Linguistic time (LT) shares some characteristics of Euclidean time, namely immediacy, universality, linearity, staticity, and unidirectionality, but lacks the features of infinity and continuity.
First, LT is discrete rather than continuous - this derives from the discreteness of linguistic forms. How finely one wishes to divide the LT dimension for analytical or expository purposes is arbitrary up to a point: No finer linguistic temporal unit may be admitted than the smallest desired linguistic unit itself - the smallest unit may not be indexed by separate times. Indeed, beyond some level not much below that of the phoneme, linguistic time cannot be further subdivided. Normal time, on the other hand, is susceptible to much finer subdivision, to the point that it is not inappropriate to describe normal time as ``continuous''.
Second, the LT dimension is finite, though unbounded, in the same way as are well-formed linguistic units like sentences or conversations. This follows from the preceding consideration: the number of times used to sequence the elements of a particular linguistic form cannot be greater than the number of smallest linguistic units in that form. Every particular linguistic form is finite (notwithstanding the claim that there is no upper bound to the length of possible sentences of given languages), and thus contains a finite number of smallest units, each of which may be indexed by at most one distinct time. For any given form, then, the dimension of linguistic time is finite. Euclidean time on the other hand is infinite, and extends indefinitely into the past and future.
Linguistic time is also distinct from normal or Euclidean time in that the ``same'' linguistic event or object may be produced or comprehended in varying amounts of normal time. Linguistic representations may be seen extratemporally, in the sense that speakers (and linguists) may examine them in normal time, somehow outside their internal temporal sequence. The possible repeated mental reviewing of a sentence during the comprehension process is not different in this connection from rereading a sentence as a linguist, perhaps while considering it, usefully, as a fixed, extra-temporal, abstract structure. In both cases, a record (whether in memory or on paper) of a temporal sequence is reexamined after the events recorded. Any record of a temporal sequence, examined after the events recorded, is indeed a static abstract structure. This is the point linguists make by avoiding reference to time in linguistic theory. Still, they are ``temporal'' sequences nonetheless, not just because they are conceived, produced, perceived, and comprehended in normal time, but because linguistic representations have a specifically temporal organization.
Despite the differences between ``real-time'' and ``linguistic time'' there are enough similarities between them to justify retaining the term ``time'', and associated temporal predicates such as ``before'' and ``after''. The five remaining properties will be considered: immediacy, universality, staticity, linearity, and unidirectionality.
First, elements in linguistic representations occur at particular locations in the sequence. This property of elements of ``occurring at'' some (temporal) location is the same as having a special time, now, associated with the element. Linguistic representations have this feature, ``immediacy''. Second, LT is ``universal''. No element in a linguistic representation lacks a (temporal) location and extent. Every linguistic element in a representation is at some time. Third, linguistic representations in which elements are composed of a sequence of other elements (as phrases are composed of morphemes, morphemes of phonemes, etc.) have the feature of staticity. A constituent which contains elements at different temporal locations may be said to be ``static'' in the sense that that constituent extends across a span of times. To this extent LT is not exclusively linear, as in the phonological representations of SPE. Fourth, linear ordering is crucial in many linguistic representations: phonemes, words, phrases, and utterances are all formed into larger units by the device of linear ordering, by which a unit is constrained not to both precede and follow any other unit.
Fifth, and most important, linguistic representations are unidirectional. Despite this fact, most generative linguistics takes the opposite, bidirectionalist, and notably incorrect position. It is commonly assumed that ``before'' and ``after'' are arbitrarily chosen and might just as well be reversed. Thus there are left-branching as well as right-branching languages, languages with prepositional as well as postpositional phrases (for consistency with this bidirectionalist viewpoint, PP's should be termed ``left-positional'' and ``right-positional'' phrases), etc. The use of the metaphor of handedness in referring to temporal direction in linguistic representations would seem to emphasize the arbitrary nature of directionality in linguistic structure, since the direction of writing is widely understood to be arbitrary.
As noted before, the linearity of linguistic representations does not in itself require unidirectionality. Linearity is a formal property of strings of linguistic forms, while unidirectionality is an empirically established restriction on possible directions of ordering of elements in linguistic structures. Unidirectionality is an empirical finding, that a certain sequence can occur in one order, but not in the reverse order. In particular languages, of course, most syntactic and phonotactic structures are restricted to one direction of ordering or another, so within the discussion of a particular language, ``before'' and ``after'' are entirely appropriate. The issue is whether Universal Grammar has unidirectional phenomena.
In phonetics it is quite clear that ordering of events is unidirectional. A burst, for example, cannot precede its associated closure. Similarly, while jaw opening and closing may be mirror images of each other, each depends on the efforts of different, antagonistic muscles. So the muscular activities behind jaw movement cannot be described without reference to the direction of the movement in time.
Similarly vocal-fold vibration may depend on the preceding state of the glottis. There is an intermediate point of glottal opening at which the vocal folds will continue to vibrate if they had been vibrating immediately before, and at which the folds will not vibrate if they had not been vibrating immediately before, despite the fact that the larynx may have the same degree of glottal opening, vocal-fold tension, etc., in both cases. That is, the voicing state is physically perseveratory, in this intermediate laryngeal position. This effect is unidirectional; it cannot be anticipatory, since physical objects are not prescient. Of course, such perseveratory physical effects must be distinguished from anticipatory voicing assimilations, etc., which must be explained in other ways. Examples of temporal irreversibility in phonetics may be multiplied indefinitely.
In phonology as well, there are unidirectional phenomena such as syllable-structure, in which it is impossible for an onset to follow a rhyme, a coda to precede a nucleus, etc. Similarly at the level of syntax, heavy-NP shift can only go in one direction, namely forward.A.5 There may be many other linguistic phenomena which are unidirectional.
The formal marking of ``linguistic time'' may be done by a discrete, finite, unidirectional, linearly ordered set of indices. The elements of any given representation are each (universality) associated with (immediacy) one or more (staticity) indices. Linguistic time thus constitutes a discrete, finite, linearly ordered dimension along which linguistic events (or elements) are organized. ``Times'' within a linguistic representation may be thought of either as formal objects associated with each element (in which case concrete ordering, such as representing order between elements by their relations as written on paper, is obviated ) or as concrete ``locations'' of elements within a spatially or otherwise laid-out sequence (which obviates formal marking of elements by indices).
The number of LT indices of a given sequence of linguistic forms is defined by the number of elements in the sequence. Of course elements may be repeated in LT without repeating their time of occurrence. In other words, there may be multiple tokens of a given type.
Type-token confusions lead to logical traps which must be avoided. LT is an indexing of the entire sequence in question. In examining a phrase in which a word repeats, the first phoneme in the first instance of the word is not indexed by the same linguistic time as the first phoneme in the second instance of that word. Of course, if the universe of discussion is limited to that word, then the ``first phoneme'' has only one index. The formal indexing of elements in a sequence depends on the sequence.
So far we have shown that certain features of time are implicit in linguistic representations. The game has been a formal one; we have not accounted for anything new that wasn't accounted for before; rather we have simply made explicit what was implicitly assumed in the past. The principles of linguistic time presented should be non-controversial, except for the issue of unidirectionality as applied to Universal Syntax. Now, we will see how the indexing power of linguistic ``space'', in conjunction with that of linguistic time, can be applied to autosegmental phonology.
Multi-linear phonology makes it necessary to add to the indexing power of LT additional features of structured simultaneity (or ``place'', in addition to time), in order to specify the organization of multiple ``tiers'' (sequences) of linguistic objects. This linking feature orchestrates multiple series of events or elements using both sequencing and overlapping.