Some parallels, accidental and expected

The articles in this issue were developed from presentations in the Intersections workshop held concurrently with NWAV44. The goals of that workshop were to widen and deepen the study of linguistic variation by creating intersections with other approaches to linguistic analysis.

Before I accepted the invitation to introduce our speakers, I asked myself the question, “why me?” Why am I particularly qualified to talk about the limitations of our field? After some thought, I realized it must be because I share those limitations as much as anyone here. I have never worked with the formal apparatus needed to study syntactic change. A majority of my work has been in phonology and sound change. I have never studied multilingual communities. I have done only one study of the acquisition of language. None of the languages or dialects I have studied have been in danger of extinction. And until recently, I had not created any corpora.

Given such a qualifying limitation, what could I say to make a useful connection with our distinguished contributors? Reading through the drafts of the plenary talks that I received prior to the workshop, I did locate in each case some good point of contact, finding, to put it informally, that we were on at least one of the same pages. To illustrate this goal, I will highlight several of these intersections here.

In the background – and often foregrounded – will be the challenge of the covariation of form and meaning. Since the early years of variation studies, the majority have dealt with grammatical factors – plural marking, past tense specification, verbal agreement, auxiliary realization, possession, do-support, question inversion, negative concord, adverb placement. Nevertheless, at each step we return to the issue raised by Beatriz Lavandera (1978). Her observation that the alternation of tense and aspect in Spanish si-clauses cannot be considered alternate ways of “saying the same thing” has been generalized to question all studies of variation above the phonological level. Sankoff and Thibault's (1981) definition of “weak complementarity” has not resolved the issue. As re-formulated by Adger (2014), the fundamental problem is how to determine the equivalence between two syntactic forms and a single semantic interpretation, a problem that does not arise at the phonological or morphophonological levels. On the other hand, Poplack and Dion (2009) observe that the search for an equivalence of form and function has led linguists as well as prescriptive grammarians to formulate semantic differences that are not supported in actual use.

Rather than tackle these general issues head on, I will look to those parallels that bring our work into contact.

I turn first to the article by David Adger, who broke new ground in the treatment of variability in syntactic theory (Adger 2014). Here he re-analyzes the findings of Nancy Dorian on the possessive in East Sutherland Gaelic, and deals, more specifically, with the alternation of inflectional and periphrastic possessives in this dialect.

1. (1) bràthair Sheumais  am bràthair aig Seumas

   Seumas's brother  the brother of Seumas

It just so happens that my own research on ways to raise reading levels in elementary schools had recently encountered the same variable. Wolford (2006) studied this variable in our recordings of 24 children in the 2^nd, 3^rd and 4^th grades in Philadelphia, Atlanta and Southern California. She observed two types of variation from the expected forms of classroom English:

1. (2) and the friend of my brother brought it back. (~ my brother's friend)

2. (3) …like when I go to my cousin house. (~my cousin's house)

Type (2) possessives, though not ungrammatical in English, suggest influence of the Spanish possessive construction, which is always periphrastic. Type (3) is parallel to the most common pattern of AAVE in which the possessor in bare form is directly followed by the possessed. The suggestion that this shows the effect of dialect contact is confirmed by Wolford's multivariate analysis. Type (2) is most favored by Latina girls in California who learned to read in Spanish first. Type (3) is most favored by African-American boys in Philadelphia. Furthermore, children from Mexican families showed a much stronger influence of Spanish than those from Puerto Rican families. This quantitative pattern shows the results of differential contact of gender and ethnicity in California and Philadelphia as studied by Poplack (1979), Baugh (1979) and Santa Ana (1991).

Language variation here seems to be the result of dialect contact, perhaps at the most superficial level. I am particularly interested in how Adger's account of variation within the minimalist framework can give us deeper insight into the mechanism and outcome of syntactic variation.

I found an even more striking parallel in the finding of Elizabeth Johnson, ¹ that infants exposed to multiple accents fail to recognize familiar words in Canadian English. This runs counter to our view that our linguistic abilities generally improve as our experience grows. The notion that new experience interferes with the core linguistic capacities seems contrary to common sense and even alarming for linguists who consistently argue that bilingualism is good for you. But it echoes in some respects one result of our experiments on Cross Dialectal Comprehension (Labov and Ash 1997, Labov 2010: Ch. 2). In these experiments, subjects in Philadelphia, Chicago, Birmingham and Pittsburgh were exposed to extreme forms of sound change to be found in their own city and in other cities, first as isolated words, then in the original phrase, then in full sentences. In most cases, subjects did rather poorly with single words, even when these were the phonetic forms they heard most commonly around them in daily life, and improved rapidly as context expanded. For example, the Chicago pronunciation of socks as [sæks] was correctly recognized by 10% of the college students from Chicago but by only 1% of the Philadelphia college students. The Birmingham pronunciation of guy as [ga:] was correctly recognized by 62% of the Birmingham college students but by only 57% of the Philadelphians.

A striking feature of these experiments is that they are some of the few examples where high school students perform better than college students. Thirty-three percent of the Chicago high school students recognized their own [sæks] correctly (as opposed to only 10% of the college students). The Birmingham high school students recognized [ga:] as guy 88% of the time (as compared to 62% for college students).

The most reasonable interpretation of these results is that exposure to a wider variety of phonetic norms interferes with the ability to operate the basic system acquired in the pre-adolescent years. This is one of the findings that show that the educational process is not entirely a gain; there are some losses along the way. The loss in the ability to acquire new languages in a native-like fashion is a more profound issue. But we can make some connection with Bohn and Flege's (1992) explanation for this fundamental feature of language learning: that our experience with language variation in adolescent years gradually categorizes all ‘new’ sounds as variants of our native categories, so that the effort to produce this new target automatically produces the older ones. Johnson's finding indicates that the accumulation of new norms operates from the earliest years.

The study of change and variation naturally began with well-known languages, where the category structure was already defined and articulated. But I have had many occasions to address the question that is so well formulated here by Meyerhoff in her article: “But what happens when you are analyzing variation at the same time as you are grappling with the fundamental structure of the language?” In recent years we have seen a rapid expansion of variation studies in indigenous and endangered languages, especially with the creation of NWAV Asia-Pacific. Here students of variation may be prepared to find that the fundamental systems may themselves vary. This may be true of such rapidly changing grammars as that of African-American Vernacular English. Though I would not be inclined to say that this is the case for the mainstream Philadelphia dialect of English, our most recent investigations of Philadelphia phonology have been involved with the variation of phonological systems rather than individual linguistic variables (Labov et al. 2016). In Meyerhoff's terms, we are dealing with an orchestra of variables rather than with a sonata for a single variable.

The systems involved in this variation are those that control the split of Philadelphia short-a into a tense and a lax category. The traditional system involves a true orchestration of the phonological resources of the dialect. Vowels with front voiceless fricative or nasal codas are tense, but those with coda /d/ are not, except for the three words mad, bad, glad; and excepting function words and the three irregular verbs ran, swam, began, and excepting polysyllabic words with zero onset, etc. etc. A new system has arisen in competition with this, among a rather large population of young speakers oriented to higher education (in non-Catholic schools with special admission requirements). This is the nasal system, rather similar to that found with increasing frequency in many parts of the U.S. It is easy to describe:

1. (4) Tense all and only all vowels before nasal consonants.

But to shift from the traditional to the nasal system requires a number of opposing moves:

1. (5)

   1. a. Tense /æ/ before intervocalic nasals (Spanish, Miami, damage)

   2. b. Tense /æ/ before velar nasals (bank, bang, hang)

   3. c. Tense /æ/ in function words with nasal codas (can, am, an)

   4. d. Lax all other /æ/

In tracing this development across the last two decades, we find that steps (5a–d) occur simultaneously. It is not a step-by-step conversion that terminates in the nasal system but rather a single shift from traditional to nasal. With the systematic approach to systems developed by Meyerhoff, we may find an increasing tendency to move from the study of linguistic variables to the variation of linguistic systems.

My last connection with our panel on intersections finds the same theme in Szmrecsanyi's contribution on corpus linguistics. He writes “Traditional research in the LVC tradition explores the conditioning of individual (socio)linguistic variables, one variable at a time. But recently researchers have taken an interest in the coherence of lects and the behavior of multiple variables.” This seems to be saying the same thing as in my discussion of Philadelphia phonological systems, if we take it to mean that these multiple variables are connected linguistically, not that they are simply found among the same speakers. It is not uncommon for analysts to create an index of features which have no inherent connection (i.e., Levenshtein numbers).

Szmrecsanyi is certainly right in saying that more and more corpus studies can be expected in the near future. Each finished sociolinguistic study leaves behind data in the form of recordings, reports, transcriptions, analyses – all of which increase in interest and value as time goes by. Let us consider for a moment the decision to be made by a linguist just entering the field: should he or she analyse a piece of a finished corpus, or enter a speech community and gather new data? Clearly a matter of personal choice and style is involved. One might not trust the ability of the earlier researcher to approach the vernacular, and it is perfectly true that some of the early corpora were missing the intensity and vivacity of interviews done by Anne Bower or Arvilla Payne. These issues have come to a head for me because the University of Pennsylvania Library has undertaken the creation of a Penn Sociolinguistic Archive, digitizing the 7,000-odd recordings created by my research from 1961 to 2012, from Martha's Vineyard to New York City, Philadelphia, including the Telsur Project interviews for the Atlas of North American English, exploratory interviews in the United Kingdom, Australia and New Zealand, studies of Parisian French and Sankoff's basic survey of Montreal. It turns out that I myself did only 1,000 of these interviews, so the knowledge I gained from personal contact cannot be a major factor. These digital recordings will become available for others’ use, transcription and analysis. Some 500 interviews have been transcribed to form the Philadelphia Neighborhood Corpus, but the plan is for an increasing number to be made available to those who want to participate in the business of tracing language development and change, past and present.

So it seems that some progress has been made in the project I set for myself in 1961 when I first entered linguistics, which I saw as an exciting and exhilarating approach to the understanding of human nature. I thought that it might become even more exciting and exhilarating, and more cumulative, if it were based on what people actually said rather than what they thought they might say. This workshop, and the articles in this issue, demonstrate that we have moved some considerable distance in that direction.