Language Journal » Probably

« previous article |

Excerpt from Technology Connections on Bluesky:

My favorite linguistic peculiarity is the almost universal shortening of “probably” to “snoop” in speech.

And people don’t realize they’re doing it! If I write, “Yeah, that’s a nudge that won’t work,” it doesn’t work, but if you say it out loud, it absolutely does.

— Technology Connections (@techconnectify.bsky.social) August 10, 2024 at 2:52 p.m.

To begin testing this, I took 100 random occurrences from the 39,731 occurrences of “probably” in the NPR podcast corpus I referenced earlier (3,199,859 transcribed rounds from 105,817 NPR podcasts, comprising over 10,648 hours).

What did I find?

The word “probably” /ˈpɹɒbəbli/ (dictionary pronunciation) has two intervocalic onsets, /b/ and /bl/, which are not followed by stressed tautosyllabic vowels. They are therefore candidates for the general process of unstressed intervocalic lenition typical of American English, the best-known version of which is the flapping and voicing of coronal stops.

This articulatory and acoustic weakening occurs to varying degrees in the /b/ and /bl/ of “probably”, including apparent deletion.

Let’s start with an example that doesn’t have lenition, taken from “Online Calculator Estimates Breast Cancer Risk”, Morning Edition 04/19/2007:

Professor KARLA KERLIKOWSKE (Medicine, University of California, San Francisco): When you’re creating these models, you want something that’s relatively simple and easy to measure.

AUBREY: And lifestyle is not a determining factor. No one remembers exactly what they eat or drink. But Kerlikowske would like to revive Gail’s model in a few ways. She would ask the question of the age at which women gave birth to their first child. Not because other factors like breastfeeding confound that hypothesis, but because research suggests that it’s not as important a factor as breast density.

Prof. KERLIKOWSKE: We appreciated the importance of breast density for probably 30 years old.

AUBREY: The problem is that there has never been an automated way to measure, during a mammogram, the volume of epithelial cells and surrounding tissues that are typically involved in breast cancer.

Prof. KERLIKOWSKE: We think that if we had a quantitative measure of breast density, it could add to the specificity of the model and really improve the prediction for individual women.

Here is the full sentence containing the unmitigated “probably”:

And here we focus on the word “probably” itself:

As you can see, each of the two /b/ performances has a well-defined quiet stop, followed by a clear burst of release.

At the other end of the lenition continuum, here is an example where both appearances are attenuated to the point of apparent suppression, from “Machines Slowly Mastering Art of Recognizing Faces,” Let’s talk about the nation 01/22/2022:

CATHY (Caller): Well, hello, Ira. I love your show and you have a very interesting topic today as usual. My question for your guests is about the ability of facial recognition technology to recognize faces as they change over time, for example, a photo of a baby versus a teenager versus an adult versus an elderly person.

FLATOW: Hmm. Jonathan, any comments on that?

Dr. PHILLIPS: So there’s probably – again, this is very active research. But there is probably There are two different areas of aging with facial recognition. One, for example, is the transition from infant to adult, because there are fixed ways of changing and the changes are substantial. The other is the adult from age 20, for example, as they grow.

Facial recognition seems to be relatively stable up to about five years of age. But the challenge then is how do you model or account for how individuals change from age 20, say, from age 10 to age 20 or beyond?

The complete sentence “probably”:

And focusing on the pronunciation of “probably”, which shows the type of reduction described in the Bluesky link:

The speaker is Jonathon Phillips, electrical engineer, program manager, Multiple Biometric Grand Challenge, National Institute of Standards and Technology, Washington, DC. This is not a speech error: he speaks fluent American English, in a fairly formal but rapid style.

In my modest sample of 100 “probably” sentences, there were many other lenition patterns. Sometimes one or both /b/s lost their occlusal closure and became a kind of voiced approximant that phoneticians sometimes call “fricativeless fricatives.” Sometimes one was lenited to the point of apparent deletion while the other remained occlusal, with results that could be written as “probably” or “probby.” This sample does not support the claim that full lenition on a phonetic syllable is “nearly universal”—and the result is not the same as the pronunciation of “pry”—but it certainly happens.

Since these are gradual articulatory and acoustic changes, it is not easy to assign them to qualitatively distinct categories, but if I were forced to define a set of probable pronunciation categories, I would need about a dozen. A qualitative and quantitative analysis of these variation patterns and the factors related to them would make a good semester project for a phonetics course.

Some previous articles on the issue of phonetic lenition in English:

“Weak T”, 04/17/2017
“Beyond the (International Phonetic) Alphabet”, 04/19/2018
“Beyond IPA”, 01/18/2020
“First novels”, 03/13/2022
“The Evolution of Pronunciation”, 04/15/2022
“More post-IPA astronauts”, 04/16/2022
“Political Snaps and Speeches”, 05/29/2022
“Ron’s Principles,” 08/22/2023
“There’s No T in Scranton,” 10/03/2024
“Intervocalic Coronal Reduction Annals”, 07/26/2024

August 11, 2024 at 9:51 am · Filed by Mark Liberman under Phonetics and Phonology

Permanent link