Using an ex post facto, correlational descriptive design, a purposive sample of 74 elderly individuals who were identified as either having a ‘good relationship’ or an ‘abusive/neglectful relationship’ with their related caregivers were interviewed in their own homes. The purpose of the study was to test two hypotheses which predicted the nature of differences between the two groups and four which predicted the magnitude and direction of theoretically deduced causal relationships. The data showed that demographic characteristics and certain theoretical variables such as anger, hostility and stress were remarkably similar between groups. The significant differences (P<0.05) between the two groups included lower expectations among the abused subjects for their caregivers, lower perceptions among the abused subjects of their caregiver's actual behaviour, differences between the social networks of the two groups and more depression in the ‘abuse’ group. Using multiple regression analysis, a total of four causal models’ were tested using anger, depression, anxiety, and abuse as terminal dependent variables. The amount of explained variance ranged from 22% for anxiety to 58% for abuse. An empirical model was generated from data available in which 2 variables, perceptions of the caregiver and family members in the house available to help, explained 54% of the variance in abuse.