© Copyright ACM 1997
But the reality of day to day life shows that paper still continues to be the preferred medium for much of our reading activity [e.g.23]. This is despite the fact that screen technologies have vastly improved in recent years, that wireless, mobile computing technology is now widely available, and that new navigational and input techniques significantly improve the flexibility of our interaction with digital documents.
We describe a laboratory study aimed specifically at discovering how reading from paper compares to reading on-line documents so that we can design better on-line reading tools. This focus on design results in a study which differs from typical laboratory comparisons of reading paper and on-line documents in three fundamental ways:
A fairly substantial body of literature comparing the reading of paper versus on-line documents can be found in the psychological, human factors, and ergonomics literature (see [4, 10] for comprehensive reviews). The majority of studies focus on "outcome" measures of reading, such as speed [16, 25], proof-reading accuracy [3, 8, 25], and comprehension [6, 16]. A lesser effort has been devoted to looking at "process" differences between reading on paper and reading on screen such as how readers look at text in terms of eye movements , how they manipulate it , and how they navigate through it .
While some of these studies illustrate differences between paper-based reading and on-line reading they are generally either unremarkable or inconsistent. Even the reliable finding that reading from a screen is significantly slower than reading from paper [9, 16] has been challenged by recent experiments which demonstrate how improvements in screen technology lessen these differences, and may even eliminate them .
One reason sometimes cited for the failure to uncover significant differences between paper and on-line reading is the insensitivity of the measures used. For example, Dillon  has argued that this may account for the fact that the literature has generally suggested no negative effects of electronic text on comprehension. An important question this raises is: If the differences in performance between paper-based reading and on-line reading are that subtle, is it the case that even the reader fails to detect them, let alone the experimental method? Such a question is not so important for theoretical purposes, but is critical from a design perspective.
Perhaps a more fundamental problem with many of the studies has to do with the approach itself. After reviewing this literature, Dillon [4, p. 1322] remarked on the distorted view many researchers appeared to have of reading: "Most [ergonomists] seem to concern themselves with the control of so many variables that the resulting experimental task bears little resemblance to the activities most of us routinely perform as 'reading'."
Taken together, a review of the experimental literature suggests a need to establish a view of reading from the reader's own perspective as a more direct way of assessing perceived differences in paper and on-line conditions. Second, it suggests that, if our findings are to be extrapolated to real reading situations, one must be quite careful about designing the experimental method. It must involve a task which is representative of real world tasks and it must employ analytic methods which capture issues of relevance to design. One might argue that laboratory methods are simply inappropriate for these reasons. While there are many issues laboratory investigations of reading cannot address, we will demonstrate that there are nonetheless many lessons for design which can be produced systematically within such an approach.
In choosing a task for the study, we wanted one that was both ecologically valid and required a comprehensive level of reading that would be rich in the demands made on the document presentation medium. We chose the task of text summarisation as fulfilling these criteria. Reading in order to write a summary is a task which requires a deep understanding of the text. Indeed, Winograd  has pointed out that some of the strategies used in summarisation may also relate to the more general process of text comprehension.
Also, because text summarisation involves reading in order to write, it has a strong connection with many of the kinds of reading tasks we have observed "knowledge workers" carrying out. For example, we found at the IMF that economists spend a great deal of their time reading and condensing information, whether it be from their own notes, or the notes and documents of other people. In more general terms, we found that, while editing documents on-line, 89% of the time they read and refer to other documents. Thus it seems that reading for the purpose of writing is a fundamental part of this kind of professional work. As such, this task seemed a good candidate for our purposes.
In choosing both the software and hardware for the on-line condition, we had at least two alternative approaches. We could attempt to optimise the configuration used in the on-line condition by choosing the "best" available software and hardware, or we could opt for a set-up which aimed to emulate a more conventional situation - a typical workstation running a commonly used word processing application. The former approach would be, in a sense, second-guessing what aspects of hardware and software best support reading, which is antithetical to our approach. We therefore chose the latter approach recognising that better systems and better interfaces may significantly alter the reading process. Nonetheless such an approach gives us insights into the benefits and drawbacks in a typical on-line reading situation.
The paper article was presented in three-column format and was black on a white background with some colour pictures. In the Paper condition, subjects were presented with 3 documents in a pile: the article to be summarised, some paper for note-making, and a blank sheet of paper on which the summary was to be written.
The On-line condition used a similar format for the article, and was run on an Apple Macintosh Quadra 950 with a ProNitron 80.19 colour screen 1152 pixels wide by 870 pixels high, and with a refresh rate of 75 Hz. Three open documents were displayed in Microsoft Word (version 6.0). The first was the article itself which was presented in page layout mode so that, although each page was scrollable, it visually resembled a single sheet of paper. On top of this was a blank word document (in Normal view mode) with the title "Notes" on the title bar. On top of this was another blank word document (in Normal view mode) with the title "Summary" on the title bar.
Subjects in both conditions were asked to create a brief summary of the source article, and were told it should be approximately 200-300 words long and provide a clear indication of the main points and ideas of the document. It was made clear that they could mark the article in any way they saw fit and that they could make as many notes as they wished. Subjects were told that they could refer back to the source article at any point during the task.
Each session was video recorded. On completion, the experimenter interviewed each subject extensively about the nature of their reading and support activites. In addition to these questions, each subject was shown, and asked to comment on and explain, various clips of their behaviour which the experimenter had noted as being interesting during the course of the task. The clips were used to cue subjects' recall as to what they were doing at these points and why.
Given the objectives of the study and the sample size, a quantitative analysis of the data was deemed inappropriate. Rather, the emphasis in the analysis was on describing subjects' strategies and activities with the aim of uncovering similarities and differences among subjects and across conditions. These descriptions were constructed by watching the tapes and transcribing subjects' comments during these key activities. Notes taken during the interviews were then used to enrich our understanding of the activities we observed. Only a selection of the findings are presented-those which we feel have the most important implications for design.
Subjects comments in both conditions indicated that annotating and note- taking while reading was important in deepening their comprehension of the text, and in helping them to form a plan for writing the summary. Essentially, it did this by drawing attention to important points and making explicit the structure between them.
Two subjects relied heavily on annotation of the source document itself. Their markings included underlining, the use of asterisks, and making notes in the margin. Marking was idiosyncratic both in the choice of marks and in the ways in which subtle style changes were used to different effect. For example, the thickness of a line was used to indicate the degree of importance of some piece of text for later reference, while asterisks were used to link disparate sections of text.
One reason for this kind of annotation was to provide a set of markings for later reference. In effect, such markings helped the subjects extract structure from the text on re-reading the article. One feature of such marks was that they relied heavily on the context of the original document, or as one subject put it:
"The first reading was quite slow and the second reading was skimming. The annotations helped me skim when I was skimming in the second reading...Whenever I finished writing about some point I skimmed forward - I looked for the next annotated bit and then I just read around it a little bit if I needed to remind myself of what they were trying to say."
Quite apart from the way these marks supported re-reading, at least one subject who marked up the source document suggested that the very act of making such marks is a process which aids understanding and facilitates the building of an internal representation of the text. This subject believed that "as you underline something you re-read the words, and this enforces it more." Previous research  offers confirmation of this finding.
Notes written separately were also used as a resource for later reference, and three of the subjects relied on this strategy. However, this was different from annotating the source document, in that it helped in re-structuring and collating information from diverse locations. As one subject put it, it provided "a pool of text and ideas that I can dip into to write real sentences". But the notes were more than this, because they were also in the form of plans or outlines which were enriched and modified iteratively during the course of reading. A key feature of this kind of note-taking mentioned was that it needed to be done without disrupting the main reading task:
"And the fact that the notes aren't in full sentences or anything, they are deliberately shortened so as not to interfere with reading and thinking and things, they are just jotting."
Indeed, a more general characteristic of this kind of note-taking on a separate piece of paper was its smooth integration with reading. Note-taking was both frequent and interleaved with reading the source document. Reasons why this was possible are elaborated when we discuss navigation and spatial layout issues.
"So the annotation was not as easy as all that... I think the whole process would have been a lot quicker on paper. Annotations are that much more flexible because you can write in the margins which you can't very easily do here. You have to establish a new text block and then have to write."
This comment emphasises the difficulty of making marks within the context of the article due to the inflexibility of interaction techniques via mouse and keyboard.
Another reason for the general reluctance to annotate the document was that annotating on-line results in making changes to the original document: emboldening, italicising or underlining all alter the original document. Subjects indicated that they wanted to regard annotations as a separate layer of the document, and felt uncomfortable not maintaining this distinction. But it was not just the fact that the methods interfered with the base document. They also expressed dissatisfaction at the fact that they could not easily make annotations that were perceptually distinct from the underlying text. This distinctiveness is, in part, what supports quick re-reading by drawing attention to points of interest.
It seems, then, that note-taking separate from the source document was more appropriate with these on-line tools. Four subjects did do this. Two of them took notes using the copy and paste function to transfer information from the source text directly into a separate document. For one of these subjects this was fundamentally different from the note-taking activity observed in the Paper condition in that the transfer of verbatim information was followed by extensive editing on the copied text. This text then became the basis of the final written summary. The other two subjects took most of their notes only after reading the whole source document, producing a plan almost entirely from memory with very little reference back to the source document. In all cases, none of the frequent back and forth movement between notes and reading was observed as in the Paper condition.
Subjects spent a good deal of time moving through their documents in both conditions. It emerged from the interviews that this served at least three important functions:
Planning. Subjects talked about the need to make connections between different parts of a document to create a plan and develop an overall sense of its structure [cf. 10]- a process supported by skim reading though the text.
"This bit is where you seem to flick through some pages." (Exp.)
"I get the 1st and 2nd pages and then look at the third...[I'm] trying to connect bits of information to write on my plan - areas that I want included in the same paragraph - or topics I want included in my paragraph." (Subject, Paper condition)
For Reference. Subjects were also seen scanning the source article to check on facts, particular expressions, and spellings:
"I was looking for the "dragon's blood tree". I was trying to find that name...I couldn't remember exactly what they called it." (Subject, On-line condition)
For Checking Understanding. There were several times when subjects needed to re-read sections of the document to confirm or clarify their understanding:
"You refer back to it after you have read something up here." (Exp.)
"Yes it was because of the name of the tree - Cinnabar. And I needed to know if it was the same tree as they talked about over here." (Subject, Paper condition)
Partly what speeded this movement through paper was the use of two hands in combination with the tangibility afforded by the paper, enabling the effective interleaving and overlapping of activities. For example, subjects used one hand to keep hold of a page while searching through the document, or referring to another page, with the other. By marking one's place with one hand, subjects could quickly return to their prior activity. Two handed manipulation also offered opportunities for other types of interactions such as writing while moving a page closer to read, or turning over a page with one hand while the other was used to feel how many pages were left. The quick flicking through pages when searching and skimming was also quite clearly a two-handed action using one hand as an anchor for the actions of the other. Aside from issues of speed, physical cues such as thickness provided important implicit information about where in the document a subject was, as well as its overall length.
A final feature of paper that showed itself to be important in navigation was the fixity of information with respect to the physical page. Consistent with previous research [22, 14, 10], subjects in the Paper condition showed that they had acquired incidental knowledge of the location of information by reference to its physical place on the page:
"I knew it was on the 3rd page 'cause I could remember that it was in the middle [column] 'cause it was this botanic bit so I knew it was on this page."
This incidental memory meant that subjects could flick through the document quickly using only surface visual features.
"I was getting very annoyed and clicking on those things and shouting at it... I just found that it took ages and ages. I was losing interest - it was distracting me from the point."
However, the sheer length of response time was only one drawback. Another feature which limited quick movement was the fact that one handed input meant that navigation activities had to be performed serially with other activities. Combining a single source of input with a significant system response time meant that interleaving any two activities was cumbersome, as well as making it impossible to perform anything concurrently.
All of this was made worse by the lack of feedback in response to many actions which meant that subjects were not supported in temporarily committing to various actions. For example, one subject tried to get to a particular location in the document using the drag function on the scroll bar, but the document contents did not move concurrently in response to the dragging. As a result, the subject had to release the page icon and then wait to find out if she was in the right place. This kind of scrolling can make finding exact locations extremely time-consuming.
Spatial constraints on the interface also interfered with quick, flexible movement. For example, in order to move an electronic window, subjects were required to access the title bar which was often obscured by another window. Similar problems were encountered by subjects attempting to resize windows or scroll - actions which were also restricted to limited active areas. The overall result of this was that subjects' attention was drawn away from reading, as they attended to the often complicated task of dealing with the mechanics of moving in these spatially constrained ways.
Unlike with the paper documents, assessing document length was difficult to do in any incidental or implicit way. For many of the reasons just discussed, subjects complained about having to scroll through to assess a document's length. Further, while information about document length and page number was available at the bottom of the active document, none of the subjects remembered using this information when asked.
Finally, whilst subjects in the Paper condition used the fixity of information with respect to the physical page to support search, it was an interesting question whether the use of Page Layout mode was used in the On-line condition for similar purposes. This mode fixes the contents of the document with respect to its electronic page. When asked, subjects said that the fact that it was difficult to view a whole page meant that they had little sense of the location of information with respect to its page. However, comments by subjects in both conditions suggested that some did use the fixed relationship between pictures and text to support navigation.
Finally, we consider differences in the way subjects in both conditions laid out their documents in space. The reasons why the spatial arrangement was important can be summarised as follows:
To Gain a Sense of Overall Structure. Subjects talked of the need to lay pages out in order to form a mental picture or overview of the document:
"I did put it into two page mode at one point just to get a sense of the how the article was structured and that was actually very helpful." (Subject, On-line cond.)
To Cross Reference. They also commented on the need to lay pages close to each other in order to check on or to relate specific pieces of information across pages.
"When I was referring back to it I didn't like the paper clip being on it...I could lay out the pages and have two pages at once or even so that I could lay them out like that...so that I could find whatever bit I was looking for and also so that I could join things together." (Subject, Paper cond.)
To Interleave Reading and Writing. Juxtaposing the article to be read with the document being written also appeared to be important in the writing process:
"I always [overlap the source document with the plan/summary page] so that you can see it and write it at the same time so you are not looking here and then writing over there You have them together and you just write from one to the other don't you..." (Subject, Paper cond.)
"You have got p1 and next to it p2 and then you have got your notes. Any reason for that?" (Exp.) "It will be there in case I need to refer back to it but also to the side so that I am concentrating on p2." (Subject)
Spatial arrangement was also dynamic. Pages would be moved in and out of the centre of attention as different parts of the document became more or less important. Hidden pages would be brought to the front, while others would be covered up or if needed for quick reference later, moved to a more peripheral region of the workspace.
The ability to quickly refer to other documents while writing was also important. Most subjects made constant shifts back and forth between reading the source document and writing during planning. The source document and the plan/summary document would be positioned close together, often overlapping, during the planning phase of writing so that sections of the source text could be close what was currently being written.
Figure 1. An example of document placement layout during the planning phase of writing (pages 1-4 being the source article).
One example of how documents were arranged during this planning stage is shown in Figure 1. The subject, when writing the summary, needed to refer to three documents at once - the source document, the notes, and the summary itself. By positioning and overlapping the documents this way, it was possible to make the desired information available in a compact space. At one point the notes were moved several times within the space of seconds as the need to refer to the source document several times in quick succession was demanded by the creation of a single sentence. This points to the need to be able to access other documents quite quickly so that reading can be effectively integrated with the process of writing.
Another important characteristic of the interplay of reading and writing was the way in which the paper could be quickly and even simultaneously accessed while reading and referring to other documents. Aside from its concurrent accessibility, another feature of separate reading and writing "spaces" is it allows independent manipulation. This is important when one considers that the optimal angles for reading and writing may be quite different from one another. Subjects placed the paper at a greater angle away from the perpendicular (approx. 30 - 45 degrees) while writing than while reading (0-20 degrees). Additionally, the ergonomics for writing with a pen required continuous minor adjustments to the paper over time, using one hand as an anchor.
By resizing and overlapping windows, however, subjects were able to do some cross-referencing of documents. But again, limitations on the space available and the lack of flexibility in the ways they could be laid out caused problems for quick access across documents. For example, there were occasions when subjects wanted to use all 3 documents for various parts of the task but this called for very small window sizes or obscured windows. While Page view allowed the display of multiple pages, it displayed them in a fixed order.
The lack of any real periphery also meant that subjects had to give some thought to selecting and displaying documents. Generally speaking the task was divided into discrete phases: first pass, note-taking, and summary writing. Window positioning and sizing tended to occur at the boundary between these different phases and not during the phases. Thus subjects were having to anticipate what they would need rather than reacting to the requirements of the task as they arose.
A commonly used arrangement, especially in the note-taking or writing phase, was placing two documents side by side with minimal overlap. This avoided problems with the currently active window being sent to the foreground which was not only time-consuming, but was sometimes undesired, such as in the case of this subject trying to alternate between taking notes in one document and navigating in another:
"It was annoying that its got to be the current window in order to be able to move, and the current window is always the one in front. So I can't hide the one I was typing into behind there and simply move the pointer into it and start typing or pasting or whatever."
This foreground/background distinction constrained the extent to which a page was accessible for writing concurrently with reading - something which was clearly important for these subjects.
On-line tools clearly offer valuable benefits for the writing process: They support fast keyboard entry, information re-use, easy modifiability of text, and provide specialised tools such as spell-checkers and word count facilities. But in the support of reading, and more specifically in support of reading for the purpose of writing, this study has shown that the benefits of paper far outweigh those of on-line tools. Further, unlike much of the existing literature comparing paper to screen, none of these benefits have do with issues of screen resolution, contrast or viewing angle. Rather, the critical differences have to do with the major advantages that paper offers in supporting annotation while reading, quick navigation, and flexibility of spatial layout. These, in turn, allow readers to deepen their understanding of the text, extract a sense of its structure, plan for writing, cross-refer to other documents, and interleave reading and writing.
In terms of design, one could make the decision not to attempt to supplant paper for reading but rather to develop scanning technologies to ease the transition of paper-based information into the digital realm for the purpose of writing. Many such technologies developed in our own laboratory  as well as others [e.g., 12] could well be applied in this way. This approach accepts that paper is, and is likely to remain, the best medium in support of reading.
But one could also choose to look to paper for improving the design of digital reading technologies, both in terms of the development of better hardware and software. This is not to say that such technologies must become more "paper-like" or support reading in the same way as paper. Rather, the use of paper can be used to help draw attention to the kinds of processes that are important to support within reading, and to suggest alternative ways in which that support could be offered. On the basis of the findings of this experiment, some examples are as follows:
Part of the problem is the lack of support for free text annotation. Not only does stylus input support variegated, idiosyncratic marks, but it also supports the making of marks in context, and as a distinct layer on top of typed text. It is clear that, for this purpose, OCR is unnecessary, since it is not so important that these notes be used verbatim for writing as it is that they support the process of helping to extract structure from a document and to speed re-reading. For this purpose, it is more important to concentrate on input techniques which maximise richness and variation in marking, such as the use of pressure-sensitivity to vary line thickness, and the use of texture and colour. Careful consideration will then need to be given to how the input technology and display device interact. For example, if emulating direct pen input onto the document through a touch-sensitive screen, then the screen will need to be quite mobile to be placed at the correct ergonomic angle for writing.
This experiment has also found that there is a clear need for markings to be functionally distinct from the "base" document so that changes to the underlying text are not actually implemented unless the reader explicitly allows it. The use of hand-written markings is a way of enforcing the idea of markings as a layer on top of a document. However, digital technologies can offer interesting advantages over paper by allowing readers to choose whether the annotations are to be permanent or temporary, by offering selective viewing capabilities according to who wrote the marks or when, or by allowing marks to actually effect changes in the underlying text if desired. Many such features can be found in a prototype collaborative writing system called "MATE" . While there are some systems such as TkMan which offer the ability to highlight text, we know of no commercial systems that adequately support these functions.
Multiple input raises another design possibility which is that kind of input can depend on the device used. For example, drawing directly from the paper case, direct input through touch could correspond to navigation activities whereas stylus input could produce markings.
Another major design issue that the findings help make clear is the need for better feedback to aid navigation. Not only does it need to be immediate, but, if it is to support quick interim actions, it needs to be continuous during the course of an action, and not simply delivered on its completion (e.g., releasing the mouse button after dragging). Further, other modes of feedback such as audio or even tactile feedback could provide more implicit cues as to where in a document one is, and how much information there is. For example, one could imagine using non-speech audio cues  to indicate the thickness of a document, or number of pages left. This feedback could be provided in the natural course of navigation so that the cues are picked up incidentally. This is not to say that visual cues have already been exploited as much as they could be. The use of perspective and stereoscopic displays may offer a "third" dimension to better represent features such as document thickness, for example.
Of additional consideration for the support of navigation is heavier reliance on a mode which fixes information with respect to a page. However, this study suggests that this will only be of benefit provided a whole page can be viewed at once. Designers might also consider emphasising endemic reference points such as page corners and edges for use in information search.
One approach is to create virtual space whereby the visible screen space forms only a part of the workspace available, such as in fvwm in UNIX, and in Hypercard. In these applications, documents can be moved around this larger, virtual workspace through a separate, miniaturised overview. This offers an advantage over paper in that it can potentially help to manage extremely large workspaces not practically possible in the paper domain. However, it is fundamentally different in that, in the paper case, what is at the focus of attention and what is at the periphery exists in a spatial continuum. In these electronic cases, peripheral documents and documents at the focus of attention exist in separate virtual spaces. Thus, drawing a new document into focus, or setting them aside, requires a shift of attention to a different representational space, and the dynamics of this will necessarily take more time and effort.
The spontaneous dynamics of spatial layout may also be facilitated by improvements in interaction techniques. For example, designers could consider doing away with the idea of constraining document movement to "active" areas such as title bars. There are various window managers in current systems that allow the whole document to be made "active" (e.g. Motif Window Manager in UNIX). However, this is done by requiring users to enter a navigation mode. Consistent with a previous suggestion, the need for these temporal modes might be handled by the use of multiple, specialist input devices such as hand versus stylus.
Another implication which arises from the experiment is the need for increased flexibility in the way that readers can arrange on-line documents in the space available. This ranges from simple suggestions like removing constraints on readers' ability to lay any two pages side-by-side, to the possibility of placing documents and pages at angles to each other. Allowing readers to make virtual "piles" is also an interesting idea, and one that interface designers have tried elsewhere .
An alternative approach to increasing the flexibility of spatial layout is to treat portable, wireless displays as a way of providing a physical embodiment for documents. As with paper, documents can be placed by moving the display itself, rather than moving documents within a display. However, unless one has many of such displays, this approach does not address the requirement of providing concurrent access to multiple documents or pages. However, the experiment has pointed to the effectiveness of independent reading and writing spaces in the paper domain. Thus, two such displays - one for reading and one for writing - might deliver valuable benefits.
Using a broadly descriptive approach, we have carried out a laboratory study which draws attention to a number of important implications for the design of better interfaces for reading digital documents. Not only was this study motivated by our own field studies of reading in real work settings, but the design of the experiment was grounded in and guided by these observations. In turn, the findings help us to understand much of what we observed in the reading practices of these organisations. For example, it gives us a better understanding of why economists at the IMF always mark up and review their colleagues documents on paper, why they choose to read important documents from paper, and why paper documents surround their workstations as they do their authoring work .
What we hope to have demonstrated is the value of contrasting paper-based and on-line reading specifically with an eye to design. By doing so, we can begin to unravel the complexity of the design challenge, and begin to make some better informed predictions about when and if we can ever expect a paperless future.
2. Bolter, J.D. (1991). Writing space: The computer, hypertext, and the history of writing. Hillsdale, N.J.: Erlbaum.
3. Creed, A., Dennis, I., and Newstead, S. (1987). Proof-reading on VDUs. Behaviour and Information Technology, 6, 3-13.
4. Dillon, A. (1992). Reading from paper versus screens: A critical review of the empirical literature. Ergonomics, Vol. 35(10), 1297-1326.
5. Dillon, A., Richardson, J., and McKnight, C. (1990) Navigation in hypertext: a critical review of the concept. In Diaper, D., Gilmore, D., Cockton, G., and Shackel, B. (Eds) INTERACT '90. North Holland, Amsterdam.
6. Egan, D. E., Remde, J. R., Gomez, L. M., Landauer, T.K., Eberhardt, J., & Lochbaum, C.C. (1989). Formative design-evaluation of SuperBook. ACM Transactions on Information Systems ,7 (1), 30-57.
7. Gaver, W. (1992). Auditory icons: Using sound in computer interfaces. Human-Computer Interaction, 2, 167-177.
8. Gould, J.D. and Grischowsky, N. (1984) Doing the same work with hard copy and cathode ray tube computer terminals. Human Factors, 26, 323-337.
9. Gould, J.D., Alfaro, L., Barnes, V., Finn, R., Grischowsky, N., and Minuto, A. & Salaun, J. (1987) Reading is slower form CRT displays than from paper: Attempts to isolate a single variable explanation. Human Factors, 29, 269-299.
10. Haas, C. (1992) Writing Technology: Studies on the materiality of literacy. Mahwah, N.J.: Lawrence Erlbaum Associates.
11. Hardock, G. (1992) Unpublished Masters' thesis. Dept. of Computer Science, University of Toronto.
12. Johnson, W., Jellinek, H., Klotz, L., Rao, R., & Card, S. (1993) Bridging the paper and electronic worlds: The paper user interface. In Proceedings of INTERCHI '93, (24-29 April, Amsterdam), 507-512.
13. Lorch Jr., R.F., Lorch, E.P. and Klusewitz, M.A. (1993) College students' conditional knowledge about reading. Journal of Educational Psychology, 85, 239-252.
14. Lovelace, E.A. and Southall, S.D. (1983) Memory for words in prose and their locations on the page. Memory and Cognition, 11, 429-434.
15. Mander, R., Salomon, G., & Wong, Y. (1992) A "pile" metaphor for supporting casual organization of information. In Proceedings of CHI '92, (Monterey, CA.), 627-634.
16. Muter, P., Latremouille, S.A., Treunit, W.C., and Beam, P. (1982) Extended reading of continuous text on television screens. Human Factors, 24, 501-508.
17. Muter, P. & Maurutto, P. (1991) Reading and skimming from computer screens: the paperless office revisited. Behaviour & Information Technology, 10(4), 257-266.
18. Newman, W., & Wellner, P. (1992). A desk supporting computer-based interaction with paper documents. Proceedings of CHI '92, 587-592.
19. Oakhill, J. and Garnham, A.(1988) Becoming a Skilled Reader. Oxford: Blackwell.
20. O'Hara, K. (1996) Towards a Typology of Reading Goals. Rank Xerox Research Centre Technical Report EPC-1996-107.
21. Richardson, J., Dillon, A., McKnight, C. And Saadat-Samardi, M. (1988) The manipulation of screen presented text: experimental investigation of an interface incorporating a movement grammar, HUSAT Memo 431, Loughborough University of Technology.
22. Rothkopf, E.Z. (1971) Incidental memory for location of information in text. Journal of Verbal Learning and Verbal Behavior, 10, 608-613.
23. Sellen, A.J., and Harper, R.H.R. (1996). Paper as an analytical resource for the design of new technologies. Proceedings of CHI '97, (Atalanta, Georgia).
24. Winograd, P.N. (1984) Strategic difficulties in summarising texts. Reading Research Quarterly, 19, 404-425.
25. Wright, P. and Lickorish, A. (1983) Proof-reading texts on screen and paper. Behaviour and Information Technology, 2, 227-235.