Certainty and Uncertainty: The Two Faces of the Hindsight Bias

“Hindsight Bias” is a person's tendency, after learning about the actual outcome of a situation or the correct answer to a question, to distort a previous judgment in the direction of this new information. In the literature, hindsight bias has been mostly discussed as an inevitable result of a “judgment under uncertainty.” We think that the hindsight bias is due to memorial as well as inferential processes: Whereas certainty about the recollection is memorial and concerns the recollective experience, certainty at the time of the judgment is inferential and concerns the individual's metaknowledge (“I know that I knew that”). In two experiments participants' feelings of certainty were measured indirectly (Koriat & Goldsmith, 1996) by giving participants the option of leaving those questions unanswered about which they felt uncertain. This free-report option was offered to half of the participants in the first estimate phase (concerning time of judgment) and to the second half in the memory phase (concerning the recollective experience). At the end of the session, participants were presented again with the questions they had skipped and were now required to answer them. This procedure allowed us to compare the amount of hindsight bias for the skipped, uncertain items to the spontaneously answered, certain ones. Both experiments demonstrated that the hindsight bias is a result of the interaction of both uncertainty and certainty.     

大学生自信发展特点的研究

本研究选取931名大学生为被试,采用自编“大学生自信问卷”?考察了大学生自信的年龄特点、性别差异及学校类型差异。结果表明：大学生自信总水平差异显著,除社交自信外,整体自信、学业自信和身体自信均存在极显著性别差异,男生高于女生;大学生在整体、学业和身体自信及其分维度上存在显著年级差异：一、四年级显著高于二、三年级,其中三年级自信度明显低于其它年级;学业、社交、身体自信及其分维度存在显著学校类型差异,非重点大学学生显著高于重点大学学生;年级和学校类型在身体自信、整体自信、学业自信、社交自信及其某些层面上存在显著交互作用;性别和学校类型在身体自信某分维度的交互作用达显著水平。     

Fluctuations of answer and confidence rating in a general-knowledge problem task: Is confidence rating a result of direct memory-relevant output monitoring?

In order to examine whether confidence rating is a result of direct memory-relevant output monitoring, a general-knowledge test was used. A question set was answered in experimental session 1, and the same question set was employed in session 2. Subjects chose one answer from two alternatives. No significant difference in answer accuracy was found between sessions, which suggests that identical information processing was carried out, and that any difference between sessions was based on probabilistic fluctuation. The major results were as follows: (a) In both sessions, answer accuracy was an increasing function of confidence. (b) Answer-change rate (rate of different answers to the same question between sessions) decreased monotonically with confidence. (c) The distributions of confidence rating in session 2, conditional on the rating in session 1, were nearly identical in no-answer-change and answer-change cases. These results suggest that confidence rating is not a result of monitoring memory-relevant output itself but an estimation of possibility of answer fluctuation.     

Response persistence under ratio and interval reinforcement schedules

In Experiment 1, rats were exposed to progressive-ratio schedules of food reinforcement while other rats were exposed simultaneously to yoked-interval schedules that arranged equivalent interreinforcer intervals but required only a single response at the end of the interval for food delivery. In Experiment 2, a within-subject yoked-control procedure was employed in which pigeons were exposed to alternating sessions (one per day) of progressive-ratio schedules and yoked-interval schedules as described above. In both experiments, responding under the yoked-interval schedule persisted beyond the point at which responding under the progressive-ratio schedule had ceased. The progressive-ratio schedules controlled break-and-run distributions, and the yoked-interval schedules controlled more even distributions of responses in time. Response rates decreased and postreinforcement pauses increased over time within individual sessions under both schedules. The results suggest that responding maintained by interval schedules is more persistent than that maintained by ratio schedules. The limitations and implications of this conclusion are discussed in the context of other investigations of response strength and behavioral momentum.     

Temporal control in rats: analysis of nonlocalized effects from short interfood intervals.

The present experiment analyzed temporal control of postreinforcement pause duration during within-session changes in the criterion for reinforcement (interfood interval, IFI). Analysis of interval-by-interval changes in the pause revealed localized and nonlocalized effects from short intervals that caused specific changes in performance. In Phase 1, rats were presented with five consecutive 15-s IFIs intercalated into a series of 60-s IFIs. The 15-s set decreased the pause in adjacent and more remote 60-s intervals. In Phase 2, two sets of 15-s intervals were intercalated. The spacing between the two sets varied so that 0, 5, 10, or 15 60-s IFIs separated the sets. The postreinforcement pause tracked all changes in the IFI duration, and the localized effect from a short set extended beyond the next interval to the next few 60-s IFIs. Effects from one set, however, did not combine with a second set: Changes in the pause after two sets were the same regardless of the spacing between sets.     

Concurrent second-order schedules of reinforcement

Responses on one key (the main key) of a two-key chamber produced food according to a second-order variable-interval schedule with fixed-interval schedule components. A response on a second key (the changeover key) alternated colors on the main key and provided a second independent second-order variable-interval schedule with fixed-interval components. The fixed-interval component on one variable-interval schedule was held constant at 8 sec, while the fixed interval on the other variable-interval schedule was varied from 0 to 32 sec. Under some conditions, a brief stimulus terminated each fixed interval and generated fixed-interval patterns; in other conditions, the brief stimulus was omitted. Relative response rate and relative time deviated substantially from scheduled relative reinforcement rate and, to a lesser extent, from obtained relative reinforcement rate under both brief-stimulus and no-stimulus conditions. Matching was observed with equal components on both schedules; with unequal components, increasingly greater proportions of time and responses than the matching relation would predict were spent on the variable-interval schedule containing the shorter component. Preference for the shorter fixed interval was typically more extreme under brief-stimulus than under no-stimulus schedules. The results limit the extension of the matching relation typically observed under simple concurrent variable-interval schedules to concurrent second-order variable-interval schedules.     

Effects of signaling on temporal control of behavior in response‐initiated fixed intervals

Behavior and events distributed in time can serve as markers that signal delays to future events. The majority of timing research has focused on how behavior changes as the time to some event, usually food availability, decreases. The primary objective of the two experiments presented here was to assess how behavior changes as time passes between two time markers when the first time marker was manipulated but the second, food delivery, was held constant. Pigeons were exposed to fixed‐interval, response‐initiated fixed‐interval, and signaled response‐initiated fixed‐interval 15‐ and 30‐s schedules of reinforcement. In Experiment 1, first‐response latencies were systematically shorter in the signaled response‐initiated schedules than response‐initiated schedules, suggesting that the first response was a more effective time marker when it was signaled. In Experiment 2, responding in no‐food (i.e. “peak”) trials indicated that timing accuracy was equivalent in the three schedule types. Compared to fixed interval schedules, timing precision was reduced in the signaled response‐initiated schedules and was lowest in response‐initiated schedules. Results from Experiments 1 and 2 coupled with previous research suggest that the overall “informativeness” of a time marker relative to other events and behaviors in the environment may determine its efficacy.     

A Test by Any Other Name: P Values,Bayes Factors,and Statistical Inference

Procedures used for statistical inference are receiving increased scrutiny as the scientific community studies the factors associated with insuring reproducible research. This note addresses recent negative attention directed at p values, the relationship of confidence intervals and tests, and the role of Bayesian inference and Bayes factors, with an eye toward better understanding these different strategies for statistical inference. We argue that researchers and data analysts too often resort to binary decisions (e.g., whether to reject or accept the null hypothesis) in settings where this may not be required.     

A Comparative Investigation of Confidence Intervals for IndependentVariables in Linear Regression

In linear regression, the most appropriate standardized effect size for individual independent variables having an arbitrary metric remains open to debate, despite researchers typically reporting a standardized regression coefficient. Alternative standardized measures include the semipartial correlation, the improvement in the squared multiple correlation, and the squared partial correlation. No arguments based on either theoretical or statistical grounds for preferring one of these standardized measures have been mounted in the literature. Using a Monte Carlo simulation, the performance of interval estimators for these effect-size measures was compared in a 5-way factorial design. Formal statistical design methods assessed both the accuracy and robustness of the four interval estimators. The coverage probability of a large-sample confidence interval for the semipartial correlation coefficient derived from Aloe and Becker was highly accurate and robust in 98% of instances. It was better in small samples than the Yuan-Chan large-sample confidence interval for a standardized regression coefficient. It was also consistently better than both a bootstrap confidence interval for the improvement in the squared multiple correlation and a noncentral interval for the squared partial correlation.     

关于假设检验的争议:问题的澄清与解决

对传统的假设检验作为心理学实验的数据分析工具的评价涉及两个标准:首先是它是否合法,其次是它是否有用。置于频率学派统计学框架中的传统假设检验在逻辑上实际上是合法的;但它在效用性方面则有着备择假设不可证伪以及只能提供定性结论这两方面的缺陷。置信区间能够集中地改进和弥补这些缺陷。对传统假设检验使用中错误的澄清也使得研究者们开始重视PSI问题,这使得心理学实验的设计和数据分析从关注总体转向关注个体。