Abstract
Objective:
We examined the effects of spatial uncertainty, field dependence/independence (FD/I), and sex on vigilance performance and perceived workload in elementary school children.
Background:
Building on previous work in which children demonstrated their ability to evaluate workload, we tested whether spatial-uncertainty manipulations in a vigilance task would elicit in children the same deleterious effects on performance and workload as it does with adults. We also examined individual difference effects associated with FD/I and sex to determine their influence on both performance and workload.
Method:
In the low-uncertainty task, stimuli appeared in the center of the computer screen; in the high-uncertainty task, they appeared in one of the four quadrants of the screen. Neutral events consisted of uppercase letter strings. Critical signals consisted of a single lowercase letter among uppercase letters. Following each vigil, children completed a workload assessment via a modified version of the NASA Task Load Index.
Results:
Children showed lower perceptual sensitivity, greater response latency variability (RTSD), and a higher response criterion in the uncertain display condition. Workload scores reflected these performance differences. Field-dependent children showed lower perceptual sensitivity and greater RTSD than did field-independent children. The two groups exhibited differing workload profiles. Despite no objective performance differences, boys reported greater workload than girls.
Conclusion:
The scale demonstrated sensitivity and diagnosticity with regard to both the task variable and individual differences.
Application:
These findings contribute to the emerging field of “educational ergonomics” and indicate that appropriate assessment tools might identify children who are experiencing increased workload.
Introduction
The field of sustained attention, or vigilance, developed during the Second World War out of concerns for human performance in the war effort. In the seven decades since, the field has grown dramatically. Early investigations into practical applications expanded to include theoretical concerns that broadened the scope of variables and range of issues encompassed by sustained attention. Perhaps more than any single volume, the classic work of Parasuraman and Davies (1984) defined the parameters of attention research. It addressed individual differences in attention, covering topics such as personality, development, and special populations. Later, Parasuraman and his colleagues went on to apply their findings to both typical and atypical aging (Berardi, Parasuraman, & Haxby, 2005; Parasuraman, Nestor, & Greenwood, 1989). Parallel to their work, child development perspectives within the vigilance literature multiplied, largely due to interest in attention deficit disorder. However, research with children has remained largely distinct from adult vigilance research (Laurie-Rose, Pempek, & Curtindale, 2015), borrowing only rarely from the extensive foundational work laid by Parasuraman and his colleagues (Parasuraman, Warm, & Dember, 1987; Warm, Dember, & Hancock, 1996). To encourage such borrowing, we explored sustained attention and mental workload in elementary school–age children, employing attention and workload findings from the adult literature as benchmarks. Doing so, we were able to identify points of intersection between child and adult vigilance studies.
Combining child and adult research in this way, our investigation also affirmed Woodcock’s (2007) call for the study of “educational ergonomics,” a field that combines the interests of human factors and education. Indeed, Young, Robinson, and Alberts (2009) have already extended the study of vigilance and workload from the workplace to the college classroom. Our research leads us to conclude that workload studies also can be of significant use in early school settings. Teachers, for example, would benefit from understanding the costs associated with mental tasks, such as mastering difficult concepts in mathematics. An appropriate assessment might identify children who are experiencing increased workload, which might in turn alert teachers to pause, slow down, or shift to a different task.
Workload, an important construct used widely in the human factors literature, refers to a participant’s self-perceived expenditure of cognitive resources. In a typical workload assessment, adult participants answer a series of questions about their experiences with an information-processing task (Warm et al., 1996). Although there exist a variety of workload vehicles, researchers consider the NASA Task Load Index, or NASA-TLX (Hart, 2006; Hart & Staveland, 1988), to be a consistently reliable, valid measure (Wickens, Hollands, Banbury, & Parasuraman, 2013) of the following subscales: Mental Demand, Temporal Demand, Physical Demand, Performance, Effort, and Frustration. The NASA-TLX meets psychometric requirements for sensitivity (the assessment’s ability to discriminate between different levels of difficulty) and diagnosticity (the assessment’s capacity for identifying the specific locus of demand), each of which provides strong evidence for its utility (Eggemeier, Wilson, Kramer, & Damos, 1991). As a multidimensional test, the NASA-TLX allows researchers to determine a workload profile for a particular task that highlights and differentiates between its distinctive demands (DiDomenico & Nussbaum, 2011).
Metacognitive studies in the cognitive development literature (e.g., Flavell, 1979
The present study examined whether children, like adults, can evaluate workload in sustained-attention tasks. In such tasks, increases in task demand are typically paired with increased workload ratings on the Mental Demand and Frustration NASA-TLX subscales (Warm et al, 1996). The nature of increased workload in vigilance may be less obvious than workload in our previous study, as those tasks specifically tapped a single dimension of workload. For example, it targeted temporal demand by varying the presentation rate of stimuli in a reaction time task. In the current study, to facilitate differential predictions of vigilance performance and workload, we introduced the variable of spatial uncertainty. In a spatially certain task, observers of a computer display know where stimulus events to be inspected for critical signals will occur with each event presentation; in a spatially uncertain task, however, observers remain unaware of where the next stimulus events will appear (See, Howe, Warm, & Dember, 1995). With adults, spatially uncertain tasks have yielded poorer performance and increased perceived workload (Warm et al., 1996; Warm, Finomore, Vidulich, & Funke, 2015; and see Funk et al., 2017 [this issue]).
Following the lead of Parasuraman, we extended our research beyond task variables to focus also on individual differences. We chose cognitive style, which is defined as a person’s consistent mode of functioning in perceptual and intellectual activities (Witkin, 1950). From the many measures of cognitive style, we adopted field dependence/independence, or FD/I (Witkin, 1950), and we assessed this style through the Embedded Figures Test (EFT) developed by Witkin, Oltman, Raskin, and Karp (1971). The EFT requires observers to identify geometric forms embedded within a more complex pattern. Field-independent (FI) individuals readily disembed forms from their background, whereas field-dependent (FD) persons experience greater difficulty with such a task. Studies of FD/I and sustained attention reveal reliable differences, with FI observers outperforming FD observers in both adults (Cahoon, 1970; Moore & Gross, 1973; Ware & Baker, 1977) and in children (Amador-Campos & Krichner-Nebot, 1999; Guisande, Tinajero, Cadaveira, & Paramo, 2012).
Theories of cognitive style have often explained differences in performance as a result of differences in resource allocation (cf., Goode, Goddard, & Pascual-Leone, 2002; Guisande et al., 2012; Miyake, Witzki, & Emerson, 2001). With regard to FD/I specifically, this approach asserts that FI observers mobilize and allocate attentional resources more effectively than do FD observers. Recent neurophysiological research has confirmed the efficiency of FI over FD observers. Thus, assessing event-related potentials (ERP), Meng et al. (2012) provided evidence of increased activity in N270 waves in FI participants, concluding, “FI subjects have better ability in mobilizing and/or allocating mental-attentional capacity” (p. 240). Other researchers of ERP make similar claims (Goode et al., 2002; Jia, Zhang, & Li, 2014). These findings, however, generate two competing predictions regarding workload in a vigilance study with FD/I observers.
First, if mobilization requires greater resources, FI children will show superior performance but pay a commensurate mental cost with heightened workload. If, however, FI children process information more efficiently, they again will demonstrate superior performance but, in this case, will experience lower workload. Our study tested these competing hypotheses.
Here it should be noted that sex differences are reported in studies using the EFT (e.g., Li, Wu, Zhu, & O’Boyle, 2014; Zhang, 2013). Such studies report that males are more likely than females to be field independent (cf. Massa, Mayer, & Bohon, 2005). Further, sex differences on the EFT are often attributed to the spatial nature of these tasks (Zhang, 2004). For that reason, Guisande et al. (2012) argue that studies using the EFT should as a matter of course include sex as a variable. Sex differences are not consistently found in studies of vigilance with adults (cf. Berch & Kanter, 1984). However, sex differences have appeared in vigilance studies in children. Boys respond more quickly, commit more false alarms (Conners, Epstein, Angold, & Klaric, 2002; Greenberg & Waldman, 1993; Pascualvaca et al., 1997), and set a more lenient response criterion when compared with girls (Pascualvaca et al., 1997). In light of the reported sex differences in both EFT studies and in the childhood vigilance literature, our analysis kept in view the possibility of an interaction of sex with other variables.
Regarding performance, we predicted that the uncertainty condition would yield decreased perceptual sensitivity, increased conservativism, and greater response latency variability, or RTSD (standard deviation of reaction time). Researchers of attention-deficit/hyperactivity disorder frequently employ RTSD as it is purportedly sensitive to group differences (cf., Klein, Wendling, Huettner, Ruder, & Peper, 2006). RTSD appears increasingly in studies of attention with typically developing children (cf., Betts, McKay, Maruff, & Anderson, 2006). We further predicted that FI children would show superior perceptual sensitivity relative to FD children and that differences between groups would become more pronounced in the uncertain display condition. Finally, we predicted that FI children would show less variability in response latency.
Regarding workload, we expected that lower performance measures would be associated with higher workload and that the uncertain display task would yield higher workload scores than those of the certain display task. Because, as pointed out earlier, there are competing hypotheses regarding differences in workload in FI and FD observers, we attempted to shed some light on those differences.
Method
Participants
Forty-eight students, 24 girls and 24 boys (range = 9 years 6 months to 11 years 9 months, M age = 10 years 6 months) were recruited from fourth (n = 21) and fifth (n = 25) grades from one school in Westerville, Ohio, a suburb of Columbus. None of the children had been diagnosed with an attentional problem. The participants reflected the predominantly Caucasian, middle-class population of the district. Children who participated in the study were each paid $10. Institutional review board approval was secured, and parents signed an informed consent.
Measures
Workload
We employed a modified version of the NASA-TLX (Hart & Staveland, 1988) that we devised from our previous research (Laurie-Rose et al., 2014). Table 1 presents the original TLX scale definitions and their simplified definitions.
Original and Adapted Rating Scale Definitions for the Subscales of the NASA Task Load Index
Adult respondents are asked to provide a numerical value (between 0 and 100 using increments of 5) to the following six subscales: Mental Demand, Temporal Demand, Physical Demand, Performance, Effort, and Frustration. We asked children to place a hash mark on a 100-mm line to indicate how much each factor contributed to their experiences on the task. Happy and sad faces placed at the appropriate end points provided reminders to the children of anchor directions. We encouraged children to use the full range of the line. Scores were determined by measuring the placement of the hash mark on the 100-mm line and then adjusting that number to correspond with the nearest multiple of 5. In traditional NASA-TLX administration, adults are asked to weight the various subscales of workload. Nygren (1991) and, more recently, Wiebe, Roberts, and Behrend (2010) reported virtually no difference between the weighted and unweighted administrations. To simplify the administration for children, we used only unweighted scores.
Vigilance
Adapting tasks developed by Mouloua and Parasuraman (1995), we manipulated display condition by designing two vigils: spatial certainty and spatial uncertainty. In both vigils, all parameters were identical except for the placement of the stimuli within the visual display. We prepared a set of 12 four-letter strings that served as the basis for both the neutral events and critical signals. The letter strings were derived from the following letters: A, B, C, D, H, T. We chose these letters because of their high discriminability from each other. Neutral events displayed each of the 12 strings in uppercase letters. Critical signals displayed the same 12 strings with a single lowercase letter appearing among the three uppercase letters. Within the critical signals, lowercase letters appeared equally often in the second or third ordinal position. The letter strings appeared in bold, black Times New Roman (font size of 46, measuring 3/8 × 3/8 inches) against a white background. The vigilance tasks were presented via a Macintosh iBook computer with participants sitting approximately 15 inches from the computer screen. The vigils ran without interruption for 14 min. The vigilance data, however, were divided into seven continuous 2-min periods of watch. All participants performed the task individually in a well-lit environment. The stimuli were presented for 200 ms at a rate of 30 events per minute with target probability set at 0.2. A total of 60 events—12 critical signals and 48 neutral events—appeared per period. For the purposes of stimulus placement, we centered an “invisible” 7.5 × 7.5-inch square on the computer screen. In the certain condition, all stimuli appeared in the center of that square. Each critical signal appeared randomly once, and each neutral event appeared randomly four times over a period of watch. In the uncertain condition, the stimuli appeared in one of the four quadrants of the computer screen, 1.5 inches above or below and to the left or right of the perimeter of the “invisible” square. Within each period of watch, the 12 critical signals were assigned at random to each of the four quadrant locations, with three appearing at each location. The 12 neutral events appeared four times each, randomly appearing once at each quadrant location. A response to a critical signal was considered a hit if it occurred before the next stimulus was presented. A response to the neutral stimulus was considered a false alarm.
Cognitive style
We used the Children’s EFT (Witkin et al., 1971) to assess cognitive style. The experimenter asked the child to identify the simple figure within the complex form. The experimenter allowed 180 s for the child to find the simple embedded figure in each of 12 cards. Per the instructions, unsolved trials were assigned a score of 180 s. A child’s final score reflected the average time of all 12 trials.
Procedure
Participants performed both the spatial-certainty and spatial-uncertainty tasks. The order in which participants performed the two vigilance tasks was balanced as closely as possible across sex and grade. Participants worked individually in a small, quiet room. Children were instructed to press a button on a keyboard whenever a lowercase letter appeared in a letter string. Children were given a 2-min practice session prior to each vigil that was identical to the main session. In the first minute, the experimenter provided feedback to correct responses, misses, and false alarms. During the second minute—upon which the criterion was set—the experimenter observed silently as the children completed the remainder of the practice session. All children completed the second practice with no more than three misses and fewer than five false alarms. Following each vigil, the children completed the modified NASA-TLX. Children were given a 5-min break between vigils during which time they could take a walk or visit the restroom. Following the administration of the second NASA-TLX, children completed the EFT.
Data Analysis
Using a split-half procedure, children were designated as FI or FD based on their EFT scores. Researchers examining individual differences in vigilance often adopt this procedure (e.g., Helton, Dember, Warm, & Matthews, 1999; Matthews, Jones, & Chamberlain, 1989). Group means for boys and girls in the FD/I categories are presented in Table 2.
Means and Standard Deviation (in seconds) for Completion Time of Embedded Figures Test for the Field-Independent (FI) and Field-Dependent (FD) Boys and Girls
Exploratory analyses of task order did not reveal any major effects, and for that reason, we collapsed over order. We employed a 2 (certainty: certain vs. uncertain) × 7 (period of watch) × 2 (cognitive style: FI vs. FD) × 2 (sex) mixed ANOVA, with certainty and period of watch serving as the repeated-measures variables for the dependent measures of d′ (perceptual sensitivity), c (response bias), and RTSD. For workload, we employed a 2 (certainty: certain vs. uncertain) × 6 (subscale) × 2 (cognitive style: FI vs. FD) × 2 (sex) mixed ANOVA with certainty and subscale serving as the repeated-measures variables. For all analyses, Mauchly’s test of the violation of the assumption of sphericity was conducted. Where applicable, the Huynh-Feldt correction was employed.
Results
Performance
Means and standard deviations for the d′, c, and RTSD performance measures in all experimental conditions are presented in Table 3.
Means and Standard Deviations of Performance Measures in All Experimental Conditions
Note. C = certain; UC = uncertain; FI = field independent; FD = field dependent; RTSD = response latency variability.
Perceptual sensitivity (d′)
We discovered a significant effect of certainty on d′, F(1, 44) = 176.08, p < .001, ηp2 = .80, with d′ significantly higher in the certain condition (M = 3.54) than in the uncertain condition (M = 2.25). We also found a significant effect of cognitive style, F(1, 44) = 9.93, p = .003, ηp2 = .18, with FI children exhibiting a significantly higher d′ (M = 3.16) than FD children (M = 2.63). The remaining sources of variance were not significant (p > .05).
Response bias (c)
We again noted a significant effect of certainty, F(1, 44) = 50.42, p < .001, ηp2 = .53, with response bias less conservative in the certain condition (M = .24) than in the uncertain condition (M = .63). In addition, we also uncovered a significant main effect of period of watch, F(5.45, 239.80) = 2.61, p = .022, ηp2 = .06, with c generally increasing over periods of watch (means for Periods 1 through 7 were .37, .38, .43, .46, .45, .45, .49, respectively). Finally, we noted a significant interaction between cognitive style and sex, F(1, 44) = 6.95, p = .012, ηp2 = .14 (see Figure 1). All remaining sources of variance failed to reach significance (p > .05).
Mean response bias (c) scores for field-independent and field-dependent boys and girls. Error bars represent standard errors.
In exploring the nature of the interaction, a simple effect analysis indicated that FI boys (M = .42) and girls (M = .49) adopted similar criteria, F(1, 22) = 1.45, p = .241, ηp2 = .06. In contrast, FD girls (M = .27) adopted a significantly more lenient criterion than FD boys (M = .54), F(1, 22) = 5.50, p = .028, ηp2 = .20.
RTSD
There were two instances in the uncertainty condition where children obtained only one hit for a particular period of watch. In these cases, we could not calculate standard deviations; therefore, degrees of freedom vary across the analysis. Participants demonstrated significantly less variable reaction time in the certain condition (M = 143) than in the uncertain condition (M = 174), F(1, 42) = 15.66, p < .001, ηp2 = .27. We also uncovered a significant effect in standard deviation across periods of watch, F(6, 252) = 2.94, p = .009, ηp2 = .07 (means for Periods 1 through 7 were 150, 142, 182, 171, 151, 162, and 154, respectively). Further, FD children were significantly more variable in reaction time (M = 174) than FI children (M = 144); F(1, 42) = 6.00, p = .019, ηp2 = .13. All remaining sources of variance were not significant (p > .05).
Workload
Means and standard deviations are presented for all experimental conditions in Table 4. Perusal of Table 4 and subsequent analyses reveal that overall workload scores were significantly higher in the uncertain condition (M = 51.4) than in the certain condition (M = 32.9), F(1, 44) = 85.62, p < .001, ηp2 = .66.
Means and Standard Deviations for Subscales in all Experimental Conditions
Note. C = certain; UC = uncertain; FI = field independent; FD = field dependent.
There was a significant effect of sex on workload ratings, F(1, 44) = 7.21, p = .010, ηp2 = .14, with boys (M = 47.8) perceiving greater workload than girls (M = 36.6). Significant main effects for subscale, F(5, 220) = 11.78, p < .001, ηp2 = .21, and certainty (noted earlier) are best interpreted in light of two interactions: a significant Certainty × Subscale interaction, F(5, 220) = 3.83, p = .002, ηp2 = .08, and a significant FD/I × Subscale interaction, F(5, 220) = 2.35, p = .042, ηp2 = .05. All remaining sources of variance did not reach significance (p > .05).
The Certainty × Subscale interaction is presented in Figure 2.
Mean workload scores for each of the six subscales for the certain and uncertain conditions. Error bars represent standard errors.
Single-factor ANOVAs revealed that the effect of subscale was significant for the uncertain, F(5, 235) = 12.32, p < .001, ηp2 = .22, and certain, F(5, 235) = 5.77, p < .001, ηp2 = .11, conditions. In the uncertain condition, subsequent pairwise comparisons using Bonferroni-corrected t tests (familywise α = .05, df = 47) indicated that Mental Demand, Temporal Demand, and Effort subscales received significantly higher ratings than Physical Demand (t = 5.12, 4.31, 6.36; d = .82, .86, 1.07, respectively) and Performance (t = 3.83, 4.09, 4.63; d = .62, .68, .90, respectively), and Frustration received significantly lower ratings than Mental Demand (t = 3.23, d = .44) and Effort (t = 3.80, d = .65). The difference between Frustration and Temporal Demand approached significance (t = 3.00, d = .50, p = .065.) In the certain condition, Temporal Demand was rated as significantly higher than Mental Demand (t = 3.16, d = .45), Physical Demand (t = 3.09, d = .55), and Performance (t = 4.09, d = .84); and Performance received lower ratings than Effort (t = 3.76, d = .72).
The FD/I × Subscale interaction is presented in Figure 3.
Mean workload scores for each of the six subscales for the field-independent and field-dependent children. Error bars represent standard errors.
Single-factor ANOVAs revealed that FI children rated their workload differently across the subscales, F(5, 115) = 14.37, p < .001, ηp2 = .38, but FD children did not manifest significant differences in ratings across the subscales, F(5, 115) = 2.21, p = .058, ηp2 = .09. Subsequent pairwise comparisons using Bonferroni-corrected t tests (familywise α = .05, df = 23) revealed that FI children rated Physical Demand and Performance significantly lower than Mental Demand (t = 4.45, 5.35; d = .91, 1.09, respectively), Temporal Demand (t = 5.67, 5.87; d = 1.21, 1.21, respectively), and Effort (t = 4.13, 5.77; d = .96, 1.07, respectively).
Discussion
Expanding upon our previous investigation of workload in children (Laurie-Rose et al., 2014), we tested whether performance and workload following a vigilance task would align with findings observed with adults. Informed by the vast literature on adult vigilance, we examined both task and individual difference effects. For our task variable, we chose spatial uncertainty, a powerful determinant of performance and workload in adults (cf. Warm et al., 1996, 2015; and see Funke et al., 2017 [this issue]). The findings confirmed our hypothesis that spatial uncertainty would exert deleterious effects on workload; these evaluations matched objective performance. To determine the impact of individual differences, we explored whether cognitive style and sex would exert an influence on performance and workload. The individual difference effects supported our hypotheses that FI compared with FD children showed superior performance. Although FI and FD children did not differ in overall workload, they provided significantly different workload profiles.
The present study points to important parallels between adult and childhood vigilance. Significantly, the psychophysical parameter of spatial uncertainty reveals itself as a significant determinant of vigilance performance in children. Like previous studies with adults, it produced lower perceptual sensitivity and a more conservative response criterion (Parasuraman, 1984). Response latency variability also proved sensitive to task manipulation, a finding that paralleled both the sensitivity and bias scores. Individual difference effects associated with cognitive style recalled earlier findings with adults (Cahoon, 1970; Moore & Gross, 1973; Ware & Baker, 1977). Compared with FD children, FI children showed superior performance as measured by both perceptual sensitivity and RTSD. It is important to note, however, that FD/I is substantially correlated with spatial ability (MacLeod, Jackson, & Palmer, 1986). Therefore, we may be observing ability, not cognitive style, differences.
The finding that FD girls set a more lenient response criterion than did FD boys conflicted with extant reports that girls tend to be more conservative responders (e.g., Pascualvaca et al., 1997). Unfortunately, signal detection measures are not widely used in childhood vigilance studies. Authors of further research should, therefore, explore whether this finding reflects a sampling issue or indicates a unique processing difference between the groups.
The current study advances our earlier research of workload in children (Laurie-Rose et al., 2014) and provides strong confirmation that children possess the ability to evaluate workload induced by vigilance tasks. In our previous study, not only were tasks brief and gamelike, but they also afforded the children the opportunity to interact with the researchers. By comparison, the present tasks were longer in duration, were monotonous, and required the children to work alone. Despite such adverse conditions, the scale demonstrated both sensitivity and diagnosticity. The average rating for the uncertain task in this study was 51.4, whereas the certainty task earned a rating of only 32.9. By contrast, depending on the task demand, average workload scores for adults often reach beyond the midrange of the scale (50). In our study, children’s scores may have been lower, in part, due to the low ratings for frustration. Adults typically rate frustration higher following a vigilance task (Warm et al., 1996). Future studies should provide a direct comparison of workload ratings between children and adults.
Several findings established the scale’s sensitivity. Children showed greater workload with lower perceptual sensitivity, greater RTSD, and a more conservative criterion in the uncertain condition. The workload scale revealed its sensitivity further by eliciting meaningful individual differences. Boys experienced greater overall workload than girls. Thus, although both boys and girls performed the tasks equally well, the tasks required higher mental expenditure in boys than girls. The behavioral requirements of the task—sitting still, looking at the computer, remaining vigilant—may be more taxing for boys. At first glance, this finding might seem counterintuitive, as it is well documented that boys, as compared with girls, spend many more hours playing video games (Homer, Hayword, Frye, & Plass, 2012). But these visually rich games display regularly occurring events that continuously advance the game’s action. By comparison, the computer-based vigilance task must seem quite boring. Under such circumstances, boys can perform at high levels but at a cost in perceived workload.
The scale demonstrated its diagnosticity by demarcating subscale dimensions important to workload. Warm et al. (1996) identified Mental Demand and Frustration as contributing most to workload in vigilance tasks. Our findings did not fall in line with this profile. We did, however, observe different profiles depending on task condition and FD/I. In the certain condition, although children rated Temporal Demand highest, the task evoked little discrepancy among remaining subscales. In the uncertain condition, children’s ratings varied across subscales, with Temporal Demand, Mental Demand, and Effort receiving the highest ratings. These diagnostic findings have an intuitive appeal. Children reported high temporal demand for the two tasks, in both of which the stimuli appeared for a brief 200 ms. With the certain display, however, the condition remained relatively easy; the uncertain display imposed the overall greater demands. As a final indication of the scale’s diagnosticity, the FI children’s ratings showed significant variation across the subscales, whereas FD children showed no such fluctuation. Regarding FI children, Temporal Demand, Mental Demand, and Effort ratings reflected the highest workload. Regarding FD children, it is possible that these children could not adequately discriminate between the varying subscales. Our earlier work (Laurie-Rose et al., 2014) revealed a similar absence of discrimination between subscales in first and second graders. Further research should explore whether certain groups differ in their ability to differentiate between the subscales.
Taken together, the findings on performance and workload give partial support to our hypotheses regarding the resource allocation model. As predicted, children, regardless of FD/I designation, performed more poorly and rated workload significantly higher in the uncertain compared with the certain condition. FI children outperformed FD children, and although we did not observe a main effect for FD/I in workload, the workload profiles differed significantly from each other. Although lack of adequate power precludes pairwise comparison of subscales between FI and FD participants, a few cautious observations can be made. We observed a trend toward higher ratings for Mental Demand and Temporal Demand in FI children, consistent with the theory that FI children are allocating more resources to the task. Additional research should seek to confirm this supposition. We did not see an FD/I × Certainty interaction in the performance or workload data. Conforming to our predictions, both groups performed the task and experienced workload similarly in the certain condition, but counter to our predictions, performance and workload profiles were similar between the two groups in the uncertain condition. Further research should examine whether manipulation of spatial certainty is diagnostic for this particular group.
Two interesting findings emerged regarding sex. First, girls set a more lenient criterion than boys, a finding that runs counter to extant data on response bias in girls (Pascualvaca et al., 1997) and underscores the need to examine individual differences within the context of other task variables (cf. Ballard, 1996). Second, boys experienced greater workload than girls, in contrast to results reported by Dittmar, Warm, Dember, and Ricks (1993) that college-age males experienced less workload than females. Although no overly broad conclusions should be drawn from these findings, they do suggest that, whenever possible, sex should be included and analyzed in studies of vigilance and workload.
Whereas the focus of our study was the workload associated with vigilance performance, the vigilance decrement bears mentioning. Response bias scores showed a general increase toward conservatism, and RTSD scores showed some evidence of increased variability across watch. The sensitivity decrement, however, was absent. We can conclude that our vigils did not impose all of the typical time-on-task effects often observed in vigilance research. Two factors that may have led to the absence of a perceptual decrement in our study are signal salience and signal probability. The difference between the target and neutral events was easy to detect, and the level of signal probability (20%) was higher than the 3% to 6% levels typically used in vigilance tasks with adults as described by Davies and Parasuraman (1982). High levels of signal salience and signal probability are important factors in modulating the perceptual sensitivity decrement in vigilance tasks (See et al., 1995; Thompson, Smilek, & Besner, 2015). We note that the absence of the sensitivity decrement in our study raises the question of the equivalency of vigilance tasks across age. We acknowledge the importance of this equivalency issue and have explored its ramifications elsewhere (Laurie-Rose, Bennett-Murphy, Curtindale, Granger, & Walker, 2005).
The present study demonstrates the important advantage of greater communication between researchers of child and adult attention (Laurie-Rose et al., 2005, 2015). In the first place, as the task variable of spatial uncertainty indicated, similarities exist between children and adults in vigilance performance and workload outcomes. Further, as our scale demonstrated, children can provide workload ratings that correspond meaningfully to their performance and so give valid measurements of those outcomes. Informed by research of atypical and typical development, the study also showed that RTSD provides a sensitive measure for the assessment of individual differences in vigilance. This measure now merits testing with adults: It might bring to the surface individual differences in adult vigilance, an area that has often yielded inconsistent findings (cf. Koelega, 1992).
In sum, our test, which focused on the application of a new methodological instrument, offered us a novel research approach in childhood attention. It strengthened our belief that workload, as it does with adults, serves as a highly useful, even essential, tool for studying and conceptualizing childhood vigilance. Finally, these findings contribute to the emerging field of educational ergonomics and indicate that human factors research may offer unique insights and practical solutions for educators in the elementary school classroom.
Key Points
The variation in the vigilance task demands confirmed the scale’s sensitivity. We observed lower perceived workload scores in a task with a certain as opposed to an uncertain display. The performance data corroborate the perceived workload data. Children showed lower perceptual sensitivity, greater variability of standard deviation of reaction time (RTSD), and an overall more conservative criterion in the uncertain display condition. We observed different subscale profiles depending on task condition.
Field-independent (FI) children showed greater perceptual sensitivity and less variability in response latency than field-dependent (FD) children. The scale demonstrated diagnosticity by delineating different workload profiles for the two groups.
The response time variability measure (RTSD) uncovered significant differences between FI and FD children. Researchers of both typical and atypical development include this measure as it is purportedly sensitive to group differences. We argue that this measure may prove useful in adult studies as well.
Boys experienced greater overall workload than girls. From the performance data, we can conclude that although both boys and girls perform the tasks equally well, the tasks require greater mental expenditure for boys than for girls.
Children can meaningfully describe their workload experiences in a sustained-attention task. We propose that determining workload demands on children can provide a more complete account of a child’s experience in clinical and school settings.
Footnotes
Acknowledgements
Cynthia Laurie-Rose acknowledges Otterbein University for a sabbatical leave and research funds for supporting this study. The authors would like to thank their action editor, Joel Warm, and two anonymous reviewers for their helpful comments on earlier drafts of this manuscript.