Dysexecutive impairment is common in patients with dementia syndromes such as Alzheimer’s disease (AD) and vascular dementia (VaD), and in patients with mild cognitive impairment (MCI) [1–3]. Past research has shown that dysexecutive impairment in VaD associated with magnetic resonance imaging (MRI) white matter alterations and among some MCI patients is “context independent”. This means that elements of a dysexecutive syndrome can be found across virtually all neurocognitive domains that might be assessed. Thus, on tests examining the production of graphomotor perseveration, patients with VaD tend to produce errors suggesting an inability to desist from a motor act. On tests that assess verbal concept formation (e.g., the Similarities subtest) patients with VaD and some MCI patients demonstrate a gross inability to assume an abstract attitude and treat each element of the word pair as a separate test stimulus (e.g., dog–lion: “one is a pet and the other roars”). When memory is assessed with the serial list-learning test, patients with executive impairment often present with retrieval, rather than consolidation problems.

By contrast, the dysexecutive syndrome associated with AD and other MCI patients is “context dependent”, i.e. subordinate to other more pervasive neuropsychological problems. Among these patients, graphomotor perseverations have been shown to be highly associated with naming and lexical access problems. The errors made on tests assessing verbal concept formation tend to be vague, but retain some context to provide a supraordinate concept that binds the word pair (dog–lion: “I like them”). Serial-list learning test performance is often replete with elements consistent with amnesia (see ref 1–3 for a complete review). Focal deficits involving language and visuospatial operations can occur; however, these syndromes are comparatively rare [4, 5].

The research reviewed above provides useful heuristics to understand the constructs that underlie executive difficulty in dementia and MCI. Another heuristic mechanism that may be used to evaluate constructs that underlie executive difficulty in dementia and MCI is to assess performance as a function test epoch [6]. Patients with VaD and some MCI patients tend to produce a negative slope on executive tests when outcome is measured as a function of time [3, 6]. Past research demonstrating this time-related, dysexecutive impairment has been interpreted within the context of Fuster’s model of executive attention [7, 8]. This model views the time necessary to bring a task to fruition as a temporal gestalt. Subordinate to this temporal gestalt are three separate, but highly integrated neurocognitive constructs that govern how time is allocated - working memory, preparatory set, and inhibitory control.

Working memory is memory for the short term and operates to marshal the necessary neurocognitive resources to initiate a response as requested. Preparatory set is involved in recruiting neurocognitive resources to sustain the necessary mental set to successfully complete the task at hand. Finally, inhibitory control denotes the capacity to filter extraneous stimuli such as perseverations and behavior resulting in the loss of mental set and to minimize the emergence of errors. Evidence supporting the existence of these three neurocognitive constructs has come from both imaging and primate research [7, 8].

Recently, Emrani and colleagues [9, 10] administered an iPad version of the Backward Digit Span Test, originally described by Lamar and colleagues [11, 12], to patients with MCI versus patients with some, but insufficient cognitive deficits to meet diagnostic criteria for MCI (i.e. non-MCI). In this research, correct 5-span test trials were analyzed. Latency, or reaction time, to repeat each individual digit backward was measured. The analysis of latency data found that when MCI and non-MCI patients scored 100 percent correct on these trials, there was no difference in the average total time to complete correct trials. However, there were distinct differences in the allocation of time [9, 10]. Specifically, non-MCI patients tended to require more time, i.e. produced a slower latency, on the second and fourth responses on digits backwards. Comparatively, the MCI patients displayed slower latencies on position three as compared to non-MCI patients. Slower initial latencies were interpreted as a means to define operationally the neurocognitive construct of working memory. Thus, the slower latency on the second response suggests patients not meeting criteria for MCI were, perhaps, allocating greater neurocognitive resources to initially re-order test stimuli as required. By contrast, the latency to respond the third number backward was slower for MCI as compared to non-MCI patients. Emrani and colleagues [9, 10] interpreted this behavior as means to define operationally the neurocognitive construct of preparatory set, or the ability to sustain the necessary mental set to generate a correct response. Slower third response latency generated by MCI patients suggest that these patients essentially needed to work harder, i.e. marshal greater resources than their non-MCI counterparts to sustain the necessary mental set for correct responding. Finally, non-MCI patients produced a slower latency for the fourth digit repeated backward. Emrani and colleagues [9] interpreted this behavior to define operationally the neurocognitive construct of inhibitory control. That is to say, slower latency on position four by the non-MCI group may reflect a greater capacity to monitor behavior to prevent the production of errors. Most importantly, these data were interpreted to support the notion that it is possible to extract specific time epochs when asked to repeat numbers backward that can operationally define separate, but highly integrated neurocognitive constructs that underlie executive abilities in MCI.

In the current research, memory clinic patients were studied and assigned to two groups—patients with MCI using Jak-Bondi criteria [13]; and patients who presented with self-reported cognitive complaints but did meet criteria for MCI. These patients were labeled non-MCI. All patients were assessed with an iPad version of Symbolic Working Memory subtest from the Wide Range Assessment of Memory and Learning 2nd Edition (WRAML-2) [14]. The overall goal of the current research was to extend the findings obtained with a standard backward digit span test, using a cross-modal test paradigm, i.e. a test that requires the management of a wider array of neurocognitive abilities. Thus, in the current research our intent was to assess how well a cross-modal test paradigm can operationally define the neurocognitive constructs that underlie the model of executive attention [7, 8] and differentiate patients with MCI from non-MCI when serial order responding is 100 percent correct.

In prior research, Emrani and colleagues [9, 10] were able to operationally define the neurocognitive constructs underlying the theory of executive attention [7, 8] with analyses of latency data obtained on a digit span backward test. In the current research, we extended these findings and demonstrated that similar patterns of behavior can be captured using the Symbolic Working Memory subtest, a multi-modal digit re-ordering test where stimuli are initially administered verbally, but responses are captured via touch.

As reviewed above, working memory is a neurocognitive construct focused on the intention to act. The cognitive operations underlying working memory include collating all past experiences necessary to complete the task at hand. The construct of working memory “looks backward,” recruiting long-term memoranda and prior experiences consistent with the task demands needed for goal directed activity. Fuster’s [7, 8] construct of working memory is consistent with constructs suggested by Atkinson and Shiffrin [25], Baddeley [26], Norman [27] and Norman and Shallice [28]. Prior functional imaging data obtained from adults during a delay visual, non-spatial, match-to-sample task [29] found considerable activation involving bilateral dorsolateral prefrontal cortical regions during the initial trial epochs. In a series of studies [30] using a variety of verbal, auditory, and visual stimuli, bilateral dorsolateral prefrontal cortical regions along with posterior cortical regions consistent with the sensory modalities of the test stimuli tended to be highly active during initial trial epochs. However, during mid- and late-delay time epochs; and, immediately preceding the requested response; activation shifts to prefrontal and premotor regions, respectively [31]; see ref 7, pp. 298–301, for a review]. These data have been interpreted to suggest that different, but highly integrated neurocognitive constructs predominant at different times during a participants response [7, 8]. In this context, the theory of executive attention [7, 8] characterizes the entire time necessary to complete the task as a superordinate temporal gestalt. Subordinate to this temporal gestalt are three neurocognitive constructs—working memory, preparatory set, and inhibitory control. However, exact operations regarding how these three constructs operate are likely variable and contingent on the task under consideration.

On the Symbolic Working Memory Test, one must leverage memoranda to call into mind attributes associated with digits as well as the cognitive resources for mentally re-order digits as requested. Recalling or repeating a circumscribed string of digits or bits of information has become a generic, everyday experience (e.g., phone numbers, passwords, etc.). Therefore, the cognitive operations associated with generic instances involving serial order recall are likely recruited. Reasonably so, on correct 4-span test trials both groups (non-MCI and MCI) took their longest time to respond to the first number to be recalled. However, MCI patients required more time than non-MCI patients for position 1 suggesting that MCI patients needed to work harder to recruit the necessary resources to initiate a response.

Similar to the neurocognitive construct of working memory, one function of preparatory set is to focus on the intention to act. In addition to this role, preparatory set sustains the ability to respond consistently with task demands. Preparatory set is, therefore, attention “towards the future”, i.e. attention focused on the prospective cognitive operations necessary to attain a goal. Evidence for this neurocognitive construct comes from electrophysiological research using delayed response tasks with primates. This research has documented the presence of slow, negative potentials (i.e. contingent negative variation; CNV) in anticipation to a motor act [32]. Further, electrophysiological activity has been shown to be proportional to the certainty that the respondent can predict a positive outcome [33]. In the current research, the latency in the middle of the test trial, i.e. position 3, was slower for MCI compared to the non-MCI patients.

Finally, inhibitory control is attention focused on excluding responses not germane to the task at hand. Any number of internal or external factors can derail correct serial order recall. Emrani and colleagues [9] noted that errors produced by MCI patients often occurred toward the latter portion of their response. Moreover, many of these errors, termed between-trial errors, were such that a digit from an immediately preceding test trial was incorrectly inserted into the current response. This type of error is consistent with recurrent perseverations as described by Sandson and Albert [34]. In addition to a slower latency in order to ‘check in’ with test demands, slower latency on position 3 may be suggestive of a concomitant, faulty inhibitory control processes, where MCI patients required more time to inhibit external information that would otherwise derail one from the task at hand. Past research suggests that the inhibition of information not germane to the task at hand is the province of the basal ganglia and orbital prefrontal cortical regions [33].

A parsimonious view of the latency data described by Emrani and colleagues [9, 10]; and the data described in the current research, is to view the entirety of patients’ response as an overall temporal gestalt whereby patients need to manage diverse test contingencies at different times during the generation of the response. This interpretation, while plausible, requires additional research. Data using imaging technology demonstrating the activation of different brain regions as a function of time-epoch in patients with suspected MCI would provide additional support for this model [7, 8]. Finally, performance on the 5-span trial was generally not statistically significant. The reason(s) for this are unclear but may revolve a restriction of range of data because the number of correct trials achieved.

Data supporting the notion that time-epochs when asked to mental order digits are related to integrated but differing underlying neurocognitive operations comes from the regression analyses described above. For example, faster position 1 latency was highly associated with better episodic memory and naming/lexical access test performance. An association between faster position 1 latency and better performance on episodic memory and naming/lexical access test comports with clinical observations. Indeed, a strategy used by many patients when asked to mentally re-order 4 or 5 bits of information is to subvocalize or actually rehearse the entire operation out loud, peeling numbers off numbers one at a time until the entire operation is completed. It is, therefore, reasonable to associate intact episodic memory and naming/lexical accessing skills with this portion of the test. On the other hand, faster position 4 latency, i.e. faster latency to bring the 4-span test trial to fruition, was associated with better executive test performance. This observation is consistent that the notion that latter response latencies when asked to mentally reorder information may be related to both preparatory set or the ability to marshal the cognitive resources to sustain mental set; and, inhibitory control or the ability to inhibit the production of errors.

Clinical drug trials designed to treat AD have not been successful [35]. Greater attention is now devoted to designing pharmacological intervention for patients with MCI rather than dementia. Therefore, identifying patients as early as possible in an emergent dementia-related disease process is highly desirable. Coupling traditional neuropsychological tests with digital technology could yield a panel of neurocognitive biomarkers that are highly sensitive to emergent illness. Among any sample of potential participants for the treatment of prodromal dementia, it is possible that analyses of the time-based data could signal the presence of emergent illness even if the actual response was correct. More research is required to test this prediction.

The current research is not without weaknesses. Our sample size and some of the effects that were obtained was modest. Also, the structure of the Symbolic Working Memory Test with trial blocks containing only three test trials could have restricted the time-based data that was obtained. Future research using cross-modal test paradigms would benefit from tasks containing more test trials. However, the current research has several strengths including a validated method by which MCI and non-MCI groups were identified, and the novel use of comparatively inexpensive, simple, but powerful technology to obtained time-based data. In sum, the current research suggests that slower response latencies, in conjunction with correct serial order recall, can operationally define neurocognitive constructs underlying the theory of executive attention [7, 8]. Moreover, time-based parameters as described by Emrani and colleagues [9, 10], and data from the current research, could eventually constitute a set of neurocognitive biomarkers that may be used to tract emergent dementia and MCI syndromes.