Anchors, scales and the relative coding of value in the brain

https://doi.org/10.1016/j.conb.2008.07.010Get rights and content

People are alarmingly susceptible to manipulations that change both their expectations and experience of the value of goods. Recent studies in behavioral economics suggest such variability reflects more than mere caprice. People commonly judge options and prices in relative terms, rather than absolutely, and display strong sensitivity to exemplar and price anchors. We propose that these findings elucidate important principles about reward processing in the brain. In particular, relative valuation may be a natural consequence of adaptive coding of neuronal firing to optimise sensitivity across large ranges of value. Furthermore, the initial apparent arbitrariness of value may reflect the brains’ attempts to optimally integrate diverse sources of value-relevant information in the face of perceived uncertainty. Recent findings in neuroscience support both accounts, and implicate regions in the orbitofrontal cortex, striatum, and ventromedial prefrontal cortex in the construction of value.

Introduction

Can a restaurant seduce you into buying an expensive wine by offering similar but more expensive alternatives? Does the wine taste better if you pay more for it? Does the taste improve when tasted alongside lower-quality alternatives? That the answer to such questions is often ‘yes’ suggests that human value systems are unlikely to be the impartial critic that economists have traditionally assumed [1], and poses important questions for how the brain constructs value from the information available to it.

Phenomena such as this are increasingly sparking interest among behavioral economists, and these and related effects have begun to make their way to the lab (Figure 1). Indeed, many imaginative experiments have illustrated just how easily primes, anchors and contexts can bias preferences and prices. For instance, students’ interest in hearing their professor recite extracts from Walt Whitman's acclaimed ‘Leaves of Grass’ anthology, for free, was found to be strongly influenced by whether they had previously be asked to pay, or receive payment, to attend [2] (Figure 1). The initial question, although subsequently irrelevant, appeared to prime people towards thinking that the recital was either a privilege or a sufferance. But although the initial valuation was malleable, subsequent behavior was far less so: once a student has come to doubt their professor's poetic prowess, they recognised that 5 min of it would be much less painful than 10 [2]. Many of these experiments display what has been termed ‘coherent arbitrariness’ [3], which describes the apparent consistency of behavior that occurs once otherwise arbitrary baseline values have been set.

This leads to two related accounts of how humans generate estimates of the value of goods in transactions. The first is largely algorithmic, and posits that humans lack stable, long-term representation of the magnitude of value, and judgments are made purely by pair-wise comparisons in an ordinal dimension. This can be formalized by relative judgment models [4, 5] and related theories (e.g. the stochastic difference model, multi-alternative decision field theory, adaptation level theory, and range frequency theory [6, 7, 8, 9]), and draws support primarily from psychophysical observations. Humans are inherently bad at making absolute judgments about the intensity of various sensory stimuli (such as the loudness of a tone), despite being very good at discrimination (e.g. [10]). Applying the relative judgment model to value, would suggest that initial experience with goods and prices generate the anchors against which subsequent experience is judged. For example, if we are led to believe a new wine from Greenland typically retails $50 a bottle, we will jump at the chance to buy a bottle for $30, simply because in relative terms it must be a good price. When extended to choice, decisions are determined by the rank position of an item in a sample rather than absolute value [11••]. That is, decisions then depend on the sample people are cued to retrieve so that if one is cued to think of wines in general, the $30 Greenland wine seems expensive and may be rejected; if cued to think of Greenland wines, it seems cheap and is eagerly chosen [11••].

The second account is computational, and posits that value scales are intact, but that the sensory information from an available option is often inherently uncertain, forcing people have to make inferences (e.g. Bayesian) from all the information presented. Informative and circumstantial cues are thereby exploited for any clues they might harbor regarding the true underlying worth of an option. This view is closely related to theories of perception [12, 13], and is well illustrated in vision. For instance, if you catch a glimpse a small round object on the ground in the middle of an orchard, you are quite likely to perceive it as an apple. If you spot the same object on a tennis court, you are more likely to perceive it as a ball. In each case, the prior distribution of beliefs strongly biases the subsequent perception. A similar mechanism might well exist for value. This would predict, for example, that wine actually tastes better when said to be the ‘best’, or ‘expensive’, because there is a strong prior belief that such characteristics offer reasonable evidence indicating quality.

Recent neuroscience research on judgment and decision-making in humans and primates has the capacity to provide evidence of the implementation of these models, and as we show below, evidence exists for both accounts.

Section snippets

Relative coding of value

The orbitofrontal cortex has a well-studied role in reward processing, and neuronal activity correlates well with the motivational value of a reward, over-and-above its sensory properties [14]. For example, activity declines for a reward (or cues that predict a reward) when an individual (human or monkey) is sated with that reward [15, 16], just as it does subjectively. Initial evidence for relative coding came from a classic experiment by Tremblay and Schultz [17], who presented monkeys with

Adaptive scaling

Recording how much better an outcome is in the context of others is clearly useful, and indeed a fully coded version of this is analogous to the prediction error, a key learning signal thought to update values as a consequence of trial-and-error experience [21]. There is good evidence that dopamine projections from the midbrain to the striatum carry this signal [22, 23, 24, 25], and striatal activity in humans recorded using fMRI concur with primate data [26, 27, 28, 29].

But theories of

Expectation, inference and placebo effects on value

In relative judgment models, contexts may provide anchors to establish scales in determining the relative positions of an option. However, in expectation and ‘perceptual’ models, they actually provide information that influences the experience of it. Expectation effects are well studied in behavioral, psychophysical and economic studies, in both the appetitive and aversive domain. Studies on the latter, which are slightly more extensive, have shown that placebo effects can be reliably induced

Equating value in transactions

Transactions of any sort involve establishing whether the value of obtaining something compares favorably with the value of losing something else. Since firing rates may not be negative and decreases from baseline firing offer limited resolution, losses and gains may be best encoded by separate populations of neurons. Indeed, this as has been shown in both the orbitofrontal cortex and striatum [39, 40, 41].

It remains largely unknown how the brain integrates and compares gain and loss

Discussion

Neurobiological studies are beginning to provide key insights into why the values people ascribe to goods, and the price they are prepared to pay for them, is often so susceptible to manipulation. First, in given contexts, the brain sets relative scales against which the ordinal position of goods is set. Second, the brain uses available and additional information to help refine judgments of value. Thus, object or price anchors can act in two distinct ways to influence trade decisions. First,

References and recommended reading

Papers of particular interest, published within the period of review, have been highlighted as:

  • • of special interest

  • •• of outstanding interest

Acknowledgements

The authors thank Nick Chater, Peter Dayan, and Ivo Vlaev for valuable comments on the manuscript.

References (50)

  • B. Shiv et al.

    Placebo effects of marketing actions: consumers may get what they pay for

    J Marketing Res

    (2005)
  • R.L. Waber et al.

    Commercial features of placebo and therapeutic efficacy

    JAMA

    (2008)
  • N.D. Daw et al.

    Uncertainty-based competition between prefrontal and dorsolateral striatal systems for behavioral control

    Nat Neurosci

    (2005)
  • D. Ariely

    Predictably Irrational

    (2008)
  • D. Ariely et al.

    “Coherent arbitrariness”: stable demand curves without stable preferences

    Q J Econ

    (2003)
  • D.R.J. Laming

    The relativity of “absolute” judgments

    Br J Math Stat Psychol

    (1984)
  • N. Stewart et al.

    Absolute identification of relative judgment

    Psychol Rev

    (2005)
  • H. Helson

    Adaptation-level Theory

    (1964)
  • A. Parducci

    Category judgment: a range-frequency model

    Psychol Rev

    (1965)
  • R.M. Roe et al.

    Multi-alternative decision field theory: a dynamic connectionist model of decision-making

    Psychol Rev

    (2001)
  • C. González-Vallejo

    Making trade-offs: a probabilistic and context-sensitive model of choice behavior

    Psychol Rev

    (2002)
  • G.A. Miller

    The magical number seven, plus or minus two: some limits on our capacity for processing information

    Psychol Rev

    (1956)
  • N. Stewart et al.

    Decision by sampling

    Cognitive Psychology

    (2006)
  • K.J. Friston

    Learning and inference in the brain

    Neural Netw

    (2003)
  • J.A. Gottfried et al.

    Encoding predictive reward value in human amygdala and orbitofrontal cortex

    Science

    (2003)
  • Cited by (0)

    View full text