Elsevier

Neuropsychologia

Volume 36, Issue 12, 1 December 1998, Pages 1355-1362
Neuropsychologia

Executive processes in Parkinsons disease—random number generation and response suppression

https://doi.org/10.1016/S0028-3932(98)00015-3Get rights and content

Abstract

In producing random numbers, subjects typically deviate systematically from statistical randomness. It is considered that these biases reflect constraints imposed by underlying structures and processes, rather than a deficient concept of randomness. Random number generation (RNG) places considerable demands on executive processes, and provides a possibly useful tool for their investigation. A group of patients with Parkinsons disease (PD) and a group of controls were tested on a RNG task, both alone and with a concurrent attention-demanding task (manual tracking). Both groups showed the biases in RNG described previously, including a strong counting tendency and repetition avoidance. Overall RNG performance did not differ between the groups, although differences were found in the counting biases in the patient and control groups, with the controls showing a bias towards counting in twos, and the patients a bias towards counting in ones. The secondary task reversed the bias shown by controls and exacerbated the bias in the patients. A network modulation model may help explain many of the features of RNG. We suggest that naturally biased output from an associative network must be actively suppressed by an attention-demanding, limited-capacity process. This suppression may be disrupted by the pathophysiology of PD and by concurrent tasks. Convergent evidence from various sources is discussed which supports a role of the dorsolateral prefrontal cortex (DLPFC) in this process.

Introduction

In attempting to produce a random sequence of responses, such as numbers, letters or manual responses, human subjects typically exhibit biases that cause their behaviour to deviate from statistical concepts of randomness. As shown by Ginsburg and Karpiuk [13]and others, there is a marked avoidance of repetitions, and a bias towards pairs or longer runs of naturally ordered sequences (e.g. number sequences, letter combinations that follow patterns occurring in language). In addition there is evidence of individual response preferences leading to unequal distributions of responses (e.g. a preference for the number 7), and there is a tendency to cycle through the elements in a response set. It is generally thought that these deviations from statistical randomness reflect structural or processing constraints with the system. Other explanations based on a faulty understanding of the concept of randomness seem unable to account for the empirical findings 34, 36.

However imperfect the results, random response generation still places considerable demands on our limited capacity cognitive processes. Brown and Marsden [4]showed that concurrently generating a sequence of random numbers led to a near doubling of response times in a version of the Stroop task [33]and a four- to five-fold increase in error rates. In contrast, simply articulating a nonsense syllable at a similar rate had no effect. Similarly, Gilhooly et al. [12]showed similar interference with performance on a syllogistic reasoning task. Of course, such interference effects tend to be bi-directional, and random generation performance itself often suffers from resource competition with the concurrent attention-demanding task [1]. Such interference effects are to be expected given the processing demands of random generation. To perform the task subjects must—(1) have on-line access to a series of rules to define randomness (however flawed);(2) based on those rules the subject will adopt and use a particular strategy to either (a) select individual responses and⧹or (b) suppress responses which violate the rules;(3) the subject needs to monitor this output, holding recent responses in memory and compare them to the concept of randomness;(4) when the sequence violates the concept of randomness, subjects must switch or modify their production strategy so that the output conforms more closely to the rules.

Such working memory and executive processes are typically assigned to higher-order cognitive structures such as the central executive [2]or supervisory attentional system (SAS) [21], and random generation may provide a useful paradigm for their investigation. The anatomical substrates of these structures have, in turn, been ascribed to regions of the prefrontal cortex. As reviewed by Brugger et al. [5], random generation performance has been assessed in a number of patients groups that display executive dysfunction. Impaired performance has been reported in patients with schizophrenia [27], Korsakoff syndrome [25], frontal cortical lesions 5, 30, Parkinsons disease (PD) 7, 26, 30and Alzheimers disease [5]. In general, the deficits revealed are an exacerbation of the biases already described in normal subjects; particularly the tendency to produce systematically ordered responses such as counting. For example, Spatt and Goldenberg [30]assessed patients with frontal lesions and patients with PD. Compared to controls, both groups showed an increased tendency to produce stereotyped counting responses. However, not all aspects of randomness were affected. The low incidence of repetitions found in normal subjects was also observed in the patient groups.

Confirmation that prefrontal regions may play an important role in random generation tasks comes from functional imaging studies using positron emission tomography (PET). In a variant of the paradigm, Petrides et al. [23]required subjects to produce randomly ordered sequences of numbers from 1 to 10, but without repeating or missing out any numbers. Compared to a control task involving simply counting 1 to 10, the self-ordered task led to significant regional cerebral blood flow (rCBF) increases in mid-dorsolateral prefrontal cortex (DLPFC) (areas 9 and 46) bilaterally, as well as premotor cortex and regions of posterior parietal cortex (areas 7 and 40). Similar results were shown in a recent study by Jahanshahi et al. [17]with a standard random number generation (RNG) task. Relative to simple counting, RNG led to increased rCBF in left DLPFC (area 9 and 46), anterior cingulate (area 32), posterior parietal cortex (area 7) bilaterally, and right inferior prefrontal cortex (areas 45 and 47). Increased activation in the cerebellum was observed.

Activation of DLPFC, particularly area 46, has been shown on a variety of tasks in which intrinsic response generation is required. These include randomly deciding which finger to move 11, 20, in which direction to move a joystick 6, 24or when to make a movement [18].

However, simply showing that an area is activated in a task does not necessarily tell you what processes it is performing. As described above, random generation potentially involves a whole series of complex executive and working memory processes. Petrides et al. [23], favoured response monitoring as the critical function being performed by the DLPFC in the self-ordered task. However, others have suggested this regions role, particularly on the left, may be more to do with the active suppression of strong habitual tendencies. In a study by Frith et al. [10]rCBF was measured during two verbal fluency tasks, generating names of jobs and words beginning with the letter a. Compared to control conditions, there was increased rCBF in the left DLPFC and decreased activation in the left superior temporal gyrus. To account for these findings, it was proposed 9, 10that the left DLPFC was actively and strategically suppressing unwanted activation in semantic associative networks in the temporal lobes. Production of one word (e.g. farmer) is assumed to lead, through a process of spreading activation across the network, to an increased probability of producing other related words (e.g. barn or tractor). In the constraints imposed by the verbal fluency task such high probability responses may be inappropriate and would have to be actively suppressed to permit a valid response to be produced.

The same model can be readily applied to random response generation tasks [19]. Indeed, in the PET study of Jahanshahi et al. [17]previously mentioned, a decrease in rCBF was observed in the superior temporal cortex, although in this case the effect was observed only for the right hemisphere. Therefore, whether the output is words, letters, numbers, or manual responses, it can be assumed that there is some distributed associative network at a cortical level, and that activation of nodes within that network leads to spreading activation to other associated representations. In a random generation task, it would be necessary (just as in verbal fluency tasks) to inhibit this spreading activation, to prevent stereotyped high probability response sequences from emerging. That such sequences do emerge, despite the subjects efforts, suggests such inhibition is difficult to achieve.

In the remainder of this paper, we will present some data that supports such a model of random generation, and which may offer some further insights into the functional system underlying strategic response suppression in man. The study involved a pair of experimental approaches. The first was a comparison of patients with PD with normal controls. Many studies have shown that patients with PD show a range of executive deficits similar to those seen by patients with prefrontal cortical damage [3]. The favoured explanation for these findings is that they result from disrupted striato-thalamic input to prefrontal regions due to striatal dopamine depletion. Direct confirmation of prefrontal dysfunction in PD comes from PET evidence showing evidence of underactivity of the DLPFC 18, 24during intrinsic response generation tasks. The second approach taken in the present study was to compare random generation with and without a concurrent attention-demanding task. As described above, dual-task paradigms provide a useful way of disrupting capacity-demanding executive processes, presumably at the level of the cortex. The pattern of interference may provide clues as to the cognitive processes underlying human random response generation.

Section snippets

Subjects

Sixteen patients with clinically diagnosed idiopathic PD were assessed (age 61.7±6.1 years, duration of illness 5.8±4.0 years). All had mild-to-moderate clinical symptoms (Hoehn and Yahr [15]stages I–III). Half had withheld their normal antiparkinsonian medication for 12–14 h prior to testing and were in a clinically defined off state, while the remainder had taken their normal prescribed dosage. In practice, medication status had no impact on the random generation findings and the subjects

Discussion

The present study replicates the findings of response biases observed in many previous studies. In generating random number sequences subjects repeated items less, counted more and had a tendency to cycle responses through the series. However, compared to most of the other studies which have looked at RNG in patients with executive dysfunction, including PD, no apparent deficits were found in the current sample on conventional measures of randomness. Several possible explanations can be

Acknowledgements

—This work was supported by the Medical Research Council (UK), the Wellcome Trust, and the European Science Foundation.

References (36)

  • R.G. Brown et al.

    Trends in Neurosciences

    (1990)
  • P. Brugger et al.

    Neurophsychologia

    (1996)
  • C.D. Frith et al.

    Neuropsychologia

    (1991)
  • P.M.J. Pollux et al.

    Neuropsychologia

    (1995)
  • J. Spatt et al.

    Brain

    (1993)
  • A.D. Baddeley

    Quarterly Journal of Experimental Psychology

    (1966)
  • Baddeley, A. D., and Hitch, G., Working memory. In The Psychology of Learning and Motivation, ed. G. Bower, Academic...
  • R.G. Brown et al.

    Brain

    (1991)
  • M.-P. Deiber et al.

    Brain

    (1991)
  • G. Ebersach et al.

    Neuropsychologia

    (1994)
  • F.J. Evans

    Bulletin of the Psychonomic Society

    (1978)
  • K.J. Friston et al.

    Proceedings of the Royal Society of London (Series B)

    (1991)
  • C.D. Frith et al.

    Proceedings of the Royal Society of London (Series B)

    (1991)
  • K.J. Gilhooly et al.

    Memory and Cognition

    (1993)
  • N. Ginsburg et al.

    Perceptual and Motor Skills

    (1994)
  • S. Heywood

    Perceptual and Motor Skills

    (1972)
  • M.M. Hoehn et al.

    Neurology

    (1967)
  • M. Jahanshahi et al.

    Effects on executive function and working memory.

    (1997)
  • Cited by (86)

    • Random number generation deficits in patients with multiple sclerosis: Characteristics and neural correlates

      2016, Cortex
      Citation Excerpt :

      Instructions of a random number generation (RNG) task require subjects to arrange numbers in a sequence “as random as possible”, implicitly asking to avoid any algorithm and to disobey any rule. Under a huge range of conditions (Brugger, 1997) healthy volunteers were found unable to follow these instructions, and so were patients with various neuropsychiatric diseases (Brown, Soliveri, & Jahanshahi, 1998; Brugger, Monsch, Salmon, & Butters, 1996; Ho, Sahakian, Robbins, & Barker, 2004; Salamé & Danion, 2007; Spatt & Goldenberg, 1993). A limited capacity of working memory and executive functions have been implied in the failure to produce unpredictable, or random sequences of response alternatives (Baddeley, 1966, 1998; Joppich et al., 2004; Maes, Eling, Reelick, & Kessels, 2011; Miyake et al., 2000).

    View all citing articles on Scopus
    View full text