Many essential quantitative characteristics of human society are gathered from surveys of individual respondents. We measure these important features via survey questions, many of which are about events or activities that occurred in the past. When adopting retrospective questions, survey researchers have to depend on the autobiographical memories of respondents, which are prone to errors of memory and comprehension, among others (Bradburn et al., 1987; Tourangeau, 2000). Nevertheless, retrospective questions are still widely used in national surveys. The American Time Use Survey (ATUS), for example, asks respondents over the phone to enumerate all activities they have done in the last 24 hours (Bureau of Labor Statistics, 2014a). Similarly, the National Household Travel Survey asks respondents to report any trips they have taken that day and time spent on walking in the last week (Santos et al., 2011). However, a number of validation studies also found that people made recall errors in other surveys that rely on such self-reports (e.g., Gray, 1955; Auriat, 1991; Means and Loftus, 1991; Graham et al., 2003; Ayhan and Isiksal, 2004; de Nicola and Gine, 2014).
To facilitate respondents? recall, this proposed research will evaluate a new data collection approach called ?Memory Bookmarking? in which respondents will be ?pinged,? i.e., signaled by text message on a smartphone, to provide a ?description? of their current circumstances in textual (text message), photographic (image) or geographical (map) format. These contemporaneous reports will primarily concern the respondent?s location when pinged. They will then be fed back to the same respondents in an individually customized survey, so that respondents will be able to see the descriptions they previously provided and use them to help recall details of the events they experienced between the time they were pinged and the present. The idea is that contemporaneous reports are analogous to ?bookmarks? in that they serve as an access point into an ongoing narrative, potentially enabling respondents to look up the identity of the intervening events as well as their relevant details, thus improving recall compared to when no bookmark is available. This approach will be tested in an experimental design, which compares the use of bookmarking to no bookmarking and also tests how the number of bookmarks, i.e., the frequency with which respondents are pinged might affect recall. Finally, the experiment will test whether bookmarks that contain more than just location information further improve recall.