Working papers

I have several ongoing projects under review that I would welcome your feedback on the preprints (contact me here).


Partial list (*: equal contribution; underlined: supervised students; ^: corresponding author):

Quick jump to sections: My research projects / Supervised research (replications & meta-analyses) / Mass pre-registered JDM replications


My research projects

Feldman, G. What is normal? Contrasting the impact of past behavior, expectations and social norms normality categories over regret in the action-effect. DOI: 10.13140/RG.2.2.17274.39364

The widely-replicated action-effect (Kahneman & Tversky, 1982a) describes a phenomenon in which negative outcomes are associated with higher regret when they are a result of action compared to inaction. The highly influential norm-theory (Kahneman & Miller, 1986) theorized that the effect can be explained using the concept of normality, arguing that inaction is more ‘normal’. We aimed to clarify the concept of normality and examine the impact on regret in the action-effect by contrasting three identified categories: past-behavior normality, role/situational expectations normality, and social-norms normality. In three exploratory experiments (N1 = 213, N2 = 300, N3 = 303) and one concluding pre-registered combined experiment (N = 403) we found that the three normality categories had distinct effects with consistent medium to strong impact on the regret action-effect (d = .51 to d =.85) and no interactions. The action-effect was significantly weakened into an inaction-effect in the joint effects of any two types of the three normality categories (d = 1.56 to 1.61) and with all three combined (d = 2.75). In addition, we conclude three replications for each of the normality categories effects, overall nine successful replications of previous findings.


Feldman, G. Personal values and moral foundations: Examining relations and joint prediction of moral variables. DOI: 10.13140/RG.2.2.32570.49600/2

We examined the relationship between personal values and moral foundations by conducting a mini meta-analysis of the values-foundations links of five large-scale cross-cultural samples (N = 32,492). We further tested whether the two theories predicted unique variance in moral variables. We found support for values and foundations as having consistent and theoretically meaningful relationships. Broadly, self-transcendence versus self-enhancement values dimensions were associated with individualizing foundations, whereas conservation versus openness-to-change values dimensions were associated with binding foundations. Links between low-level values and foundations followed the expected theoretical values theory circumplex structure sinusoidal pattern. Dimensions of the two theories predicted unique variance in morality attitudes, behavior, and individual-differences.



Supervised research (replications & meta-analyses)

*Yay, T, & ^*Feldman, G. Action-inaction asymmetries in moral scenarios: Replication of the omission bias examining morality and blame with extensions linking to causality, intent, and regret. DOI: 10.13140/RG.2.2.10240.74242

Omission bias is the preference for harm through omissions over harm through commissions. In a pre-registered experiment (N = 313), we successfully replicated an experiment from Spranca, Minsk, and Baron (1991), considered a classic demonstration of the omission bias, in an adjusted between-subject design addressing concerns regarding the generalizability of the effect beyond within-subject experiments. Participants in the harm through commission condition(s) rated harm as more immoral and attributed higher responsibility compared to participants in the harm-through-omission condition (d = 0.40 to 0.52). We examined several extensions. An action-inaction asymmetry was also found for causality and intent, in that commissions were attributed stronger action-outcome links and higher intentionality (d = 0.21 to 0.58). The effect for regret was opposite from the classic findings on the action-effect, with higher regret for inaction over action (d = -0.26 to -0.19). Overall, higher perceived causality and intent were associated with higher attributed responsibility, and with lower perceived morality and regret.


*Gao, Y., *Yao, D., & ^Feldman, G. Owning leads to valuing: Meta-analysis of the Mere Ownership Effect. DOI: 10.13140/RG.2.2.13568.33287/1

Mere ownership effect is the phenomenon that people tend to value what they own more than what they do not own. This classic effect is considered robust, yet evidence is inconsistent, effect size varies across studies, and the effect is often confused for or confounded with other classic phenomena, such as endowment or mere exposure effects. We conducted a pre-registered meta-analysis of 26 samples (N = 3132) resulting in medium effect-size for psychological ownership on valuing (g = 0.59, 95% confidence intervals [CI] [0.47, 0.70]). Moderator analyses showed that object materiality, use of replica, and valuing type were the strongest moderators, compared to weaker effects for loss aversion, type of ownership (implicit/explicit), level of exposure, and study design (between/within). Mere ownership effects were different than null across all moderator categories and publication bias corrections. We conclude that psychological owning indeed leads to valuing, regardless of endowment, loss-aversion, and mere-exposure effects.


*Yao, D., *Gao, Y., & ^Feldman, G. Impact of ownership on liking and value: Replication of three ownership effect experiments. DOI:

Ownership increases liking and the perceived value of owned objects (Beggan, 1992). In a pre-registered study, we conducted a close replication of three ownership effect experiments using different paradigms (N = 311). We successfully replicated Nuttin’s (1987) name-letter effect in Part 1, and Mandel’s (2002) endowment effect in Part 3. In these experiments, participants rated higher liking for letters of the alphabet when letters appeared in their first name (vs. when letters did not) (Part 1; d = 0.19) and set a higher price for an object when they were the owners (vs. non-owners) (Part 3; d = 0.65). However, we failed to find support for Irmak, Wakslak, and Trope’s (2013) endowment effect findings in Part 2 (target’s effect: d = 0.99) of price differences between sellers and buyers (replication effect: d = 0.10).


*Kutscher, L., & ^*Feldman, G. Impact of past behavior normality on regret: Meta-analysis of exceptionality effect. DOI: 10.13140/RG.2.2.12760.96005

Exceptionality effect is the phenomenon that people associate stronger regret with a negative outcome when it is a result of an exception (abnormal behavior) compared to when it is a result of routine (normal behavior). In this pre-registered meta-analysis, we examined exceptionality effect in 45 studies (N = 3933). An analysis of 34 experimental studies (n = 2716) showed medium to strong effect (g = 0.72, 95% confidence intervals (CI) [0.46, 0.97]) for past behavior on regret across several regret related measures (regret: g = 0.83, counterfactuals: g = 0.40, self-blame: g = 0.44, victim compensation: g = 0.44, offender punishment: g = 0.50). An analysis of 13 studies presenting a comparison of exceptional and routine behaviors simultaneously (n = 1217) revealed an even stronger exceptionality effect (converted g = 2.07, CI [1.59, 2.43]). We tested several theoretical moderators – norm strength, event controllability, outcome exceptionality, action versus inaction, and status quo. Exceptionality effect was stronger when the routine was aligned with the status quo option, when the exception represented an omission compared to a commission, and when the outcome was more common (less exceptional).


Other concluded work with former students:



Mass pre-registered JDM replications reports

See our summary page about this mass pre-registered JDM replication project. We’re in the process of sharing all materials.

Collaborative guide: Pre-registered replication project collaborative manual (DOI: 10.13140/RG.2.2.16660.14723/1)


*Chandrashekar, S.P., *Yeung, S., Yau, K., ^Feldman, G., et al. Agency and self-other asymmetries in perceived bias and shortcomings: Replications of the Bias Blind Spot and extensions linking to free-will beliefs. DOI: 10.13140/RG.2.2.19878.16961

Bias Blind Spot is the phenomenon that people tend to perceive themselves to be less susceptible to biases than others. In three pre-registered experiments with samples from Hong Kong and the United States (overall N = 969), we replicated two experiments (Study 1-Survey 2 and Study 2) from Pronin, Lin, and Ross (2002), the first published demonstration of the effect. Participants rated themselves lower than others in both susceptibility to biases (mini meta-analysis: dz = – 1.00) and personal shortcoming (mini meta-analysis: dz = – 0.34). The self-other asymmetry of susceptibility for biases was larger than that of personal shortcomings (dz = – 0.43). Thus, the replication findings provide strong empirical support for the bias blind spot phenomenon. Extending the replications, belief in free will was positively associated with the susceptibility to better than average effect, and also with a stronger self-other asymmetry in rating personal shortcomings.


*Ziano, I., *Wang, Y. J., *Sany, S., ^*Feldman, G., et al. Perceived morality of direct versus indirect harm: Replications of the preference for indirect harm effect. DOI: 10.13140/RG.2.2.29720.85760

Royzman and Baron (2002) demonstrated that people prefer indirect harm to direct harm: they judge actions that produce harm as a by-product to be more moral than actions that produce harm directly. In two preregistered studies, we successfully replicated Study 2 of Royzman and Baron (2002) with a Hong Kong student sample (N = 45) and an online American Mechanical Turk sample (N = 314). We found consistent evidential support for the preference for indirect harm phenomenon (d = 0.34, 95% confidence intervals (CI) [0.23, 0.45]), weaker than effects reported in the original findings of the target article (d = 0.70, CI [0.40, 1.00]). We also successfully replicated findings regarding reasons underlying a preference for indirect harm (directness, intent, omission, probability of harm, and appearance of harm).


*Chandrashekar, S.P., *Cheng, Y., *Fong, C.,*Leung, Y., *Wong, Y., Cheng, B., & ^Feldman. Frequency estimation and semantic ambiguity do not eliminate ‎conjunction bias, when it occurs: Replication and extension of ‎Mellers, Hertwig, and Kahneman (2001)‎. DOI: 10.13140/RG.2.2.22730.52168

Mellers, Hertwig, and Kahneman (2001) conducted an adversarial collaboration to try and ‎resolve Hertwig’s contested view that frequency formats eliminate conjunction effects, and that ‎conjunction effects were largely due to semantic ambiguity. We conducted a pre-registered ‎well-powered very close replication ‎(N = 1032), ‎testing two personality profiles (Linda and ‎James) in a four between-subject design comparing unlikely and likely items to “and” and “and ‎or” conjunctions. Linda profile findings replicated conjunction effect and were consistent with ‎Tversky and Kahneman’s (1983) arguments for a representative heuristic. We found no support ‎for semantic ambiguity. The results of James profile were a likely failure, with no conjunction ‎effect. We provided additional tests addressing possible reasons, in line with later literature ‎suggesting conjunction effects may be context sensitive. We discuss implications for research ‎on conjunction effects, and call for further well-powered pre-registered replications and ‎extensions of classic findings in judgment and decision-making.‎


So far:


Guided thesis work in the mass replication project:

Kam & ^Feldman. Wong & Kwong 2007 Escalation of commitment and anticipated regret: Replication and extension

People tend to make decisions that aim to minimise potential regret, especially when faced with failing courses of action. Wong and Kwong (2007) suggested that the effects of personal responsibility and regret possibility would increase the escalation of commitment, and that escalation of commitment would increase with the net anticipated regret about withdrawal. We conducted a pre- registered replication and extension of Studies 1a and 2a in Wong and Kwong (2007) and failed to find empirical support for the main effects of personal responsibility and regret possibility. However, we found support for a positive correlation between net anticipated regret about withdrawal and escalation of commitment in conditions with and without regret possibility. A pre-registered extension showed that whereas neuroticism may not be associated with escalation of commitment, neuroticism is associated with anticipated regret about withdrawal and persistence.


Kwan & ^Feldman. Roese & Olson 1996 JESP Counterfactuals, causality, & hindsight: Replication and extension

Hindsight bias is the phenomenon that people tend to perceive themselves as being able to predict the outcome all along after they have been presented with the outcome of a certain event. Counterfactuals involve thoughts about alternative realities leading to alternative possible outcomes and were found to heighten hindsight bias. Participants linked the pre-occurrence event (antecedent) with the outcome by causal inferences, assuming the antecedent causing the outcome. We conducted a pre-registered replication and extension of the classic experiment 3 by Roese and Olson (1996) with Americans on Amazon Mechanical Turk (N = 527). Participants presented with outcomes estimated those outcomes to be of a higher likelihood than those who were not presented with outcomes (d = -0.82, 95% confidence intervals (CI) [-1.00, -0.63]). Participants that were provided information about the main character (J.S.) drove on an unusual route (exceptional antecedent) had higher agreement ratings with J.S.’s best option would have been taking the usual route (d = 0.29, 95% CI [0.12, 0.46]), while participants presented with J.S. drove in the usual route (routine antecedent) had higher agreement ratings with J.S.’s best option would have been taking the unusual route (d = 0.29, 95% CI [0.12, 0.46]). Participants presented with unusual route also estimated higher ratings on J.S. being blamed (d = 0.84, 95% CI [0.65, 1.02]), and his action brought about his missing flight (d = 1.00, 95% CI [0.81, 1.19]), and the results did not influence by the outcome manipulation. Thus, this replication study provides strong empirical support for demonstrating causal attribution enhances the positive relationship between counterfactual thinking and hindsight bias. We extended this replication study by adding two individual predictors, rigidity, impression management, and the level of regret. Participants that received exceptional antecedent estimated a higher level of regret for J.S. than routine antecedent (d = 0.51, 95% CI [0.21, 0.81]).


Hung & ^Feldman. Hsee & Zhang 2004 JPSP Distinction Bias: Replication & extension

Hsee and Zhang (2004) suggested that evaluation modes, joint evaluation (JE) or separate evaluation (SE), affected how options were evaluated. Options are viewed concurrently in JE and in isolation in SE, so that people in SE only know one of the options. Over-examining the mere differences between options, people in JE overpredict the difference in happiness between people being presented with different options. It was termed as distinction bias in which people in JE fail to evaluate the options from SE perspective but rely on JE preference, resulting in the overprediction of difference in happiness in SE. Our pre-registered replication results of Experiments 1 and 2 using an American online sample (N = 824) largely supported the findings in Hsee and Zhang (2004). In Experiment 1, JE participants predicted that the more people buying the book, the happier the writer in SE was (80-buyer Vs. 160-buyer: d = 1.09, 95% confidence interval (CI) [0.80, 1.38] and 160-buyer Vs. 240-buyer: d = 0.59, CI [0.31, 0.87]). However, SE participants reported no significant difference in happiness (80-buyer Vs. 160- buyer: d = 0.09, CI [-0.19, 0.36] and 160-buyer Vs. 240-buyer: d = 0.03, CI [-0.25, 0.30]). Similarly, in Experiment 2, JE participants predicted that reading more negative words elicited greater unhappiness (d = 0.52, CI [0.23, 0.82]). Yet, SE participants reading 10 and 25 negative words reported similar levels of unhappiness (d = 0.04, CI [-0.25, 0.34]). Extending the replication, higher certainty was reported in prediction for qualitatively different options than quantitatively different options in JE (Experiment 1: d = 0.45, CI [0.13, 0.85] and Experiment 2: d = 0.45, CI [0.31, 0.87]).


Mok, P. & ^Feldman, G. Alicke (1985): Pre-registered replication and extension. DOI: 10.13140/RG.2.2.30198.04164

People seem to regard themselves as better than the average other. To revisit this phenomenon, we conducted a pre-registered replication and extension of Alicke’s (1985) study on the effect of trait dimensions for self versus average other judgments, collecting data from American Amazon Mechanical Turk workers in two waves (N = 670; N = 903). For more effective analyses, we switched to a correlational design after pre-testing the data. We successfully replicated the effect of trait desirability for the self-ratings in relation to average other ratings, such that participants rated more desirable traits as more descriptive of themselves than of the average American (original: ηp2 = .78, 95% CI [.73, .81]; replication: sr2 = .54, 95% CI [.43, .65]). In line with the original findings, we found that the effect was stronger for traits of higher controllability (original: ηp2 = .21, 95% CI [.12, .28]; replication: sr2 = .07, 95% CI [.02, .12]). As an extension, we measured commonness, the degree to which a trait is frequently displayed among the average American. The extension revealed that more desirable traits were rated as more common (sr2 = .04, 95% CI [-.01, .09]) and this held for the average American (sr2 = .41, 95% CI [.31, .52]) but not the self (sr2 = .00, 95% CI [-.01, .01]). Three decades after the original study, the better-than-average effect appears to remain robust. We discuss implications for future research.


Chan, W. & ^Feldman, G. Temporal Value Asymmetry: Pre-registered replications and extensions of Caruso, Gilbert, & Wilson (2008) and Caruso (2010). DOI: 10.13140/RG.2.2.33959.80801

Future events tend to elicit stronger emotions than past events (Caruso, Gilbert, & Wilson, 2008) and people therefore tend to make more extreme judgments towards a future event than an identical past event (Caruso et al., 2008). The phenomenon has been coined “temporal value asymmetry” (TVA). We ran a very close replication of three experiments from two studies of the TVA: Studies 1 and 4 of Caruso et al. (2008) demonstrating the TVA in monetary valuation, and Study 1 of Caruso (2010) demonstrating TVA in moral judgments. We also added an extension to see if TVA would display in a different context. We failed to find support for TVA in monetary and moral judgments. Our results showed that whether an event happened in the past or future had little to no effect on judgments. Further research would help establishing limits or generalization of TVA.


Lai, Y. & ^Feldman, G. Epstein, Denes-Raj and Pacini (1995): Pre-registered replication & extension. DOI: 10.13140/RG.2.2.33116.33925

Epstein, Denes-Raj & Pacini (1995) have suggested that the phenomenon of people prone to make conjunction errors in the notorious Linda problem can be explained by cognitive-experiential self-theory (CEST), which proposed that people process information with two cognitive systems, the experiential system and rational system. In this preregistered replication study, Epstein, Denes-Raj & Pacini (1995) was replicated successfully with the online MTurk workers sample (N = 903). It was argued in the original study that the conjunction heuristic responses were primarily elicited by the concrete and unnatural context property of the Linda problem, the replication study has found a smaller but still very robust effect in supporting this claim (original: w2= 0.98; replication: w2= 0.95). We have also found consistent evidential support that there were factors that would weaken heuristic responses, such as being statistically sophisticated (original: w2= 0.24, replication: w2= 0.16, 95% CI = [0.07, 0.27]), and priming intuitive statistical knowledge (original: w2= 0.17, replication: w2= 0.34, 95% CI [0.21, 0.47]. An exploratory statistical test has even revealed an effect in priming rational system in reducing heuristic thinking (w2= 0.18, 95% CI [0.05, 0.31]).


There are others, if this kind of work interests you, email me for other ongoing work.