Phonological weakness is a familiar concept in historical linguistics, but it is an ill-defined one. This volume brings together papers treating the concept from various perspectives, including the theoretical and the analytic, the phonetic and the phonological, the diachronic and the synchronic, in an attempt to shed light on what weakness is and how it is manifested in the history of English. Phonological weakness has been a central concern in the work of the volume's editor over many years, most memorably in Minkova Reference Minkova1991, Reference Minkova2003, and so the collection seems a fitting outgrowth of theoretical advances that she has played a notable role in introducing to the field of English historical linguistics. She expresses the hope (p. 6) that the book will become a standard work of reference on the topic; there is every reason to think that it will be that.

The book is divided into four parts, the first of which, ‘Phonetic and phonological aspects of weakness’, begins with ‘Treatments of weakness in phonological theory’, in which Kie Ross Zuraw outlines the various ways that weakness has been defined and accounted for in phonology. To avoid the circularity of defining weakness and weak segments interdependently, it has been widely maintained that a weak segment is diachronically intermediate between another segment and Ø: for example, since Old English final [ɑ] becomes Middle English [ə], which is lost in Modern English, [ə] may be regarded as weak. Degree of constriction, resulting in loss of airflow as it increases, seems to be a reliable indicator of relative weakness in regard to most obstruents, but it is less effective as regards sonorants and vowels. Other, less reliable factors include constriction duration, acoustic amplitude, complexity of articulatory gestures, and the treatment of weakness itself as an abstract phonological property, comparable to a phonological feature. Reduction of articulatory effort as an explanation for weakening is often invoked, though acoustic indistinctness is also a plausible explanation for certain deletion effects. The distance traversed by articulators in the transition between adjacent segments appears to play a role in differentiating why some segments are more subject to weakening in a particular environment than others; allegro/lento speech and register also correlate to weakening effects, as does prosodic structure. Certain kinds of weakening correlate, as well, to token frequency, for example deletion of the second vowel in every, as opposed to mammary.

Lisa Lavoie takes up some of the same issues in ‘Testing consonant weakness phonetically’, noting that weakness is difficult to define because it may be inherently segmental, positional, historical, or language-specific in nature. Considered in turn are four main views of weakening: (1) as an intermediate point on the way to deletion (as above); (2) as a decrease in the duration and magnitude of articulatory gestures; (3) as an increase in sonority; and (4) as a decrease in articulatory effort. These views are similar in many ways, but they make different predictions about the effects of weakening: for example, the ultimate target in (3) is a low vowel, in (4) a neutral one. Measurable features that correlate with weakness within accepted sonority hierarchies are intensity (the average root mean squared amplitude), duration (rather than voicing, which is harder to measure and may be simulated in percept tests by manipulating duration alone), degree of articulator contact, and qualitatively judged manner of articulation, the last of which, as applied to American English obstruents, reveals considerably greater variation than might be expected. The measurable characteristics of obstruent weakening discussed are presence/absence of stop bursts, of aspiration and frication, and of buccal closure, with remarks on the bias in the relevant literature toward stops, though weakening is also measurable in fricatives and sonorants. Although some types of weakening, such as decreased duration, are readily identifiable under conventional views of lenition, others, such as the association of frication with weakness, are called into question.

At the start of part 2 of the book, titled ‘Diagnosing and interpreting English schwa(s)’, Roger Lass offers a learned challenge to orthodoxy in ‘On schwa: synchronic prelude and historical fugue’, in which he argues that ‘reduction to [ə]’ as a description of the development of unstressed vowels in the history of English is fraught with difficulties. In the ‘Prelude’ he discusses the variable phonetic nature of schwa in Hebrew (the language that is the source of the concept and of the name in Western linguistics); the considerable disagreement among phoneticians, over the past century and more, in regard to specifying the nature of [ə]; and the variety of sounds actually identifiable as ‘schwa’ in his own dialect (of New York City) and in some other Germanic dialects. The conclusion to be drawn is that ‘schwa’ is never a vowel without quality in any of these linguistic varieties, though many standard definitions represent it that way. The ‘Fugue’ applies these observations to the history of English, reconsidering the standard view that all unstressed vowels fell together as a single sound beginning in the tenth century. Rather, given the considerable orthographic variation in Old and Middle English texts, he argues that unstressed vowels remained recognizable variants of stressed ones, in such verifiable variety as he earlier demonstrated for living languages and dialects. He broadly sifts the evidence of both Old and Middle English orthography and of the testimony of early orthoepists. Essentially, he concludes, there were two occasions of merger (not ‘reduction to [ə]’) of unstressed vowels in the history of English, in early Middle English and Early Modern English, merger that produced unstressed varieties of stressed vowels rather than a uniform unstressed [ə].

Touching on some of the same issues, but from a synchronic perspective, Edward Flemming argues in ‘The phonetics of schwa vowels’ that there are two schwas in English, one a true mid central vowel, found in absolute finality, and the other a more variable vowel with contextually determined features of height and frontness, found in non-final unstressed positions. The difference between the two schwas is readily detectable in a plot of the formant frequencies in the utterances of nine female speakers of American English: non-final schwas are high, with an average F1 frequency of 428 Hz, and whereas the F1 value of final schwa ranges widely, height can be correlated to variation in vocal tract sizes, as evidenced by the frequencies recorded for stressed vowels. Frontness/backness is much less constrained in non-final schwa, and this variability is best explained in assimilatory terms, since it can be pegged to consonantal environment as well as the quality of vowels in adjoining syllables. Recognizing the interaction of consonantal and vocalic influences on such assimilations is crucial, especially as contextual vowel qualities affect contextual consonantal qualities in a non-uniform fashion. Such coarticulatory effects are the result of the shortness of schwa, which hinders the achievement of any particular articulatory target. They may also result from the lack of any vowel contrasts in the environment in which variable schwa appears; but that would imply that variable schwa is qualitatively targetless, which remains to be shown. By contrast, final schwa does contrast with unstressed [i, oʊ], for example in pretty, motto vs beta, where flapping (in American English) confirms that the final vowels are unstressed. These conclusions suggest that the recommendation of Lass (above) that the different instantiations of variable schwa be discriminated carefully in transcription may not be useful, after all.

In ‘Perspectives on weakness from English /r/’ (which begins part 3 of the book, titled ‘Historical and dialectal case studies’), April McMahon argues that intrusive [r] in non-rhotic varieties of English (as in hosanna[r] in excelsis) and linking [r] (as in fea[r] of flying) are both best explained synchronically as inserted by rule. In counterevidence to accounts relying on deletion rather than insertion in regard to linking [r], she points out that in the North East of England, linking [r] may optionally be suppressed in phrases like The[r]e are some, but antevocalic /t/ mandatorily converts to [r] when in the same dialect area the preceding vowel is reduced to schwa, as in He said tha[r] it's okay, and an account relying on insertion cannot plausibly explain why the former pattern is variable while the latter is not. Historically, of course, deletion played a role, and it seems to have been conditioned by three factors: (1) relative sonority, (2) weakness of position (that is, in the coda rather than the onset), and (3) similarity between the affected segment and neighboring sounds. A comparative study of both Scottish and Northern English varieties shows that rhotacism is the most salient feature in distinguishing the two, and a similar conclusion may be reached in regard to worldwide varieties of English. Even if weakening of [r] is natural, it is not strictly predictable, and the observable patterns in the different varieties of English are best explained by functional and historical factors rather than on the basis of markedness.

Lenition of consonants in the syllable coda, though a widespread phenomenon in Irish English, varies considerably from one regional or social variety to another, according to Raymond Hickey, ‘Weak segments in Irish English’. The chief varieties are supra-regional northern and southern, and local Dublin English, the last of which is the oldest English variety, dating from the thirteenth century, and shows the greatest degree of lenition, perhaps under substratal influence. It is almost exclusively alveolar stops that are affected, most especially /t/, which may lenite to an ambidental fricative, or [h], or a glottal stop, or a flap, or a rhotic continuant, on its way to deletion, according to environment and variety. From the complex array of effects Hickey concludes that ‘movement along clines of articulatory reduction can be arrested by notions of “standardness” and the avoidance of merger with vernacular varieties. Thus Irish English represents a number of sociolinguistic scenarios in which the interplay between internal and external factors pan out in different ways depending on whether prescriptivism acts as a brake on “natural” developments such as segment weakening’ (p. 128).

The argument of Paola Crisma, in ‘Word-initial h- in Middle and early Modern English’, is that initial /h/ was a ‘weak’ phoneme, ‘in the sense of being deleted from the output in some contexts, though being present as a consonantal phoneme in the underlying representation’ (p. 130). With mostly predictable exceptions, an (rather than a) is used before words with initial h- in Middle English, excluding texts of northern origin. This does not mean, however, that /h/ was deleted from underlying forms and subsequently reintroduced, as the observable effects of restoration would be too poor in hypo- and hypercorrection for such a development to be plausible. Rather, /h/ may have remained as a feature of a vocalic onset (as argued by Schlüter, below) or as a stylistic variant within the speech of any single individual. The pattern observable in the use of a/an is replicated by the possessive pronouns my(n) and þy(n), with the exception of two texts that appear to anticipate the treatment of initial h- in Early Modern texts (Malory's Morte Darthur and Caxton's Reynard the Fox). In such Early Modern texts four stages are discernible in regard to the establishment of the modern treatment of a/an, as well as of the treatment of the possessive pronouns, suggesting the status of /h/ as a consonant – four stages that can be defined by the re-ranking of constraints in an OT analysis. By the third stage the indefinite article and the possessive pronouns pattern differently, the latter lacking final -n, and the difference may be explained on a prosodic basis, stemming from the fact that pronouns are sometimes stressed, whereas the article is never stressed. Orthographic evidence lends support: the incidence of omission of etymological initial h- in words of Germanic origin in two Middle English corpora is minuscule, while the rate in Latin and French loans is about 26 per cent. Unetymological h- is supplied in both native and borrowed words at a low rate, usually after a final consonant rather than a vowel.

Julia Schlüter considers some of the same issues in ‘Consonant or “vowel”? A diachronic study of initial 〈h〉 from early Middle English to nineteenth-century English’, which centres on the question how a sound that was putatively lost in early Middle English can have reappeared in etymologically appropriate places later in the history of the language. An analysis of a/an alternation before vowels, 〈h〉 and other consonants in several corpora covering the period 1150–1903 supports the argument of Minkova (Reference Minkova2003: 149–60) that the Old English requirement for a filled onset in stressed syllables evolved to a mere preference in early Middle English. The use of an before h- is universal in The Helsinki Corpus of English Texts (Rissanen et al. Reference Rissanen, Kytö, Kahlas-Tarkka, Kilpiö, Nevanlinna, Taavitsainen, Nevalainen and Raumolin-Brunberg1991) for the period 1150–1250, and it declines more or less steadily over the remainder of the period surveyed, though it never approaches null incidence. Since such apparent non-articulation of initial 〈h〉 thus antedates any appreciable influx of French lexis into English, the treatment of 〈h〉 cannot plausibly be attributed to French phonological influence on English. Neither can the reintroduction of 〈h〉 be ascribed plausibly to spelling pronunciation in the early modern period, as has often been assumed, since a before h- in both native and borrowed words is well attested in the Helsinki Corpus for the late Middle English period and the early Modern period to 1600. As for the pronunciation of hour, honour and heir without initial [h] to the present day, this is in part explained by their high frequency, which is again determinable from corpora. Patterns in the corpus also strongly support the argument that the reintroduction of initial [h] is far commoner in the onset of stressed syllables than of unstressed, the latter condition of course being encountered normally in borrowed rather than native lexemes. Further issues that might be illuminated by quantitative investigation of corpora include possible inflection of the indefinite article in early Middle English, differences in the impact of different French dialects on English lexis, differences between content and function words, and others. The explanation offered for the varying treatment of 〈h〉 is that early in the period surveyed it represents part of a vowel, an aspirated onset, since it lacks any supraglottal features.

Beginning part 4 of the book, titled ‘Prosodic and morphological correlates of weakness’, Julia Schlüter offers a second contribution, ‘Weak segments and syllable structure in Middle English’, in which she broadens her field of enquiry to include variable word-final consonants, final [ə] and word-initial [h] with lexemes other than a/an. The possessive pronoun my/min(e) shows a familiar pattern of development over the course of the Middle English portion of the Helsinki Corpus (covering the years 1150–1500), with final -n still predominating before vowels at the end of the period, but occurring rarely before consonants even at the start of the period. Before h-, min grows to be a minority form by 1500, but not at so precipitous a rate as before other consonants. The variant mine must be regarded as monosyllabic before vowels but disyllabic before consonants, though it has all but disappeared as a variant by the middle of the period. The determiners thy/thin(e) and no/non(e) behave similarly, except that thin in somewhat more conservative, probably because it was becoming an archaic form, being gradually replaced by your. The indefinite article a/an behaves differently, retaining both variants to the present day, and several possible explanations are offered: a/an is a weaker lexeme, being consistently a clitic (as remarked by Crisma, above); it is far more frequent than any of these determiners, and frequency contributes to the preservation of morphological irregularities; and the vowel of a/an, once reduced to [ə], could not remain distinct in hiatus before a vowel as easily as the vowels of the determiners could, which are more peripheral. Among prepositions, which might be expected to pattern like a/an, which is also a clitic, the word of is reduced to o too seldom to yield much useful information (although the reduction occurs frequently in just two texts of the period 1350–1420), but on and in pattern similarly to each other, being reduced to o and i frequently (and only before consonants, with 〈h〉 behaving like a vowel) only in the earliest Middle English texts, with only the unreduced forms (or nearly so) attested in the latest texts of the period. Final -n in infinitives is lost in similar fashion: the survey is confined to the monosyllabic verbs been, doon, goon, seen and sayn, with the expected relative rather than absolute difference between anteconsonantal and antevocalic environments maintained in a gradience from the earliest to the latest texts. Plural subjunctives pattern generally the way infinitives do in this respect, whereas plural indicatives show a notable delay in the loss of -n, and in past participles the trend is halted and reversed in the course of the period, with the result that many such participles retain -n to this day. The loss of final [tʃ] in the morphemes Ich, -lich and everich tells a similar tale of anteconsonantal patterns spreading to antevocalic ones. Thus, although the loss of weak segments depends crucially on syllable structure, being most favoured in syllable codas, morphological considerations must also be taken into account (and, it might be added, syntactic ones, as well, as with the differential loss of -n in verb forms according to syntactic function).

In ‘Orthographic indications of weakness in early Middle English’, Margaret Laing first very usefully sketches some of the differences between the methods employed in compiling the Linguistic Atlas of Late Mediæval English (LALME: McIntosh, Samuels & Benskin Reference McIntosh, Samuels and Benskin1986) and the Linguistic Atlas of Early Middle English (LAEME: Laing & Lass Reference Laing and Lass2008), the chief difference being the open-endedness of the data collected for the latter, whereas the former employs a fixed questionnaire. She also explains the nature of the tagging used in the LAEME corpus in order to show what kinds of information about verbs are retrievable. Her particular interest in the present chapter is the development of the Old English verb system inherited by Middle English, particularly with regard to syncope or its absence in the second and third persons singular of the present indicative and in the preterite/past and past participle of weak verbs. The Old English background is sketched with impressive thoroughness on the basis of treatments in the standard grammars, and then there is an examination of the differing accounts of Middle English developments in various handbooks. Some very full tables illustrate the fact that the handbooks are partly supported and partly challenged by the LAEME data with regard to the roles played by class membership and regional origin in shaping Middle English developments. The distinction between non-syncopating weak verbs of Old English class II and verbs of other classes remains robust, whereas there is considerable diversity in the development of syncope as a verb characteristic according to tense and stem-final consonant. Stems with final sonorants are particularly susceptible to syncope, and whereas syncope is to be found in all dialects, it is commonest in the South. Though rare in the North, there it may even affect verbs derived from Old English class II.

Rounding out the volume, Donka Minkova, in ‘Inflectional syncope vs. epenthesis: Notes on the history of the weak preterite in English’, builds upon Laing's findings, comparing later data from syllable-counting verse. First she sketches the morphological alternations with respect to syncope in Present-day English and Old English verbs in an OT framework. She argues that for Old English the accounts of the standard grammars can be simplified in such a way that /-d-/ is the preterite marker in all weak verbs (as in Present-day English), which means that verbs of weak class II are to be regarded as having vocalic stems throughout. In OT terms, the chief difference between consonantal and vocalic stems is that faithfulness constraints are violable for the former but not the latter. An analysis of Laing's tables shows that the incidence of unepenthesized (or syncopated) forms continued to be high in early Middle English, and phonotactics were rapidly becoming the primary determinant in the conditioning of weak past tense formation. An examination of the metrical treatment of the weak past in Chaucer and Hoccleve (i.e. c. 1400) reveals that phonotactics play a very salient role: epenthesis before /-d-/ occurs in more than half the instances in all verb stems except those ending in a vowel or a vowel plus a liquid consonant. Epenthesis is commonest after a stem-final consonant cluster (though not all clusters behave identically), and there is no remarkable relation between epenthesis and derivation from Old English weak class II. Over the course of Middle English and the development of early Modern English, the chief innovation was reduction of variability in the stem, which necessitated greater accommodation of the form of the suffix. Exceptions like hardned, opned, hindred and so forth remained in part until the eighteenth century, and their preservation, despite the reduction of stem allomorphy in other types, is due to their perceptual similarity to harden'd, open'd, hinder'd in casual speech. Frequency may have been a factor in the differentiation of past-tense forms of dental stems, such as cut, hit, burst, shed, from those like strutted, fitted, lusted, threaded. The devoicing in built, pent and so forth (cf. billed, penned) is harder to account for, but -ld is a highly constrained termination in Present-day English, occurring in a narrowly circumscribed set of forms after a lax vowel. The pattern is reinforced by comparison to forms like felt, knelt, left, lost. A great many issues remain to be explored, such as the interplay among chronology, frequency, and phonological and morphological salience; the possible mutual influence of preterites and past participles; the treatment of French borrowings in -ir and -ier; and the dialect distribution of variants (which are not differentiated on the basis of region in this chapter).

The volume includes a subject index.

The contributors to the volume are diverse, ranging from relative newcomers to some of the most senior (and eminent) scholars active in the field; yet the quality of the contributions is remarkably uniform, and uniformly high. What is particularly bracing about this collection is not just the way that the problem of phonological weakness is approached from so many different perspectives – the phonetic, the phonological, the morphological, the synchronic, the diachronic, the dialectal and so forth – but the impressively productive ways in which the contributors make use of this diversity of perspectives and continually incorporate the insights afforded by other contributions to the volume in refinement of their own arguments. The book has the distinct feel of a collaborative effort in the best sense, bringing varied forms of expertise to bear on a problem of particular interest not in an isolative manner that highlights the gaps and discontinuities that pervade English linguistics as it is currently practised, but in a notably interactive fashion that models brilliantly the strength inherent in diversity of scholarly enquiry. One consequence is that no matter what the reader's particular area of expertise, she or he is certain to find, in addition to approaches of a familiar nature, novel methodologies that will enlarge the reader's perspectives and suggest new avenues of approach to English historical linguistics. Though the topic is weakness, the book is an exemplar of strengths: it is an important accomplishment, and one that students of English historical linguistics will find indispensable.