Overall, this project provides a novel investigation of Hayling Test performance and strategy use within the stroke population. Although the Stroop Test is commonly used to assess initiation and inhibition cognitive abilities, the results of this study strongly suggest that this task taps behaviourally and neuroanatomically distinct functions than those assessed in the Hayling Test. Behavioural metrics derived from the Hayling Test were found to provide detailed insight into abnormal patterns of both general performance impairment and strategy use in stroke survivors. These findings are important when considered in the context of clinical practice, as they highlight potential avenues for improving detection and diagnosis of common post-stroke cognitive impairments.
The findings of this study are in line with previous research suggesting that the Stroop and Hayling Tests likely tap dissociable components of executive functions (Cipolotti et al., 2016). This conclusion is supported by both the behavioural and anatomical findings of this study. First, stroke patients were found to exhibit qualitative performance differences across Stroop and Hayling Test metrics intended to tap similar functions. Only minimal overlap was present between the specific patients exhibiting abnormal performance on analogous Stroop and Hayling Test measures. While Stroop Test metrics were found to have relatively high specificity to abnormal performance (specificity range = 75.0-90.5), this test was found to have very low sensitivity compared to analogous Hayling measures (sensitivity range = 10.0-66.6). This difference cannot be explained as a function of task difficulty, as performance on analogous Stroop and Hayling interference and inhibition measures were not found to be significantly correlated. In terms of neural correlates, only minimal overlap was present between the network-level correlates of analogous Stroop and Hayling Test metrics. The highest degree of overlap was present between Stroop Dot Time and Hayling Initiation Time metrics, which tap initiation, with 6 (6.97%) of the 86 implicated network edges being common across both conditions. Overall, the findings of this investigation strongly suggest that the functions tapped by the Stroop and Hayling Test are both behaviourally and anatomically dissociable.
There are several potential explanations for this dissociation. First, past research has suggested that it is possible for patients to complete the Stroop Test without relying on the initiation/inhibition executive functions this test is designed to assess (Stuss & Alexander, 2007). This is because participants can reduce interference from incongruent colour words by adopting strategies in which they do not read the full word whilst naming ink colours. This strategy may be more commonly adopted in the stroke population, as comorbid visual, spatial attention, language, or acquired dyslexia deficits would be expected to reduce the degree to which word stimuli are “automatically” read (Coslett & Turkeltaub, 2016; Leff & Starrfelt, 2014; Vallar et al., 2010). This potential confound may help explain the documented disparity between performance on purportedly analogous Stroop and Hayling Test metrics. Additionally, Stroop Test performance was found to vary widely in this study’s sample of healthy ageing controls, with many controls committing more errors than what was average within the patient sample. This is in line with previous work documenting declines in initiation/inhibition abilities within healthy ageing populations (Cervera-Crespo & González-Alvarez, 2017; Gibson et al., 2018; West & Alain, 2000). However, this high control variability does suggest that the Stroop Test may not adequately be able to distinguish between age-related performance decrements and stroke-specific cognitive impairment. These implications are critically important when considered in the context of current clinical practice.
Variations of classical Stroop tasks are commonly used as screening tools for inhibition/initiation impairments within stroke survivors (Troyer et al., 2006). The findings of this study are critically important when considered in the context of this practice, as they suggest that abnormal performance on the Stroop Test may not serve as an effective method for detecting these impairments in the stroke population. First, healthy ageing controls and stroke patients were not found to perform significantly differently on the Stroop colour-word task accuracy and interference scores. These metrics are the key measures expected to differ between clinical and control populations (Troyer et al., 2006). This lack of significant difference is likely due to the high score variance present within the control population coupled with the fact that not all stroke patients would be expected to perform abnormally on this task. This lack of difference alone does not undermine the validity of the Stroop Test but does suggest that cut-offs derived from this variable control performance may lack sensitivity to detect abnormal performance relative to other measures. This implication is supported by this study as only 5 patients met impairment criteria on Stroop-Colour Word Interference Scores, versus 14 on the analogous Hayling Response Time difference measure. This finding implies that more sensitive and targeted assessments are needed to detect inhibition/initiation deficits with high sensitivity in the stroke population.
The findings of this project suggest that the Hayling Test may offer an effective alternative screening method for initiation and inhibition deficits in the stroke population. First, Hayling Test metrics were found to be significantly different across stroke patient and control populations. Specifically, stroke survivors exhibited worse performance on Hayling Overall Scaled Scores, Global Error Scores, and Inhibition Accuracy versus control subjects. This is important when compared to the Stroop Test in which the only metric which was significantly different between controls and patients was the baseline dot-colour naming response time measure. In line with this, more patients were categorised as demonstrating abnormal performance on Hayling Test measures relative to Stroop Test measures, suggesting a comparative increase in screening sensitivity. In addition to this improved general sensitivity to potential impairment, Hayling strategy-use variables highlight additional differences between patient and control populations.
As found in previous studies, stroke participants committed more Hayling Test suppression errors than controls (Laakso et al., 2019; Nijsse et al., 2019). However, this is the first investigation to explore strategy use in stroke survivors. This analysis provides novel insight into strategy differences underlying normal and abnormal Hayling Test performance. First, a significantly higher proportion of control participant responses involved using a strategy to facilitate a correct response. Specifically, controls were significantly more likely to use strategies involving reporting visual items (See Table 3). While no other statistically significant group-level strategy-use differences emerged between patients and controls, strategy-use analyses facilitated the identification of individual patients exhibiting abnormal response types across a range of detailed, response-type metrics.
Notably, different Hayling Test strategy-use measures were found to be associated with distinct network-level neural correlates. Specifically, correct responses involving reporting visual objects, visual objects related to previous responses, and other (undefined) strategy use were each significantly associated with non-overlapping patterns of network disconnection. Errors involving reporting responses semantically related to the sentence or bizarre errors were also found to be related to significant and distinct patterns of disconnection. These identified network-level correlates were found to be relatively restricted with comparatively few significantly involved nodes compared to general task metrics. However, these results did reach statistical significance following very strict corrections for lesion volume and multiple comparisons. Future, targeted studies are needed to fully quantify the patterns of disconnection associated with different Hayling Test strategies and to interpret the functional significance of the identified network correlates. However, the present investigation is important in that it provides preliminary evidence that these different behavioural patterns are linked to distinct patterns of dysconnectivity. This is an important finding as it suggests that differences in strategy use are not just a result of patients preferred approaches but are instead may be linked to distinct patterns of stroke-specific disconnection. This finding emphasises the added utility of employing the Hayling Test in stroke populations, as it provides insight into what strategies patients are employing as well as assessing general initiation and suppression abilities.
This information can be used to direct more in-depth neuropsychological assessments and to detect abnormal strategy use patterns even in the absence of overall Hayling Test impairment. The ability to generate and implement a strategy is crucial for rehabilitation to compensate for weaknesses and to facilitate the successful completion of everyday activities. Strategy training has been shown to be effective at overcoming suppression failures (Robinson et al., 2016), as well as multitasking (Rand et al., 2009) and cognitive flexibility and disability in stroke (Skidmore et al., 2015). We propose that analysis of strategy generation via the Hayling Test not only provides clinicians information regarding another aspect of executive functioning, but also provides information to support rehabilitation planning. That is, it can provide an indication of an individual’s capacity to problem solve and compensate for deficits. In this case, strategy generation and use appear to assist with overcoming difficulties with inhibitory control, which is important for goal-directed behaviour in daily life.
Notably, no significant network level correlates of Hayling Test Category B errors were identified in this study. This finding is surprising when considered in the context of previous functional imaging studies which have suggested that suppression abilities (as quantified by the Hayling Test) are related to activity within a network of left prefrontal areas (Collette et al., 2001). There are several reasons why these previous findings were not replicated in this study. First, functional imaging is able to identify correlates involved in cognitive processes but cannot clearly distinguish whether these correlates are merely involved with or are necessary for the cognitive function of interest. It is possible that documented activation of left fronto-temporal areas may be related to the language component of the Hayling Test (e.g., verbalising responses) rather than to cognitive inhibition functions. Qualitative analysis of Hayling Test and lesion data suggest that disconnection of frontal network edges is associated with a disproportionately high prevalence of Category B Hayling errors, but further data is needed before this implication can be either confidently supported or refuted. It is also possible that this negative finding is related to variability in network-level statistical power, as the probability of committing network-edge level false negative detection varies as a function of how many patients have damage at each specific edge (Griffis et al., 2021).
Finally, it is possible that the null result produced by this analysis indicate that a wide range of lesions (and potentially underlying mechanisms) may modulate the occurrence of Hayling Category B Errors. In terms of lesions, the diversity in lesion sites associated with the occurrence of Hayling Category B errors is evidenced by Fig. 7. Contrary to expectations, many patients with posterior lesions were found to commit the highest proportion of Category B errors on the Hayling. This may indicate that critical networks underlying suppression ability may be disrupted at many, spatially distinct locations or that lesion location is not the only factor which modulates the occurrence of Hayling Category B errors. Past work has found that general measures of cortical atrophy and white matter integrity act as more effective predictors of executive function impairment than lesion-location metrics (Hobden et al., 2022). It is possible that a similar relationship may be present between pre-morbid atrophy/white matter integrity and Hayling Test performance. Additional research is needed to explore each of these possibilities in detail and to further fundamental insight into the neural correlates underlying error commission on the Hayling Test.
This novel demonstration of the Hayling Test within the stroke population suggests that this task represents an effective alternative measure which can detect initiation and inhibition deficits. The findings of this study are in line with past research indicating that the Stroop and Hayling Tests assess behaviourally and anatomically dissociable components of executive function. Behavioural metrics derived from the Hayling Test were found provide detailed insight into abnormal patterns of both general performance impairment and strategy use in stroke survivors and were able to link these patterns to distinct neural correlates. These findings suggest that the Hayling Test can be employed in acute stroke settings to provide a detailed and practical screen of initiation, inhibition and strategy use abilities in stroke survivors.
Limitations
Executive dysfunction is a highly complex disorder and any one test in isolation is insufficient to fully characterise the pattern of impairment present in a patient. The results of this study suggest that the Hayling Test may serve as an effective first-line screen for initiation/inhibition impairment, but in-depth cognitive assessment is needed to fully characterise behaviour. First-line neuropsychological screens are important and practical clinical tools which are compatible with the time and resource limitations of real-world clinical environments (Moore et al., 2022). It therefore remains important to improve the efficiency of this first-line screening by identifying the tools which can provide the most detailed information within a short testing period.
Importantly, the Hayling Test may not be a suitable screen for all stroke survivors. A high percentage of the stroke population exhibit either language comprehension or production deficits which may preclude assessing initiation/inhibition performance on this language-dependent measure (Demeyere et al., 2015, 2020), although constraining an auditory context (i.e. sentence completion) has been found to facilitate production (Berndt et al., 2002; Robinson et al., 2005). Nevertheless, future studies can aim to develop and validate novel initiation/inhibition screens which are not dependent on language and can therefore be used in a greater portion of the stroke population.
The network-level analyses employed in this investigation use normative tractography atlases which may not exactly map onto the connectivity structure of the patients included in this analysis (Gleichgerrcht et al., 2017; Griffis et al., 2021). Similarly, number of disconnected streamlines can be unrelated to the strength of connectivity between regions (Fox, 2018; Griffis et al., 2021). Despite these potential limitations, past studies employing similar normative atlas-based tractography atlases have agreed well with studies employing in-vivo tractography approaches.
Finally, past work has suggested that the non-random spatial distributions of stroke lesions may yield results which are displaced relative to the true underlying neural correlates of behavioural deficits in voxel-level lesion mapping analyses (Mah et al., 2014). It is plausible that this effect is present in network-level analyses as well but is unlikely to have significantly impacted the main conclusions drawn in this study. Importantly, the aim of this study is to identify preliminary evidence of connectivity profile differences related to select tasks rather than to quantify the exact anatomy of each considered deficit. Future investigations can aim to expand on these findings by employing in-vivo tractography methods to improve fundamental understanding of the anatomy of suppression and initiation functions.