Wagner, Sahar, Elbaum & Berliner

GRIP FORCE AS A PHYSIOLOGICAL MEASURE OF STRESS IN TRACKING TASKS

M. Wagner1, Y. Sahar2, T. Elbaum2 and E. Berliner3

1Department of Industrial Engineering & management, Ariel University, Ariel,44837, Israel.

2Faculty of Industrial Engineering & management, Technion, Haifa, 32000, Israel.

3Department of Management, Bar Ilan University, Ramat-Gan, 5290002, Israel.

Abstract

Objective stress measures in psychomotor tasks, such as flight simulator training, are strongly needed. The Yerkes-Dodson curve(inverted U relationshipcurve between performance and stress) requires monitoring of stress level if training is to be optimal.Here we examined the feasibility of joystick Grip Force as a stress measure. Nine male participants performed 2D tracking tasks composed of 3 dynamic target maneuvers, with 4 constant-velocity levels, and a secondary monitoring task. Course credits conditioned by performance level served as additional stressor. Galvanic Skin Response served as physiological validation measure. The Grip-Force sensor was embedded in the joystick grip, concealed from participants’ awareness. Our results support the feasibility of Grip-Force as a valid and reliable stress measure. Contrary to other physiological stress sensors, grip-embedded Grip Force sensor is “transparent” to the measured person and technically simple. Various applications and limitations of the Grip Force measurementtool are discussed.

Keywords:Stress, Grip-Force, GSR, Psychomotor, Tracking.

1Introduction

1.1 Motivation

This study was motivated by two elements, first is real life pilot report, the second is more scientific. Important but non-scientific source for the “Grip Force” measure, originates from unpublished military aviation medicine studies, and reported pilot experience. Pilots are fully aware of their stick-grip-force responses in stressful flight scenes. Among pilots, the “white finger syndrome” is well known, describing the extreme stress response, where the pilot’s fingers turn white resulting from extreme muscle tension. The second motivating element is the need for an objective and suitable stress measure to be applied in training facilities such as flight simulators. The idea of “Grip Force” as candidate measure stems from scientific research findings showing increased muscle-tonus as one of spontaneous stress responses(Balson, Howard, Manning Mathison, 1986; Bozdemir, SaricaDemirkiran, 2002; Liao, Zhang, Zhu, Ji& Gray,2006).

Training can be a very expensive and hazardous task, especially in the aviation field. Flight simulators are designed to solve these problems and to supplement the training procedure, providing a safe training environment in which extreme situations can be trained (MoroneyMoroney, 1999). However, simulator training time may still be an expensive resource (Smith, Gevins, Brown, Karnik & Du, 2001) and therefore training efficiency is highly valued (Salas, Bowers & Rhodenizer, 1998).

At the beginning of the 20th century Yerkes and Dodson argued that there is an inverted U shaped relationship between stress and motor task performance. Low or high sress values impair performance while mid range stress values are optimal for motorskill performance (Kramer & Weber, 2000). This phenomenon, which is called "Inverted U", has been demonstrated in many cases (Anderson, 1976; Hancock & Warn, 1989). Therefore, the need for a simulator-integrated stress measurement tool, providingreal time stress values during Flight Simulator (FS) runsis self-evident.Such a tool could aid identification of the optimal stress point, and in turn maximize training-process efficiency.

Some physiological stress measurement methods are well astablished, such as: Galvanic Skin Response (GSR), Heart Rate (HR), Pupil Diameter (PD) and more (Gopher & Donchin, 1986; Yerkes & Dodson, 1908). Although widespreadof these methods, they are unsuitable for the flight simulator environment, due to the following main reasons:

First, these measures are complex and require a subtle integration process into the FS environment (Fournier, Wilson & Swain,1999; Smith, Gevins, Brown, Karnik & Du, 2001; Van Orden, Limbert, Makeig & Jung, 2001). Second, the output of these tools requires an"over time" statistical analysis getting at the stress value,therefore not allowing a quick and real time value of the trainee's stress level (Benoit et al., 2009). Third, these measures are characterized by having some delay in reading the stress signals and therefore making online measurement impossible (Benoit et al., 2009). Forth, most of the physiological measures are salient to the trainee and therefore may cause interference to the main training simulated task performance(Van Nimwegen & Uyttendaele, 2009).The Force Sensitive Resistance (FSR) is a measure of Grip Force. This measure overcomes most of the shortcomes of the conventional physiological stress measures.

The FSR senses the force exerted upon the joystick or steering wheel surface by the trainee’s gripping hand. Since stress manifests an increased muscle tonus(Balson, Howard, Manning Mathison, 1986; Bozdemir, SaricaDemirkiran, 2002; Liao, Zhang, Zhu, Ji & Gray,2006), measuring the grip force may indicate the level of the trainees’ stress.

The FSR measure is characterized by its immediate response. In addition, its manifestation is dynamic and variable.Here, the force sensor was integrated in the flexible fiber cover of the control grip, whileparticipants were totaly unaware of its existance. This simple tool could be easily integrated in flight simulators as well as other types of trainers. The Grip Force measure -output can be easily interpreted into real time levels of trainee’s stress. Finally, this measure's face validity is strong, comparing to the other stress measures mentioned earlier.

1.2 Stress

Hans Selye(1936) initially defined stress. According to this definition, a percieved threat to an organism raises the need to mobilize resources in order to enable it to react, in one of three possible ways of action: fight the threat, flee away from it or freeze (hoping to go unnoticed).

The two most dominant theories concerning stress stem from cognitive or physiological explanation of behavior. According to the cognitive approach, stress is the outcome of a gap between a subject's perception of a threatening situation's demands and his perception of the available resources to cope with it (Lazarus, 1966; LazarusLaunier, 1978). Later definitions emphasize the role of the perceived importance of success (Staal, 2004). These theories underlie the cognitive interpretation of the Stressful stimulus.

The physiologic approach to stress defines it as a threatening situation that stimulates the hypothalamic-pituitary-adrenal axis(the HPA axis, a major part of theneuroendocrine systemthat controls reactions tostress) and the sympathetic system, resulting with the release of stress hormones, which influence the organism's behavior in order to help it cope with the threat (De Kloet, Joëls & Holsboer, 2005). These theories underlie the physiological reaction to the Stressful stimulus.

It should be emphasized that both approaches refer to performance when discussing stress.

1.3 Performance

Performance is defined as the accomplishment of a giventaskmeasuredagainstpre-set known standards of accuracy, completeness,cost, and velocity. Performance measuresare typically associated with one of four categories: measures of velocity or time, measures of accuracy or error, measures of workload or capacity demands and measures of preference (Wickens & Hollands, 2000).

1.4 Stress and performance

According to theYerkes–Dodson law, performance increases with physiological or mental stress up to a certain point. When levels of stress become too high, performance decreases(Kramer & Weber, 2000). This approach argues that optimal performance is acheived at the central levels of stress. Thus, In order to ensure that the stress levels are optimal during task execution one has to monitor the level of stress of the subject.

1.5 Stress measurement

One of the most common methods of measuring stress is through self report measures, namely questionnaires. However, this method has been criticized for discrepancies between its results and the result of validated physiological measures (Bourne & Yaroush, 2003; Gopher & Braune, 1984).

Another method of measuring stress is based on the physiological definition of stress. Since the endocrine system has a major role in the physiological stress mechanism, measurement of stress related hormones may be the most accurate and unequivocal methods of measuring stress (Baum & Grunberg, 1995). However, this procedure is expensive and may take a very long time and cost a large sum.

One of the most prevalent measures of stress is the physiological GSR measure (Kramer, 1991). It is relatively easy to use, but its sensitivity to varied psychological arousals makes it difficult to distinguish between stress and emotions such as sorrow. A possible solution for this challenge is making sure that the task at hand is designed to induce stress, by creating a psychological threat or danger (Lazarus & Eriksen, 1952). Therefore, a threatening and dangerous situation that is followed by a rise in the GSR measure may lead to the conclusion that there is a rise in the level of stress (Perala & Sterling, 2007).

1.6 Hypotheses

Hypothesis 1: In line with the physiological approach, and sincestress is manifested by increased muscle tonus, we hypothesize a positive linear correlation between FSR and GSR, so thatthe higherthe value of the FSR measurement,the higherthe value of the GSR.

Hypothesis 2: In order to validate task difficulty manipulation, we hypothesize thatcomplex target manoeuvresand higher target velocities produce lower performance (are more difficult task conditions).

Hypothesis 3: Inline with the cognitive approach, and given that hypothesize 2 is confirmed, we hypothesize thatdifficult task conditions (complex target manoeuvresand higher target velocities) induce higher FSR measurement values.

2Method

2.1General

In a specially arranged laboratory setup, participants used a joystick for tracking a 2D-manoeuvring target on a screen, and simultaneously performed a monitoring task using the joystick trigger, as a secondary task. Each trial lasted 45sec. Levels of tracking difficulty were manipulated by 4 levels of target velocity, and 3 modes of target manoeuvring profiles.

A GSR electrode was attached to the participants’ left-hand pointing finger. A special sensor recorded joystick-grip-force without Participants’ awareness. All tracking data was recorded for later statistical analysis. Participants were rewarded for their participation by bonus points in an academic course grade,conditioned to their individual performance. This added “psychological threat” in line with the cognitive approach to stress, and served as an additional stressor.

2.2Participants

The participants were 9male industrial engineering bachelor students, aged 20-30 years, all right handed and normal or corrected eye sight.The participants were not “gamers” (didn’t spend more than four weekly hours engaging in similar tasks such as computer games and domestic flight simulators). Participants who did not reach performance criterion (see section 2.5 - procedure) were ruled out so that the participants reach a uniform level of medium to high performance at the beginning of the experiment. Thus, preventing the effect of external factors such as skill differences and training effect, and strengthen the external validity regarding skilled participants (the target population).

Since the task was demanding, only 5 Participants carried out all 5 sessions, the other 4 Participants completed only three sessions.

2.3Tasks

Theparticipant performed a main and a secondary task simultaneously.The main task was tracking a 2D movingtarget on the screen, with a joystick controlled crosshair. The 12 target movement conditions were composed of 3 target maneuvering profiles and 4 constant-velocity levels.

The target velocities were:(pixel units- 50/sec (1.5°/sec), 80/sec (2.4°/sec), 110/sec (3.3°/sec), and 140/sec (4.2°/sec)). The three manoeuvring types were: P – straight lines and sharp turns, M – rounded lines and turns and D – the same as P plus 1 second target disappearances.

The secondary task wasa compensatory-tracking task: a small unstable rectangle inclined to break out of a bigger rectangle, and could be returned to center by joystick trigger-press. This task was displayed in the lower third of screen area.

Fig.1: Schematic study screen-view: Main task (track a moving target joystick controlled crosshair), red disk and black crosshair, and secondary (return the inner rectangle to center by joystick trigger-press) tasks.

2.4 Apparatus and Measures

2.4.1 hardware

The display screen was an AlienwareOptX AW2210, 21.5-Inch monitor, 1920x1080 resolution and 120Hz. refresh rate. The joystick controllerwas the Saitek X52-pro. The physiology measurement systems (GSR and FSR sensors, controllers, amplifiers and operation software) were manufacturedby “Atlas Engineering”.

2.4.2 software

The operating system we used is windows 7 SP1 x 64 bits. We engaged the development environment C# .net 4.0 and XNA 4.0 build x86 (32 bit).

2.4.3 measures

The main performance measure was percent of time the crosshair was within the target area.

2.5 Procedure

Participant attended the laboratory on two different days, for test sessions which lasted about 90 minutes each. During the first session Participants received instructions and were asked to read and sign their consent for participation.

The Participants were informed that their experimental task performance level would influence their reward (in form of bonus points at an academic course grade).Following the explanation, Participants went through several visual-tests to ensure proper vision and filled in a short questionnaire, regarding personal details

The experimenter demonstrated the experimental task and then the Participants began a training phase including several levels of difficulty. Initially they completed the easiest level. After meeting a pre-defined criterion they progressed to the next level and so on, up to meeting the final level criterion of performance.Participantswho completed 20 training trials without meeting the performance criterion(the crosshair within target area at least 70% of the time) were ruled out (see section 2.2- participants, for more details) and did not proceedto the experimental phase.

After the training phase, the Participants completed 3 to 5 experimental sessions, while monitored by the physiologic sensors:GSR and FSR. All the sessions consisted of 12 trials, 45 seconds long each. Every session included all possible combinations of four target velocities and three target manoeuvres, randomly ordered. After each session Participants were informed of their tracking performance.

Fig. 2:view of experimental setup.

3Results

3.1 Statistical Analyses

To test our Hypotheses we performedthree major analyses. First, we conducted a bivariate regression analysis of FSR vs. GSR to examine the linear correlation between the two variables (hypothesis 1).

Second, we conducted a two factors anova (4 target velocities X 3 target maneuvers) on participants’ performance, to examine weather complex target manoeuvres and higher target velocities are more difficult (hypothesis 2).

Third, we conducted a two factors anova (4 target velocities X 3 target maneuvers) on participants’FSR level, to examine whetherdifficult task conditions induce higher FSR measurement values(hypothesis 3).

3.2 Regression Analysis of FSR vs. GSR (hypothesis 1)

We performed two bivariateregressions analyses of FSR vs. GSR, in order to investigate the relationship of FSR and GSR throughout the experiment. First we conducted a regression analysis with all 9 participants who started the experiment (4 participants didn’t finish the experiment due to high task demands, (see section 2.2, participants)).Thereafter we conducted a regression analysis with only 5 participants who completed all 5 experimental sessions.

Unstandardized Coefficient, B / B standard error / standardized Coefficient, Beta / t
Bivariante Reggression with 9 participants
(no performance criteron) / 0.44 / 0.06 / 0.45 / 7.20***
Bivariante Reggression with 5 participants
(4 filltered due to performance criteria) / 0.85 / 0.08 / 0.75 / 11.31***

p<0.001***

Table 1:Bivariante regression coefficients of FSR vs. GSR, for 9 participants (including 4 participants who didn’t finish the experiment due to high task demands) and for 5 participants who completed all 5 experimental sessions.

Regression results for9participants show that there is a good linear correlation between participants’ measures of FSR and GSR (β=0.44, R²=0.20, F(1,210)=51.84, p<0.001). Thus, the higher the FSR measurement level,the higher the GSR level (see table 1 for regression coefficients).

Regression results for5participants show a very strong linear correlation between participants’ measures of FSR and GSR (β=0.85, R²=0.56, F(1,102)=127.91, p<0.001). Thus, the higher the FSR measurement level, the higher the GSR level (see table 1 for regression coefficients).According to these results, participants’ FSR values explain 56% of participants’ GSR values, through the following prediction equation:
GSR= (-148.17) + 0.85*FSR (see figure 3).

Fig. 3:Regression line of FSR vs. GSR for 5 participants who completed all 5 experimental sessions.

3.3 factorial anova of target velocities and target maneuversonparticipants’performance (hypothesis 2)

In order to examine whether higher target velocities and complex target maneuverselicit lower participants’ performance, we conducted a two factors anova analysis. The following graph reflects the participants’ performance level (percent of time the crosshair was within the target area) for each combination of the 12 task conditions (4 target velocities X 3 target maneuvers).

Fig. 4:The graph displaysthe mean performance (percent of time the crosshair was within the target area) of each combination of the 12 conditions- 3 target maneuvers (P – straight lines and sharp turns, M – rounded lines and turns and D – the same as P plus random one secondtarget disappearances)by 4 target velocities (pixel units- 50/sec, 80/sec, 110/sec, and 140/sec).

Results showa strong and significant main effect of target velocity on performance, (F(3, 213) = 209.340, p < .001). The multiple comparisons post hoc test revealed that the participants performance level was significantly different (p < .001)for each of the 4targetvelocities (pixel units- 50/sec (1.5°/sec), 80/sec (2.4°/sec), 110/sec (3.3°/sec), and 140/sec (4.2°/sec)), so that the higher the velocity the lower the performance.

Results show another significant main effectof target maneuver(P – straight lines and sharp turns, M – rounded lines and turns and D – the same as P plus one second target disappearances) on performance (F (2, 213) = 11.472, p < .001).The multiple comparisons post hoc testrevealed that: participantsperformance was significantly lower on the D maneuver, compared to the P maneuver (p< .005) and compared to the D maneuver (p< .005)(see table 2 for means and standard deviations). Theparticipants’performance on theM maneuver condition was only marginally significantly lower than the P maneuvers (p = .068).

Furthermore, Results show a significant interaction effect between targetvelocity andtarget maneuver, on performance,(F (6, 213) = 4.119, p < .001). This indicates that different target velocities affected the participants’ performances differentlyin each of the three target maneuvers (and vice versa). Specifically, the M maneuver appeared to be easier (higher performance) at the low target velocity of 50 (M = 86.54, SD = 7.973),compared tothe P maneuver (M = 79.38, SD = 8.077).However, the M maneuver appeared to be harder (lower performance) at the high target velocityof 140 (M = 34.01, SD = 9.331), compared to theP maneuver (M = 37.59, SD = 9.230).