IDEAS TOWARDS INTERFACING DIGITAL HUMANITIES RESEARCH Hans Walter Gabler What 'humanities' encompasses: Literary Studies Cultural Studies History Philosophy Musicology etc. What engagement in and with 'humanities' comprises: Scholarship verifies, secures foundations of subject Criticism creates and secures understanding of subject Research engages exploratively, learningly with subject Teaching imparts subject imparts research techniques imparts critical skills imparts methods of scholarship All this has long-standing traditions and is, as the saying goes, 'what we have always done'. The challenge to our engagement in and with the humanities today is the digital medium. This engagement moves into fresh light and focus in consequence of the medium, since, through the new mediality, 'what we have always done' is no longer a matter of course, hence remaining unreflected in itself, but demands instead reflection and questioning. This goes hand in hand, moreover, with the circumstance that we encounter the matter of the diverse humanities' subjects, and the materiality of their objects, more and more at a radical remove from themselves. They have become, or are fast in the process of becoming, virtualised—which is the main condition for their becoming accessible digitally. Virtualised, however—and this is a paradox to be savoured—the material remains from the processes and traditions of history move into closer reach than ever. This is true not only because many a journey to libraries, archives, museums, cities living and dead, need no longer be made. It is also true since a material object virtualised in many respects permits closer scrutiny than does 'the thing itself'. That is, we engage, interact, 'interface' with the virtualised object in the digital medium in other and different ways than we do, through our senses unaided, with the palpable object. Some ways to scrutinize in order to understand, it is true, are not available in the digital medium: we can't smell or touch the objects or sense an aura from their specific setting, from all contingencies in their and our world of experience. By contrast, however, we can enlarge, refract, filter, segment a virtualised object with the aid of every digital technique available and discover in it and about it what was never discernible to the naked eye. Take, for example, the case of facsimiles of pictures or of manuscript or printed pages. Photographed and printed, they were a form of visual mediation well before the age of digitization. But it is interesting to note how little thought was given to the facsimile as mediator between the material manuscript, say, and our scrutinizing eye. In the Gutenberg era, the facsimile was simply a two-dimensional material object like the page in a book or a leaf of manuscript – even though it was less than these, actually, since it provided only the front view of either. While thus defective in terms of the original, however, it was yet material just like the original. To set off the facsimile from the original, it was enough to say generally that the facsimile simply never sufficed in every respect as a shortcut, let alone a stand-in for the original. Yet just how it lost out was seldom accounted for under categories of adequacy or loss. The researcher and scholar can and does not exclusively, or on a regular day-to-day basis, work with and from original documents.1 His or her resources are commonly copies—which I'll here specify as facsimiles. The situation is conditioned by distance and loss. [John Milton would have said, 'by distance and distaste', but I will not go that far.] Yet the degree of loss differs significantly between the aspects of raw materiality, inscription, and textual record. Under the aspect of raw materiality, of course, a facsimile, by definition, preserves nothing of the original: not its size, nor its paper, paper quality, its foldings, quirings, creases and tears, its inks, crayon markings or pencilings. The textual record in a facsimile, on the other hand – that is: the conventionalized graphics of letters, numerals and marks of punctuation, doubly controlled by the conventions of the alphabet, and of grammar, syntax, and semantics – ideally loses nothing in the reproduction. Wherever the text, that is, under the double control of the conventions of writing and of language, is unambiguous, the facsimile reproduction is wholly adequate. The reason for this is that text, qua text, is allographic, meaning that it is infinitely reproducible to say always the same thing. In terms of any given original document, however, the textual record must be recognized also as a sub-category of the inscription; and the inscription, janus-faced, must consequently be recognized, too, as a sub-category of the document materiality. It is the inscription that renders an original manuscript properly iconic, 1 This paragraph takes a little further some notions previously developed in my "On Textual Criticism and Editing: The Case of Ulysses." Palimpsest: Editorial Theory in the Humanities, ed. George Bornstein and Ralph G. Williams. Ann Arbor: The University of Michigan Press, 1993, 195-224; pp. 213-214. 2 that is: autograph. The elements of a manuscript's iconicity, moreover, its positionings, spacings, shapes, and sizes of the marks on paper, importantly including, too, its doodles and all other manner of graphics, these are all random and unpredictable, non-conformant as they are to any secondary control conventions. (Nor should we forget that the actual tracings on paper of the textual record would be equally thus perceived but as random marks, as pre-Rosettastone hieroglyphics, so to speak—or, in other words, as unintelligible—were we not familiar with the control conventions of writing and of language.) It is evident that a facsimile in the Gutenberg era was simply a visual copy, and if reliable as such, then properly reliable only as copy from which to read the texts that were genuinely, that is: materially, inscribed in the original documents. They were and are adequate, in other words, for their allograph dimension alone. Their autograph dimension, however, which they also somehow, it is true, seemed to convey since they were visual copy, remained inert. For autograph material objects simply cannot (as we know) be copied into identities of themselves. Their copies have merely an illustrative, an 'almosting' quality. The facsimile of a material object, as fac-simile, can never be anything other than an illustration. This, I suggest, has been taken as a truth so self-evident that it seems seldom to have been reflected upon. With regard to the digital medium specifically, moreover (in these, the medium's early days that we still live in), so strong has as yet been the pull of habit and convention that facsimiles are mostly seen to be inserted here, too, merely as illustrations (an instance among many of how habits and modes of organisation have been and are still being deployed in the new digital medium that simply mime the world and environment of the material manuscript and book). What I wish to suggest, by contrast, is that a digital facsimile, a digitized visual copy of an original, may occupy a genuine interface position between the material object digitized and the humanities' scholar and researcher. (In terms of modeling theory, the digital visual copy might be said to perform a modeling function in relation to the material object whose place it holds in the epistemic system constructed in the digital medium.) Of course I am not saying that the fundamental difference between material facsimile and digital facsimile has never been thought of before. I merely wish to draw particular attention to some ways in which such thinking has in diverse quarters already been realized. If it has hitherto mainly been so realized intuitively and pragmatically, perhaps, without much drawing of systematic conclusions, it seems to me well worth considering as a concrete example of how the taking-over of conventions from the pre-digital era of ineluctable materiality may be profitably re-thought and so 'made new' in the transition. 3 Let us not fall into the trap, therefore, of regarding the digital facsimile as more or less the same thing as the photograph or printed facsimile. Let us instead focus on the difference. The digital image differs radically from images we are traditionally accustomed to encountering in that it is not a material counterfeit of the material original. It is a virtual image, a mirage, we might even say, projected onto that other interface surface, the computer screen. Though only in this manner virtually visualisable, it does still have a material substratum. Yet its materiality, as we know, differs from the materiality of (let us say) a replica on film of the original material object. It consists of a magnetically chargeable carrier medium for digital information from which in turn the virtual image on the computer screen gets generated. Nor do the given sets of bits and bytes have natural-language labels or headings. They have digital addresses—essential, as we shall see, for making use of the digital image, and making use of it as an interface entry to humanities' research, scholarship and teaching. A feature about a digital facsimile, seemingly innocuous, though at the same time, at least at first sight, slightly upsetting, is that you can cut it up. Imagine cutting up a material copy, let alone a manuscript original, or a painting. It were a destruction if not perhaps beyond repair, yet nonetheless beyond full re-integration. It is otherwise with a digital facsimile. Cutting it up may prove a fruitful way to activating it and putting it to use. In the pioneer project of a digital edition of Gottfried Keller's, the Swiss author's, works, for example – an edition of some twenty years' standing – the editor, Walter Morgenthaler, hit on a simple trick in aid of deciphering seemingly illegible manuscript readings for transcription. He cut out the images of single words whose graphics yielded determinable transliterations and matched them by digital image imposition with less unambiguously decipherable graphic tokens. For such adroit trickery, admittedly, the image is not absolutely required as digital image. The ploy as such, however, is a step in the direction of recognising the digital image as differing in kind from the image on a material carrier medium. What you manipulate is not visual matter (this only seems to be what you do – such is the illusion of virtuality!). It is delimited digital information that can be correlated to other delimited digital information. The correlation and the match or non-match in the relation of the sets of digital information become in their turn analyzable digitally – become determinable by machine interfacing. The digital interfacing is the counterpart to the intellectual assessment and judgement required critically to solve the task in hand within the given project of digital humanities research and scholarship. The task in my example was simply – simply! – to help deciphering a manuscript inscription correctly. 4 From here, we can make the circles widen. To this end, I'll be hovering still in the area of manuscript analysis and text study, however partial a field of humanities' scholarship at large this is. Take this image: seemingly a visualisation, an illustration merely of a James Joyce manuscript draft. But, since it is a digital image, I can begin to do things to it with impunity that would be anathema with an original. Photoshop helps to underlay pre-analysed and pre-defined areas of the image in different colours: 5 This is a preliminary, and in truth a fake step towards what is required, and possible, at the level of bits-and-bytes: namely, defining the areas digitally through coordinates and thus digitally to isolate them for scrutiny and (potential or actual) interactive engagement for research and criticism. To go beyond the photoshop stage, genuine algorithmic treatment is required. At the level of the computer interface, take this option: http://www.hypernietzsche.org/demo/bksailehwgabler-31 [ click 'synoptic'; click 'image' in the left column] In the left column is now highlighted the right-hand page from the previous visualisation (where the page was sub-segmented in shades of blue). Here, there is just one area, underlaid in pink; the upper and lower segments are brownish (and should ideally be phased out). Only the underlaid sub-segment is relevant, relating as it does to the transcription in the right-hand column of the two-column screen presentation. Thanks to the software I have had the opportunity to use – it is the original HyperNietzsche software, developed for genetic analysis of manuscripts some eight years or so ago – I am able to shift in the transcription column from one level of inscription to any other ('No'[='no overlay'], '1', '2', '3') and so study the composition process of the text. This is digital humanities' explorative research in action via the visual interface of a chosen manuscript segment that is not identical with a material page but instead digitally cut out from the source digital image of that page. If this example has demonstrated how a digital image as interface was capable of supporting an analysis of the coming into being of a text, here is another example that moves the 6 focus much closer to analysing a writing process as such. I am now looking at the upper area of the left-hand page in the double-page manuscript opening from Joyce's working draft: http://www.hypernietzsche.org/demo/bksailehwgabler-33 [ click 'wide', click 'Fit page in window', click 'Transcription' and move the magnifying glass over the digital image] By means of the software option of a digital magnifying glass, I correlate digital image and transcription differently. The transcription is here not textually oriented. It serves as an aid to visually analysing the inscription—though, mind you, the digital magnifying glass doesn't resolve what wasn't decipherable in the material original, with or without optical aids, or (as the case might be) aids from chemistry or physics (chemical analyses of inks, for instance, have been known to be used in manuscript autopsies; or so-called hyperspectral imaging in palaeography to recover sub-layered text in palimpsests2). What the arrangement of the transcription under the digital magnifying glass allows is to direct the eye again and again at the digital image of the manuscript segment so as to enable the researcher to confirm or dissent from the transcription given and, building upon such scrutiny, to arrive at interpretive conclusions about thought processes and meanings behind the writing. In terms of the intellectual input such engagement demands, the point to be made is then of course not that the interpretive conclusions couldn't have been reached in any other way: they could, for the intellectual input is not computer-dependent. However, and nonetheless: interfacing with material documents through digital images of them as interface visualisations on the computer's technical interface area, the screen, constitutes a qualitative leap in humanities research and scholarship. It contributes to the foundation of what we properly, in our day, term 'digital humanities'. At this point, critique of many a technical aspect of the procedures I have demonstrated is of course perfectly in order. As said, the software employed became operable some eight to ten years ago. As operative in my examples, it incorporates as yet very little of what are hot areas of development today, such as, for example, dynamic imaging. 2 See for example Malte Rehbein with Patrick Shiel and John Keating: “The Ghost in the Manuscript. Hyperspectral Text Recovery and Segmentation.” Kodikologie und Paläographie im digitalen Zeitalter, Norderstedt 2009: 159–174. 7 http://kundigebok.stadtarchiv.goettingen.de Just take in for a moment the entrance page of a digital edition of a late medieval book of town ordinances for Göttingen in Germany3 that kept being modified and expanded over the years it was kept: you can at a glance intuit how it represents a continuous text, and then begin to study the digital edition for the historic information it accesses in highly differentiated analytical spreads. You can click yourself into the transcriptionally edited text passages under your chosen fields of interest – beer-brewing, perhaps? or taxation? or cattle-farming? or clothing regulations? – or you can view the digital facsimiles of the book's pages, again even under a digital magnifying glass. This edition doesn't either, as you can see, take on the problem of a bidirectional imagetext correlation – another hot area of development today, along with, for instance, interactive query and response procedures; or (since what I have shown are examples of manuscript and text editions) collation modules – to name just these items from what one might expand into a substantial list of desirables. But such critique is not the direction I wish to give to this talk. For even from the way that software has been realised for these specimens such as they are, the contours are discernible of the potential for interfacing digital humanities research and scholarship in widening relational networks. Let's continue to look at what may be developed from digital images, segmented at will under the precepts of a research plan. Let's say we are interested in the development of an idea and in its progressive contextualisation. From what has been possible to explore about the genetic progression of a text in a single draft manuscript, we may assume that the progression of an idea – say, a philosophic idea with Friedrich Nietzsche – may be explored through the progression of the texts articulating, developing and variously correlating the idea through a sequence of documents. With all these documents digitized, we digitally cut out from them all relevant passages and concatenate them. We survey on the computer screen what this concatenation yields for us in the form of paths linking the segmented images, like this: 3 Malte Rehbein, 1. Kundige Bok. Digitale Edition. Göttingen: Stadtarchiv Göttingen, 2010 (http://kundigebok.stadtarchiv.goettingen.de) 8 or like this: 9 or like this: These are all so-called rhizomes of relationships of text (genetic) and ideas (thematic) – of thought inscribed in, and as, text – interfaced via digital images of material document pages, or more precisely: digital images of segments from the digital images of the real pages incorporating the member elements of these relationships. We could open one or the other of the colour-underlaid, and therefore sensitized segments. http://www.hypernietzsche.org/WS,8 [click 'Rhizome', and then click in any of the thumb-nails the highlighted segment] Working our way back, click by click, through any of these segments reveals each to be variously visualisable and analysable in the same way as my Joyce examples were. The HyperNietzsche system of which I thus profited was the original invention of Paolo D'Iorio,4 the Italian philosopher and digital humanities pioneer attached to the ITEM in Paris and working for a few years in Munich under a major German research grant – that's how we came into contact. At the conceptual core of his invention was the idea of just how to store the individual digital images, be they page or segment images, and how to concatenate them to provide a computer-interface display from which the images could function as research interfaces, as windows and doors to exploration. The idea was as simple as it is original. Each image was 4 The foundational publication was Paolo D'Iorio, HyperNietzsche. Modèle d’un hypertexte savant sur Internet pour la recherche en sciences humaines. Questions philosophiques, problèmes juridiques, outils informatiques. Paris: PUF, 2000. 10 provided with a stable URL—ah, but that sounds as if the image was the real thing, and was just provided with a label. Yet that is not the way to put the matter. To be logically correct, one should say: the real world of the materially extant objects (books, manuscripts, manuscript pages—or anything of real material existence digitizable into the virtuality of bits-andbytes storage—all such was modeled in the digital medium as a set of URLs. In the materialreality world of, say, a library, shelf-marks or signatures function analogously. It is they that essentially constitute the material as well as the intellectual order of the library. Any given book is, by comparison, an incidental appendage to the library shelf-mark—as becomes evident when a book is stolen and simply replaced by another copy. As for URLs in the system of digital storage: not for nothing are they understood as stable digital addresses. They make up an essential set of place-holders constituting the empty form and container, as it were, for a digital research site with its defined (definable) provision of discrete and individual, yet at the same time correlatable and concatenable content. To the URLs can successively be hooked up the digital images of the material objects, plus digitized transcriptions of the inscriptions they bear (or of whatever other distinguishing and identifying marks they carry), plus any discursive prose (annotation, commentary, research essays or the like) relatable to any or all units defined as and by the given set of stable URLs within the research site – or, for that matter, throughout the world wide web wherever it holds relatable (for example, semantically relatable) content.5 It now I hope becomes clear why I have insisted on the digital image being distinct in nature from the image on any material carrier medium. Within the digital medium, defined as it is through bits and bytes only, the digital image is what can be appended to a URL, and what needs to be so appended in order to provide the visualisation by which to interface with what material-reality item the URL virtually stands in for. Moreover, by not labelling the digital image with a URL, but vice versa to label the URL with an image by which to interface with what object of analysis it points to (and so, by way of the image label, making the URL humanly perceivable) one creates the digitally-technical opening for the digital modelling of segments of material reality in diversities and complexities of ever increasing density or expansion. Such as, in terms of higher specificity and granularity, by sub-dividing URLs governing images of whole material pages into sub-sets of URLs for page segments only; or else, for instance, by generating those rhizomes I have exemplified, roping in clusters of referents in the real world of material objects. 'Generating' these rhizomes, however, be it clearly 5 From an original involvement in devising the HyperNietzsche infrastructure has grown the Semantic Digital Library Framework MURUCA, to be explored for its potential at www.muruca.org 11 understood, is only seemingly an automatic performance of the software. In truth, it happens on the condition only that the links have been pre-defined. The pre-definition takes the shape – simply, you will agree, once the idea has been formed – of a selection list of URLs. The stable addresses are stepping stones for our interfacing with the simulacrum in the digital medium of real-world objects and a real-world, community knowledge we endeavour to analyse and research in and through the digital medium. With such perspectives on the nature and logic of digital humanities research and scholarship gained, let's now look, if briefly, at what is happening on your own frontiers. When opening the Cardiff Database of Mid-Victorian wood-engraved Illustration, http://www.dmvi.cf.ac.uk/ you might think (especially if you chose the 'browse' mode) you were entering a gallery of pictures. But such an attitude, which of course you do not hold, were a mite pre-Kopernikanian (that is: concluding from appearances that the sun circled round the earth – illusioning from seeing the digital image that you were looking at the real, material thing: the wood engraving in any given printed book). What you manifestly do, is of course: you enter the digital site (the databank) under a research agenda – choosing 'Keyword search' or 'Advanced search' – and while you find visualisations galore of Mid-Victorian wood-engraved illustrations, they are obviously there as digital images precisely serving as interfaces to generate, that is to call up, diverse ranges of information. Playing around most superficially as I briefly did, I entered 'George Eliot' under 'Advanced Search' as author, and 'Romola' as title in the category 'Fiction'. This gave me a set of digital images as interfaces for further enquiry. Even by surveying the list itself before going further, though, I was able to conclude that this novel of George Eliot's apparently was only in the serialised publication in the Cornhill Magazine supplied with wood engravings, but not in the subsequent book publication. This distinction could for me be an incentive to pursue the question how, in terms of illustrations, George Eliot's novels in book compared with the books of other Victorian authors. I didn't for the moment follow up that question. I switched instead to the digital images themselves, just choosing one at random, and selected the 'Iconography' rider for it. This offered me a rich array of categories defined for the image chosen, each of which I felt invited to click for further enquiry. The illustration coded ROM37, for instance: http://www.dmvi.cf.ac.uk/imageIconography.asp?illus=ROM037 represents, as its iconography categories claim, an Italian street exterior in Renaissance style with architectural features of archways, pillars and steps, and four people in it subsumable under the categories of 'men', 'women', 'children' and 'babies' (one, consequently, of each). 12 What the image interfaces in 'iconography' mode, moreover, are search options for 'activities' and 'objects', under which latter category are actually classified 'animals', and under 'animals', 'cows'. Never mind that it took me the longest time even to make out that there actually was a cow's head in the picture: clicking the category 'cows' takes me completely out of the interface image I started with and throws me instead into indeed a virtual gallery of illustrations featuring cows. I'll here refrain from herding them in. What, instead, I wish to point out: the Cardiff website of Mid-Victorian wood-engraved Illustration, in calling itself a 'Database', underplays, I feel, its own nature. 'Database' is a past-generation term denoting a kind of bulk data storage space on which the Cardiff website is a significant advance. Taking just the sequenced category list given under the 'Iconography' rider, we notice, in terms of form, that the sequencing is hierarchical—which, once we have hooked on to the image on the computer screen, is significant for efficient and successful access to the digitally stored data. Looking at the categories themselves, we recognise that the list is generated out of a systematics applicable in art history. This, needless to say, is a discipline distinct from literary studies as we know them, while we nonetheless of course include both among the 'humanities'. We might say, therefore, that even while we take our departure when looking at the wood engravings from terms provided by art history, we quickly discover that the art-historical criteria the website offers soon get intermixed with terms usable for literary and textual or book-historical exploration. The site is thus a site properly for digital humanities' scholarship, which is interdisciplinary by definition—and indeed for digital teaching in the humanities, an aspect to which I shall return. By way of an intermittent summary, then: the website's contents are opened up and rendered explorable via the interface of images sensitised at the machine interface, the computer screen. The ways in which the website's subject matter is organised and made accessible via a categorized terminology support boundary crossings between humanities' disciplines as traditionally demarcated. The site, in my understanding, is a good example, and qualifies by its specific design to set a pattern, for what may be termed a knowledge site in the fields of digital humanities' scholarship. The term 'knowledge site' was brought into circulation quite recently through Peter Shillingsburg's book From Gutenberg to Google of 20066 and has meanwhile gained widespread, though rather undifferentiated currency. It designates an advance in the digital medium beyond the kind of accumulation of positive information that tends to dominate in books 6 Peter Shillingsburg, From Gutenberg to Google: Electronic Representations of Literary Texts. Cambridge: CUP, 2006. 13 (by necessity, in that medium). The 'knowledge site', therefore, is a transformation in the digital medium of the 'information site' that books provide. The transformation is effected by means of the capability of the digital medium to store information in relational forms of organisation, in relational web structures. Moreover: to deposit information in such manner in the digital medium means to input it always already in view of retrieval. This activity itself must rely on powers and strategies of pre-organisation, which obviously cannot but already spring from humanly pre-structured knowledge. Or, in other words: the knowledge in knowledge sites, and the building of knowledge into them, grows out of creatively participatory intelligence. In consequence, of course, it also initiates intelligently generated knowledge. However, the knowledge generated from deployments of currently realised knowledge sites takes shape only at the far end, beyond the moment of output from the digital site. Present-day scholarly websites are consequently definable largely as uni-directional in design. They render usable and disposable knowledge assembled in them. Yet while they assist their users in expanding and enriching their individual knowledge, they do not at the same time also make provision for the users to respond to or expand or enrich the site. Even 'uni-directional' as they are, though, the 'knowledge site' scholarly websites increasingly put at our disposal are unquestionably of great, often stunning encyclopaedic use. And anyhow, it is (by and large) in this mode that 'knowledge-site' scholarly websites are actually operative and work (as evidenced by the Cardiff website of Mid-Victorian wood-engraved Illustration). Beyond is trial and experiment. Where the trend 'beyond' appears to be leading, is towards the bi-directional research site, dynamically progressive, capable of interactive modification, and above all sensitive and selective in terms of the semantics of its core contents and their potential aggregation and enrichment. This is where, true enough, we meanwhile have visions of grand design, which we also shall continue to require. While (I fancy) not entirely beyond my imagination, they are by and large beyond both my expertise and my practical experience. Hence I will no more than mention that to seize on the semantics of digital contents, you need procedures to establish, and/or machine-generate, what in the IT world are termed 'ontologies': semantic, that is: meaning-related, (pre-)defined filters that help to raise the efficiency of searches through bulk digital content. (The hierarchized lists of terms leading from the interface digital images to the digital contents of the Cardiff website of Mid-Victorian wood-engraved Illustration are, I would say, prototypes of such 'ontologies'. More modestly, and indeed less pretending to implications of philosophy, your term for these lists are 'vocabularies'.) Putting such ontologies, or vocabularies, to use on digital content is obviously another mode of interfacing, using not 14 digital images as windows and doors to the materials to be researched, but verbal terms. You may wonder in passing what is the difference between using a search term and a vocabulary term for data exploration. Searches in pre-semantic-web days were always string searches: your computer algorithms grazed the binary data record linearly, albeit with the speed of lightening, to pick out the binary string that represented the term you searched for. Data exploration via semantic-web potential, by contrast, uses the relational nature of the web, any web. It concatenates data to correspond not as string, but as meaning to your vocabularydefined, and thus semantically selective, query. Your query, as one might say, interfaces into the digital medium to weave, out of the bits-and-bytes of the data stored, the on-the-fly-configured web carrying the semantics, the meanings, you have (pre-)defined – and should you experience that more is handed back to you from out of the digital medium than you anticipated receiving, you will tend to wonder a little: as if there were magic in the web, like in that of Desdemona's handkerchief. The sense of wonder arises from much seemingly autonomous hardware and software operation and 'machine-intelligence', into which in turn much ingenuity has been and is constantly being invested. The initial intelligence, however, required to establish ontology semantics cannot of course but be human—or there would be no chance in the first place of interacting, interfacing, with the digital medium as it today promises to hold our research sites of the future. Such research sites will be bi-directional. Conceptually, the relational combination of content in the knowledge site should provide nodes of knowledge for the user to engage with. But then, as we observed, the engagement cannot but generate enhanced knowledge. The knowledge site must consequently open up to enlargements of content and a deepening of hermeneutic understanding. In terms of design, this would mean that the nodes of engagement be made to function as points of interface from which interactively and dynamically to feed back the knowledge modifications and enhancements into the site. That is, the knowledge site of present conception should mutate further into a genuine research site. However: the visions and conceptions tend still to be heavily biased towards the construction of research sites in terms of their technical design and content organisation. Research-site uses, by contrast, still tend to be less focussed on. Hence, where the deployment of the digital medium in the humanities will in future demand increased attention is in the areas both of technical and of subject-dependent operations and interactive interfacing procedures. A strong incentive to further developments in such directions may, I believe, come from thoroughly reconsidering and re-conceiving the role of teaching. Teaching, as we should become aware, is not a secondary and derivative task. It is instead a seminal activity of the 15 cultural endeavour of science and scholarship, whether circumscribed as 'humanities' or otherwise. A simple axiom therefore from which to start would be: research is essentially a learning process. Hence, to receive teaching as well as to teach is to lay the foundations for acquiring both research skills and knowledge by means of such skills, and so to qualify for, and to do, autonomous work of scholarship. Consequently, it would make a great deal of sense to structure the medial support of digital humanities "bottom-up", so to speak, that is, from foundations laid through teaching and in learning. Just imagine the teaching-and-learning situation as the exercise ground for engaging with the multiple ranges of the humanities digitally: to learn simultaneously to handle the medium and to comprehend the subject fields; to build the units of subject matter and the research tools in reciprocal tandem 'on the fly'; thus to interface equally with the digitally held and widely ranging subject contents and with the team of learners-and-teachers of which one is a member; and altogether to communicate throughout with the fellow learners and teachers on foundations of a shared pursuit of explorative and dynamically progressive, digitally-based research aimed to further the common endeavour of humanities' scholarship. This wishful thought concludes for today my ideas towards interfacing digital humanities research. Yet wishful though it may be, it need not be taken as mere fancy. It should be understood as a expressing an urgent demand, and need. [Delivered on 29 September 2011 in Special Collections of the University of Cardiff Library as the "Inaugural Annual Cardiff Rare Books and Music Lecture".] 16