Memory for

The ability of an eyewitness to recall a briefly-viewed license plate is an under-studied issue with great practical significance. Using large samples of adult participants (total N = 1859), we assessed the memory implications of a current proposal (EZ-ID, n.d.) to include a nameable symbol, such as a triangle or heart, on standard-issue license plates. In two experiments, no overall memory benefit was observed for license plates with symbols compared to standard designs. Although symbols conveyed no memory advantage, we show that, under certain conditions, including symbols can reduce the number of matching plates when a plate is partially remembered, enhancing the ability of investigators to identify the specific vehicle involved in a crime. We also assessed recall performance for license plates viewed under a range of conditions. Exposure time and the time delay to report have both real-world relevance and clearly predicted effects grounded in the basic memory literature. Memory increased with exposure time, but there was no effect of study-test delay. We also manipulated viewing orientation, to allow comparison of memory given a standard view with the mirror-reversed view that a witness might experience while driving. Memory was significantly worse for mirrorreversed plates. Memory for License Plates 3 Memory for License Plates United States Commonwealth of Massachusetts Senate Bill 2387 (Bill S.2387, 2012) is a call to redesign motor vehicle license plates. In Massachusetts, standard issue license plate numbers currently consist solely of 6 characters (i.e., digits and letters). The bill includes the proposal that general issue license plates could include a “commonly recognizable symbol”, such as a triangle or circle. As will be discussed below, including symbols can reduce the number of matching plates when a plate is partially remembered, enhancing the ability of investigators to identify the specific vehicle involved in a crime. Symbols may also be remembered differently than other characters. As described by the organization sponsoring the bill, a large part of the motivation behind including symbols is that “random numbers and letters on license plates are too difficult to remember in a time of crisis” (EZ-ID, n.d.). The assumption is that the inclusion of symbols would aid in license plate recognition and recall. Although the intuition behind this premise is reasonable, we could find no published studies that test this assumption. The main goal of this paper is to test the idea that the addition of a symbol would improve memory for license plates. The vast eyewitness memory literature includes hundreds of studies on witnesses' ability to describe or recognize the perpetrator of a crime (a PsychLit search with "eyewitness identification" revealed 428 peer-reviewed journal articles; 122 were returned for "eyewitness AND description"). In contrast, we found only five studies on memory for another forensically-relevant detail of a crime, namely license plate numbers, in the psychological literature. Given that a partially remembered plate is much more likely to be useful than an impoverished description of a suspect, surprisingly little research has been published in the psychological literature on the memorability of license plates. The 1 To our knowledge, no U.S. state currently includes a symbol in standard issue license plates. Vanity plates in some states do allow limited use of some symbols (Knuth, 2011). Memory for License Plates 4 studies that have been done, while valuable, are of limited scope. Two studies presented subjects with a series of slides depicting an actor who appeared to load stolen goods into a car and then drive away (Mende, MacKinnon, & Geiselman, 1987; MacKinnon, O'Reilly, & Geiselman, 1990). These studies (see also Emmett, Clifford, Young, Kandova, & Potton, 2006) primarily focused on interview techniques that might enhance recall. Consequently, they did not manipulate many encoding factors. Mende et al. did vary the presentation rate for the slides, finding that longer exposure time (5 sec/slide) yielded higher recall than shorter exposures (2 sec/slide; Mende et al., 1987), and about 3-3.5 characters were typically reported correctly (MacKinnon et al., 1990). A single license plate was used in each of these experiments (either 640VYE or 1MJT407), leaving open the possibility that memory could be better or worse for other plates. Indeed, Baerwald, Karmeier, and Herrington (1960) reported that recall exceeded 4 characters for plates comprised only of digits, and was typically about 3 characters for plates with mixed letters and digits like 640VYE. One reason for recall differences across plates appears to be that grouping letters and numbers (as in 640VYE, but not 6V4Y0E) allows the information to be more easily chunked; fewer alternations between letters and numbers yields enhanced recall (Schraagen & van Dongen, 2005). Although Al-Haboubi (1999) also manipulated study time, that study, like Schraagen and van Dongen's (2005), focused on the design of the license number. These studies both asked which particular grouping of letters and numbers yield better recall, finding that fewer alternations between letters and numbers was ideal. Previous memory research suggests that adding a symbol to license plates might help memory, if we assume that symbols are encoded and represented differently than characters (Paivio, 1963). This differential representation could result in the cognitive compartmentalization of symbols and characters resulting in either less cross-class interference or improved encoding strategies (e.g., "chunking"). Item pairs from different stimulus classes did not interfere with one another in an associative recognition task (Criss & Shiffrin, 2004). Participants were shown a series of item pairs. Each pair Memory for License Plates 5 consisted either of two words (WW), two faces (FF), or one word and one face (WF). During study, the participants were asked to judge the degree of association between items in each pair. Immediately following study, participants were given a surprise memory test. The test pairs consisted of intact (i.e., identical to a study pair) or rearranged (i.e., items from two different study pairs) pairs, and participants were asked to determine if each of the test pairs was seen during study. The number of study pairs was kept constant across participants, but the proportion of pairs of each class varied. For example, in one condition, there were 40 WF, 40 WW, and 40 FF study pairs, whereas another condition had 80 WF, 20 WW, and 20 FF study pairs. The key result was that, as the proportion of a particular pair class decreased, performance for that class improved. For example, even though the total number of pairs was kept constant, associative recognition for WF pairs was better when there were 40 WF pairs than when there were 80 WF pairs. Criss and Shiffrin (2004) concluded that different classes of pairs do not interfere with each other during retrieval. Gillund and Shiffrin (1981) performed a similar study using a free recall task for single items. Participants were presented with study lists consisting of a series of individual words and pictures. The absolute and relative number of each item class on the study list was manipulated. For example, a base study list consisted of 10 words or 10 pictures. A longer study list could then be created by adding 10 words or 10 pictures to the base list, creating lists with 20 words, 20 pictures, or 10 pictures and 10 words. Note that, for the longer lists, list length is constant, but the proportion of words and pictures varies. Consistent with research on the list-length effect (e.g., Shiffrin, 1973), recall decreased as list length increased. Furthermore, performance also decreased whether items of the same class or items of the other class was added to the base list. It is important to note, however, that in the majority of conditions, adding items of a different class decreased performance less than adding items of the same class. For example, memory for pictures was better in the 10-picture, 10-word condition than in the 2 The authors did not perform statistical tests to assess this claim, which is based on the results in Figure 3 of Gillund & Shiffrin (1981). Memory for License Plates 6 20-picture, 0-word condition, again suggesting that mixing classes of items may improve performance. There are two intuitive information categories on current license plates: letters and digits. Recall that, as the proportion of a class decreased, performance for that class improved. Because a symbol would take the place of a letter or digit, reducing the number of elements of that class, the addition of symbols could improve memory. For example, consider the plate M6C83, which has 3 digits and 2 letters. Replacing the 6 with a symbol, e.g., , produces the plate MC83. The number of digits has been reduced, potentially increasing memory for the digits in the plate. Chunking, or the grouping of items into meaningful units, is a well known way to improve memory (e.g., Bower & Springston, 1970). Like the faces, pictures, and words used in Shiffrin's studies (Gillund & Shiffrin, 1981; Criss & Shiffrin, 2004), digits and letters form natural classes of information. Schraagen and van Dongen (2005) demonstrated that grouping those item classes together can benefit memory for license plates. Following previous research (e.g., Broadbent & Gregory, 1964; Sanders & Schroots, 1968; Hull, 1976), they hypothesized that memory for a license plate would decrease as the number of alternations between letters and digits in a chunk increased. They presented participants with license plates with 0 (e.g., ABC-123), 1 (e.g., A12-BCD), 2 (e.g., A1-BC2-3), 3 (e.g., A1-B2-C3), and 4 (e.g., A1B-2C3) alternations between letters and digits within the chunks, defined by the spatially separated groups. The idea is that the spatial separation creates groups of items and that these groups would be more meaningful, and therefore easier to remember, if the items within a group were all of the same class (i.e., all letters or all numbers). Participants studied each plate for approximately 500ms and then recalled the plate after a 6 second delay. The central result was that memory performance was higher for license plates with fewer alternations, suggesting that grouping items of the same class aids memory. Note that the alternations are defined within the spaced character groups, e.g., A12 and BCD in A12-BCD. Symbols may form an efficient way to group license plate characters, akin to a space, but Memory for License Plates 7 containing meaningful, identifying information. For example, the triangle, , in the plate AB123 may create a natural grouping, e.g., AB--123. Thus, the inclusion of a symbol could help observers chunk the plate characters while at the same time serving as one of the to-be-remembered items. Our primary goal is to assess whether the inclusion of a nameable symbol enhances memory for a license plate. The second goal of this research is to explore the effect of three "real world" factors that may influence how well a license plate is remembered outside the lab: exposure time, test delay, and viewing orientation. Almost nothing is known about how these factors influence memory for license plates, although we can make some basic predictions based on the memory literature. Consider a witness to a hit-and-run incident. During the event, the witness may get anywhere from a glimpse of the plate to many seconds of viewing time. This exposure time seems likely to influence how well a plate is remembered. As in standard memory experiments that manipulate stimulus duration (e.g., Hintzman, 1970), longer viewing time provides the participant with more time to encode the license plate, which yields better recall (Mende et al., 1987; Al-Haboubi, 1999). The relative discriminability and identifiability of symbols and characters has not been explored, so it is unclear whether symbols can be encoded more quickly than characters. If they are encoded relatively quickly, then higher recall accuracy for symbols rather than characters should occur at shorter viewing times. The second factor is the recall delay, i.e., the length of time between when the plate was viewed and the participant is asked to recall the plate. Short-term memory typically declines quickly (e.g., Brown, 1958), so we expect that a longer delay will result in poorer recall. Again, it is unknown how the recall of symbols will fare relative to characters as the delay is increased. To enhance realism in our experiments, the recall delay period will be filled with a video that simulates driving down a road with the radio playing. The third factor we considered is viewing orientation, i.e., whether the plate is viewed normally or reversed as if seen in a car's rear-view mirror. Most license plates are almost surely viewed normally, but it is Memory for License Plates 8 possible, particularly for cars behind a driving witness, to be viewed in a mirror. Because the recognition of many non-character symbols has the advantage of being less affected by variation in orientation, the recall of symbols may be better preserved when viewed as in a mirror.