By Gaurav Jain
Individuals need to make various judgments in their daily lives. A lot of these judgments are quantitative estimates of entities, for instance ‘how much money to save for the post-graduation road trip’. Often, when individuals make these estimates, they are biased towards the first piece of information that comes to their mind. For instance, if I ask individuals to think of the last two digits of their phone number and then ask them to estimate the number of pages in a book, their responses will be heavily biased to the last two digits of their phone number. This first piece of information is known as the ‘anchor’. Anchoring and adjustment, a ubiquitous heuristic process in judgment and decision making, has been vastly demonstrated in the numerical domain. My colleagues and I, with the help of four studies, demonstrate the anchoring and adjustment bias in perceptual domains. Our results show that anchoring and adjustment can bias our judgments at relatively low levels of cognition. Additionally, we outline a process by which anchoring and adjustment biases individuals’ judgments in perceptual domains. This process account explains the extant data in numerical domains as well – thus, providing a way for a potential resolution to the disagreement among different existing process accounts for the anchoring phenomenon.
Sounds
In study 1, we created a tone scale to demonstrate anchoring in the domain of sounds. The tonal scale was presented on the computer to all participants, consisting of 21 unlabeled buttons, each of which produced a pure sinusoidal tone (i.e., no timbre) of increasing frequency, in steps of 100 Hz. Participants were allowed to familiarize themselves with the scale by pressing any of the buttons to generate the associated sound as many times as they wanted. Subsequently, they heard a reference tone of 1600 Hz. They were then shown the tone scale but with one of the buttons enlarged and were instructed to press the enlarged button, which served as the anchor with a frequency either above the reference tone or below it (depending on the experimental condition), and judge whether it matched the reference tone they had just heard. After this judgment, they were instructed to judge which button produced a tone that matched the reference tone they had heard earlier. At this point, all the buttons worked, including the enlarged button, and they could press any button any number of times. The participants provided with a lower frequency anchor picked a significantly lower tone as the final response compared to participants provided with a higher frequency anchor. Thus, the anchoring and adjustment phenomenon extended to the auditory domain.
Haptics
Study 2 utilized the sense of touch as the stimulus and response domain; we investigated anchoring and adjustment mechanisms in the domain of textures. Sandpapers of varying degrees of coarseness were utilized. Participants were shown a folder, inside which was a piece of sandpaper. Participants felt the sandpaper with their eyes closed. Subsequently, they proceeded to another room where sixteen pieces of sandpaper of various levels of coarseness were arranged in decreasing order of coarseness on a poster board. One of the sixteen sandpapers had two pointers, one red and one blue, above it and served as the anchor. Participants were asked to judge whether this sandpaper with the pointers above it was of the same coarseness as the one they had felt earlier in the previous room. If so, they could leave the blue pointer where it was; else, they were instructed to move the blue pointer to the sandpaper they judged to be the same as they had felt earlier. Participants were told that they could touch all the sandpapers as many times as they wanted in the process of making their judgment. A camera was used to record their decision processes and, in particular, the instances of touching the various grades of sandpaper placed on the board. The sixth sandpaper sample in the set was actually identical to the one they had felt earlier in the previous room. However, the red pointer which served as the anchor was placed over either the second sandpaper (which was coarser) or over the tenth sandpaper (which was finer). Results showed that participants who started with the coarser anchor judged a coarser sandpaper to be similar to the target sandpaper they had touched earlier, compared to participants who started with the finer.
Figure: Sandpaper boards used in Study 2. Please note the position of pointers (to mark the anchor sandpaper) in the two conditions.
Shades
Study 3 utilized shades of grey as the stimuli. This study afforded us an added advantage apart from being another domain that is not dependent on the number line – by allowing us to utilize eye-tracking as a process measure of the search process. While participants were performing the task, an eye tracker was utilized to track their eye fixations. An ordered array of twenty-six shades of grey from near-black to near-white was created on the computer. Participants were shown a shade of grey that served as an anchor (which was either ‘almost black’, ‘dark grey’, ‘light grey’ or ‘near white’, depending on the anchor condition participants were in) and were asked if that was the shade of the moon. They were then shown all the shades and were asked to pick the shade that, in their opinion, was the same as that of the moon. Results showed that the anchoring phenomenon was replicated in this novel domain as participants’ final ratings responses were significantly impacted by the anchor.
Underlying Process
Studies 1, 2 and, 3 allow us to draw the conclusion that anchoring extends to non-numeric domains as well. Also, it should be noted that judgments of sound, shades, and granularity of sandpaper are more basic perceptions – thus, the data suggest that the anchoring and adjustment bias does not depend on a numeric representation of the stimulus or response scale. Moreover, due to the use of non-numeric modalities, and unlike a response on a number line, we could surreptitiously observe the process individuals follow to get to the final response, which allowed us to decipher the processes underlying perceptual anchoring. Also, we could make important inferences from our data that shed light on the mechanisms underlying numerical anchoring, which has been a topic of conflict in the literature among various existing process accounts that explain numerical anchoring, including the three most common: insufficient adjustment selective accessibility and scale distortion theory.
In Tversky and Kahneman’s original demonstration of this phenomenon the anchor was seen to bias subsequent judgments, prompting the conclusion that adjustments away from the anchor were insufficient. Epley and Gilovich support this view and argue for the role of adjustment in the anchoring phenomenon. However, Mussweiler and Strack advanced an account that would appear to subsume the anchoring and adjustment phenomenon under an accessibility account. Specifically, they argue that individuals engage in confirmatory testing for the anchor, which makes information consistent with the anchor more available in the later judgment, resulting in the subsequent judgments assimilating towards the anchor. In other words, adjustment is not a necessary step in this account. Rather, an anchor biases the construction of the eventual judgment by making some information more accessible than others. Epley and Gilovich have argued that this is inconsistent with the influence of implausible, self-generated, or patently random anchors and argue for the continued role of insufficient adjustment. A third account, proposed by Frederick and Mochon, often referred to as the ‘scale distortion theory’ suggests that the ‘true answer’ looks extreme relative to the anchor that was just considered, prompting an adjustment towards the anchor. In other words, it proposes a contrast effect as driving the adjustment.
The extant data cannot resolve this existing conflict as past studies did not ‘observe’ how individuals were getting to the response after being biased by the anchor. Due to the use of perceptual domains, our studies could put light on the anchoring process. Our results indicated a search process dominated by adjustments to adjacent-possible responses, implying a search process constrained by selective accessibility. In other words, as suggested by the adjustment proponents, there is indeed adjustment and not construction, but the adjustment itself is constrained by positive hypothesis testing as suggested by the construction proponents. Thus, the process account proposed here paves the way to integrate the previous conflicting accounts. At the very least, the data reported here provide the most precise description yet of movement from an anchor to an answer and should thus assist to develop sharper theoretical formulations.
Even though we have used non-numeric modalities, the underlying mechanism that we propose may hold true for numerical anchoring as individuals use similar cortical metrics and cognitive hardware to assess magnitudes of different domains and humans share similar cortical metrics for different domains. Thus, past research does suggest that individuals access magnitude in similar ways independent of the domain of the magnitude. But it was still warranted to find direct evidence for our process account in the numerical domain for us to make such a claim. The last study, study 4, examined this possibility. Participants, after being given an initial cover story, were shown a screen with a question and numerical grid (going from ‘100’ to ‘490’ in steps of ten). The question asked the participants in the lower (higher) anchor condition: “How likely do you think it is that the parking lot in the Jorpati Aquarium was designed to have 160 (or ‘430’) parking spaces?”. On the next screen, participants were again shown the grid and were asked: “Exactly how many parking spaces, do you think, the parking lot at the Jorpati Aquarium was designed to have?”. Participants’ response as well as their gaze was recorded. As expected, the anchoring phenomenon was replicated. But, more importantly, the results of study 4 provided evidence that the underlying mechanism proposed by us for perceptual anchoring may hold true for numerical anchoring as well. Although it is a small step in the pursuit of ‘looking’ at the process by which individuals reach the final response from the anchor, it is among the first studies in our knowledge that utilizes eye-tracking in the domain of numerical anchoring.
Implications
The findings have practical implications in the consumption domain. Although numerical anchoring is widely applied in marketing, we do not see perceptual anchoring being used. Individuals, while making a decision in the marketplace, consider multiple features of available options where many of the features are non-numerical modes of perception (e.g., consumers may contemplate whether the crockery is ‘heavy’ enough; is the laptop ‘light weight’; is the curry too ‘salty’; is the coffee ‘warm’ enough). This work provides a systematic framework that can be utilized by marketers to anchor consumers to a particular taste, color, weight, and other perceptual factors.
Gaurav Jain