Edited by: Eduardo Salas, Rice University, United States
Reviewed by: Mathias Basner, University of Pennsylvania, United States; Gloria Rakita Leon, University of Minnesota Twin Cities, United States
This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Extreme teams help to solve complex problems outside of traditional performance environments and have significant consequences associated with failure (Bell et al.,
The National Aeronautics and Space Administration (NASA) and other space agencies seek to optimize team performance to minimize the risk of mission failure, and work with researchers from various scientific disciplines to prepare for future LDSEM missions. While meta-analytic investigations of important team relationships exist (e.g., team cognition, cohesion, composition, and performance), these investigations include traditional work team samples and findings may not necessarily generalize to the LDSEM context (Beal et al.,
Analog settings share similar characteristics of LDSEMs expected to challenge crews and possibly impinge on team dynamics. As examples, analog crews live in a confined space (i.e., small living and working spaces with minimal privacy, physical discomfort), are isolated from others (i.e., limited interaction with others outside the crew, difficulty in communicating with family), are surrounded by a harsh physical environment (i.e., an environment in which survival is not possible without special equipment), have variable workload (i.e., a high and low volume of work at different periods), and have long-duration missions (i.e., the team works together for an extended period of time). Each analog may have its strengths and weaknesses given that not all of the environmental factors may be present in a particular analog. For example, crews in Antarctic stations experience physical confinement and isolation, but are typically isolated as smaller crews for shorter periods than is expected for LDSEMs. They also have environmental cues not available in spaceflight (e.g., daylight). Crews in space simulations (e.g., HUBES, SFINCSS) may experience isolation and confinement but are typically not surrounded by a harsh physical environment.
Research on teams in analog environments has a rich history. In fact, a number of factors (e.g., compatibility and cohesion, mood, communication, conflict, performance) have been investigated in natural analogs (e.g., Antarctic; Wood et al.,
As examples, while a meta-analysis of the traditional team literature suggests that the team cohesion and team performance relationship is generally small (Beal et al.,
The psychological health of the crew is likely to be important for LDSEMs as crews will be living and working in an extreme environment for an extended duration. Communication between space crews and MC is thought to provide information about the crew's psychological health and the crew's psychological climate. Analysis of a space crew's communication with MC is the standard operating procedure of the psychological support group in Russian MC and is used to examine crews' emotional status and the communicators' coping strategies (Gushin et al.,
Some crews have reported changes in mood over time. The third quarter phenomenon is the tendency for positive mood levels to decrease while negative mood levels and conflict increase after the midpoint of the mission (Bechtel and Berning,
While a body of research examines teams in analog environments, to date, it has not been quantitatively summarized. A quantitative summary of the analog team research is important for several reasons. First, it summarizes what we know about teams in LDSEM analog environments, given the available data. Specifically, it can provide insights into how team dynamics may unfold over time for LDSEM teams, and be used to benchmark typical and atypical team dynamics in the LDSEM environment. It also can identify potential threats to LDSEM team dynamics and performance. Second, it can help guide future research in analog environments by identifying what areas are in need of more research, new areas for research, and strategies that aid with knowledge accumulation over time. Guidance for future research is particularly important given the expense and time required to collect analog research.
The primary purpose of our research was to provide an overall picture of the available data on team dynamics and performance in LDSEM-analog environments. To do this, we systematically reviewed quantitative research conducted on teams in LDSEM-analog environments. We answer two primary questions with our systematic review: (1) which factors have been examined with quantitative research, and (2) what do these studies reveal about team dynamics in LDSEM-analog environments?
Typically, meta-analysis is preferred for integrating estimates of the same relationship of interest across studies; it allows us to generate cumulative knowledge about a set of studies. The benefits of meta-analysis over narrative reviews have been widely noted (see Glass,
We sought to be as inclusive as possible while also striving to ensure that the data were relevant to understanding team dynamics in an LDSEM environment. We applied three general inclusion criteria. First, we retained sources that reported quantitative data from teams in LDSEM-analog environments, however, we excluded descriptive case studies and narrative reviews. Second, we identified and included only team-level data (as opposed to individual-level data). We excluded articles that reported individual-level data that were not tied to a particular team (e.g., Bartone et al.,
Defining an LDSEM-analog environment has challenges because a particular extreme environment (e.g., Antarctic winter-overs) may only share some of the same characteristics expected of LDSEM. All analogs are imperfect approximations of LDSEM, and researchers must weigh the importance of different features of the context in understanding the phenomena of interest. Because of this, we broadly defined LDSEM-analog research as research in which members of the focal team (e.g., the “crew” analog) live and work together for a period. We included military teams when they were expressly described as intact teams (e.g., combat teams; Ko,
We used a comprehensive search strategy to obtain quantitative research on teams in LDSEM-analog environments. Our efforts included: (1) searches of 13 databases that ranged from general databases such as Google Scholar and EBSCOhost, specialized databases such as the Military and Government Collection and space agency databases and technical report repositories (e.g., NASA, ESA, JAXA); (2) searches of specific journals such as Acta Astronautica, Aerospace Medicine and Human Performance, Human Factors; (3) contacting 29 researchers that we identified through the NASA taskbook, our project contact at NASA, or because they frequently publish in the area (e.g., Vinokhodova, Leon); (4) posts to listservs (e.g., Science of Team Science, INGRoup, relevant Academy of Management area listservs); and (5) a review of reference lists of key articles, including those from which we were able to obtain an effect size (e.g., Gunderson and Ryman,
In total, we identified approximately 309 sources (e.g., books, technical reports, dissertations, journal articles, and conference papers) for possible inclusion. To better understand the nature of the available data, we sorted the 309 sources into three categories: (1) sources that included quantitative data with a team-level sample size of 5 or greater, for which a team-level effect size between a predictor and criteria related to team functioning could be generated; (2) sources that included quantitative data on fewer than 5 teams or only data for one variable over time; and (3) sources that did not provide relevant data for our quantitative review. Sources in the third category were excluded from further review. The decision to exclude an article was agreed upon by at least two members of the research team. Seventy-two sources were retained for inclusion and coded for fidelity characteristics and other moderators, and the quantitative data on team dynamics. Of these, 11 different sources (e.g., journal articles, technical reports) provided enough information to calculate effect sizes representing the relationship between a predictor and a criterion related to team functioning, and 61 different sources reported quantitative data on team dynamics over time in LDSEM-analog environments but did not include enough data to calculate an effect size.
To extract data, two coding forms were created: one for coding effect sizes and one for coding data (e.g., means and standard deviations) related to team dynamics over time. When a source reported data on 5 or more teams and a predictor and team outcome relationships, we coded or calculated an effect size, either
Coding forms were similar in that both captured characteristics of the source, the sample, fidelity characteristics, and information about the predictor and/or criteria. In addition, a codebook with definitions of the variables and descriptions for the different categories for each variable was developed. We coded fidelity characteristics when they were described or could be reasonably assumed by two independent coders, given the descriptions provided in the sources. We used the Internet to locate information about specific simulations or Antarctic stations to complete missing fidelity information, where possible.
We coded study design as: (a) descriptive, (b) correlational, (c) quasi-experimental, and (d) experimental. We coded the degree of similarity between the sources' samples and the anticipated characteristics of LDSEM crews in terms of demographic differences (e.g., gender, national background). We coded the fidelity of the team to the characteristics expected for LDSEM crews. Studies were coded as occurring in dangerous environments when the setting had features that required individuals to use special equipment (e.g., winter-overs in Antarctic) or posed an imminent threat (e.g., polar bear threat). Studies were coded as isolated when team members were limited in physical interaction with outside parties for a substantial period of time during the study, and confined when they primarily operated in a highly restricted space. For example, winter-overs in small Antarctic stations or space simulations were coded as an isolated and confined environment. Autonomy was coded as high, moderate, low, or not reported. Many studies did not describe the level of autonomy in detail and were coded as “not reported.” Mission length was coded as the total of number of days in the team's life span. Ongoing teams such as firehouses (e.g., Kniffin et al.,
The second and third authors served as coders for this study. The primary author trained the coders on the coding scheme described in the previous section. Coders first received a coding sheet and a codebook that provided descriptive information about each category of variables. All three authors then used the codebook and coding sheet to independently code three articles. The three authors met to discuss the coding, observe areas of agreement and disagreement, and make modifications to the coding sheet and codebook. Next, all three authors recoded the initial set of articles to help establish a frame of reference that incorporated the modifications made to the coding documents. Disagreements about the coding were resolved during a follow-up meeting using a consensus approach. After the second round of coding, a common set of 5 articles was coded to determine the efficacy of the coding process and to establish decision rules. When there was little disagreement (i.e., <3 disagreements across the variables coded in the studies), two coders coded the remaining articles. A randomly sampled common set of coded articles indicated that initial agreement, prior to the consensus meeting between coders, was relatively high (mean agreement of 87% on the variables that were coded). Discrepancies between the two coders were discussed and agreement was reached using a consensus approach. When consensus could not be reached with certainty between the two coders, the coders met with the primary author to discuss how the characteristic in question should be coded. After the coding was completed, we inspected the data sets to better understand the nature of the data, to determine the appropriateness of meta-analysis for summarizing the data, and to determine the best way to summarize the available evidence.
Although we were able to locate a relatively large amount of data for our review, the small sample sizes in most studies (e.g., <5 teams) and the variety of relationships examined in the effect size studies, suggested the majority of the data did not lend themselves to traditional meta-analytic techniques. Thus, we used the following approaches. First, when the team factor and team outcome relationship could be represented using an effect size, we calculated a weighted average of the effect size from the local (analog) population and the relevant meta-analytic estimate from the traditional teams literature as a minimum-variance estimate. We used this approach as a means of balancing the precision that meta-analysis can provide in estimating a relationship across multiple settings with the high uncertainty (especially due to small sample sizes, etc.) but localness that a specific effect size generated in an LDSEM-analog environment can provide. We also calculated the average inaccuracy of the estimates and used these to create 95% credible intervals to quantify the uncertainty of the estimates.
We used equations 1, 2, 3, and A12 from Newman et al. (
Second, when the number of teams included in the study was too few to generate an effect size, and when data across studies were comparable, we descriptively summarized the data on team dynamics over time via a series of figures. We plotted team dynamics over time when data were comparable (e.g., similar scales, similar response formats), and reported for at least three different teams across at least two different data sources (e.g., articles, conference presentations). We plotted team dynamics over time in terms of mission days and over relative time. Relative time was calculated as the mission day divided by the total mission length. Relative mission time was examined given that some effects for factors such as team cohesion and conflict, are thought to different because of the point of the team in the lifespan (e.g., third quarter phenomenon) rather than the mission day itself.
PRISMA 2009 Flow diagram. Moher et al. (
Eleven sources (e.g., journal articles, technical reports) provided enough data (team
Our first research question asked: what factors related to team dynamics has quantitative research examined in analog environments? In the effect size data set, the majority of effects (i.e., 102 effects across 9 studies) represented the relationship between a predictor and team performance. Forty-seven effects across 6 studies represented the relationship between a predictor and cohesion or compatibility, and the remaining effects represented a variety of outcomes that differed across studies. The specific predictor and criterion relationship examined varied across studies. Predictors included inputs, emergent states, and team process variables (see Marks et al.,
In the benchmarking data set, team factors included emergent states, team processes, outcomes, and additional team dynamics markers. For example, emergent states included team cohesion (e.g., Allison et al.,
Our second research question asked what quantitative research reveals about team functioning in LDSEM-analog environments. We discuss the results of the weighted averages approach, and descriptive figures benchmarking team dynamics next.
We used our weighted averages approach to provide the best possible estimate of the magnitude and direction of the relationships between team factors and team outcomes in the analog environments, given the available data.
Estimated distributions for the predictor and team performance relationship in analog environments. ICE = analog team was in an ICE environment (e.g., Winter-overs in Antartica) NO = team was living and working together but not in an ICE. Hmgnt = Homogenity. Task-rel Exp = task-relevant experience. Trnsfrm Ldr = Transformational Leadership. The square represents the weighted average local validity population estimate (ρposterior) and the bar represents the 95% credible interval. Specific estimates are provided in the right column as per ρposterior [95% credible interval]. The credible interval can be interpreted as follows: there is a 95% chance that the true population predictor and team performance relationship (ρ) is between the first number and the second number. Number in the left column indicates the analog data source. 1. Gunderson and Nelson (
First, we discuss the team cohesion and team performance relationships. Studies 1, 3, and 5 (as noted in
That data suggest that with 95% certainty, we cannot speak to the direction or size of the team cohesion and team performance relationship in ICE. For example, estimate 1 reflects the estimated results for the team cohesion and team performance relationship for data collected in Antarctic stations where team cohesion was operationalized as self-rated compatibility of station members, and team performance was operationalized as self-rated station achievement. The mean estimated validity is −0.10, and with 95% certainty, we estimate that the true population validity falls between −0.46 and 0.28. This is rather imprecise, as the prediction interval includes large, moderate, and small negative effects, no effect, and small and moderate positive effects. Conversely, with 95% certainty, we can describe the team cohesion and team performance relationship in the firehouses studied as positive and small to moderate (i.e., estimate 7), and in the special operations military teams studied, as positive and moderate to large (estimate 13).
Data for a few additional relationships other than team cohesion and team performance were also available. The age homogeneity and team performance (
Taken together, there is a high degree of imprecision associated with estimates of the true predictor and team performance relationships from studies with teams in ICEs. Specifically, unlike most of the estimated relationships from teams in non-ICE, given the current data, if we retain a 95% level of certainty, we have limited to no understanding of the size or direction of the relationship of team cohesion and team performance observed in multiple ICE, age homogeneity and team performance in an ICE, and educational homogeneity and team performance in an ICE.
Next, we benchmarked team dynamics over time in studies with sample sizes too small to generate a between-team effect size, but for which data were comparable (e.g., similar measures, similar response formats) on at least three different teams from at least two different data sources (e.g., articles, conference presentations). With this requirement, we were able to generate figures on cohesion, efficiency, team conflict, communication with MC, and team mood.
While we identified several studies with cohesion data reported over time from 5 or fewer teams, these data were collected using a variety of cohesion operationalizations making it difficult to directly aggregate and make for meaningful comparisons across settings. We were able to benchmark a subset of this data by identifying 3 sources with data from 11 teams spending time together (e.g., social activities, eating meals). We classified these activities as evidence as social cohesion.
The data reported suggests some fluctuations in cohesion over time. However, two patterns are present. First, it appears team members spend more time together during shorter missions. The Concordia, Tara Drift, and Mars 500 missions lasted for 268, 507, and 520 days, respectively. In comparison to shorter missions [i.e., Emurian et al.,
Homeostat was used to collect data on team performance across a number of space simulations (e.g., HUBES, SFINCSS). Homeostat is a computer task in which, under time pressure, a team solves tasks that require the coordinated action of the whole team (Eskov,
The data suggest that three teams (i.e., a team in EXEMSI and two of the teams in SFINCSS simulations) were relatively consistent in terms of efficiency over time. The HUBES team decreased steadily in efficiency over time. One of the SFINCSS teams (Group 3) had a sharp decline in efficiency early in the simulation and then steadily increased during the remainder of the simulation.
Descriptive information on team dynamics in the HUBES and SFINCSS simulations implicate ineffective role structure and conflict as possible triggers of the performance decrements of HUBES and SFINCSS–Group 3. Specifically, in addition to measures of efficiency, the Homeostat also collects information on leadership tactics by individual team members as a means of understanding the leadership structure used while completing the task. For SFINCSS group 3, Vinokhodova et al. (
A few sources (
Gushin et al. have examined crew communication with MC in several studies (e.g., Gushin et al.,
As depicted in
It should be noted that there is a wealth of specific details (e.g., negative statements, jokes) that can be gleaned and assessed via content analysis of within- and between-group communications. Our figures here only reflect report length and total time for audio communication, which were reported in the same format across multiple teams. We refer the interested reader to Gushin et al. (
Multiple studies reported the affect of team members using Profile of Mood States (POMS; Shacham,
Team mood also has been operationalized in LDSEM-analog environments as the team mean of self-report ratings on the positive and negative mood components of the Positive and Negative Affect Schedule (PANAS; Watson et al.,
There are two key risks of bias in our systematic review. First, publication bias may be a problem, especially given the small sample sizes associated with analog research. More extreme findings are more likely to be published. Small sample sizes compound the issue because the extreme findings are less likely to replicate. Given this, we made a focused effort to obtain unpublished research. Second, there were two potential biases associated with our weighted analyses approach. Some of the effect sizes used in our weighted averages approaches were based on very small sample sizes, which may influence the normality of the local validity distribution. We based our weighted averages approach on Newman et al. (
Further, due to the limited amount of data in different analog conditions, we were unable to estimate potential bias due to certain moderators such as whether the analog study was conducted in an ICE or non-ICE environment. However, Newman et al. (
LDSEMs such as human missions to Mars are of increasing interest to NASA, space agencies, and private sector organizations. Conducting research in analog environments provides a means for understanding team dynamics for a potential LDSEM mission as well as other teams operating in similar ICE environments (e.g., oil drilling teams). Analog research on team dynamics has a long history dating back to at least the 1960's, thus it is important for researchers and agencies to learn from the past analog research to inform future analog research and prepare for future space exploration. The primary goal of this research was to summarize the existing quantitative evidence on team dynamics in LDSEM-analog environments.
Our study has three key takeaways. First, there is an extensive research base on teams in LDSEM-analog environments. We were able to locate 72 different sources reporting quantitative research. Although there are quite a few studies that have examined teams in LDSEM-analog environments, the major of the studies had too small of a sample size to generate a between team effect size. Inconsistency in how the same construct was measured across studies further limited the ability to make comparisons across studies. Second, team dynamics are dependent on specific aspects of the context. For example, the team cohesion and team performance relationship was positive and strong for teams that lived and worked together but not in isolation and confinement (e.g., special forces teams), while little could be said about the relationship between team cohesion and team performance for teams in isolation and confined environments—an important aspect of LDSEM. Further, team dynamics varied greatly over time, underscoring the importance of temporal considerations and fidelity in analog environments. Third, we were able to document and provide interesting insights into how team dynamics unfold over time. These benchmarking figures provide insights into how team dynamics may unfold over time for LDSEM teams, benchmark typical and atypical team dynamics in the LDSEM, and identify potential threats to LDSEM team dynamics and performance. More detail on specific findings is provided next.
Results from our weighted averages approach suggest that the team cohesion and team performance relationship may be operating differently in isolated and confined environments (e.g., Antarctic stations, laboratory research with ICE characteristics) than in traditional work team environments. While we can confidently state that the relationship between team cohesion and team performance in non-ICE studies (e.g., firehouses, special operations teams) is positive and small to large, and similar to previous meta-analytic estimates (Beal et al.,
As part of our quantitative review of team dynamics in LDSEM-analog environments, we also explored our benchmarking data set for trends in team dynamics over time (i.e., team efficiency, team conflict, team communication, team mood). Beginning with team efficiency, crews must coordinate and complete mission tasks in an efficient manner in order to achieve mission success (Salas et al.,
In contrast to team efficiency, intrateam conflict data greatly varied over time in LDSEM-analog settings, such that data do not show a consistent trend across teams. However, all teams reported at least one conflict within the team or with mission control by 40% of the mission completion or 90 days. Given that all teams engage in at least some conflict in extended mission, and will likely have to resolve these conflict incident rather autonomously, it is important to better understand conflict and effective conflict management strategies in LDSEM-analog settings.
With regard to team communication in LDSEM-analog settings, communication between crews and mission control is thought to provide valuable information about the psychological health of the crew and the interpersonal climate within the crew. It is interesting to note that one of the crews that demonstrated decreased efficiency over time (i.e., HUBES crew) also had shorter audio communication with mission control over time. Moreover, commanders' written communication with mission control across several missions were in line with the psychological closing phenomenon in that the length of commanders' reports to mission control decreased over time (Gushin et al.,
As for team mood—operationalized as total mood disturbance or positively affectivity—there was inconsistent support for the third quarter phenomenon (Steel,
The results described should be considered in light of the limitations of this research. In our attempt to quantitatively summarize team dynamics in LDSEM-analog environments, we were limited by the empirical research available within the extant literature (e.g., small sample size, correlational). The validity coefficients from the LDSEM-analog studies used in our analyses are based on small sample sizes. When weighted average analyses are based on smaller sample sizes, there is more uncertainty regarding how well an observed effect in a given sample reflects the true population validity. To help address this issue, based on the available data, we calculated improved estimates of the true population team predictor and team criterion relationships in an LDSEM-analog environment by inversely weighting the variances of the validity coefficients from the LDSEM-analog studies and the meta-analytic estimates of the same team predictor-criterion relationships from the extant literature. Additionally, we calculated the average inaccuracy of the estimates to generate 95% credible intervals regarding the uncertainty of the estimates. This approach afforded us the precision associated with meta-analytic estimates while accounting for the localness associated with a specific effect size from an LDSEM-analog environment.
Moreover, the studies included in our quantitative review were almost exclusively descriptive or correlational in design (see the work by Emurian and colleagues for a notable exception). With this is mind, we cannot make causal statements about the relationships examined in our review, nor can we disentangle the effects of one team predictor from another. Consequently, we encourage researchers to employ experimental and quasi-experimental designs to identify key threats to team dynamics and performance in LDSEM-analog settings. We acknowledge the limitations of this data (e.g., small sample size, correlational). Importantly, however, this is the data that we currently have for understanding team dynamics in LDSEM-analog environments.
Despite the limitations of this study, our findings provide insight into several potentially fruitful areas for research in regards to content, and research approaches related to extreme teams. In general, it seems that research should be prioritized when the nature of the relationship would be most likely to change as a function of the LDSEM context. One area in need of research is team affect. While most of the team mood data presented in this article were generated from aggregated individual-level data (for a notable exception see Šolcová et al.,
A second area in need of research is conflict management. LDSEMs provide a unique context in which conflict will need to be managed. Given the significant communication delays with those on Earth as teams travel into deep space, the teams will likely need to effectively manage conflict with at least some degree of autonomy. Our data suggest that at some point conflict is likely to occur between the crew, or between the crew and mission control. Indeed, LDSEMs are likely to be a situation where the crew will face competing or inconsistent priorities. For example, if more than one mission control is utilized for a particular mission, competing information may be given in regards to priorities (e.g., perform a function that requires the whole crew; require an individual adheres to a particular exercise schedule), which could create ambiguity in how crewmembers should allocate their time and resources. Crewmembers are likely to be diverse in a number of ways (e.g., professional, national background) which could also lead to misunderstandings or competing priorities (e.g., maintenance of the space vehicle, complete the science experiment) and potentially cause intra-team conflict (Bell et al.,
In addition to their effects on team performance, conflict management and affect are important areas for future research because they will likely play a critical role in a team's resilience. While researchers are working diligently to mitigate all potential threats to team effectiveness, LDSEM crews will inevitably face challenges. A key aspect of correctly composing, training, and providing countermeasure support to crews will include consideration of the crew's resilience, defined as the capability to withstand and recover from stressors, pressure, or challenges (Alliger et al.,
We believe the decline in team efficiency during the HUBES simulation and the dip in team efficiency for one of the teams during the SFINCSS simulation provide interesting directions for future research in LDSEM-analog settings. Several researchers (e.g., Sandal,
Finally, a number of methodological recommendations can be made for future research. First, sample sizes in high fidelity environments to LDSEM, particularly ICE, are likely to be small. Where possible, data should be collected in such a way that they can be aggregated and compared across multiple studies. Ideally, enough data should be collected to generate an effect size. The normality of the data could be reported (or even better, the raw data) to allow future summaries to ensure the data are being appropriately modeled. When the sample size is too small to allow an effect size to be generated, data on key team constructs (e.g., team efficiency, communication, mood, and cohesion) should be collected with a common set of measures. Analog research on mood has consistently relied on the PANAS and POMS which made comparisons across studies possible. Researchers at the Russian Academy of Science's Institute of Biomedical Problems and some individual researchers (e.g., Leon) have consistently collected data using the same measures, which allowed us to report many of the figures in this article. In addition, NASA's Human Research Program is adopting a standardized set of measures to be collected across NASA analogs that includes measures such as team conflict, team cohesion, and team mood as well as other constructs. For key constructs (e.g., conflict, mood, cohesion), it is essential that analog research use the same measures so that the data better lend itself to the eventual culmination of studies.
Second, continued research is needed on small sample sizes. As an example, some meta-analytic approaches (e.g., Bayesian, Fisherian) calculate sample variance as 1/
Future space exploration teams will be required to work effectively under complex and dangerous conditions to successfully accomplish their missions. With an understanding of team dynamics in LDSEM-analog environments, we can minimize potential threats to mission success while optimizing team performance. While an extensive research base exists that examines teams in LDSEM-analog environments, small sample sizes make traditional forms of meta-analysis inappropriate. Importantly, however, this is the data that we have for understanding team dynamics for future LDSEMs. Given this, we used a weighted averages approach to generate minimum variance estimates of team predictor and outcome relationships, and generated descriptive figures depicting team dynamics over time. Our systematic review of quantitative research on teams in LDSEM-analog settings summarizes what we know about team dynamics for future LDSEM, and provides guidance for future research.
STB developed the analytic approach, designed the search process and coding forms, conducted data analyses, and wrote aspects of the manuscript. SGB and TM conducted extensive literature searches, coded articles included in the review, generated tables and figures, and wrote aspects of the manuscript.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
This work was supported by contract NNJ15HK18P awarded to the first author from the National Aeronautics and Space Administration (NASA). There are some similarities between this article and our longer technical report submitted to the funding agency, NASA/TM-2016-219280. All opinions expressed herein are strictly those of the authors and not necessarily those of the sponsoring organization.
We thank Alan H. Feiveson for providing insight into potential risks of bias with our weighted averages approach. We thank Alla Vinokhodova and Gloria Leon for helping us obtain some limited circulation publications with valuable data, and Kim Binsted and Pete Roma for sharing unpublished data.
The Supplementary Material for this article can be found online at:
*References marked with one asterisk indicate effect size data set studies.
**References marked with two asterisks indicate benchmark data set studies.