Khayal is the pre-eminent genre of Hindustani vocal music. It is a dynamic ever-changing art form which, in its present form, accompanied by tabla, began to crystallise in the early nineteenth century. Many changes occurred in the twentieth century, most notably a radical slowing down of the vilambit part of performance and the gradual replacement of the bowed sarangi by the harmonium as the preferred instrument of melodic accompaniment.

In this substantial contribution to the literature on khayal, the authors have collected, transcribed, translated and analysed 492 songs (bandises) of khayal. The songs are all culled from commercial recordings of the twentieth century, beginning with the first ever recording of khayal, Gauhar Jan's rendition of rag Sur Malhar, recorded in 1902. The transcriptions utilise a modified form of Indian sargam notation, achieving an unprecedented degree of detail and accuracy with regard to rhythmic values and tonal nuance.

Chapter One defines khayal and looks at the birth of the recording industry in India and the history of khayal on records. It also presents hereditary and teaching lineages of the artists whose recordings have been transcribed. Chapter Two explains the notation system after surveying the great variety of notation systems which have proliferated since the mid-nineteenth century. Chapters Three and Four examine khayal songs from poetic, thematic and linguistic perspectives. Chapter Five looks at the rhythmic structures of khayal songs while Chapter Six explores the melodic nuances of khayal. Chapter Seven is a melodic, structural and poetic examination of a single khayal bandis as performed by five different artists.

The discussion and analysis is housed in Book One. Book Two presents the songs. The enclosed DVD contains almost 2,000 sound files, the sthayi and antara sections of each song, extracted from the original recordings. Nearly 500 additional sound files of khayal nuances illustrate the notated examples in Chapter Six. There is also a collection of alternative song texts for many of the songs and a series of charts presenting technical information from the authors' database.

Nicolas Magriel has been involved with Hindustani music since 1970 as a sarangi player and as a student of vocal music. He has a PhD in ethnomusicology from SOAS, University of London. His articles have appeared in several journals in the UK, USA and India. Since 2009 he has been working on the Growing into Music project, researching the musical enculturation of children in families of Hindustani and Rajasthani musicians.

Lalita du Perron has a PhD from SOAS, University of London. Her thesis was published by Curzon Routledge in 2007, titled Hindi Poetry in a Musical Genre: Thumri Lyrics. Since January 2009 she has been the Associate Director of the Center for South Asia at the University of Wisconsin, Madison. She teaches South Asian culture and Hindi, and specialises in song texts, dialects of Hindi, and aspects of gender in performance.

Early in 2001, Nicolas Magriel came into my office with a proposal for a research project. It was, he said, a project that he had dreamed of undertaking for many years. The project was to collect together the khayal compositions that had been recorded by Indian classical singers on 78-rpm and LP discs in the first seven decades of the twentieth century-and transcribe them.

Had he not recently completed the most thorough study yet made of the Indian bowed lute sarangi as his PhD thesis, I might have doubted his or anyone's capacity to complete such a massive undertaking. Even the proposed selection of 400 songs, recorded by a mere twenty top-ranking, singers, seemed a wildly ambitious target. The problems would be immense: how to notate the subtleties of melodic and rhythmic nuance that define khayal as a style, in a notation that would be simultaneously precise and comprehensible? How to deal with the poetry of the songs, equally hard to hear on recordings and to translate once heard? How to identify a sufficiently authoritative version of each song to merit its representation as a definitive 'text'?

Nicolas had of course prepared his answers to such questions. As I knew, he had already developed, for this PhD thesis, his own system for notating sarangi music, based on the indigenous sargam system, but going far beyond previous adaptations of that system in its capacity for representing fine detail. It would take a little further work to adapt it to the requirements of vocal music, and his many years' experience as a sarangi player, vocalist and listener meant that his ear was already finely tuned to the subtleties of the khayal style. He had furthermore enlisted Lalita du Perron, whose expertise in Hindi language and literature, and her specialism in song poetry, its language, and its performance, made her the ideal prospective collaborator and co-author. Between them, they would make a representation of each song that was authentic-being the version performed, on a particular occasion, for public dissemination, by a particular performer of unassailable reputation-but not definitive: no attempt would be made to establish an Ur-text, since such a text, in the fluid world of oral tradition, does not exist.'

The attractions of the proposal were compelling. No one writing in English had comprehensively documented the repertoire of khayal compositions on which Hindustani classical music is based. Much previous work on the theory of raga and on khayal as a genre would be illuminated afresh by such a rich anthology of exemplars from real-time performance. The songs themselves would have enormous value as cultural artefacts, and they would give a new perspective on a music too easily characterised as 'improvised'; deciphering their texts might unlock the mystery of what the songs were 'about', and reveal a rich but under- appreciated vein of Indian poetry. Even though no one performance could be regarded as the definitive realisation of a song, many of the recordings available had achieved such wide circulation, and such a degree of affectionate admiration by music-lovers, over many decades, as to confer an idealised status on them and make them models for emulation by other performers; they had become emblems of Indian culture as it evolved during the twentieth century, eminently worthy of musical and poetical investigation. In due course, after much discussion, and with the support of advisers and referees known and unknown, a research grant proposal was submitted to the Arts and Humanities Research Board (now Council), and to our great delight, accepted. The dream was about to become reality.

Starting in September 2002, for the next three and a half years the project occupied a cramped office in the Department of Music at SOAS, in which the two researchers sat back to back working on music and texts respectively far into the nights. Once the recordings had been digitised and the pre-composed sections identified and extracted, the meticulous work of transcribing the music and the words, requiring many hearings at half or even quarter speed, could begin. Periodically the researchers would leave for a period of travel in India, to consult singers, musicologists, literary scholars, record collectors and archivists; a representative community of Indian and non-Indian experts participated extensively and enthusiastically in the project. At other times there were conference papers to present, and interim publications to see through the press. So laborious was the 'odyssey of transcription' that a five-month extension was sought and granted, followed by a further two months for the complex task of seeking copyright permissions for the extracts presented on the DVD accompanying this book. This process has been largely successful, but a couple of companies have remained unresponsive.

As the project reached the end of its funding, the total number of songs transcribed had reached a target-exceeding 492, representing no fewer than 74 artists. Along the way, it had proved necessary to research a number of related topics, not all of which had been foreseen at the outset. The notation system and method of transcription had radically evolved (necessitating extensive and repeated revisions of completed transcriptions). A survey of the history of khayal recording, from the earliest recordings of 1902 to the arrival of cassettes in the 1970s, had been undertaken. Published anthologies of khayal compositions, in some cases by the same singers who recorded them, were searched and cross-indexed as sources for comparative textual study and for notated versions, and their idiosyncratic notation systems were compared. Poetic themes, linguistic structures, vocabulary, metrical structures, rag structure, and techniques of embellishment and musical expression were all meticulously analysed. These investigations, and more, took several years beyond the end of the project to write up; they are discussed in the first volume of the present publication which is followed by the transcriptions of 492 songs.

The scale of this anthology inevitably calls to mind the pioneering work of Vishnu Narayan Bhatkhande, the lawyer-musicologist who was the first, in the early decades of the twentieth century, to publish a comprehensive collection, with notation, of the songs of Hindustani art music (Kramik Pustak Malika, 6 vols.). But the aims and methods of Magriel and du Perron are diametrically opposed to those of Bhatkhande. The American musicologist Charles Seeger pointed out that writing can be used to represent music in two contrasting ways: the contrast turns on the intentions of the writer. 'Prescriptive' music writing records 'a blueprint of how a specific piece of music shall be made to sound'; 'descriptive' music writing constitutes 'a report of how a specific performance of any music actually did sound' (Seeger 1958). This distinction, which has become axiomatic in ethnomusicological circles, is not without problems, since the act of writing music notation necessarily entails both the description of a performance or performances, whether really experienced or imagined by the writer, and the prescription of a consequent reconstruction by the reader, either a real performance or an imagined one. Thus prescriptive music writing entails a descriptive element and vice versa. Nevertheless Seeger's distinction is a useful one, because a prescriptive representation typically conveys only as much information as the performer requires in order to render the specific piece represented; knowledge of how to render pieces in general, such as stylistic detail and improvisatory techniques, can be left unstated, as they are already known to the performer. This was Bhatkhande's approach, since his object was to provide didactic models suitable for use in music colleges as a basis for institutional music pedagogy. His representation of each song is an idealised version, constructed in the light of his own understanding of the raga repertoire; in a recent study, Dean Morris (2004) has shown that Bhatkhande repeatedly revised his notations in successive editions of the Kramik Pustak Malika as his concept of this ideal model evolved. Although he did include a surprising amount of pitch-nuance in his notations, much additional stylistic knowledge must be brought to bear in order to turn his notations into acceptable performances. And Bhatkhande took scant interest in the songs as poetry, treating the lyrics merely as so many syllables to be sung to the prescribed notes.

The intention in the present study is primarily descriptive and analytical. But students of Hindustani music may find the combination of detailed notations, reconstructed song texts, and sound files a valuable adjunct to oral transmission. The intention is to document particular performances of outstanding artists of the recent past, each such performance constituting one out of an infinite number of possible realisations of a song. Just as the performance itself is not definitive, neither is its representation, which draws on the notator's prior experience and makes assumptions about the prior experience of the reader. The choice of a 'digital' style of notation, based on the Indian syllabic solmization system (sargam), rather than an 'analogue' (i.e. graphic) or partly-analogue system such as Western staff notation, reflects Magriel's training in Indian music and his intention to engage with Indian readers. Music cognition is probably both' digital' and' analogue' in nature, involving both categorical perception of discrete pitch-classes and metrical units, and simultaneously an awareness of contour and of nuance within and between categories (Snyder 2000: 85-90). So it is appropriate that Magriel has extended the categorical pitch and rhythm symbols of sargam, about as far as practically possible, in order to convey singers' subtle use of inter-tonal pitch space and inter-matra rhythmic space. The complexity of the result, expecially in slow-tempo contexts, reflects the complexity of the khayal vocal style, in which even what might be initially perceived as single pitches turn out to be complex groups of pitches (or 'tonemes') in which the identity of some individual pitches may be unclear. In this situation there is no one best notational representation, and Magriel makes it clear that his proposed solutions are not definitive choices.

Both in the notated representations of songs and in the introductory chapters, this book challenges us to re-examine assumptions about many aspects of music. For example, the nature and definition of raga come under fresh scrutiny. We might have felt comfortable with the assumption that, whatever else defines a rag, its basic 'scale' or pitch-set is stable and exclusive of all other scalar pitches. Not so, says Magriel: the transcriptions reveal that many of the greatest khayal singers of the past sometimes used scalar pitches additional to the canonical scale of the rag they were singing. Dhrupad singers would no doubt see this as proof of the ignorance or laxity of khayal singers in matters of 'rag purity'. Magriel on the other hand suggests that it reflects a greater emphasis on the motivic than the scalar aspects of a rag, especially in the earlier days of khayal recording: 'The Gestalt of a rag is a much more compelling phenomenon than the sort of rigidly defined tonal structure by which we might mistakenly understand a rag to be determined'.

We might understand the 'Gestalt' or svarup of a rag, in cognitive terms, as the middle level of a hierarchical categorization of melody in Indian music: above it, the higher-level category of scale-type (thath), which may be common to a number of rags; below it, the level of specific melodies (dhun) (cf Powers 2001; Zbikowski 2002: 31-3). The middle level of a hierarchical concept, the 'basic level', is 'the level at which we most easily categorize the world, indeed, the level at which our conceptual grasp of the world is grounded' (Snyder 2000: 85). It is normally the level first learned by children, who understand the concept 'dog' before they learn the more general 'mammal' or the more specific' Alsatian'. Viewing rag as a middle-ranking 'basic level' category restores an idea of naturalness to a concept often represented as theoretically complex or ineffably mystical. The priority of this basic level concept over the higher level 'scale', suggested by Magriel's observations, might be linked to the dynamics of an oral tradition, transmitted within a social class that, before the twentieth century, had little use for keyboard instruments or music notation, both of which force the user to 'digitise' the soundstream of melody into discrete pitch-categories and scales. The gradual replacement of the fretless sarangT by the key-board harmonium as the main melodic accompanying instrument for khayal, and the proliferation of sarga-notated compositions and exercises in institutional pedagogical contexts, both occurred during the period covered by this study, and must have reinforced sastric tendencies to define music at the outer levels of scale and tune rather than at the middle level of rag svarup, Even in sastra the classification of rags according to scale dates only from the sixteenth century (for South India) and seventeenth century (for North India). Although in some respects a shift of focus from rag svarup to scale has progressed further in Karnatic than in Hindustani music, sharpened and flattened scale degrees are still not distinguished in Karnatic notation. Bhatkhande's attempt to define in writing the scalar structures of Hindustani rags remains his most controversial initiative.

As Magriel further points out, the complexities of rag Gestalt' are possibly best learned by the subconscious mind in the course of oral transmission, and are possibly indeed too complex for the conscious mind to fully grasp and categorise'. It has been shown that the kind of knowledge required for the performance of complex tasks, whether driving a car or performing music, must be implicit, that is, non-verbal. Bloch (1998: 3-21) reasons that although it may be possible in some cases to articulate such knowledge verbally, the learner must convert explicit rules back into implicit knowledge, through multiple repetition, memorization and internalization, in order to achieve the rapidity of response required in situations where many things may be happening at the same time. Bloch suggests that non-linear knowledge of this kind may be more efficiently transmitted through non-verbal than through verbal means, since the latter requires a double translation, from implicit to explicit knowledge by the teacher, and from explicit back to implicit by the pupil. The kind of distortion that such translation can produce is well illustrated by Perlmann (2004: 23-4), whose gamelan teacher explicitly verbalised a damping technique quite at variance with what he did in practice, and was unaware of the difference. We can easily extend this idea into the realm of music notation, where again a double 'translation', from implicit knowledge (of the sound of the music) to explicit (the notation) and back, must occur, with the attendant risk of multiple distortion. However, a written representation can be checked and re-checked for accuracy, in the present case at full, half and quarter playback speeds, which inspires a degree of confidence- even though no claim is made for the result being definitive.