The Evolutionary-Emergence Model: Explaining Linguistic Variation in a Transition Zone. more

The Evolutionary-Emergence Model: Explaining Linguistic Variation in a Transition Zone by Douglas S. Bigham, Ph.D. Abstract The model of language change sketched here—the evolutionary-emergence model—takes the position that language is an emergent system within which language change can be modeled as an evolutionary process applied to the emergent properties of that system. Describing language as emergent entails that the complexity of the linguistic system is built out of smaller simple processes. Though theories of language change are often juxtaposed as being in opposition to one another, the approach taken here shows that each may be considered correct, at least in part, once the separate roles and expectations for speakers and hearers are disentangled. After sketching the evolutionaryemergence model of language change in terms of previous models of language change, data taken from "emerging adults" living in a dialect transition-zone are described within the framework of evolutionary-emergence model. The data under consideration are the TRAP, LOT, and THOUGHT vowels, specifically, the relationship of TRAP-retraction to the LOT/THOUGHT merger. The data presented here show that while the presence of TRAP-retraction and LOT/THOUGHT merger are positively correlated at the community level, there is no need to find this correlation at the level of individual speakers. Introduction There are many models, theories, and suggestions as to how, and by whom, language changes. Though these theories are often juxtaposed as being in opposition to one another (e.g., functional vs. anti-functional explanations; community-based vs. speaker-based explanations), there is no principled reason why these theories could not all be considered correct, at least in part. The model of language change sketched here takes the position that to fully appreciate the many competing [ 1 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. models of language change, we must view language as an emergent system within which language change can be modeled as an evolutionary process applied to the emergent properties of that system. By calling language an emergent system we engender a way of viewing language that, I feel, is the only way to account for both the implementation of change at the level of the speaker and the continuation and organization of change at the level of the community. Describing language as emergent entails that language is a complex system overall, but that this complexity is built out of smaller simple processes. Tiny single speaker, single moment utterances which follow from and build upon past utterances, constrained by what has come before but not beholden to it. This is by no means a new view in linguistics; no one would deny that ―language‖ is built from ―sentences‖, built from ―phrases‖, built from ―words‖, built from ―phonologies‖, built from ―phones‖, built from even smaller (spoken or signed) gestural movements. But searching for language change in any one of these processes alone, without recognizing the importance of the interaction between them, is bound to fail. It is only by recognizing that each of these forces operates at a distinct level of the perception/production cognitive mechanism, and that these levels are in constant interaction with one other—providing feedback both up and down the processing mechanism—that we can begin to explain the often disparate findings regarding language change. It is only when seen as united under one emergent system that we can achieve a satisfying and complete explanation for the forces behind language change. Though this model can be readily extended for forces of syntactic change, such as grammaticalization (Hopper & Traugott, 1993), this paper deals primarily with phonetic change, drawing from, for example, Ohala’s (1993) idea of misperception, Trudgill’s (2004) notion of drift, Martinet’s (1933) notion of functional load, Lindblom’s (1990) H&H Theory, de Boer’s (2000) self-organizing vowel systems, Giles’s and Coupland’s (e.g., 1991) work on [ 2 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. accommodation, Labov’s (1972; 2001) findings regarding social class and hypercorrection, and Eckert’s (2000) work showing the indexicality of language variation and change. Overview of previous models of language change Viewing the competing theories of language change in another way, we might ask whether or not any previous models of change provide both the necessary and sufficient conditions for change. I would suggest that they do not, and many would agree with this suggestion. Thomason (2000), for example, claims that ―although it is often possible to state necessary conditions for change, it is never possible to state sufficient conditions for change‖. However, I believe that the opposite is more accurate. That is, any of the previous theories of language change contain mechanisms for change that supply conditions sufficient but not necessary for linguistic change. This echoes Lass’s assertion that ―…no change is ever necessary. If it were, it would already have happened…‖ (Lass, 1980:131). While any of the previous models and theories provide sufficient causes for language change, then, I will argue that the only force that is both sufficient and necessary for language change is that the linguistic forms of speakers be reproduced by communities—or, that hearers are also speakers. The problem with previous models and theories of language change, then, is not that they fail to accurately characterize the potential ways in which language change might happen, but rather, that none of them fail at this characterization. That is, for each of the models currently circulating in language change studies, we can find evidence that supports the conclusions of that specific model (as can be seen even in a historical linguistics introductory text, such as Campbell, 2004). As Labov (1994) points out in his discussion of the competing notions of the Neogrammarian regularity of sound change vs. dialectological evidence of lexical diffusion, ―careful consideration of the competing bodies of evidence leads to the conclusion that a higher-level theory is needed‖ (Labov, 1994:422). [ 3 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. Additionally, since ―change can occur at any and all levels of the linguistic system‖ (Thomason & Kaufman, 1988:9), any theory purporting to explain language change must be able to account for any and all levels of the linguistic system before it can be claimed to truly have described a mechanism of language change. As such, a model of language change should incorporate not only the various possible mechanisms for change (e.g., as listed above) but also be applicable the various possible levels of the linguistic system that are open to change. description for an aspect of language change. All else is simply a My suggestion for an evolutionary-emergence model of language change has three major aspects: First, the evolutionary-emergence model assumes that the object of inquiry, language, is not a stable structure or an organism, but a dynamic collection of properties which produce an emergent pattern (a similar treatment can be found in Versloot, 2008). Second, the evolutionary-emergence model recognizes the crucial importance that every level of the emergent linguistic system plays in the cause of language perception and change at every other level. In this respect, it is not enough to consider only factors of the sound system when looking at sound change, but factors of the semantic, syntactic, and social systems must be considered as well. Finally, the evolutionary-emergence model approaches language change via the principle of evolution whereby the differential success of competing variants is a result of their differential rates of reproduction and fidelity (see, e.g., Dawkins, 1982, 1989). In the end, the evolutionary-emergence model is essentially no more than an explicit formulation of King’s (1972) dictum that: It is the highly complex mechanism underlying the production and comprehension of sentences, not the units alone, that is the locus of language and language change. In other words, it is the units and the rules operating on these units in context that account for the essential facts, not only of the synchronic side of language (producing and understanding sentences), but of its diachronic aspect as well (how and why languages change). (King, 1972:928) [ 4 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. Emergence at the level of the utterance, speaker, and community The evolutionary-emergence model draws primarily from Björn Lindblom’s conception of the ―H&H Theory‖ of phonetic variation (Lindblom, 1990; see also Lindblom et al., 1995) and William Croft’s ―Theory of Utterance Selection‖ (Croft, 2000). Beginning with Croft, the Theory of Utterance Selection views each individual utterance as providing the crucial building materials for larger linguistic categories. In Croft’s terminology, these utterances are called ―linguemes‖ and the larger categories that emerge from them are the ―lingueme pools‖ (Croft, 2000:28-29). In modern society, however, we need a constraint so that every utterance one hears is not included in the lingueme pool. During the course of human evolution, small band societies were the de facto norm until very recently (Dawkins, 2004). These small bands are essentially similar to the concept in sociolinguistics of a ―speech community‖, defined as ―any human aggregate characterized by regular and frequent interaction by means of a shared body of verbal signs‖ (Gumperz, 1968:114). Croft's (2000) suggestion, then, should be specified so that the lingueme pool includes only those utterances coming from a member of one’s own speech community. This specification is supported additionally by the ―principle of density‖ (Bloomfield, 1933; cited in Labov, 2001), which claims that speaker speak most like those people most often around them. Each individual utterance by a member of one's speech community, then, contributes to the shape of the community’s ―lingueme pool‖. When a speaker hears, say, a TRAP 1 vowel, that production becomes part of her lingueme pool. She, in turn, draws from that pool when she next produces a TRAP vowel. Because speakers are also hearers, a ―feedback loop‖ develops whereby the perception of each linguistic form that goes into the lingueme pool will constrain the variants drawn from the lingueme pool for production, which in turn constrain the forms going into the pool during perception, and so on. One property of these ―feedback loops‖ in emergent systems is that stable categories will tend to remain stable while categories undergoing change will tend to continue 1 Wells’s key words will be used for phonetic classes (Wells, 1982). [ 5 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. undergoing change until a new stable point has been reached (Saperstein, 1996; Elliott & Kiel, 1996; Matthews et al., 1999). In fact, the evolution of stable phonetic systems through just this kind of feedback process has already been robustly demonstrated by de Boer in computer simulations (2000:451)2. Absent from de Boer’s simulations, however, is the inequality (both socially and phonetically) between interlocutors present in ―real-world‖ speech communities. It is this inequality that directly leads to the emergence of one form over another, or, in evolutionary terms, the ―fitness‖ of a linguistic form. For phonetic variability this inequality is described in Lindblom’s (1990) ―H&H Theory‖, though a similar speaker-hearer inequality can be found at any level of the linguistic system. In the H&H Theory, Lindblom claims that because speech is physical effort, ―default‖ speech will be produced with as little physical effort as possible. That is, speakers produce phonetic signals with only those features necessary and sufficient for accurate communication. However, an inequality arises between speakers and hearers because while speakers ―want‖ to exert as little effort as possible in speaking, hearers ―want‖ to receive the clearest possible speech signal so as to exert as little effort as possible in hearing. Because of the ways in which the phonetic level interacts with the other linguistic levels, such as lexical, semantic, and syntactic processing, this effort inequality is constantly in flux. Though speakers will ―want‖ to produce speech with as little effort as possible across situations, hearers will ―want‖ speech to be produced with more effort in situations where the linguistic signal is less recoverable from these other processing levels. This inequality that emerges between speakers and hearers also forms the basis of Labov's (1994) discussion of language change via misinterpretation and probability matching within vowel perception. Here I quote Gordon’s (2001) summary: 2 See also work by Nikolaus Ritt (e.g., Ritt, 2004) on the application of evolutionary principles to vowel systems. [ 6 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. Suppose that a token of /ɑ/ is fronted so that it appears within the usual range of /æ/. In such a case, there is a chance this item will be misinterpreted… or… not recognized … In either of these two cases, that token, because it was not identified as a token of /ɑ/, cannot influence the listener’s conception of the acoustic profile of /ɑ/…Such cases are predicted to have a conservative impact on the system, retarding any fronting of /ɑ/. If, however, the same fronted token of /ɑ/ appears after the distribution of /æ/ has moved forward and up, it is more likely to be correctly interpreted as an instance of /ɑ/… When the fronted token is correctly interpreted, it stretches the listener’s conception of the acoustic range of /ɑ/ to include variants that are more fronted. As a result of this adjustment, listeners refigure their sense of what an average /ɑ/ sounds like… [and] the mechanism serves to promote innovation. (Gordon, 2001:11-12) Though I agree with Labov’s claim for probability matching as a crucial component of language processing and change, there is much that the example of /ɑ/ fronting ignores. First, since it is not concerned with any of the linguistic processes beyond the phonetic/phonological level, this hypothetical example likely overstates the range of cases in which misinterpretation may occur. Second, it entails that language change in vowel systems can only operate by the movement of vowels into ―empty space‖ (as noted by Gordon, 2001). Therefore, the use of probability matching as described in Labov (1994) cannot function for other mechanisms of change, such as apparent push-chains or vowel mergers. Third, it predicts that in apparently linked vowel chain-shifts, each individual speaker should participate in the shifts in the same order (also noted by Gordon, 2001), or, stated in terms of the linguistic system, it predicts that the speaker-level linguistic system should mirror the community-level linguistic system. On this third point, we now have clear evidence in vowel variation research that putative chain shifts at the community level are not, in fact, always replicated at the speaker level (Gordon, 2001). Aside from the assertion that the speaker level does not matter (Labov, in Gordon, 2006), no evidence has been put forth to support this claim. In the current state of research on American English vowel variation, there exists a growing body of research showing that chain-like series of changes are [ 7 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. present at the community level but not at the speaker level and no attempt to reconcile this juxtaposition has been advanced. Framing language change within an evolutionary-emergence model, however, might lend a solution to this problem. Langstrof, in his work on vowel shifts in New Zealand English (2006), develops an evolutionary perspective on just the sort of change Labov describes above (i.e., chain shifts). In Langstrof’s reformulation of Martinet’s (1955) ―margin of security‖ concept, he shows that from an evolutionary perspective a vowel’s ―propensity of not being confused with another vowel‖ correlates directly to a vowel’s evolutionary fitness (2006:242) and that those vowels least likely to interfere with perception will be those most "fit" and most likely to survive. Returning to the /ɑ/-fronting example, Labov (1994) claims that /ɑ/ fronting will only be favored after /æ/ has moved forward and up, giving /ɑ/ a wider ―margin of security‖ for variation. In Langstrof’s terms, then, we might restate Labov’s example by saying that /æ/’s movement forward and up contributes to the positive fitness of /ɑ/. While I agree with Langstrof’s basic definition of linguistic evolutionary fitness, the fact that he states this definition axiomatically rather than deriving it from first principles brings us no closer to a unified understanding of language change. Further, since we are aware of interspeaker variation within a speech community, we must ask, after /æ/ has moved forward and up for whom—the speaker or the hearer? And with this, we return to the H&H continuum of variation. Imagine two speakers drawn from Labov’s hypothetical /ɑ/-fronting community. Speaker A, a young middle-class female, produces raised and fronted /æ/ (/æ/=[ɛ]) and concomitant fronted /ɑ/ (/ɑ/=[æ]; as explained above). Speaker B, a young middle-class male, produces an unshifted /æ/ (/æ/=[æ]) and, as such, an unshifted /ɑ/ (/ɑ/=[ɑ]). Given what we know about variation along gender lines, this is a more than plausible scenario. If Speaker A’s most common interlocutors have all raised their [ 8 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. /æ/s, a fronted-/ɑ/ (where /ɑ/=[æ]) in her speech is fine—no misinterpretation problems arise. However, what happens if Speaker A’s interlocutors include people like Speaker B, who have not raised their /æ/? In these cases, as predicted by H&H, Speaker A’s fronted /ɑ/ is no longer sufficient for communication; it sounds like an [æ] to Speaker B and is therefore misinterpreted as an /æ/; not always, but at least a not insignificant amount of times. This is the situation as described in Labov (1994). However, when we consider the implications of the H&H Theory, then when Speaker A is talking with Speaker B, we should expect Speaker A to produce an unshifted /ɑ/; she does not, after all, want to be misunderstood. Conversely, and this is the crucial point, when Speaker B is speaking with Speaker A, he is more free to vary his /ɑ/ along the H&H continuum to something more [æ]-like because this shifted form does not cause communication problems with his interlocutor, Speaker A. While it may cause problems for his own linguistic system (e.g., overlapping with his forms of /æ/), he is not the intended target of his speech and therefore any clashes his speech causes to his own system can be ignored. If Speaker B talks to enough people like Speaker A, because each vowel is an emergent category, his own /ɑ/ vowel becomes, overtime, an (ɑ) variable, with multiple forms ranging from [ɑ] to [æ]. Once these multiple forms have ―taken hold‖, the lingueme pool for /ɑ/ from which Speaker B draws will have the probability of choosing [æ]-like or [ɑ]-like forms in proportion to the number of his interlocutors who have raised or unshifted /æ/, respectively. That is, the ―choice‖ of /ɑ/ forms is directly related to both the /ɑ/ forms of the speaker’s own linguistic system and the nearest vowel forms (/æ/, /ɔ/, etc.) of the community’s aggregate linguistic system. [ 9 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. It seems, then, that the evolutionary-emergence model is able to explain both speaker-level and community-level phenomena (while likewise providing the theoretical footing to support Langstrof’s axiomatic definition of vowel fitness). Furthermore, we can now advance an account for why individual speakers appear not to exhibit putative chain shifts in the ―correct order‖—the order and degree of shifting is relative to both individual and community level lingueme pools. Additionally, the evolutionary-emergence model predicts that at incipient stages of change we should expect to find these vowel shift occurring ―out of order‖. This notion that shifts are first acquired piecemeal and only later continue advancing in the ―correct‖ order was advanced by Gordon (2001), and here an explanation of why this should be has been advanced. But does the variation we find at the level of individual speakers even matter? As Lass (1980) argues, the use of language by individual speakers is not the object of study for linguistics. Rather, the proper object of study is language at the systemic level. Labov, too, supports this notion, claiming that ―we study individuals because they give us the data to describe the community‖ and that ―there are no individuals from a linguistic point of view‖ (in Gordon, 2006:341). This is, of course, in opposition to J. Milroy’s (1992) assertion that language change is an essentially user-based phenomenon and that discussions of language variation and change must proceed from the speech of a single individuals. Is there any way in which we could reconcile these two viewpoints and explain the data? The central thesis of this paper suggests that approaching linguistic variation and change via the evolutionary-emergence model might be one such ―solution‖. Emergence at the levels of norms and fashions But what of the ―social‖ or ―stylistic‖ aspect of linguistic variation? My suggestion is that this same kind of inequality predicted at the phonetic level by Lindblom’s H&H Theory can be found at the social level of speech—the interaction between two people. Croft (2000) broaches the notion of social effects on language change within his Theory of Utterance Selection, which states: [ 10 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. …normal language change puts linguistic convention at center stage. Normal (i.e. identical) replication of linguemes in utterances is conforming to the linguistic conventions of the speech community. Altered replication of linguemes in utterances—the creation of variants—is a causal consequence of not conforming to the linguistic conventions of the speech community. (Croft, 2000:30) I would suggest, however, that both normal and altered replication of ―linguemes‖ result from conforming; the difference arises in the interplay between different interlocutors in the speech community and the changing role that members of the speech community can play at any given time (cf. Myers-Scotton, 1993). Though many sociolinguists have recently argued for finding deterministic and unintentional explanations of phonetic change before turning to intentional mechanisms (e.g., Labov, 2001; Trudgill, 2004), there is no denying the findings that phonetic variation also operates under the more ―intentional mechanisms‖ such as prestige, solidarity, indexicality, and identity-projection. I would suggest that these intentional mechanisms are not only also explained by the evolutionary-emergence model, but need no special exceptions (such as the ―performance‖ label) for that explanation. Returning to Lindblom’s H&H Theory, we have thus far been interpreting the H&H continuum as functioning only for the needs of communication (cf. Jakobson’s [1971] referential function). Lindblom explicitly defines one end of the H&H continuum as ―purpose-driven and prospectively organized‖ (Lindblom, 1990:404). Although not formulated in terms of social action, I suggest that this end of the H&H continuum can be validly interpreted as also containing those more intentional or ―stylistic‖ uses of language (cf. Jakobson’s, 1971, discussion of phatic and poetic functions 3 ). Though there are a great deal of explanations4 for the stylistic aspects of language, the common theme of these explanations can be reduced to Keller’s maxim to ―talk in such a way that you are 3 These Jakobsonian functions are also explained in Croft’s evolutionary approach to language change (Croft, 2000:73-78). However, the explanation given here is only superficially related to Croft’s position. 4 Some of which include: Bell’s (1984) ―audience design‖; Rampton’s (1995) ―crossing‖; Coupland’s (2001) ―strategic persona management‖; and Eckert’s (2000) ―indexing‖. [ 11 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. noticed‖ (Keller, 1994:101). That is, a speaker may actively and intentionally choose a linguistic form (a ―lingueme‖) that violates its category boundaries (in either the speaker’s or the interlocutor’s linguistic systems, or both) purely for effect/affect. When speakers engage in this intentional speech, sometimes they obtain the form of their target, sometimes they do not. When speakers hit their intended target, not much notice is given (they are considered adept at imitation or perhaps they have ―put one over‖ on their interlocutor, etc.). However, when speakers aim for a specified linguistic target and miss, they have engaged in hyperadaptation (Trudgill, 2004), a common form of which is hypercorrection (Labov, 1972). Originally formulated, hypercorrection is said to have the effect of promoting language change. But there is no reason this should necessarily be the case. For example, returning to our hypothetical /ɑ/-fronting community, we saw that some speakers, like Speaker A, produce /ɑ/ fronted, as [æ], while other speakers, like Speaker B, produce /ɑ/ unshifted, as [ɑ]. Although framed in terms of male/female, it could just as easily be that A-type speakers are prestigious and B-type speakers are stigmatized. When Speaker C, who comes from a conservative non-fronting /ɑ/=[ɑ] community, enters Speaker A’s /ɑ/-fronting clique, he hypercorrects and produces /ɑ/ too far front (/ɑ/=[æ<]or perhaps produces /ɔ/ as fronted as well (his original community engaged in the low-back merger [Herold, 1990], say). With enough people like Speaker C, this kind of hypercorrection indeed spurs language change. But the converse is just as possible. Suppose also that Speaker D enters Speaker A’s /ɑ/-fronting clique. Speaker D is from a non-local community where /ɑ/ is so far front that it is realized as [ɛv]. Upon entering the /ɑ/-fronting clique, he, too, wants to ―blend in‖. He therefore hypercorrects, ―misses‖ the intended /ɑ/=[æ]-target and produces /ɑ/=[ɑ] instead. With enough people like Speaker D, this would have a conservative impact on language change. Though it may very well be that the former is more common than the latter, there is nothing inherent in [ 12 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. hypercorrection (or any kind of hyperadaptation) that necessarily promotes language change over language conservativism. The point is that these intentional uses of language do not a priori have any greater or lesser impact on the linguistic system (of the speech community or the individual) than the more unintentional mechanisms of effort, interlocutor density, functional considerations, etc. They are all valid mechanisms for language change, any one of which may beget, promote, or retard language change. Once these alterations reach a certain saturation point, they may well be inevitable or, conversely, these altered forms may co-exist with original forms and enter into a kind of stable variation. On this aspect, I agree fully with Croft’s (2000) description of the evolutionary approach to the propagation of language change. Finally, by recognizing both the inherent variation in the ―lingueme pool‖ and the crucial interaction between the individual- and community-level linguistic systems, the evolutionary-emergence model can also address the actuation problem (J. Milroy, 1992; Lass, 1997). That is, while the ongoing processes of language change may be explainable, what is it that gets these processes started in the first place? But, by recognizing that language is above all else an emergent structure, the actuation problem simply vanishes. Drawing from our /ɑ/-fronting example, a common assumption when discussing the actuation problem claims that even though our hypothetical speech community currently contains both speakers who produce /ɑ/ as [ɑ] and those who produce /ɑ/ as [æ], at an earlier point in time, it must have been such that all speakers only produced /ɑ/ as [ɑ]. But this is a tacitly false assumption. Correctly formed, we must recognize that even at that past time when all speakers produced /ɑ/ as [ɑ], there would have been a variety of forms, such as [ɑ<], [ɑ^], [ɑ], [ɑv], [ɑ>], etc., as per the requirements of the H&H Theory and fully in line with the findings of variationist linguistics (e.g., work following Labov, 1963). Returning to Croft’s (2000) ―Theory of Utterance Selection‖, then, we can say that both normal and altered reproduction of linguemes in utterances are a consequence of conforming to the norms of [ 13 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. the speech community, pace Croft’s original formulation (Croft, 2000). Both normal forms and altered forms originally exist to serve to separate ―linguistic niches‖, but later become ―hybridized‖ in the lingueme pool. Though evolutionary principles of selection work as Croft described (and are therefore not rehashed here as the ―evolutionary‖ side of the model), the process of reproduction is not evolutionary, but emergent. Summary The evolutionary-emergence model, then, is not so much a new model of language change as it is as new framework for unifying what we already know, namely: The linguistic system comprised of the interaction between individual-level and community-level systems (e.g., King, 1972) is created by speakers drawing from a common source (e.g., lingueme pool: Croft, 2000) to produce variant forms; the particular form produced reflects the interplay between the distribution of the lingueme pool (e.g., probability matching: Labov, 1994), a speaker’s desire to exert as little effort as possible (e.g., H&H Theory: Lindblom, 1990), and a speaker’s desire to be noticed (e.g., Maxim of Distinction: Keller, 1994). The only force that is both necessary and sufficient for language change, then, is that the linguistic forms of speakers be reproduced by communities—or, that hearers are also speakers (e.g., Ohala, 1993). Testing the evolutionary-emergence model: TRAP-retraction and the LOT~THOUGHT merger in Southern Illinois English While the theorizing above is all well and good, the validity of the evolutionary-emergence model can only be supported through actual instances of language change. Because language change is a diachronic outcome, we must instead look for a synchronic description of a speech community wherein the findings of a speaker’s system can only be fully explained by drawing from the properties of the community’s system, not assumptions of the ―vowel system‖ in the abstract. [ 14 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. The combination of the retraction of the TRAP vowel (where /æ/ shifts to something more [a]-like) and the merger of the THOUGHT and LOT vowels (a.k.a. the low-back vowel merger; see Herold, 1990) can be found in many English dialects. In Canadian English, these features are key to the ―Canadian Shift‖ (Clarke, Elms, & Youssef, 1995), though this combination can likewise be found in Northern Californian English (Eckert, 2004). Further, though the low-back vowel merger is common in South Midland U.S. varieties of English (Labov, Ash, & Boberg, 2006), TRAP-retraction in U.S. English dialects east of the Mississippi River has not yet been reported. Gordon, relying on traditional notions about the ―margin of security‖ for vowels, has suggested that TRAP-retraction might be expected to be found anywhere the low-back vowel merger exists (Gordon, 2005). The reasoning behind this expectation is that as LOT and THOUGHT merge to a single value somewhere further back and higher than [ɑ], the ―available space‖ in which TRAP can vary is proportionally extended. Indeed, in other dialects the low-back vowel merger and retraction of /æ/ are the first two stages of a chain-shift (e.g., The Canadian Shift; see Clarke et al., 1995). The connection of the LOT~THOUGHT merger with TRAP-retraction, then, provides us with a testable hypothesis (c) from assumptions (a) and (b): (a) The amount to which the LOT and THOUGHT vowels have merged can be measured by the Cartesian Distance between them (Baranowski, 2007); (b) the amount to which the TRAP vowel has retracted can be measured by the F2 value of TRAP; (c) Therefore, a positive correlation should exist such that the F2 of TRAP decreases as the Cartesian Distance between LOT and THOUGHT decreases. [ 15 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. In order to test this hypothesis, data presented here are taken from an emerging adult (Arnett, 2000; 2001) speech community living in Southern Illinois (for a full description, see Bigham, 2008; 2010). Briefly, among Southern Illinois emerging adult speakers, we find TRAP tokens varying from a somewhat raised and fronted [æ] to a fully low and central [a] pronunciation. Likewise, the production of LOT and THOUGHT can be classified between individual speakers as being: fully merged, where there is no difference between LOT and THOUGHT classes; fully distinct, where LOT and THOUGHT classes obtain their expected phonetic values; or variable, with some tokens of LOT class taking on an [ɔ]-like pronunciation and some tokens of THOUGHT class taking on an [ɑ]like pronunciation. These three types of low-back vowel production will be called, respectively, ―merged‖, ―distinct‖, and ―occasional‖. F1 and F2 were measured for tokens of the TRAP, LOT, and THOUGHT vowels taken from word list recitation and interview speech (~35 tokens per speaker; values normalized using a variant of the Watt & Fabricius, 2001, normalization routine, see Bigham, 2008, for details). Data from Southern Illinois speakers are compared against data from emerging adults from the Chicagoland and I-55 Corridor regions of Illinois (areas that typically do not engage in low-back vowel merger nor TRAPretraction; see Labov, Ash, & Boberg, 2006). The degree of merger in the LOT and THOUGHT vowels was assessed in two ways. First, for each speaker group the LOT and THOUGHT vowels were compared in both the F1 and F2 dimensions using an unpaired t-Test comparison. As can be seen in Table 1, speakers from both the I-55 Corridor and the Chicagoland regions produce the LOT and THOUGHT vowels distinctly along both the F1 and F2 dimensions, while speakers from Southern Illinois show no significant difference between the LOT and THOUGHT vowels in either F1 or F2. Second, these statistical data are corroborated with the Cartesian Distance measure, which shows that speakers from the Southern [ 16 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. Illinois region produce LOT and THOUGHT an average of 8.8 normalized units apart (very roughly corresponding to about 30Hz), while speakers from the I-55 Corridor and Chicagoland show a greater separation of LOT and THOUGHT, at 16.5 and 21.0 normalized units apart, respectively. Finally, Table 1 presents the average F2 of the TRAP vowel for each of these three speaker groups. Although the differences between groups in the average F2 of TRAP may seem small, an ANOVA test comparing the F2 of TRAP between these three groups confirms that these differences are significant (p=.007). p-value when comparing LOT and THOUGHT Average F2 of /æ/ /ɑ/~/ɔ/ Southern Illinois .1759 .3179 8.8 107.0 I-55 Corridor .0039* .0057* 16.5 111.6 Chicagoland .0038* .0009* 21.0 115.6 Table 1: Normalized Distance of LOT~THOUGHT and average F2 of TRAP for three speaker groups F1 F2 Cartesian Distance of Summarizing, then, the data in Table 1 show that the LOT and THOUGHT vowels are merged for Southern Illinois speakers, fully distinct for Chicagoland speakers, and statistically distinct but produced more near to each other for the I-55 Corridor speakers. Regarding the TRAP vowel, we see that Southern Illinois speakers have the most retracted TRAP productions, followed by I-55 Corridor speakers, with Chicagoland speakers have the furthest front realization of TRAP. When viewed as a whole, then, these data appear to present the expected correlation between the F2 of TRAP and the distance between LOT and THOUGHT. That is, the nearer the vowels of LOT and THOUGHT are, the farther back the TRAP vowel is produced. Using a simple linear regression analysis comparing the mean Cartesian Distance between LOT and THOUGHT to the mean F2 of [ 17 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. TRAP across speakers within each group, we find that this assumption is statistically supported, though the correlation is somewhat weak (r2= .23, p=.0124, Y= -78.82 [F2 of TRAP] - 0.86* X). That is, these data appear to confirm the hypothesis that TRAP retracts as a function of the merger of LOT and THOUGHT—TRAP is most retracted for Southern Illinois speakers and least retracted for Chicagoland speakers. These data, however, are based on the aggregate values across all speakers5. When the data are viewed at the individual level, considering each speaker separately, we can see that this hypothesis begins to break down—individual speakers do not conform to the expected pattern whereby TRAP retracts as a function of low-back merger. In Figure 1, tokens of TRAP for Southern Illinois speakers have been divided into the three groups described for participation in the low-back vowel merger: distinct, merged, and occasional. 5 Southern Illinois speakers: 3 males, 4 females; Chicagoland speakers: 5 males, 4 females; I-55 Corridor speakers: 4 males, 5 females. [ 18 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. Figure 1: Southern Illinois emerging adult speakers’ TRAP tokens; split according to participation in the low-back vowel merger. As can be seen, the most retracted tokens of TRAP are produced by those speakers who have merged the LOT and THOUGHT vowels, while more of the un-retracted TRAP tokens are produced by speakers who either maintain distinction between LOT and THOUGHT or only participate in the low-back merger occasionally. This is in line with the original hypothesis and the findings from the community-level data. However, we can also see that this is by no means a categorical distinction— speakers who engage in the low-back vowel merger can produce TRAP tokens un-retracted (even slightly raised and fronted), while even speakers who do NOT participate in the low-back vowel merger may produce a retracted TRAP vowel. This disjunction within individual speakers, where TRAP may be well retracted even while LOT and THOUGHT remain fully distinct, can be seen quite clearly when looking at the vowel productions of speaker klb42241, a male from Southern Illinois, as seen in Figure 2. Here we see that speaker [ 19 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. klb42241 shows extensive evidence of TRAP-retraction while still maintaining distinction between LOT and THOUGHT. If TRAP-retraction were solely based on the margin of security generated by LOT~THOUGHT merger, how is it that this speaker can produce retracted TRAP without causing a homonymic clash with LOT? For klb42241 the supposed ―margin of security‖ does not appear to exist. My suggestion is that this apparent violation can be explained by disentangling the individual and community level (or speaker and hearer phonetic systems) via the evolutionary-emergence model. Figure 2. Speaker klb42241, average and individual tokens of TRAP, LOT, and THOUGHT vowels. Average values are represented by large connected symbols; individual tokens are represented by small symbols. [ 20 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. Conclusion The first part of the evolutionary-emergence model concerns the level of the phonetic signal as it is expressed by a recognizable ―community‖ of speakers. For the discussion of the TRAP vowel, this can be adequately represented by F1 and F2 values. Out of the aggregate of these /æ/ values, an /æ/ category emerges for the speech community. This community-level vowel category is expressed, for example, in data that draw from mean values pooled across individuals. However, because individual speakers encounter their interlocutors in somewhat different ways—such as different social circles, different families, different frequencies of common interlocutors, etc.—we can assume that the emergent /æ/ category for each individual speaker will be slightly different. These individual level vowel categories are expressed in, for example, data from individual speakers. Because individual speakers in Southern Illinois are more likely than not to participate in the lowback vowel merger (as shown in Table 1), Southern Illinoisans will be communicating with interlocutors who, more likely than not, engage in the low-back vowel merger. Therefore, as per the evolutionary-emergence model, the TRAP vowel for Southern Illinois speakers will have more room to vary in the F2 dimension because it will (more often than not) not interfere with interlocutor perception. Consequently, it should be more common than not to find Southern Illinois speakers who produce retracted forms of TRAP, regardless of whether these ―retracted /æ/‖ speakers participate in the low-back vowel merger themselves, as can be seen in Figure 1. Therefore, we see that by revising Gordon’s (2005) hypothesis through the use of the evolutionary-emergence model, we can explain both the strong correlation between TRAP-retraction and the LOT~ THOUGHT merger at the community level and the disjunction between TRAP-retraction and the LOT~THOUGHT merger among individual speakers. [ 21 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. In many respects, the evolutionary-emergence model does not diverge much from Croft’s (2000) description of the lingueme pool. However, because the evolutionary-emergence model crucially considers the difference in interaction between interlocutors' roles as speakers and interlocutors' roles as hearers, the explanatory power of the model is fully realized, and the distinction between individual- vs. community-level language shifts can now be reconciled. Though I have provided one example of how the evolutionary-emergence model sketched here can be used to explain real-world data, much more data are needed before anything as strong as a unifying theory might be developed. However, by the outline of the evolutionary-emergence model given here we may now return to the passage from King that we began with, where we find that ―language change is anything but a paradox; it is rather the predictable consequence of alterations of the mechanism that combines, and modifies in context… forms expressed in basic units.‖ (King, 1972:929). References Arnett, J.J. (2000). Emerging adulthood : A theory of development from the late teens through the twenties. American Psychologist, 55(5), 469-480. -------. (2001). Adolescence and emerging adulthood: A cultural approach. New Jersey: Prentice Hall. Baranowski, M. (2007). Phonological variation and change in the dialect of Charleston, South Carolina. Publication of the American Dialect Society (PADS), 91. Durham, NC: Duke University Press. Bell, A.J. (1984). Language style as audience design. Language in Society, 13(2), 145-204. Bigham, D.S. (2008). Dialect contact and accommodation among emerging adults in a university setting. Unpublished doctoral dissertation, University of Texas at Austin: Austin, Texas. Available online at http://dsbigham.net/files/Bigham-2008-dissertation.pdf. Bigham, D.S. (2010). Correlation of the Low-Back Vowel Merger and TRAP-Retraction. Penn Working Papers in Linguistics, 15(2). Buss, D., (ed.). (2005). The handbook of evolutionary psychology. New Jersey: John Wiley & Sons, Inc. Campbell, L. (2004). Historical linguistics: An introduction. (2nd ed.). Cambridge, MA: MIT Press. [ 22 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. Clarke, S., Elms, F., & Youssef, A. (1995). The third dialect of English: Some Canadian evidence. Language Variation and Change, 7, 209-228. Coupland, N. (2001). Dialect stylization in radio talk. Language in Society, 30, 345–375. Croft, W. (2000). Explaining language change: An evolutionary approach. London: Longman. Dawkins, R. (1982). The extended phenotype: The gene as the unit of selection. Oxford and San Francisco: Freeman. Dawkins, R. (1989). The selfish gene. (2nd ed.). Oxford: Oxford University Press. Dawkins, R. (2004). The ancestor’s tale: A pilgrimage to the dawn of evolution. New York: Houghton Mifflin. deBoer, B. (2000). Self-organization in vowel systems. Journal of Phonetics, 28, 441-465. Eckert, P. (2000). Linguistic variation as social practice. Oxford: Blackwell. -------. (2004). California Vowels. Radio interview on All Things Considered. Feb. 24, 2004. Retrieved April 1, 2008, from: <http://www.stanford.edu/~eckert/vowels.html>. Eckert, P. & McConnell-Ginet, S. (1992). Think practically and look locally: Language and gender as community-based practice. Annual Review of Anthropology, 21, 461-490. Elliott, E. & Kiel, L.D. (1996). Introduction. In L.D. Kiel & E. Elliott, (eds.). Chaos theory in the social sciences: Foundations and applications. University of Michigan Press. 1-18. Giles, H. & Coupland, N. (1991). Contexts of accommodation: Developments in applied sociolinguistics. Cambridge: Cambridge UP. Gordon, M.J. (2001). Small-town values and big-city vowels: A study of the northern cities shift in Michigan. Publications of the American Dialect Society, (PADS), 84. Durham, NC: Duke University Press. -------. (2005). The Midwest and West. In E. Schneider, (ed.). Handbook of Varieties of English: The Americas and Caribbean, Vol I: Phonology. Berlin: Mouton de Gruyter. 338-350. -------. (2006). [Interview with William Labov]. Journal of English Linguistics, 34(4), 332-351. Gumperz, J.J. (1968). The speech community. International encyclopedia of the social sciences. London: MacMillian. Herold, R. (1990). Mechanisms of merger: The implementation and distribution of the low back merger in eastern Pennsylvania. Unpublished doctoral dissertation, University of Pennsylvania. Hopper, P. & Traugott, E. (1993). Grammaticalization. Cambridge: Cambridge University Press. Jakobson, R. (1971). Linguistics and poetics. Selected Writings II. The Hague: Mouton. 18-51. [ 23 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. Keller, R. (1994). On language change: The invisible hand in language. London: Routledge. King, R.D. (1967). Functional load and sound change. Language, 43(4), 831-852. -------. (1972). [Review of Communication Complexes and Their Stages by Hans-Heinrich Lieb]. Language, 48(4), 928-929. Labov, W. (1972). Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press. -------. (1990). The intersection of sex and social class in the course of linguistic change. Language Variation and Change, 2, 205-254. -------. (1994). Principles of linguistic change: Vol. 1: Internal factors. Oxford: Blackwell. -------. (2001). Principles of linguistic change: Vol. 2. Social factors. Oxford: Blackwell. Labov, W., Ash, S., & Boberg, C. (2006). Atlas of North American English. Paris: Mouton de Gruyter. Labov, W., Yaeger, M., & Steiner, R. (1972). A quantitative study of sound change in progress. Philadelphia: US Regional Survey. Ladefoged, P. (2000). A course in phonetics (4th ed.). Oxford: Blackwell. Ladefoged, P. & Broadbent, D.E. (1957). Information conveyed by vowels. Journal of the Acoustical Society of America, 29(1), 98-104. Langstrof, C. (2006). Vowel change in New Zealand English: patterns and implications. Ph.D. thesis, University of Canterbury. Lass, R. (1980). On explaining language change. Cambridge: Cambridge University Press. -------. (1997). Historical linguistics and language change. Cambridge: Cambridge University Press. Lindblom, B. (1963). Spectrographic study of vowel reduction. Journal of the Acoustical Society of America, 35, 1773-1781. -------. (1990). Explaining phonetic variation: A sketch of the H&H theory. In W.J. Hardcastle & A. Marchal, (eds.). Speech Production and Speech Modeling. Dordrecht, the Netherlands: Kluwer Academic Publishers. 403-439. Lindblom, B., Guion, S., Hura, S., Moon, S-J., & Willerman, R. (1995). Is sound change adaptive? Rivista di Linguistica, 7(1), 5-37. Martinet, A. (1933). Remarques sur le système phonologique du français. Bulletin de la Société de Linguistique de Paris, 34, 191-202. Matthews, M.K., White, M.C., & Long, R.G. (1999). Why study the complexity sciences in the social sciences? Human Relations, 25, 439–461. [ 24 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins. Milroy, J. (1992). Linguistic variation and change. Oxford: Blackwell. Milroy, J. & Milroy, L. (1985). Linguistic change, social network and speaker innovation. Journal of Linguistics, 21, 339-384. Milroy, L. (1980). Language and social networks. Oxford: Basil Blackwell. -------. (1987). Language and social networks (2nd ed). Oxford: Blackwell. Myers-Scotton, C. (1993). Duelling languages: Grammatical structure in code-switching. Oxford: Clarendon. Ohala, J.J. (1993). Sound change as nature’s speech perception experiment. Speech Communication, 13(1-2), 155-161. Rampton, B. (1995). Crossing: Language and ethnicity among adolescents. London: Longman. Ritt, N. (2004). Selfish sounds and linguistic evolution: A Darwinian approach to language change. Cambridge: Cambridge University. Saperstein, A.M. (1996). The prediction of unpredictability: Applications of the new paradigm of chaos in dynamical systems to the old problems of the stability of a system of hostile nations. In L.D. Kiel & E. Elliott, (eds.). Chaos theory in the social sciences: Foundations and applications. University of Michigan Press. 165-198. Thomason, S.G. (2000). On the unpredictability of contact effects. Estudios de Sociolingüística, 1(1), 173-182. Thomason, S.G. & Kaufman, T. (1988). Language contact, creolization, and genetic linguistics. Los Angeles: University of California Press. Trudgill, P. (1986). Dialects in contact. Oxford: Blackwell. -------. (2004). New dialect formation: The inevitability of colonial Englishes. Edinburgh, Scotland: Edinburgh University Press. Trudgill, P., Gordon, E., Lewis, G., & MacLagan, M. (2000). The role of drift in the formation of native-speaker southern hemisphere Englishes: Some New Zealand evidence. Diachronica, 17(1), 111-138. Versloot, A.P. (2008) Mechanisms of language change: Vowel reduction in 15th century West Frisian. Ph.D. dissertation, University of Groningen. Utrecht: LOT publications. Available online at: www.lotpublications.nl/index3.html Watt, D. & Fabricius, A. (2002). Evaluation of a technique for improving the mapping of multiple speakers' vowel spaces in the F1~ F2 plane. In D. Nelson, (ed.). Leeds working papers in linguistics and phonetics, 9. 159-173. Wells, J. (1982). Accents of English. Vol. 1. London: Cambridge: Cambridge University Press. [ 25 of 25] Bigham, Douglas S. (to appear). In: De Vogelaer, Gunther & Guido Seiler (eds.). The dialect lab: using dialects as a testing ground for theories of language change. Amsterdam/Philadelphia: John Benjamins.