Multinomial Processing Tree Research Articles

On aggregation invariance of multinomial processing tree models

Multinomial processing tree (MPT) models are prominent and frequently used tools to model and measure cognitive processes underlying responses in many experimental paradigms. Although MPT models typically refer to cognitive processes within single individuals, they have often been applied to group data aggregated across individuals. We investigate the conditions under which MPT analyses of aggregate data make sense. After introducing the notions of structural and empirical aggregation invariance of MPT models, we show that any MPT model that holds at the level of single individuals must also hold at the aggregate level when it is both structurally and empirically aggregation invariant. Moreover, group-level parameters of aggregation-invariant MPT models are equivalent to the expected values (i.e., means) of the corresponding individual parameters. To investigate the robustness of MPT results for aggregate data when one or both invariance conditions are violated, we additionally performed a series of simulation studies, systematically manipulating (1) the sample sizes in different trees of the model, (2) model parameterization, (3) means and variances of crucial model parameters, and (4) their correlations with other parameters of the respective MPT model. Overall, our results show that MPT parameter estimates based on aggregate data are trustworthy under rather general conditions, provided that a few preconditions are met.

Social decision making in binge drinking: An exploration through moral dilemmas.

The continuum hypothesis proposes that binge drinking and severe alcohol use disorder (SAUD) share qualitatively similar cognitive and emotional impairments. In SAUD, these deficits have a demonstrated impact on social decision making, resulting in a utilitarian bias. Namely, when confronted with moral dilemmas, patients with SAUD tend to focus on the consequences of their actions rather than on social norms. However, social decision-making abilities remain unexplored in binge drinking. We offered the first insights on the generalization of the continuum hypothesis to social decision making, through a multinomial processing tree model applied to moral dilemmas, the "CNI model" of moral decision making. We compared 35 binge drinkers (20 females) and 36 light drinkers (21 females) on a battery of 48 moral dilemmas involving interpersonal relations from the CNI model, through multinomial modeling analyses. In each dilemma, participants were asked if they would perform the described action, generating individual scores for sensitivity to consequences, sensitivity to norms, and inaction tendency. The statistical model related to the CNI approach fits the data well. Binge drinkers and controls did not differ regarding their sensitivity to consequences nor their sensitivity to moral norms, and both groups displayed an equal inaction tendency in response to moral dilemmas. We provided insights to better understand the specific (socio)cognitive domains impaired in subclinical populations with alcohol use disorder. We showed preserved social decision making in binge drinking, which suggests that the continuum hypothesis documented for classical neurocognitive functions does not extend to complex social abilities. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Evaluating the robustness of parameter estimates in cognitive models: A meta-analytic review of multinomial processing tree models across the multiverse of estimation methods.

Researchers have become increasingly aware that data-analysis decisions affect results. Here, we examine this issue systematically for multinomial processing tree (MPT) models, a popular class of cognitive models for categorical data. Specifically, we examine the robustness of MPT model parameter estimates that arise from two important decisions: the level of data aggregation (complete-pooling, no-pooling, or partial-pooling) and the statistical framework (frequentist or Bayesian). These decisions span a multiverse of estimation methods. We synthesized the data from 13,956 participants (164 published data sets) with a meta-analytic strategy and analyzed the magnitude of divergence between estimation methods for the parameters of nine popular MPT models in psychology (e.g., process-dissociation, source monitoring). We further examined moderators as potential sources of divergence. We found that the absolute divergence between estimation methods was small on average (<.04; with MPT parameters ranging between 0 and 1); in some cases, however, divergence amounted to nearly the maximum possible range (.97). Divergence was partly explained by few moderators (e.g., the specific MPT model parameter, uncertainty in parameter estimation), but not by other plausible candidate moderators (e.g., parameter trade-offs, parameter correlations) or their interactions. Partial-pooling methods showed the smallest divergence within and across levels of pooling and thus seem to be an appropriate default method. Using MPT models as an example, we show how transparency and robustness can be increased in the field of cognitive modeling. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Evidence Against Effects of Cultural Group and Prior Knowledge on Feature Binding in Working Memory.

Feature binding is the process of integrating features, such as colour and shape, into object representations. A persistent question in the literature concerning whether feature binding is an automatic or resource-demanding process may depend on unitisation, that is, whether the to-be-bound information is intrinsic (belonging to) or extrinsic (contextual). Given extensive evidence showing that Easterners may process information more holistically than Westerners, such cultural differences may be useful to understand the fundamental processes of feature binding in visual working memory (WM). Accordingly, we recruited British and Chinese participants to complete a visual WM task wherein to-be-remembered colours were integrated within (i.e., intrinsic binding) or as backgrounds (i.e., extrinsic binding) of to-be-remembered shapes (Experiments 1 and 2). Experiment 2 further investigated the role of prior knowledge in long-term memory to facilitate feature binding in WM. During retrieval, participants decided among three probes: a target, a lure (i.e., recombination of the presented features), and a new colour/shape. Hierarchical Bayesian multinomial processing tree models were fit to the data to estimate parameters representing binding and item memory. The current results suggest that intrinsic and extrinsic binding memory are similar between the two cultural groups, with no prior knowledge benefits for either intrinsic or extrinsic binding for either cultural group. This result conflicts with the Analytic and Holistic framework and suggests that there are no cultural differences or prior knowledge benefits in feature binding.

Better source memory for remembered to-be-forgotten items than for remembered to-be-remembered items.

When participants study items one-by-one and are directed to either remember or forget the respective item directly after its presentation, retention of to-be-forgotten items is regularly worse than of to-be-remembered items. We tested whether this directed forgetting effect which is regularly observed for item memory generalizes to source memory. In three experiments, participants studied items in two different source colors (N = 101) or at two different source locations (N = 64; N = 81). Sources were manipulated orthogonally to item type (remember vs. forget). At test, we asked participants to recognize all studied items and also to identify their source. We used a multinomial processing tree model to disentangle item memory, source memory, and guessing. In all three experiments, we replicated the directed forgetting effect in item memory. Source memory for to-be-forgotten items that were recognized despite the intention to forget, however, tended to be even better than source memory for to-be-remembered items that were recognized. These results suggest that the directed forgetting effect does not simply translate from item to source memory. Rather source memory seems to be disproportionally increased in to-be-forgotten items that are remembered. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Context matters: The influence of the social situation on source memory

AbstractPrevious research has demonstrated that people remember negative reputational information particularly well. However, most of these experiments manipulated the type of information associated with each face, rather than manipulating the circumstances under which people learn this information. The present experiment examines the effect of the social situation on memory for social-exchange relevant information. Faces were paired with descriptions of cheating, trustworthy, or neutral behavior. In addition, the importance of the social situation was manipulated: Participants had either to decide if they would want to work with the described person on a student project (socially relevant scenario) or if they would want to ask this person what time it is while waiting at an airport (socially irrelevant scenario). A multinomial processing tree model was used to measure old–new item discrimination and source memory. Only in the socially relevant scenario a source memory advantage for cheaters was found.

Hierarchical multinomial processing tree models for meta-analysis of diagnostic accuracy studies

Abstract Meta-analysis represents a widely accepted approach for evaluating the accuracy of diagnostic tools in clinical and psychological investigations. This article investigates the applicability of multinomial tree models recently suggested in the literature under a fixed-effects formulation for assessing the accuracy of binary classification tools, where the study specific disease prevalences are taken into account. The model proposed in this article extends previous results to a hierarchical structure accounting for the variability between the studies included in the meta-analysis. Interestingly, by exploiting the parameter separability of the complete likelihood function, the resulting hierarchical multinomial tree model is shown to coincide, in its interest parameter component, with the well-known bivariate random-effects model under an exact within-study distribution for the number of true positives and true negatives subjects. The proposal is in line with a latent-trait approach, where inference follows a frequentist point of view. The applicability of the proposed model and its performance with respect to its fixed-effects counterpart and to the approximate bivariate random-effects model based on normality assumptions commonly used in the literature is evaluated in a series of simulation studies. Methods are applied to a real meta-analysis about the accuracy of the confusion assessment method as delirium screening tool.

I Won't Forget to Do It If It's Important: A Multinomial Processing Tree Analysis of Social Importance and Monetary Reward on Event-Based Prospective Memory.

While previous research has suggested that prospective memory may be enhanced by providing a social motive (i.e., social importance) or by promising a monetary reward for successful performance, to the best of our knowledge, the underlying mechanisms responsible for these effects are still largely unexplored. In a sample of 96 younger adults, this study investigated how social importance and promising a monetary reward influence the prospective component and the retrospective component of event-based prospective memory separately, with the application of a multinomial modeling approach. Results revealed enhanced prospective memory performance for all importance conditions compared to a standard condition. This improvement was characterized by an increased allocation of resource-demanding attentional processes in performing the prospective memory task at the expense of the ongoing task without an increase in the perceived importance of the prospective memory task. The model-based analyses showed that the beneficial effects of importance arise from an increased engagement of the prospective component, leaving the estimates for the retrospective component unaffected.

Refining Bayesian hierarchical MPT modeling: Integrating prior knowledge and ordinal expectations

Multinomial processing tree (MPT) models are a broad class of statistical models used to test sophisticated psychological theories. The research questions derived from these theories often go beyond simple condition effects on parameters and involve ordinal expectations (e.g., the same-direction effect on the memory parameter is stronger in one experimental condition than another) or disordinal expectations (e.g., the effect reverses in one experimental condition). Here, we argue that by refining common modeling practices, Bayesian hierarchical models are well suited to estimate and test these expectations. Concretely, we show that the default priors proposed in the literature lead to nonsensical predictions for individuals and the population distribution, leading to problems not only in model comparison but also in parameter estimation. Rather than relying on these priors, we argue that MPT modelers should determine priors that are consistent with their theoretical knowledge. In addition, we demonstrate how Bayesian model comparison may be used to test ordinal and disordinal interactions by means of Bayes factors. We apply the techniques discussed to empirical data from Bell et al. Journal of Experimental Psychology: Learning, Memory, and Cognition, 41, 456–472 (2015).

RT-MPTs: Process models for response-time distributions with diffusion-model kernels

We propose an extension of the widely used class of multinomial processing tree models by incorporating response times via diffusion-model kernels. Multinomial processing tree models are models of categorical data in terms of a number of cognitive and guessing processes estimating the probabilities with which each process outcome occurs. The new method allows one to estimate completion times of each process along with outcome probability and thereby provides process-oriented accounts of accuracy and latency data in all domains in which multinomial processing tree models have been applied. Furthermore, the new models are implemented hierarchically so that individual differences are explicitly accounted for and do not bias the population-level estimates. The new approach overcomes a number of shortcomings of previous extensions of multinomial models to incorporate response times. We evaluate the new method’s performance via a recovery study and simulation-based calibration. The method allows one to test hypotheses about processing architecture, and it provides an extension of traditional diffusion model analyses where multinomial models have been proposed for the modeled paradigm. We illustrate these and other benefits of the new model class using five existing data sets from recognition memory.

Cognitive effort assessment through pupillary responses: Insights from multinomial processing tree modeling and neural interconnections

The pupillary responses of humans exhibit variations in size, which are mediated by optic and oculomotor cranial nerves. Due to their sensitivity and high resolution of pupillary responses, they are used for a long time as measurement metrics of cognitive effort. Investigating the extent of cognitive effort required during tasks of varying difficulty is crucial for understanding the neural interconnections underlying these pupillary responses. This study aims to assess human cognitive efforts involved in visually presented cognitive tasks using the multinomial processing tree (MPT) model, an analytical tool that disentangles and predicts distinct cognitive processes, resulting in changes in pupil diameter. To achieve this, a pupillary response dataset was collected during mental multiplication (MM) tasks and visual stimuli presentations as cognitive tasks. MPT model describes observed response frequencies across various response categories and determines the transition probabilities from one latent state to the next. The expectation maximization (EM) algorithm is employed with MPT model to estimate parameter values based on response frequency within each category. Both group-level and individual subject-to-subject comparisons are conducted to estimate cognitive effort. The results reveal that in the group comparison and with respect to task difficulty level, that subject’s knowledge on MM task influences the successfully solve the problem. Regarding individual analysis, no significant differences are observed in parameters related to correct recall, problem-solving ability, and time constraint compliance. However, some significant differences are found in parameters associated with the perceived difficulty level and ability to recall the correct answers. MPT model combined with EM algorithm constitutes a probabilistic model that enhances pupillary responses identification related to the cognitive effort. Potential applications of this model include disease diagnostics based on parameter values and identification of neural pathways that are involved in the pupillary response and subject’s cognitive effort. Furthermore, efforts are underway to connect this psychological model with an artificial neural network.

Stimulus-triggered task conflict affects task-selection errors in task switching: A Bayesian multinomial processing tree modeling approach.

In the present study, we used a modeling approach for measuring task conflict in task switching, assessing the probability of selecting the correct task via multinomial processing tree (MPT) modeling. With this method, task conflict and response conflict can be independently assessed as the probability of selecting the correct task and the probability of selecting the correct response within a given task, respectively. These probabilities can be estimated on the basis of response accuracy in the different experimental conditions. In two task-switching experiments, we used bivalent stimuli and manipulated irrelevant-task difficulty by varying the saliency of the stimulus feature belonging to the irrelevant task. The more salient the task-irrelevant stimulus feature, the more salient the irrelevant task, leading to more task conflict. Consistent with this assumption, we observed that task conflict, but not response conflict, was larger when the task-irrelevant stimulus feature was made more salient. Furthermore, both task conflict and response conflict were larger when the task switched than when the task was repeated. On a methodological level, the present results demonstrate that MPT modeling is a useful approach for measuring task conflict in task switching and for dissociating it from within-task response conflict. Furthermore, the present results inform theories of task switching by showing that the task-irrelevant feature tends to activate the irrelevant task set rather than being associated with a specific response option via a direct stimulus-response route. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

People punish defection, not failures to conform to the majority

Do people punish others for defecting or for failing to conform to the majority? In two experiments, we manipulated whether the participants’ partners cooperated or defected in the majority of the trials of a Prisoner’s Dilemma game. The effects of this base-rate manipulation on cooperation and punishment were assessed using a multinomial processing tree model. High compared to low cooperation rates of the partners increased participants’ cooperation. When participants’ cooperation was not enforced through partner punishment, the participants’ cooperation was closely aligned to the cooperation rates of the partners. Moral punishment of defection increased when cooperation rates were high compared to when defection rates were high. However, antisocial punishment of cooperation when defection rates were high was much less likely than moral punishment of defection when cooperation rates were high. In addition, antisocial punishment was increased when cooperation rates were high compared to when defection rates were high. The latter two results contradict the assumption that people punish conformity-violating behavior regardless of whether the behavior supports or disrupts cooperation. Punishment is thus sensitive to the rates of cooperation and defection but, overall, the results are inconsistent with the idea that punishment primarily, let alone exclusively, serves to enforce conformity with the majority.

Benefits of multinomial processing tree models with discrete and continuous variables in memory research: an alternative modeling proposal to Juola et al. (2019).

Signal detection theory (SDT) and two-high threshold models (2HT) are often used to analyze accuracy data in recognition memory paradigms. However, when reaction times (RTs) and/or confidence levels (CLs) are also measured, they usually are analyzed separately or not at all as dependent variables (DVs). We propose a new approach to include these variables based on multinomial processing tree models for discrete and continuous variables (MPT-DC) with the aim to compare fits of SDT and 2HT models. Using Juola et al.'s (2019, Memory & Cognition, 47[4], 855-876) data we have found that including CLs and RTs reduces the standard errors of parameter estimates and accounts for interactions among accuracy, CLs, and RTs that classical versions of SDT and 2HT models do not. In addition, according to the simulations, there is an increase in the proportion of correct model selections when relevant DV are included. We highlight the methodological and substantive advantages of MPT-DC in the disentanglement of contributing processes in recognition memory.

Representation of shared surface information and false memory for abstract versus concrete pictures in the conjoint recognition paradigm.

An effective factor by which false memories can arise is relatedness which includes not only semantic associations but also perceptual resemblance. This issue raises questions about how patterns of perceptual features are represented in memory and how they relate to semantic representations. In five experiments, we investigated the memory processes underlying the false recognition of perceptually or semantically related pictures from the perspective of fuzzy trace theory. Multinomial processing tree model analyses for the conjoint recognition paradigm showed that the parameter representing gist trace retrieval not only contributes to false acceptances of semantically related pictures, but also underlies the false recognition of non-semantically related abstract shapes. These results challenged the hypothesis that the false recognition of non-semantically related distractors is solely due to interference with the verbatim suppression process. These experiments also showed that adding a surface feature (colour) to the category exemplars increases false recognition of related distractors by enhancing the contribution of the familiarity process, but only for pictures of real objects. Comparisons between experiments showed that different variants of the conjoint recognition model, used to analyse the effects of the same experimental manipulation, can lead to partially different conclusions.

Predicting clinical decline over 36 months with a memory‐based digital biomarker

AbstractBackgroundWord‐list recall tests are routinely employed in clinical practice for dementia screening. Process scoring and latent modeling of these tests, where underlying neurocognitive mechanisms are exploited, have shown potential to enhance test accuracy without requiring test redesign. We examined Embic’s digital biomarker M, a measure of recall derived from multinomial processing trees and hierarchical Bayesian computational methods, as a predictor of conversion to mild cognitive impairment (MCI) and Alzheimer’s disease (AD), and of increase in Clinical Dementia Rating (CDR) score.MethodsSecondary data analyses were carried out from ADNI data. M was computed from Rey’s AVLT data, and compared to standard AVLT clinical metrics (total and delayed recall). We conducted logistic regression analyses with diagnosis at 36 months (normal vs. MCI + AD) as outcome; all 169 participants were classified as normal at baseline, and 13 of these progressed to a clinical diagnosis (see Table 1). We controlled for baseline age, gender and years of education; and ran separate analyses with other metrics to avoid multicollinearity. We then repeated the same analytical framework with CDR scores (stable at 0 vs. increased to 0.5 or 1) at 36 months as outcome: stable participants were 141, and 24 increased (see Table 2). Covariates and predictors were the same as in the previous analysis.Results M predicted conversion to MCI or AD after three years (FDR‐corrected p value = 0.048), while neither AVLT total nor delayed recall did. Performance metrics showed that M generated 92% accuracy, or probability of yielding a correct classification, and 100% specificity, or no false positives (Figure 1). Analogously, M, but not standard AVLT metrics, also predicted a CDR increase after three years (FDR‐corrected p value = 0.003), with high accuracy (84%), specificity (96%), and precision (80%), or positive predictive value (Figure 2). Sensitivity analyses with ordinal regression broadly confirmed these findings.ConclusionsThese results indicate Embic’s digital biomarker M outperforms some of AVLT’s traditional metrics in identifying individuals who will progress to MCI and AD after three years, from a healthy baseline.

Remember walking in their shoes? The relation of self-referential source memory and emotion recognition

ABSTRACT Deficits in the ability to read the emotions of others have been demonstrated in mental disorders, such as dissociation and schizophrenia, which involve a distorted sense of self. This study examined whether weakened self-referential source memory, being unable to remember whether a piece of information has been processed with reference to oneself, is linked to ineffective emotion recognition. In two samples from a college and community, we quantified the participants’ ability to remember the self-generated versus non-self-generated origins of sentences they had previously read or partially generated. We also measured their ability to read others’ emotions accurately when viewing photos of people in affect-charged situations. Multinomial processing tree modelling was applied to obtain a measure of self-referential source memory that was not biased by non-mnemonic factors. Our first experiment with college participants revealed a positive correlation between correctly remembering the origins of sentences and accurately recognising the emotions of others. This correlation was successfully replicated in the second experiment with community participants. The current study offers evidence of a link between self-referential source memory and emotion recognition.

Attentional focusing and filtering in multisensory categorization.

Selective attention refers to the ability to focus on goal-relevant information while filtering out irrelevant information. In a multisensory context, how do people selectively attend to multiple inputs when making categorical decisions? Here, we examined the role of selective attention in cross-modal categorization in two experiments. In a speed categorization task, participants were asked to attend to visual or auditory targets and categorize them while ignoring other irrelevant stimuli. A response-time extended multinomial processing tree (RT-MPT) model was implemented to estimate the contribution of attentional focusing on task-relevant information and attentional filtering of distractors. The results indicated that the role of selective attention was modality-specific, with differences found in attentional focusing and filtering between visual and auditory modalities. Visual information could be focused on or filtered out more effectively, whereas auditory information was more difficult to filter out, causing greater interference with task-relevant performance. The findings suggest that selective attention plays a critical and differential role across modalities, which provides a novel and promising approach to understanding multisensory processing and attentional focusing and filtering mechanisms of categorical decision-making.

Cognitive load reduces context recollection for true sentences

ABSTRACT The dual-recollection model has been successfully applied in research on memory for truth and falsity, suggesting that “true” feedback is better recollected than “false” feedback. We used this approach to test whether the Cartesian or the Spinozan model would be a better framework to describe processes underlying memory for truth and falsity. Our sample consisted of 108 students, who performed the conjoint recognition test under: no-load, refreshing-interference or rehearsal-interference conditions. We found no difference in the rate of falsely attributing “true” label to false sentences than vice versa under cognitive load, which supported the Cartesian model. Multinomial processing tree model analyses confirmed better context memory for true than false sentences in no-load condition. Cognitive load mostly influenced context recollection for true sentences, with nonsignificant effects on context recollection for false sentences, which contradicts the Spinozan model.

How to develop, test, and extend multinomial processing tree models: A tutorial.

Many psychological theories assume that observable responses are determined by multiple latent processes. Multinomial processing tree (MPT) models are a class of cognitive models for discrete responses that allow researchers to disentangle and measure such processes. Before applying MPT models to specific psychological theories, it is necessary to tailor a model to specific experimental designs. In this tutorial, we explain how to develop, fit, and test MPT models using the classical pair-clustering model as a running example. The first part covers the required data structures, model equations, identifiability, model validation, maximum-likelihood estimation, hypothesis tests, and power analyses using the software multiTree. The second part introduces hierarchical MPT modeling which allows researchers to account for individual differences and to estimate the correlations of latent processes among each other and with additional covariates using the TreeBUGS package in R. All examples including data and annotated analysis scripts are provided at the Open Science Framework (https://osf.io/24pbm/). (PsycInfo Database Record (c) 2023 APA, all rights reserved).