Single Stimulus Dimension Research Articles

A Semi-parametric Model for Decision Making in High-Dimensional Sensory Discrimination Tasks

Psychometric functions typically characterize binary sensory decisions along a single stimulus dimension. However, real-life sensory tasks vary along a greater variety of dimensions (e.g. color, contrast and luminance for visual stimuli). Approaches to characterizing high-dimensional sensory spaces either require strong parametric assumptions about these additional contextual dimensions, or fail to leverage known properties of classical psychometric curves. We overcome both limitations by introducing a semi-parametric model of sensory discrimination that applies traditional psychophysical models along a stimulus intensity dimension, but puts Gaussian process (GP) priors on the parameters of these models with respect to the remaining dimensions. By combining the flexibility of the GP with the deep literature on parametric psychophysics, our semi-parametric models achieve good performance with much less data than baselines on both synthetic and real-world, high-dimensional psychophysics datasets. We additionally show strong performance in a Bayesian active learning setting, and present a novel active learning paradigm for the semi-parametric model.

The perceptual categorization of multidimensional stimuli is hierarchically organized

As we interact with our surroundings, we encounter the same or similar objects from different perspectives and are compelled to generalize. For example, despite their variety we recognize dog barks as a distinct sound class. While we have some understanding of generalization along a single stimulus dimension (frequency, color), natural stimuli are identifiable by a combination of dimensions. Measuring their interaction is essential to understand perception. Using a 2-dimension discrimination task for mice and frequency or amplitude modulated sounds, we tested untrained generalization across pairs of auditory dimensions in an automatized behavioral paradigm. We uncovered a perceptual hierarchy over the tested dimensions that was dominated by the sound's spectral composition. Stimuli are thus not perceived as a whole, but as a combination of their features, each of which weights differently on the identification of the stimulus according to an established hierarchy, possibly paralleling their differential shaping of neuronal tuning.

Prelimbic cortex maintains attention to category-relevant information and flexibly updates category representations

Category learning groups stimuli according to similarity or function. This involves finding and attending to stimulus features that reliably inform category membership. Although many of the neural mechanisms underlying categorization remain elusive, models of human category learning posit that prefrontal cortex plays a substantial role. Here, we investigated the role of the prelimbic cortex (PL) in rat visual category learning by administering excitotoxic lesions before category training and then evaluating the effects of the lesions with computational modeling. Using a touchscreen apparatus, rats (female and male) learned to categorize distributions of category stimuli that varied along two continuous dimensions. For some rats, categorizing the stimuli encouraged selective attention towards a single stimulus dimension (i.e., 1D tasks). For other rats, categorizing the stimuli required divided attention towards both stimulus dimensions (i.e., 2D tasks). Testing sessions then examined generalization to novel exemplars. PL lesions impaired learning and generalization for the 1D tasks, but not the 2D tasks. Then, a neural network was fit to the behavioral data to examine how the lesions affected categorization. The results suggest that the PL facilitates category learning by maintaining attention to category-relevant information and updating category representations.

Symbolism overshadows the effect of physical size in supra-second temporal illusions.

The perception of quantities has been suggested to rely on shared, magnitude-based representational systems that preserve metric properties. As such, different quantifiable dimensions that can characterize any given stimulus (e.g., size, speed, or numerosity) have been shown to modulate the perceived duration of these stimuli-a finding that has been attributed to cross-modal interaction among the quantity representations. However, these results are typically based on the isolated effects of a single stimulus dimension, leaving their potential combined effects uncharted. In the present study we aimed to investigate the joint effects of numerical magnitude and physical size on perceived time. In four complementary experiments, participants categorized six durations as "short" or "long," which were presented through combinations of Hindu-Arabic numerals in three font sizes, as well as with simple shapes (rectangles) and unfamiliar symbols (Klingon letters), the sizes of which corresponded to the font sizes of the Hindu-Arabic numerals. Our results showed temporal underestimation for the smallest numeral in the set (3), with no effects of font size on perceived duration. The perceived durations were longest for the physically smallest geometric stimuli (i.e., a rectangle), and the font size of symbol-like stimuli (i.e., Klingon letters) was not found to have an effect on perceived time. Finally, presenting only one numeral (6) instead of the rectangle once again eliminated the relationship between physical size and perceived time, suggesting an overshadowing of physical-size-based influences on temporal choice behavior, presumably by perceived symbolism. Our results point at the complex nature of the interaction between different magnitude representations.

Wide sensory filters underlie performance in memory-based discrimination and generalization.

The way animals respond to a stimulus depends largely on an internal comparison between the current sensation and the memory of previous stimuli and outcomes. We know little about the accuracy with which the physical properties of the stimuli influence this type of memory-based discriminative decisions. Research has focused largely on discriminations between stimuli presented in quick succession, where animals can make relative inferences (same or different; higher or lower) from trial to trial. In the current study we used a memory-based task to explore how the stimulus’ physical properties, in this case tone frequency, affect auditory discrimination and generalization in mice. Mice performed ad libitum while living in groups in their home quarters. We found that the frequency distance between safe and conditioned sounds had a constraining effect on discrimination. As the safe-to-conditioned distance decreased across groups, performance deteriorated rapidly, even for frequency differences significantly larger than reported discrimination thresholds. Generalization width was influenced both by the physical distance and the previous experience of the mice, and was not accompanied by a decrease in sensory acuity. In conclusion, memory-based discriminations along a single stimulus dimension are inherently hard, reflecting a high overlap between the memory traces of the relevant stimuli. Memory-based discriminations rely therefore on wide sensory filters.

Animal Learning in a Multidimensional Discrimination Task as Explained by Dimension-Specific Allocation of Attention.

Reinforcement learning describes the process by which during a series of trial-and-error attempts, actions that culminate in reward are reinforced, becoming more likely to be chosen in similar circumstances. When decisions are based on sensory stimuli, an association is formed between the stimulus, the action and the reward. Computational, behavioral and neurobiological accounts of this process successfully explain simple learning of stimuli that differ in one aspect, or along a single stimulus dimension. However, when stimuli may vary across several dimensions, identifying which features are relevant for the reward is not trivial, and the underlying cognitive process is poorly understood. To study this we adapted an intra-dimensional/ extra-dimensional set-shifting paradigm to train rats on a multi-sensory discrimination task. In our setup, stimuli of different modalities (spatial, olfactory and visual) are combined into complex cues and manipulated independently. In each set, only a single stimulus dimension is relevant for reward. To distinguish between learning and decision-making we suggest a weighted attention model (WAM). Our model learns by assigning a separate learning rule for the values of features of each dimension (e.g., for each color), reinforced after every experience. Decisions are made by comparing weighted averages of the learnt values, factored by dimension specific weights. Based on the observed behavior of the rats we estimated the parameters of the WAM and demonstrated that it outperforms an alternative model, in which a learnt value is assigned to each combination of features. Estimated decision weights of the WAM reveal an experience-based bias in learning. In the first experimental set the weights associated with all dimensions were similar. The extra-dimensional shift rendered this dimension irrelevant. However, its decision weight remained high for the early learning stage in this last set, providing an explanation for the poor performance of the animals. Thus, estimated weights can be viewed as a possible way to quantify the experience-based bias.

Dopaminergic Genetic Polymorphisms Predict Rule-based Category Learning.

Dopaminergic genes play an important role in cognitive function. DRD2 and DARPP-32 dopamine receptor gene polymorphisms affect striatal dopamine binding potential, and the Val158Met single-nucleotide polymorphism of the COMT gene moderates dopamine availability in the pFC. Our study assesses the role of these gene polymorphisms on performance in two rule-based category learning tasks. Participants completed unidimensional and conjunctive rule-based tasks. In the unidimensional task, a rule along a single stimulus dimension can be used to distinguish category members. In contrast, a conjunctive rule utilizes a combination of two dimensions to distinguish category members. DRD2 C957T TT homozygotes outperformed C allele carriers on both tasks, and DARPP-32 AA homozygotes outperformed G allele carriers on both tasks. However, we found an interaction between COMT and task type where Met allele carriers outperformed Val homozygotes in the conjunctive rule task, but both groups performed equally well in the unidimensional task. Thus, striatal dopamine binding may play a critical role in both types of rule-based tasks, whereas prefrontal dopamine binding is important for learning more complex conjunctive rule tasks. Modeling results suggest that striatal dopaminergic genes influence selective attention processes whereas cortical genes mediate the ability to update complex rule representations.

Atypical integration of motion signals in Autism Spectrum Conditions.

Vision in Autism Spectrum Conditions (ASC) is characterized by enhanced perception of local elements, but impaired perception of global percepts. Deficits in coherent motion perception seem to support this characterization, but the roots and robustness of such deficits remain unclear. We aimed to investigate the dynamics of the perceptual decision-making network known to support coherent motion perception. In a series of forced-choice coherent motion perception tests, we parametrically varied a single stimulus dimension, viewing duration, to test whether the rate at which evidence is accumulated towards a global decision is atypical in ASC. 40 adult participants (20 ASC) performed a classic motion discrimination task, manually indicating the global direction of motion in a random-dot kinematogram across a range of coherence levels (2–75%) and stimulus-viewing durations (200–1500 ms). We report a deficit in global motion perception at short viewing durations in ASC. Critically, however, we found that increasing the amount of time over which motion signals could be integrated reduced the magnitude of the deficit, such that at the longest duration there was no difference between the ASC and control groups. Further, the deficit in motion integration at the shortest duration was significantly associated with the severity of autistic symptoms in our clinical population, and was independent from measures of intelligence. These results point to atypical integration of motion signals during the construction of a global percept in ASC. Based on the neural correlates of decision-making in global motion perception our findings suggest the global motion deficit observed in ASC could reflect a slower or more variable response from the primary motion area of the brain or longer accumulation of evidence towards a decision-bound in parietal areas.

Rich stimulus sampling for between-subjects designs improves model selection

The choice of stimulus values to test in any experiment is a critical component of good experimental design. This study examines the consequences of random and systematic sampling of data values for the identification of functional relationships in experimental settings. Using Monte Carlo simulation, uniform random sampling was compared with systematic sampling of two, three, four, or N equally spaced values along a single stimulus dimension. Selection of the correct generating function (a logistic or a linear model) was improved with each increase in the number of levels sampled, with N equally spaced values and random stimulus sampling performing similarly. These improvements came at a small cost in the precision of the parameter estimates for the generating function.

Pings the Body Electric (and Means It Too). Focus on “Interval Coding. I. Burst Interspike Intervals as Indicators of Stimulus Intensity” and “Interval Coding. II. Dendrite-Dependent Mechanisms”

Structure/function questions are at the forefront of research in biological sensory systems. These questions are very elegantly addressed by Oswald et al. and Doiron et al. in a pair of articles on signal representation in electric fish in this issue of Journal of Neurophysiology (p. 2731–2743 and

Characterizing rule-based category learning deficits in patients with Parkinson's disease

Parkinson's disease (PD) patients and normal controls were tested in three category learning experiments to determine if previously observed rule-based category learning impairments in PD patients were due to deficits in selective attention or working memory. In Experiment 1, optimal categorization required participants to base their decision on a single stimulus dimension and ignore irrelevant variation on another dimension, thus emphasizing selective attention processes. In Experiment 2, optimal categorization required participants to base their decision on both stimulus dimensions using a conjunction of unidimensional decisions. Thus, this task placed less emphasis on selective attention and more on working memory. In Experiment 3, optimal categorization again required participants to base their decision on both stimulus dimensions using a disjunction of two unidimensional decisions in which an additional verbal operation was needed, thereby placing even greater emphasis on working memory. Results indicated that PD patients were impaired in the unidimensional rule-based condition, but not the other two rule-based conditions. These results are consistent with previous studies that demonstrate that PD patients are impaired in learning rule-based categories when selective attention demands are greatest, whereas these patients are normal in learning rule-based tasks when working memory demands are emphasized. Overall, these findings help to delineate the conditions under which PD patients display rule-based category learning deficits.

An empirical test of signal detection theory as it applies to Batesian mimicry

Signal detection theory (SDT) has been repeatedly invoked to understand how palatable prey might gain an advantage by resembling unpalatable prey. Here we developed an experimental test of the theory in which we sequentially presented computer-generated Mimics (profitable to attack) and Models (unprofitable to attack) to human volunteers, and asked them to forage in a way that maximized their personal scores. Both the Mimics and Models exhibited normally distributed variation in a single stimulus dimension. When we varied the mean similarity of Mimics to Models, and the proportion of all prey items that were Mimics, our human predators made foraging decisions that were close to those predicted by SDT, including the adoption of a threshold in appearance beyond which prey items were unlikely to be attacked. The fit of predictions to observations was marginally closer when including the time taken to handle the two types of prey. When Mimics and Models were allowed to evolve in appearance subject to selection, the evolutionary trajectory fitted the predictions of SDT closely. While our system was not appropriate to test all predictions of recent SDT theory, it provides strong support for the SDT framework as it applies to Batesian mimicry.

The psychophysics of perceptual memory

Psychophysical studies of short-term memory for attributes or dimensions of the visual stimulus known to be important in early visual processing--spatial frequency, orientation, contrast, motion--identify an early perceptual memory system. The proposed system, which may be part of the Schacter-Tulving perceptual representation system (PRS), is located early in the visual processing stream, prior to the structural description system responsible for shape priming but beyond primary visual cortex (V1), and consists of a series of parallel special-purpose perceptual mechanisms with independent but limited processing resources, where each mechanism is devoted to the analysis of a single stimulus dimension and is coupled to a memory store. The experimental evidence for this hypothesis is reviewed.

The Concept of Generalization in Behaviour Therapy

Several writers have been interested in the extent to which behaviour therapy techniques are derived from the principles of learning. The present paper reviews this issue with respect to stimulus generalization and response generalization in behaviour therapy. Generalization has been a problem because many therapists have reported that newly learned skills and behaviours have not transferred outside the treatment situation to settings in the patients' or clients' life. Reviewing the literature on stimulus generalization it is concluded that in research on learning theory, stimulus generalization is shown to occur along a single stimulus dimension while in the behaviour therapy literature researchers often attempt to produce generalization across many stimulus dimensions simultaneously and as a result generalization may be poor. It is suggested that subsequent therapy and research should be designed so that treatment situations and generalization situations differ on as few stimulus dimensions as possible, thus maximizing the probability of generalization. With respect to response generalization it is felt that the concept is weak as an explanation for behaviour change and it may be more useful to consider that behaviours are functionally related, so that changes in behaviour A will produce changes in behaviour B or that covert responses may account for changes in untreated behaviours. Finally, a recommendation is made that therapists be more precise about the stimulus dimensions along which they wish behaviour to generalize.

Developmental trends in children's component selection

A new method is introduced for assessing children's component selection—i.e., the disposition to attend to a single feature of multifaceted stimuli. Eight-year-old children were found to exercise component selection to a lesser degree than 4-year-olds; while children at both age levels attended primarily to one stimulus component (shape), the older children showed a moderate amount of attention to a secondary redundant feature (color) as well. However, a comparable age difference in attention deployment was not observed when a single stimulus dimension (shape) was “relevant” in two variant tasks. These results imply a developmentally increasing ability to distinguish between conditions in which attending to redundant stimulus information can and cannot be useful. That this ability undergoes little further development beyond age 8 was suggested in a second experiment with 8- and 12-year-olds in which the three tasks produced relatively similar developmental trends in performance.

Cross-modal transfer of intensity discrimination by rats.

CROSS-MODAL transfer of the effects of discrimination training has frequently been obtained with humans1,2 but not lower animals. Several studies have shown that monkeys trained to differentiate geometrical forms by touch learn little3 or nothing4,5relevant to the visual solution of the same problem. Several factors may limit transfer in these conditions. Tactile and visual versions of what is ostensibly the same problem could emphasize dissimilar stimulus features so that animals attend to different properties when learning in the two modalities3. The marked change in experimental conditions from dark to light may also disrupt performance. These difficulties can be overcome by using as discriminanda diffuse stimuli which vary along a single stimulus dimension such as intensity, and by studying visual–auditory transfer.

Effect of stimulus similarity on sensory preconditioning within a single stimulus dimension.

Effect of stimulus similarity on sensory preconditioning within a single stimulus dimension.