Pro-drop and Pronominal Subjects:
Reanalyzing features in the history of English
Elly van Gelderen – 11 February 2011
In this paper, I investigate the conditions in Old English on the types of subject
pronouns, i.e. not using a pronoun, using an h-pronoun, or using an independent
demonstrative. In Old English, (verbal) agreement features are interpretable; they
are reanalyzed as uninterpretable in Early Middle English (with first and second
person leading the way). This means that referential pro-drop is grammatical in
Old English but disappears as the reanalysis is complete. Secondly, I argue that
Old English personal pronouns are not deictic/referential but that demonstrative
pronouns are. This situation reverses itself in early Middle English, due to both
internal factors (feature reanalysis) and external ones (contact).
The existence of pro-drop in Germanic is a controversial topic. It is generally assumed that, although many of the old and modern languages have relatively rich inflection, only topic-drop occurs. For instance, Hulk & van Kemenade (1995) argue that Old English only has expletive pro-drop but not referential pro-drop. Jaeggli & Safir (1989: 33-34) suggest that verb-second and pro-drop are incompatible and hence exclude pro-drop in Germanic.
There are other views. For instance, Sigurðsson (1993) shows Old Icelandic to have both pro-drop and topic-drop and van Gelderen (2000) argues that referential pro-drop occurs in all persons and numbers in Old English, Axel (2007) argues that Old High German has pro-drop as well but that it is restricted to the main clause.
In this paper, I argue that Old English has subject pro-drop due to the interpretable nature of the phi-features on the verb (a possibility mentioned in Holmberg 2005). For the pupuses of this paper, interpretable features are those that contribute to the meaning of the sentence whereas uninterpretable features are purely grammatical and do
not contribute to the meaning. Pro-drop disappears as the reanalysis to uninterpretable features takes place. It also has object pro-drop but this is probably licensed in another way and disappears later (cf. Visser 1963: 525). I then discuss Old English personal pronouns and argue they are not deictic/referential but that demonstrative pronouns are. This situation reverses itself in early Middle English, due to both internal factors (feature reanalysis) and external ones (contact). The outline is as follows. In section 2, I first distinguish the various null subjects and in section 3 I sketch the situation in Old Icelandic, Old High German, and Old French. Old English pro-drop differs from each of these. Section 4 turns to overt pronouns and demonstratives and section 5 is a conclusion.
2. Distinguishing various null subjects
Languages freely allow non-finite verbs to have empty subjects, usually indicated as PRO, as in (1). This PRO is different from the null subject of a finite verb, usually indicated as pro, as in (2).
(1) I want PRO to go.
(2) pro quier-o PRO ven-ir Spanish
`I want to go.‟
There have been numerous accounts that I will not go through here. Some of the better known are Taraldsen (1978) and Chomsky (1981: 240ff.) who argue that rich or strong agreement, as in (2), is responsible for licensing the null subject. Rizzi (1986: 519) has claimed that certain heads (e.g. V or I) license pro under government and Ura (1994) that pro is licensed by Case or phi-features. As we will see in section 3, there is something to be said for a connection between features and pro-drop.
Apart from referential pro and PRO, expletive pro occurs, as in (3), and pseudo-referential pro, as in (4).
(3) Es muy interesante Spanish
exp is very interesting
`It is very interesting.'
(4) llueve Spanish
exp rains, i.e. `it rains.'
English lacks pro and expletive pro but has a limited form of a fourth type of null subject, namely a null topic, as in (5) and (6). These can be used if one knows the topic of the sentence, first person singular in (5) and second or third person in (6). Topic drop is only possible if the topic is at the beginning of the root clause, as evidenced by the ungrammaticality of (7) and (8). This point will be crucial in arguing that Old English has pro-drop (as well as topic drop).
(5) Hope to talk to you soon.
(6) Shouldn't have done it.
(7) *I knew that shouldn't have done it.
(8) *Later hope to talk to you.
Other modern Germanic languages are similar to English in allowing PRO and empty-topics, e.g. Dutch but here the topic left can be an object as well, as in (9). This difference, I assume, is due to the fronted verb heb `have‟ allowing the drop in Dutch.
(9) Heb ik niet gedaan Dutch
have I not done
`That, I haven‟t done.‟
Languages such as Thai, Japanese, Korean, and Chinese are well-known for leaving both subject and object topics out, as in (10).
(10) he le Chinese
drink PST `I drank the tea‟.
The verbs in Chinese are uninflected for subject or object; the empty arguments are licensed by the discourse, very much in the way that they are in (5), (6), and (9), except that Chinese doesn‟t have a positional requirement that the left out topic be the first
constituent (see e.g. Huang 1984 for an early discussion).
Apart from Thai, Japanese, Korean, and Chinese, objects can be left out in Old Icelandic and other languages (see Rizzi 1986). Ohlander (1949: 109-110) and Visser (1963: 525-528) list Old and Middle English examples, as in (11) to
(11) Me to grunde teah fah feondscaða fæste hæfde grim on grape Old English
me to ground pulled hostile dire foe safe kept angry in grasp
`an evil sea creature pulled me to the bottom, had me fast in its grip.‟
(Visser 1963: 527; Beowulf 553-555, Fulk et al edition)
(12) `We sal', he said, `do nu ful wele' Middle English
`We shall, he said, do it now very well.' (Visser 1963; Cursor Mundi, Cotton,
13492, Morris edition)
I will not have anything to say on object-drop in this paper.
Having introduced the basic distinctions among null subjects, I‟ll now turn to older languages and to Old English.
3. Old English null subjects
In this section, I first briefly discuss pro-drop in Old High German and Old Icelandic before turning to Old English.
Axel (2007) argues that Old High German has pro-drop in main clauses that is licensed by verb-movement to the left of the null subject. An example is given in (13) where the verb hahet moves to the second position and licenses a null subject to its right.
(13) Sume hah-et in cruci Old High German
some hang-2P to cross
`Some of them, you will crucify.‟ (Axel 2007: 293; Monsee Fragments)
Axel‟s evidence for the role of verb-movement is that there is no pro-drop in subordinate clauses and that there are very few verb-initial structures with pro-drop unless they are yes-no questions. Adams (1987) similarly claims that verb-fronting in the main clause is relevant in Old French. Later in this section, I will show that Old English is quite different from Old High German and Old French in having pro-drop in subordinate clauses.
Sigurðsson (1993) argues that Old Icelandic has both pro-drop, expletive-drop, and topic-drop but that, although the inflection was rich, pro-drop wasn‟t licensed by
agreement but by a preceding NP. Examples of pro-drop and topic-drop are given in (14) and (15) respectively. In (14), the left-out third person singular goes back to the first third person not to Hoskuldr and, in (15), there is strictly speaking no antecedent but an implied topic.
(14) ok kom hann þangat ok var Hoskuldr uti er reið i tún Old Icelandic
and came he there and was Hoskuldr outside when (he) rode into field
`And he came there and Hoskuldr was outside when he rode into the field.‟
(Sigurðsson 1993: 248)
(15) þa sigldi hann ihaf oc fengu veþr stór Old Icelandic
then sailed he to-ocean and (they) got-P weather big
`Then he sailed and he and his crew got bad weather.‟ (Sigurðsson 1993: 252)
The topic-drop in (15) is possible, according to Sigurðsson, even without an antecedent, whereas the pro-drop in (14) needs an antecedent.
The Old English data show expletive null subjects, as in (16), that I will not discuss further, as well as null subjects, as in (17), that have to be pro-drop and not topic-drop because the initial position is taken up by an adverb.
(16) Nap nihtscua, norþan sniwde Old English
darkened night-shadow, north-from snowed
`The shadow of the night darkened, it snowed from the North‟.
(Seafarer 31, Krapp & Dobie edition)
(17) Nu scylun hergan hefaenricaes uard Old English
Now must praise heavenly-kingdom guard
`Now we must praise the lord of the heavenly kingdom.'
(beginning of the Northumbrian version of Caedmon’s Hymn).
Further evidence that Old English has pro-drop is that subjects in subordinate clauses are frequently null. Pogatscher (1901: 261), for instance, gives 176 null subjects “im nebensatze”, as in (18) and (19).
(18) þæt ic gumcystum godne funde beaga bryttan breac þonne moste Old English
that I manly-virtue good found ring dispenser enjoyed as-long could
`that I found a noble bestower of rings and enjoyed it as long as I could.'
(19) swylcum gifeþe bið þæt þone hilderæs hal gedigeð Old English
such given be that the battle-storm unhurt endure
-300) `May it be that he will withstand unhurt the heat of the battle.' (Beowulf 299
Even as for late Old English Aelfric, Schrader (1887: 43) says "die verwendung des pronomens [ist] häufig, doch reicht . . . das verb auch allein noch aus".
In addition to pro-drop, there is probably topic-drop in (20), as in Old Icelandic (15). This topic drop is much more like Chinese than Modern Germanic in that it is not positionally licensed. In (20), the passage has Moses as the topic but the null reference is not restricted to initial position.
(20) Heah wæs þæt handlean and him hold frea Old English
high was that reward and him kind lord
gesealde wæpna geweald wið wraðra gryre,
gave weapens power against hostile terror
ofercom mid þy campe cneomaga fela
overcame with it in-fight warriors many
(Talking about Moses), `Great was the reward and God was gracious to him (=Moses) and gave him weapons against hostile terror. He overcame many warriors with it in
battle.' (Exodus 19-21, Krapp edition)
Having shown that Old English has pro-drop, let‟s look at what licenses it. Pro-
drop isn‟t licensed by Verb-movement as in Old High German because it freely appears
in subordinate clauses, as we‟ve seen in (18) and (19). In Old English, constructions such
as (17) and (21) show that pro-drop could occur sentence-initially in declarative sentences.
(21) Wæs þeaw hyra þæt hie oft wæron an wig gearwe,
was custom their that they often were one war ready
wæs seo þeod tilu. Sigon þa to slæpe
was that people good sank then to sleep
`It was their custom always to be ready for war. They were a good people. They
went to sleep.' (Beowulf 1246-51)
This means that verb-movement is not relevant to licensing null subjects though (21) could be argued to be topic-drop.
Having established there is pro-drop occurs in Old English, we need to ask what licenses it since it isn‟t verb-movement. I think the asymmetry of where pro-drop occurs
sheds some light on this question. Although all subjects are dropped, there is a person split. For instance, Berndt (1956) shows that pro-drop occurs more with third than with first and second persons. I reproduce his data in Table 1 for the Lindisfarne and Rushworth Gospels, but see also van Gelderen (2000: chapter 3). The two sections of each represent different dialect areas: Lindisfarne and the second part of Rushworth are Northumbrian and the first part of Rushworth is Mercian. The dialect difference is not so important; the main point to the table is to show that first and second person pronouns act very differently.
Two sections of Lindisfarne Two sections of Rushworth
1S 9/212 (=96%) 9/656 (=99%) 6/191 (=97%) 21/528 (=96%)
1P 0/53 (=100%) 1/120 (=99%) 1/44 (=98%) 2/100 (=98%)
2S 16/103 (=87%) 22/308 (=93%) 12/90 (=88%) 22/226 (=91%)
2P 10/206 (=95%) 21/428 (=95%) 20/168 (=89%) 62/302 (=83%)
3S 445/116 (=21%) 1292/225 (=15%) 223/246 (=54%) 995/186 (=16%)
3P 263/108 (=29%) 618/154 (=20%) 130/141 (=52%) 528/124 (=19%)
Table 1: Numbers of Null versus Overt Subject in Lindisfarne and Rushworth (Glosses)
I‟d like to argue that this person split indicates that agreement is involved in licensing pro-drop. This would make third person agreement interpretable but first and second uninterpretable or at least variable. If we look at one of the agreement patterns in Table 2, we see that all persons are marked separately in the singular.
Present S 1 -e
Preterite S 1 -de
P -dun, -don, dan
Table 2: Old English Verbal agreement in SV with weak verbs
This is also true for the dialect of the Lindisfarne and Rushworth Glosses. Table 3 shows the verbal indicative inflections for pro-drop and pronouns preceding the verb, based on Berndt (1956: 94-131), for the first part of Lindisfarne and the first part of Rushworth, i.e. the Northumbrian and Mercian dialect respectively.
Lindisfarne (first part) Rushworth (first part)
with pronoun without with pronoun without
1S -o -o -e -u
2S -as, -es, -s, -st --- -es(t), -ast, -st -est
3S -eð, -es, -as -eð, -es, -as, -æs -eþ, eth, -aþ, -eþ, -aþ, -æþ, -iþ
Table 3: Rushworth inflections on the strong indicative verb with and without a pronoun
What we see is that, even though the inflection on the verb with preceding pronoun is variable, the ones with pro-drop are typically stronger, e.g. –st in the second person, not -
es. Whether the pronoun precedes or follows seems not to be relevant in these early texts.
In later texts, verb-movement constructions, as in (22) to (24), result in less inflection when the pronoun follows. This inflection is –e and levels first and second
person in late Old English.
(22) Nelle we ðæs race na leng teon
not-want we that argument no longer extend
`We don't want to extend that argument any more.' (Ælfric, Homilies I, 88.32)
(23) a. Nu hæbbe ge gehyred ...
Now have you heard (Ælfric, Homilies I, 280.4)
b. Ac wite ge ðæt nan man ne mæg fullice ymbe god sprecan
But know you that no man not may fully/foully around god speak
(Ælfric, Homilies I, 286.15)
(24) a. Wendes ðu ðurh wuldor ðæt þu woruld ahtest
believe you through splendor that you the world own (Christ and Satan 59)
b. yfele cwæde ðu þæt ðu þa hal3æn mæ3ne to 3yrdon næmdest
evilly spoke you that you the holy power to rods named
`You spoke evil when you ascribed the holy power to the branches.' (History of
the Holy Root Tree, 16.5-6, Napier edition)
That means first and second person agreement has leveled to –e and that pro-drop is
restricted here. By the time of Aelfric, this loss is very obvious, as shown in Table 3.
Inflection: -e/Ø full suppletive total
1 P 49 2 3 54
2 P 46 2 3 51
1 P 49 1 2 52
2 P 63 2 7 72
Table 4: Loss of inflection in VS constructions in Aelfric
Such loss does not occur with third person, as shown in (25).
(25) Þonne hæfð he wiðsacen
`Then he has rejected.' (Aelfric, Homilies II, 27.278)
The account I would therefore like to suggest is that agreement features in Old English are interpretable unlike in present-day English, as the prescriptively ungrammatical (26) shows. In (26), native speakers report that neither the Case on the pronouns not the agreement on the verb tells them who the Agent or Subject is but that the word order is the main clue. This means agreement in Modern English is not interpretable in a Minimalist framework (see Chomsky 1995 and later) but that position is.
(26) me see he made-up English
`I see him‟.
Starting with first and second person agreement, a reanalysis occurs in late Old English from interpretable to uninterpretable. One can think of this as strength as well. In Old English, pro-drop is possible but there is a person split suggesting that, if we think of pro-drop as `strength‟ of the tense and agreement features, third person features are the ones that retain the interpretable features the longest. Once the interpretable features are lost