Introduction
A regional accent is more than phonological variation in speech. It also provides social information about the speaker, first and foremost that person’s place of birth or residence. How listeners perceive and interpret regional variation at different points in their lifespan has been the subject of much recent research (see Cristia et al. (Reference Cristia, Seidl, Vaughn, Schmale, Bradlow and Floccia2012) for a comprehensive review). Adults have been shown to reliably use regional variation in speech to identify a speaker’s home region (Labov Reference Labov and Linn1998; Clopper and Pisoni Reference Clopper and Pisoni2004; Clopper et al. Reference Clopper, Rohrbeck and Wagner2012), and to make social judgments of speakers based on their regional accent (Hay et al. Reference Hay, Warren and Drager2006; Campbell-Kibbler Reference 121Campbell-Kibbler2007; Staum-Casasanto Reference Staum-Casasanto2009). However, it appears that many adults have only general representations of the differences between regional varieties (Preston Reference Preston1986; Labov Reference Labov and Linn1998; Clopper and Pisoni Reference Clopper and Pisoni2007; Clopper Reference Clopper, Preston and Niedzielski2010), and their classification of dialects is subjective, based on how they perceive their own speech (Preston Reference Preston1986; Niedzielski Reference Niedzielski, Preston and Niedzielski2010).
One determining factor in adults’ performance at identifying regional accents is early childhood experience hearing regional variation. Adult listeners who lived in multiple dialect regions as children perform better at categorizing regional accents than those with no early exposure to regional accents (Clopper and Pisoni Reference Clopper and Pisoni2004). There is evidence that exposure to regional variation also enhances discrimination ability in childhood. An accent discrimination study found that 7 year olds from multi-dialectal families in Great Britain outperformed children from families speaking only one regional variety (Girard et al. Reference Girard, Floccia and Goslin2008; Floccia et al. Reference Floccia, Butler, Girard and Goslin2009).
Age may mitigate the positive effects of exposure in children. A study of children’s social preferences based on speaker’s regional accent found that 9 and 10 year olds reliably identified regional accents, and preferred speakers from their native dialect region, whereas not all 5 and 6 year olds could identify the regional accents of their home region, or express social preferences based on a speaker’s accent (Kinzler and DeJesus Reference Kinzler and DeJesus2013).
The results of these studies raise the question of how exposure to regional variation affects children’s ability to interpret, identify, and discriminate between regional accents. The course of development of awareness, and the factors that influence it, are not well understood. This chapter addresses this question by comparing how awareness differs between two groups of children who differ in their history of exposure to regional accents. One group is exposed predominately to the local regional accent, and the second receives input at home in a regional accent different from the local community accent. The expectation is that children who hear multiple regional varieties are more aware and better able to discriminate between regional accents. However, contrary to expectation, early exposure to multiple regional varieties does not improve children’s ability to discriminate between accents and has negative correlations with awareness. This study captures a stage in development in which children’s understanding of regional phonological variation is not yet fully developed, and linguistic input can clearly be seen helping to shape their interpretation of regional variation.
Background
One model of speech perception accounts for how social information is associated with linguistic variation. Exemplar Theory (Johnson Reference Johnson, Johnson and Mullennix1997; Pierrehumbert Reference Pierrehumbert2003; Johnson Reference Johnson2006) hypothesizes that tokens of speech are tagged and stored with information about the speaker’s social qualities, such as gender, age, race, socio-economic status, etc. (Johnson Reference Johnson2006). As tokens of speech accumulate, they group together with tokens tagged as belonging to a particular social category. Eventually, the accumulated tokens are abstracted to create a representation of how that social group speaks. This abstract category allows the listener to identify new speakers belonging to that social group based on similarities between their speech and the representation created from the accumulated tokens. It also allows listeners to form expectations of speakers and their speech, as the representation of a particular variety captures the associations between the speech and social characteristics of speakers (Hay et al. Reference Hay, Warren and Drager2006; Campbell-Kibbler Reference 121Campbell-Kibbler2007; Staum-Casasanto Reference Staum-Casasanto2009).
Little work has been done exploring children’s representations of social variation and the process by which they create social categories and associate them with linguistic variables. Munson (Reference 122Munson2010) hypothesizes that children’s sociolinguistic representations are initially built upon interaction with specific individuals. A child creates proto-representations of social variation by associating a known individual’s speech with the social characteristics attributed to that person. Thus, initially a representation of a regional accent may include only tokens of the speech from one individual the child knows. As the child’s range of social interactions broadens, the speech of additional individuals will be added to the representation, eventually generalizing it from specific tokens to abstract representations of social variation.
Foulkes and Docherty (Reference Foulkes and Docherty2006) expect that for children, the association of speech with social groups may proceed from the most obvious social characteristics, such as gender and age, to the least obvious, such as a person’s region of origin. They also postulate that infrequent or less-obvious traits never develop the same strong links between linguistic units and social groupings, giving underspecified representations for some kinds of social variation. Regional variation is specifically mentioned as one of the more difficult sociolinguistic categories to acquire, as in many cases it has no reliable connection to appearance and is only relatively infrequently encountered. Thus, it would be expected that children would be able to identify ethnic- or gender-based variation before regional variation in speech.
Studies of young children’s perception of social variation partially confirm these predictions. Children of 3 and 4 years of age in the United States were shown to correctly identify African-American and Standard English, assigning common stereotypes to speakers of those varieties (Rosenthal Reference Rosenthal1974). As far as regional variation is concerned, several studies show that 5-year-old children may not be able to accurately identify their own regional dialect, although they may express negative social evaluations of speakers of their dialect (Cremona and Bates Reference Cremona and Bates1977; Millar Reference Millar, Britain and Cheshire2003). Kinzler and DeJesus (Reference Kinzler and DeJesus2013) found that 5 and 6 year olds from the Northern United States showed social preferences for other Northern-accented speakers, and could identify the Northern accent as the one heard in their community. In contrast, 5- and 6-year-old children from the South showed no preferences for either accent, and also were unable to reliably identify the local accent. The difference between the two groups may be that these two accents are not equally distributed across the United States; children in the South are more likely to hear Northern accents than vice versa. Hearing two regional accents may complicate identifying one as local, or assigning social preferences to them.
Ability to identify regional accents appears to improve with age. Kinzler and DeJesus (Reference Kinzler and DeJesus2013) also show that by age 9 to 10, children from both the North and the South are able to identify regional accents and express social evaluations of both accents, indicating that this ability develops with age. As mentioned in the introduction, although Girard, Floccia, and colleagues found that 5 year olds were unable to categorize regional accents, 7 year olds were able to successfully complete the task (Girard et al. Reference Girard, Floccia and Goslin2008; Floccia et al. Reference Floccia, Butler, Girard and Goslin2009).
Exposure to regional variation also appears to affect production of regional variation. Children with parents from outside the Philadelphia region were shown to acquire some regionally specific phonological variants more slowly, or less completely, than children whose parents were from the area, although the children were being raised in Philadelphia (Payne Reference Payne1976; Roberts Reference Roberts and Ball2005). Even more interesting was the finding that success acquiring certain regional phonological features could be predicted by whether the child had one or two parents native to the Philadelphia region (Roberts Reference Roberts and Ball2005).
Children’s Discrimination of Regional Variation
Three studies(Girard et al. Reference Girard, Floccia and Goslin2008; Floccia et al. Reference Floccia, Butler, Girard and Goslin2009; Wagner et al. Reference Wagner, Clopper and Pate2013) conducted, respectively, in France, the United Kingdom, and the United States, all report children aged 5 are unable to discriminate a familiar and unfamiliar regional accent in a categorization task. Five year olds also failed using another experimental paradigm, in which they were asked to identify whether a sentence was spoken in the same accent as the sentence heard immediately prior (Girard et al. Reference Girard, Floccia and Goslin2008; Floccia et al. Reference Floccia, Butler, Girard and Goslin2009).
This is somewhat surprising, as infants have been shown to discriminate between regional accents (Nazzi et al. Reference Nazzi, Jusczyk and Johnson2000; Kitamura et al. Reference Kitamura, Panneton, Notley and Best2006; Phan and Houston Reference Phan and Houston2008; Egerova Reference Egerova2010; Butler et al. Reference Butler, Floccia, Goslin and Panneton2011). Floccia and colleagues postulate that regional variation may be less salient than second language variation, as 5 and 6 year olds were successful in discriminating native and non-native accents using the same categorization tasks (Girard et al. Reference Girard, Floccia and Goslin2008; Floccia et al. Reference Floccia, Butler, Girard and Goslin2009). However, the difference between infants and 5 and 6 year olds may be that the 5 year olds are processing the sentence stimuli for meaning, in addition to discriminating between accents, whereas infants may only be paying attention to the phonological features of the stimuli. The additional burden of interpreting meaning in sentences may divert children’s attention from the accent in which the stimuli were spoken. Given that Floccia and colleagues report that 7 year olds also participating in these studies discriminate between accents with a greater than chance accuracy (Girard et al. Reference Girard, Floccia and Goslin2008; Floccia et al. Reference Floccia, Butler, Girard and Goslin2009), this may be a memory or attention issue rather than an inability to hear regional differences in speech.
An attention-based explanation is supported by the findings of another study, in which 5 and 6 year olds successfully discriminated between regional accents when presented with shorter stimuli. Using an ABX task, and word-length, as opposed to sentence-length stimuli, Beck (Reference Beck2014) shows that 5- and 6-year-old children can reliably discriminate a familiar from an unfamiliar regional accent . Beck (Reference Beck2014) also finds that children can successfully discriminate between two regional accents in another experimental paradigm, in which children are asked to pick the speaker sounding most similar to themselves based on regional accent, again using only single word stimuli. Thus, it seems likely that sentence-length stimuli pose trouble for children between the ages of 5 and 6. This may be the result of either short-term memory limitations or processing constraints preventing them from focusing on the phonological content of the sentences.
A second consideration is that 5- and 6-year-old children are learning to interpret accents for social meaning, and have yet to acquire what regional accent represents. Therefore, a second objective of the Beck (Reference Beck2014) study was to assess whether awareness of regional accents contributed to the ability to discriminate between them. One result was the finding that children whose parents were from outside the dialect region in which the study was conducted differed from all other subjects in their awareness of regional accents. Since this finding parallels other studies in which early exposure affects children’s abilities to identify regional variation, a separate analysis was conducted to isolate and examine these effects.
This chapter presents the results of that analysis. The effect of early exposure to regional variation on 5 and 6 year olds’ awareness and ability to discriminate between regional accents is examined in the results of two tasks from the larger Beck (Reference Beck2014) study. The participants in the study were children living in a town of about 30,000 residents on the outskirts of Philadelphia, USA, and attending kindergarten at one of two public schools in the town. The performance of two sub-groups of these participants, dubbed “Insiders” and “Outsiders,” are compared. Insiders are from families where at least one parent is a native of the child’s hometown. Outsiders are children from families where neither parent is a native of the hometown, and therefore likely are not speakers of the local regional variety. Both Insiders and Outsiders were born and raised in the same town, and therefore have exposure to the same local regional variety. Insiders, however, have had most of their input exclusively in the local regional variety, and don’t have experience with other regional accents. The Outsiders have had exposure to both the local and a non-local variety from birth.
Both Insiders and Outsiders participated in two tasks. The first assesses awareness of regional accents, asking them to identify speakers of different regional varieties, and whether they recognized regional variation as a kind of social variation. This is referred to as the Awareness Task. The second task is an ABX discrimination task, contrasting the local accent and an unfamiliar non-local accent. In addition to comparing the results of the two groups, the results of the two tasks are tested for correlations with one another using a mixed effects logistical regression model. This analysis should show whether awareness of regional variation has any influence on performance on the ABX discrimination task for either group.
Experiment 1: Awareness Task
Introduction
In the Awareness Task, children are asked five questions assessing their knowledge of regional variation and the geographical concepts underlying it. No previous study, to my knowledge, has directly asked children to identify the regional accents that they are discriminating, or asked whether they understand regional variation to be a system of social variation.
As part of the Beck (Reference Beck2014) study, parents filled out an extensive questionnaire on the family’s residential and language history, as well as demographic information about the child, such as age, ethnicity, and gender. These factors were tested for correlations with children’s responses on the Awareness Task using a general linear model. Given other findings that older children perform better at identifying and categorizing regional accents (Girard et al. Reference Girard, Floccia and Goslin2008; Floccia et al. Reference Floccia, Butler, Girard and Goslin2009; Kinzler and DeJesus Reference Kinzler and DeJesus2013), increased age was expected to correlate with better performance on this task.
The general expectation was that Outsiders would answer more questions correctly than Insiders, as they have more exposure to regional variation, and may have heard more commentary on regional variation, given that their families are from other dialect areas and may point out differences between their speech and the local community’s. However, it was expected that Insider children at least would recognize the local accent, and be able to say that it was from their hometown.
Methods
Subjects
Sixty-six children (thirty-five female), aged 61 to 77 months, average age 70.4 months, or 5.10, participated in this study. All subjects resided in a town near Philadelphia, USA. For the analyses in the present chapter, children who had two parents from the town in which the study was conducted were compared to children with neither parent born in that town. Only monolinguals were included in the analyses, to avoid confounds with language proficiency. All of the subjects included in this comparison were Caucasian and had no reported hearing or speech problems.
All participants in this study were recruited from two elementary schools in the same town, and had been attending Kindergarten in these schools for approximately five months when the study took place. None of the children was reported to have any contact with the non-local accent used in the study (Southern US English), and none of the Insiders was reported to have regular or prolonged contact with speakers with a different regional accent.
Materials
Subjects were asked five questions to assess their awareness of regional accents. Two questions asking children to identify locations on a map were included to test whether knowledge of geography or relative locations in the country improved understanding regional variation. Children were provided with a colorful map of the United States to use as an aid in answering these two questions.
The audio clips used in the Accent ID questions were single-word clips taken from the set of audio stimuli used in the discrimination experiment. A complete description of those clips and how they were recorded is given in the description of the second task below. The five questions, together with the abbreviated names by which they are referred to in this chapter, are given below:
Can you show where we live (while looking at a map of the United States)? (Map ID Local Region)
Can you show me and name any other places you know (while looking at a map of the United States)? (Map ID Other Region)
Does this person sound like he lives here? (Accent ID Local)
Does this person sound like he lives here? (Accent ID Non-Local)
Can you guess why these two people talk differently? (Explanation)
Procedure
Prior to participating in this study, subjects’ parents were asked to fill out an extensive background questionnaire on the family’s residential and linguistic history. The parents were not interviewed by the experimenter, so the data collected in the questionnaire are subject to parents’ judgments and accuracy in reporting. Despite this, the analysis suggests that the data on exposure collected for the study participants help to explain differences in performance found between these two groups of subjects.
The subjects were tested at their elementary schools, at a table outside the classroom where they were often administered tests or given individualized instruction, making the setting familiar to them. The Awareness Task was consistently administered after the Discrimination Task, in order to avoid biasing their responses to the Discrimination Task. All children were offered a short break before between the two tasks.
Results
Outsider children are more accurate using a map to identify their home and other locations on a map than their Insider peers. However, they are much less proficient than Insiders at identifying local and non-local regional accents. Both groups do approximately equally well at stating that regional accent distinguishes the two sets of speakers heard in the stimuli. This question was expected to be the hardest to answer, as it requires not only knowledge of regional accent as a kind of sociolinguistic variation, but also the ability to abstract that knowledge in order to make a generalized statement about it.
Table 5.1 Comparison of Awareness Task Results, Insiders and Outsiders
| Insiders (n=24) | Outsiders (n=13) | |
|---|---|---|
| Map ID Local | 71% (17) | 85% (11) |
| Map ID Other Region | 61% (15) | 69% (9) |
| Accent ID Local | 74% (18) | 54% (7) |
| Accent ID Non-Local | 53% (13) | 31% (4) |
| Explanation | 36% (9) | 38% (5) |
Responses to the Accent ID Local item were counted as correct if the child indicated in any way that the local speakers were from their hometown; they were not required to explicitly name the town. Thus, responses such as “from here” or “from my town” were accepted as correct answers. The same is true for the Accent ID Non-Local item; any response indicating that the speaker was non-local was counted as correct. For example, some children said the non-local speaker was from another state, from “far away,” or “not from here.” All of these answers were counted as correct.
A general linear model was used to test for correlations between demographic factors, prior exposure to regional variation, and responses to the Awareness Task items. Because there were numerous demographic and exposure factors for which data were collected on the questionnaire, each factor was plotted against total number of correctly answered questions on the Awareness Task. Two factors patterned with performance on the Awareness Task and were thus included in the statistical model: age and Outsider status.
For three Awareness Task questions, Accent ID Local, Accent ID Non-Local, and Explanation, being an Outsider correlated negatively with correctly answering those three questions. Outsider children were less likely than Insider children to identify either accent correctly, or know that the sets of speakers differed by regional accent.
Increased age positively correlated correctly answering Map: ID Local. It also positively correlated with the Explanation item, the most abstract of the questions asked in the task. This may reflect the fact that older children have an easier time generalizing about regional variation than younger children.
Discussion
The results of this task show that Outsiders are at a relative disadvantage as compared to Insiders at identifying regional accents, knowing that accents mark a local/non-local distinction or recognizing regional variation in general.
Table 5.2 Can You Find and Name Where You Live on a Map? (Map ID Local)
| Estimate | Standard error | t value | Pr(>|t|) | |
|---|---|---|---|---|
| Age | 0.02917 | 0.01440 | 2.025 | 0.0471 |
| Outsider | 0.15328 | 0.06564 | 2.335 | 0.0227 |
Table 5.3 Can You Find and Name any Other Place on a Map? (Map ID Non-local)
| Estimate | Standard error | t value | Pr(>|t|) | |
|---|---|---|---|---|
| Age | 0.017686 | 0.016606 | 1.065 | 0.291 |
| Outsider | 0.004921 | 0.075677 | 0.065 | 0.948 |
Table 5.4 Does This Person Sound Like He Lives Here? (Accent: ID Local)
| Estimate | Standard error | t value | Pr(>|t|) | |
|---|---|---|---|---|
| Age | 0.009886 | 0.014865 | 0.665 | 0.50843 |
| Outsider | –0.202895 | 0.067740 | –2.995 | 0.00392 |
Table 5.5 Does This Person Sound Like He Lives Here (Accent: ID Non-local)
| Estimate | Standard error | t value | Pr(>|t|) | |
|---|---|---|---|---|
| Age | 0.01771 | 0.01559 | 1.136 | 0.2602 |
| Outsider | –0.18056 | 0.07103 | –2.542 | 0.0135 |
Table 5.6 Why Do These Two Speakers Talk Differently? (Explanation)
| Estimate | Standard error | t value | Pr(>|t|) | |
|---|---|---|---|---|
| Age | 0.05157 | 0.01347 | 3.829 | 0.000299 |
| Outsider | –0.14694 | 0.06138 | –2.394 | 0.019664 |
In terms of geographical knowledge, Outsiders have a relative advantage over Insiders. Seventy-one percent of Insiders could correctly identify their home on a map, but 85 percent of Outsiders could do so. The reason for this difference perhaps relates to the frequency with which parents from outside the area may talk about or visit other geographical regions, making maps and geography more relevant or salient to Outsider children. These questions were included in this task to assess whether children with better understanding of geography also did better at understanding regional differences in speech. Given that Outsider children did well at identifying geographical locations but poorly at identifying accents and regional variation in general, and the Insiders showed the reverse pattern, it seems unlikely that geographical knowledge is a prerequisite for identifying regional accents.
As hypothesized, both groups of subjects show the greatest relative difficulty answering the Explanation item, which asks them to report that regional accents is the kind of variation heard in the task. It is therefore unsurprising that older children perform better at this task than younger children; their ability to formulate abstract statements should improve with age.
It was expected that increased age would correlate with correctly answering all of the questions on the Awareness Task; however, this was only the case for the Explanation and Map ID Local items. The reason for this could be that the range of participant ages was 61 to 77 months, less than 15 months’ difference between the oldest and youngest participants. However, the lack of consistent correlation between age and awareness across all of the Awareness Task items indicates that age is not the only, or even chief, factor in the development of sociolinguistic competence.
The surprising result was that Outsiders were less accurate than Insiders at identifying the local and non-local accents. Outsiders were expected to have an advantage over the Insiders, given their history of exposure to regional accents. In this study, the Outsiders are less able than Insiders to identify a local regional accent, a non-local regional accent, or state why two speakers with regional accents sound different.
This parallels Kinzler and DeJesus’s (Reference Kinzler and DeJesus2013) finding that 5 and 6 year olds in the Southern United States also have difficulty identifying the local accent and assigning common social stereotypes to those accents, and suggests that hearing two regional accents in the community complicates the process of recognizing the association between accent and geographical region.
The reason for the relative disadvantage of Outsiders may be that their input does not suggest there is only one regional variety. If they hear one regional variety at home and another in the community, they would have reason to believe there are two regional accents used in their hometown, as from the child’s perspective, both accents are used by local speakers. Until they realize that the parents are not natives of the town, Outsiders would have little reason to believe the differences in accent are associated with a speaker’s place of residence.
This finding also appears to support Munson’s (Reference 122Munson2010) prediction that children construct a “library” of speakers, which forms the basis for their representations of sociolinguistic variation. However, the Outsiders have labeled some speakers in the “library” in a way that causes confusion when they have to draw on those representations in order to identify local accents. Many of the Insiders possibly have no other regional accent represented in their exemplar cloud, yet they can rule out the possibility that an unfamiliar accent is local. This finding raises the possibility that entire exemplar categories have to be reassigned labels as sociolinguistic sophistication increases throughout development, and possibly the lifespan, as listeners learn new social categories. How that reassignment of exemplars to newly learned categories proceeds and the interaction of labeling of exemplars on perception remain to be more closely examined.
The results of this task provide further evidence that ability to identify local and non-locals may depend on the kind of exposure children have to regional variation in early childhood. For children who hear a different regional variety in the home, the connection between a speaker’s place of residence and accent may not be transparent. As a result, children who hear multiple regional varieties in their input have less awareness of regional variation than children whose input is predominately in one regional variety. Age plays less of a role in awareness than expected, showing correlations with only the most abstract question asked in this task, but not with identifying regional accents or regional variation.
Experiment 2: ABX Discrimination Task
Introduction
This experiment tests whether Outsiders and Insiders differ in their ability to discriminate between a familiar and unfamiliar regional accent in their native language. Although previous studies have shown that children have difficulty with regional accent discrimination using a categorization task (Girard et al. Reference Girard, Floccia and Goslin2008; Floccia et al. Reference Floccia, Butler, Girard and Goslin2009; Wagner et al. Reference Wagner, Clopper and Pate2013), this study uses an experimental paradigm with which 5 and 6 year olds have been successful. For a detailed discussion of the methodology used in this study, consult Beck (Reference Beck2014). As with the previous task, Outsiders were expected to perform better than Insiders, given their greater exposure to regional variation.
Methods
Participants
The same group of participants who completed the Awareness Task participated in the Discrimination Task. However, some children did not complete the Discrimination Task, and as a result, only data from the ten Outsiders and twenty Insiders who completed this task were included in the analysis.
Materials
Twenty-five stimulus words representing six vowel-quality differences between the Philadelphia and General Southern accents were identified and recorded (Labov et al. Reference Labov, Ash and Boberg2006; Schneider Reference Schneider2008). The six vowel groups are characterized by the vowels in the following words (words in block letters following Wells (Reference Wells1982)): FACE, PRICE, GOAT, GOOSE, peel, and tail. In the peel and tail class, the critical difference between the accents is the reversal of the vowel quality before /l/ in minimal pairs such as heel and hill. In Philadelphia, “peel” is pronounced [pil] and in General Southern [pɪl], whereas for the word hill the pronunciation in Philadelphia is generally [hɪl] and in General Southern [hil]. The same is true for the tail class, but the two vowels that are interchanged are [e] and [ɛ]. For the GOOSE class of words, the /u/ is more fronted in Southern than in Philadelphia, and often preceded by the glide /j/, such that the pronunciation of tune becomes [tjun], for example. The PRICE class has a vowel that is pronounced as a diphthong in Philadelphia, but as a monophthong in Southern [aɪ] versus [a:]. Finally, the GOAT class of words has a vowel quality in Philadelphia of [oʊ] that is considerably more fronted in Southern.
The stimuli for this task were created from recordings of six Caucasian male speakers, three for each regional accent, 25 to 35 years of age, and all lifelong residents of their respective hometowns. The local speakers were all from the same town as the children. The non-local speakers were all from the same town in Northern Louisiana, and speakers of General Southern American English.
In the Discrimination Task, children heard a single word spoken once in each of the two accents (referred to as the A and B tokens). Then a third token of that word was played in one of the two accents (the X token). The children then were asked to match the X token with either the A or B. The order of the trials was pseudo-randomized in one of four orders, ensuring that both Southern and Philadelphia accents were the “X” token 50 percent of the time, and that the matching token for a given word appeared equally as often in both the A and B positions across the entire experiment. This was done in order to minimize effects of order, since it is possible that a short inter-stimulus interval (ISI) between the matching tokens might facilitate matching, whereas a longer ISI might make matching tokens harder.
Procedure
The children were tested on the same day and in the same location as in the Awareness Task. After receiving the children’s oral consent to participate in the experiment, the experimenter began by showing them a PowerPoint presentation with pictures of each of the stimuli words in the experiment. They were asked to name the picture shown on each slide. This ensured that the children knew the stimuli words, but hadn’t heard them pronounced by anyone except themselves immediately prior to completing the task. If the child incorrectly named a picture, or provided a synonym for the target word, they were asked to guess again until they said the target word.
The experiment was also presented on Microsoft PowerPoint. Small pictures of radio speakers represented each token that the child heard. The speaker icon for each token on the PowerPoint slides were labeled as one of A, B, or ?. After listening to all three tokens, the child could either point to one of the icons or say “A” or “B” to indicate his or her response. Responses were recorded by the experimenter, together with any commentary on the accents made during the testing session.
The children were given three warm-up trials that did not include any of the target vowel contrasts, in order to introduce them to the format of the experiment. No feedback was given in the experimental trial, and when giving instructions, no reference was made to matching accents, regions, or any other suggestion of how children should match the speakers. This was done to test whether regional accent was a salient feature by which to match speakers, and not to bias children’s responses.
Results
The Insiders (n=20) averaged 16.7/25 (67 percent) correct matches. The Outsiders (n=10) averaged 16.3/25 (65 percent) correct matches. The averages for the two groups were not significantly different from one another (t= 0.296, p=0.7695 in a two-tailed t-test). Both groups performed above chance (Outsiders t= 30.268, p=2.294 x 10–10 in a one-tailed t-test), Insiders (t= 25.430, p=3.888 x 10–16 in a one-tailed t-test).
Using a mixed effects logistic regression model, the Awareness Task questions were examined for correlation with performance on the Discrimination Task. This model takes into account independent variables selected by the researcher, in this case, the Awareness Task questions and subject age, as well as random effects of variation across subject and items in the task. A binomial link function was used in the model, allowing for data that are not normally distributed, such as the binary responses in the present task, to be used in the model.
As the Tables 5.7 and 5.8 show, only one correlation between Awareness Task items and performance on the Discrimination Task was found for either of these two groups. The Outsiders showed a positive correlation between ability to identify the local regional accent and discrimination between the accents. The Insiders showed no correlations between any Awareness Task items and discrimination ability.
Table 5.7 Correlations with Awareness Task and Discrimination, Insiders
| Estimate | Std. error | Z value | Pr(>|z|) | |
|---|---|---|---|---|
| Map ID Local | 0.13817 | 0.43340 | 0.319 | 0.750 |
| Map ID Other Region | 0.17877 | 0.53813 | 0.332 | 0.740 |
| Accent ID Local | 0.37474 | 0.50739 | 0.739 | 0.460 |
| Accent ID Non-Local | –0.02486 | 0.43191 | –0.058 | 0.954 |
| Explanation | –0.15770 | 0.50258 | –0.314 | 0.754 |
| Age | 0.05043 | 0.06801 | 0.742 | 0.458 |
Table 5.8 Correlations with Awareness Task and Discrimination, Outsiders
| Estimate | Std. error | Z value | Pr(>|z|) | |
|---|---|---|---|---|
| Map ID Local | 0.65919 | 0.53125 | 1.241 | 0.2147 |
| Map ID Other Region | 0.42077 | 0.48382 | 0.870 | 0.3845 |
| Accent ID Local | 0.57306 | 0.27067 | 2.117 | 0.0342 |
| Accent ID Non-Local | –0.15593 | 0.43244 | –0.361 | 0.7184 |
| Explanation | 0.07388 | 0.38086 | 0.194 | 0.8462 |
| Age | 0.06561 | 0.04274 | 1.535 | 0.1247 |
Discussion
In this task, both Outsiders and Insiders performed better than chance at discriminating between regional accents. The performance of both groups indicates that 5- and 6-year-old children can reliably discriminate between regional accents based solely on the acoustic differences between the accents. The average scores of the two groups were not significantly different, showing that exposure to regional variation does not enhance or degrade the ability to discriminate acoustically between regional accents for the participants in this study. This finding is in contrast to some earlier studies (Girard et al. Reference Girard, Floccia and Goslin2008; Floccia et al. Reference Floccia, Butler, Girard and Goslin2009; Wagner et al. Reference Wagner, Clopper and Pate2013), which found that 5 year olds had difficulty discriminating between regional accents.
Subject age shows no correlations with discrimination ability. As in the Awareness Task, one possible explanation is that the range of subject ages is narrow. A second possibility is that this task was relatively easy, and as a result age does not affect performance, whereas in tasks using longer stimuli or more difficult paradigms, increased age provides an advantage.
The one correlation found in the analyses between the Awareness Task items and performance on the Discrimination Task was in the Outsiders group. Outsider children who correctly identified the local regional accent in the Awareness Task correlated with greater accuracy discriminating between regional accents.
Beck (Reference Beck2014) finds other participants in that study also show correlations between awareness and performance using the same discrimination task as described in the present chapter. For the African-American participant group, not included in the present analyses, but described in Beck (Reference Beck2014), the correlations between Awareness Task items and performance on the Discrimination Task depended on which accent was the matching accent in a given trial. Correctly identifying the local regional accent in the Awareness Task correlated positively with correctly answering Discrimination Task items in which the local speakers matched. For the Discrimination Task items in which the non-local speakers matched, there was a correlation with the Awareness Task Explanation item, which asked children to identify regional variation as the reason why the two sets of speakers sounded different. The analysis suggested that the correlations may represent strategies children used to find matching speakers, and that they used different strategies depending on which accent matched in each trial.
It is possible that the correlation between correctly identifying the local accent and correctly matching speakers also represents a strategy used by some Outsider participants to discriminate between the accents in this study. Outsiders who could identify the local accent may have used this heuristic to identify and match local speakers. Remember, however, from the Awareness Task that Outsiders performed poorly as a group at identifying the local regional accent. Thus, this correlation reflects only those few Outsiders who were aware that there was a local accent and could correctly identify it.
Interestingly, although proportionally more of the Insiders than Outsiders could correctly identify a local accent, there was no correlation between that knowledge and performance on the Discrimination Task for the Insider group. There were also no correlations between any other Awareness Task items for the Insiders, suggesting that they were not utilizing social knowledge as a heuristic for discriminating between the accents heard in the Discrimination Task.
General Discussion
From the two tasks described here, the effect of exposure to a non-local regional variety in the home in early childhood becomes apparent. While exposure to multiple regional accents doesn’t affect the ability to discriminate between two regional accents, it does affect children’s awareness of regional accents between the ages of 5 and 6. Outsider children perform worse at identifying accents as local or non-local than their Insider peers, and being an Outsider negatively influences the ability to recognize regional accent as a kind of linguistic variation between speakers.
A possible explanation for the differences in awareness between these two groups is that Outsiders have not understood the association between accents and geography. Outsiders can hear the regional variation in their input, but don’t understand that regional accent corresponds to the speaker’s place of origin, given the distribution of regional accents in their input. Insiders, on the other hand, receive the bulk of their input in a single regional variety, shared by the majority of their community. The correspondence between a familiar accent and the hometown is more straightforward, and therefore more easily identifiable, for Insiders.
Possibly for this same reason, Insider children are also more likely than Outsiders to know that an unfamiliar regional accent is not from their hometown. Insiders’ representation and understanding of the local regional accent may provide a reliable means of contrasting familiar and unfamiliar accents, allowing them to more readily identify unfamiliar accents as non-local.
Foulkes and Docherty (Reference Foulkes and Docherty2006) make a prediction that regional variation may be one of the last kinds of speech variation identified by children, given that its source, the speaker’s place of origin, is not as visually apparent as other social qualities, such as gender. They also predict that awareness of social variation improves with exposure to it, adding to children’s difficulties with identifying it. The present study has shown that despite the lack of transparency, many 5 year olds have no trouble identifying regional variation in speech, or recognizing regional variation. Ironically, and in contradiction to Foulkes and Docherty’s prediction, the children with the least amount of experience with contrasting regional varieties perform best at identifying it. The hypothesized opacity of the source of regional variation only holds for Outsiders, who must sort out the source of multiple varieties, indicating that exposure does not always entail awareness of a variety.
Munson’s (Reference 122Munson2010) “library of speakers” hypothesis of sociolinguistic category formation could predict these findings. A category for a regional accent is founded on the association of the speech tokens from a familiar individual with the labels for social qualities, in this case place of origin, of that speaker. Outsiders may have labeled the speech of their parents as local. When asked to identify local speech, they access the categories labeled as local, only to find multiple, dissimilar kinds of speech. This creates a conflict when children are asked to identify other local speakers based on accent, as there is only a partial match. Insiders do not share this conflict and their library of speakers supports an identification of regional accents as local or non-local.
Presumably, as Outsiders learn about their parents’ place of origin, they re-label their representations of the parents’ speech, and they are able to identify regional varieties. However, this raises the question of whether representations of social categories develop independently from linguistic categories, and what role labels play in defining which categories develop. Based on the findings in the present chapter, it appears that at this stage of development there can be a disconnect between the development of social and linguistic categories under certain circumstances. Because no other social knowledge beyond labels for local and non-local speakers was addressed in this study, this remains a question for future investigations.
From the differences in awareness between the two groups, it is clear that input plays a role in forming categories for regional variation in speech. For 5 and 6 year olds in this particular context, less regional variation in the input helps to create a representation of the local accent, associated with the child’s hometown. Exposure to multiple regional varieties obfuscates the link between geographical location and accent, which doesn’t impede discrimination, but complicates identification of regional accents. This raises the question of how labels for linguistic varieties and social categories develop and influence linguistic representations across the course of development – a question for future studies.
Conclusion
This comparison of 5- and 6-year-old children with different histories of exposure to regional variation shows that they can reliably discriminate between a familiar and an unfamiliar regional accent at this age, regardless of their linguistic background. However, their awareness of regional variation depends on their exposure to regional accents. Children who hear multiple regional accents at home and in their community have a harder time identifying the local variety. They also have more difficulty explaining that regional variation identifies speakers as being from different places. This is likely to be because their experience with regional varieties at home has not permitted them to associate accent with a speaker’s place of origin. Despite this, they are able to discriminate between regional accents of their native language as well as children without exposure to multiple regional varieties. Children who hear predominately one regional accent in their input perform better at associating the local accent with their hometown, although this does not provide any advantage in discriminating between regional accents. The lack of effect of accent identification on discrimination suggests that the 5 year olds in this study were not heavily influenced by social knowledge in perception of regional accents, or at least in tasks that don’t explicitly reference the connection between the two. More work is needed to better understand when and how social knowledge influences speech perception in children across different social contexts, and the relation between the formation of linguistic and social categories over the course of language acquisition.
