Brian Cerney of Hand And Mind Publishing received the contract for the photography of the RID National Conference held in Atlanta Georgia from July 17th until July 22nd, 2011. These photos are available publicly through various Picasa accounts.
Professional Discussion (1.5 hours) – This presentation assumes the participant has a general familiarity with the literature and professional practice within the topic area. The focus is “increased understanding and application by the participant.”
Overview of Presentation
Why do most interpreters switch every twenty minutes? Interpreting team members commonly switch their working roles of providing primary service (the “A” interpreter role) and that of monitoring the interpretation (the “B” interpreter role) every twenty minutes in order to maintain consistency and avoid deterioration of the integrity of the interpretation due to fatigue. Brasel’s 1976 research provided supporting evidence that this approach provides for best practices within teamed interpreting. Thirty-five years later, Brasel’s research has not been satisfactorily replicated until this current study, which investigated a pool of professional interpreters working within a post-secondary educational setting.
This presentation reviews Brasel’s original research, the current practices within our profession, and the results of the study with specific implications for how to modify our current best practices for optimum consumer satisfaction with interpreting services while reducing our own stress and overuse symptoms as professional service providers
Additional Information
Interpreters working in teams generally transition every twenty to twenty-five minutes between direct provision of service (“A” role) and supporting and monitoring the interpretation (“B” role). The only research that supports this pattern is Brasel’s 1976 study, which indicated that interpreter fatigue begins to negatively impact the quality of signed language interpretations after twenty minutes.
This presentation provides a review of a new and on-going study, which seeks to partially replicate Brasel’s original study in order to determine what the current best practices in interpreting should be for switching roles. Subjects were staff ASL/English interpreters at the National Technical Institute for the Deaf, in Rochester, NY who were video recorded during their normal interpreting assignments as part of regular ongoing professional development. Multiple recordings for each interpreter within several settings provided baseline comparisons as an effort to establish consistency of performances and eliminate poor work due to having an “off day”. Recordings were systematically analyzed to determine the first onset for a certain class of errors (repaired slips of the hand) and the subsequent performances.
At the time of this proposal the data collection process is on-going and results have yet to be determined, but they are guaranteed to reveal statistics regarding average durations of interpretations prior to performance deterioration. These results will inform the remainder of the presentation and provide specific recommendations for generating successful interpretations of longer durations with greater accuracy. Additionally, the study will provide insight to identify particular error patterns, providing evidence that may help professional interpreters advocate for teaming interpreted assignments.
Brasel, B. 1976. The effects of fatigue on the competence of interpreters for the deaf. In H.J. Murphy (ed.), Selected Readings in the Integration of Deaf Students at C.S.U.N. Centre on Deafness series (#1). Northridge, CA: California State University.
Educational Objectives
Working professionals in the field of interpreting will gain comprehension of Brasel’s original research and how it has affected our field’s current expectations of best practices regarding the duration of interpreting segments within a teamed approach to interpreting. Additionally, audience members will learn about the presenter’s recent research regarding these topics.
Audience members will be able to apply practical knowledge regarding the recognition of symptoms of fatigue that appear within signed ASL target texts.
Audience members will attain strategies for problem solving regarding interpreting assignments in which teamed interpretation is not provided. These strategies will include suggestions for how to advocate for providing services through teamed interpretation as well as suggestions for how to reduce stress and enhance target-text integrity during solo interpreting conditions.
Information about the Presenter
Brian Cerney, Ph.D., CI, CT, ASLTA-Professional has been a nationally certified interpreter since 1991 and is currently an Associate Professor within the ASL - English Interpreting Program at Keuka College in New York. Dr. Cerney’s research interests include interpreting processes, human-cognitive responses to stress, and ASL/English/Interpreting pedagogy. He has presented on topics ranging from interpreters serving as language models in mainstream environments, teaching methodologies for ASL, and evaluation methodologies for target texts.
In short, the answer is "yes" but let's start by looking at how we got here.
Background on Brasel's (1976) study
Barbara Babbini Brasel (B3) conducted her research at CSUN. Her research was published in 1976 as part of a Center on Deafness Publication Series (1976 was the first year of the publication series). Brasel's study was the fourth article in the section titled "USE OF INTERPRETERS". The fifth article, however, mentions the exact same title - "The Effects of Fatigue on the Competence of Interpreters for the Deaf" in its bibliography, but provides a publication date of 1969 (and also identifies the source as "San Fernando Valley State College" - the former name of the place now known as the California State University at Northridge). This means that the 1976 publication date actually refers to research conducted in the 1960s, no later than 1969.
Within the 1976 publication there are references to information "that will be discussed later in this paper", but no subsequent discussion occurs. It is this researcher's assumption (educated guess) that the 1976 publication was actually a summary of the original 1969 research rather than a complete reproduction of it.
The original study involved five subjects (three female, two male) but one (male) was eliminated from portions of the study due to insufficient interpreting skills. Each subject's skills were evaluated using a 13-item evaluation tool (not revealed in the study's 1976 publication) that rated skills based on five minutes of interpreting work. who interpreted different lengths of lecture content (pre-recorded on audio tape): 0 minutes (control subject), 20 minutes, 30 minutes, 60 minutes and 90 minutes.
The 1976 publication clearly identifies that the control subject was female and that her skills were evaluated differently (based on her participation in a separate study that had used the same evaluation tool). This leaves two female and two male interpreters for the remainder of the study. Of all five subjects Brasel indicates that three had "good/outstanding skills," one had "acceptable skills" and one had "high unacceptable skills". Further description in the 1976 publication reveals that the 60-minute and 90-minute interpreters had "good/outstanding skills" and that the insufficiently-skilled interpreter performed for 20 minutes. Thus the resulting table can be created:
It is worth noting that this research was not exclusively related to the duration of interpreting assignments and interpreter fatigue. The study actually had several other assessments including a trigram test, a math test and a typing test. The trigram test involved revealing ten cards with three-letter sequences for 3-second intervals followed by the subject writing down as many trigrams as they could remember. The math test required subjects to correctly add a column of ten three-digit numbers. The typing test allowed subjects five minutes to re-type a source (typed) text. Each subject completed all of these assessments prior to their interpreting task, thus the need for the control subject to account for test/te-test effects.
As with any study the low number of subjects gives room to doubt the general applicability of the results to the general population. The study is still useful, however, because it sheds light on the question of assignment duration related to interpreter fatigue.
The evaluation of interpreter fatigue affecting the interpretation itself was accomplished using a four-person panel of two Deaf evaluators (Brasel being one of them) and two skilled Interpreter evaluators. Deaf judges looked for errors (mis-articulation, semantic mismatch, omissions) and also noted any time that they felt "baffled or confused or did not understand a fingerspelled or signed word". The Interpreter judges focused on omissions, substitutions, or other skewing where the source and target texts did not match. Duration of the interpretation was divided into 5-minute segments by a time-keeper, visible to the judges but behind the subject who indicated five-minutes, ten-minutes and so forth.
There are several points to notice about the study. First, there was no video recording made of the interpreting performances. All evaluation was accomplished by the written comments of four panel members watching the live performance. The second point to consider is that the subjects had live audience members watching their work, but every time they wrote something down, it was to note an error (negative feedback). Third, the source text is an audio recording that cannot be stopped or interrupted for clarification. The speaker is not physically present and no visual aids (gestures, notes, pictures, etc) accompany the source message.
The most significant results that Brasel reports are that "for up to 30 minutes, there are no significant differences in interpreting competence, errors or quality –– although a deterioration can be noted beginning at about 25 minutes. After 30 minutes there is a slow but steady increase in error rate and after 60 minutes this increase becomes significant." In other words, we can interpret 30 minutes without significant deterioration in the message (but fatigue is noticeable at about 25 minutes). Interpreters can only work about an hour before their work begins to significantly suffer.
These results largely depend on the interpretations generated by the two subjects who worked longer than 30 minutes.
So 1976 gives us support to request team interpreting or at the very least, switching interpreters for assignments of one hour or longer. No other studies specifically investigating assignment duration and interpreter fatigue were conducted until Gabrian and Williams published their 2009 research "The Effect of Interpreter Fatigue On Interpretation Quality"
The 2009 study was conducted at Gallaudet University. The researchers were able to recruit a single subject and chose a case-study approach. The subject was video recorded interpreting two different extended source texts (80+ minutes each); one was an ASL source text interpreted into English, the other an English source text interpreted into ASL. Because of the overwhelming amount of data it was reduced to four discrete five-minute segments for each task (ASL-to-English and English-to-ASL): from 10 to 15 minutes; from 30 to 35 minutes; from 50 to 55 minutes; and from 74 to 79 minutes (due to avoiding a change from monologue to dialogue during a question-and-answer session at the end of a lecture).
Both the ASL to English and English to ASL samples demonstrated deterioration of the message over time as measured through OMISSIONS. The interpretation into ASL showed a gradual increase in omissions across all four segments. The interpretation into English showed a constant rate for the first three segments followed by a dramatic increase in the final segment.
CONCLUSION - The research conducted by Gabrian and Williams support the general conclusions that Brasel had found: That sustained interpretations show signs of deterioration by 30 minutes and the quality is more seriously diminished after one hour.
So now we are ready to understand the current study.
Abrams, Cerney, Hoock, Marble, Prestano & Staehle (2011)
In January of 2011 41 staff interpreters employed by the National Technical Institute for the Deaf (NTID) were asked to participate in a study on Interpreting, based on their Winter Quarter schedules placing them within a lecture setting for at least one hour on their own, without a team interpreter. Eighteen of these interpreters consented to participate in the study. Four were not able to provide the study's goal of three forty-minute (or longer) samples of work with the same consumers and setting. Of the fourteen subjects that remained in the study nine provided a single data set, three provided two data sets each, and one subject completed four different data sets. This resulted in a grand total of 20 completed sets of three different recordings (each one 40 minutes or longer) within the same setting/participants.
Ten additional sets of at least one recording 40 minutes or longer (four of them additional subjects not in the primary data set, six are interpreters already in the primary data set but with different participants/settings). These data sets were incomplete due to scheduling of midterm exams; field trips; or interpreter, student, or faculty absences. Additional data was also discarded due to the subject being the second interpreter in a two-hour class that dismissed prior to a full forty minutes of class time in the second hour.
The data was collected by four Interpreting students at Keuka College as part of their Field Period experiences. Student researchers were available for the first three and a half weeks of January which was when all of the data was collected. Students met with the interpreters prior to beginning data collection to ensure a smooth integration with existing classroom behaviors (introduction to instructor and students, location of the camera, etc.) All subjects regularly (at least annually if not quarterly) videotaped their own work within classroom settings so the provision of a student researcher to run a video recording was generally perceived as a benefit to the subjects, who were able to obtain copies of their own performances to review privately. The cameras were all HD Vivitar cameras with fixed lenses and digital zoom. The pixilation of the digital zoom on these cameras caused poor quality video and the researchers were instructed to refrain from depending upon it for capturing the subjects' interpretations. The focus of the entire study was on the ASL production of English-to-ASL interpreting. All of the recordings were captured on SD memory cards. Recordings were copied from these cards into a Macintosh computer running Quicktime to play the video files. Backup copies were made onto an external hard drive and the SD memory cards were erased and used again for subsequent recordings.
In addition to the video recording of interpreting work each subject was asked to complete a six-item questionnaire. Three items asked the subjects to rate (on a 1 - 5 scale) their own mental alertness, physical stress and topic preparation prior to and after each interpreting sample. Student researchers were also asked to note any possible indications of fatigue that they noticed.
The analysis of the data is on-going. Two initial explorations have been completed. Both were considered productive and may result in a more widespread analysis in the future. Neither of these initial explorations of the data have included the questionnaire or student researcher notes. Both of the initial analyses were conducted directly by the lead researchers, Abie Abrams and Brian Cerney, in June and July of 2011.
Analysis #1 - Biomechanics and Evidence of Fatigue
The first analysis was conducted on only the first data sample (of the three recordings made) for each of six subjects. The subjects were chosen based on their relative tenure as staff members. Two subjects represented "RECENT HIRES" and had worked as NTID Staff interpreters for less than five years. Two subjects were "MEDIAL HIRES" and had worked as NTID Staff interpreters for between ten and fifteen years. Two subjects were "LONG-TERM HIRES" and had worked as NTID Staff interpreters for more than twenty years.
Nine unique one-minute segments were analyzed for four factors: 1) "Behaviors indicative of fatigue", 2) "Behaviors contra-indicative of fatigue", 3) "Positive biomechanic behaviors" and 4 "Negative biomechanic behaviors". The one-minute video samples were viewed starting at the twenty-minute mark and proceeding as follows in the order indicated:
20:00 - 21:00
25:00 - 26:00
30:00 - 31:00
35:00 - 36:00
40:00 - 41:00
45:00 - 46:00
05:00 - 06:00 (this allowed for direct contrast of end-of-sample with beginning-of-sample data)
10:00 - 11:00
15:00 - 16:00
20:00 - 21:00 (second viewing of this segment to verify the first analysis)
Information noted regarding Positive or Negative biomechanic behaviors did not significantly change across the data sample. If subjects were demonstrating good biomechanics early, they generally continued to demonstrate them later. If subjects demonstrated poor biomechanics early, they generally continued to demonstrate them later (with the singular exception of some subjects making use of more microbreaks later rather than earlier in their data sample).
The more interesting results from this first analysis came from looking for behaviors indicative of fatigue (such as posture changes, stretches and restrained yawns) and behaviors contra-indicative of fatigue (such as extra effort/movement or emphasis/animation in the target text).
The information at the bottom of the chart indicates the relative indicators of fatigue between the RECENT (less than 5 years) hires, the MEDIAL (10-15 year) hires, and the LONG-TERM (20+ years) hires. Indications of fatigue decrease with employment duration. Contra-indicators of fatigue also decrease with employment duration. This analysis is inconclusive but may demonstrate that fewer indicators AND contra-indicators of fatigue are revealed across greater durations of employment. In other words, interpreters with longer employment durations reveal less information about their level of fatigue in general (good or bad) and perhaps this merely means that they are better at "pacing" themselves across assignments.
The information from the right-most column provides the cleanest indication of interpreting fatigue over time. Interpreters demonstrated a fairly steady increase of indicators of fatigue over time, with an interesting pattern of temporary recovery at around twenty minutes. This is interesting in light of the fact that this segment was viewed twice to ensure the accuracy of the first analysis. Repeated viewings would likely REDUCE the score rather than enhance it because additional fatigue could be noted in the second reviewing of the sample. Generally the scores for the first viewing were not altered as a result of the second viewing. In other words, it appears that the effect of temporary reduction in the indicators of fatigue at around twenty minutes is a real phenomenon and merits further study.
Please remember, this is an INITIAL analysis of the data and it only involves six subjects. The exact behaviors exhibiting an INDICATION of fatigue or a CONTRA-INDICATION of fatigue are not yet precisely defined and ultimately will still be largely subjective.
The second analysis was conducted on only the second data sample (of the three recordings made) for all fourteen subjects with complete data sets. This analysis was inspired by an original observation by this researcher in the Spring of 2010 - it was observed that the first incidents of SELF-REPAIR within an interpretation among very skilled interpreters occurred around the twentieth minute (regardless of the target language - ASL or English). After this first incidence of self-repair the observed interpreters regained their accuracy that they had previously had for several more minutes. These unstructured observations lead to the desire to re-investigate interpreter fatigue and the pursuit of the current study.
From the data sets mentioned above, each was reduced to a fifteen-minute sample starting at 10:00 minutes and continuing until 25:00 minutes. Incidents of self-repair were fairly obvious to note because they consistently co-occured with non-manual signals marking the self-repairs. Two forms of self-repair were noted, and each presented with distinctly different non-manuals:
Self-Corrections were lexical errors that were replaced (or attempted to be replaced) with the more conceptually correct sign. The non-manual signals that co-occured with this version consistently included an eye-blink at the point of halting the errant sign. Twenty incidents of this type were observed and almost half (9 incidents or 45%) were related to self-correcting fingerspelling errors.
Re-Statements were processing substitutions that sometimes restated the exact same vocabulary but with a different word-order. These self-repairs did NOT include eye-blinks at the point of restatement (a variety of other behaviors were observed including eye-gaze rolling upward (as in thought) or a brief pause while waiting for more source text information to be processed. Seven instances of this type of Self-Repair were noted.
Overall 27 incidents of Self-Repair were recorded across twenty fifteen-minute samples. Four samples had zero Self-Repairs (20%). Seven samples had one Self-Repair (35%). Six samples had two Self-Repairs (30%). One sample had more than three Self-Repairs (5%). Overall the average number of Self-Repairs was less than 1.5 per fifteen-minute sample.
Looking at the data across time there is no obvious pattern to the numbers of incidents of Self-Repair as related to the duration of the assignment. This data set indicated two peaks at around ten minutes and around 14 minutes but otherwise the data indicates a fairly even number of incidents of Self-Repair across the entire time of the data samples.
The fact that most interpreters demonstrated at least one Self-Repair in their work, and that only one interpreter demonstrated more than two incidents suggests that looking at Self-Repairs may not be a productive evaluation of fatigue. One interpreter exhibited more than two Self-Repairs within one fifteen-minute sample (but in another sample from the same interpreter only one incident of Self-Repair occured). The shear number of incidents within the sample (six incidents within fifteen minutes, or an average of one Self-Repair every two-and-a-half minutes) may indicate interpreter fatigue for that interpreter, for that class, on that day. In fact, a number of other behaviors were noted (stretching, non-manual behaviors exhibiting self-frustration) within the same fifteen minute sample that clearly indicated fatigue. Thus a high number of Self-Repairs would likely co-occur with other indicators of fatigue. In other words, Self-Repairs are a normal part of the interpreting process and noting them in isolation is not likely to be a useful indicator of fatigue.
Given the research by Brasel (1969/1976), Gabrian & Williams (2009), and the current study what conclusions can be make regarding assignment duration and interpreter fatigue?
The first observation is that there is evidence of fatigue setting in after twenty-five minutes across ALL THREE of these studies and that the evidence of fatigue becomes greater as the total assignment duration increases. This supports the idea that interpreters should switch interpreter roles between twenty and twenty-five minutes of sustained lecture content.
This study did NOT investigate interpreting situations where interpreters DID switch, however, and so it cannot speak to the effectiveness of sustained TEAM interpreting as a hedge against fatigue. In other words, both members of an interpreting team may very well BOTH become fatigued within a sustained interpreting environment if both team members are actively processing the informational content for more than 25 minutes, even though only one of the interpreters is actively interpreting that content.
The second observation, so far, is that overall the message quality does not appear to be significantly impacted within forty-five minutes of sustained interpretation. During the data collection process each data sample was briefly viewed at random segments to verify that each recording was successful. At no time were any of the interpretations determined to be inaccurate or less than useful to their prospective consumers. This observation is not supported by a thorough analysis across all the subjects, however, so it remains in doubt.
Other observations that occurred within the data collection process but not yet directly analyzed include the following:
Interpreting for an inattentive Deaf consumer can be fatiguing. Interpreting for an extremely attentive Deaf consumer can serve as a counter-balance to fatigue.
Interpreting a disorganized message or a message generated in less-than-natively-fluent English (common examples come from student presentations) can be fatiguing.
Monologic (lecture) content is often less fatiguing than dialogic (conversational) content.
Content that is familiar, of interest to the interpreter or otherwise predictable is generally less fatiguing than uninteresting/unknown content or presentation styles where it is difficult to make predictions about the upcoming content.
Interpreting group discussions is generally perceived by interpreters as the most difficult work for interpreters to be effective.
In other words - Source and Target consumers can significantly impact the interpreter's own sense of fatigue. Organized source texts and attentive target-language consumers have positive effects. Disorganized/disfluent source texts and in-attentive target-language consumers have negative effects.
