Elsevier

Cortex

Volume 44, Issue 7, July–August 2008, Pages 794-805
Cortex

Research report
The multiple dimensions of sustained attention

https://doi.org/10.1016/j.cortex.2007.04.002Get rights and content

Abstract

Sustained counting (or temporal numerosity judgements) has been one of the key means of investigating anterior attentional processes. Forty-three patients with localised lesions to the frontal lobes were assessed on two tests of the ability to count the number (8–22) of stimuli presented at either a slow (roughly one per 3 sec) or fast (roughly three per sec) rate. Patients with lesions to the Superior Medial (SM) region (particularly Brodmann areas 24, 32, and 9) were impaired both in the Slow condition and also in the Fast condition, where they underestimated the number of stimuli. Patients with Right Lateral (RL) lesions (8, 45, and 46) also had difficulties in the Fast condition, especially when the number of targets was greater than 15. The results are considered from the perspectives of alternative positions on anterior attentional processes developed by Posner and Petersen (1990) and by Stuss et al. (1995). The most plausible interpretation is in terms of energising processes which involve the SM frontal cortex and monitoring processes which involve the RL frontal cortex.

Introduction

Attention varies on a number of dimensions (see Evans, 1970). It may be voluntarily or spontaneously drawn to an object, in the latter case either under conscious control or not. Attention may be oriented towards action, thought or perception. It may be shared between activities or concentrated on one task, and for perception it may be narrowly focused on one part of space or diffusely oriented towards a wide area.

To what extent do these different aspects of attention involve the same or different neural mechanisms? Posner and Petersen (1990), developing ideas from Posner (1978), innovatively proposed the existence of three interacting neural systems of attention. The first was the orienting system which directed attention spatially to critical stimuli. A second system was initially characterised in terms of response selection, for instance for focused attention in selection-for-action, as in Stroop tasks, or when rapidly occurring signals have to be responded to, as in serial reaction time paradigms. Later, this second system became characterised as ‘executive attention’ (Posner and DiGirolamo, 1998, Fernandez-Duque and Posner, 2001). A third system was the ‘alertness’ system, which ensured the alert state necessary for carrying out tasks efficiently when vigilance is needed because environmental stimulation is low. Alerting can be ‘phasic’ as in increases in response to a warning signal or ‘tonic’ as in decreases when low levels of stimulation lead to problems in vigilance situations.

The three components of attention were initially characterised as ‘systems’ with discrete localisations: orienting in the parietal lobes, superior colliculi and pulvinar, response selection in the anterior cingulate (ACG), and alertness in the Right Lateral (RL) prefrontal region. In later formulations (e.g., Fan et al., 2003, Fan et al., 2005) the components of attention were described as ‘networks’ with broader localisations: orienting became parieto-frontal, the response selection (or as it was now called ‘executive’) network became more extensively medial frontal than just ACG plus lateral prefrontal, and the alerting network became right fronto-parietal (Fernandez-Duque and Posner, 2001, Fan et al., 2003). It is unclear if these different localisations are meant to be the evolution of the original hypothesis or a distinct new hypothesis about localisations. The most recent empirical evidence for three networks comes from neuroimaging (Fan et al., 2005). The Attention Network Test (ANT) contains three variables which are held to be independently related to the functioning of the orienting, executive and alertness networks (Fan et al., 2002). The orienting variable activated parietal sites as expected but also the frontal eye fields. The executive control variable activated the ACG but also several other brain areas. The alerting variable activated the thalamus and “anterior and posterior cortical sites” specifically right prefrontal.

That spatial orienting involves the inferior parietal lobes, particularly on the right, is supported by both functional imaging of normal subjects and lesion studies of the spatial neglect syndrome (see e.g., Corbetta and Shulman, 2002, Mort et al., 2003, Driver et al., 2004; but see also Karnath et al., 2001). However we have proposed an alternative method of characterizing more anteriorly located systems relevant to attention as a cluster of processes. In our initial theoretical paper on this topic (Stuss et al., 1995), we only suggested some possible localisations of these attentional processes and noted the similarities and differences of functional characterisation of the systems with the Posner and Petersen proposals. In subsequent papers, we have defined some specific and consistent anatomy for these processes. Two of these processes are critical in the present context and these correspond closely anatomically to the localisations of Posner and Petersen's alerting and executive systems. One of our systems is held to be ‘energising’; it is required for initiating and maintaining preparation to respond and for sustaining the intention to respond (Stuss et al., 2002b, Stuss et al., 2005, Alexander et al., 2005) for any action which is not highly over learned or which must occur at a particular time. After any action schema has been activated sufficiently to be selected for operation, its continuing operation after its initiation requires energization. This ‘energising system’ may correspond to the ‘cognitive effort’ system (Hockey, 1993) proposed within information-processing psychology. Evidence that lesions of superior, medial frontal structures including, but possibly not limited to the ACG, impair “energising” processes is provided by the grossly slowed performance of such patients in a variety of reaction time tasks (Stuss et al., 2002b, Stuss et al., 2005, Alexander et al., 2005). Moreover manipulations that slow reaction time in the normal subject show it even more in Superior Medial (SM) patients (Stuss et al., 2005).

A second critical process we proposed to be that of the control of active monitoring or checking of the consequences of one's actions or of the state of the environment against the goals of the task, so that compensatory action can be taken if they deviate (see Shallice, 2006). We held this process to involve predominantly the RL prefrontal cortex. From the perspective of the Posner–Petersen theory there is evidence from both lesion studies and fMRI for a role of right prefrontal structures in “alertness”, from studies where vigilance is required because the rate of stimulus presentation is low (Pardo et al., 1991, Wilkins et al., 1987, Rueckert and Grafman, 1996, Paus et al., 1997, Fernandez-Duque and Posner, 2001; see also Coull et al., 1996). A recent fMRI study of simple reaction time to signals occurring at a rate of once every 3–5 sec again demonstrated activation of a right fronto-parietal network (Mottaghy et al., 2006); however in this case the anterior cingulate was viewed as having a central coordinating role for the network, a point we will return to later. However, a failure in vigilance tasks can also be explained on the Stuss et al. theory as due to a problem in active monitoring. Moreover, lesions of the RL, prefrontal cortex also lead to a number of other impairments of attention and action that seem more compatible with a disturbance in a second anterior attentional process concerned with actively monitoring the appropriateness of the on-going course-of-action. Thus, we have shown how lesions to this region lead to diminished sensitivity to distinctions between targets and non-targets (Stuss et al., 2002b) and to poor control of timing behaviour (Picton et al., 2006). In addition we have shown that such lesions lead to problems in the variable foreperiod reaction time paradigm in which the time between a warning signal and the target stimulus varies (Stuss et al., 2005; see also Vallesi et al., 2007 for confirmatory TMS findings). Normal subjects show a speeding of the response when the foreperiod is longer; RL patients show no such effect. When the foreperiod is long, the patient is held to fail to actively monitor during the foreperiod that no stimulus has yet occurred and so does not increase preparation appropriately for later-occurring stimuli: no analogous deficit occurs if the foreperiod is fixed.

There is, in addition, evidence from functional imaging that the RL prefrontal cortex is involved not only when increasing alertness is required because the subject is in a vigilance situation but also that it has a key role in active monitoring in fully alert subjects, such as checking memory judgements (e.g., Henson et al., 1999, Henson et al., 2000; see Shallice, 2002, Shallice, 2004 for discussion). In line with these findings are analogous results in neuropsychological and TMS studies. Patients with RL frontal lesions make excessive repeated responses in free recall of word lists (Stuss et al., 1994) and many more capture errors in a concept attainment task (Reverberi et al., 2005), both behaviours suggesting a lapse in the monitoring of the suitability of on-going behaviour. Moreover, rTMS to RL prefrontal regions leads to subjects failing to set criteria appropriately in memory recognition studies and so to make many false positive errors (Rossi et al., 2001).

The models of Posner and Petersen and Stuss et al. have in common that there are at least two anterior, attentional processes involving SM and RL structures, but they differ in the characterisation and localisation of the processes. To address these unresolved issues, we returned to one of the critical experimental tasks: sustained counting of a string of clicks presented at a slow rate. In this task, there are at least two procedures available for determining how many items have been presented in a string. One is by estimation, which both in monkeys and in humans seems to involve a fronto-parietal system, which is predominantly right hemisphere based in humans (Nieder et al., 2006, Piazza et al., 2006). The other is counting, which involves individuation of each item in the set, the assignment of an attentional index to each item and then the use of phonological codes to keep a running total (Gelman and Gallistel, 1978). In sustained counting paradigms the optimal strategy is indeed to count, as estimation leads to a higher error rate (Cordes et al., 2001).

In the sustained counting task the rate of stimuli is much slower than the optimal rate for counting of one per second and the number of clicks is well above that available from any subitising process based on the contents of the immediate present. The critical challenge of the sustained counting task is then of maintaining accuracy when an essentially simple task has to be performed at a much slower rate than optimal for a time that is sufficiently long such that alertness may wane. Thus we will argue is a task loading on vigilance, which we operationally define as a task which is cognitively simple and yet which shows deterioration over blocks in its performance (see Sanders, 1998). Moreover, by contrasting performance on the Slow condition with that on a Fast condition where the rate of presentation is an order of magnitude faster, a control task can be obtained which differs only in its attentional requirements (vigilance or high-rate processing). A model such as that of Posner and Petersen would predict different patterns of deficits in the two conditions.

The functional imaging data on sustained counting judgements when alertness is stressed also tend to support a RL prefrontal component. For instance, when subjects had to count touches to the foot that occurred at a very slow (roughly once every 20 sec) rate, there was increased activation of RL prefrontal cortex (Pardo et al., 1991). However, Ortuno et al. (2002) found no significant increase in activation of prefrontal areas comparing a counting task at an easy one per sec rate to an uncontrolled second task, namely simply listening to the clicks.

Lesion studies also suggest a role for the right prefrontal cortex in sustained counting. Wilkins et al. (1987) investigated sustained counting with rates of either one or seven per sec. The low rates are cognitively undemanding, so maintaining vigilance (maintaining alertness in the terminology of Posner and Petersen) appears to be the main process required for correct responding given a slow, long train (10–22 targets). When the presentation rate was slow, there was a significantly greater impairment for patients with anterior compared with posterior lesions in the right hemisphere but not in the left, but the result was not clear because the difference between the effects of left and right frontal lesions was not significant. This result thus provides some support for the Posner and Petersen position of a right frontal alerting system.

The two theoretical frameworks for the specification and localisation of components of attention outlined above represent different ways of describing two distinct processes of attention which are localised in the RL and SM prefrontal cortex, but they differ in how the theories characterise the operations of the two brain regions and the predictions they make concerning the sustained counting paradigm. First, as far as the RL prefrontal cortex is concerned, any role of boosting alertness as on the Posner–Petersen theory would predict a deficit in the Slow condition rather than the Fast one. Active monitoring, the Stuss et al. position, makes no such prediction; monitoring is as relevant, if not more so, in highly demanding rapid tasks as in vigilance conditions. Thus the Stuss et al. theory would see RL system as involved in both the Fast and Slow conditions.

Second, as far as the SM structures are concerned, an executive role as in the Posner–Petersen account would see them as being required more in the more demanding Fast condition than in the Slow one. However, on the energising function which is ascribed to these regions by the Stuss et al. theory, they would be required in both the Slow and the Fast conditions. Energising is held to be needed when the rate of occurrence of stimuli differs from the optimal one for the lower-level systems operating without top–down control, namely unmodulated contention scheduling (Norman and Shallice, 1986). This can be because the rate of presentation is too low, leading to insufficient exogenously induced arousal or because it is too high when the demanding nature of the task also entails increased cognitive effort.

The present study in neurological patients of the effects of lesions in different parts of prefrontal cortex on sustained counting attempted to take advantage of some of the methodological lessons from earlier studies. To compare cognitively demanding situations – rapid counting – to ones requiring vigilance, there are two contrasting rates of presentation and long trains of stimuli are compared to short ones. The slow rate was set at one per 3 sec to increase the demand for maintaining attention. This is a slower rate than the one per sec one used in the previous study in view of the negative results of others using that rate (Ortuno et al., 2002). The fast rate (three per sec) is slower than the corresponding rate in the earlier study (Wilkins et al., 1987) to allow higher accuracy. The tasks were presented twice, with the two Slow condition blocks following one another to check that that condition had a vigilance component as planned; conditions which stress vigilance lead to a decrement in performance over blocks (Sanders, 1998). Moreover, longer strings are compared with short strings as the former clearly loads more on any vigilance component. Prior studies have not analysed error types so we have expanded the assessment of errors to identify any systematic underestimation or bias in strategy.

To analyze the effects of lesion sites, we classified the lesions in much greater detail than the left versus right dichotomies of prior studies. The frontal lobes contain four major regions based on differences in cortical and limbic connectivity; these functional subdivisions are readily identified for analysis of lesions. We have demonstrated ample evidence to support this approach (Stuss et al., 1995, Stuss et al., 1998, Stuss et al., 2002a). When there is a significant effect for a large region, a second analysis is made of the relative significance of individual architectonic regions (Petrides and Pandya, 1994).

Section snippets

Patients

Forty-three patients with a focal frontal lobe lesion with no evidence of diffuse brain pathology were assessed in the post-acute stage of recovery (range, 2–109 months post-onset, M = 22 months) together with the same number of control subjects (see Table 1). Aetiologies were all acute acquired disorders: infarction, haemorrhage (including ruptured aneurysms), trauma, and resection of a benign tumour. There were 4/11 trauma patients in the left frontal group and 8/15 in the Inferior Medial (IM)

Overall analysis

The basic results for stimuli presented at the slow rate are shown in Fig. 1 and for the fast rate in Fig. 2. An analysis of the whole set of findings with five groups and three different within-group factors (slow/fast; first/second block; and high/low number) using the statistical procedure described above was carried out on the mean Number Correct for each subject. There was a main effect of speed of presentation, performance on the slow rate being better than on fast rate [F(1,76) = 31.70; p < 

Discussion

Patients with lesions to certain regions of the frontal lobe have significant difficulties in counting the number in a train of auditory stimuli when stimuli are presented either at a slow or at a fast rate. This was established in a group of patients who showed no deficits on baseline neuropsychological tests which do not stress executive functions, vigilance or rapid responding. Although certain of the patients had traumatic lesions which could have created diffuse axonal injuries, these were

Conclusion

These findings extend our understanding of the role of prefrontal structures in three ways. They support Posner and Petersen's model that SM structures, including the anterior cingulate, and the RL region play key roles in attention. However, Posner and Petersen's proposal that the RL frontal region is critical for low rates (vigilance) is not confirmed. Lesions in SM structures impair both high and low rates because both tasks require an attentional energising mechanism. We previously proposed

Acknowledgements

This study was funded by the Canadian Institutes of Health Research, #MT-12853 and #MRC-GR-14974. M. Alexander is partly supported by NIH (NS 26985), the Heart and Stroke Foundation Centre for Stroke Recovery and the Louis and Leah Posluns Centre for Stroke and Cognition at Baycrest. T. Shallice is partly supported by a PRIN grant. D. Stuss is the Reva James Leeds Chair in Neuroscience and Research Leadership. T. Picton is the Anne and Max Tanenbaum Chair in Cognitive Neuroscience. We are

References (48)

  • D.T. Stuss et al.

    Wisconsin card sorting test performance in patients with focal frontal and posterior brain damage: effects of lesion location and test structure on separable cognitive processes

    Neuropsychologia

    (2000)
  • A.J. Wilkins et al.

    Frontal lesions and sustained attention

    Neuropsychologia

    (1987)
  • M.P. Alexander et al.

    Impaired concentration in patients with frontal damage: deficits from two anatomically distinct lesion sites

    Neurology

    (2005)
  • M. Corbetta et al.

    Control of goal-directed and stimulus-driven attention in the brain

    Nature Reviews Neuroscience

    (2002)
  • S. Cordes et al.

    Variability signatures distinguish verbal from nonverbal counting for both large and small numbers

    Psychonomic Bulletin & Review

    (2001)
  • H. Damasio et al.

    Lesion Analysis in Neuropsychology

    (1989)
  • J. Driver et al.

    Spatial neglect and extinction

  • P. Elsass et al.

    Reaction time and brain disease: relations to location, etiology and progression of cerebral dysfunction

    Acta Neurologica Scandinavica

    (1985)
  • C.O. Evans

    The Subject of Consciousness

    (1970)
  • J. Fan et al.

    Testing the efficiency and independence of attentional networks

    Journal of Cognitive Neuroscience

    (2002)
  • D. Fernandez-Duque et al.

    Brain imaging of attentional networks in normal and pathological states

    Journal of Clinical and Experimental Neuropsychology

    (2001)
  • R. Gelman et al.

    The Child's Understanding of Number

    (1978)
  • R.N.A. Henson et al.

    Confidence in recognition memory for words: dissociating right prefrontal roles in episodic retrieval

    Journal of Cognitive Neuroscience

    (2000)
  • R.N.A. Henson et al.

    Recollection and familiarity in recognition memory: an event-related fMRI study

    Journal of Neuroscience

    (1999)
  • Cited by (104)

    • Fluency test generation and errors in focal frontal and posterior lesions

      2021, Neuropsychologia
      Citation Excerpt :

      Within frontal lobe functions, Stuss and colleagues have conceptualized this process as energization, based on “neurophysiological observations that there is an internal tendency for any neural activity to become quiescent in the absence of input.” ( p.903, Stuss and Alexander, 2007; also referred to as cognitive effort - Shallice et al., 2008). Energization has been associated with the superior medial region, particularly on the right (Stuss and Alexander, 2007), using a simple serial five-choice task from the Rotman-Baycrest Battery to Investigate Attention (RoBBIA; Alexander et al., 2005).

    • Executive/cognitive control

      2021, Encyclopedia of Behavioral Neuroscience: Second Edition
    View all citing articles on Scopus
    View full text