| Home | Feature Inventory | Related Issues | Search | Contact |
|
Grammatical Features Home > Feature Inventory > Tense
TenseAnna Kibort
1. What is 'tense'Following Comrie (1985:viiff), we take tense to be the grammaticalisation of location in time. Three parameters are traditionally cited as relevant in defining tense and indentifying tense distinctions: (a) the location of the deictic centre (either at the present moment - in the so-called 'absolute tenses', or at a different point in time - in the so-called 'relative tenses'); (b) the location of the situation with respect to the deictic centre (i.e. prior to, subsequent to, or simultaneous with the deictic centre); and (c) the distance in time at which the situation referred to is located from the deictic centre (though see §4 below for comments on this point). In the discussion of the category of tense, the term 'situation' is understood as an event, process or state, without consideration of its internal temporal contour. The internal temporal contour of the situation provides the conceptual basis for the notion of aspect (see §5 below, and the entry on 'Aspect' in this Inventory). This conceptualisation of time, which appears to be adequate for an account of tense in human language including all time location distinctions found in natural language, can be represented diagrammatically as follows: 
The diagram represents time as a straight line, with the past represented conventionally to the left and the future to the right. The present moment is represented by a point on that line, labelled S (mnemonic for 'speech time'). Several things are intentionally left unspecified. One of them is whether the time line is bounded at either the left or the right, including whether it bends to form a circle. While this is an important philosophical issue, it does not seem to be relevant for the grammaticalisation of time. Similarly, conceptualisations of time as cyclic are found in all cultures, but on a macroscopic scale which does not have a bearing on tense distinctions. Furthermore, the diagram does not represent the flow of time - that is, it does not indicate whether S (or, Ego) moves along a stationary time line, or whether time flows past a stationary reference point S (or, Ego). This is another important philosophical question, but again it does not seem to play a role in the analysis of tense as a linguistic category. However, many of these culture-specific conceptualisations of time are metaphors that are important sources of time expressions across languages. For temporal-deictic distinctions, the speech situation S projected onto the time line serves as the basic orientation point - that is, is always the primary point of reference (R0). One of the extra-linguistic presuppositions for an utterance is constituted by the speaker's consciousness of the relation of the speech situation S to the reported situation E (mnemonic for 'event') along the time line (Thelin 1978:37). A past situation is therefore located in the time before and not including the present moment, a future situation (a prediction, imposition, or an instance of pre-planning) is located in the time after the present moment, and a present situation, whether continuing or repetitive, is located in the time that includes the present moment, regardless of whether it encompasses a shorter or longer strech of time. This concept of time, which relates a situation to the time line, is essential to the linguistic category of tense (Comrie 1985:6; see §3 below). It is interesting to note that a further parameter that could theoretically be posited for tense distinctions - that of a specific location in time, or a specific time lapse - does not seem to be grammaticalised as tense. In those cultures which care about, and are able to capture, very precise location in time and very fine distinctions of time, these are usually expressed using existing grammatical patterns and the appropriate lexical items which may be combined with mathematical expressions in order to gain precision (e.g. 10.45 am on Friday, 9 June 2006; nanosecond; 10-6 seconds). On the other hand, in cultures which lack the technology to capture precise temporal locations, or attach little value to precision in temporal location, such precision may not be attainable even lexically. For example, in Yidiny, an Australian language, it is impossible to distinguish lexically between the concepts of 'today' and 'now' (Comrie 1985:7-8; Yidiny information from Dixon 1977:498-499). Jump to top of page/ top of section 2. Expressions of 'tense'Location in time can be achieved linguistically in many different ways ranging from purely lexical to grammatical. Comrie (1985:8-9) suggests grouping them in three classes:
In order to be regarded as a (grammaticalised) tense, the expression of location in time has to be integrated into the grammatical system of the language. In contrast, a lexicalised expression of the location in time indicates its integration into the lexicon of the language, but does not entail any necessary consequences for the language's grammatical structure. Grammaticalisation, as opposed to lexicalisation, of the location in time, correlates with two parameters: obligatory expression and morphological boundness. The very rough rule is that a tense is grammaticalised if its morphological expression is obligatory even if the information carried by the exponent is redundant. For example, in the English sentence Last year I bought a new car "the choice of a tense other than the simple past would make the sentence anomalous, although the information that the event took place in the past is expressed unambiguously by last year" (Dahl & Velupillai 2005:266). The obligatoriness of tense can also be demonstrated in languages which allow for tense marking to be omitted in some circumstances. This has been argued for Central Alaskan Yup'ik whose tense system is relative (the deictic centre can shift even within one sentence between the time of speech and another reference point) and where the absence of an overt tense suffix has to be recognised as a meaningful zero (Mithun 1999). Morphological boundness is perhaps a slightly more problematic criterion, which is not necessary in itself - as in Yemba (Niger-Congo), where tense is expressed primarily by means of auxiliaries (i.e. not bound morphemes) which are also clearly not separate lexical items which could be seen as contributing to the compositional meaning of tense (Comrie 1985:11, 86-87). (For more detailed discussion of grammaticalisation and lexicalisation in general, and the grammaticalisation and lexicalisation of the location in time, see Lyons 1977:234-237, and Comrie 1985:10-12). Tense is typically a morphological category of the verb, or verbal complex, and it can be expressed either by verbal inflection (on the main verb or the auxiliary - as in English), or by grammatical words adjacent to the verb (such as the Yemba tense auxiliaries; see above). It can also be analysed as a grammatical category of the clause. Anderson (1992) classifies tense as 'phrasal inflection' (distinct from 'configurational inflection' such as case, 'agreement inflection' such as number concord on English verbs, or 'inherent inflection' such as gender on Latin nouns), because it is a property that is "assigned to a larger constituent within a structure" (a clause) but "realized on individual words" (verbs). Similarly, Booij (1994:30) argues that tense has scope over a whole clause (however, he classifies it as 'inherent inflection', opposed to 'contextual', because "the tense of the verb is not determined by syntactic structure"; see §3 below). It is, therefore, not unexpected that tense can also be expressed by a marker placed in the position of sentence particles (as in the sentence-second position in Warlpiri, Australian; Hale 1973, cited in Comrie 1985:12), that languages may allow multiple marking of the tense value on different elements in the same clause (see §3 for discussion, and §5 for some examples of tense markers found outside the domain of the verb), or that tense systematically does not occur on non-finite verb forms. Finally, it is important to note that there is often no agreement between scholars about the classification of individual grammatical forms, especially when the verbal forms and constructions simultaneously involve semantic elements from different domains, typically those of aspect and mood (see Dahl 1985), but also case (e.g. in Australian languages - see Evans 1995; 2003) and other categories. In fact, there are seldom grammatical markers that express just the temporal location of the situation. Jump to top of page/ top of section 3. The status of 'tense' as a featureTense is often assumed to be and frequently referred to as a 'morphosyntactic category' or 'morphosyntactic feature'. As summarised in Mithun (1999:23), tense is "one of the most frequently cited examples of a prototypical inflectional category", seen as having "relevance to syntax". However, as pointed out by Mithun, the definition of syntactic relevance depends on the particular view of syntax that is assumed. Although tense is typically an inflectional category, Booij (1994, 1996) classifies it as 'inherent inflection' (alongside number on Dutch nouns), rather than 'contextual inflection' (such as number agreement on Dutch verbs) because "the tense of the verb is not determined by syntactic structure" (1994:30). We propose that, for a grammatical feature, to be 'relevant to syntax' means that it is involved in either syntactic agreement or government. In many familiar languages the feature 'tense' encodes regular semantic distinctions and is an unquestionable inflectional category. However, it is not required by syntax through the mechanisms of either agreement or government: syntax is not sensitive to the tense value of the verb. Therefore, the familiar instances of the feature 'tense' are morphosemantic, but not morphosyntactic. Thus, tense is typically a morphosemantic feature. Corbett (2006:138-141) investigated whether tense, aspect, mood and polarity could function as features of agreement - and therefore be morphosyntactic features - but did not find convincing examples and concluded that "the argument for tense, aspect, mood and polarity as agreement features remains open". Two types of construction in Kayardild which show elements of a clause matching in tense are probably better analysed as instances of multiple marking of the same value of tense in the clause. The first type of construction involves so-called 'verbalising cases' (Evans 2003; earlier 'verbal cases', Evans 1995). Besides ordinary cases, Kayardild has several verbalising cases that are marked on all parts of the selected noun phrase. Verbalising case is involved in various domains; once an element is marked by verbalising case, it takes tense-aspect-mood and polarity inflection matching that of the main verb. For example, beneficiaries take the verbalising dative case marker -maru which appears on all parts of the noun phrase; the elements marked with the verbalising case then take verbal inflections for tense, aspect, mood and polarity. In the example below, which Corbett (2006:80, 139) quotes from Evans (2003:215), the items marked with verbalising dative (V_DAT) are also matching in tense:
In a dependency approach to syntax, which implies asymmetrical marking, it can be argued that tense, aspect, mood and polarity are primarily features of the verb, and if other items are marked secondarily, we have asymmetrical marking and an instance of agreement (Corbett 2006:139). However, if one believes that tense, aspect, mood and polarity are features of the clause, then marking of these features on items other than on the verb is symmetrical marking and hence does not qualify as agreement. This argument is paralell to the argument about case marking within a noun phrase. On the dependency view, the adjective can be seen as agreeing in case with the noun which is head of the phrase (this view is expressed, for example, in Mel'čuk 1993:329, 337). However, on the constituency view, the matching of case within the noun phrase results from the case being imposed on the noun phrase (often through government by the verb), and the case value being shared within the noun phrase. At this point we will, therefore, treat instances of tense matching - as in (1) - as multiple marking, or sharing, of the tense value within the clause. The other type of construction from Kayardild involves what Evans (2003:215-216) calls a 'verbal group', which is a sequence of serialised verbs consisting of an obligatory main verb plus up to two further verbs functioning as markers of associated motion, adverbial quantification and aspect. They appear in a fixed order in a single intonational group, and the meaning of the group may be non-compositional. In Kayardild, all verbs in a verbal group take identical values for tense, aspect, mood and polarity, as in the following example (Evans 2003:223, also cited in Corbett 2006:140), where the two verbs of the verbal group match in tense:
(Note that the verb thaa-tha 'return' here means 'go off and V' rather than 'V and return', therefore (2) could be uttered in a situation when someone has gone off for a swim and not necessarily returned from the swimming yet.) Evans emphasises that the past tense "is used because of a rule that all words in a verbal group must agree in TAMP [tense, aspect, mood and polarity - AK], not because it is independently locating 'returning' in the past" (2003:223-224). Examples like (2), of tense (and other verbal features) matching within a serial verb complex, are common in serial verb constructions, and the following sentence illustrates the same phenomenon from Paama (Paamese) (Oceanic, a language of Vanuatu; Crowley 2002:68)
As suggested by Corbett (2006:140), in examples such as (2) and (3) it is still possible to view the verbal group as a semantic and syntactic unit, and tense (as well as aspect, mood and polarity) as being assigned to this unit. This is consistent with viewing a serial verb construction as "a sequence of verbs which act together as a single predicate, without any overt marker of coordination, subordination, or syntactic dependency of any other sort. Serial verb constructions describe what is conceptualised as a single event" (Aikhenvald 2006:1; though see Baker & Harvey forthcoming for discussion of serial verb constructions versus co-verb constructions). Therefore, we conclude that tense matching in verbal groups, as in (2) and (3), is also better treated as an instance of multiple marking, or sharing, of the tense value within the verbal group - unless one adopts a dependency view of syntax, or reformulates the definition of agreement and revises the typology of controllers, domains and features. (For more discussion of tense as a morphosemantic, but not a morphosyntactic feature, see Kibort 2010.) Jump to top of page/ top of section 4. The values of feature 'tense'Time itself does not provide any landmarks - such as the beginning or the end of time, or any other absolutely specified point - which would enable us to locate situations. Therefore, relating a situation to the time line is possible only if we establish some arbitrary reference point (note that this in a sense makes all time location relative, even though the location of situations relative to the present moment is usually referred to as 'absolute tense'). The point on the time line that is universally the basic reference point is the time of speech, i.e. the present moment - which makes tense a deictic category. Another logical possibility of a reference point - a 'famous event' in the history of the language community - is apparently never used for tense as a grammatical category, though it can be used for lexically composite expressions or even lexical items (e.g. pre-Revolutionary, post-Reformation; Comrie 1985:14), and for non-linguistic purposes (e.g. the calendar system). Furthermore, despite the invention of writing and sound recordings that enable temporal dislocation of speaker and hearer, as far as the lexicon and the grammar are concerned, human language makes the assumption that there is only one deictic centre common to both speaker and hearer (Comrie 1985:15-16). As was mentioned in §1 above, one of the parameters that is often regarded as contributing to tense distinctions is the distance in time at which the situation referred to is located from the reference point. In languages which code different degrees of remoteness, these are usually labelled as different tenses. The most prolific systems coding several different degrees of remoteness have been found in sub-Saharan Africa (especially Bantu languages), Australia, and the Americas. Since temporal distance is relevant only with respect to the parameters of 'before' and 'after', we find distinctions of temporal distance only among past tenses and future tenses. The reference point from which the temporal distance is measured can be the present moment or a different point. The most common systems make two or three distinctions with regard to temporal distance, but five-way oppositions have been found in Africa, Australia and the Americas (e.g. in Yagua, an official language of Peru), while Kiksht (a nearly extinct Lower Chinook language of America) has been reported to have around seven oppositions (Comrie 1985:87,99-100, after Hymes 1975). For detailed discussion of various languages coding the degrees of remoteness, see Comrie (1985:83-101). Although the degrees of remoteness are usually referred to as tenses, alternatively, this parameter could be seen as an expression of a different category, say 'remoteness' or 'distance', which is orthogonal to the category of tense. (Note also that the same concept of 'remoteness' or 'distance' could be seen as intersecting with the category of 'person' in some languages, yielding for example the proximate versus obviative distinction in the third person; see the entry on 'Person' in this Inventory for more details.) This explanation is compatible with the observation that some languages, e.g. Burarra and some other Australian languages, and also Kiksht, have 'tenses' that appear to refer simultaneously to more than one time segment, and thus not fit well within most current conceptions of tense (Comrie 1985:50,88-80, after Glasgow 1964, and Dixon 1977:498-500). In fact, the tense system in Burarra results from "the combination of two oppositions, one an absolute cut-off point between today and earlier than today, the other between recent and remote within each of these two time frames" (Comrie 1985:89). This makes Burarra's two tenses that code the degree of remoteness ('close' and 'remote') alternate cyclically. They receive present time reference within the 'today' frame, and reference to the most recent period that is not today (i.e. the last few days) within the 'not today' frame. Thus, in order to derive all tense meanings found in natural language, it is sufficient to examine the parameter coding the location of the situation with respect to the reference point. It is immediately obvious that the relative positions of just two points on the time line - the speech time S, and the event time E - are not sufficient to account for all the different tense meanings found in language. Therefore, since Reichenbach (1947) most tense theories have used a third point in time, labelled R (mnemonic for 'reference point'), to capture all possible tense distinctions. Although the speech time (S) is always the primary reference point for tense, R serves as the temporal standpoint for the clause and may, but does not have to, coincide with the basic orientation point S. The reference point R indicates the speaker's chosen psychological or imaginary temporal location and alters, in a systematic way, the viewing of the temporal location of a situation whose actual location represented by E remains constant with regard to S. The tense system of a language results from a selection of the following distinctions identified with the three semantic primitives: the time of speech (S), the time of the event (E), and the reference point (R):
To sum up, tense meanings that are possible in human language result from the possible arrangements on the time axis of the three primitives {S,E,R} plus the multiplications of this set. Tense values are grammaticalisations of particular tense meanings or distinctions. Despite the wide range of logical possibilities (which do not have an obvious limit), not all of them are found; it is very common to find neutralisations of various temporal distinctions, often in combination with other (aspectual, modal) dinstinctions. In some linguistic traditions, combinations of tense, aspect, and mood values are referred to as 'screeves' - see the 'Screeve' entry in this Inventory. Neutralisations of temporal distinctions have the effect of segmenting the time axis in different ways:
Jump to top of page/ top of section 5. Oddly behaving tense markers
Nordlinger & Sadler demonstrate that the encoding of TAM on nominals can have one of two broad functions (2004:778-779ff):
Specifically, the proprietive, ablative and oblique suffixes here are being used 'modally': the proprietive and the 'potential' verb inflection express futurity; the ablative and the 'past' verb inflection express 'prior occurrence'; and the oblique and the 'desiderative' verb inflection express strong emotion (in this instance desire) towards the event. The allative may also be used modally, as in (5), to express 'that the event is spatially oriented towards the speaker, or that it is just beginning, or just coming into the speaker's awareness' (Evans 1995:108). Butt (2006:10) notes that "this phenomenon is not confined to Australian languages, but that languages like Finnish have also been implicated in this type of case usage." Jump to top of page/ top of section 6. Problem cases
Smith (2005) suggests that languages fall into three categories with respect to expressing grammatical tense: (1) there are fully tensed languages in which all main clauses have an obligatory tense marking (as in English, French, German, or Hindi - though all these differ in how aspect is realised); (2) mixed-temporal languages have some of the characteristics of tensed languages: they have inflectional markers and/or temporal particles and clitics that give direct temporal information, but these are syntactically optional; thus, a given sentence may or may not convey temporal information; examples of such languages are Navajo and other Athabaskan; (3) finally, there are languages without temporal inflections or particles, such as Mandarin Chinese and Thai, some Mayan languages, and probably others - the languages in this category are tenseless and use indirect means to locate a situation temporally. See Smith (2005) for an outline of a unified approach to direct and indirect temporal interpretation.
Jump to top of page/ top of section 7. Key literature
REFERENCES
Jump to top of page/ top of section How to cite this entry:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Occasionally, tense markers can be found outside the domain of the verb.
For example, in Malagasy, Malayo-Polynesian, certain spatial
adverbs (locatives), temporal adverbs, and prepositions obligatorily
agree in tense with the verb (Comrie 1985:13, after Randriamasimanana 1981:355-367;
also Keenan & Polinsky 1998:566-567). Semantically tense is not a
property of these elements, but rather of the verb phrase or the clause.
Therefore, this can be analysed as an instance of multiple marking of
the tense value on more than one element in the clause, as was
discussed in §3.
The question of tenselessness: distinguishing tense from
mood and aspect.
Burmese and Dyirbal appear to
be good examples of tenseless languages, with the grammaticalisation of
mood (realis and irrealis) but
no grammaticalised time reference (Comrie 1985:50-52). Similarly
Manipuri uses realis mood and other temporal means
to encode the relationship between speech time and event time, as it
does not have a grammatical way of marking tense (Poudel 2007).
Lillooet and other Salishan languages spoken in Canada
are also reported not to have tense distinctions, although the issue is
debated.