Skip to main content

Exploring Subtitle Reading Process with Eyetracking Technology

Periodic Reporting for period 1 - SURE (Exploring Subtitle Reading Process with Eyetracking Technology)

Reporting period: 2016-07-15 to 2018-07-14

In the modern world, we are surrounded by screens, captions, and moving images more than ever before. Subtitling is ubiquitous in many different forms and formats like TV broadcasting, DVD, cinema, and streaming services. Subtitling is a perfect tool to promote EU policy on language learning, linguistic diversity and multilingualism as well as a key element of making audiovisual culture accessible to people who are deaf or hard of hearing.

Understanding the way people watch subtitled films has become a central concern for subtitling researchers in recent years. As more and more people watch subtitles on different devices and in different environments, there is a pressing need to verify the current subtitling standards, some of which were developed many years ago and were not based on empirical research.
In the SURE Project, we studied how people read subtitles depending on their speed (how fast subtitles appear and disappear) and text segmentation (the way the text is divided between the two lines in a subtitle). Appropriate subtitle speed and segmentation allow viewers to follow the text in the subtitles comfortably and to have enough time to look at the on-screen action. If subtitle speed is too fast or too slow, and segmentation does not adhere to linguistic rules, viewers may find it difficult to follow and understand the information contained in the subtitles. Today’s changing audiovisual landscape calls for more up-to-date research on how fast different groups of users can read subtitles and on whether text segmentation in subtitles has a direct impact on the reading process.

A combination of different research methods used in the SURE Project: eye tracking, a set of questionnaires and semi-structured interviews, has enabled us to isolate the impact of subtitle speed and segmentation on the processing of subtitled videos modulated by different linguistic backgrounds of viewers. No work to date has investigated subtitle speeds and segmentation using such mixed methods approach. Our approach provided a unique research opportunity to determine that modern viewers are able to keep up with fast subtitles and that they prefer uncondensed subtitles when watching films in a language they are familiar with. We also found that people’s viewing experience depends on the language of the soundtrack and their familiarity with subtitling. Finally, we established that linguistic-based text segmentation in subtitles results in more efficient subtitle processing.
WORK PERFORMED
We tested 97 people. Our participants were either hearing native speakers of English (30 people), Polish (21 people) and Spanish (26 people), or English speakers that were hard of hearing (10) or deaf (10).

Three eye tracking experiments were conducted:
Experiment 1: Subtitle speed in films in a language unknown by viewers
Viewers watched videos in an unfamiliar language (Hungarian) with slow (12 cps), medium (16 cps) and fast (20 cps) subtitles.
Experiment 2: Subtitle speed in English-language films
Viewers watched English-language videos with slow and condensed subtitles (12 cps) and fast and uncondensed (20 cps) subtitles.
Experiment 3: Subtitle segmentation
Part 1: Viewers watched subtitled videos with syntactically-segmented and non-syntactically segmented subtitles.
Part 2: Viewers chose their preferred line breaks in subtitles displayed using screenshots.
Part 3: Viewers watched videos with two- and three-line subtitles and chose their preferred option.

OVERVIEW OF THE RESULTS
The most important findings on subtitle speed:
- When watching films in English, young, well-educated viewers proficient in English are able to follow subtitles displayed at a speed of up to 20 characters per second, which is higher than a typical subtitling speed. They prefer text to be less condensed in subtitles in films in English.
- When watching foreign films in an unknown language, most people prefer dialogue to be condensed and subtitles displayed at a slower speed.
- Subtitles displayed at slow speed, recommended by traditional subtitling standards, result in more re-reading.
- When watching English-language films with slow subtitles, viewers tend to notice discrepancies between the dialogue and the subtitles, resulting from text reduction.

The most important findings on subtitle segmentation:
- Linguistic units should be kept together and should not be split across the lines.
- Non-syntactically segmented text in the subtitles may increase cognitive load and induce more re-reading.
- It is more important to keep closely related linguistic units together than to create a certain subtitle shape, like a pyramid or rectangle.
- Subtitles should not have more than two lines.

The results of the project show the need to revise the current subtitling guidelines in terms of subtitle speeds and confirm the rules on text segmentation in subtitling. When viewers are familiar with the language of the film soundtrack, they can follow faster subtitles with less text reduction, but when the language of the soundtrack is unknown, it is better for subtitles to be slower and more condensed. Text is subtitles should follow linguistic units.
The SURE Project addressed some of the key issues in modern audiovisual translation research: revisiting the old truths, replicating previous studies, improving research methodology, introducing transparency in research, and working together with stakeholders, such as viewers and professionals. Some of the ‘old truths’ in subtitling may no longer be relevant to 21st century subtitling. One such truth is the six-second rule, which seems to be outdated given that many modern viewers are able to read fast subtitles and in fact prefer such subtitles when watching films in English. Some of the results of the SURE Project question the core of traditionally conceived subtitling and show the necessity for modern subtitling to re-establish itself in the context of new technological developments and changing viewer preferences.

The SURE Project was also focussed on improving subtitling research methodology and on making it more transparent. All data from the project, including raw files, have been made available in an open access repository:
Szarkowska, A., & Gerber-Morón, O. (2018). SURE Project Dataset. RepOD. http://dx.doi.org/10.18150/repod.4469278

The results of the project are directly applicable to current subtitling practices. They can become a springboard for discussions on policy changes and revising subtitling guidelines. Project results were disseminated in four high-impact journals and communicated to the general public, professional subtitlers and regulatory bodies: Ofcom in the UK, the National Broadcasting Council (KRRiTV) in Poland and CESyA in Spain.

The project also raised awareness of the importance of good-quality subtitling. Higher education institutions and subtitler trainers may also benefit from project findings, particularly through open access publications and communication actions addressed at the general public and professional subtitlers, such as a video tutorial on subtitle speed:
Szarkowska, A. (2018) Subtitle speed. Video tutorial. https://avt.ils.uw.edu.pl/files/2018/07/ReadingSpeed.mp4
On a wider societal scale, the project may impact not only the field of subtitling and audiovisual translation, but also media accessibility, second language acquisition and teaching, disability studies and reception studies.
SURE Project Logo