Cerebral hemovelocity and the vigilance decrement: Sample-level data may not represent individuals

Abstract

Vigilance, or sustained attention, tasks involve monitoring environments or displays for critical signals, which tend to be embedded amid a background of neutral signals (Davies & Parasuraman, 1982; Funke et al., 2012). Research has shown that vigilance performance tends to decline over time, known as the vigilance decrement (Helton & Warm, 2008). This poses a serious threat due to the high-risk nature of many of the domains requiring vigilance, including air traffic control, cockpit monitoring, medical screening, TSA baggage screening, and cybersecurity (Sawyer et al., 2014; Shaw et al., 2009; Vidulich, Wickens, Tsang, & Flach, 2010; Warm, Finomore, Vidulich, & Funke, 2015; Warm, Parasuraman, & Matthews, 2008). As long as the human remains a critical component in operations requiring vigilant attention, it is necessary to develop a system to ‘monitor the monitor,’ i.e., an objective detection system that can reliably predict a decline in performance before it happens. One theoretical model proposed to explain the vigilance decrement is resource theory (Warm et al., 2008). In this theory, task performance requires mental activity, such as information processing or attention, which consumes a pool of renewable cognitive resources of limited supply. Vigilance has been shown to require mental effort (Warm et al., 2008), and consequently, it has been argued that the vigilance decrement is rooted in resource depletion (Davies & Parasuraman, 1982; Warm et al., 2008). Using a noninvasive neuroimaging technology, transcranial Doppler sonography (TCD; Aaslid, 1986), to measure cerebral hemovelocity (CBFV) several studies have provided support for the resource theory by showing the vigilance decrement paralleled by a temporal decline in CBFV, or a decline in metabolic activity (Shaw et al., 2009; Shaw et al., 2013; Tripp & Warm, 2007). Such trends have prompted researchers (including the current authors) to speculate that CBFV could provide a useful indicator of operator performance in real time, and could serve as a key feature of a closed-loop adaptive aiding system to indicate when participants are in need of support. While studies of CBFV in vigilance have focused on the parallel temporal decline between CBFV and performance efficiency, the reliability of CBFV as an indicator of vigilance task performance at the individual level and across measurement occasions has not been established. Thus, due to the need for cognitive augmentation for human-in-the-loop systems, the present study was conducted to explore this possibility through an examination of the reliability of CBFV, in relation to vigilance performance, across two different sessions at the individual level. Participants in this study came in for two days of testing, and on both days performed a 40-minute vigilance task while having their middle cerebral artery CBFV monitored with TCD. It was hypothesized that vigilance performance would decline over time on both days of testing, and that this decrement would be mirrored by a decline in CBFV. Twenty-five right-handed participants (13 women, 12 men) recruited from a university in the Dayton, Ohio area participated in this study. Participants (M = 22.8 years old, SD = 4.8) completed the two data collection sessions separated by a mean of three days (SD = 2.0). Participants gave informed consent and completed electronic pre-task surveys before being fit with the TCD headset and sitting quietly for a five minute baseline recording. Participants then viewed PowerPoint training slides followed by two five-minute practice sessions with performance feedback. The participants’ goal in this task was to send a warning when displayed RPAs were on collision flight paths (i.e., critical signals) by pressing the spacebar on the computer keyboard and to withhold a response from non-collision paths (i.e., neutral signals; Hitchcock et al., 2003). The display updated 40 times per minute with a dwell time of 80 ms. For each minute of watch, eight critical signals and 32 neutral signals occurred at random, allowing for finertuned temporal analyses than the coarse ten-minute periods frequently analyzed in prior research. After training and practice, participants completed the 40 minute experimental vigil, followed by post-task questionnaires. Other than the informed consent procedure, which was administered at the beginning of the first session, both session one and two proceeded the same way for all participants. Consistent with prior research, a decrement in both vigilance performance and CBFV were evident across both sessions, and a strong correlation between sample-level performance and CBFV over time in both sessions was observed. However, when calculated at the individual level, the correlations between performance and CBFV were weak. The present results indicate that the relationship between performance and CBFV in sample-level data may not map well onto individual-level responses, and that a smooth vigilance decrement function and its relationship to a decline in CBFV may be an artifact of averaging a large group of individuals over a long period of time. Most Proceedings of the Human Factors and Ergonomics Society 2019 Annual Meeting