Professional Discretion and Accountability in the Welfare State

The discretionary powers of welfare state professionals are in tension with the requirements of the democratic Rechtsstaat. Extensive use of discretion can threaten the principles of the rule of law and relinquish democratic control over the implementation of laws and policies. These two tensions are in principle ineradicable. But does this also mean that they are impossible to come to grips with? Are there measures that may ease these tensions? We introduce an understanding of discretion that adds an epistemic dimension (discretion as a mode of reasoning) to the common structural understanding of discretion (an area of judgment and decision). Accordingly we distinguish between structural and epistemic measures of accountability. The aim of the former is to constrain discretionary spaces or the behaviour within them while the aim of the latter is to improve the quality of discretionary reasoning. The focus in this article is on epistemic measures that are internally related to the main characteristic of accountability, namely justified use of discretionary power.     

Allocation of attention in the visual field

This research investigated whether attentional resources can be simultaneously allocated to several locations in a visual display, whether the mode of processing (serial or parallel) can be switched within a trial, and the nature of the costs when attentional resources are concentrated on an invalid location. Subjects were required to determine which of two target letters was present in eight-letter circular displays. In precue conditions, a primary and a secondary target location were designated 150 ms before target onset by an indicator that varied in validity. In the control conditions no cue was provided. A second experiment verified several assumptions that had been made in interpreting the data of Experiment 1. Modifications in Jonides' (1983) two-process model were suggested in terms of a zoom lens model of attentional resources. Instead of two alternative processing modes, attentional resources are conceived as capable of distribution over the visual field, but with low resolving power, or as continuously constricting to small portions of the visual field with a concomitant increase in processing power.     

Pre- and poststimulus activation of response channels: a psychophysiological analysis

To examine mechanisms of response activation, we asked subjects to respond differentially to the central letter of one of four arrays--HHHHH, SSHSS, SSSSS, and HHSHH--and measured event-related brain potentials (ERPs) and electromyographic activity (EMG). For very fast responses, accuracy was at chance level for all arrays, suggesting that subjects were guessing. For intermediate latency responses, accuracy was above chance if the noise was compatible with the targets and below chance if it was incompatible, suggesting that these responses were based on partial stimulus analysis. For slow responses, accuracy was above chance for all arrays, suggesting that these responses were based on complete stimulus analysis. The occurrence and accuracy of fast responses could be predicted by examining motor potentials preceding the presentation of the array. Measures of the motor potentials in the period following the presentation of the array suggested that partial analysis of stimulus information could activate responses and that the level of response activation at the time of the EMG response was constant for trials with different response latencies. The data are discussed in terms of a response channel conception.     

Filtering versus parallel processing in RSVP tasks.

An experiment of McLean, D. E. Broadbent, and M. H. P. Broadbent (1983) using rapid serial visual presentation (RSVP) was replicated. A series of letters in one of 5 colors was presented, and the subject was asked to identify the letter that appeared in a designated color. There were several innovations in our procedure, the most important of which was the use of a response menu. After each trial, the subject was presented with 7 candidate letters from which to choose his/her response. In three experimental conditions, the target, the letter following the target, and all letters other than the target were, respectively, eliminated from the menu. In other conditions, the stimulus list was manipulated by repeating items in the series, repeating the color of successive items, or even eliminating the target color. By means of these manipulations, we were able to determine more precisely the information that subjects had obtained from the presentation of the stimulus series. Although we replicated the results of McLean et al. (1983), the more extensive information that our procedure produced was incompatible with the serial filter model that McLean et al. had used to describe their data. Overall, our results were more compatible with a parallel-processing account. Furthermore, intrusion errors are apparently not only a perceptual phenomenon but a memory problem as well.     

Filtering versus parallel processing in RSVP tasks

An experiment of McLean, D. E. Broadbent, and M. H. P. Broadbent (1983) using rapid serial visual presentation (RSVP) was replicated. A series of letters in one of 5 colors was presented, and the subject was asked to identify the letter that appeared in a designated color. There were several innovations in our procedure, the most important of which was the use of a response menu. After each trial, the subject was presented with 7 candidate letters from which to choose his/her response. In three experimental conditions, the target, the letter following the target, and all letters other than the target were, respectively, eliminated from the menu. In other conditions, the stimulus list was manipulated by repeating items in the series, repeating the color of successive items, or even eliminating the target color. By means of these manipulations, we were able to determine more precisely the information that subjects had obtained from the presentation of the stimulus series. Although we replicated the results of McLean et al. (1983), the more extensive information that our procedure produced was incompatible with the serial filter model that McLean et al. had used to describe their data. Overall, our results were more compatible with a parallel-processing account. Furthermore, intrusion errors are apparently not only a pereptual phenomenon but a memory problem as well.     

Selective encoding from multielement visual displays

When a multiletter display is preceded by a bar designating one of the letters for report, reaction time (RT) to voice the indicated letter decreases. Previous research had indicated that the efficiency of this selective mechanism decreases as the number of display elements increases. Two experiments were conducted to determine whether the effect of display size could be eliminated when the indicator precedes the display at long intervals. Results indicated that the display size effect was maintained. The results could not be attributed to eye movements, but were interpreted in terms of a central encoding mechanism that is limited in its precision of localization and exclusion.     

The extent of processing of noise elements during selective encoding from visual displays

Circular visual displays of 12 elements, consisting of A, H, M, and U, were presented to S. She responded with a lever movement in one direction if a given letter designated by a bar indicator was a member of the set A-U and in the opposite direction if it was from the other set. The principal experimental variables were the SOA by which the indicator preceded the display and whether the target letter was flanked by letters of the same or opposite set. The results indicated that the interference produced by noise letters is primarily on the response as opposed to the processing side. Also, there is a limit to the precision of selective attention in the visual field. The interference of opposite-set letters was inversely proportional to their distance from the target.