In order to investigate lapses in attention during a sustained boring attention task Robertson et al (1997) formulated The Sustained Attention to Response Task. The task involves a computer generated single occasional target presenting alongside frequent non – targets. Participants are required to respond to the non-targets and inhibit their response to the target. This task requires constant sustained attention to ensure that the correct non-response is selected for the target. Additionally the task is designed to be very boring. Robertson et al (1997) analysed the reaction times for each set of four correct presses, preceding the correct withholding response to the target with the reaction times of four correct presses preceding an erroneously pressed target. They found that everyone displayed faster reaction times before an error in the low probability condition i.e. infrequent non-targets, while reaction time after an error was slower. They found that in the high probability condition there was less of a tendency to go into autopilot as the frequent
presentation of a non-target supported the participants’ ability to maintain attention.
Robertson et al (1997) investigated and found a significant correlation with SART performance and the Cognitive Failures Questionnaire, which is a 25-item measure of self-reported daily errors (Broadbent, Cooper, Fitzgerald & Parkes, 1982). Respondents indicate on a five-point scale if they have ever experienced a specified cognitive failure. Robertson et al (1997) found that a significant number of respondents who scored high on the CFQ, indicating a high rate of cognitive failures, were also both quicker in response to non targets before an error in SART and did not slow down as much after an error (see Manly et al, 1999). Both measures correlate highly thus giving credence to their reliability and validity in measuring lapses of attention.
Arthur, Barrett & Alexander (1991) provide evidence to suggest that high scores on the CFQ are related to the occurrence of accidents. This is of particular interest in the light of the safety implication of absentminded errors. It must be noted that Larson, Alderton, Neideffer & Underhill (1997) have pointed out that the accident and CFQ scores are self-report measures, which are open to participant hindsight evaluations. Despite this criticism, the CFQ is still regarded as a valid and reliable measure of cognitive failures (Manly, Robertson, Galloway & Hawkins, 1999.
It is interesting to note that Martin and Jones (1984) believed that stress induced mild psychiatric symptoms are associated with cognitive failures. Hence reducing stress in high-risk occupations (i.e. occupations in which lapses in attention could be catastrophic), which are by definition stressful because of their important safety role, must be imperative.
The intention of this study is to replicate the findings of Robertson et al, 1997 to investigate whether attention slips are preceded by faster reaction times before an error and slower reaction times after an error with high Cognitive Failures Questionnaires scores acting as forecaster of SART errors of erroneously pressing a the number three. The following hypotheses are proposed:
Hypothesis 1- there will be faster reaction times before an error in the low probability condition.
Hypothesis 2 – reaction times will be slower after an error is made.
Hypothesis 3 – Those participants who score high on the Cognitive Failures Questionnaire will display significantly faster reaction times than all others.
Hypothesis 4 – After an error those participants scoring high on the Cognitive Failures Questionnaire will not significantly slow down their reactions as much as the low scorers.
Words 1079
METHOD
DESIGN
The design was a between groups and within groups study. The participants’ task was randomly assigned to a SART measure. The within groups design was used by all SART groups responding to a response task. The between groups design was used when the 4 groups were separated into 2 additional groups, of which groups 1 & 2 responded to an infrequent show of the target - number three, 10% of the total numbers presented, while groups 3 & 4 responded to a more frequent show of the target – number three, or 50%.
PARTICIPANTS
Sixteen participants were collected by opportunity sampling methods. There were twelve females and four males, all from Glasgow Caledonian University. All were assumed to have no visual impairments that could not be corrected with the use of spectacles, although this was not explicitly asked. In order to assess whether age or gender had an incidental effect on scores, information on both was taken and analysed to act as a control factor.
APPARATUS
The participant completed the experiment in a quiet room with the use of a personal computer. The computer had four versions of the Sustained Attention to response Task (SART) on the desktop. The SART programme asked the participant to press the space bar, with their favoured hand, on the presentation of every number between one and nine except on presentation of the number three.
This experiment included one informed consent form (filled in after reading the computers instructions) and a Cognitive Failures Questionnaire which measured the rate of cognitive failures in the participants recent past (Broadbent at al, 1982). This measure is regarded as a reliable and valid measure (Larson, Alderton, Neideffer & Underhill, 1997).
Participants were randomly assigned to either group one or two or group three or four. Each session was preceded by a practice session.
SART 3 & 4 introduced the number three at the rate of one stimulus every two seconds.
PROCEDURE
Before the task began, each participant was sat before the computer. The individual’s initials were marked on the participants’ form, which indicated the version of SART they were to complete.
The overall procedure for the task was explained, including the need to complete the questionnaire when the computer prompted them. Verbal standardised instructions were not used as the computer programme clearly explained all the relevant details in a standard format. The participant was told to type their initials into the computer when prompted. There then followed a practice block of 20 stimuli containing 2 targets. There were 6 experimental blocks containing 80 stimuli, 3 of which were in the low probability condition (20%) and 3 of which were in the high probability condition (50%). The stimuli were presented at the rate o 1 stimuli every 950 milliseconds. The inter stimulus interval was 250 milliseconds. The stimulus order was counter balanced in an AB design.
After completion of the task they were asked to leave the door open. The participant was then fully debriefed.
RESULTS
For the raw data chart see appendix i. In order to assess whether gender or age had an influence on the Cognitive Failures Test the following analysis were performed.
Table 1: descriptive statistics for age over two levels of age
In order to assess whether age had an influence on the Cognitive Failures Test scores, namely if the scores would vary significantly depending on whether one was in the young or old condition, a t–test for independent groups, was utilised with the following results: t = -.325,df = 14, ns. The results indicate that age does not affect scores on the Cognitive Failures Test.
Table 2: Gender CFQGP Crosstabulation
In order to analyse whether gender had a significant effect on CFQ scores a Chi square test was used with the following results: χ2 = 4.8593, 1df, p>0/05.
Table 3: reaction time before an error in high probability and low probability conditions.
A 2 by 2 Mixed ANOVA was used to test whether reaction time before an error in both the low and high probability conditions and whether there were any significant interactions in both conditions to the Cognitive Failures Questionnaire scores. Table 3 shows the mean scores indicating the low probability condition was faster before an error (.370lp compared to .342hp; .394lp compared to .343hp).
The statistical results of the 2 by 2 Mixed ANOVA was as follows: in the within subjects design there is a significant main effect in times before an error (F (1, 14) = 6.011, p <0.05). In the between subjects design there is no effect (f (1, 14) = .380, ns). Additionally, there was no interaction between the high and low probability conditions and the Cognitive Failures Questionnaire scores (F (1, 14) = .492, ns).
Table 4: CFQGP and reaction time before and after an error in the low probability condition.
A Mixed ANOVA was used to investigate whether reaction time before and after an error in the low probability condition would differ and whether scores on the CFQ would be related. The results were as follows: in the within subjects design there was a significant effect of time (f = (1, 14) = 15.282, p < .05). In the between subjects design there was no significant effect (f (1, 14) = .289, ns). Taking in consideration with table fours similar mean scores for low before and after this indicates that after an error there was a slowing down of response. Table four also indicates that there was a slight difference between CFQ 1 & 2. The results of the 2 by 2 ANOVA show there was no significant interaction (F (1, 14) = 4.507, ns)
Table 5 : reaction time before and after an error in the high probability condition
A Mixed ANOVA was used to investigate reaction time before and after an error in the high probability condition. Table 5 shows mean reaction time after an error in the high probability condition increased in both CFQ groups although CFQ group 2 was only slightly longer after an error (.370HB4 & .488haf compared to .394haf & .397haf).
The result of the ANOVA was as follows: there was a main effect of time (f (1, 14) = 5.436, p < .05). The between subjects design did not have a main effect (f (1, 14) = .211, p>0.05). The results revealed time and scores on the CFQ showed no significant interaction (f (1, 14) = 4.525, ns).
DISCUSSION
The main results of this study were that hypothesis 1, there will be faster reaction times before an error in the low probability condition, was supported (F = (1,14) = 6.011, p = <0.05). There was a slowing down of response after an error in all the conditions (f = (1, 14) = 15.282, p < .05) supporting hypothesis 2. This is in agreement with both Robertson et al
(1997) and Manly et al’s (2001) studies. This also demonstrates support for Robertson’s results in that those participants in the low probability condition (infrequent non-response target) did indeed experience a greater degree of automaticity and subsequent attentional slips as evident in the faster reactions times before an error.
However there was no support for hypotheses 3 & 4. High scorers on the CFQ did not display significantly faster reaction times (f (1, 14) = 5.436, ns) or significantly slow down less after an error (F = (1, 14) = .492, p = ns). It must be noted that the fact that the interactions just failed to meet the significance criteria indicating that there was movement towards the replication of Robertson et al (1997) results. This implies that hypothesis 4 is almost supported, that is, the high CFQ group do tend to not slow down as much, after an error, as the low CFQ group. As already mentioned, Robertson et al (1997) found that a significant number of respondents who scored high on the CFQ, indicating a high rate of cognitive failures, also scored high in lapses of attention in SART (see also Manly et al, 1999). Their results were not fully supported by this study, although it came close to significance.
An independent group’s t-test and a chi-square was performed in order to test for gender or age effects in the CFQ; there were none indicating it is a specific measure of cognitive failures only. As the validity and reliability of the Cognitive Failures Questionnaire is verifiable, but the hypotheses have been only partially supported then it is reasonable to conclude that something else must have been the cause of failure to replicate Robertson’s findings. As the measures of SART are significant, yet the Cognitive Failures Questionnaires scores fail to show a significant interaction then it is also reasonable to conclude whatever went wrong related to the CFQ.
Possible reasons for this failure may be that participants did not answer truthfully or failed to remember slips of attention. As attention lapses may be because of a higher stress rate (Martin and Jones, 1984) then it may be that either the participants did not encode the memory of their slip into long term memory or that they were too stressed during the task to answer accurately (unlikely although not totally impossible). An alternative reason could be that participants did not respect the ethics of the experiment and may have gone into the room with a friend and failed
to give full attention to the questionnaire, resulting in an incorrect score leading to placement in the wrong CFQ group.
Further most participants were psychology students and they may have not wanted to disclose any information that made them appear (even though it was anonymous) stupid. Demand characteristics such as trying to please the experimenter, may have been in action. Furthermore, participant expectations as to the purpose of the study may have influenced their reactions in an unknown manner.
Responses in the SART may have been corrupted by variables unknown. As in completing the CFQ, these could have been, bringing another person into the room when the experiment was in progress or perhaps previous participants loitering and talking loudly outside the room used for the experiment.
It is admitted that the reason for the failure to replicate Robertson et al’s (1997) results are not known. Consideration of the original study (Robertson et al, 1997) and the methodology and comparing both studies for small (but maybe significant) differences may be of help in determining what, if anything, went wrong. Some possible explanations for failure to replicate Robertson et al’s study have been examined. Alternatively, and in all probability, the most likely answer for the failure to replicate his study is the size of the sample. Sixteen people are too small a sample to achieve significant effects. This is the most feasible reason for failing to achieve significant results. The fact that significance was almost achieved supports this point.
The failure to replicate Robertson et al (1997) has meant that time has not been spent discussing the safety implications of the SART and CFQ. Investigating SART sensitivity to CFQ scores may yield insight into how to deal with boredom in repetitive jobs that could have disastrous effects if responses become automatic. Consequently as lapses in attention sometimes have catastrophic consequences, the need for ongoing investigation into the causes of absentmindedness is still imperative (Reason, 1984).
Words 814
REFERENCES
Arthur, W., Barnett, G. V., & Alexander, R. A. (1991). Prediction of vehicular accident involvement: A meta-analysis. Human Performance, 4, 89-105. Cited in Larson, G. E., Alderton, D. E., Neideffer, M. & Underhill E. (1997). Further evidence on dimensionality and correlates of the Cognitive Failures Questionnaire. British Journal of Psychology, 88, 29-38.
Bernstein, D., Penner, L. A., Clarke – Stewart., & Roy, E. R. (2003). Psychology (6th ed.). New York: Houghton Mifflin Company.
Broadbent, D. E., Cooper, P. F., Fitzgerald, P. & Parkes, K. R. (1982). The Cognitive Questionnaire (CFQ) and its correlates. British Journal of Clinical Psychology, 21, 1-6.
Cardwell, M. (1996). The Complete A – Z Psychology Handbook. Oxon: Hodder & Stoughton.
Larson, G. E., Alderton, D. L., Neidefer, M., & Underhill, E. (1997). Further evidence on dimensionality and the Cognitive Failures Questionnaire. British Journal of Psychology, 8, 29 – 38.
Manly, T., Robertson, I. H., Galloway, M. & Hawkins, K. (1999). The absent mind: further investigations of sustained attention to response. Neuropsychologia, 37, 661 – 670.
Martin, M. & Jones, G. V. (1983). Distribution of attention in cognitive failure. Human Learning, 2, 221 – 226. Cited in Martin, M. & Jones, G. V. (1984). Cognitive Failures in Everyday Life. In J. E. Harris & P. E. Morris (Ed.), Everyday Memory, Actions and Absent-Mindedness, pp. 173-190. London: Academic Press.
McNay, E. C., McCarty, R. C., & Gold, P. E. (2001). Fluctuations in brain glucose concentration during behavioural testing: Dissociations between brain areas and between brain and blood. Neurobiology of Learning and Memory, 75, 325 – 337. Cited in Bernstein, D., Penner, L. A., Clarke – Stewart., & Roy, E.R. (2003). Psychology (6th ed.). New York: Houghton Mifflin Company.
Reason, J. & Mycielska, K. (1982). Absent-Minded? Englewood Cliffs, New Jersey: Prentice-Hall.
Reason, J. (1984). Absent-mindedness and Cognitive Control. In J. E. Harris & P. E. Morris (Ed.), Everyday Memory, Actions and Absent-Mindedness, pp. 113 – 132. London: Academic Press.
Reason, J. (1990). Human Error. New York: Cambridge University Press. Cited in Cohen, G. (1996). Memory in the Real World (2nd Ed.). East Sussex: Psychology Press.