Fifteen blind and sighted subjects, who featured as a control group for acceptance, were asked for their requirements for automated teller machines (ATMs). Both groups also tested the usability of a partially operational ATM mock-up. This machine was based on an existing cash dispenser, providing natural speech output, different function menus and different key arrangements. Performance and subjective evaluation data of blind and sighted subjects were collected. All blind subjects were able to operate the ATM successfully. The implemented speech output was the main usability factor for them. The different interface designs did not significantly affect performance and subjective evaluation. Nevertheless, design recommendations can be derived from the requirement assessment. The sighted subjects were rather open for design modifications, especially the implementation of speech output. However, there was also a mismatch of the requirements of the two subject groups, mainly concerning the key arrangement.

An existing cash dispenser was implemented with speech output to give access to blind and visually impaired people. Additionally, the screen graphics and the function access were modified. The hardware was not changed. Blind and visually impaired subjects performed a usability-test, and experts in the field of human-computer-interaction evaluated the dispenser system's usability heuristically. The results showed that the modifications help blind and visually impaired people to access such machines, but adaptations of the hardware are necessary to maintain usability. The two evaluation methods did not produce consistent results.