Sociolinguistics has incorporated methods from phonetics enthusiastically over the past thirty years. However, while it has examined speech production extensively, studies of perception have lagged far behind. This situation stands in contrast to the practice in phonetics, in which studies of production and perception have an equal status. Nevertheless, perception experiments have gained a foothold in sociolinguistics and several lines of research have emerged.
The first of these lines involves identification of demographic features of speakers. One group of studies has investigated how accurately listeners can identify speakers’ regional dialects. These studies include studies in which listeners try to identify the dialect and those in which listeners rate how typical of a particular dialect a sample of speech is. Another group of studies has examined whether listeners can identify the ethnicity of speakers. Most, but not all, focus on identifications as African-American vs. White. Several have experimentally tested for what cues listeners use to make ethnic identifications. A third group of studies has examined the ability of listeners to name the socioeconomic class of speakers.
The second line of research has looked at how stereotypes about speakers affect listeners’ perceptions. Work on this topic is limited, but it has shown that information about a speaker’s sex and geographical origin can affect what listeners think they hear.
The third line involves research into phonological mergers and splits, most of which has focused on mergers. Some of this research has tested whether listeners can distinguish sounds that are nearly merged in their own or other dialects. There is also some discussion of whether sounds can be merged in perception but not in production. Investigation of splits involves whether listeners identify certain allophones with other allophones.
The fourth line involves differences in how speakers of different dialects categorize phones. One sort of study has examined differences in the perceptual boundaries between sounds in different dialects. Other types of studies investigate the cues used by different listeners to identify sounds, differences in goodness judgments of synthetic stimuli, and the ability of listeners to identify sounds from their own and other dialects.
The final line includes studies on how listeners assess personal traits of speakers in order to determine listeners’ attitudes toward particular groups. A large number of subjective reaction studies, in which listeners rate samples of speech for various personality features, have been conducted, mostly by social psychologists. Other studies have examined attitudes by having listeners rate voices for, e.g., job suitability.
I illustrate the applicability of perception experiments to sociolinguistics with two perception experiments that I have conducted. The first is an investigation of how well listeners can identify African-American speakers who speak atypical African-American English. African Americans from Hyde County, North Carolina, show a mixture of typical features of African-American English (e.g., prosodic patterns) and features more typical of white Pamlico Sound English (e.g., fronted /o/, as in so). 5-second excerpts were extracted from field recordings; for each speaker, one excerpt featuring local vowel variants and one not doing so were taken. Each of these excerpts was subjected to three treatments: no modification, lowpass filtering at 330 Hz (to eliminate segmental information), and monotonization (to eliminate intonational information). Different groups of subjects heard each treatment and were asked to identify the ethnicity of each stimulus. Several prosodic and voice quality factors were measured for the stimuli. For the unmodified and monotonized stimuli, only the presence of diagnostic vowels showed a clear effect on identifications, indicating that the local vowel variants, especially fronted /o/, misled listeners. Results for the lowpass-filtered stimuli were more complicated.
A different sort of experiment involved examining how Anglos and Chicanos perceive variations in the quality of the glide of /ai/, as in tide and tight. In most dialects, the glide is higher before voiceless obstruents than before voiced obstruents. It was found, however, that Chicanos in southern Texas barely show this difference at all, while Anglos in most parts of North America show it robustly. Synthetically modified stimuli representing continua from tide to tight for length of the diphthong, the length of the murmur, and the quality of the glide were created. These stimuli were played to bilingual Chicanos from southern Texas and an Anglo group from Ohio. Identifications by both groups were affected by the length of the diphthong and by the quality of the glide. However, it was found that the two groups treated variations in the glide differently. Anglos used it as a cue to following /d/ vs. /t/, while Chicanos used it as a cue to following /d/ vs. null. Evidence suggested that Chicanos may rely more on stop releases to distinguish final /d/ from /t/.

The Jews, throughout history, have traditionally spoken a Jewish variety of the language of the surrounding community. Elements from their religious, cultural, and linguistic backgrounds contribute to differentiating their language use from that of their non-Jewish contemporaries. Judeo-Spanish is one such Jewish language. After the Expulsion from Spain in 1492, many Sephardim lived in the Ottoman Empire, where they continued to speak Spanish while incorporating words and structures from Hebrew, Greek, Turkish, and other languages with which they came in contact. It remains unclear, however, whether the Jews already spoke a distinctive variety of Spanish before the Expulsion, when they still lived in the Iberian Peninsula.
One text that offers some intriguing evidence of what Jewish language may have been like prior to the Expulsion is the taqqanot , or statutes, written by Jewish leaders in Valladolid in 1432. These statutes were composed to regulate life in the Jewish communities of Castile; they address topics including education, taxation, traitors, the election of community officials, and sumptuary law. This document is aljamiado - Spanish written in Hebrew script. The Spanish is heavily mixed with Hebrew, in the form of formulaic expressions (‘ may his Rock and Redeemer protect him ’), borrowings ( kašer ‘kosher;’ Talmud Torah ‘Jewish elementary school’) and code switches (que sea ha-rešut b-yado ‘ the authority shall be in his hand’). In this talk, I will describe the patterns of language use in the taqqanot and address the question of whether the taqqanot support the notion of a distinct Jewish Spanish in fifteenth-century Castile.
 
 

One of the most serious problems of the United States is the failure of the schools to teach basic reading skills to children whose home dialect is different from standard English. The Urban Minorities Reading Project at the University of Pennsylvania is engaged in testing an Individualized Reading Program that is based on five years of research in Philadelphia schools with African American elementary students. The program begins with a linguistic diagnosis of reading errors that maps each child's knowledge of the relation between sound and spelling, and directs the reader to the instructional sequence that is needed to advance in decoding skills. Direct instruction on alphabetic principles is then applied to narratives that develop the concerns, emotional interest and cultural perspective of African American youth. This program is now being expanded to Latino, African-American and Euro-American struggling readers in Philadelphia, Atlanta and California. We hope to discover how linguistic differences among ethnic groups affect progress in reading, and how to use our linguistic knowledge to raise reading levels on a national scale.
 
 

Sometimes what a sentence says varies from context to context, in the sense that different utterances of that sentence semantically encode different information. Such sentences are context-sensitive. For example, the sentence 'I am hungry' means one thing in your mouth and another in mine. Sometimes, however, different utterances of the same sentence convey different pieces of information even though both utterances semantically encode the same information. In such cases the speaker means more with her utterance than, or something all together different from, what the sentence she utters says. When we find different utterances of the same sentence conveying different pieces of information, should we say that it is because the sentence uttered is context-sensitive or should we say that it is because one or both of the speakers meant more than what was said? I propose a stark conception of context-sensitivity. In my view, context-sensitive expressions are few and far between. I think that only pure indexicals are context-sensitive. I propose a variety of ways in which a speaker might use a sentence to mean more than is said. In my view the most important, and philosophically interesting, ways do not require minimally competent speakers to realize that they are saying something different from what they mean.
 
 

It has been observed that when-questions are one of the last wh-questions produced by children learning English either as L1or L2. Explanations proposed for the late appearance of when -questions in L1 acquisition have been mostly based on cognitive factors. However, the cognition-based approach to when-questions faces problems in explaining L2 acquisition data, which show that L2 children, who are cognitively more mature than L1 children follow the same developmental sequence. In this paper, I propose a possible explanation based on internal linguistic factors. According to Enç (1987), Tense is a referential expression and temporal adverbials are antecedents of Tense. I develop Enç’s theory further and propose that in a when-question, Tense is a bound variable, which is bound by the quantificational interrogative when . Thus, in order to produce when-questions, children must be at a stage where they understand bound variable readings. According to Roeper and de Villiers (1991), English-speaking children learn a bound variable reading approximately after three years of age, and the learning continues through the kindergarten years. The age at which a bound variable reading first appears corresponds to the point when when-questions begin to occur. I propose that the complexity of the interaction between the quantificational when and Tense, a bound variable, causes the delayed production of when -questions in developing grammars.
 
 

L2 learners have linguistic knowledge they have internalized through communication and can use in spontaneous communication. They also have linguistic knowledge that they have learned from formal instruction but often fail to use in spontaneous speech production. While the distinction of these two types of linguistic knowledge in L2 learners has been long and widely recognized, it plays a very limited role in SLA because of the lack of a research paradigm that can be used to empirically differentiate the two types of knowledge. In this talk I will report a study that was intended to discover such a paradigm and use it as a tool for investigating the issue of what linguistic knowledge can be integrated in the process of SLA.
 
 

The majority of research in the field of language acquisition is based on data from a small group of European languages. As a result, a number of theories have developed around this small empirical base with the (sometimes-explicit) assumption that such theories can be extended universally. In this talk, I present several prominent theories of the Root Infinitive phenomenon in child language, and show that they are based on primarily five languages: English, French, German, Dutch and Italian. I then present data from four Swahili speaking children, and show that the facts exhibited in the Swahili data are compatible with only one of these theories: the ATOM Model (Schütze & Wexler, 1996). I show that while the ATOM Model is based entirely on English data, it is better exemplified in Swahili. The research paradigm in language acquisition that I advocate is articulated in this detailed study of child Swahili and its application to contemporary theory, and I propose a similar use of Swahili (and other understudied languages) in the field of Second Language Acquisition.
 
 

My research investigates how a language might change from an agrammatical, irregular pidgin to a regular and fully grammatical creole, in particular, examining the contributions of adult and child language learners to this process. It further examines the nature of the leaning mechanisms involved, asking whether they are necessarily specific to language, or might instead be of a more general nature.

Most current theories of pidgin and creole formation propose that adult L2 learners are primarily responsible for the formation of creole languages. Investigations of L2 acquisition consistently demonstrate that adult L2 learners’ grammars contain unpredictable variation. It follows that if pidgin and creole languages were created by adults, they should also contain this same unpredictable variation. Historical records of pidgins and creoles contain some evidence for this proposition. Creoles, however, do not typically contain unpredictable variation, and it is this aspect of the creolization process that my research addresses. In particular, I investigate whether adults could be responsible for the regularization of structures, or whether children are crucial contributors to the creolization process, smoothing out the inconsistencies left by adults. To investigate this, Elissa Newport and I have been running a series of artificial language experiments. The input languages mimic real pidgins by containing some unpredictable variation typical of adult L2 speakers. Our main question concerns whether adult or child learners of this ‘pidgin’ will maintain the inconsistency when they learn it, or rather will regularize the structure of the pidgin.

In our first study, adult participants were exposed to a language where the inconsistency was defined as the probabilistic presence or absence. After several days of exposure, participants were tested on their general knowledge of the language, as well their knowledge of the inconsistent items. Both production and judgment measures showed that participants learned the language, despite the presence of inconsistency. Of greater interest, however, was their use of the inconsistent items. First, participants maintained them in their speech. Second, participants’ productions reflected their input to a high degree. This study, then, suggests that when exposed to variation of this type, adult learners do not regularize it.

Pidgin languages also exhibit scattered inconsistency. This occurs when several different forms are used in the same way in a language, in the absence of any categorical or deterministic conditioning factors (linguistic or otherwise) to account for the variation. The second experiments examined whether scattered inconsistency produces more regularization. In particular, we hypothesized that the extreme unpredictability and complexity of multiple forms might necessarily lead learners to choose one form and regularize the language. To investigate this, we exposed subjects to a language that contained competitive inconsistency. Findings from these studies show a different picture than the original study. In these new studies many adult learners used the most frequent determiners more often than they had heard them in the input. Thus it seems that certain kinds of inconsistencies in the input may induce adult learners to regularize, at least a bit. Further, as the number of alternating forms (and the amount of unpredictable variation) increases, adults show a modest tendency to regularize the main form.

These results are very much what we would expect from the literature on probability (c.f. Estes, 1964). The results from most experiments in this literature show that, after very little exposure, adult participants' predictions begin to match the probabilities in the input (called probability matching). However, in a few studies where the probabilistic patterns were made more complex (e.g. by increasing the number of competitors), adults overmatched, predicting the most frequent competitor slightly more often than it actually occurred. The results of these studies, then, predict exactly the pattern of results we have found in our work, without recourse to language-specific mechanisms, suggesting that at least some aspects of creolization might be due to the operation of learning mechanisms not specific to language.
 
 

Loanword adaptation is sometimes considered to be a process of phonetic approximation, thus making it of marginal interest to phonologists. This presupposes that the phonology of L2, the language from which a loanword is drawn, is inaccessible to the borrower. The borrower is thus forced to rely on the phonetic form of the loanword, interpreting it through the filter of a perceptual system attuned exclusively to L1, the native language. It is this faulty perception that is held responsible for the sound changes seen in a word borrowed into another language (e.g. Silverman 1992; Yip 1993).

Our goal is to show that loanword adaptation is highly relevant to phonology, providing a rich source of evidence and insights not otherwise readily available. The borrowing process is overwhelmingly phonological because borrowers are mainly bilinguals, a claim supported by sociolinguistic studies and the systematicity of the adaptations in our database of loanwords that contains 45,738 malformations from 12 large corpora. In pursuit of our goal, we 1) show that cross-cultural perception studies, including studies of second language learners’ errors, predict sound substitutions that we do not find in our loanword adaptation data and 2) circumscribe phonetic approximation in loanword adaptation, showing that it is quite limited compared to phonological adaptation.

Our first point is that phonetic approximation makes unattested predictions. This has already been exploited, to some extent, in LaCharité & Paradis (2000) and in Paradis & LaCharité (2002) where it is shown that phonetic variants in the source and the borrowing language have no impact on loanword adaptation. For example, the English flap is identified as /t/ or /d/ in English loanwords in Spanish, not as the Spanish rhotic tap, which is phonetically closer. Here, we extend this line of argument by comparing some predictions made by perception studies of monolinguals with the adaptations that we find in our loanword corpora. For example, the Voice Onset Time (VOT) associated with the English voiced stops corresponds to the VOT of voiceless stops in Spanish. This predicts that monolingual Spanish speakers will identify English voiced stops as voiceless, something that perception studies with monolinguals confirm. However, in the adaptation of English loanwords in Spanish we do not see English voiced stops replaced by voiceless ones. Neither do several other sound confusions predicted by phonetic approximation occur other than rarely in loanword adaptation, though they are known to occur in the early stages of second language acquisition, diminishing with increasing exposure to L2 (Jameson 1967; Williams 1979; etc.).

With respect to our second point, we do not claim that phonetic approximation has no role whatsoever in loanword adaptation. In fact, it may well play a more important role in the adaptations of monolingual borrowers (consider, for example, the pronunciation of foreign proper names by monolingual speakers). In order to have at our disposal a sufficient number of phonetic approximations we looked for loanwords introduced suddenly into an almost exclusively monolingual society with no bilingual circles to speak of. Finding such a corpus is not easy, but the Glossaire du Parler Français au Canada which contains many English borrowings (with their phonetic transcriptions) introduced at the beginning of the English conquest of New France in 1763 and collected in the 19th century, provided the possibility for such a study. We culled from this source a corpus of 648 loanword forms that can be compared to two other corpora of recent English loanwords in Quebec French, that of Quebec City (2,412 loanword forms) and that of Montreal (2,245 loanword forms). The comparison indeed reveals some contrasts that we attribute to the greater influence of phonetic approximation in the early corpus. For instance, the rates at which foreign segments are imported are markedly different, as are the patterns of segment adaptations.

Nonetheless, overall, our corpora suggest that borrowers are rarely monolinguals. Cross-cultural perception studies, including studies of second language learners’ errors, predict sound substitutions that are not seen in our loanword adaptation data. This indicates that borrowers are standardly neither monolinguals, nor are they in the early stages of L2 acquisition, and that in equating sounds they look at their phonological identity, not their phonetic realization. Apart from certain well circumscribed situations, it does not appear that foreign words are adapted phonetically.
References

Jameson, G. (1967). The Development of a Phonemic Analysis for an Oral English Proficiency Test for Spanish-speaking School Beginners. Ph.D. dissertation, University of Texas, Austin.

LaCharité, D. and C. Paradis (2000), “Phonological Evidence for the Bililngualism of Borrowers”, J. Jensen and G. van Herk (dir.), Proceedings of the 2000 Annual Meeting of the Canadian Association of Linguistics, Cahiers de linguistique d’Ottawa. Université d’Ottawa, p. 221-232.

Paradis, C. and D. LaCharité (2002), “Guttural Deletion in Loanwords. Phonology 18.2: 255-300.

Silverman, D. (1992), “Multiple Scansions in Loaword Phonology: Evidence from Cantonese”. Phonology 9: 298-328.

Williams, L. (1979). “The Modification of Speech Perception and Production in Second-language Learning”. Perception & Psychophysics 26: 95-104.

Yip, M. (1993). “Cantonese Loanword Phonology and Optimality Theory”, Journal of East Asian Linguistics 2: 261-291.

Recent developments in our understanding of how the human brain handles information both linguistically and non-linguistically are giving new insights into how language began, and how "literature" is constructed in both literate and nonliterate societies.

Designed to be of interest to those in linguistics, cognitive science, archaeology, literary studies and myth/religion.

In the West Germanic languages, the voicing of the Germanic fricatives f, th, x and Indo-European s is attested at various times for the different members of the series and different word positions. Medial voicing in a voiced environment is attested in or generally posited for Old English, Old Frisian, Old Low Franconian, Old Saxon, and Old High German. Initial voicing is attested in or posited for Middle English, Middle Dutch, and Old High German. Although there are exceptions, voicing is generally reflected in orthography for the f, but not for s. The picture is further complicated by the individual developments of the various fricatives. Gmc. th became a stop in Dutch and High and Low German but remained in English; Gmc. x became an aspirate initially and intervocalically. Loss also occurred in both cases. For Gmc. s there is the further complication in High German of the development of a new sibilant through the second consonant shift. Also influencing the distribution of fricatives is the positional occlusion of the voiced Germanic fricatives which varied for each member of the series and geographically. Orthographic considerations also play a role in the attestation. Despite these complications and further developments in the later stages of West Germanic, it seems the voiceless fricatives generally underwent a common development in the early stages of West Germanic. It seems that the voicing is an old, rather than a younger, development and is, to a great extent, dependent on the development of the stop series.

Generalizations about the discourse mechanisms whereby women achieve cooperativeness have been largely based on data from informal conversations. This paper aims to give a more complete picture of female discourse by comparing topic development in all-women conversation in the context of business meetings to previous findings based on casual conversation. The findings show a clear effect on the mechanisms and nature of topic development for the business meeting context.
        Coates (1989), in a study of women’s informal conversation, provided empirical evidence in support of the frequent claim (e.g. by Maltz and Borker 1982) that women develop topics progressively in conversation, by opting for continuity over discontinuity and shifting topics gradually rather than abruptly, and identified specific discourse sections in the process of topic development, to which speakers and hearers showed sensitivity. This study analyzes the speech of women in the business meetings of the ladies’ auxiliary of an outdoor recreation club in the American midwest, with a focus on how the structure of the conversations reflects the speakers’ awareness of the business meeting as relevant context.  A comparison with the conversation of Coates’ speakers yields both similarities and differences. While gradual topic introduction and cooperative development do occur, more topics are introduced and on average they are much more briefly covered. Also, topic introduction tends to be abrupt rather than gradual, which is accounted for by the topics’ relevance to the purpose and structure of the meeting, and the right or responsibility to introduce topics is allocated with reference to the speaker’s role in the club. Further, many topics, particularly those in which the contributions are relatively predictable from the pre-established structure of the meeting, are developed with little or no collaboration; however, topics that require group decisions and other topics in which the contributions are relatively unpredictable are developed at length, with structure and collaborative features similar to those in Coates.
        These findings both support and expand on the findings of Coates, in that the women in this study use the collaborative mechanisms that Coates describes but also show sensitivity to the contextual features inherent in a business meeting. Comparing the speech of the women in this context to those in Coates' study, with the specific question of how their status as club members or officers, the relative formality of the purpose of their gathering, and the business-related as opposed to personal goals of their interaction may affect their speech can be a step toward a more complete understanding of female discourse.