<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-1404">
  <Title>Hypertext and Deixis</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
2. Background
</SectionTitle>
    <Paragraph position="0"> The earliest major work on deictics (also called indexicals) is by Btihler. He begins with an elegant image: The arm and the finger gesture of man, to which the index finger owes its name, recurs when the signpost imitates the outstretched &amp;quot;arm&amp;quot;...and the question \[is\] posed as to whether spoken language contains signs that function as signposts. The answer is yes, deictic words such as here and there have a similar function. \[1934\] Biihler did not originate the idea of deixis. He notes that classical Greek grammarians were aware of the concept. In the modern era, he credits turn-of-thecentury linguists such as Brugmann and Wegener, who, while reconstructing Proto-Indo-European, described four types of deixis: I, you, this, and that. Biihler notes that land you are referred to as personalia (Latin persona), which is from the Greek for mask or role. The meaning is that I and you, like actors, switch roles. BiJhler then advances a twofield theory: a deictic field, in which de ictics point to the actual entity in the world, and a symbolic field, in which all other terms refer symbolically to entities in the world.</Paragraph>
    <Paragraph position="1"> Fillmore \[1975\] gives a very thorough taxonomy of deixis. He enumerates five types: person, place, time, discourse, and social.</Paragraph>
    <Paragraph position="2"> Person deixis refers to the speaker, addressee, and audience of the utterance. Examples are pronouns and possessives such as I, we, yours, my, etc.</Paragraph>
    <Paragraph position="3"> Place deixis, including the locative adverbs here and there and the demonstratives this and that, are divided into two types. Gestural place deixis requires the addressee to physically monitor the utterance for an accompanying gesture: &amp;quot;Put the flower pot here&amp;quot;. Symbolic place deixis requires only knowledge of the conditions of the utterance, without physical monitoring. My friend from London may write, &amp;quot;It's raining here again.&amp;quot; Though I did not see her write the letter, I know where she was when she wrote it, and thus I know where here is. As an interesting sidelight, Fillmore notes that there is only one word in English which requires gestural deixis for its use: yea, as in &amp;quot;She is yea tall&amp;quot;.</Paragraph>
    <Paragraph position="4"> Time deixis, including the adverbs now, last night, and tomorrow, involves three different times. The encoding time is the time at which the utterance is made. The decoding time is the time at which the utterance is received (which will be delayed in written text). Finally, the reference time is any time referred to in the topic of the sentence. A typical voice-mail message might illustrate all three types.</Paragraph>
    <Paragraph position="5"> &amp;quot;This is Dan calling at 9 AM (encoding time). Give me a call back when you get this message (decoding time). I'd like to set up a meeting for Tuesday the 14th (reference time).&amp;quot; Discourse deixis, according to Fillmore, &amp;quot;has to do with the choice of lexical or grammatical elements which ... refer to some portion ... of the ongoing discourse -- something like, for example, 'the former'.&amp;quot; Discourse deixis (also called textual deixis by Lyons \[1979\]) can be both written C/' This sentence is written in English&amp;quot;) and spoken (&amp;quot;She spoke about this loud&amp;quot;). Fillmore notes that discourse deixis is de-Hypertext and Deixis 31 scribed temporally, as &amp;quot;any point in discourse can be thought of as a point in time&amp;quot;. For this reason, we can say &amp;quot;in the preceding paragraph&amp;quot;, or &amp;quot;the following announcement is a paid advertisement&amp;quot;. In a:kl ition, written discourse has unique terms due to pages being read top-to-bottom: thus, we say, &amp;quot;as mentioned above&amp;quot;, or &amp;quot;see below&amp;quot;.</Paragraph>
    <Paragraph position="6"> Finally, social deixis encodes, in Fillmore's words, &amp;quot;the social relationships on the part of the partic ipants in the conversation that determine, for exa mple, the choice of honorific or polite or intimate or insulting speech levels&amp;quot;. In addition to honorifics, examples include nicknames, informal names, and titles of respect.</Paragraph>
    <Paragraph position="7"> Fillmore also notes that certain verbs, such as come and go, are also deictic, as they imply a here and a there.</Paragraph>
    <Paragraph position="8"> Considerable work in the field has also been by Lyons. He makes note of the deictic center- the &amp;quot;zeropoint of deictic context&amp;quot;, which is always egocentric for the speaker \[1968, 1977\]. This zero-point is in relation to person, time, and space, with the center representing I, here, and now. Deictic terms are expressed in relation to this deictic center. Here and this are closer than there and that, and now is closer than tomorrow. It is illustrative that the first, second, and third person radiate outward from the dei ctic center. Lyons reminds us that deictic terms may also have non-deictic uses \[1973\]. &amp;quot;This idea is not new&amp;quot; refers anaphorically to some previously expressed idea, while &amp;quot;This book is dedicated to my husband&amp;quot; refers deictically to the book containing the dedication. Finally, Lyons discusses tensed verbs as time deictics \[1977\]. Note that we need to know the time of utterance to work out any truth conditions of a tensed sentence.</Paragraph>
    <Paragraph position="9"> I have thus far summarized work on deixis. As for deixis in relation to hypertext, no previous work exists that I am aware of. I hope that this study (a  deictic!) will shed some light on this new phenomenon. null 3. Forms of Hypertext Anchors  Before addressing my first question the forms authors use for hypertext anchors - I will briefly discuss my general methodology. Methodology specific to each of the other three questions will be covered in that question's section, in turn.</Paragraph>
    <Paragraph position="10"> The data used for all four questions consists of pages from the World Wide Web collected in April 1996.</Paragraph>
    <Paragraph position="11"> There are of course many types of Web pages (Roberts \[1997\] discusses the different genres found). For this study, I collected pages from two types: the &amp;quot;home pages&amp;quot; of college students, and the &amp;quot;home pages&amp;quot; of commercial enterprises. A home page is the initial, top-level page at a &amp;quot;web site&amp;quot;, with a web site typically consisting of multiple related pages. For consistency, I used home pages only.</Paragraph>
    <Paragraph position="12"> For college student data, 126 different Web pages (containing roughly 1000 anchors) were collected randomly from three institutions: Cornell University, the University of Texas at Austin, and the University of Oregon. For commercial data, 31 Web pages (containing roughly 500 anchors) were collected.</Paragraph>
    <Paragraph position="13"> The commercial businesses included well-known firms in varying fields: examples are Apple Computer, Prudential Insurance, Pepsi-Cola, Sony, Whirlpool, Disney, and the American Stock Exchange. null After each page was downloaded, an automatic script stripped out the words used as anchors.</Paragraph>
    <Paragraph position="14"> (Fortunately, Web pages are written in HyperText Markup Language (HTML), which has specific tags surrounding the anchors). The anchors were saved to a separate file, as was the original Web page, for reference. Graphical anchors were ignored, unless they had &amp;quot;alternate&amp;quot; text associated with them, for the use of non-graphical Web browsers.</Paragraph>
    <Paragraph position="15"> To answer question (1), I then went through the list of anchors, and manually c~ed them per the following categories:  The results are shown in Table 1, which lists in roughly decreasing order the forms authors used for hypertext anchors, by number and percentage. The table is further broken down by source (college student and commercial), for comparison. Note that the &amp;quot;Full Sentence&amp;quot; category has been subdivided into &amp;quot;Imperative&amp;quot;, &amp;quot;Question&amp;quot;, and &amp;quot;Declarative&amp;quot;.  (Percentages do not exactly total 100.0, due to rounding) We can now answer our first question, as to the forms hypertext anchors take. The overwhelming majority are noun phrases, with full sentences taking a distant second. Note however, that commercial Web authors tend to use fewer noun phrases and more full sentences than college students (the diffe rence is significant, per a chi-squared test, with p &lt;= 0.01). Perhaps this is because commercial sites try to present a more polished image.</Paragraph>
    <Paragraph position="16"> Our first answer regarding hypertext immediately raises another questio n: Why are the overwhelming majority of anchors noun phrases? An intuitive reply is that authors are viewing hypertext anchors as referring expressions. Since referring expressions in more familiar non-hypertext are almost always noun phrases, authors choose noun phrases for their anchors. Continuing this theory, to the extent that authors are not using noun phrases as anchors, to just this extent are they &amp;quot;branching out&amp;quot; in the new phenomenon of hypertext, and discovering linguistic forms other than noun phrases for encoding hypertext links. Confirmation of these speculations, however, must await fu~er research.</Paragraph>
  </Section>
class="xml-element"></Paper>