M.A. Thesis 2014 for program:
Linguistics of European Languages: English
The Many Faces of /t/:
An argument against /t/-dropping
and lenition in English.
-Wesley D. Tucker
Universiteit van Amsterdam
Supervisor: Dr. Silke Hamann
Table of Contents 1. Introduction 2. /t/-dropping
2.1. The Process and Defining the term 2.2. Alternate Realizations
3. On Lenition
3.1. Definitions and Hierarchies 3.2. Redefining the term 4. On Weakening
4.1. Sonority, Intensity, and Duration 4.2. Effort
5. Context
5.1. Word Position 5.2. Stress 6. Allophones
7. Lenition versus Allophones 7.1. Infrequency and Variability 7.2. The Path toward /Ø/ 8. Other Factors
8.1. Language and dialect contact 8.2. Contact-influenced change in English 9. The Experiment
9.1. Set-up and method 9.2. Possible Results 9.2.1. Possible Result A 9.2.2. Possible Result B 9.3. Additional Discussion 10. Conclusion 11. References 1 2 2 3 6 7 10 13 14 16 17 17 19 21 23 24 25 27 27 28 30 31 32 32 33 33 34 36
1. Introduction
Languages change. This fact is not disputed; it is the theories and processes of language change which are under scrutiny. The theories and processes which point to language change phenomena include those of semantic, syntactic, and lexical bases. This paper will focus on phenomena in the phonetic and phonological categories, specifically ‘lenition’ and its usage surrounding /t/ in all varieties of English (also referred to as ‘/t/-dropping’). In America (the United States and Canada) very few studies have been conducted concerning the geographical diaspora of the various pronunciations and even fewer concerning the exact motivation for the phenomena. A common British example would be the pronunciation of little [lɪtl̩ ] as [lɪʔl̩], and similarly is the American alternative better [beɾəɹ] in place of [betəɹ]. Preceding a discussion of the term “lenition,” I will argue the term ‘/t/-dropping’ is both incorrect and misleading. Lenition is the favored explanation for the phenomenon, while other research leans in favor of an explanation that is not lenition-based. I will argue that the alternate pronunciations of /t/ should be treated as allophones. In order to work with such a well-known, linguistic term, a definitive meaning of ‘lenition’ must first be constructed. However, “Theory-specific definitions model lenition differently […], and phonologists sometimes even disagree about which types of segmental change or process instantiate the concept” (Honeybone 2008: 11). I will first describe the phenomenon and its realizations in Section 2. Also discussed will be the articulatory movements which define the various realizations (/ʔ/, /ɾ/, etc.). Then section 3 will accumulate the uses of ‘lenition’ in modern research, and possible explanations for its occurrence, as well as presenting the varying circumstances which bring about the change. This last element will be discussed in reference to lenition hierarchies in Section 3.1. Section 4 will define the articulatory characteristics of lenition, including sonority, intensity, duration, and effort. Section 5 will end the basic discussion of lenition by better explaining contextual causes of lenition. Section 6 will introduce and discuss allophones. By the end of Section 7, I will have debunked the theories of /t/-dropping being a case of lenition, preferring to favor theories of multiple /t/ allophones. Section 8 will briefly discuss language and dialect contact that may have had an impact on lenition processes in English. Section 9 will contain a small-sized experiment proposal, which will help to back up the allophonic claim and discredit any lenition theories on the matter. The discrimination task experiment will challenge participants to categorize given item sets, which contain nonsense English words with /t/ and alternate
/t/-realizations, as the same or different. However, this experiment is as faulty in its possible results as is the discussion on lenition in relevant literature—further recognizing the confusion previous research has attempted to overcome.
2. /t/-dropping
Research into /t/-dropping is prevalent when looking at well-known and clearly attested examples of lenition in major world languages. Moreover, it has been documented in every major dialect of English: from the British varieties of Scottish (Dyer 2002), Irish and Scouse (Honeybone 2012; Hickey 1996) to individual variances among the many American accents (Kenyon 1994; Eddington and Taylor 2009). What exactly causes the phenomenon will be left for section ##. This section will focus on describing /t/-dropping in its many observed forms.
2.1. The Process and Defining the term
The term ‘/t/-dropping’ has been used to describe any process performed by a speaker of English which, when pronounced, reduces the traditional pronunciation of /t/ to a different form. These different forms are often referred to as weaker to /t/ in some abstract phonological sense or as easier to pronounce from an articulatory standpoint. Section 4 will focus on what exactly accounts for the idea of weakness in phonology.
One large study by Eddington and Savage (2012) on the specific dialect of Utah speakers in the United States was concerned with the different realizations of /t/ among speakers. In particular, Eddington and Savage (2012) focused on determining the production rates of two very different sounds when speakers were choosing the same realization of /t/ in words with a t followed by a syllabic nasal. The realization was a glottal stop [ʔ] (to be discussed below in section 2.2), but how the glottal stop was released had two variants. First was a nasal release, in which the speaker would release the glottal stop directly into the following nasal sound; for example kitten as [kɪʔn]1.The second was an oral release, as in [kɪʔən]. More clearly: “Oral releases appeared as a
discernible vowel between the glottal stop and nasal…In contrast, nasal releases transition from the glottal directly into the nasal” (2012: 340). At the time of Eddington and Savage’s (2012) study, the Utah accent had a reputation for a higher rate of /t/-realization as an oral release.
1 All instances of phonetic transcriptions in this paper were based on the Cambridge Online Dictionary: dictionary.cambridge.org
On the topic of the term /t/-dropping, Tucker (2013) argues that “Eddington and Savage were mistaken in the usage of their term “/t/-dropping,” which suggests that the phoneme is not only weakening, but weakening to such an extent as to be completely lenited (/Ø/)” (p. 4). While ‘lenition’ is sometimes seen as a trajectory towards elision, which is omission of the actual acoustic realization, this claim by Eddington and Savage (2012) would seem to imply that eventually words like mountain and kitten, and any other words with a /t/ followed by a syllabic nasal, would either already or are eventually going to “drop” the /t/ altogether, leaving such pronunciations as [maʊnɪn] and [kɪən]. This assumption is incorrect. The /t/ is still present in the items; it has just been realized in a different way. “I would suggest to the linguistic community that “/t/-dropping” as a term for the lenition of the /t/ phoneme needs to be change.” (Tucker 2013: 4). Tucker suggests the use of the term ‘/t/-leniting’ to refer to the phenomenon in place of ‘/t/-dropping.’
2.2. Alternate Realizations
The English phoneme /t/ is a common sound in most languages. Although it is categorized as a coronal stop, it may vary between dental and alveolar. It is produced by a quick movement of the tongue against the teeth or alveolar ridge with concurrent, unvoiced, airflow. The English /t/, as argued by many researchers, has undergone processes which alter this known pronunciation. The main alternate realizations I will focus on in this paper are the flap /ɾ/ and the glottal stop /ʔ/. First let’s look at the flap. Sometimes referred to as a tap, the flap is a quick articulatory gesture in which the tongue is thrown against the alveoloar ridge. There are two common descriptions of this gesture. Lavoie (2009) writes “ When a /t/ is realized as a flap, there is decreased articulator contact such that the flap is often fully voiced and there's no seal behind which pressure is built up" (p. 37). So one definition of a flap involves the lack of air pressure behind the contact of articulators and the other the duration of said contact. So a flap is barely, if at all, plosive, and it can result in voicing. Common examples of a flap include whatever [wɒɾevər] and at all [æɾɔl]. “[It] is relatively common at word boundaries, and to some extent within high frequency words such as British” (Ashby and Przedlacka 2010: 3-4), i.e. [brɪɾɪʃ]. Ashby and Przedlacka however forget to include a key detail on the flapping of /t/ in specific word positions; while common at word boundaries, word initial placements, such as tall and teeth, are not commonly lenited, even when preceded by a vowel. This intervocalic element will be better defined in section 5.1.
There are two other elements of flapping which point toward explaining the phenomenon as lenition. First, as mentioned above, flapping can result in voicing. Voicing, as we will see later in section 3.1., is another sign of lenition. Secondly, “[Flapping] is lenition as it is a temporal reduction in the articulation of a segment. As an uncontrolled ballistic movement it is shorter than the alveolar stops which it replaces" (Hickey 1996: 2). This idea of reduction will be addressed in section 4.1.
As for the source of flapping, in American English the flap “[…] originally arose from the weakening of /t/ and /d/ in certain contexts, it is now, arguably, a distinct articulatory target” (Hualde 2011: 2230). Research has proven the flap to be a chosen realization in the American dialects due to “Speakers [showing] awareness of its existence as a different sound” (p. 2230). So as it may not have been a phoneme in the original inventory of American English, it has developed over time to be a particular segment which is produced by speakers in certain contexts.
Next is the glottal stop. The process of /t/ /ʔ/, also called glottalization, “is a type of lenition in which the oral gesture of a stop is removed” (Eddington and Taylor 2009: 298). Contrary to other consonant gestures, glottal stops are formed by the obstruction of airflow with the glottis. Glottalization has been seen as a predominantly British variant with common examples such as little [lɪʔl̩ ] and better [beʔə]. But Eddington and Taylor (2009) point out that this may be due to the different environments in which /t/ lenites to /ʔ/ in British versus American informal speech: “The idea that glottals are nonexistent in American speech may come from the abundance of glottals in British English prevocalically […] where American varieties tend to have a flap” (pp. 298-299); compare the British better [beʔə] to the American [beɾər]. Tucker (2013) found a similar result when comparing the rates and contexts of flaps and glottal stops between British and American English dialects. So one could not claim that the phenomenon of glottalization as a whole is completely British since it occurs—albeit in different environments—in American English as well. The difference in American and British environments which permit lenition will be discussed later in section 5.1.
One additional element of glottal stops that must be mentioned is the two distinct types of glottals: implosive and ejective. Implosive consonants are articulated by a lowering of the glottis, an absence of air pressure, and an ingressive or stationary airflow (Clements 2002); this is contrary to egressive glottal stops which are produced by an outward airflow and. The glottal stops of
English are specifically egressive, as implosive glottal stops are virtually absent from Indo-European languages (Clements 2002). This topic will be addressed again later in section 8.2.
An important element of the process /t/ /ʔ/ is that it has often been labeled “stigmatized.” Glottalizing /t/’s has been “generally associated with low prestige varieties […], although in Cardiff, Wales, it is more common in higher social classes” (Eddington and Taylor: 2009 298). It is because of this variance one cannot claim glottalization is socially stigmatized across the board—it varies by locale. When stigmatized, /t/ /ʔ/ is avoided by speakers, especially in certain circumstances under which they are able to monitor and edit their speech moment by moment, such as reading texts allowed. For instance, in their “Reading Task” experiment, Eddington and Taylor (2009) follow this thought; they write, “a drop in glottalization rates when reading would indicate that people are self-conscious about their use of glottal stops” (p. 309). However, their data earlier in the study provide sufficient evidence to claim that “glottalization is on the rise in U.S. English” (p. 307). For example, female participants of the study were much more frequent in their use of /ʔ/ in place of /t/, and since women tend to lead sound change, this is a valid argument for the steady rise in the phenomenon of glottalization. The “Reading Task” study corroborated this idea. Subjects glottalized items 55.8% of the time. “This high degree of glottalization in a task that tends to elicit very formal language clearly demonstrates that the use of the glottal is not stigmatized” (p. 310)—at least it is not stigmatized in the region from which this study was taken (Utah) or in the regions from where the subjects originated (predominantly Western states). Interestingly, Eddington and Savage (2012) found a similar bias against glottalized /t/’s, but the participants of the study were actually confused as to which realization was truly the stigmatized version. This social aspect is elaborated in section 7.1.
Other than the stigma associated with the nasal/oral release dichotomy in Utah, social discouragement of glottalization has been commonly found in Scottish dialects like Glaswegian and varying locales across England; Milton Keynes (Dyer 2002) is an example. So even though there is a social attitude which discourages the glottal realization, it is still quite prevalent across the Anglophonic world. Additionally, the speakers associated with t-glottalization varies on the locale. Stuart Smith (1999) writes that “T-glottalling tends to be associated with male working class speech, although not always” (p. 183). The fact that the phenomenon is not always based in the speech of a particular group simply goes to show that it is a present characteristic of English no matter the dialect.
The above glottalization process in one dialect—Utah of the American states—is only one element that divides the many dialects of American English, and there are arguably even more variances between dialects on the other side of the Atlantic. If we take the alternate forms of /t/ as separate happenings, we can come upon many instances of flaps (/ɾ/) in American English, with glottalization (/ʔ/) occurring more frequently in the west than the east. The flap is usually marked as a more American pronunciation, while the glottal is more British. But this is not an inflexible rule, and the occurrences need to be observed with such a wide range of other circumstances (i.e. stress, word position, etc.) that there is little use in stating such broad claims.
3. On Lenition
What is lenition? The term, used widely over the last hundred years, has referred to a wide range of phonetic and phonological processes which alter the realization of a consonantal sound. For the purposes of this paper, only plosives will be considered as for valid discussion due to their articulations which share factors with English /t/; vowel reduction will be excluded from discussion due to their obvious difference in articulation and their associated processes of sound change. The changes involving plosive consonants are best displayed as a series of IPA symbols; for example, the hierarchy of /t/: /t/ > /d/ > /ʔ/ > /Ø/. Other plosives (i.e. /p/, /k/, etc.) have been shown in similar environments to display successful actuation of lenition. However, what exactly constitutes “successful lenition” is a large topic of debate. There is a disagreement among researchers as to what exactly counts as lenition; for instance, is it the element of the segment in question which is changing or the motivation behind such a change that qualifies the phenomenon as ‘lenition’? Honeybone (2012) also notices this variance: “some see lenition as the loss of segmental material; others as an increase in sonority; still others try to link lenitions to an increase in ‘ease of articulation’” (p. 2). In this section, I will discuss lenition hierarchies and present the ideas of various researchers who attempted to define the term.
3.1. Definitions and Hierarchies
“It is common to claim that 'lenition' is a synonym of 'weakening' in phonology, and this clearly implies a notion of consonantal 'strength'. (Honeybone 2008: 9). This idea of ‘consonantal strength’—and consequently ‘weakness’—is difficult to define when it comes to language change without addressing the hierarchies which have been proposed by various researchers, especially
when hierarchies in fact simply demonstrate the observed phenomena, and do not explain why such changes occur—more on this ‘why’ factor later.
Venneman’s definition for phonological weakness can be found in multiple sources, most often cited by Hyman (1975); it has been applied by Kirchner (2001), Honeybone (2012), and Bauer (1988) among others. This definition is as follows: "a segment X is said to be weaker than a segment Y if Y goes through an X stage on its way to zero" (Hyman 1975: 165). This definition has been widely accepted due to its flexibility and applicability to any sound segment in any environment and only assumes one thing: the end result of lenition is elision. One disclaimer must be added, however: if the process is not halted somewhere between the initial realization and its complete elision, which could happen for any number of reasons, lenition of a segment X can result in zero. The issue with this assumption is also noticed by Szigetvári (2008): "we can call the change lenition if the segment is eventually going to disappear, that is to tell whether a change is lenition or not presupposes that we know the future fate of the given segment" (p. 101). This is problematic because explaining a phenomenon does not necessarily mean the goal is to predict future similar phenomena, or again that all processes of lenition will result in a /Ø/ segment.
However, one method of attempting to explain the phenomenon of ‘lenition’ is to set up a lenition hierarchy, which displays the many processes through which a segment may ‘weaken.’ Honeybone (2008) seems to prefer the term ‘trajectory’ over ‘hierarchy,’ and here I must agree. ‘Hierarchy’ implies that one segment may have authority over another or possible that one is better than another. The less problematic use of ‘trajectory’ implies only direction—that direction being toward elision—and this falls in line with the definition given by Venneman above.
One issue with lenition trajectories is that they “cannot be used for any other purpose than what they were devised for, which greatly decreases their explanatory value" (Szigetvári 2008: 101). However, this purpose is very useful when we observe similar happenings across language systems. If /p/ lenited to /b/ in one language, then perhaps it will happen in similar environments in another language; or the reverse: "since we frequently see that the change Y > X is a step in a trajectory leading to the loss of Y in the history of a number of languages, we consider the same change lenition in the particular language we are analysing" (p. 101). This implies that predictive methods can be used in lenition studies. The difference between the above statements is whether or not a case of lenition in one language can be applied to another, or is it language specific? This could be why some instances of “obvious” lenition which appear to be lenition on a certain
trajectory have been assumed to be such, when really the phenomenon in question is a separate occurrence. In order to grasp the applicability of a certain trajectory, we must understand how such a trajectory is designed and how segments are aligned in their supposed order of strength and/or weakness. Both concepts are intertwined and must be addressed simultaneously.
The phonological elements of a sound segment can be described in many ways, including ‘strong’ and ‘weak.’ Honeybone (2008) lists a set of processes which he qualifies as being associated with ‘weakening;’ these are “spirantisation, approximantisation, gliding, debuccalisation, voicing and vocalization” (p. 16-17); similarly, Kirchner (2001) writes of ‘weakening:’ “[…] the core idea […] is some reduction in constriction degree or duration” (p. 3), and then lists the processes included under the term lenition (reproduced below).
degemination, or reduction of a long consonant to a short one (e.g. t: t);
flapping, or reduction of a stop to a flap (e.g. t r);
spirantization, or reduction from a stop (or affricate) to a fricative or approximant continuant (e.g. t {θ, θ̞}
[approximantization,] reduction of other consonants to approximants (e.g. r ɹ, s s); debuccalization, or reduction to a laryngeal consonant (e.g. t ʔ, s h); […]
complete elision (e.g. t Ø) (Kirchner 2001: 3).
The process of voicing differs from the processes above because it “[involves] an adjustment in laryngeal specification rather than reduction of constriction” (Kirchner 2001: 3); it is, however, considered one of the lenition processes because the occurrence of voicing relies on some of the same contextual cues and a closer examination of the articulations remains in line with the reductions of the above processes (Kirchner 2001). An example of voicing would be a shift from /t/ to /d/. Vocalization is another process of lenition in which a consonant is reduced to the form of a vowel or semi-vowel, such as ‘old’ /oʊld/ /oʊd/ (Petyt 1985). The final process of lenition to be discussed here, gliding, will be discussed further in section 7.2. It is interesting to note the differences between each process and that in describing each process one can come across the term ‘reduction,’ or in other words ‘weakening.’
While a sense of ‘weakening’ may be apparent among each of these processes, any measurement of ‘strength’ here could be confused due to the differences in the manifestation of the phonetic elements of each process. Foley (1977: 144) writes, “phonological strength: reflects
the unequal relation among phonological elements. It does not refer to the phonetic strength of the phonetic manifestation of the phonological element, but rather simply to their abstract relation.” This is to say that different phonetic segments, even if described implicitly as stronger than another segment, do not in fact overlap in their manifestations. For instance, a voiced stop and an approximate appear on trajectories—presented below—to be directly related in phonological strength. However, the voiced stop is stronger than the approximate, not because their shared phonetic characteristics (such as duration and intensity) display this difference in strength, but because their behaviors in similar environments and the resulting realizations can be compared via sonority, articulatory effort, and other qualities. So trajectories cannot be considered phonetic or phonological law; they are the illustrative display of observed phenomena and possible future changes that may occur therein. In other words, phenomena are observed and charted in relation to the behavior of other processes. For instance, "if plosives spontaneously change into fricatives, as is often claimed, then fricatives are claimed to be weaker than plosives" (Honeybone 2008: 19); hierarchies are then set up to “represent lenition trajectories, that is, they show the route that a segment will take if it lenites" (p. 13).
The above explanation could be seen as a disclaimer for the following hierarchy (Figure 1), so that readers will not take the presented image as proven fact. It is simply the visual representation of lenition as created by one researcher. Figure 1 is the most widely used hierarchy, by Lass (1984).
Figure 1 Lenition/fortition Scale by Lass (1984: 187)
As mentioned above, the focus sound segment of this paper will be /t/, and it follows that any other examples used to explain the phenomena will also be plosive. Stop consonants have been a popular choice when observing the behavior of sounds with lenition in mind. Lavoie (2009)
writes, "It's not surprising that stops are the stars of weakening: they show off the steps of weakening very clearly. Starting from an articulatory structure with a closure and a release, as they lenite, they show numerous easy-to-label intervening stages" (p. 41). In Figure 1 we can see these stages. A chosen segment can begin in any position on the trajectory and has two to three possible paths, depending on that position, toward /Ø/; note here again I do not write “a path to /Ø/” implying this is where all processes of lenition end. The movement is simply toward elision on this trajectory. Sticking with our example set of /t/: /t/ > /d/ > /ʔ/ > /Ø/, the voiceless stop /t/ starts in position 5a, a less sonorant and voiceless stop, then (in this situation) moves to a voiced stop /d/ (5b), to a fricative (3b), and to /Ø/ (1). This example is important in that it recognizes that stages on the scale can be skipped. Also, other processes could occur which could lead to other paths being taken. For example, a voiceless stop (5a) could aspirate (4a) and proceed through to a glottal fricative (2a); or /t/ > /ts/ > /h/.
3.2. Redefining the term
For ‘lenition,’ the extensive above information does not begin to outline a useable term, leaving readers to discern a true definition. Harris’ (1990) work is, similar to this thesis, irritated by a lack of certainty; he writes that "one researcher's lenitions frequently turns out to be another's fortition" (p. 257). This is problematic. Precise theories and terms in a soft science such as Linguistics are of utmost importance because the evidence is unlikely to be truly tangible in nature. Considering the speculative nature of the science, the theories which are crucial to research-central and academic discussions cannot be so unclear in that researchers differ in their opinions on the matter. A phenomenon with blurred lines like this can prevent further research from being carried out clearly. This disagreement needs to be remedied.
Ashby’s and Przedlacka’s (2010) use of the term ‘lenition’ includes ‘weakening.’ They write: “[lenition] refers to one or a number of phonological processes, which render a consonant weaker, the weakening in question being that of strength of resistance to the airflow” (1). Here we find that ‘lenition’ can be seen as a large category of phonological processes, all of which consistently result in a weakening of a segment; what exactly constitutes weakness will be addressed in the next section. Honeybone (2012), like Ashby and Przedlacka (2010), recognizes the range of processes that could be included under the umbrella term ‘lenition.’ “One central aim of lenition theory has been to formally unify these types of change as the ‘same kind of thing’” (p.
1). And in order to unify these processes, one would need to recognize their similarities. For example, each is concerned with a whole class of segments that have changed or are in the process of changing in the same articulatory way. So then one way to see ‘lenition’ is as a set of processes which affect a language’s sound system, and though each may differ slightly in their journey from segment Y to segment X, each process is markedly similar because they each result in altered— ‘weakened’—realizations of segments.
This idea of strength and weakness is consistent throughout lenition literature. Watson (2002) equates lenition “with the reduction of strength of a phonological segment" (196). For this researcher, ‘lenition’ is the actual process of reduction, not an over-arching label for a group of processes. Here is where the lines of ‘reduction’ and ‘weakness’ tend to blur—the concept of ‘reduction’ being synonymous with ‘weakness’. If, in fact, ‘lenition’ is concerned with both of these terms—being both ‘reduction’ in the articulatory factors and referencing processes that affect ‘weak’ segments—then a couple questions may be asked: (1) in what ways and to what end can a segment be ‘reduced’? (2) What constitutes weakness and, consequently, nominates a segment to be eligible for reduction?
If one were to exclude the problematic terms of ‘weakening’ and ‘reduction,’ perhaps a less abstract view of the topic could be solidified. Hickey (1996) has done this: “Lenition is a phenomenon which is common in many languages which usually manifests itself as a shift from stop to fricative or a shift from voiceless to voiced with obstruents” (p. 173). The operative word here—‘shift’—implies change but does not leave any mysterious undertones which could point toward a possible motivation. ‘Shift’ is focused on the result, which for the current discussion is useful in that it does not muddle us with intangible ideas. Hickey does, however, imply that there is a general direction of the ‘shift:’ stop becomes fricative, voiceless becomes voiced, as addressed in the above section on hierarchies. This is why Hickey (1996) is a good example on the topic of lenition because his working definition is much more inclusive.
There is a need, in this paper as well as for ongoing research on the subject, to develop a definition that is both coherent and sufficiently covers all aspects of the phenomenon. Honeybone (2008) addresses this need: "We can safely conclude […] that 'lenition' is now standardly assumed to be the same thing as phonological 'weakening', and that the concept groups together a smallish set of processes” (p. 16-17). This follows the suggestion that the idea of 'lenition' is more than just
a term for a single phenomenon but a concept which includes many terms for processes in language change.
So, a possible definition, from the information provided thus far, is:
Definition (1): Lenition is both an independent term for a phenomenon in sound change and an umbrella term for a set of phonological processes, all of which result in ‘weaker’ or ‘reduced’ realizations of sound segments.
This definition is problematic due to its ability to apply to two very different processes, one of a synchronic nature and the other diachronic. Hualde (2011) writes of this difference in methodology: "The goal of the diachronic analysis of a phonological pattern is to discover how it developed through time. A synchronic analysis of the same phenomenon, on the other hand, may be concerned with providing a succinct and precise statement of the facts and/or with modeling speakers 'knowledge of the phenomenon in question" (1). This difference in synchronic and diachronic studies is understood across the field of linguistics. However, it can be confusing to use a term like ‘lenition’ (among others) which differs in its application depending on the research in which it is being used. Referring back to our definition (1), the first part, “an independent term for a phenomenon in sound change,” refers to lenition as a diachronic change, which can be made up of many sequential, synchronic changes. These synchronic changes—the second part of the definition, “a set of phonological processes”—are rules which affect, at first, a single speaker, but spread throughout a language population.
The following sections will deal with the extensive diachronic processes which alter segments, resulting in ‘weaker’ or reduced forms, as well as the synchronic processes of the sounds as they sit in the minds of the individual speaker. Thus, for the purposes of this paper, our working definition stands as:
Definition (2): Synchronic lenition is a general term for a set of phenomena that are described and categorized as phonological processes, all of which result in ‘weaker’ or ‘reduced’ realizations of sound segments in the minds of individual speakers or the system of a specific dialect. Diachronic lenition aims to explain the series of synchronic lenition processes over time.
Thus, a segment, such as /t/, which has been proposed to undergo processes that can each be seen as separate phenomena and be labeled as such (i.e. voicing then frication then glottalization then elision), can be assumed to have been in one of the many processes of synchronic lenition. And research regarding the origin of the phenomenon and its journey over time refers to diachronic lenition.
4. On Weakening
But what exactly constitutes the ideas of ‘strength’ and ‘weakness’ when it comes to sound segments? What exactly sets researchers up to layout the trajectories from section 3.1.? Two terms, referring to the auditory realization of the segment or the amount of effort associated with a particular sound that appear frequently in lenition resources are ‘reduction’ and ‘weakening.’ However, some linguists consider these terms separate (i.e. Hickey, 1996), while others define one with the other (i.e. Escure, 1977; Hualde, 2011). Watson (2002) barely attempts a definition, preferring to move past definition debates and writing, "It is sufficient for this paper to describe lenition as a process generally equated with the reduction of strength of a phonological segment" (p. 196). This is not a clear definition at all. One must ask, while considering both the element of strength and the trajectories from section 3.1., created by linguists via observing phenomena and developing said trajectories, what qualifies these segments to change in the first place? To address this issue I will refer back to Honeybone (2012) (numbering is my work): “some see lenition as (1) the loss of segmental material; others as (2) an increase in sonority; still others (3) try to link lenitions to an increase in ‘ease of articulation’” (p. 2). If we look at the trajectories of lenition presented above as the directional stages leading toward the elision or deletion of a specific segment, we have already accounted for (1). The following sections will address the other two possibilities separately.
4.1. Sonority, Intensity, and Duration
Sonority refers to the intensity and the quality of a sound segment to be resonant. If we consider ‘intensity’ to be separate for this moment, ‘resonant’ means that a sound segment uttered without obstruction is more sonorant, which would place the category of ‘obstruents’ quite low on a scale of sonority because of the articulatory strictures involved that interrupt or ‘obstruct’ the flow of air and, consequently, sound.
Like those of lenition, hierarchies of sonority have also been developed by various researchers. Lavoie (2009) has collected multiple scales of proposed sonority hierarchies, and writes that a “[…] consonant strength hierarchy mirrors the ordering of a sonority hierarchy” (p. 33). So to get an idea of a sonority hierarchy one can simply consider a lenition trajectory (Figure 1) in reverse. However, one caveat must be mentioned; sonority and lenition hierarchies do not
traditionally include all of the same categories in which to place sounds. To compare the inclusion of descriptive categories and observe the ordering the processes of lenition. Lavoie’s (2009) compilation of sonority hierarchies has been reproduced in Figure 2 below, “[…] with the segment of the highest sonority at the bottom of each column” (p. 32).
Zec (1995) Clements (1990) Jespersen (1904) Venneman (1988)__
Obstruent Obstruent Voiceless plosives
Fricatives Voiced plosives
Voiced stops Voicelss fricatives
Voiced fricatives Voiced fricatives
Sonorant Nasal Nasals, laterals Nasals
Liquid Lateral liquids
r-sounds Central liquids
Glide
High vowels High vowels
Vowel Vowel Mid vowels Mid vowels
Low vowels Low vowels
Figure 2: Compilation of sonority hierarchies (Lavoie 2009: 32)
Again, the main difference between Figures 1 and 2 are the categories included, as the ordering is almost identical. This is understandable considering Lass’s (1986) lenition trajectory is meant to account for the many processes which may occur, while Figure 2 is much more focused on the grouping together of similar sounds. For example Figure 2 includes vowels and their three possible placement locations as separate categories, which are irrelevant for a hierarchy of consonant lenition.
But if we take the general layout of the scales to be the same, the Lass trajectory from section 3.1. (Figure 1) can be viewed as a sonority scale as well as a scale of lenition in general. In fact, the y-axis in Figure 1 is labeled sonority. So sonority was included in the thought process and data collection for developing this hierarchy. However this is problematic because Figure 1 only displays sonority as having two levels. While it is true that voiced stops (5a) and are less sonorant than voiceless stops (5b), the former cannot so simply be considered equal to affricates, fricatives, and approximates—the entire second row. Lass’s (1984) scale of lenition only allows for a two-leveled understanding of sonority, which is grossly over-simplified.
Setting aside resonance, we can now discuss intensity. “Lavoie (2001) and Parker (2002) both found intensity to correlate with sonority. So, increased consonant intensity can be taken as
evidence of weakening” (Lavoie 2009: 33). This is saying that an “increase” in the intensity of a particular sound segment could be a marker of its current weakening in progress and possibly its eventual elision. This adds yet another element to consider when looking at Figure 1; a trajectory of lenition could as well be seen as both a hierarchy of sonority and a hierarchy of intensity.
Last we have ‘duration.’ Duration is the actual measured time it takes for a sound to be initiated and ceased. One could assume then that a shorter consonant is closer to being lenited due to its length, meaning its actual time to be pronounced is closer to not being pronounced at all. However, the opposite has proven to be true. “Lavoie (2001) found that American English phonologically voiced consonants are consistently shorter than voiceless” (Lavoie 2009: 36). So between voiceless and voiced consonants (positions 5a and 5b in Figure 1), voiced consonants are shorter than voiceless; consonants with a longer duration are actually farther along in the lenition process than the shorter forms. It follows that regardless of its sonority or intensity, the actual time it takes to pronounce a consonant could be an indicator of its ‘strength.’ The shorter the time it takes to pronounce the segment, the stronger and less prone to further lenition it is. Interestingly, some studies such as Cole and Cooper (1975) note that the duration of a fricative could be a major element in the minds of the listeners in deciding if the sound had been voiced or not. Shorter durations were perceived as voiced even if the speaker did not intend to voice it.
Our running example of /t/ applies here. The difference between /t/ and /d/ in their sonority is the voiced characteristic of /d/ that is missing from /t/. So, /t/ is less sonorant than /d/, and if we consider intensity, again /d/ is seen to be more intense than /t/ (Lavoie 2009). In all, while the terms are not identical in meaning, the clear correlations of ‘sonority,’ ‘intensity,’ and ‘duration’ are not to be ignored.
4.2. Effort
The idea that lenition is related to articulatory effort is not new. Some researchers have termed this element of the lenition phenomenon as ‘laziness,’ which is clearly a problematic word for a science. ‘Laziness’ differs from the other problematic terms above (like ‘strength’ and ‘hierarchy’) because it does not refer to the actual sound segment or its position. Applying the ideas of laziness and effort are difficult, again like lenition, because there is hardly a clear definition of what effort actually is. This is due to such a large variety of factors within which a segment may fluctuate. For example, “the actual effort involved in a given utterance may vary
with speech rate, loudness, the size of the speakers jaw, the amount of air in the speaker’s lungs, the presence of chewing gum in the speaker’s mouth, etc.” (Kirchner 2001: 29). When pronouncing a sound, a speaker is activating many mechanisms with many changing variables. And what concerns us is what exactly in these variables is considered by the speaker to involve less effort in producing them. In Kirchner (2001) an attempt is made “to provide an explicit, physically-based notion of articulatory effort” (p. 30). Neurologically and muscularly speaking, it is the call and response between nerve impulses and muscle fibers which results in some form of a contraction or tenseness. And these exchanges between the brain and muscles are repeated by speakers and become second nature. So over time “[it] is therefore plausible that, as speakers acquire experience in making articulatory gestures, they develop knowledge of the effort required for those gestures, and can therefore anticipate the effort required to produce any given set of gestures” (Kirchner 2001: 30). This is the reason why articulatory effort is viewed differently between dialects and languages. For an L2 speaker of a language to perfectly perform the gestures as a native speaker is complicated and arguably impossible, not to mention difficult to measure. Thus comparison of the effort needed to produce different phonemes is difficult. It follows that even in one’s own language to say which gesture requires more effort than another can be similarly difficult.
Attempts have been made by researchers to quantify the effort for a particular gesture by means of neuromuscular estimates of force and from more biomechanical perspectives. It is my opinion that the key factor in measuring and comparing effort of two gestures is force—and that is force throughout a gesture, not at one particular point, taking into account velocity and time. And there have been elaborate logarithms developed on this idea which include variables of rest position, timing, and magnitude. Due to the wide range of strictures the human vocal tract can manipulate, forming a specific gesture can vary. And any slight alteration of each of the aforementioned elements can have extreme consequences on the phonetic realizations.
In sum, effort is not a characteristic of acoustic realization which is ideal for measuring because observing and recording such measurements of effort cannot effectively be operationalized. This is not only due to the neurological and muscular effort exerted but also, because the effort of any focus phoneme would be difficult to clearly separate from its context. A particular target phoneme is not limited to only the segment in question because where the gesture begins and ends (i.e. the preceding and following environments) can cause slight alterations to said
target. This is why context, in the next section, is another crucial element of effort in lenition and lenition in general.
5. Context
Elements that fall outside the realm of auditory realizations like that of duration and intensity can also suggest the idea of ‘weakness’ and ‘strength’ in phonological segments of language. This section will discuss the context of a sound segment. It is widely believed that not only are some sound segments inherently weaker (i.e. a fricative versus a stop), but also the space the segment occupies in relation to the word or phrase of the utterance can largely affect the probability of the sound being lenited. The context of each sound, including the stress and the quality of the preceding or following segments can be deciding factors on the realized form of the sound during production. Honeybone (2012) puts it this way: "[…] weak positions don't promote lenition—they just inhibit it less than strong ones" (p. 10). But what exactly qualifies as a weak position?
5.1. Word Position
One of the strongest positions—meaning lenition is seen least in said position—noted by researchers (i.e. Honeybone 2012) is word-initial position. Most instances of /t/, and other plosives for that matter, are prevented from leniting if they occur at the beginning of a word. This also applies to word-internal, syllable initial plosives that do not follow a vowel. For example, in addition to word initial /t/ items like ten and tremor, the /t/ in elliptical would not be likely to lenite because it follows the closed syllable /lɪp/ which ends in a consonant. There is little else to be said about consonants in word initial or syllable initial (post-coda) positions due to the overwhelming agreement between researchers and concurrent observations that lenition simply does not often occur in these positions.
Hickey (1996) writes of environmental factors on lenition that "Weakness in Irish English is defined by position relative to stressed nucleus […] and secondly by the sonority values of flanking segments" (p. 4). These qualifying criteria for a possibly lenited segment are based in its word position as well as the quality of the neighboring segments (we will address stress in section 5.2.). Szigetvari (2008) agrees “that different environments promote different types of lenition, movement along different lenition trajectories" (p. 112). To argue his case he presents yet another
hierarchy designed by Escure (1977); this time it is an environmental hierarchy for lenition (reproduced below).
(1) Final
(a) V-C ## or VC -##
(b) V-C #
(c) V-#C
(d) V-##
(2) Intervocalic
(e) V-V
(f) V-#V
(g) V#-V
(3) Initial
? ##-V
Figure 3: Environmental Hierarchy by Escure (1977: 58)
Escure labels three overall categories which refer to the position of the segment in question. Each level of the hierarchy implies a segment’s level of possible lenition, or as Escure puts it “deletion,” in certain positions, with the first level (Final) being more prone to lenition than the second (Intervocalic) or the third (Initial). Additionally, the subhierarchies (a - ?) are listed in order of favorability to lenite. As an example, the hierarchy will be applied to three English words: cat,
later, and tin. Taking only the /t/ phoneme into consideration, cat /kæt/ represents a word final
consonant or position (d) in Figure 3. It is more likely to lenite than the other two examples and could result in many realizations depending on which process is occurring from Figure 1. The t in
later /leɪtər/ represents position (e) on the hierarchy, and the intervocalic /t/ could result in similar
realizations to those found in cat. Lastly, the t in tin /tɪn/ is the least likely to experience one of the lenition processes because it falls in the final row of Escure’s hierarchy, word initial position. Escure (1977) also remarks that the scale is implicational, in that lenition in the “weaker” or higher levels is a precondition for lenition in “stronger” or lower positions. She writes: "Intervocalic deletion presupposes final deletion: all those dialects which retain intervocalic consonants also retain final consonants" (p. 59). So if a lower level example were to occur in one dialect of English, for example if the t in later were to lenite, then it can be assumed that the t in cat will as well.
But Szigetvari (2008) has multiple problems with this context-specific hierarchy. Szigetvari writes “Escure’s collection of environments […] ignores the fact that consonant clusters cannot be treated on a par with single consonants” (p. 112). Also, “adjacent consonants may “protect” their neighbours against lenition […], or they may “promote” the deletion of their neighbours […]” (p. 112). The former, findings of environment-specific restrictions found by other
researchers before her, is not taken into consideration by Escure; and the latter, more recent research, would need to be considered in an updated version of an environmental hierarchy for lenition. An attempt at creating an environmental hierarchy for English, or any language for that matter, would be a large undertaking, worthy of its own thesis; so this paper will not attempt to do so. However, the factors that relate to word position are so key in understanding the processes that are occurring, we must consider them in the present discussion. Of specific interest is not only word position but the relative position of t to the stress of the word.
5.2. Stress
The stress patterns of individual languages vary greatly, and the patterns of stress throughout a phrase, interacting with the relative strength or weakness of a particular segment in a particular position can radically influence the possibility of the segment to lenite. And because unstressed syllables are commonly seen to be weaker than stressed ones, a segment located in an unstressed position only enhances the possibility of cuing the process. However, it is interesting to note that not only does stress play a part in the possibility of a segment leniting at all, but stress can also affect which of the set of processes included under the term ‘lenition’ will be cued. Watson (2002) momentarily touches on the variance in the stress of a particular lexical item and how that can change the realization of the segment in question, for example permit /`pərmɪt/ the noun and
permit /pər`mɪt/ the verb. Watson writes to observe the occurrence of the processes /t/ /h/ and
/t/ /ts/ in Liverpool English. The /t/ /ts/ process refers to a segment which is released as a fricative but actually began as a plosive, which is a move from position (5a) to position (4a) on Lass’s (1984) scale, Figure 1; a common example of an affricate in English is ch /tʃ/. Due to the difference of stress, Watson (2002) found that it is more likely for a speaker of this Scouse dialect to realize the final t of permit depending on the stress of the item. If the t appears finally in the stressed syllable it is more likely to be realized as a /ts/; and if the t appears finally in the unstressed syllable it is more likely to be realized as a /h/: /`pərmɪh/ and /pər`mɪts/. By looking at Figure 1, we can see that the affricate /ts/ would be stronger than /h/, and it follows that the stronger segment would appear in the more strongly stressed syllable. So the ideas presented in Lass’s (1984) scale and Watson’s (2002) analysis of the Scouse dialect are in agreement.
In Irish English, there is a tendency for /t/ to be pronounced as an affricate as well. Hickey (1996) provides us with pairs of words in Irish English which display this process in conflicting
environments, for instance compare tight /`taits/ and titanic /tai`tænɪk/. This is an interesting set of examples because from them “one could argue that the distribution of stops and fricatives is determined by position in a word” (3). However, if we compare Italy [`ɪtsəli] and Italian [ɪ`tæliən] it is clear “that position to the vowel of the stressed syllable is the determining factor” (3). In Italy the /t/ is affricated because it follows the stressed vowel; this frication disappears when the stress follows the /t/. Hickey (1996) concludes that syllable coda position is the only position in which /t/ can be lenited. This is different for Honeybone (2012) who found possible affrication in word-initial and foot-word-initial positions in Liverpool English; in order for /t/ to be glottalized or flapped, it must occur stress-internally or in in a word’s coda. However, it must be said again that comparing dialect-specific rule environments will not be without its complications.
Many of the American dialects of English share a similar relationship between flaps and a word’s stress. The flap was conventionalized in American English dialects in many words and phrases, both word internally like better [beɾəɹ] and across word boundaries like at all /æɾɔ:l/. But this process has additional rules based on the presence and lack of stress. For instance, Hualde (2011) writes, “across word-boundaries […] flapping is not conditioned by stress and may occur immediately before a stressed vowel, as I in ate apples; this is also the case in a compound such as whatever” (16). Hualde delves deeper into this phenomenon by attempting to discover its root cause. He finds: “it seems reasonable to assume that after [ɾ] was conventionalized as a replacement for [t] before unstressed vowels it was analogically extended to the context before stressed vowels when word final” (16). So here, stress played an important factor in the initial occurrence of the process but also in its development over time into other environments.
6. Allophones
The above sections define lenition as a term to explain the “weakening” of /t/ to other realizations in English. Another possible explanation for the variation of /t/ realizations is based in allophones. Allophones are different phonetic realizations that are recognized by speakers as the same as the underlying phoneme. In order to understand the behavior of allophones, a few terms must first be explained. First to be discussed herein is the difference between overlapping (contrastive) distribution and complementary distribution. Complementary distribution refers to sounds which are never found in the same environment. Yavaş (2006) concerns himself mostly with the Spanish example allophones [d] and [ð] (i.e. ‘dedo’ /deðo/); “[ð] occurs between two
vowels or after a nasal, [d] occurs in the remaining environments” (p. 33)—a perfect example of complementary distribution. Overlapping distribution, on the other hand, is most easily displayed by minimal pairs, which “are pairs of words that have exactly the same sounds in the same order except for a single difference in sounds, and have different meanings” (Yavaş 2006: 32). English examples of this are fight and night, bed and bad, and tease and team. These word pairs share most of their phonemes in the same order and differ only in one phoneme; those pairs of phonemes are said to be in overlapping distribution because they occur in the same word position. If this overlapping distribution is found then the segments in question are not allophones of one phoneme. However the above information is not enough to pick out all allophones of a language because one crucial detail is missing. The phonetic features of each segment must be taken into consideration as well. This is because “different changes can and will be stimulated by different environments” (Yavaş 2006: 34). For more on contexts and environments, look to section 5. Yavaş (2006) attempts at creating guidelines for discerning phonetic similarity. For obstruents, Yavaş is sure to include voicing, manner of articulation, and place of articulation as features which must be considered as possible allophonic groups. At least two of the three previous features must be found between sounds for them to be considered allophones. for example: “Pairs of sounds with same voicing and place of articulation but different manner of articulation (e.g. [t-tʃ], [k-x], [p-ɸ])” (p. 35). The English [ʔ] allophone, for example, is similar to /t/ in that they are both unvoiced stops, but their placement of articulation is different.
So, again, in order for a set of sounds to qualify as allophones they must be able to be described as follows: multiple phonetic realizations in complementary distribution and non-contrastive in meaning. As an example, if [b] is an allophone of [p] in a language and [b] only appears in intervocalic position (which commonly stimulates voicing), it can be concluded that the two sounds are most probably allophones of the same phoneme. This is because they fulfill the requirements of complementary distribution and having similar phonetic features (Yavaş 2006). With this information it would be understandable to confuse allophonic variation of a phoneme for lenition because contextual motivation is such a large factor for both. For English, it is a difficult task, as seen above in section 5, to discern concretely where exactly /t/ will be replaced by another realization, and this is mostly because the environments for each allophone differ between dialects.
Vayaş (2006) compares the phonological systems of English, Spanish, and Malayalam (a Dravidian language). Malayalam speakers are completely aware of a difference between an
Formatted: Font: Italic
alveolar and dental nasals, while English speakers do not make a phonemic distinction. This is because the two sounds are allophones in English but separate phonemes in Malayalam. Similarly, “In English [d - ð] is contrastive, but it is allophonic is Spanish” (Vayaş 2006: 35). So the alternate forms, or allophones, of one phoneme are categorized as equal in the phonological system of speakers of a specific dialect, and when one of those forms is chosen, it is recognized in its underlying form. For example, if a speaker of the Utah American English dialect (from Eddington and Savage 2012) hears a /ʔ/ or /ɾ/ (where a /t/ usually occurs), they understand this segment as a /t/. The allophonic scheme of this example is seen below in Figure 4:
/t/ [ʔ] / Vn_V
[ɾ] / V_V
Figure 4: Allophonic scheme for Utah American English /t/
This scheme is the representation of the idea that /t/-lenition in English is actually a process of allophonic variation. The alternate realizations of /t/ are allophones and not a process of sound change whereby /t/ is ultimately changing. So an allophonic view of the processes of /t/-lenition is that /t/ is not actually being reduced or changing at all. If the realizations were being understood as different from /t/ underlyingly, then it could be said that the segment in question is being recategorized by listeners as a separate sound. However, in the phonological system of English, /t/ remains a distinct phoneme, and the allophones of /t/ are still very connected to the underlying form of /t/. In other words, there is no threat to [ʔ], [ɾ], and the others from being recategorized as a different phoneme. An example of recategorization is found in Hualde (2011). Hualde indicates phonological recategorization as the cause for the Latin word lupu wolf to change to lobo in Portuguese. If we set aside the change in vowel, the change of the intervocalic stop /p/ /b/ was at first a gestural reduction that was then conventionalized in casual speech. So /b/ was one allophone of /p/. Over time, the new realization was “no longer attributed to underlying /p/” (Hualde 2011: 10). Phonological recategorization of /ʔ/, /ɾ/, /ts/, etc. separate from the English phoneme /t/ does not appear to be in process, but additional research would be needed to prove this statement. Another area of additional research that could help us to understand this phenomenon would be to discover to what extent the allophones of /t/ are recognized as /t/; in other words, if the speakers of English are aware of the alternate forms (which seems to be the case) and
Formatted: English (United States)
can consciously alternate between realizations in any context, it would support the idea that the allophones are in no way being recategorized. However, these are only a couple of the many ideas that oppose lenition and support allophones as a more probable explanation for the varying realizations of /t/. The next section will deal with this argument.
7. Lenition versus Allophones
The fact that there is equal possibility that /t/-lenition is a phenomenon based in the processes of lenition or allophonic theories is not something that can be cleared up with one thesis. However, it is the purpose of this paper to lay the two theoretical perspectives side-by-side and choose the more probable occurrence. This section aims to present and discuss both ideas.
7.1. Infrequency and Variability
The instances of supposed lenition in English have been observed and explained in many different ways. Davidson (2011), for one, writes:
“Both phonetic surveys and textbook descriptions of American English often make the claim that stops are frequently unreleased in American English when preceding either an obstruent or nasal consonant or a pause at a phrasal boundary (i.e. napkin; They crack nuts) […]. Yet, there is some evidence that stop release in these environments does occur, albeit infrequently and variably” (p. 1042).
However, it is this infrequency and variability that needs a closer look.
For variability we will look to Davidson’s (2011) study in which American English native speakers were asked to pronounce a set of “69 target words containing a stop-stop or stop-nasal sequence in word-internal position” (p. 1051). The study was not restricted to the instances of /t/ and instead included any stops (/p t k b d g/) that may be lenited in any cluster (e.g. /bd/, /pk/). Examples of included tokens are milkman, rugby, and elliptical. Davidson found that “approximately 90% of the realizations of stop releases in word-internal pre-stop and pre-nasal position are divided between glottalized (29%), unreleased (29%) and released variants (30%). The remaining tokens were either quasi-released (7%), lenited (4%) or deleted (1%)” (2011: 1053). Two things are clear from this quote: (1) the term “lenited” is included in a set of labels whereby the tokens could be categorized, adding to the confusion of the term and its misuse since the other labels could be included under the umbrella term “lenition” (this was addressed in section 3.2);
and (2) all stops in English can show some form of lenition depending on the position and context of the sound segment.
If lenition is an ongoing process with the ultimate goal of elision and all stops in English can display one process of lenition or another, then it is simply enough to state that lenition as a goal-oriented concept is not possible. So one could surmise that the current realizations of any stop in all positions where the “traditional” pronunciation is not favored are still leniting. This means that any pronunciation of a word with a word-medial or word-final /t/ that is not pronounced as a dental or alveolar stop is still progressing toward elision. If all stops in English were to eventually lenite completely (elide), then all stops in the language would be reduced. Additionally, if all fricatives, approximates, and the other categories of consonants are constantly in different processes of lenition, language could be completely reduced to /Ø/—an extreme statement to say the least.
Referring back to the Utah study from section 2.1., Eddington and Savage (2012) found a dichotomous variance which was prominent in one particular U.S. state. Interestingly the socially stigmatized form was in actuality the more produced form and was perceived differently by the listeners; this is again referring to the difference in the release of a glottal stop as nasal or vocal (i.e. [kɪʔn] versus [kɪʔən]). With the traditional pronunciation of the items also possible (kɪtən), this is similar to Davidson’s (2011) study in that multiple realizations for the “same” phonological sound was discovered and recorded. However, an important element of this study which was not discussed was the comprehensibility of each realization. If, for instance, the glottal stop with a vocal release occasionally led to misunderstandings between speakers, this could subconsciously discourage the usage of this particular pronunciation. However, since no record of misunderstandings were occurring, one could ask if this is simply the rise of a particular pronunciation or an increase in the rate of lenition of the sound segment? If this were a case of lenition, then one form would appear to take precedence over the other. A longitudinal study of the same topic could help to support the lenition theory, because if the lenited forms (or preferably one specific form) was to be documented as rising in popularity and /t/ was falling out of fashion, the idea of lenition in this situation could be supported. However, because this Utah variance includes the rise of multiple, varying forms which do not result in misunderstandings, this could be considered an allophonic variation—the speakers produce different acoustic sounds but they are phonologically the same in the focus dialect.
7.2. The Path toward /Ø/
As mentioned in section 3.2, if lenition is a process wherein the end goal is complete elision, then it could be fair to assume that the end state of all lenited segments will eventually be /Ø/. This is a very difficult statement to support due to the wide variance in the behavior of the sound segment in different contexts. For instance, in English the phoneme /t/ is realized in multiple forms depending on word position, stress, and other factors; however, the traditional pronunciation is still found, especially in word initial position. It would be a much easier claim to make that certain segments of particularly common phrases or lexical items could lenite much easier due to the habitual and repetitive articulatory gestures the speaker is making. Of course it is possible for lenition to happen in one environment but not in similar environments, because no one has ever claimed that lenition is lexically abrupt in its spread through a language. Two well-known example of this would be the English phrases want to and going to, which are traditionally pronounced /wɒnt tu/ and /ɡoʊɪŋ tu/ but are also commonly pronounced as wanna /wənə/ and gonna /ɡənə/. So in these particular phrases the segment has been reduced to /Ø/ but it is not an across-the-board process due to the pronunciation of the phrases sent to and moving to as /sent tu/ and /muvɪŋ to/. Since the pronunciations of /sənə/ and /muvənə/ are not possible here and in fact the variants /senʔ tu/ or /muvɪŋ ɾə/ are more likely, an allophonic view of the change is preferred here.
Let’s look at one realization in particular. The glottal stop does not appear to be weakening further. Glottals are never listed in a high position on a lenition trajectory or sonority hierarchy. And considering the lack of a orthographic symbol in written language for a glottal stop, it is fair to assume that the glottal stop has always been (in English at least) considered as an allophone of another segment. Ashby and Przedlacka (2010) analyzed tokens of teenage British English speakers from three different dialect regions of England. In looking at the glottal stops of these participants, they observed and measured the durations of the segment in question. They found that “the glottal event is relatively localizable in time, and has a duration commensurate with that of conventional consonantal segments” (8). In other words, the glottal stop is showing no further signs of elision in the positions where it was measured. So, if a certain sound segment is said to be part of a lenition trajectory, yet the specific segment does not appear to be weakening further when under scrutiny, could it not be assumed that the process of lenition (if present in this situation at all) has halted?
If we consider for a moment Figure 1, we can see that there are many different paths a particular sound can take wherein the final stage would be /Ø/. Watson (2002) focuses on the processes and the particular path that /t/ has taken in Liverpool English (t > ts > s > h > Ø). “Each of the realizations possibilities (and a few others) […] are found as allophones of /t/ in Liverpool English, conditioned by different phonological environments” (197). So while there are plenty of examples of a /t/ being realized as a /ts/, there are other instances where /s/, /h/, and /Ø/ also appear. This corroborates with Ashby and Przedlacka (2010) because neither claim all instances of /t/ are leniting. They point to different realizations in different contexts with little to no evidence of lenition (still) being in progress. It is interesting to note that while Watson approaches the Liverpool accent with the assumption that it is lenition, he also mentions allophones; this is the first researcher to hold a divided opinion.
Ashby and Przedlacka (2010) include another realization of /t/ found nowhere else in the research at hand. This is the previously unheard of allophone /j/. This sound is called a glide, which is a segment “in which the tongue and lips move during the production” (Coxhead, 2012), like the English ‘you’ /ju/. In observing speakers on various British public television channels, a Yorkshire speaker and a Welsh speaker were recorded in separate instances pronouncing /t/ as a /j/ in intervocalic, word-medial position; examples included exciting [ɪksajɪn] and relegated [relɪgeɪjɪd]. This palatal glide has been sput in place of /t/ in order to separate the diphthong that would be created by side-by-side vowels if the /t/ were to be completely elided. It is as if the choice of /j/ instead of /t/ as an “intervocalic stop articulation is a planned articulatory movement” (Ashby and Przedlacka 2010: 5). And if the articulation is, as they say, “planned,” then it could be assumed to be one of many choices the speakers could have made in order to articulate the desired item— again, an allophone.
Other interesting examples Ashby and Przedlacka (2010) provide of alternate lenition realizations are Portobello [pɔ:beləʊ], Battersea [bæsi], and better [bɛə]. At first these pronunciations may seem a little too far reduced, so that, for instance, better as [bɛə] could be misheard as bear or bare. “Impressionistically, it appears that […] the lenition of the intervocalic glottal does not proceed all the way to zero but the presence of the underlying stop is signaled by creaky voicing” (p. 6). Creaky voicing has been researched significantly in recent years from many different angles, such as pathological, phonological, conversational, and pragmatic) (Benoist-lucy and Pillot-loiseau, 2013). Specifically large topics of discussion focus on how creaky voicing
