Correlates and effects of delayed responses to job interview questions
Adrian Bangerter, Julie Brosy, and Paloma Corvalan, University of Neuchâtel
Disfluencies preceding responses to questions can reveal much about the further time course of the response or about respondents’ internal states. For example, respondents who are unsure they know the answer to a question produce more fillers like uh or um in the course of responding (Smith & Clark, 1993), and listeners are able to use this information to estimate how much respondents know (Brennan & Williams, 1995). Similar phenomena occur in interview settings. In survey interviews, pauses, fillers, repair and other disfluencies are indicators of respondents’ need to clarify questions (Schober & Bloom, 2004).
Job interviews are a common everyday setting where candidates for a job position answer questions asked by recruiters. Their answers are used by recruiters to make judgments of suitability for the position and, ultimately, hiring decisions. Recent innovations in job interview best practices involve so-called past-behavior questions, where candidates are asked to talk about a past job-related episode they mastered well. Such questions typically require producing a narrative about the situation, actions, and events that transpired. Not all candidates can produce such narratives on demand, and responding to past-behavior questions may thus involve a tradeoff between (1) delaying a response to search memory for a suitable episode to narrate at the risk of appearing inept and (2) responding quicker but perhaps less appropriately (e.g., in non-narrative form). Therefore, pause durations in job interviews may be related to the type of response (narrative or non-narrative) produced by the candidate.
We investigated this question in a corpus of 62 job interviews. Candidates responded to four past-behavior questions. We transcribed each question and the initial responses, and everything in between. In other words, all pauses, fillers, discourse markers and embedded question-answer sequences before the initial response were transcribed, as well as sighs and laughter. We coded initial responses as narrative (e.g., a story) or non-narrative (e.g., a self-descriptive utterance like “I am good at communicating with people”).
The mean total pause duration is 3.8 s (SD = 4.9 s), with wide variations (min = 0.1 s, max = 35 s). We analyzed the typical patterns of pauses (results will be reported in detail in the talk). A generalized linear mixed model analysis was run in R 3.0 to predict the probability of a narrative or non-narrative initial response as a function of the total pause duration, with each individual response clustered at the candidate level. Increased pause duration significantly increased the probability of a non-narrative response. At the candidate level, we also found that the total pause duration in responding to all questions was a negative predictor of hireability ratings by professional interviewers, even while controlling for several ancillary variables.
The results extend our understanding of disfluent responses in several ways: (1) job interview settings constitute situations affording large variations in response delays, and in particular, some very long delays, (2) delays also predict response parameters, not only in survey interview settings, but also in job interview settings, and (3) delays are interpreted with regard to personal competencies.
Brennan, S. E. and Williams, M. (1995). The feeling of another’s knowing: Prosody and filled pauses as cues to listeners about the metacognitive states of speakers. Journal of Memory and Language, 34, 383-398.
Schober, M. F., & Bloom, J. E. (2004). Discourse cues that respondents have misunderstood survey questions. Discourse Processes, 38, 287-308.
Smith, V. L., & Clark, H. H. (1993). On the course of answering questions. Journal of Memory and Language, 32, 25-38.