Evaluating Visual Information Provided by Audio Description E. Peli, E.M. Fine, A.T. Labianca Abstract: Short segments of two TV programs without audio description (AD) were presented to 25 subjects with low vision and 24 subjects with normal vision, and 29 additional subjects heard only the standard audio portions. The subjects then answered questions based on the visual information contained in the AD of the programs. The subjects with normal vision performed the best. followed by those with low vision and those who heard only the audio portion; all performed at better than chance levels. The results indicate that although AD may provide information on visual details to visually impaired audiences, some of the information in the AD may be obtained from the standard audio portion. Television (TV) and other video outlets (Cronin & King, 1990). Descriptions of de­ are an important means of obtaining tails concerning such aspects of a scene as information and sharing in the culture of a clothing styles, body language, colors, and society. For people with visual impair­ landscapes are inserted during pauses in the ments, the visual aspects of these media dialogue or narration of normal programs. are impoverished or absent. Nevertheless, The AD is typically broadcast via the sepa­ most visually impaired people watch TV rate audio program (SAP) channel that is a with their families and use other video standard feature of most stereo TVs and vid­ sources, such as videotaped movies and eocassette recorders (VCRs). ADs of pro­ computer video displays (Berkowitz, Hiatt, grams broadcast on the SAP are available deToledo, Shapiro, Lurie, 1979; Joseph­ free of charge and do not interfere with.other son, 1968). In fact, visually impaired viewers' enjoyment of the programs. AD is people watch as much or more TV as do also available with VCRs, for which narra­ sighted people (Cronin & King, 1990). tion is simply added to the videotapes, which Audio description (AD) provides descrip­ are played on standard equipment. Although tions of the visual elements of TV programs the reviews by visually impaired audiences without interfering with the programs' audio have been positive, AD was not formally evaluated with visually impaired people un­ This research was supported by NIH Grant EY05957 to E. Peli. The authors wish to thank Julie Irish for til the study presented here. her help in developing the questions used in this Rabbitt and Carmichael (1993) exam­ study; Barry Cronin and Gerry Field, of WGBH. ined the comprehension by older adults Boston. for providing the program videos, software. with normal vision of short segments of and valuable technical help: and Laurie A. Everett. of WGBH's Descriptive Video Service. for her programs. as well as of an entire 30-minute comments on an earlier draft of the article. program. presented to half the subjects A378 )ounlal of Visual Impairmenr & Blil1dness. Sept-Oct lQ<lc- ([1190(:> .-\FB. A" Rights Reserved with AUDETEL, a British AD service, and volunteers with low vision and those with to half without AUDETEL. They tested the normal sight to answer questions devel­ subjects' comprehension of each segment oped on the basis of visual information by asking them to identify characters and provided by Descriptive Video Service locations that appeared in the segment and (DVS), an AD service developed by to elaborate on their descriptions. They WGBH, the public television station in also measured the subjects' overall com­ Boston. Two different programs broadcast prehension of the program using 20 with DVS (one from the Mystery! series true-false questions. and one from the Nature series) were Rabbitt and Carmichael found that chosen because the important visual ele­ comprehension of the segments (but not ments they contained were different and overall comprehension) was greater for the spanned the type of programming gener­ group that was exposed to the AD and ally aired with AD. Observers were surmised that AD helped cue the subjects presented only with the video and standard to features of the program that were the audio portions of the films..They then focus of the open-ended questions. The answered multi-choice questions on the primary impetus for their study was to content of the audio descriptions without determine whether the A UDETEL descrip­ actually hearing them. It was hypothesized tions were disruptive to elderly viewers. that if the descriptions contain visual On the basis of their data, they concluded information available to normally sighted that AUDETEL descriptions did not viewers but not to those with low vision, detract from the enjoyment or interfere then observers with normal vision would with the comprehension of the program answer more of the questions correctly. by elderly viewers with normal vision. To verify that the questions could not be The current study was designed to answered on the basis of the standard audio evaluate only whether and to what extent of the program or common knowledge, the the visual information supplemented by authors asked an additional group of volun­ AD is npt available to persons with visual teers the same questions after hearing only impairments. (AD occasionally includes the standard audio portion of the programs; information that is not strictly visual, such these volunteers neither saw the video nor as period and temporal references, rela­ heard the AD. Although this group was in­ tionships, and poetic expressions, but these cluded primarily to verify that visual infor­ types of information were not included.) mation was needed to answer the questions, Although it seems evident that any such the data they generated can also be used to description would be useful for blind estimate the minimum level of performance audiences, the authors were interested in that could be expected from blind audiences the level of transfer of information from engaged in the same task. the visual content of the programs only to audiences with low vision. Methods To evaluate the potential for transfer of SllBJECTS visual information from AD to visually Adults aged 55 or older who were impaired audiences, the authors asked recruited to participate were separated into 1<l<l6 AFB. All Rights Reserved * as r""""",,,,*-- kl't I ,,,'WiMi6 'yt'n' #1'.'1 ')T' spm ? 1 two groups on the basis of the acuity in measured. It should be noted that the AO their better eye. Normal vision (NV) was condition is not a simulation of a truly defined as acuity of 20/40 or better in the blind audience, which may have adapted to better eye and low vision (LV) as acuity long-term vision loss by gathering infor­ between 20/1 00 and 20/800 in their better mation through other sensory inputs (pri­ eye. Adults who did not meet the criterion marily auditory). Thus, it is likely that the for either group were excluded. AO group would have been outperformed A total of 24 NV subjects (9 who viewed by a group of blind observers who are only Nature; 9, only Mystery!; and 6, both) presumably more skillful at gaining infor­ and 25 LV subjects (10 who viewed only mation from nonvisual sources. Nature; 10, only Mystery!; and 5, both) participated. There was no difference in ApPARATUS mean age between the vision groups (NV An Apple Macintosh SE was used to run = 71, SEM 1.63; LV = 75, SEM a HyperCard® stack, deScriptor version 2.10; t(45) -1.51, P = .14). The 1.2.1 (DVS, Boston). Modifications to this median acuities for the NV and LV groups stack allowed the authors both to control were 20/30 and 20/200, respectively. the presentation of the video and to record Twenty-nine additional subjects (2 who subjects' responses. The computer "was heard only Nature; 7, only Mystery!; and 20, interfaced with a Sony U-matic/SP VCR both) were recruited to participate in the au­ via two V ideo media Serial I/O units that dio-only (AO) condition. The age of this sam­ synchronized the time code between the ple [mean age = 51, SEM 4.15] was not computer and the VCR. restricted, and the authors continued to col­ DVS deScriptor stacks were originally de­ leet data from the AO group beyond the num­ veloped to add AD to TV programs. Using ber of subjects in the NV and LV groups for this system, a describer can control the VCR another study. Since both the audio portions to navigate precisely through and display par­ and the ADs of programs rely on hearing, ticular scenes. In addition, a description can and many visually impaired people are el­ be added by advancing to the appropriate scene derly and may have some hearing loss, the and adding the narration to a script field des­ authors explored whether age might affect ignated for that scene. The authors added ques­ performance on this task. There was no sta­ tion-and-response fields in lieu of the narra­ tistical difference in performance between tion script field that DVS created. They also the 55-or-older group and the younger group modified the software to export response datil [t(47) < 1.0. n.s.], and so decided to main­ automatically from the response fields to tain the larger sample size for the analyses. spreadsheet format. These modifications re­ The AO group neither saw the video nor duced data-entry errors, as well as any risk heard the AD, so their responses were that the experimenter would ask questions or based only on the standard audio of the record responses that did not correspond to program. Since the purpose of this condi­ the correct scene or question. tion was to verify that questions could not The video segments were displayed on a be answered on the basis of the standard 27-inch Sony Trinitron color TV monitor.