Effects of similarity and practice on speeded classification response times and accuracies: Further tests of an exemplar-retrieval model

Observers were tested in a perceptual category-learning experiment in which they were instructed to make classification decisions as rapidly as possible without making errors. Nosofsky and Palmeri's (1997b) exemplar-based random walk (EBRW) model of speeded classification was tested for its ability to fit the classification response times and accuracies. The authors demonstrated that the EBRW model provided good quantitative fits to the mean response times and accuracies associated with individual objects as a function of their locations in a multidimensional similarity space and as a function of practice in the task. Preliminary evidence was also obtained that stimulus-specific adjustments in the random walk response criteria may have occurred during the course of learning.     

Testing Evidence Accrual Models by Manipulating Stimulus Onset

Many models of response time assume that subjects accrue stimulus "evidence" samples in time (e.g., random walk models, counter models). In this paper, the concept of one stimulus dominating another is used to construct a test of the whole class of evidence accrual models. For an example of dominance, consider stimuli that are presented either virtually instantaneously (stepped) or in a gradually increasing manner (ramped). Ramped stimuli are presented such that the ramped portion precedes the stepped onset of stepped stimuli. In this case ramped stimuli dominate stepped stimuli. In this paper the class of evidence accrual models is formalized. It is shown that under appropriate assumptions evidence accrual models do predict more accurate responses to dominating stimuli. However, this result does not hold for response latencies. There are anomalous cases where an evidence accrual model (the accumulator model of Vickers (1970, Ergonomics 13, 37-58)) predicts slower mean correct response latencies to dominating stimuli. It is shown through extensive computer simulation that these anomalous cases occur only when response criteria are so asymmetric that there are exceedingly extreme response biases. For experiments where response biases are not exceedingly extreme, random walk and accumulator models predict more accurate and quicker correct responses to dominating stimuli. In sum, manipulating the time course of stimuli in accordance with the concept of dominance can provide empirical tests of the class of evidence accrual models. Copyright 2001 Academic Press.     

Short-term memory scanning viewed as exemplar-based categorization

Exemplar-similarity models such as the exemplar-based random walk (EBRW) model (Nosofsky & Palmeri, 1997b) were designed to provide a formal account of multidimensional classification choice probabilities and response times (RTs). At the same time, a recurring theme has been to use exemplar models to account for old-new item recognition and to explain relations between classification and recognition. However, a major gap in research is that the models have not been tested on their ability to provide a theoretical account of RTs and other aspects of performance in the classic Sternberg (1966) short-term memory-scanning paradigm, perhaps the most venerable of all recognition-RT tasks. The present research fills that gap by demonstrating that the EBRW model accounts in natural fashion for a wide variety of phenomena involving diverse forms of short-term memory scanning. The upshot is that similar cognitive operating principles may underlie the domains of multidimensional classification and short-term old-new recognition.     

When do those who know more also know more about how much they know? The development of confidence and performance in categorical decision tasks

The study traces the development of confidence and its relation to performance as participants learn a new skill and develop from laymen to experts. We describe a new hierarchy of Categorical Decisions Structures that distinguishes between screening, discrimination, and classification modes. Each decision mode offers a different representation of information and provides a different type of feedback. We propose simple models of performance and confidence, and derive their predictions under all three modes. We report results of an empirical study designed to evaluate the models’ predictions. Ninety participants performed 1200 categorical decisions under the three modes and three distinct base rates. Results show that (a) decisions made in the screening mode tend to induce over-confidence; (b) the discrimination mode leads to fast learning and high correspondence between performance and confidence; and (c) the detailed feedback provided in the classification mode results in slow and steady improvement of the correspondence between confidence and performance.     

Gambling on the unconscious: A comparison of wagering and confidence ratings as measures of awareness in an artificial grammar task

We explore three methods for measuring the conscious status of knowledge using the artificial grammar learning paradigm. We show wagering is no more sensitive to conscious knowledge than simple verbal confidence reports but is affected by risk aversion. When people wager rather than give verbal confidence they are less ready to indicate high confidence. We introduce a “no-loss gambling” method which is insensitive to risk aversion. We show that when people are just as ready to bet on a genuine random process as their own classification decisions, their classifications are still above baseline, indicating knowledge participants are not aware of having. Our results have methodological implications for any study investigating whether people are aware of knowing.     

Retrieving texts from an external store: the effects of an explanatory context and of semantic fit between text and address

Summary External-information storage may reduce the load on human memory. But if information is preserved in an external store the task of retrieving information from that store puts new demands on memory. This creates a trade-off between internal and external storage. One of the prerequisites for external retrieval is learning an address. In this study, the trade-off between internal and external retrieval is exemplified by contrasting the memorization of written texts with the memorization of addresses under which these texts are stored in a computer file. In Experiment 1 Ss were tested for the retention of addresses. In Experiment 2 they were tested for the retention of texts, but had the choice between memorizing the texts and memorizing an address under which the texts were available. Independent variables were the availability of an explanatory context, and the semantic fit between address and text. An explanatory context improved address learning in Experiment 1 and unaided recall in Experiment 2. It did not influence the utilization of addresses as tested in Experiment 2. The semantic fit between address and text had a considerable impact on both the learning and the utilization of addresses. Whereas some subjects were assigned standard addresses, others generated their own addresses. Self-generated addresses had effects similar to standard addresses selected for high semantic fit. It is suggested that learning an external address involves an internal representation of the material stored under that address.     

Overconfidence effects in category learning: a comparison of connectionist and exemplar memory models

Exemplar and connectionist models were compared on their ability to predict overconfidence effects in category learning data. In the standard task, participants learned to classify hypothetical patients with particular symptom patterns into disease categories and reported confidence judgments in the form of probabilities. The connectionist model asserts that classifications and confidence are based on the strength of learned associations between symptoms and diseases. The exemplar retrieval model (ERM) proposes that people learn by storing examples and that their judgments are often based on the first example they happen to retrieve. Experiments 1 and 2 established that overconfidence increases when the classification step of the process is bypassed. Experiments 2 and 3 showed that a direct instruction to retrieve many exemplars reduces overconfidence. Only the ERM predicted the major qualitative phenomena exhibited in these experiments.     

Age differences in memory retrieval shift: governed by feeling-of-knowing?

The noun-pair lookup (NP) task was used to evaluate strategic shift from visual scanning to retrieval. We investigated whether age differences in feeling-of-knowing (FOK) account for older adults' delayed retrieval shift. Participants were randomly assigned to one of three conditions: (1) standard NP learning, (2) fast binary FOK judgments, or (3) Choice, where participants had to choose in advance whether to see the look-up table or respond from memory. We found small age differences in FOK magnitudes but major age differences in memory retrieval choices that mirrored retrieval use in the standard NP task. Older adults showed lower resolution in their confidence judgments (CJs) for recognition memory tests on the NP items, and this difference appeared to influence rates of retrieval shift, given that retrieval use was correlated with CJ magnitudes in both age groups. Older adults had particular difficulty with accuracy and confidence for rearranged pairs, relative to intact pairs. Older adults' slowed retrieval shift appears to be attributable to (1) impaired associative learning early in practice, not just a lower FOK; but also (2) retrieval reluctance later in practice after the degree of associative learning would afford memory-based responding.     

Modeling memory and perception

I present a framework for modeling memory, retrieval, perception, and their interactions. Recent versions of the models were inspired by Bayesian induction: We chose models that make optimal decisions conditioned on a memory/perceptual system with inherently noisy storage and retrieval. The resultant models are, fortunately, largely consistent with my models dating back to the 1960s, and are therefore natural successors. My recent articles have presented simplified models in order to focus on particular applications. This article takes a larger perspective and places the individual models in a more global framework. I will discuss (1) the storage of episodic traces, the accumulation of these into knowledge (e.g., lexical/semantic traces in the case of words), and the changes in knowledge caused by learning; (2) the retrieval of information from episodic memory and from general knowledge; (3) decisions concerning storage, retrieval, and responding. Examples of applications include episodic recognition and cued and free recall, perceptual identification (naming, yes–no and forced‐choice), lexical decision, and long‐term and short‐term priming.     

Situation models and aging

Younger and older adults were tested for their ability to process and retrieve information from texts. The authors focused on the construction and retrieval of situation models relative to other types of text representations. The results showed that during memory retrieval, younger adults showed superior memory for surface form and textbase knowledge (what the text was), whereas older adults had equivalent or superior memory for situation model information (what the text was about). The results also showed that during reading, older and younger adults were similar in their sensitivity to various aspects of the texts. Overall, these findings suggest that although there are age-related declines in the processing and memory for text-based information, for higher level representations, these abilities appear to be preserved. Several possibilities for why this is the case are discussed, including an in-depth consideration of one possibility that involves W. Kintsch's (1988) construction-integration model.     

A method in modeling interactive pedestrian crossing and driver yielding decisions during their interactions at intersections

Investigating pedestrian crossing and driver yielding decisions should be an important focus considering the high risks of pedestrians in exposed to motorized traffic. Limitations, however, exist in previous studies – variables considered previously have been limited; how their behavior affect each other (defined as interactive impacts) were not sufficiently considered. This paper aims to provide a methodological approach for pedestrian crossing and driver yielding decisions during their interactions, considering of different variable types including interactive impact variables, traffic condition variables, road design variables, and environment variables. A Distance-Velocity (DV) framework proposed in an earlier study is introduced for definitions and concepts in studying pedestrian-vehicle interactions. Logistic regression, support vector machines, neural networks and random forests, are introduced as candidate models. A case study involving six crosswalk locations is conducted, focusing on interactions between pedestrians and right-turn vehicles. The proposed methodological approach is applied, with the performance of the four machine learning methods compared in terms of model generalization and confusion matrix. The model with the best performance is further compared to the typical gap-based model. Results show that random forest and logistic regression models performed the best in modeling pedestrian crossing and driver yielding decisions respective, in terms of model generalization. Besides, the DV-based modeling method (average accuracy of over 90% for pedestrians and 80% for drivers) outperformed the traditional gap-based method in all test seeds. As a key finding, interactive impacts from each other (the pedestrian and the driver) act as a key contributing variable on their decisions.     

Does opposition logic provide evidence for conscious and unconscious processes in artificial grammar learning?

The question of whether studies of human learning provide evidence for distinct conscious and unconscious influences remains as controversial today as ever. Much of this controversy arises from the use of the logic of dissociation. The controversy has prompted the use of an alternative approach that places conscious and unconscious influences on memory retrieval in opposition. Here we ask whether evidence acquired via the logic of opposition requires a dual-process account or whether it can be accommodated within a single similarity-based account. We report simulations using a simple neural network model of two artificial grammar learning experiments reported by that dissociated conscious and unconscious influences on classification. The simulations demonstrate that opposition logic is insufficient to distinguish between single- and multiple-system models.     

Does Providing Prompts During Retrieval Practice Improve Learning?

The purpose of this investigation was to identify ways to prompt retrieval practice to make recall even more effective at producing meaningful learning. In two experiments, subjects read educational texts and practiced retrieval across two periods. During prompted retrieval, subjects were cued to explain and describe concepts from the text, whereas during free recall, subjects recalled as much of the material from the text as they could. A reading control condition was also included. Learning was assessed using both verbatim and higher‐order questions 1 week later. Practicing retrieval improved learning relative to the control on both types of questions; however, whether subjects practiced free or prompted retrieval did not matter for learning. Subjects rated prompted retrieval as less enjoyable and interesting than the other retrieval conditions. Results demonstrate practicing retrieval promoted meaningful learning, and that subjects' initial retrieval success was highest when they used their own retrieval strategies during recall. Copyright © 2016 John Wiley & Sons, Ltd.     

Towards end-to-end likelihood-free inference with convolutional neural networks

Complex simulator-based models with non-standard sampling distributions require sophisticated design choices for reliable approximate parameter inference. We introduce a fast, end-to-end approach for approximate Bayesian computation (ABC) based on fully convolutional neural networks. The method enables users of ABC to derive simultaneously the posterior mean and variance of multidimensional posterior distributions directly from raw simulated data. Once trained on simulated data, the convolutional neural network is able to map real data samples of variable size to the first two posterior moments of the relevant parameter's distributions. Thus, in contrast to other machine learning approaches to ABC, our approach allows us to generate reusable models that can be applied by different researchers employing the same model. We verify the utility of our method on two common statistical models (i.e., a multivariate normal distribution and a multiple regression scenario), for which the posterior parameter distributions can be derived analytically. We then apply our method to recover the parameters of the leaky competing accumulator (LCA) model and we reference our results to the current state-of-the-art technique, which is the probability density estimation (PDA). Results show that our method exhibits a lower approximation error compared with other machine learning approaches to ABC. It also performs similarly to PDA in recovering the parameters of the LCA model.     

Problem structure and the use of base-rate information from experience

This article is concerned with the use of base-rate information that is derived from experience in classifying examples of a category. The basic task involved simulated medical decision making in which participants learned to diagnose hypothetical diseases on the basis of symptom information. Alternative diseases differed in their relative frequency or base rates of occurrence. In five experiments initial learning was followed by a series of transfer tests designed to index the use of base-rate information. On these tests, patterns of symptoms were presented that suggested more than one disease and were therefore ambiguous. The alternative or candidate diseases on such tests could differ in their relative frequency of occurrence during learning. For example, a symptom might be presented that had appeared with both a relatively common and a relatively rare disease. If participants are using base-rate information appropriately (according to Bayes' theorem), then they should be more likely to predict that the common disease is present than that the rare disease is present on such ambiguous tests. Current classification models differ in their predictions concerning the use of base-rate information. For example, most prototype models imply an insensitivity to base-rate information, whereas many exemplar-based classification models predict appropriate use of base-rate information. The results reveal a consistent but complex pattern. Depending on the category structure and the nature of the ambiguous tests, participants use base-rate information appropriately, ignore base-rate information, or use base-rate information inappropriately (predict that the rare disease is more likely to be present). To our knowledge, no current categorization model predicts this pattern of results. To account for these results, a new model is described incorporating the ideas of property or symptom competition and context-sensitive retrieval.     

计算机化分类测验终止规则的类别、特点及应用

计算机化分类测验(Computerized Classification Testing, CCT)能够高效地对被试进行分类, 已广泛应用于合格性测验及临床心理学中。作为CCT的重要组成部分, 终止规则决定测验何时停止以及将被试最终划分到何种类别, 因此直接影响测验效率及分类准确率。已有的三大类终止规则(似然比规则、贝叶斯决策理论规则及置信区间规则)的核心思想分别为构造假设检验、设计损失函数和比较置信区间相对位置。同时, 在不同测验情境下, CCT的终止规则发展出不同的具体形式。未来研究可以继续开发贝叶斯规则、考虑多维多类别情境以及结合作答时间和机器学习算法。针对测验实际需求, 三类终止规则在合格性测验上均有应用潜力, 而临床问卷则倾向应用贝叶斯规则。     

Retrieval practice and retention of course content in a middle school science classroom

We investigated the effect of retrieval practice (i.e., the use of testing to enhance learning) by middle school students on science learning in an authentic educational setting. For initial study, all students (n = 39) read a science text and made study notes about the text, a regular study activity in their course. For restudy (2 days later), students either copied their notes or did retrieval practice. For the final test (an additional 2 days later), all students did a recall task from memory. Students in the retrieval practice condition showed better retention of information than students in the copy condition. These results add to a growing body of research on the use of retrieval practice as an effective learning tool for retention within authentic school settings.     

Quantization of molecular motion at the synapse and its consequences as contrasted with the generalized turing machine

Using the most general models, we show that molecular motion concerned with transmission of the nervous impulse across the synapse is indeterminate (in the sense of Heisenberg). This occurs despite the assumption of a minimal uncertainty product and the taking into account of the nature of the random walk. Consequences of the interposition of an indeterminate process in nervous system function are detailed by investigation of the operation of a “quantized” Turing machine. Such a generalized algorithm has been taken by some investigators to be the starting point for building analogues to brain function. The brain, with a demonstrated indeterminate basis for at least one of its “levels” of operation, has capabilities or modes of operation not able to be mimed by the most general calculating engine.     

Confidence in accuracy moderates the benefits of retrieval practice

ABSTRACT

The effect of Retrieval Practice refers to the phenomenon that taking a practice test is more effective for learning than re-study, probably due to the benefit from processes underlying successful retrievals during practice. However, it is rarely studied whether other processes (e.g., metacognitive monitoring) during retrieval practice may also play an important role. In two experiments, we examined whether the effectiveness of retrieval was affected by subjects’ confidence in their retrieval success. Subjects studied word-pairs and rated their confidence after each practice test trial (cued-recall in Exp.1, multiple choice in Exp.2), or re-study trial. In both experiments, we observed the classic retrieval practice effect. Moreover, in Exp.2, the benefits differed as a function of subjects’ confidence in their retrieval performance: the effect of retrieval practice only emerged for high-confidence trials with higher than 56% confidence. In summary, the retrieval practice only facilitates the retention of correct answers with high confidence.