The cross-sectional studies include data from children and adults (4-94 years old) in four sets of experiments related to the perception and production of dialect variation in American English:
- dialect categorization, including classification, discrimination, labeling, and locality judgment tasks
- social attitudes, including ratings of talker intelligence and wealth (status dimensions) and friendliness and honesty (solidarity dimensions)
- cross-dialect speech intelligibility
- speech production, including picture-prompted word list and story-telling tasks
All of the perception tasks involve speech samples produced by adult native speakers of four major regional dialects of American English: New England (blue), North (red), Midland (green), and South (yellow).
1a. Dialect classification (Jones et al., 2017)
Participants listen to brief speech tokens from a set of talkers and are asked to arrange the talkers into dialect groups based on the way they talk. Category labels are not provided and participants are free to make as many groups as they wish so that their perceptual categories can be assessed. The results of this experiment demonstrate that children as young as 4-5 years old can classify New England talkers into a separate category from Midland, Northern, and Southern talkers. By 6-7 years old, children can separate local Midland and Northern talkers from non-local New England and Southern talkers. By 8-9 years old, children’s classification performance is comparable to adults: they can reliably separate talkers into three categories: New England, Southern, and a mixed Midland and Northern group. Our data are the first demonstration that children as young as 4 years old have the ability to classify talkers on the basis of regional dialect, suggesting that they can use phonetic detail in the speech signal to group talkers by region in a task that is appropriate for their cognitive development. Specifically, the memory demands in the classification task are low because they can listen to the talkers as many times as they want while making the groups. The world knowledge demands are similarly low because the participants are not required to apply category labels to their talker groups.
1b. Dialect discrimination and labeling (McCullough et al., 2019b)
In the discrimination task, participants listen to two speech tokens and decide if the talkers are from the same place or different places. They indicate their choice by choosing a pair of same-colored smiley faces (same place) or different-colored smiley faces (different places). In the labeling task, participants are taught to associate regional dialects with different colored smiley faces and are then asked to assign new voices to the appropriate color group. Consistent with the classification results, discrimination of the New England dialect emerges first (beginning at 4-5 years old) and discrimination of the Southern dialect emerges later (beginning at 6-7 years old), but adult-like discrimination is not observed until 12-13 years old. The discrimination task requires keeping two voices in memory, which may explain why performance in this task lags several years behind performance in the classification task, which is less cognitively demanding. Accurate labeling performance emerges even later than discrimination: adult-like performance is not observed until 14-15 years old. The task requirement that children assign talkers to learned arbitrary labels may explain why previous studies using similar tasks have concluded that young children cannot categorize talkers by regional dialect of their native language. Our results demonstrate that adult-like performance in this kind of explicit labeling task emerges later than adult-like performance in less cognitively demanding tasks.
1c. Locality judgments (McCullough et al., 2019a)
Participants are asked to rate how likely it is that a particular talker is from Ohio. These locality judgments parallel the classification results: 4-7 year olds identify New England talkers as non-local and children 8-9 years old and older perform comparably to adults, identifying New England talkers as the least local, Midland and Northern talkers as most local, and Southern talkers in between. These results suggest that children acquire an understanding of how speech can mark a talker as local or non-local during the preschool years and have refined this knowledge to adult-like levels by the time they reach elementary school.
2. Social attitudes (McCullough et al., 2019a)
Participants are asked to rate how smart, rich, friendly, and honest the person sounds. The social attitudes data show a different pattern of development than the dialect categorization data (1a-c). Children do not begin to apply the status stereotypes (smart, rich) or the solidarity (friendly, honest) stereotypes to talkers of non-local dialects until 6-7 years old and they do not show adult-like stereotypes until at least 12-13 years old. Thus, the social attitudes that accompany non-local dialects are not acquired until elementary school and are critically learned after the geographical knowledge about localness is mastered.
3. Cross-dialect speech intelligibility (Clopper & Wagner, 2019)
In the speech intelligibility task, participants listen to short phrases mixed with noise and are asked to identify what they hear. The data demonstrate that local Midland talkers are the most intelligible and New England talkers are the least intelligible for Midwestern listeners, with Northern and Southern talkers in between. The results also show adult-like patterns of dialect intelligibility by 4-5 years old, although adult-like accuracy is not achieved until 12-15 years old. The results also reveal an intelligibility difference between Midland and Northern talkers, which is critically different from all of the categorization tasks, in which these two dialects were not strongly differentiated, suggesting that intelligibility differences can emerge even in the absence of explicit awareness of dialect differences.
4. Speech production (Nelson, 2017)
In the production tasks, participants are asked to name colors presented on the screen and to tell two familiar stories (Goldilocks and the Three Bears and Little Red Riding Hood) with the aid of wordless picture books of the stories. Preliminary analysis of the color word productions confirms that this task elicits regional variation, especially evidence of the Northern dialect among 4-7 year old children from the Northern dialect region.