Working memory controls involuntary attention switching: evidence from an auditory distraction paradigm

Introduction

Working memory permits the storage and processing of information in the service of higher cognitive functions such as text comprehension (Baddeley, 1986; Daneman & Merikle, 1996; Cowan, 1999). This includes protection from distraction as is the case, for example, when reading in a noisy environment. By contrast, working memory should not completely be encapsulated from information unrelated to the current task in order to react adequately to unexpected and possibly dangerous events. Such events are registered by a system preattentively scanning the environment for potentially relevant information (Näätänen, 1992). A prerequisite for normal behaviour is therefore to adjust the optimal balance between focusing on task demands and the involuntary switching of attention as a basis of distraction. This view is in accordance with working memory models, suggesting a close link between working memory and attention (Baddeley, 1986; Cowan, 1999; Engle et al., 1999; Kane et al., 2001). Whereas research on working memory has focused on the aspect of keeping information active, or on how attention is divided between two parallel tasks (Baddeley, 1986, 1996), little is known about the interaction between working memory functioning and distraction by involuntary attention. The present experiment aimed at investigating the interaction of involuntary attention and working memory in the auditory system.

The processes involved in the preattentive scanning of the environment for new information, the involuntary orienting of attention, and the subsequent reorienting on task-relevant information can be investigated with an auditory distraction paradigm consisting of a duration discrimination task and task-irrelevant pitch changes (e.g. Schröger & Wolff, 1998a, b; Berti & Schröger, 2001; Roeber et al., 2003; Sussman et al., in press). These changes elicit the mismatch negativity (MMN) component of the event-related brain potential (ERP). It is generated when a discrepancy between the memory record for the regular, standard sound and the current, deviant sound is detected. Importantly, the MMN is elicited preattentively, that is when the subject has no explicit intention to detect the deviancy (e.g. Näätänen & Winkler, 1999; Näätänen et al., 2002). The MMN is followed by the P3a which is assumed to indicate the switching of attention towards a perturbing event (e.g. Friedman et al., 2001; Alho et al., 2003). Subsequent to the P3a, the reorienting negativity (RON) is elicited in conditions where subjects discriminate duration, but does not occur when frequency changes were task relevant. It is assumed to be related to the reorienting of attention towards task-relevant aspects of stimulation following distraction (Escera et al., 2000; Schröger & Berti, 2000; Sussman et al., 2003).

Here we investigate whether the working memory system has some control over these distraction effects by manipulating task load in the auditory distraction paradigm. If working memory has no effect on involuntary attention, distraction effects should not be modulated by an increase in task demands. Alternatively, it seems possible that working memory might affect involuntary attention in order to cope with increasing task demands. Behavioural data were collected to determine the presence and the relative magnitude of distraction. The ERPs were measured to determine which of the underlying processing stages (preattentive change detection, involuntary attention switching, voluntary reorienting) are affected by task load and which are not.

Materials and methods

Subjects and experimental design

Ten subjects (age range 19–26 years) performed an auditory duration discrimination task under two experimental conditions. According to the Declaration of Helsinki, subjects gave informed consent about their participation in this study.

During the experiment, subjects were tested in a low-load and a high-load condition (see Fig. 1). In both conditions subjects were presented repetitively every 1800 ms with sinusoidal tones of 200 or 400 ms duration including a 5-ms rise and fall time. In addition, the stimuli were presented with a repetitive frequency (standards), or a rare, deviant frequency (deviants): standard stimuli (P = 0.9) had a frequency of 1000 Hz, and deviant stimuli (P = 0.1) had a frequency of either 950 Hz or 1050 Hz. Standard and deviant stimuli were presented binaurally (through headphones) with equal probability as short or long stimuli.

In both experimental conditions, the subjects' task was to discriminate between short and long tones and indicate the duration by a button-press. To increase the working memory load in the high-load condition, subjects were asked to postpone their response until the next tone was presented. Therefore, in the low-load condition subjects had to respond to the duration of the current stimulus whereas in the high-load condition subjects were instructed to react to the duration of the preceding stimulus (see Fig. 1). As noted before, the rare frequency changes were irrelevant for the task, and therefore subjects were instructed to attend only to the duration of the stimuli. In both experimental conditions subjects started with a training block consisting only of standard stimuli to practice the task.

Results

The behavioural data show that subjects can perform the discrimination task under the low- and the high-load condition. However, as expected, overall percentage of correct responses is reduced in the high-load (83.3%) relative to the low-load condition (95.4%). This indicates that the delayed response task is cognitively more demanding (F_1,9 = 7.47, P < 0.05). Moreover, subjects are faster in the low-load than in the high-load condition irrespective whether standard or deviant stimuli were presented (Fig. 2). This result is qualified by the two-way anova showing a significant main effect of the load factor (F_1,9 = 16.25, P < 0.01). More important, deviant stimuli prolonged the reaction time in the duration discrimination task in both conditions as reflected by the main effect of the factor stimulus type (F_1,9 = 36.46, P < 0.001). Interestingly, the extent of distraction is greater in the low-load (48 ms) than in the high-load (20 ms) condition. Correspondingly, the significant main effects are modified by a significant interaction in the anova (F_1,9 = 6.37, P < 0.05) which is because of a reduced distraction effect in the high-load condition. Additional t-tests for the individual reaction time-prolongation confirmed the existence of a distraction effect in both conditions (low-load condition t₉ = 5.17, P < 0.001; high-load condition, t₉ = 3.23, P < 0.05).

The ERPs and the corresponding difference waves obtained in the experiment (Fig. 3) show distinct deviance-related effects in both conditions: The MMN, at the 100–200 ms range relative to the onset of the deviant, is followed by the P3a around 350 ms and the RON peaking around 500 ms. Importantly, MMN amplitude as a correlate of the sensory change detection system does not differ between both conditions, whereas P3a and RON amplitudes are reduced in the high-load condition. This is reflected by the statistical analysis revealing interactions of the load and the stimulus type factor for the P3a window and the RON window but not for the MMN window (MMN: load, F_1,9 = 16.42, P < 0.01; stimulus type, F_1,9 = 50.85, P < 0.0001; load × stimulus type, F_1,9 = 1.31, P > 0.1; P3a load, F_1,9 = 0.37; stimulus type, F_1,9 = 26.33, P < 0.001; load × stimulus type, F_1,9 = 5.64, P < 0.05; RON load, F_1,9 = 1.23, P > 0.1; stimulus type, F_1,9 = 15.35, P < 0.01; load × stimulus type, F_1,9 = 5.19, P < 0.05). It is noteworthy that the main effect of the load factor in the MMN window is because of the decrease in N1 amplitude in the high-load condition for both stimuli (see Fig. 3, left). As the N1 is modulated with the amount of allocation of attention (e.g. Hillyard et al., 1973), this effect corroborates the efficacy of the manipulation of task load also indicated by the behavioural data.

Discussion

The present results show that deviating information is processed when it is task-irrelevant. Moreover, the finding that irrelevant deviations impair performance even in the high-load condition indicates that the working memory system did not completely protect current task-related processing from distraction triggered by the preattentive change detection system. This vulnerability to distractors enables the organism to react adequately to potentially important but unexpected changes in the environment even under higher task demands. By contrast, performing tasks with more complex demands might become impossible if ongoing mental processing is disrupted too easily. Obviously, these two functions, namely the maintenance of distractibility and the focusing on task relevant information, should be coordinated in order to be able to realize adaptive behaviour. In other words, the working memory system would need some control over involuntary attention switching. Indeed, this is the case as indicated by the smaller reaction time effects in the high-, relative to the low-load condition. This finding suggests that involuntary attention switching is not fully determined in a bottom-up manner but is under the control of working memory.

The deviance-related ERP effects (MMN, P3a, RON) give a clue about the stage at which the deviance-related processing is suppressed by the working memory system; in other words, they indicate at which stage the control of the working memory system over the involuntary attention occurs. The finding that MMN was not modulated by task load suggests that the working memory system did not exert top-down control over the activity of the initial change detection system. This result is consistent with findings showing that higher-level cognitive processes related to expectancy or conscious prediction of auditory events do not play a role in the MMN-generating process (Ritter et al., 1999; Rinne et al., 2001; Sussman et al., 2003) although the sensory input to the MMN system might be sensitive to top-down influences (e.g. Sussman et al., 2002). However, the working memory system did partly suppress the consequences of the preattentive change detection system; that is, the involuntary orienting of attention indicated by deviance-related ERP effects following the MMN. The P3a and RON – and also the behavioural distraction effect – were reduced with an increase in task load. This limited control over involuntary attention switches at stages subsequent to the initial change detection system most probably helps to focus on the task-relevant information under higher situational demands without losing the ability to scan the environment preattentively. This could be useful as large (presumably important) distractors would still be able to elicit a call for attention.

How does the working memory system execute its control over distraction? The presence of involuntary attention switch effects in both conditions and the absence of an effect on the latency of the deviance-related ERP components between conditions suggest that working memory does not influence the elicitation and timing of the processes of the preattentive change detection system. The balance between working memory functioning and distraction seems rather to be achieved by the extent of how resources are divided between the primary task and the perturbing event.

Our results show that the ability to react flexibly to unexpected and potentially important events is preserved even when attention is focused on another task. Interestingly, the balance between distraction by involuntary attention switches and focusing on task demands seems not to be fixed but adjustable to the situational demands posed by the primary task. In our opinion, this coordination of resources justifies the concept of executive functions within working memory (Baddeley, 1996; Engle et al., 1999). Moreover, together with the findings that working memory might also control selective attention in the visual domain (Downing, 2000; De Fockert et al., 2001), our findings suggest that there is a close link between working memory and attention, as proposed by the controlled attention view of working memory stressing the importance of the coordination of voluntary and involuntary attention as an important feature of working memory (Engle et al., 1999). In addition to brain imaging studies that investigate brain structures involved in attention and executive working memory functions (Smith & Jonides, 1999), our approach elucidates the timing of different processes of attentional orientation and reorienting in the context of distraction. As the P3a and RON most probably include frontal generators (Escera et al., 2000; Friedman et al., 2001), our finding is compatible with the idea that the frontal lobes are involved in central executive functions of the working memory system (Smith & Jonides, 1998; Engle et al., 1999).

Acknowledgements

We thank D. Friedman and M. Tervaniemi for helpful comments on this paper and N. Wetzel, T. Ohlenbusch, and T. Horenkamp for help during data acquisition. This study was sponsored by the Working Memory Study Group of DFG (German Research Foundation) at the University Leipzig.

Abbreviations