More recent studies have examined the effects of moderate intensity noise on cognitive vigilance tasks .sx An example of this type of task is the Bakan task where single digits are presented on the screen and subjects have to detect a particular sequence such as three successive odd digits .sx Poulton ( 1977 ) suggested that all the detrimental effects of noise could be explained either by the masking of acoustic cues or , in the case of cognitive tasks , by the masking of inner speech .sx The first part of this view has been strongly opposed ( see Broadbent , 1978) .sx The second part led to an interest in the effects of noise on cognitive vigilance , for post-experimental reports from subjects having completed the Bakan task clearly suggest that they say the digits to themselves .sx If noise masks internal speech such tasks should be impaired and this issue was examined by Jones , Smith & Broadbent ( 1979) .sx They carried out four experiments using slightly different versions of the Bakan task .sx In each case there was evidence of a detrimental effect of noise , although the exact nature of the noise effect depended on specific features of the task .sx This , and the finding that the noise had no differential effect on digits which were rehearsed together and those which were rehearsed separately , argues against noise producing effects solely by interfering with subvocal articulation .sx These results do , however , show that vigilance tasks involving memory are susceptible to noise below 95 dB , which is consistent with results obtained by Benignus , Otto & Knelson ( 1975) .sx This finding has also been confirmed in subsequent studies by Miles , Auburn & Jones ( 1984 ) and Smith ( 1988 a) .sx The importance of the nature of the tasks will be discussed again in a later section , after results from other attention paradigms have been summarized .sx Noise and the five-choice serial reaction time task .sx Another task which has been widely used to study the effects of noise is the five - choice serial reaction time task .sx Broadbent ( 1979 ) has reviewed the effects of very loud noise on this task and concluded that noise increases momentary inefficiency , in that there are more errors and gaps ( occasional very long reaction times ) than in quiet .sx Again , this increased momentary inefficiency depended on the noise being over 95 dB and the subject being in the noise for at least half an hour .sx Recent research has shown that a long exposure of moderate intensity noise can produce comparable results to the older studies using levels over 95 dB .sx For example , Smith & Miles ( 1985 ) asked subjects to carry out a serial choice reaction time task when they had been in 75 dB noise for two hours and when they had been in the noise for five hours .sx The results showed that when subjects had been in the noise for five hours there was an increase in the number of errors .sx Poulton ( 1977 ) has tried to explain the effects of noise on the five-choice serial reaction time task in terms of masking .sx The explanation of the early results has been opposed by Broadbent ( 1978 ) , and recent data also argue against it .sx For example , Jones ( 1983 ) was able to confirm the effect of noise on errors and show that this was still apparent after the noise had been switched off .sx It is clear that something other than the masking of acoustic cues was responsible for this effect of noise on the accuracy of performing the task .sx Smith ( 1985 ) and Smith ( 1986 b ) changed the nature of the serial reaction time task and examined the effects of noise on a biased probability task and a version where the stimulus-response arrangement was incompatible .sx The results showed that the effects of noise depended on the task parameters , with noise slowing responses to stimuli which were presented less frequently , but reducing reaction times to those with a high probability of occurrence .sx Incompatible responses were found to be slower in noise and it is difficult to explain this result , or the previous one , in terms of masking of acoustic cues .sx Noise and selectivity in attention .sx In multiple tasks or tasks with several subcomponents , noise often leads to increased concentration upon the dominant or high probability component at the expense of other features .sx For example , Hockey ( 1970 ) showed that noise improved performance on a central tracking task but led to slower reactions to those lights which had a lower probability of occurrence .sx Broadbent ( 1971 ) cites several lines of evidence to support the view that noise influences selectivity in memory and attention .sx For example , studies of the effects of noise on the Stroop task have shown that noise may reduce the amount of interference from irrelevant colour names ( e.g. Houston & Jones , 1967) .sx Hockey & Hamilton ( 1970 ) have also demonstrated that noise aids intentional recall but impairs incidental recall , and Smith ( 1982 ) has replicated this effect using priority instructions rather than the intentional/incidental manipulation .sx Unfortunately , other studies have failed to replicate the effects of noise on selectivity in memory and attention .sx For example , Forster & Grierson ( 1978 ) and Loeb & Jones ( 1978 ) were unable to replicate the results of Hockey ( 1970 ) , although it should be noted that they used tasks with a different level of difficulty from the original study .sx Similarly , Smith ( 1982 ) demonstrated that the effects of task priority are easily modified , which in turn changes the effects of noise .sx Indeed , Smith ( 1982 ) suggested that Broadbent's ( 1971 ) view needs modifying and that noise biases the allocation of effort towards the operation which appears best to repay the investment of effort .sx This may take the form of a bias towards the high priority task but it is also going to depend on other factors such as the difficulty of each part of the task , the subject's prior experience and the salience of the stimuli .sx This suggests that most theories which argue that performance is shifted by noise in an invariant or mechanical fashion are going to be inadequate , and that it may be more profitable to consider the strategies of performance in noise .sx Strategies of performance in noise .sx Smith ( 1983 ) has reviewed evidence which demonstrates that , when subjects carry out a task which can be performed in different ways , noise may lead to the adoption of certain strategies in preference of others .sx If task parameters are changed then the method of doing the task often changes , and this plausibly explains why noise effects vary when different versions of a task are used .sx It is also known that choice of an appropriate strategy can eliminate the detrimental effects of noise .sx Pollock & Bartlett ( 1932 ) carried out an experiment in which subjects were given groups of letters and told to form as many words as possible from these letters .sx At first the subjects were impaired by noise but later they were able to perform as well as in the quiet condition .sx The subjects reported that as time went on they discovered rules and mechanical techniques of solving these problems , and once they had developed these strategies the noise ceased to have an effect .sx In many tasks it is obvious that one strategy should be used in preference to others .sx This may be because of instructions , previous experience , or some other feature of the task .sx Studies by Smith ( 1982 ) and Wilding , Mohindra & Breen-Lewis ( 1982 ) have shown that noise often reinforces the use of the dominant strategy .sx Smith ( 1988 a , 1990 a , b ) has examined resource allocation in noise by studying dual task performance .sx The difficulty of each task was varied , the probability of having to do the tasks was manipulated , and in other conditions one of the tasks had a higher priority than the other .sx The effect of noise depended on the nature of the task , in that monitoring tasks involving active strategies were impaired by noise whereas tasks performed passively were unimpaired .sx Changing task parameters only influenced the noise effect if the two tasks competed for common resources .sx It was also important whether task parameters were constant or changed rapidly .sx This last point is described in more detail in the following section on noise and flexibility of changing strategy .sx There is evidence that noise may influence the efficiency of the control processes which monitor and alter performance .sx For example , Rabbitt ( 1979 ) has suggested that the effects of noise on the five-choice serial reaction time task ( an increase in errors and/or gaps ) can be explained in terms of noise producing inefficient control of the speed-error trade-off function .sx The effects of noise on control processes may be of short duration and reflect initial coping with the task .sx However , other effects may continue even when the noise is switched off ( see Cohen , 1980 , for a review of the after-effects of noise ) and may still be apparent when the subject is transferred to the quiet condition ( see Poulton , 1982 , for an account of asymmetric transfer) .sx Poulton ( 1982 ) considers asymmetric transfer to represent the transfer of the strategy used in one condition to another condition .sx Dornic & Fernaeus ( 1982 ) have shown that subjects in noise are less able to switch between strategies .sx Smith ( 1990 a ) asked subjects to carry out a dual task involving a running memory task and a proportion perception task .sx The difficulty , probability and priority of the tasks changed frequently and noise produced a general impairment of the running memory task .sx Overall , these results suggest that subjects in noise are rather inflexible and unable to deal efficiently with changing task demands .sx Another area which is relevant to the present discussion is the effects of noise on response criteria .sx Broadbent & Gregory ( 1963 , 1965 ) found that very loud noise ( over 90 dB ) produced more confident assertions or denials in the second half of a vigilance task , but reduced the number of reports of intermediate confidence .sx Jones , Thomas & Harding ( 1982 ) examined the effects of noise on recognition memory for prose items .sx They used a signal detection analysis and showed that noise decreased values of beta for rare names and increase beta for common names .sx Smith ( 1989 b ) also found that noise influenced the use of response categories , although , again , the nature of the noise effect was easily modified by changing task parameters .sx Many of the studies of strategies of performance and noise have been carried out with verbal memory tasks , because they offer a variety of strategies , and shifts of dominance or preference can be easily effected .sx An alternative approach is to develop tasks where certain strategies are clearly used and examine whether noise influences the efficiency of the different types of processing .sx This technique was adopted here to investigate whether noise has selective effects on different aspects of attention .sx The following section gives a brief rationale for the choice of tasks in this study of attention .sx Dimensions of attention .sx Many of the details of the mechanisms of selective attention are controversial ( see Broadbent , 1982) .sx However , two main types of task have been used to study selective attention ( see Kahneman & Treisman , 1984 ) , the first involving focused attention or 'filtering' , and the second categoric search or 'pigeonholing' .sx The first type of task involves selecting on the basis of some characteristic simple feature such as location in space .sx The other type requires the person to select a member of a category of events .sx If one is trying to measure the efficiency of attention it is clearly important which kind of task one uses .sx Indeed , Broadbent , Broadbent & Jones ( 1989 ) have shown that the two paradigms respond differently to changes in state ( in their experiment this was examined by comparing performance at different times of day ) and the main aim of the present study was to examine the effects of noise on these different aspects of attention .sx The experiment reported here obtained a number of measures of attention , the principal one being the difference between a filtering and search paradigm .sx The filtering paradigm was based on the method of Eriksen & Eriksen ( 1974) .sx They required subjects to perform a choice reaction time task to visual letters in a known location .sx If distracting non-target letters were also presented , the latency of the response to the target was increased .sx This effect was greatest if the distractors were close to the target , and this result has given rise to the analogy with a zoom lens or spotlight ( Broadbent , 1982) .sx