored. I strongly oppose this position; we have learned a great many very useful things about humans with research that cannot, at the time, be clearly tied to well-understood neurological processes.

Of course, the criticisms of stigmatizing prostitutes, the particular choice of events or stimuli used, and--most importantly--the theoretical model providing the conceptual framework are quite relevant. What I would really like to see would be a strong test of this theoretical framework by pitting it directly against a different framework that had different, but also plausible, predictions.

At April 28, 2013 5:24 PM, Blogger The Neurocritic said...

B. Fiend - The problem with not having a clear neurophysiological explanation for the findings stems from the authors' own rationale, which relies on differential activation in the dorsal ACC.

I agree that I would like to see a stronger test of their hypothesis. One way to do this would be as a within-subject crossover design, like many typical drug studies. For this to work, the authors would have to expand their dependent measures beyond one per experiment, and come up with other surrealist film clips. Neither of these would seem that hard to do. Come up with a series of 10 moralistic & norm violating questions and 10 control questions. In separate placebo and drug sessions, watch Lynch or Man Ray or perhaps Buñuel and Dalí. Answer 5 moralistic and 5 non-moralistic questions after each.

At April 28, 2013 5:57 PM, Blogger The Neurocritic said...

Lew - Thanks for the link on the "Clenching Right Fist May Give Better Grip On Memory" study. It's caused quite a stir in the comments section of PLOS ONE.

I ended up doing a post on the study myself: Want to remember something? Clenching your fist doesn't help!


Post a Comment

<< Home

eXTReMe Tracker