Introduction to Dialect Syntax: core concepts David Hall Outline and questions to tackle • What is a dialect? Are there linguistically meaningful differences between “dialects” and “languages”? • What different kinds of variation and variability do we see? • How do generative linguists and variationist sociolinguists differ in their approach to the study of language and dialects? • Principles and Parameters approaches have been quite successful in describing dialectal variation. • There is something more than cross-dialectal variation: there is inter-speaker and intra-speaker variability which a blanket parametric approach struggles to capture. • Given this, what are alternative generative approaches to dialectal and idiolectal variability, which can capture idiolectal variability? • From a methodological point of view: what are the pros and cons of different types of linguistic data used in different traditions, and how should you go about tackling complex variable data? 1 Language, Dialect, Idiolect: what’s the difference? • What are often called “languages” are often more similar to each other than what are referred to as dialects. (1) Separate languages? (2) Separate dialects, same language? a. Serbian and Croatian a. Mandarin and Cantonese b. Hindi and Urdu b. Yiddish and German c. Swedish and Norwegian c. Sicilian and Italian • From a purely linguistic perspective, Hindi and Urdu are the same language (same syntax, phonology, morphology, etc., but different script), with some lexical differences. They are mutually intelligible. • Mandarin and Cantonese, on the other hand, are not mutually intelligible: they have different phonology, and to some extent syntax, aside from lexical differences. 1 Dialect Syntax D. Hall • Mutual intelligibility seems to be the best we can do, but it’s not good enough: there is the problem of gradience in mutual intelligibility. • It seems that “languages” and “dialects of a language”, as separate entities, are not really definable linguistically. From a linguistic perspective, everything is a dialect. • BUT the terms have sociopolitical reality. A sociopolitical definition seems more fitting. Hindi and Urdu are considered to be separate languages because of complex historical and sociological reasons. • Famous quote, attributed to Max Weinreich, who attributed it to an auditor at one of his lectures: “A language is a dialect with an army and navy.” Summary • There is no linguistically meaningful difference between a ‘language’ and a ‘dialect’. Everyone speaks a dialect, but some dialects have more social and political capital associated with them than others. 2 Differing conceptions of the object of study: what’s the ‘ling’ in ‘linguistics’ ? • Generative Linguistics (Chomsky), also to some extent Structuralist Linguistics (e.g., Saussure) – The object of study is linguistic competence: the rule governed mental system that generates well-formed linguistic expressions (what is the nature of the structure of linguistic expressions, and what is the nature of the system that gives rise to them). – This involves an abstraction away from the nuances of what actually comes out of someone’s mouth at any given point: a variety of non-linguistic factors can give rise to ‘performance issues’ (memory, motor-control, environmental factors). Homogeneity across speakers is assumed. “Let us refer to this ‘notion of structure’ as an ‘internalized language’ (Ilanguage). The I-language, then, is some element of the mind of the person who knows the language, acquired by the learner, and used by the speakerhearer... The statements of a grammar are statements of the theory of mind about the I-language, hence statements about structures of the brain formulated at a certain level of abstraction from mechanisms.” (Chomsky 1986 pp21-23) • Sociolinguistics (Labov) – The object of study is linguistic performance, the actual use of language in concrete situations, and the (orderly) heterogeneity of language in use. 2 Dialect Syntax D. Hall – Further, the object of study is the sociolinguistic variable, and what constrains the variability present in speech within and across communities (more on this later). – This extends to language change: what/who drives it, when and why. – All of these things are tied to the social meaning of different variables, and communicative competence: our implicit understanding of the socio-pragmaitc rules of language use. – “The concept of idiolect might be employed here: we might isolate each style as a separate idiolect, and attempt to describe that. However, the over all [sic] result of this study indicates that idiolects are not the most consistent, most explicable unit of linguistic behaviour in New York City. On the contrary, the speech of many individuals appears as studded with oscillations and contradictions, and it is only when it is placed against the overall framework of social and stylistic variation of the speech community that we can discern the regular structural pattern that governs this behavior.” Labov (1966, p11) • Generative I-language approach appears at first glance to have nothing much to say about variation/variability, as it abstracts away from local differences, and takes the main object of study to be the cognitive system that allows us to generate and interpret complex linguistic expressions. • The sociolinguistic perspective, on the other hand, is focused pretty much solely on variability. • Is there a way to bring together sociolinguistic ideas about variation (the sociolinguistic variable) and generative approaches to the language faculty? Can the I-language perspective be enriched by including the variability that sociolinguistics focuses on? • Well, yes! Summary • From an I-language perspective, it makes sense to go one step further and say that there are not even really dialects: we’re after a characterisation of an individual’s mental make-up. It’s probably the case that no two speakers have the same grammar, but of course there does exist generally consistent behavioural data (judgments) across a community in clear cases. • From a sociolinguistic perspective, there are certain regularities that can only be gleaned from studying the entire speech community, and interactions between social and linguistic variables. • We should be looking for ways to unite the two approaches: we want to be able to abstract away from variability at a certain level, but explain it when it is rule governed. 3 Dialect Syntax 3 D. Hall Kinds of Variation and Variability • We can think of variation at three levels: – Variation between “languages” (cross-linguistic variation): is a phrasal category in a language head-initial or head-final? (3) a. I [VP went to Scotland]. (4) a. [VP V [PP P DP] ] b. boku wa [VP sukottorando ni itta]. I top Scotland to went b. [VP [PP DP P] V ] – Variation across dialects: does a dialect have negative concord, where another doesn’t? (5) a. I didn’t do anything. b. I didn’t do nothing. – Intra-speaker (idiolectal) variability: does a speaker occasionally realise word medial intervocalic /t/ as [P], and sometimes as [t]? (6) water a. wO:t@ b. wO:P@ • Note that each of these types of differences could also appear at each other level: negative concord can be a regular feature of a standardized language (e.g., Italian, Hungarian), and it is possible that word-order in some phrases could be a dialectal difference (e.g., She gave a book the man in some Northern English dialects). Phonological Morphological Syntactic Inter-language Initial aspiration of stops Direction of affixation Word order; Agreement; Negative concord Dialectal l-vocalization; t-glottalization; rhoticity Idiolectal t-glottalization Agreement; Negative concord; word order? Agreement; Negative concord • What causes cross-dialectal variation and intra-speaker variability? A number of factors: sociological factors, geographical factors, discourse factors, processing factors, prosodic factors, grammatical factors... 4 Dialect Syntax D. Hall • We should note that from a generative perspective, we should expect no difference between the limits of variation between “languages” and between “dialects”, because they are one and the same. • There are, however, two different kinds of variation that we need to separate out, but which could potentially be explained in the same way: 1. Variation within a community which largely shares the same judgments: e.g., group A always requires 3sg -s with 3sg subjects, group B does not (both groups share a broadly similar grammar). 2. Variability in output within an individual: e.g., sometimes I say “we were...”, sometimes I say “we was...” Summary • There is variation across ‘langauges’, dialects, and also within an individual speaker. • For the most part, the types of variation that we see across languages, dialects, and within idiolects is the same. • This fits well with the idea that there really is no distinction between languages and dialects. We can perhaps go further and say that only idiolects are real in a linguistic sense: this is what the I-language approach says, but the sociolinguistic approach doesn’t quite. • We should probably distinguish between within-group variation and withinindividual variability. • We’ll come back to the notion of sociolinguistic variable later, where we’ll see that there are some kinds of variability which seem to go beyond the individual system. 4 Covariation of phenomena and parameters • One classical way of thinking about variation across languages and dialects comes from the Principles and Parameters framework of generative linguistics. • Fundamental idea: – there is a set of universal principles governing the nature of the linguistic system, and these principles hold across all human language users. Proposed principles: structure dependence of rules, binding theory, move α, cyclicity; and later on in GG history: Merge, Agree, Phases (cyclic spell-out). – Alongside this, there are a set of parameter settings, whose values vary across speakers. They can be thought of as a set of ‘switches’ or binary choices between two options, and determine the differences in the surface grammar that we see.1 1 As we’ll see later, the ‘switch’ analogy is not the only way to think about parametric variation. 5 Dialect Syntax D. Hall – Comparative parametric syntax “accounts for observed clusterings of syntactic properties by showing that the several properties in question can all be traced back to a single relatively more abstract parameter setting.” (Kayne 2000, p3). • For example: there could be a set of ‘headedness’ parameters which determine whether a phrase is head initial or head final. The assumption is that at some point in acquisition, the child gets to a point where the primary linguistic data points to a particular setting of a parameter, and switches it to the appropriate setting.2 • We might expect certain parameters to interact with each other, so that certain patterns would be expected to be common/rare/impossible (and this could explain Greenberg universals/tendencies). (7) Example parameter hierarchy (taken from Boeckx and Leivada 2013) • A microparametric approach would go one step further and suggest that differences between dialects are also accounted for in the same way, just with more subtle parameters. 2 There are a variety of ways to think about how these parameters get set: see, e.g., Yang (2002), the Cambridge ReCoS project (e.g., Biberauer, Holmberg, Roberts, and Sheehan 2010, Bazalgette 2015), Kayne (2000), among many others. 6 Dialect Syntax D. Hall • It is out of this kind of conception of linguistic variation that we get a more meaningful use of the term ‘dialect’: you might see syntacticians using the term to refer to groups of people who mostly share grammaticality judgements where there is some variability within a language group. The ‘dialect’ is the grammar that results from a set of parameter settings interacting. • Richard Kayne: “My own experience in observing the syntax of English speakers, both linguists and nonlinguists, makes me think that it is entirely likely that no two speakers of English have exactly the same syntactic judgments . . . in fact, if it is true that no two English speakers have the same syntactic grammar, then the number of varieties of English/distinct grammars of English must be at least as great as the number of native speakers of English . . . Extrapolating to the world at large, one would reach the conclusion that the number of syntactically distinct languages/dialects is at least as great as the number of individuals presently alive.” (Kayne 2000:7-8) • This idea goes back at least 100 years: “... we have, strictly speaking, to differentiate as many languages as there are individuals.” (Hermann Paul 1890, Principles of the History of Language) 4.1 A concrete example: parameters applied to dialects and idiolects • Henry (1995) — Inverted imperatives in Belfast English (8) a. You go away! b. Go you away! c. Read you that book! • For some (typically old) speakers, any verb can do this. Henry proposes that this is a case of optional verb movement to C. 7 Dialect Syntax D. Hall (9) CP C[imp] TP T vP you vP v VP V read DP that book • For some (typically younger) speakers, only unaccusative verbs and passives can do this: Henry proposes that there is no verb movement to C for these speakers: the subject is just structurally base-generated in post-verbal (object) position, and never moves out of VP. (10) a. Go you away! b. *Read you that book! (11) CP C[imp] TP T VP VP V go away DP you 8 Dialect Syntax D. Hall • Some additional evidence that this is the case comes from the position of adverbs: VP adverbs never intervene between the verb and the subject (12) a. Quickly run you home b. *Run quickly you home • A parameter could exist which determines whether the verb can move to C or not. For younger speakers, the setting is [− move to C], for the older speakers it is [+ move to C]. • This is a case of the slow disappearance of grammars where verb movement to C is possible. Henry claims that this is because standard English does not allow this generally, and the influence of Standard English on the younger generation has become overwhelming. Thus the language change can also be explained. • Note: to capture the intra-speaker variability, we still need something like an optional feature/rule. Summary • The generative enterprise has tried to explain the vast variation that we see whilst maintaining that there is also a core set of principles that underlie the faculty of language. • The universal characteristics are called Principles: they do not vary, and underlie the language system of all humans. • The variation that we see is due to different sets of parameters, which are set based on the PLD that the child is exposed to. • The idea of Parameters can be applied to explain cross-dialectal differences. 5 The Sociolinguistic Variable • Some linguistic differences are not differences in truth conditional meaning, but do convey different social meanings.3 • Cheshire (2005), pp86-87: A sociolinguistic variable is “a structural unit with two or more variants involved in co-variation with social variables.” (13) I bought a bottle of water a. ai bO:t @ b6tl @v wO:t@ " b. ai bO:P @ b6PO»: @v wO:P@ 3 The lack of difference in truth conditional meaning is less clear in the case of syntactic variables, but we can see a similar effect nonetheless. 9 Dialect Syntax D. Hall (14) a. We were going to Sainsbury’s when ... b. We was going ∅ Sainsbury’s when ... • The (a) and (b) examples are equivalent in truth conditional meaning, but people will assign different social meanings to them. • The (b) variants are often proscribed, by teachers, politicians, social commentators, and other authority figures – but what is the driving force behind negative attitudes? Is there any sense in which one form is more degraded than the other? • Chambers (2002) points out that this kind of attitude has been around since some of the earliest recorded writings: (15) Sirach, Old Testament moralist (16) Cicero, 55 bc “When a sieve is shaken, the rubbish is left behind; so too the defects of a person appear in speech. As the kiln tests the work of the potter, so the test of a person is conversation” (Ecclesiasticus 27: 4-5) “learn to avoid not only the asperity of rustic pronunciation but the strangeness of outlandish pronunciation” (De Oratore III, 12) • Possible variables which interact with use of a particular variant include age, class, gender, ethnicity, relationship, relative standing in some hierarchy, language attitude etc. • Sociolinguistic variables can be sensitive to some social variables and not others: it might be the case that t-glottalization covaries with class and age, but not ethnicity. • Example from Labov (1966): percentage of ‘constricted /r/’ use across class groups and across speech styles. (17) Class Working Middle Upper A 2.5 4.0 12.5 Style B C D D0 10.5 14.5 23.5 49.5 12.5 21.0 35.0 55.0 25.0 29.0 55.5 70.0 Styles: casual speech (A), careful speech (B), reading style (C), pronunciation of isolated words (D), pronunciation of minimal pairs distinguished only by the variable in question (D0 ). 10 Dialect Syntax D. Hall • (r) is the variable, and has two variants, /r/ and ∅. X is the frequency with which any variant occurs, and X is correlated with social variables and stylistic levels. (18) a. (r) = X /r/ b. X = f(S,C), where S = stylistic level, C = class status. • The data in the table suggests a linear model: (19) f(S,C) = aS + bC + c • “Some linguistic variables are functions of other linguistic variables, and are not directly influenced by any factors outside of the linguistic system” (p16) • Certain speakers might not ever even produce /r/ in speech, but the claim is that they nonetheless participate in the sociolinguistic structure of the community, and that the variable (r) is a part of that system. • The important take-home message: there are multiple forms which have the same (compositional) meaning, but which have different social meaning. Choice of a particular variant appears chaotic when focusing only on the individual, but can be understood as structured and rule-governed when viewed as embedded in the interactions of the wider speech community. Choice of a particular variant is a function of certain social (and linguistic) variables. Summary • The idea of variable output throws a spanner in the works: differences are not just across individuals, but also within individuals. • This also presents a problem for the generative ‘rule-based’ approach to linguistic output. Whether one variant or another is used does not depend only on linguistic factors. • To capture this, it seems we need to be able to talk about “multiple grammars” producing different outputs, but is that really what we want? Is there some other way to think about variability in the output? 6 A different approach to parameters — it’s all in the lexicon. • The parametric system that we discussed earlier would have to say that a speaker who exhibits variable output of a particular syntactic variable has two different parameter settings, and that the speaker can switch between them. But this goes against most ideas about how parameters are set from an acquisition perspective. 11 Dialect Syntax D. Hall • A slightly different implementation of the idea of parameters became popular with the advent of the Minimalist programme. This alternative view of variation in minimalism has been influenced by a proposal by Hagit Borer (originally in 1984), which later became the Borer-Chomsky conjecture (Baker 2008): (20) The Borer-Chomsky Conjecture All parameters of variation are attributable to differences in the features of particular items (e.g., the functional heads) in the lexicon. • Therefore, the idea of parameter switches at a global level has been replaced by the idea of different feature specifications on heads at a very local level. • Take the Belfast English example earlier. We might imagine that Imperative C has a lexical specification for a strong feature which will enter into an Agree relation with, and attract, the verb. • Adger and Smith (2005) propose that we can do this by making use of interpretable vs uninterpretable features in the syntax. If the difference in the two lexical items lies only in the uninterpretable features, then we have the same compositional meaning (the syn-sem interface ignores these features), but we leave open the possibility that they can carry different social meanings (and also that phonological outputs could be different). • We can imagine that for the older generation of speakers, there are two possible different lexical items for Imperative C: one with a strong feature which attracts the verb, and the other without this feature (* indicates a strong feature, which attracts the goal: see Adger 2003, Chomsky 1995). The choice of particular lexical item is then influenced by a variety of language external factors. (21) Feature specifications for the two imperative C heads: a. inversion-C: [i imp, u*V] b. non-inversion-C: [i imp, uV] • Importantly, though, the speaker only has one grammar, just some extra lexical items. Summary • An influential way to think about variation is the Borer-Chomsky Conjecture. • Different features on functional items means syntax appears to behave differently across individuals. • We can model sociolinguistic variability using different uninterpretable feature specifications. 12 Dialect Syntax 7 D. Hall Other generative approaches to idiolectal variability • There are a variety of other ways of thinking about variability from an generative perspective. I give a short summary of two approaches here. 7.1 Kroch 1994: competing grammars • Variability in some grammatical feature invariably leads to a stable grammar where one of the options wins out in the end (examples Ancient Greek SOV to SVO, Late Middle English V2 to normal SVO). • There is no stable variability: variability is the reflex of competing grammars, where one eventually wins out. • Evidence for the notion of competition between grammars can be seen in the Constant Rate Effect, whereby grammatical changes, regardless of the differing frequencies of variants of a variable, follow the same rate of change of frequency (the slope of the regression model is the same). • Competing grammars are only ever posited by the learner when there is an appropriate level of evidence in the PLD. • Morphological variants which are not functionally distinguished are disallowed (competition and blocking): “so we should not expect to find variation between semantically non-distinct syntactic heads”. • The same can be said of functional heads in the syntax. • This is broadly similar to the Borerian approach that Adger and Smith take, with a slightly different execution: it is mainly focused on the idea of unstable variability across a community. 7.2 Henry 2005: Back to phrase structure rules • There are a bunch of different factors which affect whether agreement on the verb in existential-there sentences in English is sg or pl, and these factors vary across individuals. This gives rise to quite a lot of variation across speakers. (22) Sentence There are three books on the table There’s three books on the table There is three books on the table There was three books on the table Group 1 Group 2 Group 3 X X X Group 4 X X X * * X * * X * X X * • A strict minimalist conception of agreement between verb and subject (and associate), modelled as Agree/feature valuation cannot straightforwardly capture the massive variation here. 13 Dialect Syntax D. Hall • What we need is some set of conditions on agreement which is differently specified for each speaker, and a blanket parameter based approach cannot capture the data (note also that this is structure specific, the same factors do not affect non-expletive subjects). • Henry’s suggestion is that variability in the input leads the individual to posit a set of rules that captures the data, and these rules are construction specific in some cases. The grammar, it seems, is a set of rules: Henry argues that our research project should be to work out the limits on the type of rule that language can make use of (e.g., agreement is never dependent on the number of words away from the verb the associate is). 8 Methodological considerations 8.1 Data types: mining corpora vs judgment tasks • Pros and cons of different types of data. Mining Corpora Possible performance errors Possible annotation/coding errors Possibly skewed in non-linguistic ways Frequency data available Unpluggable gaps Mostly divorced from observational bias Non-standardized Introspection/Judgment tasks Can control for at least some performance errors Little to no need for annotation/coding of data Can control for skewing No frequency data available Gaps can be plugged Possible subjective skewing Can be affected by standardization • An important point on frequency data: when is a zero really a zero? – You might find that you have a cool corpus full of interesting spontaneous speech. You are looking at one variable in particular (say Preposition-drop in Multicultural London English), and upon a sophisticated statistical analysis, you discover that P-drop never occurs when the verb go is in an embedded clause of depth 2 (John said that Mary thinks that she’s going ∅ church on Sunday). – Does this mean that you have a categorical absence of P-drop in such an environment? Is this a real zero or just an artefact of the data? – This is where judgment tasks really pull their weight! You can test this. – The moral is: never assume that zero means *. Unfortunately, there is no negative data in corpora, but you can use the positive data to form hypotheses which are testable with judgment tasks. 8.2 What to do when confronted with variability • Take all of this into account when confronted with interesting dialect data – What does the corpus data (if any) suggest? 14 Dialect Syntax D. Hall – What might your (internal) linguistic variables be? What might your (external) sociological variables be? – Is the appearance of a certain variant conditioned by hard (categorical) or soft (probablilistic) constraints? – Is it a mixture of the two? – Is variability in one domain isolated, or does it correlate with variability in another domain? – Do frequency data alone suggest any hypotheses that you can test with judgment tasks? Overview and Conclusion • The relevant target of analysis of language variation is the idiolect: every person has their own linguistic system, but this system is likely to be similar to those around them. • There is not only variation between speakers/groups of speakers: there is also intra-speaker variability. • This intra-speaker variability is traditionally the topic of sociolinguistic analysis, because use of a particular variant often covaries with certain sociological variables. • However, we as generative linguists also have to say something about how to model this variability. • There are a number of ways to do this: multiple grammars, optional rules, stochastic weighting of rules, differing lexical feature specifications. • Great guiding principle from Barbiers (2005, p235): “Generative linguistics and sociolinguistics are complementary in that it is the task of sociolinguistics to describe and explain the patterns of variation that occur within a linguistic community, given the theoretical limits of this variation uncovered by generative linguistics.” • It is quite possible to gain insight into the underlying nature of the human linguistic system through studies of variation. References Adger, David (2003). Core Syntax: A Minimalist Approach. Oxford: Oxford University Press. Adger, David and Jennifer Smith (2005). Variation and the minimalist program. In Syntax and Variation: Reconciling the Biological and the Social. John Benjamins. 15 Dialect Syntax D. Hall Baker, Mark (2008). The macroparameter in a microparametric world. In Theresa Biberauer (ed.), The Limits of Syntactic Variation, 351–373. Amsterdam: John Benjamins. Barbiers, Sjef (2005). Word order variation in three-verb clusters and the division of labour between generative linguistics and sociolinguistics. In Leonie Cornips and Karen Corrigan (eds.), Syntax and Variation: Reconciling the Biological and the Social. John Benjamins. Bazalgette, Timothy (2015). Algorithmic acquisition of focus parameters. Ph.D. thesis, Cambridge University. Biberauer, Theresa, Anders Holmberg, Ian Roberts, and Michelle Sheehan (eds.) (2010). Parametric Variation: Null Subjects in Minimalist Theory. Cambridge: Cambridge University Press. Boeckx, Cedric and Evelina Leivada (2013). Entangled parametric hierarchies: Problems for an overspecified universal grammar. PLoS ONE 8(9). Chambers, J. K. (2002). Studying language variation: An informal epistemology. In J. K. Chambers, Peter Trudgill, and Natalie Schilling-Estes (eds.), The Handbook of Language Variation and Change. Blackwell. Cheshire, Jenny (2005). Syntactic variation and spoken language. In Leonie Cornips and Karen Corrigan (eds.), Syntax and Variation: Reconciling the Biological and the Social. John Benjamins. Chomsky, Noam (1995). The Minimalist Program. Cambridge, MA: MIT Press. Henry, Alison (1995). Belfast English and Standard English: Dialect Variation and Parameter Setting. Oxford: Oxford University Press. Henry, Alison (2005). Idiolectal variation and syntactic theory. In Leonie Cornips and Karen Corrigan (eds.), Syntax and Variation: Reconciling the Biological and the Social. John Benjamins. Kayne, Richard (2000). Parameters and Universals. Oxford University Press. Kroch, Anthony (1994). Morphosyntactic variation. In Katharine Beals (ed.), Papers from the 30th Regional Meeting of the Chicago Linguistics Society: Parasession on Variation and Linguistic Theory. Labov, William (1966). The linguistic variable as a structural unit. Washington Linguistics Review 3:4–22. Yang, Charles (2002). Knowledge and Learning in Natural Language. Oxford: Oxford University Press. 16
© Copyright 2026 Paperzz