|Navindra Persaud (2009), Scholarpedia, 4(1):7428.||doi:10.4249/scholarpedia.7428||revision #69234 [link to/cite this article]|
Post-decision wagering is a method for measuring consciousness or awareness in psychology and neuroscience experiments. A person makes a decision (e.g. a visual discrimination) and then places a wager. If the decision is correct (e.g. the visual discrimination is accurate) she earns the amount of money wagered while if the decision is incorrect (e.g. the visual discrimination is inaccurate) she loses the amount of money wagered. Awareness can be assessed by assuming that the person will place a high wager if she is aware that her decision (e.g. visual discrimination) is correct.
Rationale and context
The motivation problem
- The easiest way to determine whether knowledge is conscious is to ask people to freely report their knowledge. Don Dulany and David Shanks and others have pointed out that free report often under-estimates conscious knowledge. Participants may be unsure of their strategy, too embarrassed to make a declaration that may be false, may momentarily forget, or they may not be able to properly articulate their strategy.
- Alternatively, awareness can be measured by simply asking people whether they have awareness, or by having them rate their confidence in the accuracy of a decision at the time they make it, side stepping the above problems. As described by Zoltan Dienes and colleagues, a lack of correlation between decision accuracy and confidence rating can be taken as evidence of unconscious knowledge; similarly, above baseline performance when people say they are literally guessing also can be taken as indicating unconscious knowledge (guessing criterion).
- Nonetheless, people may not be motivated to articulate precisely how confident they actually feel. Post-decision wagering addresses the problem of motivation by offering people financial incentives to report their awareness. That is, tendencies to under report awareness are arguably counteracted by cash incentives.
The miscommunication problem
- Post-decision wagering addresses difficulties caused by how people interpret the questions asked by experimenters and how experimenters phrase questions. Different people may interpret the same question (e.g. “Did you see the pattern on the screen?”) differently. This is especially true when the questions asked by the experimenter involve terms such as “awareness” or “consciousness” since these terms are difficult to define. Similarly, a confidence rating of 8 out of 10 or a high confidence rating may also mean different things to different people. Post-decision wagering avoids these problems by avoiding terms liable to misinterpretation such as “confidence”, “awareness”, or “consciousness”. In subjective measures, the problem should be addressed by defining terms as precisely as is needed. But in post-decision wagering participants only have to understand how to place a wager.
- Several classic experiments show that people make different declarations of strategy awareness depending on how questions are asked. In 1911, Édouard Claparède described an amnesic patient who learned to withdraw her hand whenever she saw him after he pricked her with a pin hidden in his hand, despite denying any recollection of being pricked. However, after persistent questioning about why she withdrew her hand she finally responded, “Well, you never know who might have a pin in his hand,” suggesting some awareness of the event previously described as “forgotten”. More recently it has been shown that the quality of explanations depends on the questioning style; with multiple choice questions yielding higher estimates of awareness than open-ended questions. Post-decision wagering addresses this problem by avoiding questions altogether; people merely place wagers. Nonetheless, differing ways of wagering or gambling may also produce different answers, highlighting the importance of empirically confirming the usefulness of different methods of measuring consciousness.
Vision and blindsight
- Post-decision wagering has been used to study blindsight, the ability to make accurate visual discriminations in the absence of visual awareness. A well studied person with blindsight, GY, made visual discriminations and then placed wagers of either 1 or 2. GY did not consistently place high wagers when he made accurate visual discriminations in his blind field.
- Post-decision wagering has also been used to study visual discriminations without awareness in healthy people. Craig Kunimoto, Jeff Miller and Harold Pashler had healthy people make visual discriminations and then place small wagers. They concluded that their results were consistent with performance without awareness. However, the specific methodology employed has been criticized.
Post-decision wagering has also been to study implicit learning or learning without strategy awareness in the artificial grammar task. Participants were shown a series of strings of letters (e.g. XMTVR) and asked to remember as much about them as possible. They were then told that each of the letter strings they were shown obeyed a complicated set of rules and that they were going to be shown new strings that either obeyed those or not. Participants then decided whether each new string obeyed those rules and placed a wager of either 1 or 2 on each decision. Participants performed significantly better than chance but yet failed to consistently wager high.
The intuitiveness of post-decision wagering lends it to use in children. Ted Ruffman and colleagues had three year old children predict where story characters would go and then wager counters which were exchanged for stickers at the end of the experiment. The children tended to wager high after making incorrect verbal predictions, while eye gaze made the right predictions, consistent with the control of eye gaze being made without awareness.
Iowa gambling task
Post-choice wagering was also used in an attempt to resolve an apparent contradiction in findings from the Iowa gambling task. In this task, participants select cards from four decks of cards for pretend cash rewards and penalties. As they select cards they learn which decks are better and which are worse. Antoine Bechara, Hanna Damasio, Daniel Tranel, and Antonio Damasio claimed to demonstrate participants “deciding advantageously before knowing the advantageous strategy” in the Iowa gambling task having used open-ended questions to establish a lack of strategy awareness. In contrast, Tiago Maia and James McClelland asked participants in the same task more quantitative questions and concluded that they were aware of the strategy they were using when making advantageous choices. Experiments employing post-decision wagering suggest that the conflicting findings of Bechara et al. and Maia & McClelland is a consequence of differential influences of the different questioning styles on strategy awareness - questions can change the internal state that they are supposed to be measuring.
- As post-decision wagering employs cash incentives, interpretations of the results involves making assumptions about how participants value money. People who value the amount be wagered less may be less motivated to reveal their awareness. This problem can be overcome by intra-participant control experiments.
- Post-decision wagering can be costly for experimenters.
- People who object to wagering on moral or religious grounds may choose not to participate in experiments employing post-decision wagering. Further, ethics committees may not approve research involving participants being paid large sums of money.
Anil Seth argued that post decision wagering “cannot supply a ‘direct measure of awareness’. Because conscious content is ontologically subjective, it is a simple fact of the matter that no such direct behavioral measures exist.” He later explains that “the simple reason why [post-decision wagering] cannot supply a direct measure of consciousness is that no behavioral measure, subjective or objective, is up to the job.” He stresses the importance of “explicit theoretical frameworks” to complement methods such as post-decision wagering.
Signal detection theory
- Colin Clifford and colleagues questioned the reasoning behind taking GY’s failure to maximize his earning after making correct visual discriminations in his blind field as evidence of a lack of awareness: “If that conclusion was logically sound, we reasoned, it should be impossible for the reported pattern of results to have arisen from a situation in which GY based the magnitude of his post-decision wager on knowledge of the same sensory evidence upon which the Yes–No decision had been based.” They go on to show that GY’s behaviour on the task could be mimicked by a signal detection theory model in which visual discriminations and wagers are made using the same sensory information.
- Post-decision wagering does not make any assumptions about what information is used to make the wagers. Specifically, it does not assume that different information is used to make visual discriminations and wagers. If GY used the same information to make his visual discriminations and his wagers, he apparently used the information differently: he made the profitable decision in the former but not the latter case. It is difficult to explain this without positing a lack of awareness.
If there is a 50 % chance of being correct on the task decision (e.g. the visual discrimination), the expected outcome of a guess on the task decision is zero regardless of the wager. The expected gain thus does not explain why a participant would wager low if she believes that she is guessing. Aaron Schurger and Shlomi Sher suggested that participants may tend to wager low because of loss aversion, valuing the avoidance of a loss of a certain amount of money more than the gain of that same amount. They suggest that participants should choose between having the financial outcome of each trial determined by the accuracy of their task decision (e.g. whether the visual discrimination is correct) or by chance (e.g. the flip of a coin) instead of choosing between a low and high wager. Dienes and Seth measured risk aversion as an individual difference and showed it correlated with wagering but not with verbal confidence in an artificial grammar learning task. Further, following methods in the subjective probability literature, they introduced a new "no-loss" gambling method in which people could choose to bet either on the correctness of their grammaticality decision or on a random process. No-loss gambling was not found to be sensitive to risk aversion. When people bet on a random process, their grammaticality decisions were above chance, indicating unconscious knowledge by the guessing criterion.
- Broadbent, D. E., Fitzgerald, P., & Broadbent, M. H. P. (1986). Implicit and explicit knowledge in the control of complex systems. British Journal of Psychology, 77, 33-50.
- Claparède, É. (1911). Recognition et moite. Arch Psychol Geneva, 11, 79-90.
- Clifford, C., Arabzadeh, E. and Harris, J. (2007). Getting technical about awareness. Trends in Cognitive Sciences, 12(2), 54-58.
- Dienes, Z., Altmann, G., Kwan, L., & Goode, A. (1995). Unconscious knowledge of artificial grammars is applied strategically. Journal of Experimental Psychology: Learning, Memory, & Cognition, 21, 1322–1338.
Dienes, Z., & Seth, A. (submitted). Betting on the unconscious.
- Evans, S., & Azzopardi, P. (2007) Evaluation of a 'bias-free' measure of awareness. Spat Vis, 20(1-2), 61-77.
- Kunimoto, C., Miller, J., Pashler, H. (2001).Confidence and Accuracy of Near-Threshold Discrimination Responses. Consciousness and Cognition, 10, 294–340
- Ruffman, T., Garnham, W., Import, A., Connolly, D. (2001) Does eye gaze indicate implicit knowledge of false belief? Charting transitions in knowledge. Journal of experimental child psychology, 80(3), 201-224.
- Schurger, A., & Sher, S. (2008). Awareness, loss aversion, and post-decision wagering. Trends in Cognitive Sciences, 13, 682-684.
- Seth, A. K. (2007). Post-decision wagering measures metacognitive content. Consciousness and Cognition, doi:10.1016/j.concog.2007.05.008
- Shanks, D. R. & St John, M. F. (1994) Characteristics of dissociable human learning systems. Behav. Brain Sci., 17, 367–447.
- Tunney, R. J., & Shanks, D. R. (2003). Subjective measures of awareness and implicit cognition. Mem Cognit., 31, 7, 1060-1071.
- Persaud, N., McLeod, P., Cowey, A. (2007). Post-decision wagering objectively measures awareness. Nature Neuroscience, 10, 257-261.
- Koch, C., & Preuschoff, K. (2007). Betting the house on consciousness. Nature Neuroscience, 10, 140-141.
- Nav Persaud's homepage
- Consciousness may be a sure bet by Nikhil Swaminathan in Scientific American Mind
- Put your money where your mind is by Kaspar Mossman in Scientific American