Sections:

Publications
Thesis
Presentations

Publications

Copyright Notice: Some of the documents listed below are available for downloading. These have been provided as a means to ensure timely dissemination of scholarly and technical work on a noncommercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be re-posted without the explicit permission of the copyright holder.

Forthcoming

• Speekenbrink, M., & Shanks, D.R. (in press). Decision making. In: D. Reisberg (Ed.), Oxford handbook of cognitive psychology. [ abstract | preprint ]
This chapter reviews normative and descriptive aspects of decision making. Expected Utility Theory (EUT), the dominant normative theory of decision making, is often thought to provide a relatively poor description of how people actually make decisions. Prospect Theory has been proposed as a more descriptively valid alternative. The failure of EUT seems at least partly due to the fact that people’s preferences are often unstable and subject to various influences from the method of elicitation, decision context, and goals. In novel situations, people need to infer their preferences from various cues such as the decision context, memories, and emotions. Through repeated experience with particular decisions and their outcomes, these inferences can become more stable, resulting in behavior that is more consistent with EUT.

2012

• Berry, C. J., Shanks, D. R., Speekenbrink, M., & Henson, R. N. A. (2012). Models of recognition, repetition priming, and fluency: Exploring a new framework. Psychological Review, 119, 40-79. [ abstract | pdf  ]
We present a new modeling framework for recognition memory and repetition priming based on signal detection theory. We use this framework to specify and test the predictions of four different models: 1) a SS (single-system) model, in which one continuous memory signal drives recognition and priming, 2) a MS1 (multiple-systems-1) model, in which completely independent memory signals (such as explicit and implicit memory) drive recognition and priming, 3) a MS2 (multiple-systems-2) model, in which there are also two memory signals, but some degree of dependence is allowed between these two signals (and this model subsumes the SS and MS1 models as special cases), 4) a DPSD1 (dual-process signal detection) model, one possible extension of a dual-process theory of recognition (Yonelinas, 1994) to priming, in which a signal-detection model is augmented by an independent recollection process. The predictions of the models are tested in a continuous identification with recognition (CID-R) paradigm in both normal adults (Experiments 1-3) and amnesic individuals (using data from Conroy et al., 2005). The SS model predicted numerous results in advance. These were not predicted by the MS1 model, though could be accommodated by the more flexible MS2 model. Importantly, measures of overall model fit favoured the SS model over the others. These results illustrate a new, formal approach to testing theories of explicit and implicit memory.
• Osman, M., & Speekenbrink, M. (2012). Prediction and control in a dynamic environment. Frontiers in Psychology, 3:68. [ abstract | pdf  ]
The present study compared the accuracy of cue-outcome knowledge gained during prediction-based and control-based learning in stable and unstable dynamic environments. Participants either learnt to make cue interventions in order to control an outcome, or learnt to predict the outcome from observing changes to the cue values. Study 1 (N = 60) revealed that in tests of control, after a short period of familiarization, performance of Predictors was equivalent to Controllers. Study 2 (N = 28) showed that Controllers showed equivalent task knowledge when to compared to Predictors. Though both Controllers and Predictors showed good performance at test, overall Controllers showed an advantage. The cue-outcome knowledge acquired during learning was sufficiently flexible to enable successful transfer to tests of control and prediction.
• Speekenbrink, M. (2012) Amnesia and learning. In: N. Seel (Ed.). Encyclopedia of the Sciences of Learning. New York: Springer.
• Speekenbrink, M., Twyman, M.A., & Harvey, N. (2012). Change detection under autocorrelation. In N. Miyake, D. Peebles, & R. P. Cooper (Eds.), Proceedings of the 34th Annual Conference of the Cognitive Science Society. Austin, TX: Cognitive Science Society. [ abstract | preprint ]
Judgmental detection of changes in time series is an ubiquitous task. Previous research has shown that human observers are often relatively poor at detecting change, especially when the series are serially dependent (autocorrelated). We present two experiments in which participants were asked to judge the occurrence of changes in time series with varying levels of autocorrelation. Results show that autocorrelation increases the difficulty of discriminating change from no change, and that observers respond to this increased difficulty by biasing their decisions towards change. This results in increased false alarm rates, while leaving hit rates relatively intact. We present a rational (Bayesian) model of change detection and compare it to two heuristic models that ignore autocorrelation in the series. Participants appeared to rely on a simple heuristic, where they first visually match a change function to a series, and then determine whether the putative change exceeds the variability in the data.

2011

• Gerstenberg, T., Lagnado, D. A, Speekenbrink, M. & Cheung, C. (2011). Rational order effects in responsibility attributions. In L. Carlson, C. Hölscher, & T. Shipley (Eds.), Proceedings of the 33rd Annual Conference of the Cognitive Science Society. Austin, TX: Cognitive Science Society [ abstract | pdf  ]
Two experiments establish a rational order effect in responsibility attributions. Experiment 1 shows that in a team challenge in which players contribute sequentially, the last player's blame or credit for a performance is reduced if the team's result is already determined prior to his acting. However, credit and blame attributions still vary with quality of performance in these cases. This finding is at odds with Spellman (1997) who proposed that a person's perceived contribution varies with the degree to which it changes the probability of the eventual outcome. Experiment 2 illustrates that the rational order effect does not overgeneralize to situations in which the experienced order of events does not map onto the objective order of events. The quality of the last person's performance is only discredited if she knew that the result was already determined.
• Osman, M., & Speekenbrink, M. (2011). Cue utilization and strategy application in stable and unstable dynamic environments. Cognitive Systems Research, 12, 355-364. [ abstract | preprint  ]
We took a novel empirical approach to investigating dynamic decision making behavior by examining the profiles of individuals' information sampling behavior and strategy application under conditions in which the control task was unstable as well as stable. Participants were presented with a dynamic system which they interacted with by intervening on three cues in order to reach and maintain a specific outcome (goal). The system was manipulated so that in the Stable condition participants controlled an outcome that fluctuated steadily overall trials, and in the Unstable condition the outcome fluctuated erratically over trials. In general, unstable fluctuations in the outcome led people to sample all the cues most of the time, even those which had no effect on the outcome. In contrast, under Stable conditions people were more conservative in their cue sampling behavior. The implications of these findings are discussed with respect to previous work on dynamic decision making and the Monitoring and Control (Osman, 2010a, 2010b) framework.
• Osman, M., & Speekenbrink, M. (2011). Controlling stable and unstable decision making environments. In L. Carlson, C. Hölscher, & T. Shipley (Eds.), Proceedings of the 33rd Annual Conference of the Cognitive Science Society. Austin, TX: Cognitive Science Society [ abstract | pdf  ]
In the present study we ask: Are people sensitive to the stability of a dynamic environment under short exposure to it? To examine this we investigate people's cue manipulation and strategy application when instructed to learn to control an outcome in a dynamic system by intervening on three cues. The system was designed in such a ways that in the Stable condition participants controlled an outcome that fluctuated steadily overall 40 trials, and in the Unstable condition the outcome fluctuated erratically over 40 trials. In the present study we show that people tended to intervene more frequently on all three cues when the system was Unstable compared to the when the system was Stable. Overall, the evidence from this study supports the general prediction made from the Monitoring and Control framework (Osman, 2010a, 2010b). It claims that people are sensitive to the underlying stability of dynamic environments in which they are required to control the outcome, but are insensitive to autonomous characteristics of the system.
• Speekenbrink, M., & Shanks, D.R. (2011). Is everyone Bayes? On the testable implications of "Bayesian fundamentalism". Behavioral and Brain Sciences, 34, 213-214. [ abstract | preprint  ]
  note: The published version of this commentary has suffered from mysteriously dissappearing characters so I recommend reading the preprint.
A central claim of Jones and Love's article is that Bayesian fundamentalism is empirically unconstrained. Unless constraints are placed on prior beliefs, likelihood and utility functions, all behaviour - it is proposed - is consistent with Bayesian rationality. Although such claims are commonplace, their basis is rarely justified. We fill this gap by sketching a proof, and we discuss possible solutions that would make Bayesian approaches empirically interesting.

2010

• Lagnado, D.A., & Speekenbrink, M. (2010). The influence of delays in real-time causal learning. The Open Psychology Journal, 3(2), 184-195. [ abstract | pdf ]
The close relation between time and causality is undisputed, but there is a paucity of research on how people use temporal information to inform their causal judgments. Experiment 1 examined the effect of delay variability on causal judgments, and whether participants were sensitive to the presence of a hastener cue that reduced the delay between cause and effect without changing the contingency. The results showed that higher causal ratings were given to cause-effect pairs with less variable delays, but that conditions with an active hastener actually reduced participants' ratings of the causal cues. The latter finding can also be explained in terms of people's sensitivity to variability, because an undetected hastener leads to greater variability in experienced delays. Experiment 2 followed up previous research showing that people give higher causal ratings to cause-effect pairs with shorter delays. We examined whether this finding might be due to the greater probability of intervening events rather than the length of delay per se. The results supported the former conjecture: participants' causal ratings were influenced by the probability of intervening events in the cause-effect interval and not the mere length of delay. The findings from both experiments raise questions for current theories of causal learning.
• Speekenbrink, M., Lagnado, D. A, Wilkinson, L., Jahanshahi, M., & Shanks, D. R. (2010). Models of probabilistic category learning in Parkinson's disease: Strategy use and the effects of L-dopa. Journal of Mathematical Psychology, 54, 123-136. [ abstract | pdf ]
Probabilistic category learning (PCL) has become an increasingly popular paradigm to study the brain bases of learning and memory. It has been argued that PCL relies on procedural habit learning, which is impaired in Parkinson's Disease (PD). However, as PD patients were typically tested under medication, it is possible that L-dopa caused impaired performance in PCL. We present formal models of rule-based strategy switching in PCL to re-analyse the data from Jahanshahi et al. (submitted for publication) comparing PD patients on and off medication (within subjects) to matched controls. Our analysis shows that PD patients followed a similar strategy switch process as controls when off medication, but not when on medication. On medication, PD patients mainly followed a random guessing strategy, with only few switching to the better Single Cue strategies. PD patients on medication and controls made more use of the optimal Multi-Cue strategy. In addition, while controls and PD patients off medication only switched to strategies which did not decrease performance, strategy switches of PD patients on medication were not always directed as such. Finally, results indicated that PD patients on medication responded according to a probability matching strategy indicative of associative learning, while PD patients off medication and controls were consistent with a rule-based hypothesis testing procedure.
• Speekenbrink, M., & Shanks, D. R. (2010). Learning in a changing environment. Journal of Experimental Psychology: General, 139, 266-298. [ abstract | pdf ]
Multiple cue probability learning studies have typically focused on stationary environments. We present 3 experiments investigating learning in changing environments. A fine-grained analysis of the learning dynamics shows that participants were responsive to both abrupt and gradual changes in cue-outcome relations. We found no evidence that participants adapted to these types of change in qualitatively different ways. Also, in contrast to earlier claims that these tasks are learned implicitly, participants showed good insight into what they learned. By fitting formal learning models, we investigated whether participants learned global functional relationships or made localized predictions from similar experienced exemplars. Both a local (the associative learning model) and a global learning model (the Bayesian linear filter) fitted the data of the first 2 experiments. However, the results of Experiment 3, which was specifically designed to discriminate between local and global learning models, provided more support for global learning models. Finally, we present a novel model to account for the cue competition effects found in previous research and displayed by some of our participants.
• Visser, I., Jansen, B.R.J., & Speekenbrink, M. (2010). A framework for discrete change. In: Newell, K.M. & Molenaar, P.C.M. (eds.) Individual Pathways of Change. Washington: APA (pp. 109-123). [ abstract| pdf ]
A class of models is developed for measuring and detecting discrete change in learning and development. The basic model for detecting such change is the latent or hidden Markov model. Traditionally, these models were restricted to categorical, mostly binary, observed variables, placing severe restrictions on possible measurement models. In this paper, the basic model is extended to include various distributions for the observed variables, including mixed multi-variate distributions. Moreover, there is an option to include time-varying predictors on the observed distributions. In effect, this model consists of mixtures of generalized linear models with Markovian dependencies over time to model the change process. In addition, the transition parameters can be estimated with covariates, such that the switching regime between states depends on characteristics of the individual or the experimental situation. The model is illustrated with examples of participants' learning in the Iowa Gambling Task and in the Weather Prediction Task.
• Visser, I., & Speekenbrink, M. (2010). depmixS4 : An R-package for hidden Markov models. Journal of Statistical Software, 36(7), 1-21. [ abstract | pdf ]
depmixS4 implements a general framework for defining and estimating dependent mixture models in the R programming language (R Development Core Team 2009). This includes standard Markov models, latent/hidden Markov models, and latent class and finite mixture distribution models. The models can be fitted on mixed multivariate data with distributions from the glm family, the logistic multinomial, or the multivariate normal distribution. Other distributions can be added easily, and an example is provided with the exgaus distribution. Parameters are estimated by the EM algorithm or, when (linear) constraints are imposed on the parameters, by direct numerical optimization with the Rdonlp2 routine.

< 2010

• Speekenbrink, M., & Shanks, D.R. (2008). Through the looking glass: A dynamic lens model approach to MCPL. In: Chater, N., & Oaksford, M. (Eds.). The probabilistic mind: Prospects for Bayesian cognitive science. Oxford: Oxford University Press. (pp. 409-429). [ abstract | preprint ]
We present a generalisation of lens model analysis which deals with non-stationarity of the environment and/or of the response system.
• Speekenbrink, M., Channon, S., & Shanks, D.R. (2008). Learning strategies in amnesia. Neuroscience and Biobehavioral Reviews, 32,292-310. [ abstract | pdf ]
Previous research suggests that early performance of amnesic individuals in a probabilistic category learning task is relatively unimpaired. When combined with impaired declarative knowledge, this is taken as evidence for the existence of separate implicit and explicit memory systems. The present study contains a more fine-grained analysis of learning than earlier studies. Using a dynamic lens model approach with plausible learning models, we found that the learning process is indeed indistinguishable between an amnesic and control group. However, in contrast to earlier findings, we found that explicit knowledge of the task structure is also good in both the amnesic and the control group. This is inconsistent with a crucial prediction from the multiple-systems account. The results can be explained from a single system account and previously found differences in later categorization performance can be accounted for by a difference in learning rate.
• Speekenbrink, M. (2004). De ongegronde eis tot consensus in de methodologie. Nederlands Tijdschrift voor de Psychologie, 59(1), 1-11. [ abstract | preprint ]
  note: This paper is in Dutch. An English version can be found as Chapter 6 in my thesis.
The imperative of scientific consensus, as proposed by Ziman and De Groot, is without a proper foundation. If the imperative has any appeal, consensus should function as a scientific goal, means, or criterion. It is argued that consensus fulfills none of these roles. In a goal of rational consensus, it is the rationality of the concerned parties that is of primary importance. Reaching agreement has no added value. Psychological research shows that, as a means for the improvement of judgment and decision, a standard of consensus is less effective than one of rational disagreement. As a definition of truth or an epistemic criterion, consensus has strong objections. One is that, in order to preserve its possible indicative function, a consensus criterion may not be applied. For these reasons, consensus has no place in normative methodology.
• Speekenbrink, M. (2003). The hierarchical theory and statistical model selection. In H. Yanai, A. Okada, K. Shigemasu, Y. Kano & J.J. Meulman (Eds.). New developments in psychometrics. (pp. 331-338). Tokio: Springer. [ abstract | preprint ]
This paper analyses the applicability of the hierarchical theory of justification as a normative theory of statistical model selection. The hierarchical theory prescribes a path from aim to evaluation criterion to model choice. Although the hierarchical theory has limited applicability as a normative theory due to underdetermination problems with equivalent models, it does provide an insightful framework for the area of statistical model selection.
• Speekenbrink, M. (2003). Book review of Gigerenzer (2002) Adaptive Thinking: Rationality in the Real World. The EADM Bulletin: Spring Edition 2003, 16-18. [ abstract | pdf ]
A not entirely positive, nor entirely negative, review of Gigerenzer's book.
• Speekenbrink, M., & Koele, P. (in revision). Weighting by competence: Group decisions in multiple cue tasks with distributed information [ abstract ]
Previous research suggests that groups usually adopt a simple majority rule to arrive at collective decisions. This can result in a situation in which the group fails to integrate decision-relevant information adequately. In a medical diagnosis task, we investigated whether groups adopt more complex decision strategies, in which decisions are weighted by individuals' decisional competence. When information is completely distributed over group members, groups adopted such a weighting-by-competence rule. When information was partly shared, partly unique, groups adopted a weighting-by-competence, or weighting-by-confidence rule, depending on the validity of the shared information. Only when information was entirely shared, did we find evidence for a simple majority rule. By adopting a lens model approach, we also investigated how the group decision process affected information utilization. While group decision processes that integrate preferences rather than information do no result in optimal information utilization, by adopting a weighting-by-competence rule, preference integration can closely mirror information integration.

Thesis

• Speekenbrink, M. (2005). Consensus and methodology. PhD thesis, Faculty of Social and Behavioral Sciences, University of Amsterdam. [ summary | press | pdf ]
This thesis investigates the role of consensus in psychological methodology from a variety of viewpoints, including empirical psychology, philosophy of science, decision theory, and statistics.

Consensus has been a central concept in Western thinking on science. Classically, consensus was taken as a consequence of the scientific method, which demands unanimous consent amongst those who adopt it. Later theories in the philosophy of science, starting with Kuhn (1970), assigned a more pivotal role to consensus, arguing that consensus is itself arbiter in scientific decision problems. An important reason for raising the status of consensus is the general problem of underdetermination. Underdetermination is discussed in Chapter 2, which proposes a more general version of underdetermination, called axiological underdetermination. In the decision-theoretic framework introduced in Chapter 2, scientific inference is regarded as goal-directed behaviour. Scientists pursue epistemic aims, such as descriptive and predictive adequacy and simplicity, and the choice between competing theories or methods is based on an evaluation of the utility of these theories and methods for those aspired aims. In such multi-attribute decision problems, underdetermination arises when multiple theories and methods have an identical utility.

While axiological underdetermination is, as empirical underdetermination, a problem of possible existence, Chapter 3 shows how axiological underdetermination can actually arise in practical problems of statistical model selection, and how methodological rules in general underdetermine theory choice. As such, the idea that theoretical consensus is a necessary consequence of methodological consensus should be abandoned.

Rather than being a direct consequence of the scientific method, some have argued that consensus is itself a basis for solving scientific decision problems. Since methodological rules are insufficient to determine theory choice, theorists such as Hesse (1980) have argued that social factors should be taken into account when explaining scientific decisions. A similar point is made in a major theory in social psychology. According to Festinger's (1950, 1954) theory of social comparison, when objective evidence is insufficient, individuals will attempt to validate their opinions by comparing them to those of others. From this theory, it follows that, if the choice between competing hypotheses is underdetermined by objective evidence, peer consensus on one of the alternative hypotheses should be a strong impetus to adopt this hypothesis. This prediction was tested in two experiments described in Chapter 4.

Chapter 5 investigates collective behaviour in nonmetric probability learning tasks. In general, there are two important reasons why a group can outperform individuals when it comes to making good judgements or decisions. The first is that individuals may possess (partly) non-overlapping information, so that the group as a whole can base its judgement or decision on more information than any individual alone. The second is that idiosyncratic biases may affect the group judgement or decision to a lesser extent than individual ones, because the idiosyncratic biases may cancel each other out in a group judgement or decision. While these two reasons render the assumption that groups are advantaged over individuals plausible, previous research has shown that groups often do not realise their potential. The two experiments of Chapter 5 were conducted in order to further scrutinise the assumption.

Chapter 6 addresses the possible roles of consensus in a normative methodology. Three such roles were distinguished: consensus as a goal, as a means, and as a criterion. It was argued that consensus fulfills none of these roles adequately.

Presentations

• Speekenbrink, M. (2011). Adaptive design for model discrimination. Paper presented at the London Judgement and Decision-Making group, London, February 2, 2011.
• Speekenbrink, M. (2011) Learning in a changing evironment. Invited talk, Department of Psychology, University of Swansea, Swansea, January 23, 2011
• Speekenbrink, M., Chater, N. & Shanks, D.R. Adaptive design for model discrimination. Paper presented at the Experimental Psychology Society meeting, London, January 2011.
• Speekenbrink, M. Adaptive design for model discrimination. Paper presented at the Annual Scientific Meeting of the BPS Mathematical, Statistical and Computing Section, Nottingham, 14 December 2010.
• Speekenbrink, M, Chater, N. & Shanks, D.R. Adaptive tests of cognitive models. Poster presented at the 51st Annual Meeting of the Psychonomic Society, St. Louis, 18-21 November 2010
• Speekenbrink, M. & Lagnado, D.A. (2010). Real time causal inference. Paper presented at the 1st joint meeting of the EPS and SEPEX, Granada, 15-17 April 2010
• Speekenbrink, M. (2010). Learning in a changing environment. Paper presented at the London Judgement and Decision-Making group, London, March 3, 2010.
• Speekenbrink, M. (2009). Cue learning in a changing environment. Paper presented at MathPsych 2009, Amsterdam, 2 August 2009.
• Speekenbrink, M. Chater, N. & Shanks, D.R. (2009) Adaptive tests for model discrimination. Paper presented at the International Meeting of the Psychometric Society, Cambridge, 24 July 2009.
• Speekenbrink, M & Shanks, D.R. (2008). Learning in a changing environment. Paper presented at the XXIX International Conference of Psychology, Berlin, 23 July 2008.
• Speekenbrink, M. (2006). A Lens Model Approach to the Study of Learning in Multiple Cue Tasks. Paper presented at the London Judgement and Decision-Making group, London, October 18, 2006.
• Speekenbrink, M. & Shanks, D.R. (2006). A Lens Model Approach to the Study of Learning in Multiple Cue Tasks. Paper presented at the 39th Annual Meeting of the Society for Mathematical Psychology, Vancouver, B.C., Canada, July 29 - August 1, 2006.
• Speekenbrink, M., Channon, S. & Shanks, D.R. (2006). Learning in Amnesia. Poster presented at the 28th Annual Meeting of the Cognitive Science Society. Vancouver, B.C., Canada, July 26-29, 2006.
• Speekenbrink, M. & Shanks, D.R. (2006). Through the Looking Glass: A Dynamic Lens Model Approach to Learning in MCPL. Invited presentation at: The probabilistic mind: Prospects for rational models of cognition, London, U.K. 27 - 28 June 2006.
• Speekenbrink, M. (2003). A social validation model of group decisions. Paper presented at the 13th international IOPS conference, Free University, Amsterdam, 11-12 December 2003.
• Speekenbrink, M. (2003). A dynamic group decision model. Paper presented at the 13th International Meeting of the Psychometric Society IMPS-2003, Cagliari, Italy (Sardinia), 7-10 July 2003.
• Speekenbrink, M. (2002). A hypothesis testing game. Paper presented at the SFPML colloquium, University of Amsterdam, The Netherlands, 14 October 2002.
• Speekenbrink, M. (2001). The hierarchical model and model selection. Paper presented at the International Meeting of the Psychometric Society IMPS-2001, Osaka, Japan, 15-19 July 2001.
• Speekenbrink, M. (2000). Conditions for deriving scientific consensus. Paper presented at the 10th International IOPS conference, University of Leuven, Belgium, 14-15 December 2000.