Hopelessness and Interpersonal Conflict: Antecedents and Consequences of Losing Hope

Abstract

Research on intractable conflict indicates that difficult conflicts often are associated with a sense of hopelessness about the potential for resolution. We argue that hopelessness arises from individuals who want another to change and whose arguments evidence self demand/other withdraw and mutual hostility. The sense of hopelessness arising from these patterns is related to avoidance after an argument and withdrawing support and affection. We conducted a survey of undergraduates about their arguments with parents and romantic partners. As predicted, the correlations between desire for change and withdrawal of support/affection and desire for change and avoidance were positive and statistically significant. A mediation analysis confirmed our hypotheses. The relationships between desire for change and withdrawal of support/affection and avoidance were mediated by self demand/other withdraw and hopelessness. Similarly, the relationships between desire for change and withdrawal of support/affection and avoidance were mediated by mutual hostility and hopelessness. We discuss implications for research on serial arguing, intractable conflict, and destructive communication behaviors.

Keywords:

• Demand/Withdrawal
• Hopelessness
• Mutual Hostility
• Relational Disengagement

Notes

Note. N = 301.

^a Reliability coefficients for multi-item scales reported along diagonal.

^b 0 = romantic partners; 1 = parents.

^c 0 = male; 1 = female.

*p < .05. **p < .01.

Note. N = 301. Number of bootstrap samples for bias corrected bootstrap confidence intervals = 5000.

^a  = Unstandardized coefficients.

^b 0 = romantic partners; 1 = parents.

^c 0 = male; 1 = female.

^d  = Bias corrected 95% bootstrapped confidence intervals.

*p < .05. **p < .01. ***p < .001.

Note. N = 301. Number of bootstrap samples for bias corrected bootstrap confidence intervals = 5000.

^a  = Unstandardized coefficients.

^b 0 = romantic partners; 1 = parents.

^c 0 = male; 1 = female.

^d  = Bias corrected 95% bootstrapped confidence intervals.

*p < .05. **p < .01. ***p < .001.

Note. N = 301. Number of bootstrap samples for bias corrected bootstrap confidence intervals = 5000.

^a  = Unstandardized coefficients.

^b 0 = romantic partners; 1 = parents.

^c 0 = male; 1 = female.

^d  = Bias corrected 95% bootstrapped confidence intervals.

*p < .05. **p < .01. ***p < .001.

Note. N = 301. Number of bootstrap samples for bias corrected bootstrap confidence intervals = 5000.

^a  = Unstandardized coefficients.

^b 0 = romantic partners; 1 = parents.

^c 0 = male; 1 = female.

^d  = Biased corrected 95% bootstrapped confidence intervals.

*p < .05. **p < .01. ***p < .001.

In clinical research hopelessness is linked with theories of depression (e.g., Abramson et al., 1989; Beck, 1967) and is a key factor in predicting suicidal behavior (e.g., Brown et al., 2005). Our focus is on hopelessness regarding conflict resolution. We are using the term in the same way that intractable conflict scholars have used it to characterize intractable conflicts (Coleman, 2000; Pruitt & Olczak, 1995). Gottman (1999) and his colleagues (Driver et al., 2003) also use hopelessness in this manner.

We conducted a confirmatory factor analysis (CFA) to determine whether our scales reflect distinct constructs. Our measurement model has six latent factors including desire for change (2 indicators), hopelessness (3 indicators), self demand/other withdraw (5 indicators) mutual hostility (4 indicators), withdrawal of support/affection (5 indicators) and avoidance (3 indicators). We acknowledge that our sample size is not optimal for conducting a CFA. The model had a total of 87 estimated parameters and our sample size of 301 falls below the minimum recommended ratio of 10 subjects per parameter and ideal recommended ratio of 20 to 1 (Kline, 2011). Although creating parcels of indicators would reduce the number of parameters, it also would decrease the number of indicators for each latent factor below the recommended minimum of three (Hair, Black, Babin, Anderson, & Tatham, 2006). Hence, we conducted the test only to show relationships within this sample. The following standards were used to determine fit: nonsignificant goodness of fit χ², Bentler's (1990) comparative fit index (CFI) no lower than .90 and near .95, and Steiger and Lind's (1980) root mean square error of approximation (RMSEA) near .05 and no greater than .08 (Schumacker & Lomax, 2010). We began with a model in which the error terms were uncorrelated. The six-factor model failed to meet one of the standards, and barely met the other two, χ2 (104, N = 301) = 584.16, p < .001, CFI = .90, RMSEA = .08, RMSEA: low 90% confidence interval = .07, upper 90% confidence interval = .09. We examined the modification indices that indicate how much fit might be improved if error terms were allowed to correlate. Seven sets of error terms had very high modification indices (greater than 10), which indicated that fit would be improved if they were allowed to correlate. Three of the error terms were among items in the self demand/other withdraw scale, one error term was between two items composing the mutual hostility scale and three were among the items in the withdrawal of support/affection scale. Correlated error terms often result from items that are adjacent to each other in the questionnaire or are phrased similarly. When the seven sets of error terms were allowed to correlate, two measures of fit improved, CFI = .95, RMSEA = .06, RMSEA: low 90% confidence interval = .05, upper 90% confidence interval = .06 and a chi square difference test indicated that the goodness of fit significantly improved, χ² difference (83, N = 301) = 116.60, p < .009. However, because the chi square goodness of fit statistic remained significant, χ² (187, N = 301) = 367.56, p < .001, we looked for standardized residuals having values greater than 2.58 which indicates misspecification among items (Byrne, 2010). None were found.

We conducted another analysis to see if including two secondary latent factors would replicate the results of the composite tests. Using Mplus 7.0, we created a model that included two latent factors (desire for change and hopelessness) and two secondary latent factors (lack of influence, composed of self demand/other withdraw and mutual hostility and relational disengagement, composed of withdrawal of support/affection and avoidance). We examined whether the relationship between desire for change and relational disengagement was mediated by lack of influence and hopelessness. The model fit was acceptable on two of three standards, χ² (227, N = 301) = 501.42, p < .001, CFI = .92, RMSEA = .06, RMSEA: low 90% confidence interval = .06, upper 90% confidence interval = .07. A positive relationship was found between desire for change and the relational disengagement factor (withdrawal of support and avoidance) that was mediated by the relationship between the failed influence factor (self demand/other withdraw and mutual hostility) and hopelessness (coefficient = .024, bootstrapped biased corrected 95% lower confidence interval = .009, upper confidence interval = .055) and a shorter positive relationship between desire for change and relational disengagement that was mediated by the failed influence factor (coefficient = .254, bootstrapped biased corrected 95% lower confidence interval = .163, upper confidence interval = .371). However, it should be noted that the model included 94 parameters and our sample was far short of the required 10 to 1 ratio. Hence, these results should be interpreted with caution.

Additional information

Notes on contributors

Courtney Waite Miller

Courtney Waite Miller (PhD, Northwestern University) is an Associate Professor at Elmhurst College.

Michael E. Roloff

Michael E. Roloff (PhD, Michigan State University) is a Professor in the Communication Studies Department at Northwestern University.

Rachel M. Reznik

Rachel M. Reznik (PhD, Northwestern University) is an Associate Professor at Elmhurst College.