Neurocognitive Constructs Underlying Executive Control in Statistically-Determined Mild Cognitive Impairment

Abstract

Background:

The model of executive attention proposes that temporal organization, i.e., the time necessary to bring novel tasks to fruition is an important construct that modulates executive control. Subordinate to temporal organization are the constructs of working memory, preparatory set, and inhibitory control.

Objective:

The current research operationally-defined the constructs underlying the theory of executive attention using intra-component latencies (i.e., reaction times) from a 5-span backward digit test from patients with suspected mild cognitive impairment (MCI).

Methods:

An iPad-version of the Backward Digit Span Test (BDT) was administered to memory clinic patients. Patients with (n = 22) and without (n = 36) MCI were classified. Outcome variables included intra-component latencies for all correct 5-span serial order responses.

Results:

Average total time did not differ. A significant 2-group by 5-serial order latency interaction revealed the existence of distinct time epochs. Non-MCI patients produced slower latencies on initial (position 2-working memory/preparatory set) and latter (position 4-inhibitory control) correct serial order responses. By contrast, patients with MCI produced a slower latency for middle serial order responses (i.e., position 3-preparatory set). No group differences were obtained for incorrect 5-span test trials.

Conclusion:

The analysis of 5-span BDT serial order latencies found distinct epochs regarding how time was allocated in the context of successful test performance. Intra-component latencies obtained from tests assessing mental re-ordering may constitute useful neurocognitive biomarkers for emergent neurodegenerative illness.

Keywords

Digit span executive control intra-component latency mild cognitive impairment temporal organization

INTRODUCTION

Alzheimer’s disease (AD) is an insidious onset neurodegenerative dementia characterized by impairment in cognitive and functional abilities, thought to arise as many as 20 years before the clinical manifestation of symptoms [1 –3]. Between the years 2000 and 2017, there has been a 145%increase in deaths from AD. Without medical breakthroughs to modify, prevent, or cure AD, the number of adults ages 65 and older with AD is projected to reach 13.8 million by 2050 [4, 5]. As such, there is great interest in developing assessment procedures capable of identifying emergent neurodegenerative disease as soon as possible.

Mild cognitive impairment (MCI) is viewed as a prodrome of dementia illnesses including AD and has become an important target for early intervention [6, 7]. Recent neuropsychological and neuroimaging research have been instrumental in disambiguating MCI subtypes such as amnestic and mixed/dysexecutive MCI [8 –12]. Further, digital biomarkers have gained considerable traction in identifying and diagnosing cognitive decline associated with MCI [13 –19]. While changes in memory are commonly associated with the early emergence of cognitive decline [6], recent research shows that executive functions may also be a marker for emergent cognitive impairment [20 –23]. In the current research, memory clinic patients were assessed using a standard neuropsychological test combined with digital technology to elucidate long-advocated theoretical constructs that underlie executive functions [12] to show how these constructs can be useful as potential digital biomarkers for emergent neurodegenerative illness.

The Fusterian model of executive attention

Executive control is a complex, multifactorial, top-down mental process involving attention and concentration, inhibition/self-control, working memory, interference control, mental manipulation and flexibility, and concept formation, that allow us to utilize reasoning, problem-solving, and mental planning skills for effective responding [24 –26]. The model of ‘executive attention’ [12] is built on a framework by which neural networks integrate and temporally organize information to optimize goal-directed behavior. The neural scaffold upon which ‘executive attention’ [12] functions are mediated involves the prefrontal cortex (PFC), comprising dorsolateral, orbitofrontal, medial, and frontal/anterior cingulate areas. These PFC regions connect and process cognitive and emotional information by coordinating sensory and motor responding from multiple brain areas; sharing smaller neurocognitive networks, or “mini-networks” of memoranda [12 , 27–29]. The role of these “mini-networks” is to recruit and retrieve long-term memory for specific behaviors. As such, ‘executive attention’ is dependent on previously established associations and the temporary activation of long-term memoranda [12].

The model of ‘executive attention’ [12] posits the existence of three separate, but highly integrated mechanisms: working memory, preparatory set, and inhibitory control [12 , 30–32]. Working memory is the ability to temporally and retrospectively reclaim and retain items from recent and past experiences. In this context, working memory is memory for the short term, rather than short-term memory and is best understood as attention focused on the internal representation of the task at hand [31]. It is here that mental set is initially established and preexisting networks of long-term memoranda are activated [12]. Preparatory set sustains the necessary mental set for expected actions contingent on previous events and information from working memory [12 , 32]. Working memory can be seen as attention directed toward the past, while preparatory set is attention directed toward the future. Simply put, preparatory set is the prospective intention(s), and associated behaviors, to act according to the task parameters. Both constructs require the integration of higher-order neurocognitive schemas, gestalts, and rules of actions that cascade to subordinate non-prefrontal cortical areas that ultimately execute partial goals and concrete actions [12]. Embedded or nested within the constructs of working memory and preparatory set are ‘partial’ sets, comprised of components of complex behavior, that monitor and correct goal-directed activity at every stage to insure correct responding [33 –35].

Finally, inhibitory control is a construct that is able to discriminate and/or suppress extraneous internal and external inputs that can derail or interfere with the goal-directed behavior that bring tasks to successfully completion [12 , 32]. Inhibitory control is, therefore, an exclusionary mechanism that functions to maintain and protect selective focus from interfering stimuli not germane to the present task [32, 36]. Although these three constructs are associated with unique, separate underlying neurocognitive abilities, their operation can be understood as a unified gestalt. When, say, working memory ends and preparatory set begins; or, when inhibitory control is deployed, may vary as task demands change.

Temporal organization and executive attention

The model of executive attention suggests that pre-frontal cortical regions are recruited to the extent information is novel, complex, and non-routinized. However, a factor that is highly determinant regarding the recruitment of pre-frontal neurocognitive resources is the time necessary to bring tasks to fruition [12]. Indeed, the construct of temporal organization denotes the capacity of pre-frontal regions to effectively coordinate the necessary neurocognitive operations for successful task completion as a function of time. The construct of temporal organization, therefore, assumes a superordinate role in coordinating the three subordinate constructs described above, working memory, preparatory set, and inhibitory control, for effective goal-directed behaviors [37, 38]. Data supporting the importance of temporal organization come from primate research showing that neural processing often begins in the PFC and ends in the motor cortex, i.e., narrowing from global to concrete actions [38]. Therefore, as one brings executive tasks to fruition, behavior becomes increasingly selective. With continuous performance, neurons in the PFC begin to associate novel sensory stimuli with an expected response [31]. It is hypothesized that the process by which brain networks involved in executive attention are activated, and eventually become a learned response, is through “top-down” cortico-thalamic-striatal loops [36 , 40]. This occurs downward, in a feed-forward fashion, through an executive hierarchy, and simultaneously monitors and receives feedback from each level to its precursor level.

Temporal organization: dementia and mild cognitive impairment

Evidence illustrating how the construct of temporal organization can explain dysexecutive behavior in patients with dementia has been provided by Lamar and colleagues [41]. In one analysis, patients clinically diagnosed with AD and vascular dementia (VaD) were assessed with the Boston Revision of the Wechsler Memory Scale Mental Control subtest. Performance was divided into three test epochs: the initial, middle, and latter responding. When performance was analyzed as a function of these three test epochs, patients with VaD produced a striking negative slope. That is, performance worsened from test epoch 1, to test epoch 2, to test epoch 3. By contrast, patients with AD declined from test epoch 1 test to epoch 2; however, there was no further decline.

In a second experiment, the percent of total output on the letter fluency test (letters ‘FAS’) was examined as a function of each of the four 15-s test epochs. In this analysis, patients with AD, VaD, and Parkinson’s disease (PD) were studied. Patients with VaD and PD tended to produce their maximum output during the first 15-s test epoch after which there was a steep decline in output throughout the remaining portion of the test, a profile consistent with the striking negative slope seen on the Mental Control subtest described above. Equally interesting was the observation showing little difference between AD and normal control participants where percent of total output was more evenly distributed across all test epochs. These data were interpreted to suggest that derailed temporal organization underlie the aggregate Mental Control and letter fluency total scores that differentiated between dementia groups. Eppig and colleague [11] documented similar behavior in patients with MCI. The current research is an extension of these prior studies and sought to provide operational definitions of the three subordinate constructs associated with temporal organization: working memory, preparatory set, and inhibitory control.

Backward Digit Span Test

The Backward Digit Span Test (BDT) described by Lamar and colleagues [41, 42] has been used to operationally define executive dysfunction in dementia and MCI through the analysis of serial order recall. Lamar and colleagues [41, 42] found that performance on the BDT was able to differentiate patients with VaD from those with AD based on less accurate serial order recall. Moreover, reduced serial order recall was associated with greater MRI-defined white matter disease in parietal and prefrontal regions. Emrani et al. [43] found that BDT serial order recall was able to dissociate mixed/dysexecutive MCI patients from other MCI groups by an absence of a recency effect. Bezdicek et al. [44] administered the BDT to patients with PD. Using functional MRI technology, these researchers found that better serial order recall performance was associated with increased functional connectivity between the bilateral dorsolateral PFC and left insula, inferior frontal gyrus, and putamen in patients with PD-MCI and controls. Collectively, these studies can infer derailed working memory, preparatory set, and inhibitory control as suggested by the model of executive attention [12]. However, prior research lacked the granularity to characterize how each of these three constructs contributes to derailed performance on executively-mediated tests.

The current research was designed to test two hypotheses: 1) that the analysis of 5-span Backward Digit Span intra-component latencies would reveal the presence of distinct time epochs; and 2) these time epochs could be used to operationally define the subordinate constructs of working memory, preparatory set, and inhibitory control. The model of executive attention [12] suggests that successful task performance relies on the capacity for correct temporal ordering of test stimuli. This is to say, in order to correctly complete an executively-mediated test, all three constructs of working memory, preparatory set, and inhibitory control must work appropriately and under the direction of temporal organization. For this reason, correct test trials were analyzed separately from incorrect test trials. Latency data, i.e., individual reaction times for each backward digit span response was obtained for correct trials. Both intra-component latencies from correct tests, and errors made on incorrect test trials were employed to provide operational definitions of the constructs of working memory, preparatory set, and inhibitory control; and to assess how patients with MCI versus patients not meeting criteria for MCI differ.

METHODS

Participants

Patients in this current research study (n = 58) were recruited from the New Jersey Institute for Successful Aging Memory Assessment Program (MAP). All MAP patients underwent a comprehensive neuropsychological evaluation. Patients were also examined by a social worker and a board-certified geriatric psychiatrist. An MRI study of the brain and appropriate blood serum tests were obtained to evaluate reversible causes of dementia. A clinical diagnosis was determined for each patient at an interdisciplinary team conference. Patients diagnosed with MCI presented with evidence of cognitive impairment relative to age and education, preservation of general functional abilities, and the absence of dementia. Exclusion criteria included: history of head injury, substance abuse, and major psychiatric disorders including major depression, epilepsy, B12, folate, or thyroid deficiency. For all patients, a knowledgeable family member was available to provide information regarding functional status. This study has been approved by the Rowan University institutional review board with consent obtained consistent with the Declaration of Helsinki.

Neuropsychological assessment

The neuropsychological protocol used to classify MCI subtypes is the same as described by Emrani et al. [43]. Nine parameters, from three domains of cognition (executive control, naming/lexical access, and declarative memory), were expressed as z-scores derived from normative data (see Table 1). While we acknowledge that other neuropsychological tests/domains of cognitive functioning could have been used, our rationale for using the current protocol was based on extensive prior research showing that these tests are able to illustrate key neurocognitive constructs and differentiate between MCI subtypes [see 45–48; Table 1].

Table 1

Neuropsychological domains

Executive function domain	Language/lexical access domain	Episodic memory domain
WMS –Mental Control Subtest (Boston revision) [41]	60-item Boston Naming Test [63, 66]	CVLT (short form) - Immediate Free Recall [65]
Letter Fluency –‘FAS’ [60, 61]	‘Animal’ Fluency [60, 61]	CVLT (short form) Delayed Free Recall [65]
Trail Making Test –Part B [60, 62]	WAIS-III Similarities Subtest [64]	CVLT (short form) - Delayed Recognition [65]

CVLT, California Verbal Learning Test; WAIS-III, Wechsler Adult Intelligence Scale-3rd edition.

Patient classification

Mild cognitive impairment

Jak-Bondi et al. [48] criteria was used to classify patients as presenting with MCI. Accordingly, single domain MCI was diagnosed when participants scored > 1.0 standard deviation (SD) below normative expectations on two or more of three measures within any single cognitive domain. Mixed MCI was diagnosed when participants score > 1.0 SD below normative expectations on two of three measures within two or more cognitive domains. In the current research, all MCI patients were aggregated into a single MCI group.

Non-mild cognitive impairment

Some memory clinic patient did not meet Jak-Bondi [48] criteria for MCI. These patients scored above 1 SD above all nine neuropsychological parameters or scored 1 SD below the mean on no more than one neuropsychological parameter from each of the three domains that were assessed. These memory clinic patients were labeled as non-MCI.

iPad administration of the BDT

The current research collected data in real-time via iPad-administered BDT through voice recognition. Thus, the BDT, comprised of seven trials of 3-, 4-, and 5-digit span lengths for a total of 21 trials, was administered as described elsewhere [41, 42] with the exception that the iPad verbally speaks numbers and records the patient’s spoken response, i.e., they repeat the numbers backwards. The utility of recording responses on the iPad included the ability to measure total time to complete each trial, as well as time to complete each intra-component latency. Thus, iPad technology collected intra-component latency for each response, defined as the time to begin a response for each digit position (i.e., time zero to first response, time from the end of the first response to the beginning of the second response etc.), across each serial order position for each span. Average total time is the aggregated time for all correct trials of a specific span divided by the number of correct trials.

Outcome variables

Correct test trials

Outcome variables included the five intra-component latencies, total average time to completion, and number of correct trials.

Incorrect trials

Outcome variables also included the five intra-component latencies, total average time to completion, and the number of correct trials. Consistent with Emrani et al. [43], total number of out-of-sequence or transposition errors were tallied.

Transposition errors

Transposition errors were expressed by measuring the degree of displacement as related to their correct serial position. Anticipation transposition errors were scored using a negative displacement value because they occurred in advance of its correct serial position. Postponement transposition errors were scored using a positive displacement value because they occurred after the correct serial position.

Dysexecutive errors

Four types of non-out-of-sequence errors were tallied. Non-out-of-sequence errors included between-trial perseverations, or a number from the preceding two trials that was pulled into the current response; and within-trial perseverations, or a number within a trial that is repeated. Between-trial capture errors were scored when a number from either of the preceding two trials was pulled into the current response creating a contiguous, automatized string of digits. Within-trial capture errors were scored when number(s) within the same trial were incorrectly repeated, also creating a contiguous string. Because of the low frequency of some of these errors, all perseveration and capture errors were summed and labeled dysexecutive errors.

Statistical analyses and outcome variables

All continuous variables were screened for outliers and evaluated for departures of normality through quantitative examination of skewness and kurtosis, as well as visual inspection of frequency distributions. To address non-normality issues, where indicated, we assigned outliers a lower weight, i.e., the data point was modified to the next closest value in the set [49].

Between- and within-group latency analyses:Using IBM SPSS, between-group t-tests were used to assess differences on number of correct trials, average total time for correct trials, and transposition/item errors. Separate 2 MCI/non-MCI, (between-group) x 5 intra-component latencies (within-group) ANOVAs were calculated for correct and incorrect trials. Follow-up analyses included either within- or between-group t-tests (as indicated) to compare significant results. We also transformed correct intra-component latencies to a fraction using the formula: correct intra-component latency/correct average total time and performed a second 2 x 5 mixed design ANOVA.

RESULTS

Demographic characteristics

Table 2 lists demographic and clinical information. No between-group differences were found on age, education, the Geriatric Depression Scale [50], projected premorbid general intellectual abilities assessed with the Wide Range Achievement Test Reading subtest-IV (WRAT-IV) [51], gender, or Instrumental Activities of Daily Living [52]. There was statistical significance (t (56) = 2.18, p < 0.035) on the Mini-Mental State Examination (MMSE) [53] (Table 2).

Table 2

Demographic and clinical information

	Non-MCI (n = 36)	MCI (n = 22)	Significance
Age	73.19 (7.15)	72.45 (5.62)	ns
Education	15.81 (2.45)	15.23 (2.60)	ns
MMSE	28.61 (1.48)	27.68 (1.73)	MCI < non-MCI; p < 0.035
WRAT-IV Reading subtest	115.61 (14.71)	109.50 (17.06)	ns
IADL abilities	15.83 (2.09)	14.86 (2.61)	ns
GDS	2.78 (2.72)	2.64 (2.17)	ns
Gender	22 Female; 14 males	17 Females; 5 males	ns

MCI, mild cognitive impairment; IADL, instrumental activities of daily living; WRAT-IV, Wide Range Achievement Test-IV; ns, not significant.

5-Span Backward Digit Span latency

Correct trials and average total time to complete trials

Between-group differences for total correct trials was statistically significant such that the non-MCI group produced more correct trials than the MCI group (non-MCI; mean = 4.20, SD = 1.54; MCI; mean = 2.38, SD = 1.92, t (60) = 4.05, p < 0.001, Cohen’s d = 1.05). By contrast, independent sample t-test assessing between-group differences for the average total time of correct and incorrect responses was not statistically significant (correct trials: non-MCI; mean = 7.34, SD = 3.94; MCI; mean = 6.72, SD = 2.46; incorrect trials: non-MCI; mean = 8.81, SD = 3.06; MCI; mean = 7.91, SD = 2.40).

Group by serial order latency: correct trials

Group by serial order intra-component latencies was analyzed using a mixed-design ANOVA with a within-subjects factor (latency for correct positions 1–5) and a between-subject factor (non-MCI = 36, MCI = 22; Fig. 1). Mauchly’s test indicated that the assumption of sphericity was violated (χ² (9) = 176.14, p < 0.001); therefore, the degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity (ɛ= 0.374). A main effect of group on latency for each serial position was significant (F[1.50, 83.71] = 33.77, p < 0.001, η_p² = 0.376; see Fig. 1). There was no significant interaction between serial order position latency and group.

Fig. 1

BDT latency correct positions 1 through 5.

Follow-up independent sample t-tests were used to measure differences between-group on correct intra-component latency positions one through five. Correct latency positions two (t (53.52) = 2.66, p < 0.011, Cohen’s d = 0.66), three (t (56) = –2.63, p < 0.012, Cohen’s d = 0.71), and four (t (56) = 2.10, p < 0.012, Cohen’s d = 0.59) were statistically significant. Non-MCI patients required more time, i.e., displayed slower latencies on positions two and four. By contrast, MCI patients required more time, i.e., produced a slower latency for position three (Table 3; Fig. 1).

Table 3

Serial order position latency: Means and standard deviations (s)

Serial order position	Non-MCI	MCI
Position 1	1.96 (1.78)	1.93 (1.91)
Position 2	0.47 (0.50)	0.21 (0.24)
Position 3	0.71 (0.69)	1.19 (0.67)
Position 4	0.76 (0.51)	0.48 (0.43)
Position 5	0.26 (0.30)	0.22 (0.27)

MCI, mild cognitive impairment.

Serial order latency as a fraction of average total time: Correct trials

Each correct intra-component latency was transformed to a fraction by dividing each correct intra-component latency by the correct average total time and examined using a mixed-design ANOVA. Mauchly’s test indicated that the assumption of sphericity was, again, violated (χ² (9) = 101.84, p < 0.001); therefore, the degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity (ɛ= 0.501). A main effect of group on latency for each serial position was significant (F[2.00, 112.16] = 47.63, p < 0.001, η_p² = 0.460; Fig. 2). Moreover, there was a significant serial order position latency by group interaction (F[2.00, 112.16] = 3.88, p < 0.024, η_p² = 0.07). Follow-up independent sample t-tests revealed that groups continued to statistically differ on positions two (t (55.08) = 2.83, p < 0.008, Cohen’s D = 0.70), three (t (56) = –3.75, p < 0.002, Cohen’s D = 1.07), and four (t (56) = 2.30, p < 0.026, Cohen’s D = 0.63; see Table 5). Again, non-MCI patients continued to display slower latencies on positions two and four with MCI patients displaying a slower latency on position three (Fig. 2). Similar analyses examining incorrect trials revealed neither a significant main effect for group nor a group by serial order interaction.

Fig. 2

5-Span BDT (fraction of average total time).

Table 5

Transposition and items errors (means and standard deviations)

Serial order position	Non-MCI	MCI	Significance
Anticipation:
4 position before correct response	0.00 (0.00)	0.00 (0.00)	n/a
Anticipation:
3 position before correct response	0.02 (0.16)	0.09 (0.29)	n/a
Anticipation:
2 position before correct response	0.36 (0.59)	0.81 (1.00)	n/a
Anticipation:
1 position before correct response	2.25 (1.71)	3.36 (2.61)	n/a
Postponement:
1 position before correct response	1.47 (1.08)	2.40 (1.79)	n/a
Postponement:
2 position before correct response	0.30 (0.52)	0.63 (0.84)	n/a
Postponement:
3 position before correct response	0.05 (0.23)	0.22 (0.42)	n/a
Postponement:
4 position before correct response	0.00 (0.00)	0.00 (0.00)	n/a
Average anticipation displacement	0.65 (0.44)	1.06 (0.76)	F[1, 56] = 6.61, p < 0.013
Average postponement displacement	0.47 (0.28)	0.81 (0.53)	F[1, 56] = 10.31, p < 0.002
Item errors	2.63 (1.85)	4.00 (2.44)	F[1, 56] = 5.75, p < 0.020

n/a, not analyzed; MCI, mild cognitive impairment.

Table 4

Transformed serial order position latency: Means and standard deviations

Serial order position	Non-MCI	MCI
Position 1	0.25 (0.14)	0.25 (0.17)
Position 2	0.07 (0.07)	0.03 (0.03)
Position 3	0.09 (0.08)	0.18 (0.09)
Position 4	0.11 (0.07)	0.07 (0.05)
Position 5	0.04 (0.04)	0.03 (0.04)

MCI, mild cognitive impairment.

Transposition and dysexecutive errors

Means and standard deviation for all anticipation or postponement transposition errors found very few 2, 3, or 4 position errors. For this reason, between-group analyses were obtained by averaging anticipation and postponement transposition displacement errors. As displayed in Table 5, both analyses were significant with patients with MCI generating greater transposition displacement (anticipants; F[1, 56] =6.61, p < 0.013; postponements; F[1, 56] = 10.31, p < 0.002). Patients with MCI also produced more dysexecutive errors (F[1, 56] = 5.75, p < 0.020; Table 5).

DISCUSSION

Intra-component latency and the analysis of errors

The two goals of the current research were to test the hypotheses that intra-component latencies on the 5-span backward digit task would yield distinct time epochs between MCI and non-MCI groups; and, that these time epochs could be used to operationally define the subordinate constructs of working memory, preparatory set, and inhibitory control. As described above, patients with non-MCI produced more correct trials than patients with MCI. However, no between-group difference was found for average total time to completion for correct 5-span test trials. Nonetheless, consistent with our hypothesis, the analysis of intra-component latencies suggests the presence of distinct time epochs. Specifically, between-group analyses showed that the non-MCI group required more time or produced slower intra-component latencies than the MCI group on positions two and four. By contrast, the MCI group required more time to respond to position three (see Figs. 1 2). These test results suggest differences in how time was allocated to produce a correct response. These patterns of responding were found for both raw and fractionated intra-component latencies for correct trials. Similar analyses for incorrect trials were largely not significant. Not surprisingly, patients with MCI produced more dysexecutive errors. Also, greater anticipation and postponement transposition displacement was noted for the MCI group compared to the non-MCI group.

Temporal organization and intra-component latency

The second goal of the current research was to test the hypothesis that 5-span time intra-component time epochs may be a means by which to operationally-define the constructs of working memory, preparatory set, and inhibitory control [12]. As described above, working memory is related to the cognitive resources to establishing the necessary mental set for correct response. Working memory is attention focused retrospectively on the internal representation of the task at hand, in this case, the instructions to repeat numbers backwards. In order to repeat correctly digits backward, one must mentally re-order all test stimuli before responses are forthcoming. The first and second intra-component latencies are likely measuring the cognitive operations that underlie working memory/preparatory set. Each group required almost two seconds before initiating their response. However, differences emerged on intra-component 2 where the non-MCI groups produced a slow latency or required more time to generate a correct response. These data suggest that a characteristic that distinguished between non-MCI versus MCI groups is a greater capacity to deploy the necessary neurocognitive resources to for effective mental re-ordering early in the test trial, i.e., a cognitive operation consistent with effective working memory operations.

Preparatory set denotes cognitive operations that start and sustain the necessary mental set initially established by working memory to bring the task to successful completion. It has been suggested that to facilitate backward digits recall, a visual mental image, or template of the numbers to be repeated is formed and numbers are subsequently “read” backwards [54, 55]. Thus, success requires both the coordination of considerable temporal and visuospatial information [54, 55], and the ability to synchronize this information with recent and long-term memory to establish prospectively an effective response set [12]. Thus, a second characteristic that differentiated groups was the slower latency produced by MCI groups during the middle responding or position 3. Slower intra-component response latency for position 3 produced by MCI groups suggests that these patients needed more time to deploy the necessary neurocognitive resources to sustain the necessary mental set.

Inhibitory control is an exclusionary mechanism that works to maintain and protect selective focus from interfering stimuli not germane to the present task and prevent the production of errors that can derail test performance [32, 36]. The intra-component latencies during the latter responding, i.e., positions 4 and 5, and the production of errors could provide an operation definition for inhibitory control. Thus, a third characteristic that distinguished between groups was that slower response latency produced by non-MCI groups for position 4. Slower latency during the latter, final time epoch could represent a greater capacity for an iterative ‘check in,’ whereby patients revisit working memory and preparatory set to ensure correct implementation of instructions, intentions, and behaviors. Patients with MCI also produced a greater number of dysexecutive errors and greater transposition displacement than patients without MCI. The emergence of these errors very likely suggests defective operations related to inhibitory control [12 , 32]. Previous studies have shown derailed recency effects in patients with a dysexecutive/mixed MCI [11, 43]. This is likely why the MCI group generated fewer correct trials. As such, it can be extrapolated that derailed serial order recall may reflect dysfunctional inhibitory control, where internal or external stimuli interfere with the behavior to produce a correct response. Therefore, slower latency on position 4 produced by non-MCI patients may reflect better or more effective inhibitory control. Taken as a whole, the deployment of digital technology to measure individual intra-component latencies clearly shows the presence of discrete time epochs likely related to separate, but integrated underlying neurocognitive constructs.

As stated above, working memory, preparatory set, and response inhibition represent separate neurocognitive operations. Nonetheless, all three of these constructs are highly integrated and necessary for successful goal-directed behavior; therefore, these constructs are best understood as a unified response set. For example, having deploy maximum resources during the initial time epoch (position 2), the non-MCI group may not have needed to call upon additional neurocognitive resource during the middle time epoch (position 3), and had sufficient neurocognitive resources to monitor their response during the latter time epoch (position 4). By contrast, the MCI group may not have been able to deploy maximum neurocognitive resources during the initial time epoch (position 2) but were able to re-coup and deploy the necessary neurocognitive resources during the middle time epoch (position 3). However, faster latencies produced by the MCI group during the latter time epoch (position 4) could suggest less capacity to monitor their behavior. The incapacity to monitor behavior could underlie the greater number of errors generated by the MCI group. In sum, it is important to understand that the current research does not rigidly assign each of the five intra-component latencies to a single construct. Rather, these individual latencies measures are best understood as heuristics by which to understand how separate, but integrated neurocognitive mechanisms operated to produce goal directed behavior.

Alternative explanations

In prior research, Emrani and colleagues [43] found that derailed serial order recall in patients with MCI could be explained using constructs drawn from Competitive Queuing (CQ) models of executive control [56 –59]. CQ models suggest the existence of two integrated constructs: 1) an excitatory parallel planning mechanism and 2) an inhibitory competitive choice or response suppression mechanism. Parallel planning is the construct associated with the initial excitatory activation of all elements in the sequence to be recalled. The construct of parallel planning is thought to consist of neural nodes corresponding to all memoranda to be recalled. The strength of activation for each node might vary depending upon task parameters. After neural nodes are activated, competitive choice/response suppression governs the actual output and order of recall. The item with the greatest activation is selected for recall. After competitive choice, an inhibitory feedback system, or response suppression, removes items from the planning layer so the next strongest activated item can be recalled. This process continues iteratively until all items are recalled [23 , 44].

The model of executive attention [12] is not inconsistent with CQ theory. The initial activation of memoranda to be recalled is consistent with the necessary neurocognitive operations to establish an initial mental set or working memory. Inhibitory feedback and response suppression permitting subsequent correct responding is consistent with neurocognitive operations necessary to sustain mental set or preparatory set, and the ability to inhibit errors or inhibitory control.

Limitations, conclusions, and future work

The current study has several strengths including a priori predictions regarding the relation between time-based behavior and underlying neurocognitive constructs; novel technology to measure latency, or time to generate a response; and the use of objective criteria to classify MCI and non-MCI groups. Nonetheless, several limitations must be acknowledged. First, our overall sample and effects size was modest. Clearly, additional research is necessary. Second, our definition of MCI was limited to three neurocognitive domains. Despite these limitations, our findings provide preliminary evidence that assessing latency for serial order recall in executively-demanding tasks follow a behavioral pattern consistent with model of executive attention [12]. Moreover, the BDT was able to dissociate MCI from non-MCI group by assessing the proportion of each response time as a function of total time to complete correct trials.

To expand upon the current findings, future work should investigate whether and how the analysis of intra-component latencies can differentiate MCI subtypes, and between other neurologic groups such as non-dementia patients with PD. Moreover, replicating and extending these findings in larger patient samples may yield methods to calculate discrete neurocognitive biomarkers to identify and track neurodegenerative illness earlier on in the disease process. Specifically, time-based data obtained from both correct and incorrect test trials, along with tallying correct and incorrect test trials, may be combined to calculate a neurocognitive biomarker(s) to predict cognitive decline in those with MCI. Finally, studying both time-based and non-time-based parameters using machine learning may provide additional information regarding variables that are most likely to predict cognitive decline that can ultimately be applied in primary care settings. In sum, the current research demonstrates that coupling traditional paper and pencil neuropsychological tests with digital technology yields meaningful empirical and theoretical information.

Footnotes

ACKNOWLEDGMENTS

This research was undertaken to fulfill, in part, requirements for a Doctor of Philosophy degree in Psychology, Rowan University, for Sheina Emrani. Funding for this research was obtained from the Osteopathic Heritage Foundation and the New Jersey Health Foundation (ISFP 23-19).

Authors’ disclosures available online ().

References

Bateman

, Xiong

, Benzinger

, Fagan

, Goate

, Fox

, Marcus

, Cairns

, Xie

, Blazey

, Holtzman

(2012) Clinical and biomarker changes in dominantly inherited Alzheimer’s disease. N Engl J Med 367, 795–804.

Braak

, Thal

, Ghebremedhin

, Del Tredici

(2011) Stages of the pathologic process in Alzheimer disease: Age categories from 1 to 100 years. J Neuropathol Exp Neurol 70, 960–969.

Villemagne

, Burnham

, Bourgeat

, Brown

, Ellis

, Salvado

, Szoeke

, Macaulay

, Martins

, Maruff

, Ames

(2013) Amyloid β deposition, neurodegeneration, and cognitive decline in sporadic Alzheimer’s disease: A prospective cohort study. Lancet Neurol 12, 357–367.

Hebert

, Weuve

, Scherr

, Evans

(2013) Alzheimer disease in the United States (2010–2050) estimated using the 2010 census. Neurology 80, 1778–1783.

Gaugler

, James

, Johnson

, Marin

, Weuve

(2019) Alzheimer’s disease facts and figures. Alzheimers Dementia 15, 321–387.

Petersen

, Doody

, Kurz

, Mohs

, Morris

, Rabins

, Ritchie

, Rossor

, Thal

, Windblad

(2001) Current concepts in mild cognitive impairment. Arch Neurol 58, 1985–1992.

Wilson

, Leurgans

, Boyle

, Bennett

(2011) Cognitive decline in prodromal Alzheimer disease and mild cognitive impairment. Arch Neurol 68, 351–356.

Chao

, Pa

, Duarte

, Schuff

, Weiner

, Kramer

, Miller

, Freeman

, Johnson

(2009) Patterns of cerebral hypoperfusion in amnestic and dysexecutive MCI. Alzheimer Dis Assoc Disord 23, 245–252.

Delano-Wood

, Abeles

, Sacco

, Wierenga

, Horne

, Bozoki

(2008) Regional white matter pathology in mild cognitive impairment: Differential influence of lesion type on neuropsychological functioning. Stroke 39, 794–799.

10.

Delano-Wood

, Bondi

, Sacco

, Abeles

, Jak

, Libon

, Bozoki

(2009) Heterogeneity in mild cognitive impairment: Differences in neuropsychological profile and associated white matter lesion pathology. J Int Neuropsychol Soc 15, 906–914.

11.

Eppig

, Wambach

, Nieves

, Price

, Lamar

, Delano-Wood

, Giovannetti

, Bettcher

, Penney

, Swenson

, Lippa

(2012) Dysexecutive functioning in mild cognitive impairment: Derailment in temporal gradients. J Int Neuropsychol Soc 18, 20–28.

12.

Fuster

(2009) Cortex and memory: Emergence of a new paradigm. J Cogn Neurosci 21, 2047–2072.

13.

, Piers

, Devine

(2017) How technology is reshaping cognitive assessment: Lessons from the Framingham Heart Study. Neuropsychology 31, 846–861.

14.

Binaco

, Calzaretto

, Epifano

, McGuire

, Umer

, Emrani

, Wasserman

, Libon

, Polikar

(2020) Machine learning analysis of digital clock drawing test performance for differential classification of mild cognitive impairment subtypes versus Alzheimer’s disease. J Int Neuropsychol Soc 26, 690–700.

15.

Dagum

(2018) Digital biomarkers of cognitive function. NPJ Digit Med 1, 10.

16.

Dion

, Arias

, Amini

, Davis

, Penney

, Libon

, Price

(2020) Cognitive correlates of digital clock drawing metrics in older adults with and without mild cognitive impairment. J Alzheimers Dis 75, 73–83.

17.

Kourtis

, Regele

, Wright

, Jones

(2019) Digital biomarkers for Alzheimer’s disease: The mobile/wearable devices opportunity. NPJ Digit Med 2, 9.

18.

Piers

, Devlin

, Ning

, Liu

, Wasserman

, Massaro

, Lamar

, Price

, Swenson

, Davis

, Penney

(2017) Age and graphomotor decision making assessed with the digital clock drawing test: The Framingham Heart Study. J Alzheimers Dis 60, 1611–1620.

19.

Rentz

, Dekhtyar

, Sherman

, Burnham

, Blacker

, Aghjayan

, Papp

, Amariglio

, Schembri

, Chenhall

, Maruff

(2016) The feasibility of at-home iPad cognitive testing for use in clinical trials. J Prev Alzheimers Dis 3, 8–12.

20.

Clark

, Schiehser

, Weissberger

, Salmon

, Delis

, Bondi

(2012) Specific measures of executive function predict cognitive decline in older adults. J Int Neuropsychol Soc 18, 118–127.

21.

Emrani

, Lamar

, Price

, Wasserman

, Matusz

, Swenson

, Heilman

, Libon

(2020) Neuropsychological profiles in Alzheimer’s/vascular spectrum dementia. In Vascular Disease, Alzheimer’s, Disease and Mild Cognitive Impairment Advancing an Integrated Approach, Libon DJ, Lamar M, Swenson RA, Heilman KM, eds. Oxford University Press, New York, NY.

22.

Harrington

, Chiang

, Pogoda

, Gomez

, Thomas

, Marion

, Miller

, Siddarth

, Yi

, Zhou

, Lee

(2013) Executive function changes before memory in preclinical Alzheimer’s pathology: A prospective, cross-sectional, case control study. PLoS One 8, e79378.

23.

Kirova

, Bays

, Lagalwar

(2015) Working memory and executive function decline across normal aging, mild cognitive impairment, and Alzheimer’s disease. Biomed Res Int 2015, 748212.

24.

Collins

, Koechlin

(2012) Reasoning, learning, and creativity: Frontal lobe function and human decision-making. PLoS Biol 10, e1001293.

25.

Lamar

, Zonderman

, Resnick

(2002) Contributions of specific cognitive processes to executive functioning in an aging population. Neuropsychology 16, 156–162.

26.

Schoenberg

, Scott

(2011) The little black book of neuropsychology: A syndrome based approach. Springer, New York.

27.

Hebb

(1947) Spontaneous neurosis in chimpanzees: Theoretical relations with clinical and experimental phenomena. Psychosom Med 9, 3–10.

28.

Miller

, Cohen

(2001) An integrative theory of prefrontal cortex function. Ann Rev Neurosci 24, 167–202.

29.

Sakai

, Passingham

(2003) Prefrontal interactions reflect future task operations. Nat Neurosci 6, 75–81.

30.

Fuster

, Alexander

(1973) Firing changes in cells of the nucleus medialis dorsalis associated with delayed response behavior. Brain Res 61, 79–91.

31.

Fuster

(2002) Frontal lobe and cognitive development. J Neurocytol 31, 373–385.

32.

Fuster

(2003) Cortex and mind: Unifying cognition. Oxford University Press.

33.

Badre

, D’Esposito

(2007) Functional magnetic resonance imaging evidence for a hierarchical organization of the prefrontal cortex. J Cogn Neurosci 19, 2082–2099.

34.

Koechlin

, Hyafil

(2007) Anterior prefrontal function and the limits of human decision making. Science 318, 594–598

35.

Koechlin

, Ody

, Kouneiher

(2003) The architecture of cognitive control in the human prefrontal cortex. Science 302, 1181–1185.

36.

Fuster

(2008) The Prefrontal Cortex (3^rd ed). Lippincott Raven, Philadelphia, PA.

37.

Luria

, Karpov

, Yarbuss

(1966) Disturbances of active visual perception with lesions of the frontal lobes. Cortex 2, 202–212.

38.

Fuster

(1997) Network memory. Trends Neurosci 20, 451–459.

39.

Kastner

, Ungerleider

(2000) Mechanisms of visual attention in the human cortex. Ann Rev Neurosci 23, 315–341.

40.

Alexander

, Fuster

(1973) Effects of cooling prefrontal cortex on cell firing in the nucleus medialis dorsalis. Brain Res 61, 93–105.

41.

Lamar

, Price

, Davis

, Kaplan

, Libon

(2002) Capacity to maintain mental set in dementia. Neuropsychologia 40, 435–445.

42.

Lamar

, Price

, Libon

, Penney

, Kaplan

, Grossman

, Heilman

(2007) Alterations in working memory as a function of leukoaraiosis in dementia. Neuropsychologia 45, 245–254.

43.

Emrani

, Libon

, Lamar

, Price

, Jefferson

, Gifford

, Hohman

, Nation

, Delano-Wood

, Jak

, Bangen

(2018) Assessing working memory in mild cognitive impairment with serial order recall. J Alzheimers Dis 61, 917–928.

44.

Bezdicek

, Ballarini

, Albrecht

, Libon

, Lamar

, Ružička

, Roth

, Hurlstone

, Mueller

, Schroeter

, Jech

(2021) SERIAL-ORDER recall in working memory across the cognitive spectrum of Parkinson’s disease and neuroimaging correlates. J Neuropsychol 15, 88–111.

45.

Bondi

, Edmonds

, Jak

, Clark

, Delano-Wood

, McDonald

, Nation

, Libon

, Au

, Galasko

, Salmon

(2014) Neuropsychological criteria for mild cognitive impairment improves diagnostic precision, biomarker associations, and progression rates. J Alzheimers Dis 42, 275–289.

46.

Thomas

, Edmonds

, Delano-Wood

, Bondi

, for the Alzheimer’s Disease Neuroimaging Initiative (2017) Longitudinal trajectories of informant reported daily functioning in empirically-defined subtypes of mild cognitive impairment. J Int Neuropsychol Soc 23, 521–527.

47.

Libon

, Bondi

, Price

, Lamar

, Eppig

, Wambach

, Nieves

, Delano-Wood

, Giovannetti

, Lippa

, Kabasakalian

(2011) Verbal serial list learning in mild cognitive impairment: A profile analysis of interference, forgetting, and errors. J Int Neuropsychol Soc 17, 905–914.

48.

Jak

, Bondi

, Delano-Wood

, Wierenga

, Corey-Bloom

, Salmon

, Delis

(2009) Quantification of five neuropsychological approaches to defining mild cognitive impairment. Am J Geriatr Psychiatry 17, 368–375.

49.

Dixon

(1960) Simplified estimation from censored normal samples. Ann Math Stat 31, 385–391.

50.

Yesavage

, Brink

, Rose

, Lum

, Huang

, Adey

, Leirer

(1982) Development and validation of a geriatric depression screening scale: A preliminary report. J Psychiatr Res 17, 37–49.

51.

Wilkinson

, Robertson

(2006) Psychological Assessment Resources I. WRAT 4: Wide Range Achievement Test. Psychological Assessment Resources, Lutz, FL, USA.

52.

Lawton

, Brody

(1969) Assessment of older people: Self-maintaining and instrumental activities of daily living. Gerontologist 9, 179–186.

53.

Folstein

, Folstein

, McHugh

(1975) Mini-Mental State: A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–198.

54.

Hoshi

, Oda

, Wada

, Ito

, Yamashita

, Oda

, Ohta

, Yamada

, Tamura

(2000) Visuospatial imagery is a fruitful strategy for the digit span backward task: A study with near-infrared optical tomography. Cogn Brain Res 9, 339–342.

55.

Larrabee

, Kane

(1986) Reversed digit repetition involves visual and verbal processes. Int J Neurosci 30, 11–15.

56.

Hurlstone

, Hitch

, Baddeley

(2014) Memory for serial order across domains: An overview of the literature and directions for future research. Psychol Bull 140, 339–373.

57.

Grossberg

(1978) Behavioral contrast in short-term memory: Serial binary memory models or parallel continuous memory models. J Math Psychol 17, 199–219.

58.

Grossberg

(1978) A theory of human memory: Self-organization and performance of sensory motor codes, maps, and plans. In: Progress in Theoretical Biology, Vol. 5, Rosen R, Snell FM, eds. Academic Press, New York, pp. 233–374.

59.

Houghton

(1990) The problem of serial order: A neural network model of sequence learning and recall. In Current Research in Natural Language Generation, Dale R, Mellish C, Zock M, eds. Academic Press, London, pp. 287–319.

60.

Heaton

, Miller

, Taylor

, Grant

(2004) Revised comprehensive norms for an expanded Halstead-Reitan Battery: Demographically adjusted neuropsychological norms for African American and Caucasian adults scoring programs. Psychological Assessment Resources.

61.

Spreen

, Strauss

(1990) Compendium of Neuropsychological Tests. Oxford University Press, New York.

62.

Reitan

, Wolfson

(1993) The Halstead-Reitan Neuropsychological Test Battery: Theory and Clinical Interpretation. Neuropsychology Press.

63.

Goodglass

, Kaplan

, Weintraub

(1983) Boston Naming Test. Lea & Febiger.

64.

Wechsler

(1997) WAIS-III: Administration and scoring manual: Wechsler adult intelligence scale. Psychological Corporation.

65.

Delis

, Kramer

, Kaplan

, Ober

(2000) The California Verbal Learning Test-II. Psychology Corporation, New York.