Continuing the dialogue between corpus linguistics and grammaticalization theory: Three case studies

3.1 Looking for parallels between ontogenetic and diachronic grammaticalization – Case study 1: Existential there

The first case study here concerns the grammaticalization of existential there. It is generally agreed that existential there developed from the distal deictic adverb there, meaning ‘in/at that place’ (see, among many others, Lyons 1967: 390; Bolinger 1977: 91–92; Breivik 1983, 1997: 32; Johansson 1997; Pfenninger 2009: 49–53), ^[2] through processes of grammaticalization and reanalysis, probably before the first written records of English (see Breivik 1977: 346). If the beginning of these processes of change indeed pre-date Old English times, lack of data for pre-literary periods may make it difficult (if not impossible) to plot the split of existential there from locative there. Nevertheless, the Old English period clearly represents a transitional stage in the historical development of existential there. Notice, for instance, that existentials with no introductory element represented the most common existential pattern at this early stage (see Breivik 1983: 278, 319; López-Couso 2006: 182, among others), while there-existentials became the default type of existential construction only during the Middle English period (see Breivik 1983: 320–321).

Looking at the ontogenetic development of existential there, and using longitudinal corpus evidence from the CHILDES archive (MacWhinney 1995), Christopher Johnson (1999, 2001, 2005) has shown that, in the learning process, children use locative there as a source for existential there, via overlap utterances which share formal and semantic-pragmatic properties of either construction. Taking Johnson’s analysis as a starting point, I examine in López-Couso (2011) whether the developmental relation between deictic there-constructions and there-existentials explained by Johnson’s theory of “constructional grounding” or “developmental reinterpretation” for child language acquisition is valid also in diachronic terms.

My analysis of the prose texts of the Old and Early Middle English sections of the HC ^[3] suggests that the emergence of existential there in the history of the language ran parallel to the three developmental stages proposed by Johnson for Child English (see Johnson 1999: 95–114; 2001: 131–134).

Stage 1. At the initial stage, both ontogenetically and diachronically, there is a distal deictic adverb, meaning ‘in/at that place’, as shown in (2a–b):

Stage 2. Over the course of time, locative there begins to be found in overlap contexts, which allow both a deictic and an existential reading, as in (3a–b). Johnson describes such contexts as “overlap deictics”, i.e., statements in which there performs an existence-informing function in addition to its original pointing-out function. Historically, as mentioned above, the transition from Stage 1 to Stage 2 is most likely to have taken place before the Old English period.

A particularly relevant subtype of overlap deictics identified by Johnson in Child English is the so-called “double-locative overlap deictic” construction (Johnson 1999: 85), illustrated in (4a) below; it is characterized by the co-occurrence of there with an additional locative expression in the clause. Constructions of this kind seem to play a prominent role in the process of change in first language acquisition, since children need to look for a new (non-locative) function of there, which would otherwise be perceived as redundant. As in early child speech, double-locative overlap constructions seem to have served as bridging contexts (Heine 2002) in the diachronic grammaticalization of existential there (cf. (4b)):

Stage 3. After the overlap stage, children produce the first unambiguous instances of there-existentials. Notice, for example, that in (5a) a locative reading of there would not make sense, given the co-occurrence of the proximal deictic here in the same clause. Historically, the differentiation between locative there and existential there is finally accomplished, most probably in post-Old English times (cf. the Early Middle English example in (5b)). Interestingly, it seems that unequivocal instances of existential there appear earlier in affirmative clauses both in Child English and in the history of the language, and later spread to negated contexts (cf. (5c–d)):

Figure 1 summarizes the development of the overlap deictics (cf. (3)), double-locative overlap deictics (cf. (4)), and unambiguous existentials (cf. (5)) in the Old and Early Middle English sections of the HC: ^[4]

Figure 1:

Overlap deictics, double-locative overlap deictics, and unambiguous existentials in the prose texts of the Old and Early Middle English sections of the HC (normalized frequencies per 100,000 words).

The close similarity between the ontogenetic and diachronic processes of grammaticalization of existential there just outlined is probably more than mere coincidence. Similar corpus-based studies may reveal the existence of further parallels in the way grammaticalization works in child language acquisition and in the history of the language.

3.2 Recognizing and documenting incipient or ongoing grammaticalization – Case study 2: Like-parentheticals

My second case study brings us forward in time to the present day. From as far back as the early 1990s, one of the controversial issues in grammaticalization has been whether it is possible to recognize and document incipient and ongoing grammaticalization processes. The question was addressed by Christian Lehmann (1991), who reviewed, from the perspective of grammaticalization, a number of constructions in both the nominal and verbal spheres in contemporary non-literary German; these constructions were in vogue at the time, and although not accepted by standard grammars they did show some of the features characteristic of grammaticalization. ^[5] In the closing section of his article, Lehmann admits that the study of ongoing grammaticalization processes “is subject to a serious problem of verification” and that “it is next to impossible to know which of the changes that speech habits currently exhibit are synchronic manifestations of ongoing genuine language change, and which of them are but ephemeral fashions” (Lehmann 1991: 532).

In spite of this note of skepticism – and similar ones – the last few years have seen a number of studies on various grammaticalization phenomena which have convincingly demonstrated that, at least for English, ^[6] the use of corpora can help to pinpoint incipient or ongoing grammaticalization. Evidence here can be found in the analysis of the emergence of the epistemic/evidential parentheticals looks like, seems like, and sounds like in contemporary American English, which I investigate in collaboration with my colleague Belén Méndez-Naya (López-Couso and Méndez-Naya 2014a, 2015). Examples of the pattern we are interested in are given here:

Notice that this parenthetical type differs from prototypical clausal parentheticals, such as I think, I believe, you know, and it seems (López-Couso and Méndez-Naya 2014b, 2014c), in that the conjunction like has become bonded to the verb as a result of a process of fusion, one of the characteristics Brinton and Traugott (2005: 27–28) associate with grammaticalization processes (see also Brinton 2008: 50).

For the analysis of the parentheticals under discussion we make use of the Corpus of Contemporary American English (COCA, Davies 2008–), which contains about 450 million words of material from 1990 to the present day, and is therefore an excellent source for the study of recent and ongoing developments in American English. ^[7]Figure 2 shows the distribution of relevant instances in the corpus from the early 1990s to the year 2009, arranged in five-year periods. The data show a clear increase in the frequency of like-parenthetical structures with look and sound over the last two decades. By contrast, like-combinations with seem remain rather stable over time, probably because of the high frequency of the bare pattern it seems, which constitutes the default parenthetical type with this predicate (see López-Couso and Méndez-Naya 2014a, 2014c).

Figure 2:

Like-parentheticals across time in COCA (1990–2009).

In order to determine whether the development of the parentheticals looks like, seems like, and sounds like can be conceptualized in terms of processes such as grammaticalization, subjectification, and/or intersubjectification, we have compared them with related complement constructions featuring the predicates look, seem, and sound and the “comparative complementizer” like of the type shown here: ^[8]

Looking closely at the morphosyntactic and semantic-pragmatic characteristics of the like-parentheticals attested in COCA, it can be seen that these combinations illustrate a number of features which are typical of grammaticalization and related processes of language change. For example, they have undergone a process of decategorialization: originally complement-taking clauses (see (7) above), they are downgraded to parenthetical constructions which no longer show a complete syntactic structure (see (6a–c)). Once they have achieved parenthetical status, these combinations become syntactically and prosodically independent and may therefore occupy different positions in the clause (e.g. medial (6a) and final (6b–c)), ^[9] thus coming close to adverbs. ^[10] The coexistence in contemporary American English of the different related constructions (the complement-taking clause, the parenthetical structure, and the quasi-adverb) points at another of the criterial features of grammaticalization, namely layering (see Hopper 1991).

Hand in hand with decategorialization, the parentheticals at issue show a high degree of morphosyntactic fixation; this is in clear contrast to the related complementation patterns, which display a wider range of variability. For one, as regards tense, aspect, and mood (TAM) distinctions, the like-parentheticals in our data show a marked tendency to occur in the present tense (75.6%; see the examples in (6) above) and, to a much lesser extent, in the preterite (24.4%), as shown in (8). On the other hand, when used as matrix predicates in complementation structures, the verbs look, seem, and sound not only occur in the present and the preterite, but also in combination with modals and other auxiliaries (cf. (9a–b)) as well as in non-finite forms in catenative constructions (see (9c)):

Besides their limited range of TAM distinctions, like-parentheticals also show a significant restriction as regards polarity. Our data from COCA indicate that these parenthetical constructions tend to occur in the affirmative (99% of all relevant instances), while negated like-parentheticals are only sporadically attested; sentence (10) below is one of these exceptional instances. In all such cases, the anchor clause is also negative, in accordance with the claim that a negative parenthetical can only occur in combination with a negative host (Kaltenböck 2007: 37; Schneider 2007: 149–150).

Further indicators of the high degree of morphosyntactic fixation of the like-parentheticals are their inability to show adverbial modification or to occur in interrogative structures. As shown in (11a–b), these two patterns are available for the related complementation structures:

Finally, even though the predicate and the conjunction making up the like-parentheticals remain orthographically separate, they do show fusion (see above): the predicate and like are bonded, and come to be understood as constituting a single chunk, thus disallowing intervening material. By contrast, the occurrence of intervening elements between the predicate and like is possible in the related complementation structures, as shown in (12):

In addition to the morphosyntactic features discussed so far, like-parentheticals also undergo interesting semantic-pragmatic changes. On the one hand, when used in the combinations currently under discussion, the predicates look, seem, and sound show reduced semantic content, their sensory meaning being bleached. On the other hand, lacking propositional semantics, like-parentheticals specialize in subjective and intersubjective meanings and functions (see López-Couso 2010). Thus, they can be used, for instance, to mitigate the speaker’s commitment to the truth of a proposition. Consider in this connection our earlier examples (6a–b), where the parentheticals it seems like and looks like indicate that the speakers do not fully commit themselves to what is said in the respective anchor clauses, probably suggesting a lack of definite evidence for their statements. Moreover, in conversation (see (13) below), like-parentheticals may also serve intersubjective functions as face-saving devices: they soften the speaker’s assertion and hence allow the interlocutor(s) to disagree: ^[11]

3.3 Corpora as a source for qualitative analysis of grammaticalization phenomena – Case study 3: Optional marker of expository apposition namely

The development of the like-parentheticals discussed in Section 3.2 shows that large databases like COCA can help us to identify, describe, and analyze incipient and ongoing grammaticalization. However, not all grammaticalization processes “leave a statistical imprint” (Mair 2004: 133) in corpora in the way existential there or like-parentheticals do. This is especially problematic for historical analyses of low-frequency phenomena such as the one in my third case study: ^[12] the development of the optional marker of expository apposition namely, illustrated in (14).

According to the Oxford English Dictionary (OED) and the Middle English Dictionary (MED), the origin of the adverb namely is found in the combination of the noun name and the derivational suffix -ly (see OED s.v. namely; MED s.v. nam(e)li adv.). A look at the entries for namely in these two dictionaries reveals the existence of three different meanings for the form over the course of time. The original meaning of namely seems to have been that of ‘particularly, especially, above all’ (see OED s.v. namely adv. 1; MED s.v. nam(e)li adv. 1). The first occurrence of the form I have been able to trace (example (15) below) dates from the late twelfth-century text of the Lambeth Homilies:

Similar instances survived into the Early Modern English period, but became obsolete after 1700. Example (16) is the last occurrence of this kind recorded in the OED:

In examples like (15) and (16), namely is used as a particularizer, whose function is to “restrict the application of the utterance predominantly to the part focused” (Quirk et al. 1985: 604; original emphasis).

A second possible meaning of namely in earlier stages of the language is ‘at least, at any rate’ (see OED s.v. namely adv. 2; MED s.v. nam(e)li adv. 3), as in example (17) from the MED. It seems, however, that this use of namely was rather rare and did not survive the Middle English period.

During Late Middle English, namely acquired a third possible meaning: ‘to wit, that is to say, videlicet’ (see OED s.v. namely adv. 3; MED s.v. nam(e)li adv. 2). Example (18), dating from the mid-fifteenth century, is one of the earliest attested unambiguous occurrences of namely with this meaning:

In such cases, namely is not a particularizer but rather functions as an optional marker of apposition (see Quirk et al. 1985: 1307–1316; Meyer 1992: 97), which can be omitted without affecting the grammaticality of the sequence. More specifically, in (18) namely marks a relation of expository apposition, where “the units are referentially or semantically equivalent” (Meyer 1992: 97). This is the only meaning and function that the adverb namely retains in Present-day English.

Figure 3 shows the distribution of namely in the HC and ARCHER 3.1 according to function. ^[13] As can be seen, in the Middle English material namely is recorded only as a particularizer, and in this function it is still sporadically attested by the end of the Early Modern English period. On the other hand, namely as a marker of expository apposition occurs in the corpus data only from the early seventeenth century onwards. This latter function is the only one attested in the Late Modern and Present-day English data from ARCHER.

Figure 3:

Functions of namely in the HC (M1–E3) and ARCHER 3.1 (eigtheenth- to twentieth-century British English) (normalized frequencies per 100,000 words).

Examples (19) and (20) from the HC show namely in its function as a particularizer and as a marker of expository apposition, respectively:

Namely is also attested sporadically in the Early Modern English data from the HC in the combination as namely, which functions as an optional marker of exemplifying apposition meaning ‘for example’ (cf. OED s.v. namely adv. 3b), as shown here: ^[14]

Two important features distinguish the particularizer namely from the marker of expository apposition namely:

(i) As happens with other items with a particularizing function, the particularizer namely can occupy different positions in relation to the part of the utterance that is focused. Although it most commonly occurs in pre-position, as shown in (19) above, it can also occur postposed, as in (22) below. Postposition, however, seems a highly marked option; in my data it is only attested in four instances out of a total of 55 (7.3%).

(ii) When occurring in pre-position, the particularizer namely can be preceded by the conjunction and. An analysis of the HC reveals that and, though optional, very frequently occurs with the particularizer namely in both Middle and Early Modern English (35 out of 51 relevant cases; 68.6%). Our earlier instance (16) and example (23) below from the HC illustrate this pattern:

Through univerbation, the combination and namely sometimes resulted in the form anameli/anamely, as in (24) from the MED. No instances of this kind, however, have been attested in my corpus data.

The core syntactic, semantic, and pragmatic features of namely in its two basic functions over time are summarized in Table 1.

Given that optional markers of apposition are rather infrequent in discourse, ^[15] tracing the origin of the appositive marker namely through statistical analysis of historical data can turn out to be a difficult task. However, “corpora can be (and should be) sources for qualitative as well as quantitative analyses” (Lindquist and Mair 2004: xiii).

A close reading of the very few attestations of namely in the HC suggests a possible grammaticalization route from the particularizer namely to the optional appositive marker namely. A comparison of the two uses of the form as shown in Table 1 suggests that this change involved both semantic and functional discontinuity. As happens with other particularizers, namely in its particularizing function could occupy different positions in relation to the part of the utterance that is focused. Although it most commonly occurred in pre-position (see (19) above), it could occasionally appear postposed, as in (22). As shown below, instances of the latter type may have served as bridging contexts (Heine 2002) or critical contexts (Diewald 2002) ^[16] in the development of namely from particularizer to optional appositive marker. Consider in this vein example (25), from Dan Michel’s early fourteenth-century text Ayenbite of Inwyt, a translation from the late thirteenth-century French work La Somme le Roi, a manual of moral instruction written by Friar Lorens d’Orleans at the request of King Philip the Bold. Here, namely functions as a particularizer, it is not preceded by and, and is postposed to the part of the utterance which is focused (þri þing). Example (25) can be compared with (26), which offers the same passage in The Book of Vices and Virtues, ^[17] a later independent translation of the same French original. The punctuation of this version suggests that the change from particularizer to marker of expository apposition could already have taken place by the mid-fifteenth century, possibly as a result of faulty scribal analysis and resegmentation, as suggested by Pahta and Nevanlinna (2001). Example (26) is the only instance in the Middle English section of the HC where namely would allow an interpretation as an optional marker of expository apposition.

It appears, then, that the development of namely as seen here – from a marker of particularization in Early Middle English to an optional marker of expository apposition in Late Middle English – shows that corpora may “provide insights into grammaticalisation processes that go beyond the statistical” (Mair 2004: 138).