Broadcast media and sound change

The role of broadcast media in sound change is widely debated, and direct evidence is scarce. This study tested whether television exposure can change the pronunciation of second-language speakers in a setting where media are the main source of native-speaker input: German secondary-school students learning English.

Linguistic variables

The study looked at two variables. Most American English accents are rhotic, realising post-vocalic /r/ (e.g. car, nurse, letter), whereas most accents of England are non-rhotic. In American English, intervocalic /t/ after a stressed vowel is also realised as a flap [ɾ] (e.g. water, better), known as /t/-flapping; British English typically has [t] or a glottal stop. Each token was coded for the presence (1) or absence (0) of the American variant.

Results: media exposure

/t/-flapping increased significantly in the group exposed to the American series; the decrease in the group exposed to the British series was not significant. Rhoticity changed in the expected direction in both groups, but the interaction of group and exposure was not significant. The chapter suggests that the difference between the two variables may be related to /t/-flapping being a more “off-the-shelf” variant than rhoticity.

Who changed?

In the figure below, each line is one participant, which shows how much variation lies behind the group means. For /t/-flapping, each percentage rests on only 11 to 17 tokens per session, so a single token moves it by six to nine percentage points.

In the American-series group, flapping went up for five of the nine participants, three of them by more than 30 points (AE6, AE7, AE9), and went down for three. For rhoticity, one of the largest changes in the study is in the British-series group, where BE6 went from 57% to 16%. BE6 also reported the highest involvement with the series of any participant, a link the section on involvement below returns to.

Results: involvement with the series

Three individual measures were tested: English aptitude, engagement (regularity and frequency of viewing) and emotional involvement with the series. Only involvement explained variation in the data. For rhoticity, the three-way interaction of group, exposure and involvement approached significance (χ² = 3.81, p = .051); in a combined model of both variables, it was significant (χ² = 4.72, p = .02).

The plot below shows each participant’s change in the proportion of American variants against their involvement score. If involvement moderates accommodation, the slope should be positive in the American-series group and negative in the British-series group. Points can be moved or removed to show how much a single participant affects a slope with nine participants per group. These per-participant regressions are for illustration only; the inferential analyses in the chapter use mixed-effects logistic regression.

feature rhoticity /t/-flapping both combined

fit one line per group one line for everyone reset points

watched Atypical (American) watched TEOTFW (British)

Drag a point to move it. Double-click (or double-tap) a point to remove it.

Each point is one participant. y: change in % American variants, after minus before (percentage points). x: emotional involvement with the series, z-scored (0 = average). Lines are simple least-squares fits, lm(change ~ involvement), with p for the slope. They are for illustration only; the chapter's analyses use mixed-effects models.

Reference

Rathcke, T., Castellano, C., & Canzi, M. (2024). Understanding the role of broadcast media in sound change. In F. Kleber & T. Rathcke (eds.), Speech Dynamics: Synchronic Variation and Diachronic Change, 425–452. De Gruyter. doi:10.1515/9783110765328-015

Analyses in R with mixed-effects logistic regression (lme4, lmerTest, emmeans). The plots on this page use the study’s final data, summarised per participant.