3 Case notes: Independent factors
3.3 Level of formality
Besides the intralinguistic factors mentioned so far, an important extra-linguistic factor restricting preposition placement mentioned in virtually every publication on preposition-stranding is the level of formality: prep-osition-stranding is usually associated with speech and informal written contexts, whereas pied-piping is preferred in formal writing (see e.g. Biber et al. 1999: 107; Leech 1996: 375). This stylistic distribution of preposition-stranding has its foundations in the prescriptivist tradition of the eighteenth century:
The Preposition is often separated from the Relative which it governs, and joined to the verb at the end of the Sentence … as, ‘Horace is an author, whom I am much delighted with’ … This is an idiom which our language is strongly inclined to; it prevails in common conversation, and suits very well with the familiar style of writing; but the placing of the Preposition before the Relative is more graceful, as well as more perspicuous; and agrees much better with the solemn and elevated style. (Robert Lowth, A Short Introduction to English Grammar, 1762; cited in Aitchison 1991: 11)
The prescriptive grammarians’ preference of preposition pied-piping over stranding resulted from their admiration for Latin, which ‘was widely regarded as the most perfect of languages’ (Aitchison 1991: 9) and there-fore considered a model for the English language. Now, in Latin preposi-tions cannot strand but instead always have to precede their complements (cf. also the etymological source of preposition: Lat. praeponere ‘put before’;
Chalker and Weiner 1994: 310). Consequently, it is not surprising to find that Lowth calls pied-piping, which follows the Latin model, ‘more grace-ful’, while he would like to restrict the ‘un-Latin’ stranding of prepositions to ‘common conversation’ and ‘the familiar style of writing’. Nevertheless, considering that ‘Lowth’s approach was essentially proscriptive’ (Tieken-Boon van Ostade 2006: 544), the above statement is actually fairly lenient since it does not generally proscribe preposition-stranding. In this context, it is interesting to take a look at other prescriptive grammars from the eight-eenth century.
As Alston (1965) points out, the most popular eighteenth-century gram-mars were Murray (1795), Lowth (1762), Ash (1760) and Fisher (1750) (cf.
also Tieken-Boon van Ostade 2006: 542). Concerning preposition place-ment, Murray’s view is almost a verbatim quote of Lowth’s statement above:
The preposition is often separated from the relative which it governs; as
‘Whom wilt thou give it to?’ instead of, ‘To whom wilt thou give it?’ [‘]He is an author whom I am much delighted with’; ‘The world is too polite to shock authors with a truth, which generally their booksellers are the first that inform them of.’ This is an idiom to which our language is strongly
inclined; it prevails in common conversation, and suits very well with the familiar style in writing: But the placing of the preposition before the relative is more graceful, as well as more perspicuous, and agrees much better with the solemn and elevated style. (Murray 1795: 122; bold emphasis added) The first thing to note is that upon discussing relatives, Murray also includes the interrogative example Whom wilt thou give it to? vs To whom wilt thou give it?. As the quote shows, he furthermore agrees with Lowth as to pied-piping being more ‘solemn and elevated’ but stranding being common in conver-sation. The most notable difference is Murray’s correction of Lowth’s This is an idiom which our language is strongly inclined to to This is an idiom to which our language is strongly inclined. Concerning Lowth’s stranded version Tieken-Boon van Ostade claimed that ‘[i]t is the kind of joke purists enjoy making, such as “a preposition is a word you cannot end a sentence with,”
or “this is something up with which I will not put.”’ (2006: 551). While that may have been the case, the joke seems to have been lost on Murray.
Moving on to the other two important grammarians of the time, it appears as if Ash (1760) did not comment on preposition placement, while Fisher (1750) explicitly argues that
The Preposition is frequently transposed as, who do you dine with? For, with whom do you dine? [TH: italics sic!] what Place do you come from? For, from what Place do you come?
Q. May Words in Sentences be placed in what Order we please?
A. No; but we must in this, as well as in all other parts of Grammar, follow the Use of the best Speakers and Writers.
☞ The clearest and best Writers in Prose have the fewest Transpositions in their Discourses; and in Poetry, they are never used but when the nature and Harmony of the Verse require it (Fisher 1750: 123)
Thus, Fisher also acknowledges the fact that prepositions are often stranded (note how she considers this a case of ‘transposition’, i.e. for her pied-piping is the underlying normal structure). However, since the best speakers and writers predominantly favour pied-piping, people should try to follow their model (though Fisher also concedes that in poetry prepositions might need to be stranded because of metrical constraints). Even more striking is the fact that despite her preference for pied-piping she herself occasionally strands prepositions: ‘What do you learn Grammar for? … What does Grammar treat of?’ (Fisher 1750: 1). Unlike Lowth’s stranded example above, which might be interpreted as a linguistic joke, these two sentences appear on page 1 of Fisher’s grammar, long before she covers preposition placement. Yet, both examples are interrogative clauses. Thus functional constraints (cf. 3.1.2) might have led to the choice of stranding, leaving the wh-word at the front of the clause to signal interrogative force. (In addition to this, the what … for sequence in the first example might already have acquired its status as a dis-continuous lexical item meaning ‘why’; cf. 3.2.3.2.)
Having surveyed the most important eighteenth-century grammars, it needs to be emphasized that these acknowledge stranding as a grammat-ical structure of spoken, informal English. Rather surprisingly, however, the prescriptive tradition emerging from these grammars then endorsed the slogan ‘it is incorrect to end a sentence with a preposition’ (Pullum and Huddleston 2002: 627). Pullum and Huddleston correctly point out that
‘[s]tranded prepositions often, but by no means always occur at the end of the sentence’ (2002: 627). Nevertheless, lacking the metaphorical label ‘stranded preposition’, prescriptivists came up with the above hard-and-fast rule. The rule itself is attributed to Dryden, who criticized Ben Jonson’s phrase The bodies that those Souls were frighted from (Catiline, 1611): ‘The Preposition in the end of the sentence; a common fault with him and which I have but lately observ’d in my own writings’ (John Dryden, Defence of the Epilogue 1672, ll. 55–76; taken from Görlach 1994: 202).
Usually it is implied that the simpler rule, i.e. the strict ban on stranded prepositions, has dominated the prescriptive tradition (cf. Sundby, Bjørge and Haugland 1991: 426–8; Pullum and Huddleston 2002: 627; Yáñez-Bouza 2006):
Instead of being dismissed as unsupported foolishness, the unwarranted rule against stranding was repeated in prestigious grammars towards the end of the eighteenth century, and from the nineteenth century on it was widely taught in schools. The result is that older people with traditional educations and outlooks still tend to believe that stranding is some kind of mistake.
(Pullum and Huddleston 2002: 627)
Yet, in addition to the fact that the most influential eighteenth-century grammarians were very well aware of the influence of the level of formality, Görlach also stresses the fact that there was a tendency in the nineteenth century to sever ‘the ties that bound English to Latin’ (1999: 76). The most prominent prescriptivist of the early twentieth century, Henry Fowler, even claimed that ‘[t]hose who lay down the universal principle that final preposi-tions are “inelegant” are unconsciously trying to deprive the English lan-guage of a valuable idiomatic resource, which has been used freely by all our greatest writers except those whose instinct for English idioms has been over powered by notions of correctness derived from Latin standards’ (Fowler 1926: 458). Modern usage guides also reject the strict ban on stranded prep-ositions, calling the prescriptive rule ‘one of the most persistent myths about prepositions in English’ (Burchfield 1996: 617), which ‘should be disregarded’
(Weiner 1983: 166). Instead, as pointed out above, the choice between prep-osition-stranding and pied-piping with wh-pronouns is seen as a matter of style and obligatory stranding contexts such as passives are pointed out (e.g.
Leech 1996: 375; Weiner 1983: 166).
Whenever there is a choice with respect to preposition placement, present-day usage guides thus emphasize that pied-piping is more or less restricted
to formal registers. This has led some researchers to propose that in present-day English stranding is in fact the unmarked option, and that preposition pied-piping is only acquired, along with other formal writing skills, through formal education. Working on the acquisition of relative clauses, McDaniel, McKee and Bernstein (1998: 309), for example, claim that ‘preposition pied-piping is not a natural option in English, but rather a prescriptive artifact probably picked up during schooling’.
As will be recalled, in section 3.1.2 it was argued that from a processing perspective relative clauses are probably the context which favours pied-pip-ing most. Thus, for present-day English, McDaniel, McKee and Bernstein’s study is of particular interest: could it really be that even in wh-relative clauses (for which Van den Eynden claims that ‘stranding is not really an option’ (1996: 444)), pied-piping is only acquired during formal education?
Investigating the relative clause production of 115 (American) chil-dren from the age of 3 to 11 with an elicitation test,20 McDaniel, McKee and Bernstein found that even the youngest subjects already produced that+Pstranded constructions like that one that Baby Bob is jumping over, but that none of their subjects employed the alternative Ppiped+wh-pronoun option (indeed, their subjects only produced that-introduced relative clauses; see McDaniel, McKee and Bernstein 1998: 315). In a second set of experiments, the researchers then decided to test the subject’s grammaticality judgements with regard to wh-relative clauses. For this, an experimenter would perform a short scene with toys, and then ask the subject whether a sentence related to the observed situation was grammatical; e.g. ‘Is it right or wrong to say,
“This is the robber who Dorothy talked to”? … “This is the robber to whom Dorothy talked”?’ (McDaniel, McKee and Bernstein 1998: 318). As their results showed, virtually all subjects considered the stranded variant gram-matical. In contrast to this, the pied-piped alternative enjoyed a relatively low grammatical status: only 3% of the youngest subjects (3–5 years old) and 6% of their ‘Middle’ group (6–8 years old) classified the pied-piped relative clauses as grammatical. Even within the oldest subject group (9–11 years) only 54% of the children considered the pied-piped alternative grammat-ical. On the other hand, 90% of an adult control group identified the pied-piped relative clauses as grammatical (cf. McDaniel, McKee and Bernstein 1998: 323).
All in all, McDaniel, McKee and Bernstein’s study seems to imply that stranding, and not – as expected from cross-linguistic data (cf. section
20 Basically, a first experimenter performed a story with a set of toys which the sub-ject watched together with a second experimenter. Once the story was over, the second experimenter covered her eyes, and the first experimenter pointed to one of two identical toys, which could only be distinguished from each other by what had happened to them in the story. Then the child was asked to instruct the second experimenter to pick up the toy which the first experimenter had pointed to. Thus, children were forced to produce relative clauses like the girl that the giraffe is sitting on (McDaniel, McKee and Bernstein 1998: 313f.).
3.1.4) – pied-piping is the natural choice with wh-pronouns in present-day English. In the light of these results, it obviously appears as if pied-piping would not be part of the grammar originally acquired by children, but a pre-scriptive rule ‘learnt’ later via formal education.
However, taking a closer look at McDaniel, McKee and Bernstein’s study reveals that it might not be possible to draw such far-reaching conclusions from their experiment. First of all, even though only a small number of their subjects considered the pied-piped variant grammatical, it follows that the grammar of this small group of children (e.g. 3% of the 3–5 year olds) already includes this structure.
Secondly, in general, the relatively low percentage of subjects accepting pied-piping might be due to the type of sentences used for the grammatical-ity judgements: the material only included complement PPs (talk to) but no sentential adjuncts. In fact, at no point in their article do McDaniel, McKee and Bernstein discuss the potentially different behaviour of different PP types. However, within their experiment it would have easily been possible to test sentences like This is the house in which Dorothy met the robber vs This is the house which Dorothy met the robber in, which might have produced a different result.
Finally, it might be that both pied-piping and the wh-relative pronouns are strongly associated with formal contexts (cf. section 3.1.2.2). Then it is to be expected that younger children simply have not encountered enough formal situations in which these phenomena occur frequently. While for-mal instruction in school is clearly such a situation, this need not be the only context. The increase in acceptability of pied-piping with increasing age might also have to do with the fact that older children have been to more church services or other formal occasions. It would definitely be surprising if explicit formal education on preposition placement should be the most important factor.
The above qualifications should be particularly true for Britain, where grammar teaching has undergone considerable changes over the past fifty years: up to the 1960s there was indeed a focus on explicit grammar teaching which was essentially prescriptive and assumed that ‘there is a correct stand-ard form of the language normally only found in writing’ (Philp 2001: 724).
From the early 1960s onwards, however, the focus has shifted towards more descriptive and usage-based grammatical descriptions (cf. Philp 2001: 727).
The Cox Report (Department of Education and Science 1989), for example, emphasizes that the grammatical description of English taught in schools should be ‘a form of grammar which can describe language in use [as well as be] able to describe the considerable differences between spoken and written English’ (Department of Education and Science 1989: 4.28). Thus, gener-ally speaking, British citizens who started their schooling in the 1960s or later (i.e. speakers of ages about 50 and under) should not have received the same kind of prescriptive, explicit grammar teaching as their predecessors.
Consequently, for such speakers it seems unreasonable to assume that their us of pied-piping in wh-relative clauses can simply be accounted for by a taught prescriptive ‘never end a sentence with a preposition’ rule.
Nevertheless, if despite these objections, it actually turned out that chil-dren strongly favour stranding over pied-piping with wh-relative pronouns with all PP types, then it would obviously be interesting to see in what way formal education influences these preferences. Since modern usage guides tend to limit pied-piping to formal written registers, it should prove insight-ful to compare the results of studies on preposition stranding with wh-pro-nouns in written adult language, with studies examining spoken corpora.
Bergh and Seppänen (2000: 306f.), for example, present an overview of five corpus studies21 on preposition stranding in wh-relative clauses, presented in Table 3.5.
Upon comparing the two subtables, the most striking finding concerns the fact that preposition pied-piping is always preferred over stranding,
21 The data analysed by these authors comprises the following corpora: Johansson and Geisler’s study (1998) has already been mentioned in section 3.2.1. Bengtsson (1996) is an unpublished paper which investigates an 85,000-word corpus consisting of writings of G. B. Shaw; while Trotta examined relative clauses in a corpus of American written English (BROWN). Van Eynden’s corpus consisted of texts from British newspapers (tabloids as well as quality newspapers), and Quirk analysed tape-recordings of English adults. All these studies – with the exception of Trotta, who looked at American Standard English – used corpora representative of Standard British English.
Table 3.5 PP types used for stranding and pied-piping with wh-relativizers in studies of written and spoken PdE (adapted from Bergh and Seppänen 2000: 307;
supplemented by additional data from Johansson and Geisler 1998: 70)
written corpora stranding % pied-piping %
Bengtsson (1996) 3 5 62 95
Van den Eynden (1996) 6 3 179a 97
Johansson and Geisler (1998) [LOB] 27 3 1053 97
Trotta (2000) [BROWN] 12 1 1054 99
total (written) 48 2 2348 98
spoken corpora stranding % pied-piping %
Quirk (1957) 18 17 86 83
Johansson and Geisler (1998) [LLC] 50 21 190 79
Johansson and Geisler (1998) [spoken BIRM] 129 14 799 86
Johansson and Geisler (1998) [spoken BNC] 65 31 147 69
total (spoken) 262 18 1222 82
a In Bergh and Seppänen’s table, Van den Eynden is claimed to have found 166 piped tokens.
The actual figure for piped WH-relative clauses, however, was found to be 179 (cf. Van Eynden’s 1996: 442).
regardless of the particular mode: on average, 98 per cent of the prepositions are pied-piped in the written corpora and 82 per cent in the spoken cor-pora. If stranding were indeed the natural choice with wh-relative pronouns, then, in the light of Table 3.5, the influence of formal education would be immense: it would not only affect a speaker’s formal written register, but would apparently install pied-piping as the default choice for both spoken and written language. Again, it seems doubtful that explicit prescriptive teaching should be responsible for such an across-the-board effect. Instead, it makes more sense to assume that pied-piping with wh-relative clauses is preferred due to processing constraints (cf. section 3.1.3). In addition to that, as pointed out in section 3.1.2.2, the wh-relative pronouns themselves are felt to be more formal than the that- and Ø -relativizers. Once speakers encounter more formal occasions, they will also come across more wh-pro-nouns, which favour pied-piping. This mutual co-occurrence of wh-relative pronouns and pied-piping thus explains the preposition placement prefer-ence in spoken and written registers. If pied-piping with wh-relative clauses really was nothing but a reflex of the prescriptive ban on stranding allegedly taught at schools, one would expect that in the written mode speakers have enough planning time for this rule to affect their language. Since the spoken channel leaves less time for conscious rules to interfere however, the strong preference for pied-piping in this mode (as indicated by the data in Table 3.5), does not lend itself to such an explanation. In the words of Halliday (1988: 38): ‘Our ability to use [spontaneous spoken] language depends crit-ically on our not being conscious of doing so – which is the truth that every language learner has to discover, and the contradiction from which every language learner has to escape.’
Besides, as Weiß notes, ‘modern standard languages … are not only learned by instruction (e.g., in schools), but to a large extent acquired as L1 too’ (2001: 94). Standard British English, for example, is not only acquired via formal education, it is also the sociolect associated with the middle and upper classes (cf. Leisi 1985: 191f.). As such, considering that adult speak-ers of Standard British English employ the ‘Ppied-piped+wh-pronoun’ strategy quite frequently in spoken language (see Table 3.5), one might even expect social factors, such as class, to have a considerable effect on the acquisition of pied-piping.
Turning back to the data in Table 3.5, a closer look at the actual data of the individual studies reveals that the underlying distribution is again a lot
Turning back to the data in Table 3.5, a closer look at the actual data of the individual studies reveals that the underlying distribution is again a lot