2 Basic valency, basic valency orientation and valency changes

Nichols et al. (Reference Nichols, Peterson and Barnes2004) propose basic valency orientation as a new typological parameter. The basic valency of a language is its preference for intransitive or transitive verbs as basic in the sense of underived lexical units. In parallel, the basic valency orientation of a language is ‘the preferred or predominant or most common form of lexicalization or valence-related derivation’ (Nichols et al. Reference Nichols, Peterson and Barnes2004: 150–1). Orientation is thus embodied in a language’s derivational processes.

Nichols et al. distinguish transitivising and detransitivising languages, and two more types, neutral and indeterminate languages (2004: 161, 163). Transitivising languages are those that favour valency-increasing derivation. As an example of valency-increasing derivation of the causative type consider the following Vedic and PDE examples:

For the present purposes, morphological and syntactic coding are not distinguished; it is assumed that make laugh is an established expression in English, and hence illustrative of a grammatical operation.

A brief note on terminology is necessary. Haspelmath et al. (Reference Haspelmath, Calude, Spagnol, Narrog and Bamyaci2014: 590 and fn. 4) detect confusion between formal and semantic terms in some studies on causatives. They propose a more rigorous terminology, which will be followed in this article. They use the terms causal/noncausal to refer to the lexical semantics of the verbs and reserve causative/anticausative for the morphosyntactic coding. In the English pair above, laugh is noncausal and make laugh is the corresponding causal expression. Since the noncausal pair is basic, i.e. non-derived, the coding is causative.

Detransitivising languages, on the other hand, are those that favour valency-decreasing operations. Anticausative coding is frequent in these languages. Consider the following German verb pair:

Here, ‘sich erinnern’ shows anticausative coding by means of the reflexive pronoun sich.

There are other formal correspondences between causal and noncausal verb pairs, one of which is of particular interest for this study. The intransitive and transitive members of the verb pair can have the same form, with no extra coding, as in (1) above, repeated here for the sake of clarity:

I refer to this type of verb as labile in line with a long tradition recently followed in Haspelmath et al. (Reference Haspelmath, Calude, Spagnol, Narrog and Bamyaci2014: esp. 590–1). Nichols et al. classify languages favouring this coding as indeterminate (Reference Nichols, Peterson and Barnes2004: 161).

The valency of a verb or a group of verbs can change over time, affecting the basic valency orientation of the language in question. For the purposes of this article, the process of a verb becoming labile is particularly relevant. A case in point is example (5). The English verb burn is the reflex of OE bærnan (weak class I),Footnote ⁷ which was a ja-causative originally in Proto-Germanic and therefore used only transitively, as in the following OE sentence:

However, it was increasingly used in an intransitive sense, too, as in the following late OE example:

OE bærnan has become labile through detransitivisation. The opposite process, transitivisation of an intransitive verb, is of course also possible. The term labilisation will be used in this article to refer to the process by which a verb with originally only one valency value (i.e. either transitive or intransitive) acquires a second value and becomes both transitive and intransitive. The direction of the change, although deserving of attention, is of little relevance in the scope of this study.

Note that coinciding with Nichols et al. (Reference Nichols, Peterson and Barnes2004: 151), causatives are not investigated here for their own sakes, but as representatives of the major valency-changing operation in early Germanic languages (see section 5).

3 Previous studies on the basic valency of Old English

Previous studies on the basic valency orientation of OE have made use of one or several of the possible strategies to measure it: (1) computation of verbs according to valency; (2) the assessment of prevalent derivational valency-changing operations and (3) diagnostic questionnaires of verb glosses, a common practice in typological studies.Footnote ⁹ This section will briefly describe the main findings of these studies and identify the strategies they use, as one of the goals of this article is to examine their arguments critically (see sections 4 and 6 below).

Work on the basic valency of Old English pre-dates the term valency itself. In his 1952 book Reflexive und intransitive Verba im älterem Westgermanischen, Lars Hermodsson made a solid contribution to establishing the basic valency of OE. He compiled lists of verbs that can be used both as intransitives and transitives in OE (Verba mit Doppelfunktion), as well as intransitive and reflexive verbs. Hermodsson’s work is also helpful as a typological tool, as OE is compared with other (West Germanic) languages. He observed that the pronounced development of labile verbs in PDE is in fact inherited and can be traced back to Ingvaeonic (comprising Old Frisian, Old English, Old Saxon and Middle Dutch).

Although he comments here on Present-day rather than Old English, it is obvious that he assumes that the latter, as an Ingvaeonic language, shares the trait. Elsewhere, however, he observes that intransitive verbs are more frequent in North-West Germanic (including OE) than in High German (1952: 185).

Visser, in An Historical Syntax of the English Language (Reference Visser1963), remarks that whereas in OE intransitives are much more numerous than labile verbs, which he calls ‘double-functioned or amphibious’ (Visser Reference Visser1963: 97), the opposite is true in PDE, where the labile construction is much more frequent (Visser Reference Visser1963: 99).

Hermodsson and Visser differ in the basic valency they assume for OE: Hermodsson points out the strong spread of labile verbs, whereas Visser emphasises the abundance of intransitives (Reference Visser1963: 99).Footnote ¹¹

McMillion (Reference McMillion2006) applies to OE Haspelmath’s (Reference Haspelmath1993) questionnaire of 31 verb meanings that can take part in an inchoative/causative alternation, such as ‘open (intr.) / (tr.)’ or ‘rise’ / ‘raise’. He concludes that only ten were labile in OE, which is about the same number as in Present-day Swedish and German, and that:

Thus, according to McMillion, the basic valency of OE, like that of Swedish or German, is not particularly labile, and PDE’s labile orientation is a later development which cannot have had its origin in OE. In this, McMillion diverges from both Hermodsson (Reference Hermodsson1952) and the next study to be reviewed, Ottosson (Reference Ottosson2013).

Ottosson (Reference Ottosson2013) charts the development of various valency-coding mechanisms in early Germanic languages. He rightly asserts that transitivising mechanisms, i.e. ja-causatives, are prevalent over any detransitivising ones, such as na-verbsFootnote ¹² or what he calls ‘reflexive Middles’, in Proto-Germanic (Ottosson Reference Ottosson2013: 347, 356, 378). This makes Germanic a transitivising language. He assumes that Old English makes extensive use of labile verbs, and considers this a North-West Germanic (i.e. Ingvaeonic) tendency, linked to the lack of a reflexive pronoun proper in these languages (Ottosson Reference Ottosson2013: 332, 356; this has also been suggested by Poppe Reference Poppe2009: 260). Note that Ottosson (Reference Ottosson2013: 374–5) considers that the causative formation is still moderately productive in OE.

Van Gelderen’s comprehensive study (Reference Van Gelderen2011) follows the changes in valency orientation throughout the history of English, including both agent-changing and object-changing strategies. She advances her views on the basic valency of OE cautiously. Having applied Nichols et al.’s questionnaire of verb meanings alternating in valency to Old English, she rightly warns that the absence of native speakers of the language makes it difficult to identify the most common gloss (Reference Van Gelderen2011: 115). She interprets the results as suggesting that ‘there is no basic valency in Old English’ and that OE was ‘not as intransitive-oriented as sometimes argued’ (Reference Van Gelderen2011: 117).

Van Gelderen also presents as evidence Hermodsson and Visser’s lists of intransitive, transitive and labile verbs and concludes that, as Visser had pointed out, OE had many intransitive verbs that became transitive (Reference Van Gelderen2011: 118, 119, 122), and that ‘even though we have to be cautious with all the verbs in [Visser’s and Hermodsson’s lists of labile verbs], labile verbs already seem common in Old English but not as common as in Modern English’ (Reference Van Gelderen2011: 122). In this she coincides with Hermodsson. Like Ottosson, van Gelderen detects some productivity in the causative ja-formation in OE (Reference Van Gelderen2011: 123–4). This view is disputed in section 6.1 below.

Summing up section 3, the various methods used to assess the basic valency of OE yield different results. Diagnostic tables do not seem to lead to any clear conclusions, other than that labile coding is not conspicuous. Computation of verbs reveals a large proportion of intransitive verbs, with a small but notable number of labiles. Finally, with respect to the main valency-changing mechanism in OE, it is suggested that the causative formation retains some productivity, which is consistent with the high proportion of intransitives. Crucially, as will be argued in section 6.1, this would make OE transitivising, like its Germanic ancestor, and not labile, as Hermodsson and van Gelderen conclude.

Scholars differ in their interpretation of the data. Hermodsson points out the spread of labile verbs in OE as compared to other Germanic languages, while acknowledging, like Visser, the large number of intransitive verbs. In the recent literature on valency, Ottosson and van Gelderen, following Hermodsson, trace back the PDE tendency to favour labile coding to OE, while maintaining that the causative construction was still productive in that period. McMillion, however, claims that OE is not especially labile.

The following sections re-examine the effectiveness of two of the procedures for testing the basic valency orientation of OE: computation of verbs and assessment of its main valency-changing operations, in this case, the causative ja-formation. I will argue for the superiority in diagnostic power of the latter and describe how, in both cases, a historical approach based on detailed structural analysis of textual evidence may inform synchronic analysis and thus underpin typological statements.

Diagnostic lists are not analysed further, as they do not seem to lead to any clear results for OE and it is doubtful that they can be applied with any reliability to languages with no native speakers on whom to test the accuracy of the glosses, as van Gelderen remarks (see above).

4 The contribution of etymology in establishing morphological typologies: what bare word lists don’t tell us

This section discusses the computation of verbs as a strategy to assess valency tendencies from a synchronic and diachronic perspective. For the latter, I have established the word-formation pattern of OE verbs classified as labile in previous literature and carried out a comparative-historical analysis of weak verb classes and their valency expression potential in OE and other Germanic languages. The analysis reveals that many of the labile verbs in OE texts might be inherited, so that they do not necessarily demonstrate labilisation tendencies in the period itself.

The main tools used in this section are etymological dictionaries and grammars of the Germanic language (Krahe & Meid Reference Krahe and Meid1969; Seebold Reference Seebold1970; Heidermanns Reference Heidermanns1993; Ringe Reference Ringe2006), OE grammars (Campbell Reference Campbell1997 [1959]; Brunner Reference Brunner1965; Hogg & Fulk Reference Hogg and Fulk2011) and OE dictionaries (Bosworth & Toller Reference Bosworth and Toller1898; Hall & Meritt Reference Hall and Meritt1970; The Dictionary of Old English (DOE); the NerthusV3 database). Doubtful valency values have been confirmed through a process that will be detailed in the following section.

From a synchronic point of view, the lists of labile verbs in Hermodsson (Reference Hermodsson1952: 196–207) and Visser (Reference Visser1963: 99) are reliable. Overall, the valency assigned to each verb is accurate, even by current standards using updated tools.Footnote ¹³

A diachronic analysis reveals that morphophonetic processes in Pre-OE, that is, during the period preceding the first attestations,Footnote ¹⁴ may have influenced the valency of whole verb classes. A large proportion of the labile verbs included in the lists by Hermodsson (Reference Hermodsson1952) and Visser (Reference Visser1963), repeated by McMillion (Reference McMillion2006) and van Gelderen (Reference Van Gelderen2011) as evidence, may be labile because of formal conditioning, as noted in passing by Hermodsson (Reference Hermodsson1952: 62, 195). Many belong to the second weak class,Footnote ¹⁵ which historically contains both transitive and intransitive verbs (see lists in Brunner Reference Brunner1965: 330), as opposed to the first weak class, with only transitive-causative verbs. By way of illustration, consider the following OE weak verbs, class II:

Some of the ambivalence of weak class II verbs is inherited and applies in principle to all Germanic languages. However, the situation was aggravated in OE as the vast majority of the ē-verbs, which constitute the third weak class in Germanic, as it is traditionally called, shift to the second weak class (ō-verbs) (Campbell Reference Campbell1997 [1959]: 342). The ē-class contains mainly intransitive verbs with stative or inchoative meaning and is especially productive in Old High German (OHG) (Krahe & Meid Reference Krahe and Meid1969: 249; Ringe Reference Ringe2006: 236, 256–7; see example (10) below). In Gothic and in Old Norse (ON), where the ē-class is not productive, derived intransitive verbs could be formed with the na-suffix (Krahe & Meid Reference Krahe and Meid1969: 250, 253; Ringe Reference Ringe2006: 259; and especially Ottosson Reference Ottosson2013passim), as in the following Gothic example (Braune & Heidermanns Reference Braune and Heidermanns2004: 166):

In OE ē- and na-verbs survive only fragmentarily (see further Ottosson Reference Ottosson2013, esp. 377), and the second weak class is a catch-all and hence heterogeneous in all aspects, including valency. Whereas in other Germanic languages deadjectival verbs can distinguish valency depending on whether they are built with the ja-suffix (transitivising) or with the ē- or na-suffixes (intransitivising), this distinction cannot be expressed in OE. In this language, next to a transitive-causative ja-verb there is often a labile class II verb (ō-verb), instead of an intransitive ē- or na-verb. The following example may serve to clarify the argument (Germanic forms follow Heidermanns Reference Heidermanns1993: 280–1):

In example (10), whereas the intransitive/transitive opposition is expressed by two different verbs in OHG (ē- and ja- respectively), OE lacks an intransitive-only form, and the ō-verb heardian expresses both valencies.Footnote ¹⁶ This verb is rightly listed as having a ‘double function’ (‘Doppelfunktion’) by Hermodsson (Reference Hermodsson1952: 198). However, neither heardian nor the other labile class II verbs necessarily testify to tendencies towards the labilisation of single-valency verbs in OE, but to quirks of word-formation type in the pre-history of OE verbal classes. Crucially, the numerous labile verbs of the second class, or ō-verbs, may have been labile from their inception in OE (thus also Ottosson Reference Ottosson2013: 357) rather than the result of a process of transitivisation or detransitivisation in historical OE.

In summary, as highlighted by previous studies, there are a substantial number of labile verbs in OE. However, this does not necessarily imply that OE favours labile coding, since prior morphophonetic processes may have influenced the valency of a whole verb class. Many of the labile verbs in Old English belong to the second weak class, which contained intransitive and transitive verbs already in common Germanic, and which moreover had been affected by class syncretism in pre-OE. Their being labile may well be related to a morphophonetic development that took place before historical OE, rather than the result of a process of labilisation (detransitivation or transitivisation) within OE. They do not point to a tendency to increase labile coding in OE.Footnote ¹⁷ Note that I am not referring here to verbs of the first weak class in historical, textually attested OE, which will be addressed below.

5 Main valency-changing operations and the basic valency orientation of Old English

The computation of verbs according to valency has been shown to be a not entirely reliable indicator of the valency orientation of OE. In the remainder of this article another strategy to assess valency tendencies will be put to the test: a language’s main valency-changing operation(s), in this case, the causative formation, as will be justified in the paragraphs below. In this section, causative verb-pairs in OE will be introduced, comprehensive lists provided and the importance of this construction in Germanic will be discussed.

The data for this and the following section are part of an ongoing project on causatives in OE and other early Germanic languages from a historical-comparative perspective, which has given rise to several publications (García García Reference García García2005, Reference García García2012, Reference García García2016). The data have been revised and updated, with particular attention paid to valency. Bammesberger (Reference Bammesberger1965) and García García (Reference García García2012) provide the initial corpus, namely OE deverbal ja-verbs. The probable causative verbs among them were identified using the definitions and text examples in the DOE where possible, that is, for entries A to H.Footnote ¹⁸ For these verbs, the information given here is comprehensive. From H onwards, Bosworth & Toller (Reference Bosworth and Toller1898), Hall & Meritt (Reference Hall and Meritt1970), and the NerthusV3 database served as preliminary sources of data. Since the information they provide is necessarily incomplete and sometimes misleading with respect to the syntactic pattern of verbs (their valency), it was necessary to ascertain the usage of verbs in actual OE texts. This was done by conducting spot checks in the Dictionary of Old English Corpus (DOEC; diPaolo Healey et al. Reference diPaolo Healey2009) (a comprehensive examination of the whole corpus of OE is, needless to say, well beyond the frame of a journal article. This includes exact token frequency counts). As the DOEC is not lemmatised, searches were carried out using as many forms of their paradigm as possible. An example might serve to clarify the procedure. In order to find the attestations for the verb licgan ‘lie’, the forms læg- (past), leg- (past participle), liþ and lit (present) were used as search strings. Subsequently, the clauses were analysed and translated to determine both valency and meaning.

The information obtained is not exhaustive (except for verbs from A to H), but it is representative of the syntactic patterns of the verbs in question. As a result of this analysis, it was possible to compile an updated list of OE causative ja-verbs and their non-causative counterparts, containing both intact and deteriorating causative pairs, as well as unclear cases. The lists include approximate frequency counts, which have been carried out using the DOE for entries A to H, and the aforementioned dictionaries and searches in the DOEC for the rest, as described above. Finally, the data used to establish a comparison between OE causatives and those in other Germanic languages have been drawn from Schnieders (Reference Schnieders1938), Seebold (Reference Seebold1970), Riecke (Reference Riecke1996) and García García (Reference García García2005).

In the previous section, some labile verbs in OE were discussed. Besides these verbs, which do not code valency, there is a relatively large and conspicuous group of verb pairs in Old English that are coded for valency, namely causative ja-pairs, such as:

At first sight, OE could be classed as transitivising in view of the number of causatives and their frequency, both in type and token frequency. From the 106 deverbal ja-verbs that can be identified in OE (see García García Reference García García2012), some 70 may be causatives. Of those, 45 have a clear causal meaning with regards to their primary base (Appendix A), whereas for 11 verb pairs a causative interpretation can be neither discarded nor confirmed (Appendix B). Finally, 14 verb pairs were in an intact causative opposition in Pre-OE and underwent valency changes in OE, although the former relationship between the two members is still recognisable (Appendix C). These are decisive for the argument in section 6.3 and will be discussed there.

All in all, there are 59 secure causatives in OE (45 in Appendix A + 14 in Appendix C), some of them belonging to the core vocabulary and extremely frequent. Some examples are drencan ‘give drink to’, ferian ‘carry’, gremman ‘enrage’, lǣdan ‘lead’, lecgan ‘lay’, rǣran ‘raise’, settan ‘set’, swencan ‘harass’. These data suggest a large incidence of this type of construction in OE, larger, incidentally, than recent studies acknowledge. However, the value of causative verbs as evidence for the basic valency orientation, following Nichols et al.’s definition given above,Footnote ¹⁹ does not lie in their exact frequency, be it in token or type, but in the relevance of the causative construction as a valency-changing operation. Thus, although causative verbs are frequent in all Germanic languages, it is because causatives clearly constitute the main derivational strategy for code valency changes in Germanic that they can be relied upon as indicators of the language’s valency orientation, which can be then safely described as transitivising. About a third (184) of the 643 Germanic strong verbs posed by Seebold (Reference Seebold1970) have a ja-causative in one Germanic language or another (this is my own estimation after surveying the entirety of Seebold’s examples). Moreover, the ja-formation was productive in Germanic. Most of the ja-causatives lack cognates elsewhere in Indo-European, an indication that they have been coined in Germanic itself. Other valency-changing strategies did exist, but are not nearly as widespread. This is the case with anticausative na-verbs, which form a separate class in Gothic, are well represented in Old Norse, but minor in West-Germanic languages (Ringe Reference Ringe2006: 260). Ottosson (Reference Ottosson2013: 347) concludes that na-verbs might not have existed as a class in Proto-Germanic (citing Schwert Reference Schwert2001). The primacy of causatives as valency-changing operation suggests a transitivising orientation and hence a preference for intransitive verbs in this language stage (also Plank & Lahiri Reference Plank and Lahiri2015: 47).

The specific way in which the causative formation develops in early Germanic languages is crucial to assessing their basic valency types.Footnote ²⁰ The reason for this is that, as has been argued in the above paragraph, it was the predominant valency-changing mechanism in Germanic.

The valency coding mechanisms in OE will be examined against the background of its Germanic ancestor. The innovations or resilience of OE with respect to Germanic give a clearer indication of the tendencies in valency coding at work at that particular stage than a purely synchronic description, including the quantification of verbs of different valencies.

Concerning valency-changing strategies in OE, the two key questions are (1) whether other mechanisms have arisen that compete with the causative formation and (2) to what extent the causative formation is still productive or not. The answer to the first question is no. As explained above (section 4), detransitivising suffixes have all but disappeared in OE, in contrast to other Germanic languages. On the other hand, the use of reflexive middles to form anticausatives (see example 4), fairly widespread in other Germanic languages, is scarcely attested in OE (see Hermodsson Reference Hermodsson1952: 193–5; Ottosson Reference Ottosson2013: 353–4).

Once the absence of other valency-changing mechanisms in OE has been established, it is necessary to decide whether the ja-formation is still productive in OE; see sections 6 and 6.1. Obviously, if this was the case, OE should be considered transitivising, given the hegemony of the causative formation as a valency-changing operation. However, as we will see, estimating productivity proper might not be sufficient. As recent literature has shown, lack of productivity does not preclude some morphological markers being effective. Hence, a more fine-grained definition of the status of the ja-formation as a derivational mechanism in OE is needed in order to ascertain to what extent ja-verbs were recognised and used by the speakers of OE as derived, higher-valency counterparts to lower-valency basic verbs. If this was the case, OE would still be best described as transitivising. This will be addressed in section 6.2.

6 The opacity of the causative ja-formation and the rise of labile verbs in OE

As mentioned above, there are 59 secure causative pairs attested in OE, some of which are not intact:

The meanings of both weallan and its causative wyllan show metaphorical transfer in different directions – weallan does not mean ‘be tormented, agitated’, nor does wyllan express ‘create in large numbers’.

At the formal level, the ja-suffix is no longer visible in OE, except in a few verbs with light root ending in -r such as ferian ‘carry’, causative to faran ‘go’. The morphological complications of OE ja-verbs have been emphasised by many authors, from Bammesberger (Reference Bammesberger1965: 12–13) and Krahe & Meid (Reference Krahe and Meid1969: 118) to Lass (Reference Lass1994: 166–7). Compare the following ja-pair in OE and Gothic:

The distinction between causal and noncausal verb in OE rests on a sound alternation and a change of verb class, whereas it has the additional support of the ja-suffix in Gothic. The ja-formation seems to be already disintegrating in Old English, as its form became more opaque and its function less distinctive (see van Gelderen Reference Van Gelderen2011: 123–4, discussed below; see also García García Reference García García2012: 135). A clear example is the verb myltan, originally causative ‘melt (sth.)’, which already in OE adopts the additional intransitive meaning ‘melt’, like its very similar strong counterpart meltan ‘melt (intransitive)’ (see also section 6.3 below). An example of the intransitive use of myltan is:

The example of OE myltan illustrates how a labile verb can result from a scarcely distinctive causative opposition (that between meltan and myltan). This link has been detected before, recently by van Gelderen (Reference Van Gelderen2011: 122), who affirms that ‘the loss of intransitives and the resulting increase in lability are due to the loss of a productive causative affix’, without specifying when this occurred. This conclusion seems in broad terms plausible, although some details are disputable.Footnote ²¹

It would seem that the right conclusion has been reached using slightly flawed arguments. In the first place, van Gelderen’s dating of the loss of the causative suffix in Early Middle English (Reference Van Gelderen2011: 127) is too late, as will be argued in the next section. Secondly, the visibility of the causative ja-suffix in the OE period is illustrated with two examples, both inaccurate.Footnote ²² The first one is OE gladian (van Gelderen Reference Van Gelderen2011: 121, 123–4). This verb belongs to the second weak class and was therefore never formed with a *ja-suffix in the first place. The -i- that surfaces in some forms of its paradigm is not related to *-ja-, but to Germanic *-ō-/-ōja- (see e.g. Krahe & Meid Reference Krahe and Meid1969: 238–43). As was argued in section 4, this class, unlike class I, contains both intransitive and transitive verbs already before the OE period and many of its verbs show labile coding from the beginning of the textual transmission.Footnote ²³ In fact, OE gladian means both ‘to make glad’ and ‘to be glad’ (thus DOE s.v., pace van Gelderen Reference Van Gelderen2011: 121, 124).

A second questionable choice on van Gelderen’s part is the verb byrnan ‘burn’. She wonders whether byrneð could show ‘evidence of an -i causativizer through the fronted vowel of the stem’ (van Gelderen Reference Van Gelderen2011: 122). Her example (22b) is repeated here:

There can be no causativiser involved in byrneð, as it belongs to the paradigm of the strong verb. The weak causative verb is bærnan (see DOE). The form byrneð is a third-person singular present of an infinitive form byrnan or beornan, both well attested for the strong verb (see DOE s.v. byrnan ‘burn’). The root-vowel in byrneð is either a continuation of the vowel in the infinitive byrnan, or the result of the fronting of -eo- by a following third-person singular present Germanic suffix *-iþi- (beorn- + *-iþi- > byrneð). As is well known, this type of palatal mutation affects only strong verbs. In fact, what example (15) illustrates is the opposite of what the author understands: not the resilience but the demise of the causative distinction. It attests the causative-transitive use of the originally only intransitive strong verb byrnan ‘burn’. Thus, it contradicts what Bammesberger (Reference Bammesberger1965: 38–9) holds about the verb pair beornan ∼ bærnan, which van Gelderen (Reference Van Gelderen2011: 121) endorses, namely that beornan is only intransitive and bærnan only causative-transitive. The fact is that they are both labile. The DOE contains several instances of causal-transitive and intransitive uses for both verbs. It is one of the 14 verb pairs crucial to evaluating the rise of labile coding in OE, as will be discussed in section 6.3.

So far in this section two isolated cases of causative verb pairs that become labile have been discussed: byrnan ∼ bærnan and meltan ∼ myltan. They seem to point to a link between the increase of labile coding and the demise of the causative formation in OE. At this point it is necessary to widen the picture and inspect the whole group of causative verbs. Two issues will be dealt with in the next two sections. The first one is whether the causative formation was still productive in OE, as some authors claim (section 6.1). It will be argued that it was not. The second issue is whether this matters at all for an assessment of valency orientation. In this respect, the relationship between the lack of productivity of the causative formation and the increase of labile verbs has been challenged in a recent study by Plank & Lahiri (Reference Plank and Lahiri2015) on sound alternations in noncausal/causal verb pairs. The study has shown that an unproductive causative formation does not necessarily entail the language in question, German in the case of their study, changing its orientation towards labile coding. This will be discussed in detail in section 6.2.

6.1 Dating of loss of productivity of the ja-formation

Ottosson (Reference Ottosson2013: 374–5) affirms that ‘although ja-causatives were probably more opaque in Old English than in [Gothic and Old Norse], the formation seems to have retained some productivity in that language’. Van Gelderen shares Ottosson’s view, citing a previous version of his 2013 article (van Gelderen Reference Van Gelderen2011: 123–4). She dates the loss of the ‘j-affixes’ [sic] to the Early Middle English period (Reference Van Gelderen2011: 127).

The question is what is meant by the term productive. It is not difficult to prove that the causative ja-formation is not productive in Old English in the strict sense of new ja-causatives being formed (see e.g. Haspelmath & Sims Reference Haspelmath and Sims2010: 114; see also García García Reference García García2012: 135). In fact, the number of ja-verbs decreases in Old English with respect to Germanic, both in absolute terms – 59 causatives in OE against c. 184 that have been reconstructed for Germanic in Seebold’s 1972 dictionary of strong verbs – and in comparative terms – there are 72 ja-verbs attested in OE which had a causative function in Germanic, but 13 have lost this function in OE.Footnote ²⁴ One of these is:

Moreover, of the 59 secure causative pairs, just 7 (12%) are attested in OE only (dyrfan ‘bring into danger’, smīcan ‘smoke’, þwǣnan ‘reduce the size’, flȳgan ‘put to flight’, āhryran ‘destroy, cause to fall’, sȳcan ‘suckle’, ācwencan ‘extinguish’). Compare the figures with those for OHG, a similarly attested West Germanic language, with 87 secure causatives (Riecke Reference Riecke1996: 673), 31 of which (35%) are documented in OHG only (Riecke Reference Riecke1996: 675).Footnote ²⁵ The proportion of causatives is even higher in Old Norse and Gothic (see Schnieders Reference Schnieders1938; García García Reference García García2005).

From a theoretical point of view, if the causative formation was even minimally productive, there would be little chance of OE favouring labile formations, as the two tendencies run counter to each other (see e.g. Kulikov Reference Kulikov2009: 88 for Indo-Iranian). Kulikov describes the following process for Indo-Iranian:

The decay of the labile patterning essentially runs parallel with two processes: the rise and development of new valency-changing categories, causatives with the suffix -áya- ... and passives with the suffix -yá- ... , which brings the language to a more overt morphological marking of the transitivity oppositions. (Kulikov Reference Kulikov2009: 88)

In conclusion, contrary to other accounts, the ja-formation is not productive in OE at all. Furthermore, with a productive causative formation, even if only ‘to some extent’, Old English would have to be classed as a transitivising language, given the predominance of causatives as valency-changing mechanism in OE and Germanic.

6.2 Productivity vs efficiency of the causative ja-formation: the case of Present-day German causatives

Though the ja-causative formation is no longer productive, it may still be operative, that is, able to signal that causal verbs with the former suffix *-ja- are derived from a noncausal base in OE. An operative causative formation would also reveal a transitivising orientation in Old English, rather than indeterminate or labile, following Plank & Lahiri’s (Reference Plank and Lahiri2015) study on valency in German. These authors have detected a transitivising tendency in that language (‘a deep current of transitivisation’; Reference Plank and Lahiri2015: 45) on the basis of a substantial number of intransitive/transitive-causative verb pairs whose coding relies on sound alternations which are no longer productive (Reference Plank and Lahiri2015: 12), but which have a clear derivational direction, namely from basic lower-valency to higher-valency (Reference Plank and Lahiri2015: 47). Plank & Lahiri define the coding as ‘derivedness, rather than productive, online derivation’ (Reference Plank and Lahiri2015: 18).

They consider two types of sound alternations, of which only the first is relevant here. This is the alternation of the ablaut type, whereby ablaut characterises the noncausal member of causal/noncausal verb pairs, which is strong, as opposed to the causal member, which is non-ablauting and weak (Plank & Lahiri Reference Plank and Lahiri2015: 11). They cite among others the following pairs (the strong noncausal member precedes the weak causal):

These are examples of a relatively large group of 21 causal/noncausal verb pairs, which are, for the most part, reflexes of Germanic ja-causative pairs. The authors show that in these verbs, and a few others, there is an identification of intransitive=strong; transitive-causative=weak in German. The morphological coding in these pairs consists of an alternation in the root vowel and in the inflectional class, where the intransitive verb follows the strong and the causal-transitive the weak declension. Note that it is identical to that in OE causative pairs, as manifested by a comparison between the following verbs and those in the previous example:

Plank & Lahiri’s compelling conclusions raise the question of whether a similarly consistent co-variation holds for Old English ja-causative pairs, in spite of the formation no longer being productive. To test this, and in the absence of native speakers with whom to conduct the kind of experiments that Plank & Lahri undertake, we can assess the formal coding and the meaning and valency alternations that characterise each ja-causative verb pair in Old English in search of any repeating patterns that may point to systematic correspondences. In particular, we will be looking for consistent matching of strong and ablauting verbs with intransitive meaning and weak, non-ablauting verbs with transitive-causative, as Plank & Lahiri suggest exists in German. This is the focus of the next section.

6.3 Correspondences between form and valency in Old English causatives

The correlation is the expected one (strong + ablauting=intransitive, and weak – ablauting=transitive-causative) in 45 causative pairs (see Appendix A), in spite of idiosyncratic semantic changes, as in weallan above.Footnote ²⁶ Among these 45 are the few remnants of causative pairs that have survived into PDE, most of which are subject to lexicalisation and/or frequently not used in a historically faithful fashion: bite/bait, drink/drench, fall/fell, lie/lay, rise/rear/raise (Brinton & Traugott Reference Brinton and Traugott2005: 98, 153).

However, in 14 ja-causative pairs (over 31%) one or both members undergo valency changes by adopting the valency of their counterpart in addition to their own. That is, the intransitive strong verb can be used as a transitive as well, and, more often, the transitive weak ja-verb can also be used intransitively.Footnote ²⁷ They become labile verbs. Below are two well-attested verb pairs of this kind, with etymological information:

The members of the former causative ja-verb pairs wecgan (weak) ~ wegan (strong) and bærnan (weak) ~ strong byrnan (strong) can be used in both an intransitive and a transitive sense. A list of OE causative verb pairs which become labile is included in the Appendix. A few of them have not been recognised as labile in previous studies; bærnan, calan, dyrfan, hwyrfan, lǣfan, myltan, smēocan, swengan, weccan, wegan are missing from both Visser’s and Hermodsson’s lists of labile verbs in OE. However, again, what is relevant is not only that these new ascriptions increase the number of labile verbs in OE, but also, and more decidedly, the fact that these verbs had been part of a causative opposition which they have abandoned in favour of labile coding.

The spread of labile verbs that neutralise the inherited alternation expressed by the ja-formation reveals the tendencies in valency coding that were at work in OE itself. It also precludes the interpretation of the ja-formation as an effective transitivising operation in Old English, as there is no consistent co-variation between intransitive bases and transitive derived ja-verbs. This sets OE apart from German, in spite of the fact that causative verb pairs in both languages are characterised by similar morphophonetic alternations and, hence, share the same degree of intransparency.Footnote ²⁸ Intransparency of the formation cannot be, therefore, the (sole) reason for the loss of the causative alternation, in spite of what previous accounts appear to assume (Hermodsson Reference Hermodsson1952: 210; van Gelderen Reference Van Gelderen2011: 123–5; García García Reference García García2012: 135).

We turn now to labile verbs in later periods. Labile coding increases after OE. The tendency to labilisation is well established by Middle English, when only the following 12 causative pairs survive, four of which show signs of labilisation, e.g. setten ‘make (sb.) seat, sit’ (list compiled with data from Middle English Dictionary):

In summary, the basic valency orientation of Proto-Germanic can be defined as transitivising on account of the relevance of its causative formation. Causative verbs are not only frequent in type, but they also belong to the core vocabulary of early Germanic languages. Although they are widely attested in OE, there are fewer of them than in other languages of the family, such as Gothic, ON and OHG. The causative formation is definitely no longer productive at this stage. It is not even operative in the sense that Plank & Lahiri describe it, as the correspondence between strong=intransitive / weak=transitive-causative is not stable, in opposition to what seems to be the case in Present-day German. Members of causative oppositions becoming labile is an unmistakable pointer to a tendency towards labilisation in OE, as the verbs involved lose their inherent, inherited specification for valency. Attributing this loss to the opacity of the causative formation is a circular argument, as is made evident by the resilience of causative verb pairs in Present-day German in spite of their being characterised by similar morphophonetic alternations as those in OE.