Signal detection and threshold models of source memory

Recently, several authors claimed that the curvilinear shape of rating-based source memory receiver operating characteristics (ROCs) refutes threshold models. However, rating-based ROCs are not diagnostic to disprove threshold models. Furthermore, source memory ROC-analyses ignore influences of other processes like old-new-detection and old-new-response-tendencies, so direct estimation of parameters is preferable. Five source monitoring experiments with different response bias manipulations and materials were conducted. We fitted the Two High Threshold Multinomial Model of Source Monitoring by Bayen, Murnane, and Erdfelder (1996) and the Multivariate Signal Detection Model for Selection DeCarlo (2003). We also included rating-based ROCs. The results suggest that both models are at least equally valid as measurement tools which capture bias processes in the corresponding parameters, and they can perhaps be integrated theoretically.     

Animal models of source memory

Source memory is the aspect of episodic memory that encodes the origin (i.e., source) of information acquired in the past. Episodic memory (i.e., our memories for unique personal past events) typically involves source memory because those memories focus on the origin of previous events. Source memory is at work when, for example, someone tells a favorite joke to a person while avoiding retelling the joke to the friend who originally shared the joke. Importantly, source memory permits differentiation of one episodic memory from another because source memory includes features that were present when the different memories were formed. This article reviews recent efforts to develop an animal model of source memory using rats. Experiments are reviewed which suggest that source memory is dissociated from other forms of memory. The review highlights strengths and weaknesses of a number of animal models of episodic memory. Animal models of source memory may be used to probe the biological bases of memory. Moreover, these models can be combined with genetic models of Alzheimer's disease to evaluate pharmacotherapies that ultimately have the potential to improve memory.     

Hemispheric asymmetries for temporal resolution: A signal detection analysis of threshold and bias

Divided visual field techniques were used to investigate hemispheric asymmetries for (a) the threshold of fusion of two flashes of light and (b) the detection of simultaneous versus successive events for a group of normal, right-handed adults. A signal detection analysis revealed a higher level of accuracy for the right visual field-left hemisphere (RVF-LH) relative to the left visual field-right hemisphere (LVF-RH) for both tasks. These results were interpreted in terms of a general left-hemisphere advantage for the discrimination of fine temporal events. The implications of these results for models of temporary asymmetry that describe the left hemisphere's advantage in terms of an exclusive specialization or relative superiority are then discussed.     

Similar threshold functions from contrasting neural signal models

Three neural signal models of increment threshold detection are compared. All assume that the criterion for threshold is the attainment of a critical, minimum neural signal (or difference between two neural signals), and that the signal due to a test flash of intensity λ in the absence of a background light is $\text{λ}\text{(λ + σ)}$ (where σ is the semi-saturation constant). The models differ in the manner in which a background light of intensity θ is assumed to affect the signal. One model (due to Alpern et al., 1970a, Alpern et al., 1970b, Alpern et al., 1970c) assumes that the test flash signal, $\text{λ}\text{(λ + σ)}$, is attenuated by the multiplicative factor $\text{θ}{}_{\mathrm{D}}\text{(θ + θ}{}_{\mathrm{D}}\text{)}$ (where θ_D is a constant interpreted as sensory noise); another model specifies that the test flash signal is simply reduced (by subtraction) by the amount $\text{θ}\text{(θ + K)}$ (K a constant). One main result of this paper is that in the absence of pigment bleaching, these two models imply indistinguishable increment threshold functions. Further, a necessary and sufficient condition for each model guaranteeing the absence of saturation with steady backgrounds is found to be empirically satisfied. A third model is considered where the background field is assumed both to contribute to the neural signal and simultaneously to attenuate it (via a gain change). These assumptions are closely related to theoretical accounts of color induction and color perception. Though this model needs further investigation, it appears to be in better accord with actual increment threshold data than the others.     

ROC and z-ROC analysis in recognition memory: continuous, threshold, and asymmetric models

In this research, we applied the operating characteristic of the receiver (ROC) to evaluate item-recognition memory by means of open class words (CA) and closed class words (CC). The purpose was to evaluate memory processes and models, according to the type of stimuli used: words in Spanish, classified according to grammar characteristics. To register the data, a rating scale was applied to obtain the appropriate response levels. Statistically significant differences between the processing of both types of words were obtained. We evaluated models of threshold, continuous, and unequal variance. As a result, we could contribute to the resolution of the present controversy about the models adapted to evaluate the memory of item recognition. Lastly, the differentiation and integration of the models from an empirical point of view is discussed.     

Auditory threshold,signal-to-noise ratio and signal detection theory

The use of Signal Detection Theory has demonstrated that there is not an auditory threshold at the level supposed by classical psychophysics, but has not disproved the existence of an auditory threshold at a lower level. Following the recommended signal detection theory procedure fairly closely, an experiment obtained evidence either of the existence of a signal-to-noise ratio audi tory threshold, or a nonlinear relationship between stimulus and response which could not be statistically distiguished from such a threshold , Using the same apparatus and subjects, audi tory thresholds were also obtained using the classical method of limits. The acoustic pressure for the classical thresholds was about five times as great as for the signal detection theory method.     

An application of signal detection theory with finite mixture distributions to source discrimination

A mixture extension of signal detection theory is applied to source discrimination. The basic idea of the approach is that only a portion of the sources (say A or B) of items to be discriminated is encoded or attended to during the study period. As a result, in addition to 2 underlying probability distributions associated with the 2 sources, there is a 3rd distribution that represents items for which sources were not attended to. Thus, over trials, the observed response results from a mixture of an attended (A or B) distribution and a nonattended distribution. The situation differs in an interesting way from detection in that, for detection, there is mixing only on signal trials and not on noise trials, whereas for discrimination, there is mixing on both A and B trials. Predictions of the mixture model are examined for data from several recent studies and in a new experiment.     

A comparison of binary and rating techniques in the signal detection analysis of recognition memory

Forty-eight subjects were run on a recognition memory task employing low, medium, or high meaningfulness (m) CVC items and utilizing either a binary (yes—no) or rating response technique. Significant differences were obtained between levels of m with higher m items resulting in larger d' values. The interaction term was also significant reflecting a large difference between the binary-rating techniques with high meaningfulness items. The results indicate that the binary and rating techniques do not reveal identical findings in the analysis of recognition memory data and that the frequent use of the rating technique by researchers in the analysis of memory data may not be appropriate.     

Understanding memory impairment with memory models and hierarchical Bayesian analysis

The study of human episodic memory is a topic that interests cognitive and mathematical psychologists as well as clinicians interested in the diagnosis and assessment of Alzheimer’s disease and related disorders (ADRD). In this paper, we use simple cognitive models for the recognition and recall tasks typically applied in clinical assessments of ADRD to study memory performance in ADRD patients. Our models make use of hierarchical Bayesian methods as a way to model individual differences in patient performance and to facilitate the modeling of performance changes that occur during multiple recall tasks. We show how the models are able to account for different aspects of patient performance, and also discuss some of the predictive capabilities of the model. We conclude with a discussion on the scope to improve on our results by discussing the link between memory theory in psychology and clinical practice.     

Picture-word differences in decision latency: An analysis of single and dual memory models

Semantic and perceptual size decision times for pictorial and verbal material were analyzed in the context of a unitary memory model and several dual memory models. Experiment 1 involved a same-different categorical judgment task. The results showed that picture-picture response latencies were 185 msec faster than the corresponding word-word latencies, and word-picture and picture-word latencies equaled the mean of these two extremes. Similarity of subcategory for “same” judgments led to faster decision latency for all presentation conditions. Additionally, a linear relationship was found between picture-picture and word-word latencies for individual item pairs. Experiment 2 involved a comparison of pictures and words across a. categorical judgment and a size judgment task. Pictures produced faster decision latencies in both tasks, and the latency diflerence between pictures and words was comparable across tasks. These data fit the predictions of a unitary memory model. Several variants of a dual memory model are rejected and those which fit the data require assumptions about storage and/or transfer time values which result in a functional regression to the unitary memory model.     

An introduction to Bayesian hierarchical models with an application in the theory of signal detection

Although many nonlinear models of cognition have been proposed in the past 50 years, there has been little consideration of corresponding statistical techniques for their analysis. In analyses with nonlinear models, unmodeled variability from the selection of items or participants may lead to asymptotically biased estimation. This asymptotic bias, in turn, renders inference problematic. We show, for example, that a signal detection analysis of recognition memory data leads to asymptotic underestimation of sensitivity. To eliminate asymptotic bias, we advocate hierarchical models in which participant variability, item variability, and measurement error are modeled simultaneously. By accounting for multiple sources of variability, hierarchical models yield consistent and accurate estimates of participant and item effects in recognition memory. This article is written in tutorial format; we provide an introduction to Bayesian statistics, hierarchical modeling, and Markov chain Monte Carlo computational techniques.     

Some models of observer behavior in two-channel auditory signal detection

Five models of 0 behavior in a two-channel exclusive-or (XOR) detection task were evaluated. The models included two types of single-channel and three types of two-channel O. Only the most efficient two-channel model adequately described human performance in a set of monaural and dichotic XOR conditions. Detectability measures derived from the XOR task matched those obtained from separate single-channel control conditions. We concluded that. in this two-channel task. the O’s performance was not limited by any inability to monitor signals arriving simultaneousl.v in two different earphone or two different frequency channels. The implications of this result for two-channel information processing and for multicomponent and sequential signal detectmn are discussed.     

Three-stimulus procedures in olfactory psychophysics: An experimental comparison of Thurstone-Ura and three-alternative forced-choice models of signal detection theory

The experimental differences between the triangular method and three-alternative forcedchoice (3-AFC) signal detection theory procedure are briefly reviewed, and the conditions to be met for a valid application of the respective procedures under the Thurstone-Ura and 3-AFC models are described. The assumptions of both of these unidimensional probabilistic models are given, and the functions relating the probability of a correct response to d, a parameter of sensory stimulus distance, are specified. Since the two models both contain d, the hypothesis of invariability of this parameter with the triangular method and the 3-AFC procedure was tested using olfactory stimuli. This hypothesis was strongly confirmed, since the functions relating d to the logarithm of the ratio of the physical stimulus values were virtually identical for the two methods and their concomitant models. The results are important not only for the field of olfactory psychophysics; they also have wider methodological and psychophysical interest. The bias against the middle stimulus of a triangle in the triangular method observed when olfactory stimuli are used can be explained as the resultant of sensory adaptation. Finally, some practical implications of the present results for application of the triangular method are discussed.     

A direct test of the unequal-variance signal detection model of recognition memory

Analyses of the receiver operating characteristic (ROC) almost invariably suggest that, on a recognition memory test, the standard deviation of memory strengths associated with the lures (σ_lure) is smaller than that of the targets (σ_target). Often, σ_lure/σ_target ≈ 0.80. However, that conclusion is based on a model that assumes that the memory strength distributions are Gaussian in form. In two experiments, we investigated this issue in a more direct way by asking subjects to simply rate the memory strengths of targets and lures using a 20-point or a 99-point strength scale. The results showed that the standard deviation of the ratings made to the targets (s _target) was, indeed, larger than the standard deviation of the ratings made to the lures (s _lure). Moreover, across subjects, the ratios _lure/s _target correlated highly with the estimate of σ_lure/σtarget obtained from ROC analysis, and both estimates were, on average, approximately equal to 0.80.     

Preference structures and threshold models

Preferences are often represented in terms of a function, in the deterministic case as well as in the probabilistic case. In the present paper we develop a new numerical representation of preference structures for which the strict preference relation (P) is without circuit but not necessarily transitive. Moreover, we investigate the consequences of the representation for the usual preference structures. In particular, we propose new formulations for the numerical representation of the interval order structure.     

Comparisons of some learning models for response bias in signal detection

The effects of variations in signal probability and varying degrees of correct feedback on response bias were studied in a yes-no auditory signal detection task. The main finding was that the bias towards saying yes was an increasing function of the frequency of signal feedback events, but did not depend on the correctness of the feedback. Several learning models coupled with a simple psychophysical and decision model yielded predictions about overall biases and certain sequential statistics. Only one model, which can be decribed as an “informational” model, gave a good account of both observed overall biases and sequential statistics. This model assumes the observer’s response bias is strengthened for the feedback-reinforced response when the observer’s sensory information is ambiguous or is contradicted by the feedback information.     

An experimental analysis of memory processing

Rhesus monkeys were trained and tested in visual and auditory list-memory tasks with sequences of four travel pictures or four natural/environmental sounds followed by single test items. Acquisitions of the visual list-memory task are presented. Visual recency (last item) memory diminished with retention delay, and primacy (first item) memory strengthened. Capuchin monkeys, pigeons, and humans showed similar visual-memory changes. Rhesus learned an auditory memory task and showed octave generalization for some lists of notes--tonal, but not atonal, musical passages. In contrast with visual list memory, auditory primacy memory diminished with delay and auditory recency memory strengthened. Manipulations of interitem intervals, list length, and item presentation frequency revealed proactive and retroactive inhibition among items of individual auditory lists. Repeating visual items from prior lists produced interference (on nonmatching tests) revealing how far back memory extended. The possibility of using the interference function to separate familiarity vs. recollective memory processing is discussed.