by Cardozo ( Cardozo and Leopold 1963). For letters, the best length was four, and for digits, the best length was five.
Domain Length. Klemmer ( Klemmer 1969) showed that groups of three and four were the fastest to recall and groups of two and five were the second fastest. Howe ( Howe 1970) showed that errors were not distributed equally within a series of letters. Among the results, errors occur more frequently at the fifth or sixth position.
Symbols. Martin et al. ( 1977) examined the performance of copying digit strings. They found that there was no difference in performance between using space, full-stop, or hyphens as separators in a digit string.
Mnemonics. Organizational mnemonics are knowledge structures in memory ( Bellezza, 1987), if a knowledge structure is well organized, its activation in memory will provide mental cues to associating new information. Moses & Ehrenreich ( 1981) summarized that truncated abbreviations, in which the first letters of the word are used, are processed better than contracted abbreviations, in which letters within the word are deleted.
The current seven generic Top-level domains (TLD) are easy to recall, because there are only seven domains. Expanding the space of the Internet by expanding domains and increasing the number of generic TLDs in the current system is likely to result in more confusions and failures of recall than now. The storage function of computers can help users storing and retrieving electronic mail addresses. But computers are not so helpful for information association task. E- mail addresses can provide useful information such as organizational and geographical information for users.
The purpose of the user survey of e-mail addresses is to obtain information on user's likes, dislikes, preferences, usage and difficulties associated with the usage of e-mail address systems. The questionnaire was electronically mailed on June 6, 1997. Responses dated July 7, 1997 or earlier, were counted. The survey was mailed to two Human-Computer Interaction (HCI) mailing lists and thirty news groups in which readers discuss computers, science, and psychology issues through the Internet. The number of individuals in the mailing lists and discussion groups varies from day to day. Best estimate is that the total number for all the groups varies from 500 to 1,500.
The survey had 17 questions, which were divided into three groups. The first two questions asked the experiences of respondents using e-mail. The following