Visual and Vocal Interview Cues: How Your Voice in a Job Interview Predicts the Verdict and the Job

By Noam Shemla · The Studies

TL;DR DeGroot and Motowidlo measured the voices of managers in recorded job interviews with a computer, and compared them with how the interviews were judged and how well the managers actually performed. A combined score for five vocal cues predicted both. The link ran through people’s personal reactions: listeners who heard only the voice liked, trusted and credited some managers more than others, and those reactions tracked both the interview verdict and the job ratings.

• Paper: DeGroot, T., & Motowidlo, S. J. (1999). Why visual and vocal interview cues can affect interviewers’ judgments and predict job performance. Journal of Applied Psychology, 84(6), 986–993.
• Study 1: 60 managers from four utility companies, interviewed on video as if applying for their own jobs; their supervisors rated their performance.
• Study 2: 110 managers in a news-publishing company, interviewed the same way with nine structured questions.
• Voice: Five cues measured by a computerised voice analyser from short speech samples: pitch, pitch variability, speech rate, pauses and amplitude variability.
• Looks: Physical attractiveness, smiling, gaze, hand movement and leaning in, rated by students from silent video.
• Reactions: How much people would like, trust and credit each manager, rated separately from sound only and from silent video only.
• Outcomes: Interview judgments by 26 employees of the company, and job performance ratings from each manager’s supervisors and colleagues.

Interview advice usually concentrates on what to say. This paper is about how it sounds. Timothy DeGroot and Stephan Motowidlo recorded managers answering structured interview questions, measured their voices with a computer rather than by ear, and asked two questions: does the voice sway the interviewer, and does it predict how well the person actually does the job?1

What they measured

The managers were already in their jobs. They were asked to answer the interview questions as if they were applying for the position they held, which is what made it possible to set the interview against real performance ratings. From short samples of each interview, a voice analyser produced five measures.

The five vocal cues, and what the authors expected of each

To test the voice as a whole, the authors combined the five cues into one score: each cue standardised, the two expected to help added, the three expected to hurt subtracted. They also combined repeated samples of each cue, and estimated the reliability of the combination with a classic formula.

Study 1: a pilot that pointed the right way

With 60 managers, no single cue was significantly related to supervisors’ ratings of performance. But every one of them pointed in the expected direction, and the combined score correlated .18 with performance. The authors treated this as a feasibility check and as a reason to run a larger study with more speech per person.

Study 2: voice, looks and the reactions in between

The main study added two things: judgments of the interview itself, made by 26 employees of the company watching the tapes, and a way to test why nonverbal cues might matter. Separate groups of students rated how much they would like, trust and believe each manager, one group from the sound alone and another from silent video alone. That kept the two channels apart: the voice-based reactions could only come from the voice.

How the cues, the reactions and the two outcomes correlated (Study 2, N = 110)

Three things stand out in that matrix. The voice score predicted both the interview verdict (.32) and job performance (.20). Reactions formed from the voice alone predicted both even more strongly (.41 and .26). And the interview verdict itself predicted performance (.35), so the interviews were doing their job.

Why: the mediation tests

The authors then asked whether the reactions explain the link, using the three-step test of mediation standard at the time. First, the cues must predict the outcome; second, the cues must predict the reactions; third, with both in the model, the reactions must still predict the outcome, and the cues' own effect should shrink.

All three mediation tests

So reactions carried the effect of looks on the interview, and of voice on both the interview and the job. The difference matters: an attractive, smiling candidate impressed interviewers, but that impression did not predict how they would perform. The voice predicted both, and in the same way.

The authors' explanation is that managing is interpersonal work. A manager whose voice makes people like, trust and believe them is likely to find it easier to win cooperation, and interviewers may be picking up on exactly that.

Has it held up?

The authors were unusually cautious about their own result, and their reasons are worth repeating:

• As far as they knew, it was the first study to link specific vocal cues in an interview to job performance, which they called too exotic to act on without extensive replication.
• An alternative explanation was not ruled out: both interviewers and performance raters might be biased by the same cues, whether or not those cues have anything to do with managing well.
• The correlations are modest, and there was no evidence that voice adds anything beyond established predictors such as the structured interview itself.

They explicitly declined to recommend using voice in hiring decisions.

What this study cannot tell you

• The interviews were simulated. Managers answered as if applying for jobs they already held, which is what made the design possible, and may not sound the same as a real candidate with something at stake.
• The effects are small. A correlation of .20 means voice accounts for about 4% of the differences in performance ratings.
• Performance ratings are judgments too. If raters respond to the same vocal cues, voice would predict ratings without predicting performance.
• Some cues were measured with proxies. Speech rate, for example, was estimated from pitch-period length rather than counted syllables.

What a speaker can take from it

Two practical points, both modest, which is how the authors would want them.

The voice is heard, even when nobody is listening for it. The reactions that carried the effect were formed by strangers who heard short clips of the interviews without a picture. How you sound registers as liking, trust and credibility before anyone weighs what you said.

The voice that helped here is learnable. Faster speech, fewer pauses and more pitch movement were the directions that pointed the right way. None of them is fixed, and all of them respond to rehearsing your answers out loud until you no longer have to search for the words.

This article summarises a published paper. The correlations, reliabilities and regression coefficients are the authors’ own, from Tables 1 to 3 and the method sections. The confidence intervals in the Study 1 plot are ours.

Measured

• Five vocal cues, from a computerised voice analyser
• Visual cues and personal reactions, rated by students from silent video or from sound alone
• Interview judgments by company employees, and performance ratings by supervisors and colleagues

Inferred, not measured

• Mediation by personal reactions: a correlational test, consistent with the explanation but not proof of it
• That voice in the interview reflects voice on the job, which the authors assumed rather than measured

References

1. DeGroot, T., & Motowidlo, S. J. (1999). Why visual and vocal interview cues can affect interviewers’ judgments and predict job performance. Journal of Applied Psychology, 84(6), 986–993. https://doi.org/10.1037/0021-9010.84.6.986