Read or observe like a scientist and you see the post title Truth: Bad is stronger than Good. People have a built in find and flee function for all things Bad and overweigh any bad news more than any good news. Now, you expect and find that the Bad also leads to Badder beliefs, attitudes, and behaviors, but today we find a variation on the Bad Is Stronger theme and present circumstances by which Bad makes for more Good whether belief, attitude, or behavior. And, this isn’t a simple trick where you zag Good against a Bad zig, but rather a contradiction to the headline. More Bad information makes More Good persuasion. I don’t play murder mystery and name the killer before the crime: Dial down the dimmer switch to Low WATT, then provide a little Bad with a lot of Good.
Ein-Gar, Shiv, and Tormala in a JCR 2011 paper demonstrate in four experiments that a little Bad plus a lot of Good is actually Better than just Good alone, but with that WATTage twist. These experiments randomly dial the dimmer switch either through manipulation (1, 2, and 4) or with a personality trait (3). Processors then are exposed through random assignment to a product that contains either only positive information or the same positive information plus one piece of small negative information. Ein-Gar et al. provide a variety of products (hiking boots, chocolate bars, and champagne glasses – Cheers, my Dear; Toujours L’Amour, Tonight For Sure!) So, we’ve got a range of products, several different dimmer switches, and varying content for the Bad and the Good news about each product. We also have experimental designs (although the personality trait is not controlled, but its inclusion provides nice external validity for the WATTage variable – science is always a dynamic tension that requires thoughtful trade offs). And just to put a little murder and suspicion into this post, we’ll add Order Effects to this howdoneit. I’ll detail just Study 1 and only brief the others. Consider the basics.
Participants and Design. One hundred forty-one participants volunteered to take an online survey and received $5 for their time (M age = 26). Participants were randomly assigned to conditions in a 2 (processing effort: high or low) × 2 (informational content: positive only or positive + weak negative) between participants factorial design.
Let’s manipulate WATTage now.
In the low capacity condition (low effort), participants were told that for purposes of experimental control we would like them to do their best to not shift their eyes from the computer screen. Instructions indicated that each time they accidentally looked away from the screen they should look back immediately, and at the end of the task they would be asked to report the number of times they had looked away. Tracking their own side-glances was expected to reduce participants’ processing capacity in this condition. In the high capacity condition (high effort), we simply instructed participants to process the information as if it were online.
This is a cognitive load task that requires more resource from people who track their glance behavior. It sounds trivial, but prior research demonstrates something this simple is sufficient to disable higher WATTage. Now, the product and the Bad and Good news.
In the positive only condition, participants read about a new pair of hiking boots that had a designer orthopedic sole to protect feet, that came in many colors, that was waterproof, that had a five-year warranty, and that included two spare shoelaces. In the positive + weak negative condition, participants received the exact same information in the same order, with the exception that “comes in many colors” was replaced with “comes in only two colors.”
Note that the Bad news is not Terrible, Awful, Dreadful, and on and on, but is only mildly Bad. It is a weak, but not WEAK, Argument.
The ideal result from this theorizing would be no main effect for WATTage or Good versus Good+Bad but only an interaction between the two. You know you are living right with the Lord and the gods of research when you fire up your favorite stat program, run the ANOVA, and read . . .
. . . there were no main effects (Fs < 1), but we did find a significant interaction (F(1,137) = 8.64, p < .01). Under low effort conditions, participants were more interested in purchasing the boots when they were presented with a weak negative attribute (M = 4.26) than when they were presented with only positive attributes (M = 3.29; t(1, 137) = 2.14, p < .04). Conversely, under high effort conditions, participants were more interested in purchasing the boots when they were presented with only positive attributes (M = 4.24) than when they were presented with a weak negative attribute (M = 3.36; t(1, 137) = 2.02, p < 0.05).
Those specific t-tests that make comparisons under the two different WATT states produce Small plus effects with d’s around .40 and a Windowpane of 40/60. Not quite obvious and practical, like a Medium effect size, but given the experimental design, I feel more confident about the effect than if this was an observational design with a d of .40. Here’s a graphic display that nicely demonstrates the beautiful crossover interaction. Amen.
Of course, we obtain this miraculous interaction only with this congenial combination of variables and conditions. So, being hardheaded persuasion detectives, we find more victims and see if we get the same killer. In Study 2 . . .
To vary processing effort, we approached students either immediately prior to taking an exam (low effort condition) or while they were simply walking outside on campus (i.e., high effort condition).
And in Study 3 . . .
Finally, after a brief delay and filler task, participants completed an abbreviated version the analytic-holistic scale (Choi et al. 2007; see Appendix). This scale measures the general tendency to think in a holistic or analytic manner. Participants responded to each item on a scale ranging from 1 (does not describe me at all) to 5 (describes me very much; alpha = .69).
Here are the items.
1. The whole, rather than its parts, should be considered in order to understand a phenomenon.
2. It is more important to pay attention to the whole than its parts.
3. The whole is greater than the sum of its parts.
4. It is more important to pay attention to the whole context rather than the details.
5. It is not possible to understand the parts without considering the whole picture.
6. We should consider the situation a person is faced with, as well as his/her personality.
Through these first 3 experiments, the basic effect replicates: An interaction of WATTage and a lot of Good plus a little Bad to produce more favorable persuasion. Note the different dimmer switches: Glance counting (cognitive load), test pressure (motivation and cognitive load), and personality trait. And, the different products and attributes: Boots, chocolate, or champagne glasses. Different conditions, but always the same crossover interaction. Now, what is driving this Low WATT effect? Let’s consider Study 4 for some evidence on that.
Study 4 uses the same WATTage manipulation as Study 1 (count your glances!) and hits the trail with hiking boots then . . .
. . . All participants received a minor negative feature in this study, which was presented graphically. Specifically, the picture that accompanied the description showed the boots resting aside a shoe-box with minor damage. In the positive + weak negative condition, participants first read the positive description of the hiking boots’ features and then saw the picture with the damaged box. In the weak negative + positive condition, participants first saw the picture with the damaged box and then received the positive description of the boots’ features.
Notice here, we’ve got a test of Order Effects or is first better than second or is second better than first? Ein-Gar et al. argue that Order Effects may explain why a Lotta Good plus a Little Bad works. Typically, primacy (early) trumps recency (late), especially with Low WATT processors who simple grab the first train that pulls into the station and ignore following information. What happens here?
Under low effort conditions, participants evaluated the boots more favorably in the positive + weak negative condition (M = 2.17) than in the weak negative + positive condition (M = 1.45; t(1, 132) = 2.96, p < .01). Under high effort conditions, however, participants rated the boots equivalently under positive + weak negative (M = 1.87) and weak negative + positive (M = 1.88) conditions (t(1, 132) = .04, p = ns).
This is again a Small Plus effect, a Windowpane of 40/60. And, it is entirely consistent with prior research on primacy effects and Low WATT processors. Here’s a graphic.
If you’re thinking, who cares about Order? You see that High WATT effect on the right side of the panel as both Orders (the white and black bars) lead to the same favorable scores. When you’re not High WATT (the left side of the panel), you fall for the first good line you hear. And, alternatively, if you hear a Bad whisper, it kills all the Good shouts that follow. Clearly, you see the Low WATT effect here. They just grab the first Cue available and move along to the next scene.
Let’s summarize this.
When people are Low WATT, a lot of Good with a little Bad produces more persuasion. When they are High WATT, a little Bad reduces the persuasion, as human nature demands. Bad is a lot worse than Good and you’d better look out, but that means High WATT. What else is interesting here is the impact of Order. Low WATT processors will jump all over that little Bad if it comes first and clearly miss all the following Good. If Bad is that Bad, then why don’t Low WATT people react with avoidance when the Bad is late in the list? They’ve made their choice and stop even reading, much less thinking.
Pull back now and consider this evidence. Experiments with random assignment, careful control, great comparison, and proper counting. Variations in how WATTage and Good+Bad are operationalized. Variations in the products. Variations in the setting of the research from lab to field. Also, variations in the age and background of the participants that I didn’t report. Lots of good quality for both internal and external validity.
And, since this is linked into classic ELM-ish theorizing, you can connect this research into a huge literature of dual process effects.
Do you see the good science here both in this specific paper and in the larger body of work? And this is just persuasion, not molecular genetics or evolutionary biology or even Sheldon Cooper’s physics. Yet is this not science?
Danit Ein-Gar, Baba Shiv and Zakary L. Tormala (2011). When Blemishing Leads to Blossoming: The Positive Effect of Negative Information. Journal of Consumer Research.
Stable URL: http://www.jstor.org/stable/10.1086/660807
P.S. For a funny take on the perils of primacy for Low WATT processors, read this transcript from a Richard Pryor sketch on the original Saturday Night Live (1976, the earlier, funnier period for SNL). Imagine a soldier (Pryor) getting orders from a general (Dan Akroyd) about a dangerous mission. The general points the soldier to a variety of equipment including one, saying, “Take that pill.” Sorry, no video on the WWW.
P.P.S. That SNL episode also featured Gil Scott-Heron as the musical guest where he performed “Johannesburg”. Say everybody now what’s the word, Johannesburg, now that’s the word.
And you think the world doesn’t change?