A body of linguistic research on ethnic identification dating back nearly 70 years has found that U.S. listeners are generally rather accurate (70–100%) at distinguishing black speakers from white speakers (cf. Thomas & Reaser, 2004). Recent studies have attempted to unpack the role of suprasegmentals in ethnic identification. Thomas and Reaser (2004) found that listeners were equally accurate at ethnic identification for monotonized and unmodified stimuli, suggesting that listeners do not rely solely on F0 cues in ethnic identification. They also discovered that some cues relevant to pitch accents are recoverable even from monotonized stimuli (i.e., amplitude, duration, and segmental qualities), so it is conceivable that pitch accents may aid identification even in monotonized stimuli. Holliday and Jaggers (2015) examined listeners’ ability to identify the ethnicity of U.S. politicians based on single-word stimuli, in order to assess the effects of voice quality on listener judgments. Building on some of the earlier findings of Purnell et al. (1999), Holliday and Jaggers found that several suprasegmental variables, including jitter and harmonics-to-noise ratio, influenced ethnic identification, though they note that a combination of multiple speakers and contexts may cause challenges in isolating speaker-specific variables influencing ethnic identification. For this reason, in the present study, we attempt to control for the effect of speaker-specific voice quality variation and more carefully isolate the prosodic variables that may affect ethnic identification by employing stimuli from a single speaker.

This study focuses on one particular type of intonational variable as a starting point for understanding how listeners may react to ethnically-linked suprasegmental features, using methods based in the auto-segmental/metrical (AM) intonational framework (Pierrehumbert, 1980). Essential to the AM theory is the idea that movements in fundamental frequency (F0), the main correlate of what we perceive as pitch, result from an underlying sequence of tones that determine their structure. In the AM theory, these tones are either low or high, and all movements of the pitch contour are composed of a series of low and high sequences. The labeling system for intonational phenomena that is based on the AM theory is called the Tones and Breaks Index system (ToBI). Each language, and indeed a number of dialects and varieties, have distinct ToBI systems that reflect the variety’s intonational specifications (Beckman & Ayers-Elam, 1997). The ToBI system for Mainstream American English (MAE), originally developed by Beckman and Ayers-Elam (1997) and based on the findings of Pierrehumbert (1980), is the only ToBI system generally in use for examining variation within American English. MAE-ToBI has previously been used for descriptions of Jewish English (Burdin, 2015), Appalachian English (Reed, 2016), as well as African American Language (AAL) (Holliday, 2016; Jun & Foreman, 1996; McLarty, 2018).3

MAE-ToBI contains two types of pitch movements: pitch accents, which occur on some stressed syllables, and edge tones, which occur at phrase boundaries. The current study focuses only on the movement of pitch accents, though it is important to note that we also tested for the perceptual effects of edge tones. This study focuses on the difference between two types of pitch accents in MAE: a simple high tone, labeled as H*, and a fall-rise, labeled as L+H*. Though other types of pitch accents exist, H* and L+H* are by far the most common pitch accents in most varieties of U.S. English, including AAL (Burdin et al., 2018).

Earlier studies have shown that pitch accents are perceptually salient for listeners and that naïve listeners can be trained to identify them quickly (McLarty, Vaughn, & Kendall, 2017; Thomas, 2011). Especially relevant to the current study, studies such as Loman (1975), Holliday (2016), and McLarty (2018) have found that MAE and AAL exhibit different rates and contexts of use for H* versus L+H*. In particular, Loman (1975) and McLarty (2018) each found that L+H* pitch accents are more common in some varieties of AAL.

Recent work by Holliday (2016), Burdin (2015), and Reed (2016) inter alia has also found that a greater rate of use of the L+H* pitch accent may also be a resource in production for performance of different types of ethnic identity. For example, Holliday (2016) recorded 25 men (age 18–32) with one black parent and one white parent in Washington, DC to examine their rates of use of different types of pitch accents in ethnic identity performance. The participants were recorded in casual peer dyad conversations, and the analysis of their intonational patterns was taken from these recordings. A sociolinguistic interview also elicited ideologies about race and self-identifications. The participants who identified more as black, as opposed to multiracial or mixed, were more likely to use a greater quantity of L+H* accents than H* accents. This finding supports Loman’s (1975) and McLarty’s (2018) findings that L+H* is more prevalent in AAL than in MAE; also relevant for the current study, this finding demonstrates that speakers’ production of intonational variation is gradient in terms of frequency.

This study’s focus on intonational and suprasegmental variation presents an opportunity to address questions about phonetic detail and social meaning. One of the most significant recent advances in sociolinguistic theory has been the advent of sociophonetics (e.g., Foulkes & Docherty, 2006), with the notion that paying attention to phonetic detail can enrich our understanding of sociolinguistic variation—especially for variables that have traditionally been considered binary or categorical (Labov, Ash, & Boberg, 2006).

Although the binary treatment of phonetic variables reveals structure in sociolinguistic variation, a sociophonetically informed approach recognizes that the distribution of these variables’ continuous acoustic correlates is not always compatible with discrete categorization. For example, Jacewicz and Fox (2018) use a continuous measure of /aɪ/ monophthongization (trajectory length) to analyze preadolescent Appalachian English speakers. They find that these preadolescents produce variants that are more diphthongal than Appalachian adults but less diphthongal than central Ohio adults. The authors’ continuous approach pays off, in other words, by revealing finer-grained phonetic variation than is suggested by the monophthong/diphthong binary.

At the same time as research on production in sociolinguistics has increasingly turned to phonetic detail, the role of such detail remains under-theorized and under-investigated in the study of social meaning. To that end, Podesva (2011) proposes a framework for salience in sociolinguistic variation that reconciles the roles of frequency and phonetic detail. He hypothesizes that salience takes one of two linguistic forms: ‘categorial salience’ (frequent productions of a marked feature are salient) and ‘phonetic salience’ (more extreme productions are salient). In particular, with respect to phonetic salience, Podesva argues that a more extreme production signals a stronger social meaning: “If an axis of phonetic variation indexes a particular social meaning, then outliers on that axis can be understood as the strongest indicators of meaning” (pp. 254, emphasis added).

These predictions about categorial and phonetic salience have been supported by a handful of findings on the distribution and social meaning of intonational variation in production. For example, Podesva (2011) found that one speaker constructed a ‘life of the party’ persona by using acoustically extreme falling contours to imbue partying-related narrative elements with extra emphasis. Burdin et al.’s (2018) comparison of L+H* pitch accents in Jewish English, AAL, and Appalachian English showed that both categorical and continuous properties of pitch accents are sites for sociolinguistic differentiation. The authors found that, across communities, L+H* pitch accents differed in both rates of use and acoustic properties (e.g., peak F0, peak offset).

As far as we are aware, only a handful of perceptual studies have investigated how social meanings are affected by phonetic detail. Plichta and Preston (2005) presented U.S. listeners with a synthesized continuum from monophthongal to diphthongal /aɪ/ and asked listeners to identify the speaker’s geographic origin along an axis running from the U.S. north to the U.S. south. Listeners not only associated monophthongal /aɪ/ with the south and diphthongal /aɪ/ with the north, they also placed successive continuum steps linearly along the north–south axis. D’Onofrio (2018) found that labeling a speaker as a ‘Business Professional’ or ‘Valley Girl’ cued U.S. listeners to classify more ambiguous [æ~ɑ] tokens as /æ/, with ‘Valley Girl’ being especially associated with backer /æ/, compared to a ‘Chicago Bears Fan’ label or no label at all. In an experiment with Californian listeners, Villarreal (2016) found significant correlations between speakers’ raising of /æ/ in bad and glass and listeners’ ratings on the scales ‘accented,’ ‘doesn’t speak like me,’ ‘unfamiliar,’ and ‘not Californian.’ Foulkes, Docherty, Khattab, and Yaeger-Dror (2010) found that listeners’ identification of Tyneside children’s gender was affected by two continuous measures (amplitude and F0) as well as several categorical measures; however, the authors also report significant correlations between amplitude and F0 in stimuli, suggesting potential issues with collinearity in the modeling procedure. In terms of voice quality, Szakay (2012) found that in New Zealand, ethnic identification was affected by several continuous voice quality measures; speakers with higher mean H1–H2 (a measure of creakiness) were likelier to be identified as Māori.

The present study seeks to expand our understanding of the relationship between phonetic detail and social meaning by investigating this relationship through the lens of intonational variation. Building on Podesva (2011), we hypothesize that the social meanings of continuous variables will exhibit what we call incrementality: a monotonic relationship between the variable’s phonetic extremeness and the strength of the social meaning it elicits in perceivers.4 We focus on pitch accents, which are ideally suited to this question as they vary both in category (e.g., H* versus L+H*) and phonetic shape (e.g., peak offset, rise slope).

This task used the ‘open-guise technique’ (OGT) (Soukup, 2013); as in the more common matched-guise technique (MGT), OGTs offer insight into the social meanings of a focal feature, variety, or language, by comparing listeners’ reactions to stimuli differing only by the focal linguistic structure (e.g., Campbell-Kibler, 2009). Unlike the OGT, the MGT axiomatically hinges on listeners’ belief that they are listening to different speakers (Giles & Billings, 2004; Purnell et al., 1999); otherwise, it is assumed that listeners will not differentiate guises on personal characteristics that are considered intrapersonally stable qualities (e.g., intelligence). In OGTs, by contrast, listeners are openly informed that they are hearing the same speaker in different guises. Soukup (2013) shows that listeners responded differently to standard versus dialectal Austrian German guises in both OGT and MGT settings (with the OGT actually yielding stronger effects for some scales), undermining MGTs’ key assumption about different speakers.

In the present study, we assumed that listeners (all from the United States) were highly likely to recognize our stimulus speaker, President Barack Obama, necessitating an OGT rather than MGT approach. We openly informed our listeners, “This study is designed to test how people respond to different speech excerpts from the same speaker.” In so doing, we rejected the type of instrumental task framing often used in MGTs, such as evaluating prospective radio newsreaders (Labov et al., 2011; Villarreal, 2018). By contrast, Obama represented an ideal stimulus speaker to test our hypotheses, as his ability to command both AAL and MAE is well-known by the general public (Alim & Smitherman, 2012); our use of the OGT took advantage of this awareness. In the discussion, we make recommendations about the appropriateness of OGT versus MGT.

The 120 stimuli were based on excerpts of President Barack Obama’s spontaneous speech from two different 2016 television interviews with Gayle King, a black broadcast journalist who co-anchors the CBS This Morning news program (Kaplan, 2016). Each stimulus excerpt was based on a single Intonational Phrase (IP) unit, ranging from 0.4 to 2.3 seconds in duration (median 0.9 seconds). Following Pierrehumbert and Hirschberg (1990) as well as subsequent works utilizing their methods, we identified IPs through looking for pausing and phrase-final lengthening, as well as the presence of characteristic boundary tones and smaller intermediate phrase units contained within the IPs. We attempted to select short phrases that were fairly semantically bland to avoid overly tilting responses in one direction, though it is impossible to completely control for content in listening tasks.

Sixty excerpts were selected: 20 critical excerpts and 40 filler excerpts. Ten critical excerpts were ‘H* phrases,’ which contained between 1–3 H* pitch accents and 0 L+H* accents; and ten were ‘L+H* phrases,’ which contained between 1–3 L+H* pitch accents and 0–2 H* accents. This imbalanced definition of H* versus L+H* phrases was necessary since L+H* pitch accents are relatively rarer, even in AAL (Burdin, Holliday, & Reed, 2018), so it was not possible to find enough excerpts that contained only L+H* accents. Filler excerpts contained 1–3 H* pitch accents and 0 L+H* accents.

In choosing excerpts, we intentionally sacrificed a degree of experimental control for the sake of presenting listeners with natural, spontaneously produced stimuli rather than unnatural, lab-like speech. The benefit of using spontaneous stimuli is that it more closely models real-world perception conditions, as listeners perceive spontaneous and read speech (including oratory) differently (Campbell-Kibler, 2009; Holliday & Jaggers, 2015). The drawback is that the distribution of H* versus L+H* pitch accents across stimuli prevented us from addressing Podesva’s (2011) hypothesis about categorial salience; at the same time, total experimental control over stimuli is impossible to obtain, as features co-occurring in the stimuli can always shape interpretation of features of interest (Leach, Watson, & Gnevsheva, 2016), including propositional content (Campbell-Kibler, 2009). (We explore this issue further in a post hoc analysis of L+H* phrases.)

The critical stimuli were created by manipulating critical excerpts to four manipulation steps, with the original excerpt as Step 1. Steps 2, 3, and 4 were created by making pitch accents’ F0 minima and maxima successively more extreme. With each manipulation step, H* and L+H* maxima were increased by a semitone, and L+H* minima were decreased by a half-semitone. For example, the H* pitch accent in the top panel of Figure 1 has an F0 maximum at 118.3 Hz in step 2 and 125.2 Hz in step 3, a one-semitone difference; the L+H* pitch accent in the bottom panel has an F0 minimum at 101.6 Hz in step 1 and 99.1 Hz, a half-semitone difference. (In some cases, it was not possible to make the manipulations exactly one or one-half semitone.) We based F0 manipulations on semitones rather than constant magnitudes because semitones are psychoacoustically comparable regardless of the pitch accent’s initial F0 (e.g., the difference between 100 and 105 Hz sounds much larger than the difference between 200 and 205 Hz). The first author created stimuli by hand using the Manipulation utility in Praat (Boersma & Weenink, 2015). Both authors listened to all manipulated critical stimuli and confirmed that they sounded natural.

Filler stimuli were created by modifying the final syllable of filler excerpts to include percepts of creaky voice: low F0 and damped pulses (Keating, Garellek, & Kreiman, 2015). A Praat script modified alternating cycles of the final syllable by lengthening their duration and lowering their amplitude. As with critical stimuli, both authors listened to all manipulated filler stimuli and confirmed that they sounded natural.

The task was administered via an online survey hosted by Qualtrics. In each of 120 randomly ordered trials, listeners heard a single stimulus auto-play twice and responded to the question “How black or white does Obama sound here?” on a continuous unit-less slider bar with “very black” and “very white” on opposite poles. As the recognizability of President Obama’s voice would have likely rendered ineffective the type of instrumental task framing often used in MGTs (e.g., rating prospective radio newsreaders, as in Labov et al., 2011), we eschewed such framing; we instead informed listeners, “This study is designed to test how people respond to different speech excerpts from the same speaker.” Listeners then completed a demographic questionnaire and were invited to comment on the task (see Appendix A).

The survey was distributed via social network sampling in May 2017, with a raffle incentive for one randomly selected listener to win an Amazon.com gift card. The listener sample for analysis contains 79 American English-speaking listeners who self-identified as black and/or white. Of these listeners, 24% self-identified as black and 77% as white (one listener identified as both); 65% identified as female and 35% as male. The majority of listeners also self-identified as politically liberal and indicated that they overwhelmingly approved of Obama’s presidency; in particular, on a 1–7 scale (where 7 indicated “very liberal” and “strongly approve of Obama”), the median rating was 6 on both scales, and 91% of listeners rated 5 or above on both scales. In this respect, the listener sample is not representative of the United States voting population; however, our intent was not to survey a sample spanning the political spectrum but rather to determine how some listeners judge ethnicity based on intonational and voice quality variation (we return to this point in the Discussion).

As mentioned above, both authors listened to all stimuli and confirmed that they sounded natural. As a further check on stimulus naturalness, we coded listeners’ responses to the final two questionnaire items: “How did the clips sound to you?” and “Do you have any other comments on the clips or on the survey?” Based on listeners’ responses to these questions, the second author developed eight true-or-false codes that described sentiments listeners expressed in their responses and coded responses accordingly (with a single response capable of being coded “true” in multiple categories). For example, 21% of listeners reported something amiss with the quality of the clips (although numerous listeners commented positively about the clips’ quality). More information about these codes, including examples, can be found in Appendix B. As we discuss below, however, none of these codes significantly improved our model of intonation results, so we did not find evidence that they impacted listeners’ perceptions of the speaker’s blackness.

Slider-bar positions were converted to real numbers between 0 (“very white”) and 100 (“very black”) and standardized by listener to control for variable usage of the continuous slider bar. All results are reported in unit-less standard deviations (i.e., z-scores); the average listener’s standard deviation was 16.6, so a difference of 1 standard deviation can be interpreted as a difference of roughly one-sixth of the length of the slider bar for the average listener.

Our task was specifically designed to address the first two research questions, about the role of pitch accents and phonetic incrementality in affecting listener judgments of ethnicity; we first present the analysis of intonation features. We then describe a post hoc analysis of voice quality characteristics that addressed the third research question, about the role of other voice quality features in affecting listener judgments of ethnicity.

The model indicated no main effect of phrase type on listener ratings of blackness, indicating that pitch accent alone did not trigger different blackness ratings. Figure 2 shows this result, with results for H* stimuli in the left panel and L+H* stimuli in the right panel. As is evident in this figure, listener ratings of blackness were remarkably similar for the H* phrases at each step, though the L+H* phrases showed greater differences between manipulation step. There was also no main effect of manipulation step on listener ratings of blackness.

Though the main effect of phrase type failed to reach significance, the model indicated a significant interaction between phrase type and manipulation step, with more extreme L+H* phrases rated as sounding blacker than less extreme L+H* phrases, and no perceived blackness difference for H* phrases regardless of step. Figure 3 presents these results, with each panel representing a manipulation step. This figure indicates that listeners appear to interpret the more phonetically extreme L+H* realizations (greater difference between F0 minimum and F0 maximum within a L+H* pitch accent) as blacker, but this is not the case for the more extreme H* realizations (which only had higher F0 maxima).

This model also implies that listener judgments of blackness are affected by more than just pitch accent type and phonetic shape. As mentioned above, these results must be interpreted with caution, especially in light of the fact that the model’s fixed-effects predictor structure accounted for less than 1% of the variance in ratings, while random effects—the effect of individual excerpts—accounted for 11.3% of the variance. In other words, listeners were much more attuned to features varying by excerpt, such as segmental, semantic, pragmatic, or voice quality characteristics, than the type and phonetic shape of pitch accents. However, this small effect size may represent an inherent challenge to studies of prosody, since the highly nested nature of such variables causes them to be difficult to isolate from one another. Despite this challenge, the finding of a significant difference here may be a step in the direction of discovering how these variables may operate both independently and together. The small effect size of this intonation effect motivated the post hoc analysis of voice quality features.

Summaries of fixed effects for the voice quality models are in Appendix C. The voice quality model for H* critical trials revealed that few voice quality measures significantly affected listener perceptions of blackness: phrase speech rate and the interaction of peak delay and HNR. Phrase speech rate (seconds per syllable) had a positive effect on listener perceptions of blackness, with slower phrases rated blacker. While the model returned positive estimates for the effects of peak delay and HNR, neither of these main effects reached significance. Rather, the effect of peak delay on listener perceptions of blackness was constrained by the phrase’s HNR. As Figure 4 shows, phrases with longer peak delay were rated blacker, but only if HNR was sufficiently high. This co-patterning of variables suggests that listeners may be attuning to a threshold of combined characteristics in order to make judgments, particularly in the absence of ethnolinguistically-salient intonational differences such as the L+H* pitch accent. Together, the roles of peak delay and phrase speech rate suggest a possible salience effect, as both relate to vowel duration; conceivably, longer vowels (which co-pattern with slower speech rates) with longer intonation rises can better carry indexes of social meaning.7

As with the H* model, few predictors reached significance in the L+H* model—including just one voice quality measure, jitter. Among L+H* stimuli, phrases with less jitter were rated blacker, suggesting that listeners are sensitive to the interaction of F0 movement and local periodic perturbations. Notably, the measures affecting listener perceptions of blackness did not overlap for H* versus L+H* phrases; the jitter term in the H* model, and the phrase speech rate & peak delay × HNR terms in the L+H* model, did not even approach significance. This finding provides additional evidence that listeners may respond to different intonation and voice quality cues in phrases containing L+H* pitch accents than those not containing L+H* pitch accents. As L+H* accents are far less common than H* accents, it is possible that L+H* accents cue listeners to adjust their expectations as to markers of ethnic identification.

In addition, manipulation step was significant in the L+H* voice quality model (manipulation steps 3 and 4 were rated blacker than steps 1 and 2) but not the H* voice quality model. This finding corroborates the generalization that the percept of blackness is subject to phonetic incrementality only with respect to the more socially marked L+H* pitch accent. However, this finding is tempered by the fact that the fixed effects in the L+H* model accounted for less than 3% of the variance, as compared with 10.2% for the fixed effects in the H* model (Table 2); in other words, there remain properties of the stimuli with L+H* accents that listeners are reacting to, above and beyond the intonational and voice quality features that previous studies have suggested are implicated in differentiating AAL from MAE.

In short, the voice quality analysis found that listeners relied on multiple acoustic cues—beyond those pertaining to pitch accents’ type or phonetic shape—in making judgments of perceived blackness; crucially, in the presence of an L+H* pitch accent listeners not only relied on different voice quality cues than in the absence of one, but they apparently relied to a much greater degree on cues other than those relating to intonation or voice quality. This finding suggests a fundamental difference in how listeners judge phrases in the presence of an L+H* pitch accent, although this is an open question for future study. More broadly, this finding further supports the claim that understanding the interrelated nature of prosodic variables is a necessary part of their description.

This study’s results show that in a perception task, listeners appear to be sensitive not only to the phonological category of pitch accents, but also their phonetic realization, as listeners appear to be sensitive to increasingly extreme manipulations of F0 within a single pitch accent type. In the traditional AM model of intonational phonology, pitch accent and edge tones have largely been binned into discrete categories, with meaning presumed to be attached to those categories and their combinations (Pierrehumbert & Hirschberg, 1990). The results presented here provide further motivation for considering intonational variation on a phonetic as well as a phonological level. This study also provides further motivation for the development of ethnolinguistic variety-specific ToBI models as well as phonetic methods for studying intonational variation cross-dialectically. While we have employed the MAE-ToBI conventions (Beckman & Ayers-Elam, 1997) in this study, the nature of the intonational system of AAL has not yet been fully described (McLarty, 2018; Thomas, 2015). As the current study’s results provide evidence that listeners are sensitive to differences in the realization of F0 and timing of the L+H* pitch accent, future studies should examine whether the tonal inventory of AAL differs from MAE, as this could be one element that triggers the observed differences in listener judgments.

Relatedly, as much of the work on prosody has focused on the meaning of intonational contours in an imagined Standard American English as opposed to in specific varieties, it is clear that much more work is needed on both variation in speaker production and listener perception of contour meaning. Though the current study did not reveal differences in perception of ‘sounding black’ conditioned by listener demographics, future work should explore how such perceptions could potentially be affected by listeners with different backgrounds and sociolinguistic experiences.

This point about the role of demographics is especially relevant because (as mentioned above) the listener sample was overwhelmingly liberal and approving of Obama’s presidency, more so than the US population at large. While this is not an issue for the present study—our aim was not to achieve political representativeness but rather to ascertain how intonational variation affected perceptions of blackness within a population of US listeners—it does contextualize the results. Theoretical frameworks that take as primary the role of experience in forming linguistic representations (e.g., Exemplar Theory, Pierrehumbert, 2016) would take the standpoint that listeners who are more inclined to listen to President Obama would have a greater opportunity to hear him in multiple situations, thus facilitating their awareness of his style-shifting; it is thus conceivable that the small-sized effects uncovered in this study would not reach significance in a sample more representative of the US political spectrum. This is a question open for future work to address.

In their 2004 study and summary of the body of research on ethnic identification of white and black speakers and the U.S., Thomas and Reaser reveal gaps in our knowledge about what triggers judgments of speakers as ‘black’ or ‘white.’ Most ethnic identification studies have focused on segmental features, at least in part due to the fact that so little is known about how non-standard varieties of American English employ intonational variation, though it is the case that such studies on prosodic variables have been carried out outside the U.S. (Szakay, 2012; Todd, 2002). While a number of segmental features, such as vowel quality, have been identified as important in triggering listener judgments, researchers still know relatively little about how suprasegmental features may contribute to these judgments. Dating back to the 1970s, researchers such as Tarone (1973) and Loman (1975) have suspected that suprasegmental features played a serious role in triggering these judgments, though few studies have been able to isolate the specific intonational and suprasegmental features involved. The results of the current study, especially those related to the fact that speakers are able to provide consistent judgments of how a speaker whose race is known to them adheres to their ideologies about what it means to ‘sound black,’ provide evidence that it may be possible to isolate the variables of interest using a single-speaker model. This has the advantage of eliminating other types of variation that are inherent in studies with multiple speakers, for whom it is impossible to control every level of linguistic variation, which may be important especially in light of our findings on the effects of voice quality.

This study also builds on the findings of Purnell et al. (1999) as well as Thomas and Reaser (2004), and Holliday and Jaggers (2015) by providing further evidence that a number of voice quality features, including jitter, HNR, and speech rate may be involved in triggering ethnicity judgments. The pattern that we observed wherein there appear to be important interactions of intonational and voice quality features obviates the need for more controlled studies that simultaneously focus on a number of suprasegmental features. Listeners appear not only to be sensitive to both intonational and voice quality features but also the ways in which they combine to create sociolinguistic meaning.

It is worth reiterating here the small effect size that we found in our intonation model, in which fixed effects accounted for just 1% of the variance in listener ratings of blackness. Some readers may interpret this small effect size and the proximity of the sole significant intonational model term’s p value (0.0434) to our predetermined α level (0.05) as casting doubt upon the generality of the result. Although this effect size is modest, it is not without precedent in studies of sociolinguistic perception. Clopper (2010, p. 212), describing Clopper and Pisoni’s (2007) study of free classification of regional dialects of American English, notes “grouping accuracy was still rather poor overall, which may indicate attention to talker-specific differences instead of dialect-specific variation.” This greater attention to talker-specific differences parallels our finding that random effects accounted for a much greater percentage of the intonation model’s variance. Likewise, Villarreal (2018) found that out of 12 ratings scales, a vocalic guise manipulation yielded only three significant differences, compared to eight significant differences for both speaker region and speaker gender and eleven significant differences for speaker ethnicity. In other words, while the effect revealed by the intonation model is modest, it is possible that this is a general property of phonetic guise manipulations, as well as an artifact of the interconnected nature of suprasegmental features in general and the resulting challenges in isolating them from one another.

These findings support the notion that listeners attend to phonetic detail in constructing social meanings of sociophonetic variation, given that listener ratings of blackness for L+H* increased stepwise as L+H* pitch accents became more phonetically extreme. In other words, there is some evidence that listeners map continuous social meanings to continuous variation, supporting our incrementality hypothesis; contra Podesva’s (2011) phonetic salience hypothesis, these findings suggest that greater social meanings are not only attached to phonetic outliers, but also to phonetically intermediate realizations of L+H* pitch accents. This research also sheds light on how phonetic salience works in context. While the intonation analysis found a jump between manipulation steps 2 and 3 in listener ratings of blackness for L+H* phrases, the analysis of the L+H* voice quality model’s random effects found considerable differences in step 1 ratings across L+H* phrases. That is, for some stimuli smaller differences in intonation were sufficient to trigger higher listener ratings of blackness; for others listener ratings of blackness only increased with larger differences in intonation. Thus, just as context shapes the social meaning of a variant’s presence or absence (Campbell-Kibler, 2009; Gumperz, 1982; Leach et al., 2016; Pharao, Maegaard, Møller, & Kristiansen, 2014), context also shapes the way that phonetic detail affects social meanings.

These findings expand our understanding of methods for probing language attitudes, countering the received wisdom in MGT research that these tasks only work if listeners believe they are judging different speakers (Giles & Billings, 2004). This work expands on the findings of Soukup (2013) in demonstrating additional support for the OGT: Listeners were aware that they were hearing the same speaker, but the guise manipulation nevertheless yielded a difference in listener responses. Soukup (2013) finds that the OGT yielded larger effects than the MGT on ‘superiority’ scales (Zahn & Hopper, 1985), while the MGT yielded larger effects on ‘social attractiveness’ scales. However, her comparison did not address socio-indexical traits like ethnicity that fall outside the superiority-versus-social-attractiveness rubric, but which nevertheless form an important part of listeners’ awareness of language variation (e.g., Hay & Drager, 2010; Koops, Gentry, & Pantos, 2008; Niedzielski, 1999). Although it is impossible to determine how the results of this study would compare to a hypothetical companion MGT (as the MGT simply wouldn’t work with such a recognizable stimulus speaker)—and the small effect size we found suggests that a hypothetical companion MGT could yield larger effects—the present study indicates that a socio-indexical trait, ethnicity, can also work in an OGT context.

Moreover, whereas stimulus speakers in typical MGTs are anonymous to listeners, representing blank attitudinal canvases save for small bits of contextual information provided via stimulus text and/or explicit labels, listeners in this study likely had salient prior impressions of President Obama and his racialized speech. The finding that the guise manipulation affected listener perceptions of Obama’s blackness is even more persuasive against that backdrop. Indeed, among the qualitative questionnaire codes that failed to significantly improve the model was ObamaIsBlack (see Appendix B); that is, we found no evidence that listener perceptions of blackness were affected by whether listeners found it difficult to rate Obama as ‘sounding white.’

Although the OGT worked in the present study, we caution readers against the assumption that the OGT will necessarily apply to any context, feature, or trait. First, while both Soukup’s study and the present study intentionally violated the assumption that listeners should believe they are judging different speakers, in both studies listeners were not told which feature was manipulated; we argue that this remains an important element of methodological opacity in speaker evaluation tasks. It is likely that doing so would produce rather different results than if listeners are not informed, especially for those few sociolinguistic variables that attract public commentary. Indeed, only 20% of the listeners in the current study reported that they could detect the guise manipulation (DetectManip, Appendix B), and this failed to significantly improve the model; this is helped by the fact that, aside from high rising terminal (Tyler, 2015), intonational variation is generally not a subject of public commentary in American English.

Second, we argue that there remain contexts in which it is important to conceal the fact that the same speaker is behind both or all guises. While the majority of speaker evaluation tasks involve cognitive and/or affective responses, we predict that tasks involving behavioral responses (e.g., making a hiring decision) are likelier to hinge on listeners believing they are hearing different speakers. For example, if the landlords in Purnell et al. (1999) knew they were hearing John Baugh in multiple guises, they might have been on their ‘best behavior’ to avoid prosecution under the Fair Housing Act.

Third, we argue that the use of an OGT rather than MGT approach must be justified by a plausible style-shifting context. For example, this task relied on listeners’ awareness of President Obama’s style-shifting to sound more black in some contexts and less black in others (Alim & Smitherman, 2012); as mentioned above, it is conceivable that listeners’ awareness in this respect was facilitated by their generally positive attitude toward Obama’s presidency making them more likely to hear Obama’s public speaking. In a similar justification of a plausible style-shifting context, Soukup (2013) relied on her observation that speakers routinely shift between standard and dialectal Austrian German in stylistic practice. If a speaker evaluation task involves styles that do not coexist in stylistic practice ‘in the wild’ (e.g., the same speaker commanding both an L1 and an L2 accent), the OGT is not likely to work.

Caveats about the OGT notwithstanding, it is clear that traditional approaches to linguistic perception do not give listeners enough credit for being aware of style-shifting; indeed, explicit public awareness of style-shifting (e.g., Meraji, 2013) indicates that listeners may be willing to accept reacting to the same speaker using different features, styles, or languages. Future research should explore the extent to which style-shifting itself, not just the individual styles involved in shifting, affects listeners’ judgments of speakers.