Recognition Performance as a Function of Detection Criterion in a Simultaneous Detection-Recognition Task

104

In two experiments, event-related brain potentials (ERPs) were recorded while subjects performed a simultaneous detection and recognition task. Ten subjects listened to pure tones in noise and reported both whether a target tone had occurred (using a four-category confidence rating scale) and whether the target was one of two (Experiment 1) or four (Experiment 2) tones differing in frequency. The amplitudes of three ERP components were found to be differentially related to detection and recognition performance. The early NI00 component varied with processing related only to detection, while the late P300 varied with both detection and recognition, and a later slow positive shift varied only with recognition and not with detection. While the latencies of both NIOO and P300 increased for less confident target detections, there were no differences in the latencies of these ERP components between correctly and incorrectly identified targets. Recognition performance was above the level expected by chance even when subjects reported that no target had been presented. The results indicate that brain potential components can be used to disclose temporal features of the processing of a stimulus by the nervous system and support the view that detection and recognition are partially independent, concurrent processes in perception.

Section: F~tor'contrasting

confidence: 55%

Detection and recognition: Concurrent processes in perception

Parasuraman

Richer

Beatty

1982

104

“…She found reasonable support for a threshold-type model. Lindner (1968) examined a situation similar to Shipley's, but found that a threshold-type model did not fit his data. In particular, he found that the probability of a correct modality judgment, conditional upon a "no" detection response, was above the guessing level, contrary to the predictions of a simple threshold model.…”

Section: Resultsmentioning

confidence: 98%

“…We propose perhaps the simplest possible model for this situation. This model is a threshold-type model which is similar to those discussed by Shipley (1965) and Lindner (1968). We assume attention has no effect.…”

Section: Resultsmentioning

confidence: 99%

Can attention be allocated to sensory modalities?

Shiffrin

Grantham

1974

Three experiments examined the effects of attention allocation to the modalities of vision, audition, and touch. The first two experiments utilized a simultaneous-successive comparison. The simultaneous procedure involved simultaneous monitoring of all three sensory modalities for the presence of a near-threshold stimulus. The successive condition allowed S to give his full attention to each sensory modality in turn. There was no advantage for the successive condition, whether the task consisted of detection of a single stimulus (Experiment I) or detection of the absence of one of many stimuli (Experiment II). Experiment III used a different paradigm to extend these results and bridge the gap between these results, those of Eijkman and Vendrik (1965), and those of Moray (1973a, b). We concluded that selective allocation of attention to sensory modalities does not affect the early stages of perceptual processing. Rumelhart (1970) and Norman and Rumelhart (1970) have proposed a model of this type based primarily on experimentation in visual processing. It should be noted that both types of limited-capacity attentional models require a "preattentive" mechanism to direct attention to important channels (see Neisser, 1967, for a discussion of the need for preattention in these models). The final class of models is seen in the lower panel of Fig. 1. These models assume minor limitations of This paper seeks to determine whether S can divert attention to a particular sensory modality and thereby improve the quality with which a signal in that modality is processed. Furthermore, if an improvement occurs, will it be associated with a corresponding decrement for signals presented on a nonattended sensory modality? On the other hand, if no change in processing results from changes in attention allocation, it should be possible to localize the site of attention within the memory system. All current views of information processing agree that a host of selective attention effects operate in short-term memory following perceptual processing. These were termed "control processes" by Atkinson and Shiffrin (1968), and include rehearsal and coding. We wish to determine if selective attention to sensory modalities operates during the initial processing of information prior to short-term store.Three classes of models of attention during sensory processing are shown in Fig. 1. The top panel illustrates a single-channel model in which information enters the recognition system from only a single source at anyone moment. It is assumed, in this model, that attention is switched rapidly among various possible sources of sensory information. A model that has been interpreted in this fashion was proposed by Broadbent (1958) in the context of split-span dichotic listening experiments. Franzen, Markowitz, and Swets (1970) proposed a model of this type for the processing of near-threshold vibrotactile information. Taylor (1964, 1966) proposed a model of this type for visual processing. Moray (1970a, b) proposed such a model for dichotic t...

“…For example, it is not clear whether signals must always be recognized in order to be detected. Shipley (1965) argues from the results of her simultaneous detection and recognition experiment that this is true, but Lindner's (1968) data from a similar experiment do not support such a hypothesis. Macmillan (1971) presented evidence of a different sort that the detection of signals was not completely dependent on their recognition.…”

mentioning

confidence: 91%

Detection and recognition of intensity changes in tone and noise: The detection-recognition disparity

Macmillan

1973

Recognition of increments vs decrements in auditory intensity improves with signal duration, relative to detection of increments and decrements. This effect obtains whether the background stimulus is a tone, noise, or a tone with noise masker, and is largely uninfluenced by the rise time of the signals. These data are inconsistent with detection models in which the 0 makes only one sensory observation during each observation interval, but can be described in terms of a neural timing model in which transients playa critical role.The problems of detection (distinguishing a signal from the absence of a signal) and recognition (distinguishing signals from each other) are frequently investigated in similar paradigms and explained by similar theories (Luce , 1963: Tanner, 1956). Yet basic facts concerning the relation" between the two are not well established. For example, it is not clear whether signals must always be recognized in order to be detected. Shipley (1965) argues from the results of her simultaneous detection and recognition experiment that this is true, but Lindner's (1968) data from a similar experiment do not support such a hypothesis.Macmillan (1971) presented evidence of a different sort that the detection of signals was not completely dependent on their recognition. At short durations (less than about 100 msec), as' ability to recognize whether a small increment or decrement in a pure tone background had been presented was comparable to their ability to detect those signals. At longer durations, their recognition performance exceeded their detection performance by approximately the amount predicted by a signal detection theory model in which the sensory effects due to the three signals (decrement, no change, and increment) are assumed to lie upon a single decision axis. This effect of duration on recognition, relative to detection, will be referred to as the "detection-recognition disparity."The interpretation of these data suggested by Macmillan was as follows: In addition to an integrative detector (ID), which accumulates information about the stimulus over time and offers an identification of the signal, the a also has a change detector (CD), which 65operates on signal onsets (and perhaps offsets). and thus does not integrate over time, and which only conveys information about the presence or absence of a signal. not about its identity. Both mechanisms combine in an unspecified manner to yield detection performance. but in recognition experiments only the ID is useful. In detection experiments, the CD will be relatively more important at short durations. where the ID is less reliable, so recognition performance will increase relative to detection as duration increases. i.e.. the detection-recognition disparity wiJI obtain. A model of this sort differs from most existing models of detection in proposing that detection (1) is in part nonintegrative , (2) is in part nonidentifying. and (3) involves more than one process.The positing of more than one process is perhaps the most serious departure fro...