Background: Verbal intrusion errors are irrelevant responses made in the course of verbal memory retrieval or language production that have been associated with disruption of executive functions and the prefrontal cortex. They have been observed to occur more frequently both with normal aging and with neurodegenerative diseases such as Alzheimer's disease (AD). The purpose of this study was to longitudinally examine the production of verbal intrusions among middle-aged adults with Down syndrome (DS) and unspecified intellectual disability (ID) to determine whether producing verbal intrusions at one point in time was related to subsequent verbal memory performance. Because of the combination of a relative deficit in verbal working memory (WM), premature aging, and higher risk of AD among adults with DS, it was predicted that they would make more verbal intrusions than adults with unspecified ID. Methods: Word List recall (WLR), the Selective Reminding Test (SRT), and the Cued Recall Test (CRT), were administered three times at 18-month intervals during a 3-year period. In Analysis 1, aetiology differences in making intrusion errors were examined. Twenty-three adults with unspecified ID in the moderate to mild range [time 1(T1) mean age = 47.2 years] and 42 adults with DS (T1 mean age = 44.3) participated. WLR is a serial WM task beginning at two word sequences and progressively increasing by one word every three trials. WLR intrusions were analysed because they were least likely to include "educated guesses" because this test is not based on semantic categories. In Analysis 2, we only examined participants with DS. They were divided into two groups, 16 individuals who made at least one intrusion error at T1 (T1 mean age = 45.8) and 26 who did not (T1 mean age = 43.3). Longitudinal performance for these groups was analysed to determine whether the group that intruded at T1 did more poorly on subsequent memory tests. Results: A higher proportion of responses comprised intrusions for the group with DS and a higher percentage of the participants with DS made at least one intrusion error when compared with participants with unspecified ID (74% and 44% respectively). Those participants with DS who made at least one intrusion error at T1 showed a subsequent decline in performance on both WLR and the SRT. Conclusions: The production of intrusion errors during a verbal WM task is a characteristic of middle-aged adults with DS. This suggests compromised executive function and control of inhibition within the verbal modality for this group. Further, verbal intrusions are a qualitative aspect of verbal processing that merit attention in considering the issue of deficiencies of language and verbal WM abilities among people with DS. Last, and perhaps most importantly, although not definitive diagnostically, an increase in verbal intrusions is a potentially noteworthy signal when evaluating the cognitive health of adults with DS.