To evaluate the temporal stability of maternal attachment representations obtained using a word-prompt task, a sample of mothers (N = 55) was assessed on two occasions, 12 - 15 months apart. Each mother responded to six word-prompt sets on each assessment occasion (4 word-prompt sets were designed to prime secure base themes, 2 word-prompt sets were designed to prime different themes), and the resulting stories were scored in terms of the presence and quality of the secure base scripts evident in each story. The story scriptedness scores (average across four stories) were internally consistent at each assessment (alphas >.85) and the mean difference in scores was not significant across assessments. The cross-time correlation for the composites (aggregates of scores at each age) was positive and significant, r53 = .54. Other aspects of maternal stories were also stable (e.g., number of words used, number of sentences per story, use of words from the prompt list). Controlling for stable stylistic features of the stories did not reduce the magnitude of association for scriptedness scores across time. These results suggest that the presence and quality of secure base scripts is a stable aspect of maternal representations of attachment and that the word-prompt task is useful for prompting the script in narrative production.