Dimensionality and optimal combination of autonomic fear-conditioning measures in humans.

Fear conditioning, also termed threat conditioning, is a commonly used learning model with clinical relevance. Quantification of threat conditioning in humans often relies on conditioned autonomic responses such as skin conductance responses (SCR), pupil size responses (PSR), heart period responses (HPR), or respiration amplitude responses (RAR), which are usually analyzed separately. Here, we investigate whether inter-individual variability in differential conditioned responses, averaged across acquisition, exhibits a multi-dimensional structure, and the extent to which their linear combination could enhance the precision of inference on whether threat conditioning has occurred. In a mega-analytic approach, we re-analyze nine data sets including 256 individuals, acquired by the group of the last author, using standard routines in the framework of psychophysiological modeling (PsPM). Our analysis revealed systematic differences in effect size between measures across datasets, but no evidence for a multidimensional structure across various combinations of measures. We derive the statistically optimal weights for combining the four measures and subsets thereof, and we provide out-of-sample performance metrics for these weights, accompanied by bias-corrected confidence intervals. We show that to achieve the same statistical power, combining measures allows for a relevant reduction in sample size, which in a common scenario amounts to roughly 24%. To summarize, we demonstrate a one-dimensional structure of threat conditioning measures, systematic differences in effect size between measures, and provide weights for their optimal linear combination in terms of maximal retrodictive validity.

Saccadic scanpath length: an index for human threat conditioning.

Threat-conditioned cues are thought to capture overt attention in a bottom-up process. Quantification of this phenomenon typically relies on cue competition paradigms. Here, we sought to exploit gaze patterns during exclusive presentation of a visual conditioned stimulus, in order to quantify human threat conditioning. To this end, we capitalized on a summary statistic of visual search during CS presentation, scanpath length. During a simple delayed threat conditioning paradigm with full-screen monochrome conditioned stimuli (CS), we observed shorter scanpath length during CS+ compared to CS- presentation. Retrodictive validity, i.e., effect size to distinguish CS+ and CS-, was maximized by considering a 2-s time window before US onset. Taking into account the shape of the scan speed response resulted in similar retrodictive validity. The mechanism underlying shorter scanpath length appeared to be longer fixation duration and more fixation on the screen center during CS+ relative to CS- presentation. These findings were replicated in a second experiment with similar setup, and further confirmed in a third experiment using full-screen patterns as CS. This experiment included an extinction session during which scanpath differences appeared to extinguish. In a fourth experiment with auditory CS and instruction to fixate screen center, no scanpath length differences were observed. In conclusion, our study suggests scanpath length as a visual search summary statistic, which may be used as complementary measure to quantify threat conditioning with retrodictive validity similar to that of skin conductance responses.

Calibrating the experimental measurement of psychological attributes.

Behavioural researchers often seek to experimentally manipulate, measure and analyse latent psychological attributes, such as memory, confidence or attention. The best measurement strategy is often difficult to intuit. Classical psychometric theory, mostly focused on individual differences in stable attributes, offers little guidance. Hence, measurement methods in experimental research are often based on tradition and differ between communities. Here we propose a criterion, which we term 'retrodictive validity', that provides a relative numerical estimate of the accuracy of any given measurement approach. It is determined by performing calibration experiments to manipulate a latent attribute and assessing the correlation between intended and measured attribute values. Our approach facilitates optimising measurement strategies and quantifying uncertainty in the measurement. Thus, it allows power analyses to define minimally required sample sizes. Taken together, our approach provides a metrological perspective on measurement practice in experimental research that complements classical psychometrics.

Representation of probabilistic outcomes during risky decision-making.

Goal-directed behaviour requires prospectively retrieving and evaluating multiple possible action outcomes. While a plethora of studies suggested sequential retrieval for deterministic choice outcomes, it remains unclear whether this is also the case when integrating multiple probabilistic outcomes of the same action. We address this question by capitalising on magnetoencephalography (MEG) in humans who made choices in a risky foraging task. We train classifiers to distinguish MEG field patterns during presentation of two probabilistic outcomes (reward, loss), and then apply these to decode such patterns during deliberation. First, decoded outcome representations have a temporal structure, suggesting alternating retrieval of the outcomes. Moreover, the probability that one or the other outcome is being represented depends on loss magnitude, but not on loss probability, and it predicts the chosen action. In summary, we demonstrate decodable outcome representations during probabilistic decision-making, which are sequentially structured, depend on task features, and predict subsequent action.

Psychophysiological modelling and the measurement of fear conditioning.

Quantification of fear conditioning is paramount to many clinical and translational studies on aversive learning. Various measures of fear conditioning co-exist, including different observables and different methods of pre-processing. Here, we first argue that low measurement error is a rational desideratum for any measurement technique. We then show that measurement error can be approximated in benchmark experiments by how closely intended fear memory relates to measured fear memory, a quantity that we term retrodictive validity. From this perspective, we discuss different approaches commonly used to quantify fear conditioning. One of these is psychophysiological modelling (PsPM). This builds on a measurement model that describes how a psychological variable, such as fear memory, influences a physiological measure. This model is statistically inverted to estimate the most likely value of the psychological variable, given the measured data. We review existing PsPMs for skin conductance, pupil size, heart period, respiration, and startle eye-blink. We illustrate the benefit of PsPMs in terms of retrodictive validity and translate this into sample size required to achieve a desired level of statistical power. This sample size can differ up to a factor of three between different observables, and between the best, and the current standard, data pre-processing methods.

Computational optimization of associative learning experiments.

With computational biology striving to provide more accurate theoretical accounts of biological systems, use of increasingly complex computational models seems inevitable. However, this trend engenders a challenge of optimal experimental design: due to the flexibility of complex models, it is difficult to intuitively design experiments that will efficiently expose differences between candidate models or allow accurate estimation of their parameters. This challenge is well exemplified in associative learning research. Associative learning theory has a rich tradition of computational modeling, resulting in a growing space of increasingly complex models, which in turn renders manual design of informative experiments difficult. Here we propose a novel method for computational optimization of associative learning experiments. We first formalize associative learning experiments using a low number of tunable design variables, to make optimization tractable. Next, we combine simulation-based Bayesian experimental design with Bayesian optimization to arrive at a flexible method of tuning design variables. Finally, we validate the proposed method through extensive simulations covering both the objectives of accurate parameter estimation and model selection. The validation results show that computationally optimized experimental designs have the potential to substantially improve upon manual designs drawn from the literature, even when prior information guiding the optimization is scarce. Computational optimization of experiments may help address recent concerns over reproducibility by increasing the expected utility of studies, and it may even incentivize practices such as study pre-registration, since optimization requires a pre-specified analysis plan. Moreover, design optimization has the potential not only to improve basic research in domains such as associative learning, but also to play an important role in translational research. For example, design of behavioral and physiological diagnostic tests in the nascent field of computational psychiatry could benefit from an optimization-based approach, similar to the one presented here.

Psychophysiological modeling: Current state and future directions.

Psychologists often use peripheral physiological measures to infer a psychological variable. It is desirable to make this inverse inference in the most precise way, ideally standardized across research laboratories. In recent years, psychophysiological modeling has emerged as a method that rests on statistical techniques to invert mathematically formulated forward models (psychophysiological models, PsPMs). These PsPMs are based on psychophysiological knowledge and optimized with respect to the precision of the inference. Building on established experimental manipulations, known to create different values of a psychological variable, they can be benchmarked in terms of their sensitivity (e.g., effect size) to recover these values-we have termed this predictive validity. In this review, we introduce the problem of inverse inference and psychophysiological modeling as a solution. We present background and application for all peripheral measures for which PsPMs have been developed: skin conductance, heart period, respiratory measures, pupil size, and startle eyeblink. Many of these PsPMs are task invariant, implemented in open-source software, and can be used off the shelf for a wide range of experiments. Psychophysiological modeling thus appears as a potentially powerful method to infer psychological variables.

Beyond p-values in the evaluation of brain-computer interfaces: A Bayesian estimation approach.

BACKGROUND: To statistically evaluate the performance of brain-computer interfaces (BCIs), researchers usually rely on null hypothesis significance testing (NHST), i.e. p-values. However, over-reliance on NHST is often identified as one of the causes of the recent reproducibility crisis in psychology and neuroscience. NEW METHOD: In this paper we propose Bayesian estimation as an alternative to NHST in the analysis of BCI performance data. For the three most common experimental designs in BCI research - which would usually be analyzed using a t-test, a linear regression, or an ANOVA - we develop hierarchical models and estimate their parameters using Bayesian inference. Furthermore, we show that the described models are special cases of the hierarchical generalized linear model (HGLM), which we propose as a general framework for the analysis of BCI performance. RESULTS: We demonstrate the effectiveness of the proposed models on three real datasets and show how the results obtained with Bayesian estimation can give a nuanced insight into BCI performance data. Additionally, we provide all the data and code necessary to reproduce the presented results. COMPARISON WITH EXISTING METHOD(S): Compared to NHST, Bayesian estimation with the HGLM allows more flexibility in the analysis of BCI performance data from nested experimental designs, and the obtained results have a more straightforward interpretation. CONCLUSIONS: Besides gains in flexibility and interpretability, a wider adoption of the Bayesian estimation approach in BCI studies could bring about greater transparency in data analysis, allow accumulation of knowledge across studies, and reduce questionable practices such as "p-hacking".

Asynchronous detection of kinesthetic attention during mobilization of lower limbs using EEG measurements.

OBJECTIVE: Attention is known to modulate the plasticity of the motor cortex, and plasticity is crucial for recovery in motor rehabilitation. This study addresses the possibility of using an EEG-based brain-computer interface (BCI) to detect kinesthetic attention to movement. APPROACH: A novel experiment emulating physical rehabilitation was designed to study kinesthetic attention. The protocol involved continuous mobilization of lower limbs during which participants reported levels of attention to movement-from focused kinesthetic attention to mind wandering. For this protocol an asynchronous BCI detector of kinesthetic attention and deliberate mind wandering was designed. MAIN RESULTS: EEG analysis showed significant differences in theta, alpha, and beta bands, related to the attentional state. These changes were further pinpointed to bands relative to the frequency of the individual alpha peak. The accuracy of the designed BCI ranged between 60.8% and 68.4% (significantly above chance level), depending on the used analysis window length, i.e. acceptable detection delay. SIGNIFICANCE: This study shows it is possible to use self-reporting to study attention-related changes in EEG during continuous mobilization. Such a protocol is used to develop an asynchronous BCI detector of kinesthetic attention, with potential applications to motor rehabilitation.

Sample size determination for BCI studies: How many subjects and trials?

Sample sizes and, consequently, statistical power have a large influence on the reliability of statistical results, but they are often neglected when planning and reporting studies of brain-computer interfaces (BCIs). This may be in part due to the limitations of classical power calculations, which do not apply to nested experimental designs, that are usually employed in BCI research. In this paper we introduce the methodology of simulation-based sample size determination (SSD) for the planning of BCI studies. We show how the proposed method can be used to determine the necessary number of subjects and trials to obtain a precise estimate of BCI accuracy, when the cost of sampling needs to be constrained by a budget. Furthermore, the method is fully general and can be applied in different experimental designs and in different statistical frameworks.