The diachronic development of zero complementation: A multifactorial analysis of the that/zero alternation with think, suppose, and believe

2.1 That/zero alternation and the emergence of discourse formulas and parentheticals

In usage-based approaches to the that/zero alternation (Thompson et al. 1991a, 1991b; Aijmer 1997; Diessel and Tomasello 2001; Thompson 2002), frequently occurring subject–verb combinations, e.g., I think and I guess, are considered to have developed into conventionalized “epistemic phrases” (Thompson et al. 1991a, 1991b) or “discourse formulas” (Torres Cacoullos and Walker 2009). Torres Cacoullos and Walker (2009) argue that such discourse formulas have reached a high degree of autonomy (see Bybee 2003, 2006) from their productive complement-taking source construction. The frequency with which the zero complementizer is used is seen as an indication of this increasing autonomy. Following this rationale, Thompson and Mulac (1991b) argue that the absence of that points toward the blurring of the distinction between matrix clause and complement clause, i.e., to a reanalysis of this [matrix + complement clause] construction as a monoclausal utterance in which the complement clause makes the “main assertion” (Kearns 2007a), for which the matrix clause provides an epistemic or evidential “frame” (Thompson 2002). ^[1]Thompson and Mulac (1991b) show that the subject–verb collocations with the highest frequency of occurrence have the greatest tendency to leave out the complementizer that. It is exactly these sequences that “are most frequently found as EPAR [epistemic parenthetical] expressions” (Thompson et al. 1991b: 326), ^[2] which occur in clause medial or final position with respect to the (erstwhile) complement clause.

These synchronic, frequency-based findings lead Thompson and Mulac (1991b) to propose that that complementation (1), zero complementation (2), and parenthetical use (3) embody three degrees or three stages in a process of grammaticalization into epistemic phrases/parentheticals. ^[3] A study on the use of I think in Middle and Early Modern English (EModE) by Palander-Collin (1999) adds support to the diachronic validity of this grammaticalization path. Her data show an increase in the use of I think with the zero complementizer and a concomitant rise in parenthetical use.

Brinton (1996), on the other hand, takes issue with what she calls the “matrix clause hypothesis” and presents an alternative model which posits a paratactic construction with an anaphoric element rather than a complement-taking construction as the historical source construction. Brinton’s proposal is consistent with Bolinger (1972: 9), who states that “both constructions, with and without that, evolved from a parataxis of independent clauses, but in one of them the demonstrative that was added.”

Along similar lines, Fischer (2007) posits two source constructions for present-day parentheticals: what Quirk et al. (1985: 1111) have called subordinate clauses of proportion and the seeming zero complementation patterns that Gorrell (1895: 396–397; cited in Brinton 1996: 140 and Fischer 2007: 103) designates as “simple introductory expressions like the Modern English ‘you know,’” which stand in a paratactic relationship with the ensuing clause. Fischer (2007: 106) classifies the anaphoric connective element introducing such independent clauses as an adverbial derived from a demonstrative pronoun.

The notion of reanalysis, on which Thompson and Mulac’s (1991a, 1991b) account of epistemic parentheticals is based, has been subject to additional criticism. An important point here is the role of zero complementation. Kearns (2007a), for example, does not regard the occurrence of the zero complementizer with epistemic phrases/parentheticals as a diagnostic of the syntactic reanalysis involved in their formation; rather, she accounts for zero complementation in strictly pragmatic terms: it signals a shift in information structure such that the complement clause conveys the main assertion while the matrix clause loses prominence and has a modifier-like use (see also Diessel and Tomasello 2001; Boye and Harder 2007). These studies allow for a hybrid analysis in which some occurrences with zero complementation are adverbial in terms of function while syntactically retaining their matrix clause status. A further criticism regarding reanalysis concerns the necessity of that omission to the use of I think (and similar epistemic phrases) as discourse formulas. Both Kearns (2007a) and Dehé and Wichmann (2010) argue that complement-taking predicates followed by that, e.g., I think that, may also be analyzed as discourse formulas, the whole sequence having become routinized as a whole. In addition to providing prosodic evidence for this position, Dehé and Wichmann (2010: 65) remark that this view is supported by the historical origins of that as a demonstrative pronoun (see the discussion of Brinton 1996 and Fischer 2007 earlier). ^[4]

In this study, we adopt the matrix clause hypothesis insofar as we aim to test Thompson and Mulac’s grammaticalization hypothesis that there is a tendency across time for the zero complementizer to be preferred over the complementizer that, i.e., that the verbs under investigation in this study (think, suppose, and believe) have tended toward higher frequencies of the zero complementizer as conditioned by the factors presented in Section 3. Ascertaining the main effects of these conditioning factors, we determine which ones are good predictors of the zero form. The present study is innovative in approaching the that/zero alternation from both a quantitative and a diachronic point of view. While Tagliamonte and Smith (2005) and Torres Cacoullos and Walker (2009) have performed multifactorial analyses of the synchronic conditioning of that and zero complementation, the current paper adds a diachronic dimension along with a parallel analysis of diachronic spoken and written data sets and investigates, by means of a stepwise regression analysis, whether the zero form is on the increase and how time affects the predictive power of the factors. In addition to interactions with time, this study seeks to lay bare any other significant interactions between factors, notably mode (i.e., spoken vs written data), and to identify any resulting similarities and/or differences between the three verbs of cognition.

2.2 A concise history of the that/zero alternation

There is general agreement on the historical development of the complementizer that from an Old English neuter demonstrative pronoun (see, e.g., Mitchell 1985), but the question which of the two complementation patterns, that or zero, is older is strictly speaking impossible to answer as both the that and the zero complementizer occur in the earliest extant texts (Rissanen 1991). ^[5] This renders the notion of “that-deletion” or “omission” somewhat problematic. On the other hand, it should be observed that in Old English and throughout most of the Middle English period, occurrences of zero are scant. In Warner’s (1982) study of the Wycliffe Sermons, for example, that is used 98% of the time. It is not until the Late Middle English period that the zero complementizer gradually takes off (Rissanen 1991; Palander-Collin 1999), a trend that continues in EModE. Rissanen (1991) notes a steady increase between the fourteenth and the seventeenth centuries, but the most dramatic rise in the zero complementizer can be observed in the second half of the sixteenth century and in the early seventeenth century, when its frequency jumps from 40% to 60%. In addition, Rissanen (1991) shows that the zero form is more common in speech-like genres (i.e., trials, comedies, fiction, and sermons) and that its increase is more pronounced with think and know than with say and tell. Finegan and Biber (1985), too, find that the zero complementizer is more frequent in the more colloquial genre of the personal letter than in the formal genres of medical writing and sermons. ^[6] In the eighteenth century, we witness a temporary drop in zero use. Both Rissanen (1991) and Torres Cacoullos and Walker (2009) attribute this change to the prevalence of prescriptivism, which advocated the use of that out of a concern with clarity.

2.3 Conditioning factors in the literature

Jespersen puts the variability between that and zero down to nothing more than “momentary fancy” (1954: 38, cited in Tagliamonte and Smith 2005: 290); as will be seen, this is a claim that several scholars have tried to refute through an examination of a wide range of conditioning factors. ^[7] Some of these factors are of a language-external nature; many are language-internal.

Many previous studies have tried to account for that/zero variability from the point of view of register variation (Quirk et al. 1985: 953; Huddleston and Pullum 2002: 317; see Rohdenburg 1996 for more references); that tends to be regarded as the more formal option while zero is associated with informal registers (see Kaltenböck 2006: 373–374 for reference). For example, Kearns (2007b) observes some significant differences across varieties in newspaper prose and attributes these to different degrees of sensitivity to some of the conditioning factors discussed further down in this section.

There is also a wide range of language-internal factors. One semantic factor is discussed in Dor (2005), who notes that the semantic notion of the “truth claim” is crucial to the that/zero alternation, in that that clauses denote “propositions” while zero clauses denote “asserted propositions.” Also, particular semantic classes of verbs, notably “epistemic verbs” (Thompson et al. 1991a) or “propositional attitude predicates” (Noonan 1985; Quirk et al. 1985), turn out to have a stronger preference for zero complementation than other complement-taking verbs, such as utterance or knowledge predicates (Thompson et al. 1991a; Tagliamonte and Smith 2005; Torres Cacoullos and Walker 2009).

Importantly, various studies have shown certain high-frequency subject–verb collocations to be strongly associated with zero use (among these are “epistemic verbs” mentioned earlier). Torres Cacoullos and Walker (2009: 32) therefore hypothesize that the conditioning factors for complementizer choice should be different for these highly frequent “discourse formulas” (viz. I think, I guess, I remember, I find, I’m sure, I wish, and I hope) than for the (relatively more) productive complement-taking construction, and indeed they find a number of differences in terms of significance and effect size.

Finally, a wide array of language-internal, structural factors operating on the selection of zero or that have been proposed in previous studies, some of which employ statistical methods, of diverse levels of refinement, to ascertain the import of these factors. In the following three sections, the structural conditioning factors favoring the use of zero will be discussed on the basis of the literature. The factors have been divided into three groups depending on whether they concern matrix clause features, complement clause features, or the relationship between the two. At the end of each section, a table provides a summary of the factors discussed. For each factor, we indicate whether previous studies have or have not statistically tested the factor’s predictive power, and if so, whether it came out as significant or not.

4.1 Interactions with verb

First, we gauge that effects are verb specific (as these effects are aggregated over all time periods, we can call them “panchronic”). The significant factors are presented in Figures 7–11.

Figure 7:

Verb: matrix clause person.

Figure 8:

Verb: matrix clause number.

Figure 9:

Verb: matrix clause tense.

Figure 10:

Verb: absence of intervening elements.

Figure 11:

Verb: harmony of polarity between the matrix and complement clause.

In our first interaction plot analyzing the interaction between verb type and matrix clause person, we see that for think and suppose, the order from best to poorest predictor of the zero form is first, second, and third persons. For think, the difference between the three levels is minimal. Suppose has similar values to think for first and second persons, but a much lower rate of zero with third person. The rate of zero use for both first and second persons is lower overall for believe in comparison to think and suppose. Third-person suppose has approximately the same value as third-person believe. One possible explanation why third-person suppose leads to a much lower probability of that may be that in the third person singular simple present, suppose is trisyllabic, which gives it more phonetic weight. This extra syllable sets supposes apart from all other finite forms of suppose, and may lead to more that-use. This tentative explanation is consistent with the finding that when additional material is added to the matrix clause, that becomes more likely; the third-person ending can be regarded as extra material adding more weight to the matrix clause and making it sound more marked than the default disyllabic verb form.

Although believe behaves differently from think and suppose in having a higher zero rate in the third person than in the second person, this difference is not significant.

In Figure 8, we see that the singular form (sg) more strongly predicts the zero form across all three verbs. The predictive power of singular matrix clause subjects is stronger for think and suppose than it is for believe. The zero rates for both singular and plural (pl) believe are lower than those of the other two verbs.

An analysis of tense across all three verbs indicates that past and present tenses do not differ significantly as predictors of the zero form. The effects of verb forms other than present, past, or future, i.e., non-finite forms, or the use of auxiliaries (n/a) are lower for the verb believe than for the verbs think and suppose; in fact, all three verbs can be seen to differ significantly with regard to the n/a form. The future tense, then, is unreliable for all three verbs, as indicated by the large confidence intervals (cf. the dotted lines). Finally, this plot shows that for each of its tense forms, the verb believe is less predictive of the zero form than think and suppose, i.e., the believe values are lower than those of the others verbs.

In Figure 10, we see that absence of intervening elements between matrix and complement is a very strong predictor of the zero form, it is significant across all three verbs and that its effect is similar for think and suppose. However, believe is most sensitive to the presence or absence of intervening material. When intervening material is present between a matrix clause with believe and the ensuing complement clause, the presence of the explicit that form is more likely than the zero form; the plot shows that the zero rate for believe with intervening material is below 0.5.

The fifth and final interaction with verb type is harmony of polarity between the matrix and the complement clause. The results show that think and suppose slightly favor the zero form in disharmonious patterns and this effect was even more pronounced with believe, which has a lower zero rate in harmonious patterns than think and suppose but comparable values when there is no harmony of polarity.

We will now compare the extent to which these factors predict the use of the zero form in the spoken and written modes.

4.2 Interactions with mode

The interactions of the factor “mode” (i.e., spoken vs written mode) with other factors (see Table 12 in Appendix) are also panchronic, i.e., all periods are conflated. In this section, we will see that mode plays a more important role in the zero/that alternation, in that it has an impact on the strength of the other factors: some factors may be better predictors of the zero form in one mode as opposed to the other.

Figure 12 allows us to compare the conditioning effect of intervening elements between matrix clause and complement clause in the spoken and written modes. In both panels of Figure 12, we observe a dramatic difference in complementizer use between presence and absence of intervening material. A notable difference, however, resides in the extent to which the presence of intervening material in the written mode predicts the zero form. When there is intervening material in the written mode, we are much less likely to get the zero form than in the spoken mode, so much so that the explicit complementizer that in fact becomes more likely; the zero rate drops to below 0.4. It may be that writers are more led by the complexity principle than speakers and feel the need to insert that to make clause boundaries clearer when intervening material risks impairing clarity.

Figure 12:

Mode: absence of intervening elements.

In Figure 13, we examine the effect of matrix clause person in the two modes. In the spoken mode, first-person subjects (P1) predict more zero use than second- and third-person forms (P2 and P3). In the written mode, the difference between first- and second-person subjects is not significant, but the difference between these values and the much lower zero rate with third-person subjects is significant. Also, compared to the spoken mode, third-person subjects in written data are much less likely to be used with a zero complementizer.

Figure 13:

Mode: matrix clause person.

The analysis of tense relative to mode follows the pattern of tense and its interaction with verb; once again, in both the spoken and written data, past and present tense forms are not significantly different from one another. The n/a forms condition for zero use, but much less so in the written mode than in the spoken mode. Due to the sparseness of future forms and resultant large confidence intervals, we cannot make any claims about the effect of the future on zero use in spoken versus written data (Figure 14).

Figure 14:

Mode: matrix clause tense.

Figure 15 presents the results for the effect of elements between the matrix clause subject and verb in the spoken and written modes, respectively. Seeing that this interaction only approaches significance, we can say that the absence of matrix-internal elements is a fair predictive factor for both spoken and written data. The steepness of the plot lines shows that the difference in predictive power between presence and absence of matrix internal elements is comparable for both modes, but in the written mode, the zero form is used less overall, as the lower points for both levels indicate.

Figure 15:

Mode: absence of matrix-internal elements.

An analysis of the effect of the coreferentiality of person between the matrix and complement clause subjects reveals an interesting difference with regard to mode. Coreferentiality of person leads to higher levels of zero in the spoken data. In the written data, when subjects are coreferential, zero is slightly less likely than when they are not coreferential. However, the difference is minimal and the confidence interval of the non-coreferential written data point is rather large (Figure 16).

Figure 16:

Mode: coreferentiality of person between the matrix and complement clause subjects.

The final factor that we will examine in this section is the predictive effect of the length of the complement clause subject on the zero form relative to mode. The plot in Figure 17 shows that within the spoken data the following cline exists: it > pro > np-short > np-long (which is in line with the main effect observed for this factor). A comparison between the two modes shows that short and long NPs tend more strongly toward that in the written mode than in the spoken mode. Overall, length of the complement clause subject has a stronger effect on written data than on spoken data. Again, the complexity principle, i.e., the need to mark off clause boundaries, may motivate writers’ choice of the that complementizer as opposed to the zero form. In addition, the concern with clarity fostered by standardization and prescriptivism may also play a role.

Figure 17:

Mode: length of the complement clause subject.

We now will turn to the final stage of our analysis and look at the effect of the structural factors across the eight time periods. In the following sections, we will discuss the interactions with period that came out as significant.

4.3 Interactions with period

The interaction effects with period are significant for the following factors: verb, length of the complement clause subject, matrix clause person, and harmony of polarity. This final step in the analysis offers a diachronic perspective; it shows whether the import of a given factor becomes stronger or weaker over time.

Figure 18 shows the diachronic development of the zero form for each of the three verbs. It reveals that think, suppose, and believe start out at roughly the same zero rate and that all three verbs indeed exhibit a loss of zero. However, there are notable differences as to the extent of this downward trend. The decrease is minimal for suppose and even more so for think. Believe, by contrast, is characterized by a dramatic drop in zero use, especially from 1780 up to the present day. In the most recent data set (period 8), it even plummets below 50%, i.e., in most recent times, believe has come to prefer that over zero. ^[18]

Figure 18:

Period: verb.

An analysis of the effect of the length of the complement clause subject over time shows that all types of subject used to be better predictors of the zero form than they are now, but that NPs lose more of their zero forms than pronouns (it and other pronouns). Thus, over time, short and long NPs tend more strongly toward that than it and other pronouns (Figure 19).

Figure 19:

Period: length of the complement clause subject.

The loss of the zero form over time is also seen in Figure 20, which represents the diachronic effect of person. It can be observed that zero rates with both first and third persons decline gradually over time, with third person dropping off more dramatically than first person. By contrast, no such decrease in zero can be observed in the second-person data.

Figure 20:

Period: matrix clause person.

The final significant effect over time is the interaction between harmony of polarity between matrix and complement clause and period. Figure 21 shows that when there is harmony of polarity, there is a distinct tendency toward more that over time; in other words, harmony of polarity used to be a stronger predictor of the zero form than it is now. This trend is absent from the non-harmonious data; here, the level of zero use remains more or less stable.

Figure 21:

Period: harmony of polarity between the matrix and complement clause.