Abstract
This study discusses textual variants at the sentence level in 1QIsaa, in which syntactic phrases in MT, mainly objects, seem to have been interpreted as overt subjects in the scroll. These variants are attested in two syntactic contexts: (a) clauses with non-canonical word order; and (b) subjectless clauses with third-person masculine verbs. In both contexts, an overt subject does not appear in its expected linear position. Consequently, an unintentional interpretation of the first suitable syntactic phrase as an overt subject at the closest proximity to its predicate is sometimes observed in the scroll. This (mis)interpretation is evaluated from a sentence-processing perspective and explained by (a) economic processing; and (b) the probability of comprehenders’ interpreting linguistic input as a simpler and more plausible structure in their language. A preference for subject interpretation is well attested in other Subject-before-Object languages, indicating a general bias of comprehenders toward the simplest structure of the linguistic input, in which all syntactic phrases are realized in their canonical positions with no moved or absent elements.
1. Introduction 1
Textual variants in the Great Isaiah Scroll from Qumran (1QIsaa) have been recorded and discussed extensively in scholarly literature since the magisterial study by E.Y. Kutscher, The Language and Linguistic Background of the Great Isaiah Scroll (1974; first published in 1959). 2 Most variants attested in 1QIsaa, compared with the Masoretic Text of Isaiah (henceforth: MT) and other textual witnesses of Isaiah, are traditionally described at the word or sub-word level and evaluated from philological and text-critical perspectives. 3 However, a word or sub-word description may sometimes overlook types of variation occurring at the clause or verse level. As an example, we may consider the syntactic variant in Isa. 58.11.
[1] Isa. 58.11: וְנָחֲךָ֣ יְהוָה֮ תָּמִיד֒ וְהִשְׂבִּ֤יעַ בְּצַחְצָחוֹת֙ נַפְשֶׁ֔ךָ וְעַצְמֹתֶ֖יךָ יַחֲלִ֑יץ ‘And the L 1QIsaa: בצצחות נפשכה ועצמותיךה ‘And the L
At the word level, the change of יחליץ (“make strong,” 3.m.sg yiqtol) in MT to יחליצו (“make strong,” 3.m.pl yiqtol) in the scroll is merely a change in number. At the sentence level, however, as also noted by Kutscher (1974: 394–395),
4
this variant may reflect a different interpretation of the verse: in MT, עצמותיך (“your bones,” f.pl), is the object of יחליץ, whereas in 1QIsaa, עצמותיך is probably the subject of יחליצו. In other words, it seems that the L
In this study, I discuss the type of variant attested in Isa. 58.11, in which syntactic phrases in Isaiah, mainly objects, seem to have been interpreted in 1QIsaa as overt subjects in the closest proximity to their predicates. This type of variant is evidenced by various textual changes in the scroll, e.g., changes in number or gender, changes in syntactic category of a word, and the addition or omission of function words. Alongside philological and text-critical explanations, I suggest a sentence processing-based motivation for these different changes, which in these cases probably represent a later version than that preserved in MT. 5
The rest of the paper is arranged as follows. In the next section, I present the two syntactic environments in which this type of variant is generated: (a) non-canonical, scrambled 6 word-order clauses; and (b) subjectless clauses with third-person masculine verbs. In the third section, I discuss the relevant theory and findings from the psycholinguistic literature, pointing to a subject-first interpretation of comprehenders when reading scrambled word strings in their Subject-before-Object language. On the basis of these findings, I propose that a similar preference may be found in subjectless clauses, in which, as in scrambled clauses, the subject is absent from its expected linear position. Consequently, the first available fitting phrase in the clause may be interpreted as the subject in 1QIsaa. This unintentional interpretation will be explained in the relevant model of language comprehension. In the fourth section, I discuss the variants found in 1QIsaa, including Isa. 58.11. I conclude that such variants, which are not a deliberate intervention in the text but rather reflect the way a written input is interpreted and reproduced in copying, should also be evaluated from a sentence processing perspective.
2. Scrambled and third-person subjectless clauses in BH
Because biblical Hebrew (BH) displays a relatively free word order for clauses, there has long been a debate regarding its canonical linearization. 7 According to the prevalent approach, finite verbal clauses in BH are VSO; therefore, BH is typologically a VSO language. 8 A few scholars hold a different view and see BH as an SVO language. 9 According to Moshavi (2000: 17), in nonverbal and participial clauses, the canonical order seems to be Subject-Predicate. 10 In both competing approaches, word order variation, that is, the occurrence of scrambled clauses, is treated within different frameworks and offered semantic and pragmatic explanations. 11
According to both of these possible word orders, however, BH is clearly an SO language, in which the subject canonically precedes the object (or the predicate in nonverbal and participial clauses), as in example [2]. Consequently, scrambled clauses as in [3], in which an object is topicalized, deviate from the canonical SO linearization of the BH clause.
[2] Isa. 41.7: וַיְחַזֵּ֤ק חָרָשׁ֙ אֶת־צֹרֵ֔ף Lit: ‘and-encouraged artisan ACC goldsmith’ ‘And the artisan encourages the goldsmith’ [3] Isa. 50.6:גֵּוִי֙ נָתַ֣תִּי לְמַכִּ֔ים lit: back-my gave-I to-strikers ‘I gave my back to those who strike me’
In addition to scrambled clauses, the second relevant syntactic context for subject interpretation in 1QIsaa is that of subjectless clauses with third-person masculine verbs in MT. These clauses are of two types; the first is impersonal structures. According to Notarius (2021), in impersonal structures with active verbs, 12 the animate pronominal subject is not realized in the surface structure. 13 The verbal form in such structures may be singular third-person masculine, as in [4]; or plural third-person masculine, as in [5]. ([NS] represents an absent subject.) 14
[4] Isa. 7.24: בַּחִצִּ֥ים וּבַקֶּ֖שֶׁת יָ֣בוֹא שָׁ֑מָּה ‘With bow and arrows [NS] will go (3.m.sg) there’ ‘With bow and arrows one will go there’ [5] Isa. 59.19: וְיִֽירְא֤וּ מִֽמַּעֲרָב֙ אֶת־שֵׁ֣ם יְהוָ֔ה ‘And [NS] will fear (3.m.pl) from the west the name of the L ‘And those in the west shall fear the name of the L
Although both constructions exemplified in [4] and [5] are used in BH, Notarius (2021: 13) argues that the plural form in MT is not just a variant of the singular form, but also, the two forms differ syntactically, semantically, and pragmatically. For example, a singular verb usually denotes a non-referential, generic subject, and its predicate is rarely anchored in time; whereas the subject of a plural verb usually denotes a collective or group, and its predicate is anchored in time, “commonly in a quite specific sense” (2012: 28).
In the second type of subjectless clauses, the referential subject is overt in the prior input, thus serves as an antecedent for coreferential absent subjects (null pronouns) with its consecutive verbs, as in [1] above 15 (imperfective) and [6] (perfective). (Overt subjects are marked by SUB; [pronoun]x represents coreferential absent subject; x (i/j) indexes coreferential elements.)
[1] Isa. 58.11: וְנָחֲךָ֣ יְהוָה֮ תָּמִיד֒ וְהִשְׂבִּ֤יעַ בְּצַחְצָחוֹת֙ נַפְשֶׁ֔ךָ וְעַצְמֹתֶ֖יךָ יַחֲלִ֑יץ ‘And [the L [6] Isa. 6.6–7 … וַיָּ֣עָף אֵלַ֗י אֶחָד֙ מִן־הַשְּׂרָפִ֔ים וּבְיָד֖וֹ רִצְפָּ֑ה בְּמֶ֨לְקַחַ֔יִם לָקַ֖ח מֵעַ֥ל הַמִּזְבֵּֽחַ׃ וַיַּגַּ֣ע עַל־פִּ֔י וַיֹּ֕אמֶר ‘And [one of the seraphs]SUBi flew to me, in hisi hand a live coal (that) [he]i took from the altar with a pair of tongs. And [he]i touched my mouth (with it) and [he]i said …’
In Isa. 58.11, an overt NP subject יהוה (the L
It is thus apparent that when a BH clause has no overt subject NP or pronoun, and the verb is third-person masculine singular or plural, the construction has two possible interpretations: an impersonal construction, as in [4] and [5], or a coordinate construction involving a coreferential absent subject with an overt antecedent, as in [1] and [6].
So far we have seen that in two environments in BH, the subject does not appear in its expected canonical position: (a) scrambled clauses and (b) subjectless clauses with third-person masculine verbs. These two types of clauses in BH may be syntactically distinct, but in both of them the subject is absent: in scrambled clauses, the object (or another phrase) precedes it, and in subjectless clauses, even if structurally postulated, the subject is absent from the surface structure. In what follows, I discuss the implications of the properties of such clauses for their comprehension by the scribes 17 who copied 1QIsaa from some earlier type of the book of Isaiah. 18
3. Subject preference in canonical SO languages
In the psycholinguistic literature, it is generally argued that canonical structures are the simplest, unmarked structures of a language and are therefore easiest to process. By contrast, scrambled structures are inherently more difficult to process than canonical structures. 19 In behavioral experiments in SO languages, this complexity correlates with longer reading times and decreased comprehension accuracy in scrambled OS sentences compared with their canonical SO versions. 20
As an example, consider the German complement clauses in [8a] and [8b], in which die Oma (“the granny”) is temporarily ambiguous, since it might be interpreted as either subject or object until the disambiguating verb at the end of the sentence resolves the ambiguity in favour of an object interpretation.
21
(Disambiguating verbs
[7a] Fritz hat erzählt daβ die Oma einige der Kinder besucht ‘Fritz has said that the granny visited some of the children’ [7b] Fritz hat erzählt daβ die Oma einige der Kinder besucht ‘Fritz has said that some of the children visited the granny’
German readers show a consistent preference for the subject reading in [7a], evidenced by increased reading times at the last verb in [7b], where a reanalysis to the object interpretation is needed.
The preference for the canonical word order is also evident in unambiguous contexts, that is, when a case unambiguously marks the syntactic role of a phrase. In the following example from German, 22 the SO clause [8a] is read faster than its scrambled OS counterpart [8b].
[8a] Der Mann sah den Fisch TheNOM man saw theACC fish ‘The man saw the fish’ [8b] Den Fisch sah der Mann TheACC fish saw theNOM man ‘The man saw the fish’
Several approaches account for the bias toward canonical structures, all postulating a greater processing cost for scrambled clauses due to the elements moved from their canonical positions. De Vincenzi (1991), for example, attributes the preference for subject interpretation, specifically in relative clauses and wh-questions, 23 to “a general preference to posit unmoved elements over moved ones” (1991: 124). In line with this view, Gorrell (2000) argues that canonical structures reflect the base-generated structure of a language, in which each argument is realized in its unmarked, expected position. Accordingly, the subject preference is a general bias toward base-generated structures, which are the simplest structures compatible with the grammar and the input (2000: 26). 24
Gibson (1998; 2000) offers a memory-based explanation for the subject preference. According to Gibson’s Dependency Locality Theory, 25 scrambled word order creates an expectation for a required element that has not yet been encountered. Maintaining this expectation, and then integrating the moved element to its original position in the structure, increases memory load. Gibson (1998: 1) argues that the longer a predicted element is kept in memory before the prediction is satisfied, and the greater the distance between an encountered element and the position where it is interpreted, the greater the processing cost. 26
The preference for the canonical SO structure is also found in unambiguous contexts in Modern Hebrew. In a recent study by Keshev and Meltzer-Asscher (2021), comprehenders were presented with sentence beginnings as in [9].
[9] … חיבבנו את התלמידות שלמרות החששות מצאה ‘We liked ACC the students (f.pl) that despite the concerns found (3.f.sg) …’
Two completion options followed [9]: the grammatical object-extracted relative clause [9a], which presents the non-canonical VSO word order in modern Hebrew (the object is extracted and leaves a trace/gap, thus the surfaced elements are VS); and the ungrammatical subject-extracted relative clause [9b], which maintains the canonical SVO word order (the subject is extracted and leaves a trace/gap, thus the surfaced elements are VO).
[9a] המדריכה בקצה המסלול ‘The guide (f.sg) at the end of the trail’ = grammatical completion [9b] את החתול בריא ושלם = ungrammatical completion ‘ACC the cat in good shape’
Because התלמידות (“the students”) is a plural feminine NP, and מצאה (“found”) is a singular feminine verb, [9] is not ambiguous, that is, [9a] is the only grammatical option to complete it. Interestingly, nearly 80% of the comprehenders completed [9] with [9b]. The authors concluded that Hebrew readers are willing to compromise Subject-Verb agreement for the simpler and more plausible SVO word order in Hebrew, and that the bias toward the simpler structure applies even when only one grammatical reading is available. 27
The bias toward the canonical structure in Keshev and Meltzer-Asscher was accounted for by the Noisy Channel model of language comprehension, which suggests that comprehenders tend to assume an error or corruption in the input to accommodate a more plausible or simple message of the input. This tendency is also shown in pragmatically implausible sentences. For example, the sentence The mother gave the candle the daughter is usually interpreted by English readers as meaning The mother gave the candle
At this point, we may conclude that (a) scrambled clauses are more difficult to process than their canonical counterparts; (b) the greater the distance between a moved element and its original position, the greater the processing cost; (c) the preference to interpret the first available phrase as a subject is evident in different contexts in SO languages, 30 explained by the bias toward the simplest interpretation of the input.
On the basis of the findings from SO languages, including Hebrew, we may expect some evidence for subject preference in 1QIsaa. If a clause in MT presents a fitting candidate for a subject, and a minor change of the input can accommodate such an interpretation, the parser may adopt it. Turning now to the variants in 1QIsaa, let us see whether the findings confirm this expectation.
4. Locally-interpreted overt subjects in 1QIsaa
We shall now examine the variants attested in 1QIsaa. In all of them except the last, the subject in MT is not realized in its canonical position due to a scrambled word order, an absent subject, or both. Accordingly, the first available candidate seems to have been interpreted locally as an overt subject in 1QIsaa. Note that when an overt subject is interpreted over a coreferential absent subject, as in most of the examples, a new discourse referent is being introduced. According to Gibson (1998: 12–13), the building of a new discourse referent/event is costly. However, these new referents are overt subjects that are interpreted locally; thus, their integration cost should not be substantial and, crucially, they are interpreted at the expense of other new referents, mainly objects.
All variants presented herein involve a minor change of the input; thus, they may be accounted for by Noisy Channel processing. For example, when a phrase is (mis)interpreted as subject in 1QIsaa and the verb is changed to agree with it, we may assume that the scribe perceived the verbal form he produced as intended in his Vorlage.
Although adopting Kutscher’s view that 1QIsaa was copied from an earlier MT-type text (1974: 2–3) and Tov’s view (2012: 274) that it reflects a later version of Isaiah than the one preserved in MT, I do not argue that the MT text served directly as the Vorlage of 1QIsaa. Subject interpretation could have occurred at any prior stage of the textual transmission of Isaiah and may have already appeared in the scribes’ Vorlage.
I do, however, postulate that variants that indicate subject interpretation, in any textual witness, are likely to reflect a later version of their more complex variants in another textual witness; from a language processing perspective, the unintentional analysis of a linguistic input as a more complex structure is not very likely. Accordingly, subject interpretations shared with the LXX or other witnesses do not necessarily indicate a common Vorlage, as these interpretations could have been generated independently through the same processes of language comprehension.
The examples are presented according to their syntactic contexts: scrambled subjectless clauses (4.1); non-scrambled subjectless clauses (4.2). Two counterexamples are discussed in 4.3.
In all of the examples, a relevant overt subject is represented by SUB; a coreferential absent subject by [pronoun]x; an impersonal null subject by [one]x; and an uncertain absent subject by [ ]?. A relevant object is represented by OBJ, and a locative argument by LOC. x (i/j) indexes coreferential elements. In some variants, brackets marked a, b, and so on are used to outline clauses. The variant reading in 1QIsaa is bolded. The LXX translation is also presented. 31
The first example is Isa. 58.11, repeated here as 4.1.1.
4.1 Scrambled subjectless clauses
4.1.1
Isa. 58.11: וְנָחֲךָ֣ יְהוָה֮ תָּמִיד֒ וְהִשְׂבִּ֤יעַ בְּצַחְצָחוֹת֙ נַפְשֶׁ֔ךָ וְעַצְמֹתֶ֖יךָ יַחֲלִ֑יץ ‘And [the L 1QIsaa: בצצחות נפשכה ‘And [the L LXX: καὶ ἔσται ὁ θεός σου µετὰ σοῦ διὰ παντός· καὶ ἐµπλησθήσῃ καθάπερ ἐπιθυµεῖ ἡ ψυχή σου, καὶ τὰ ὀστᾶ σου πιανθήσεται
In Isa. 58.11, יהוה (“the L
4.1.2
Isa. 2.17–18: וְנִשְׂגַּ֧ב יְהוָ֛ה לְבַדּ֖וֹ בַּיּ֥וֹם הַהֽוּא׃ וְהָאֱלִילִ֖ים כָּלִ֥יל יַחֲלֹֽף ‘And [the L 1QIsaa: ונשגב יהוה לבדו ביום ההו֯א ‘And [the L LXX: καὶ ὑψωθήσεται κύριος µόνος ἐν τῇ ἡµέρᾳ ἐκείνῃ. καὶ τὰ χειροποίητα πάντα κατακρύψουσιν
The overt antecedent of the absent subject of יחלף (“pass away,” 3.m.sg yiqtol) must be יהוה (“the L
4.1.3
Isa 29:13: וַיֹּ֣אמֶר אֲדֹנָ֗י יַ֚עַן כִּ֤י נִגַּשׁ֙ הָעָ֣ם הַזֶּ֔ה בְּפִ֤יו וּבִשְׂפָתָיו֙ כִּבְּד֔וּנִי וְלִבּ֖וֹ רִחַ֣ק מִמֶּ֑נִּי ‘And the L 1QIsaa: ויואמר אדוני יען כי נגש העמ הזה בפיו ובשפתיו כבדוני ‘And the L LXX: Καὶ εἶπεν κύριος Ἐγγίζει µοι ὁ λαὸς οὗτος τοῖς χείλεσιν αὐτῶν τιµῶσίν µε, ἡ δὲ καρδία αὐτῶν πόρρω ἀπέχει ἀπ’ ἐµοῦ
In Isa. 29.13, the overt subject העם הזה (“these people,” m.sg) is locally attached to the first verb נגש (“draw near,” 3.m.sg qatal) but is not realized again with כבדוני (“honor me,” 3.m.pl qatal) and רחק (“keep far,” piʕel, 3.m.sg qatal). In the last scrambled clause/VP, the object לבו (“his heart”) precedes its transitive verb רחק. The reading of the scroll points to an overt subject interpreted locally: לבו was interpreted as a subject, as evidenced by the change of רחק to the adjective רחוק (“far”). It is possible that the scribe perceived the more frequent form רָחוֹק (16 occurrences in Isaiah out of 96 in MT) instead of רִחַק, attested only twice in MT (Isaiah). The adjective rules out the interpretation of לבו as an object in 1QIsaa. The LXX also translates לבו as a subject (see Penner, 2020: 164).
The next example presents misinterpreted conjoined objects that were disjoined in order for one of them to be realized locally as an overt subject. Similar analysis in non-scrambled subjectless clauses are discussed in examples 4.2.4 and 4.2.5 below.
4.1.4
Isa 17.3: וְנִשְׁבַּ֤ת מִבְצָר֙ מֵֽאֶפְרַ֔יִם וּמַמְלָכָ֥ה מִדַּמֶּ֖שֶׂק וּשְׁאָ֣ר אֲרָ֑ם כִּכְב֤וֹד בְּנֵֽי־יִשְׂרָאֵל֙ יִֽהְי֔וּ ‘And a fortified city will disappear [from Ephraim]LOCi, and a royal power (will disappear) [from Damascus and (from) the remnant of Aram]LOCj, [they]? will be (3.m.pl) like the glory of the Israelites’ 1QIsaa: ונשבת מבצר מאפרים וממלכה מדרמשק ‘And a fortified city will disappear from Ephraim, and a royal power (will disappear) from Damascus, LXX: καὶ οὐκέτι ἔσται ὀχυρὰ τοῦ καταφυγεῖν Εφραιµ, καὶ οὐκέτι ἔσται βασιλεία ἐν Δαµασκῷ, καὶ τὸ λοιπὸν τῶν Σύρων ἀπολεῖται· οὐ γὰρ σὺ βελτίων εἶ τῶν υἱῶν Ισραηλ καὶ τῆς δόξης αὐτῶν
Isa. 17.3 is an elliptical structure, in which the verb נשבת (“disappear,” 3.m.sg wəqatal) is elided from the second conjunct. In one possible reading of this verse, indicated by the cantillation marks in MT (see also Targum Jonathan, 34 and Ibn Ezra), the locative argument of the ellipsis clause is conjoined, that is, דמשק ושאר ארם (“Damascus and the remnant of Aram”). The change of the final verb from the plural יהיו (“be” 3.m.pl yiqtol) to יהיה (3.m.sg yiqtol) indicates that שאר ארם was interpreted as the overt subject of יהיה in the last clause of the verse, in which the subject is absent in MT and its inferred antecedent is uncertain (probably Ephraim, Damascus and the remnant of Aram). Parry (2020: 142) notes that “the subject of יהיה (1QIsaa) is “the remnant of Syria” […] MT’s verb is pl. יהיו, and its subject is either the children of Israel […] or Ephraim and Syria.”
The conjunction of דמשק and שאר ארם, as may be the case in MT, is expected from a processing perspective. When an NP is temporarily ambiguous between conjoined NP analysis and conjoined sentences analysis―in this case, if שאר ארם may conjoin to דמשק or be the subject of the next clause―the former is preferred (e.g., Engelhardt and Ferreira, 2010). However, in Isa. 17.3 (also in 4.2.4 and 4.2.5 below), a revision of this possible analysis was apparently made by the scribe. Note that this type of revision entails relatively low reanalysis costs, as the first verb (in this case, the elided נשבת) is not left with an unfilled argument slot for an object. 35 The LXX translates שאר ארם as a subject (see Penner, 2020: 118, 120).
In the next three examples, possibly impersonal constructions with active third-person masculine singular verbs in MT are in the plural from in 1QIsaa. As discussed in §2 above, both singular and plural third-person masculine active verbs are used in impersonal constructions in BH, but, as Notarius (2021) argues, the plural is not to be considered a variant of the singular in BH. In 1QIsaa, it seems the change of singular to plural in the impersonal construction is mostly evident in verbs found in the passive voice in MT. In fact, in active verbs used for the impersonal, Kutscher has listed three changes from singular in MT to plural in 1QIsaa versus three changes in the opposite direction, 36 whereas for the passive voice, the change is probably unidirectional singular passive > plural active.
In examples 4.1.5, 4.1.6, and also 4.2.2 (non-scrambled subjectless clause), I propose that the scribe did not interpret impersonal constructions but rather clauses containing overt subjects.
4.1.5
Isa. 14.32: … וּמַֽה־יַּעֲנֶ֖ה מַלְאֲכֵי־ג֑וֹי כִּ֤י יְהוָה֙ יִסַּ֣ד צִיּ֔וֹן “And what [one] answers (3.m.sg) (to) [messengers of a nation]OBJ? ‘For the L 1QIsaa: ומה ‘And what do [ LXX (only the first clause): καὶ τί ἀποκριθήσονται βασιλεῖς ἐθνῶν
Isa. 14.32 begins with a wh-question, namely, an obligatory movement of the wh-element מה (“what”) to the beginning of the clause. In this case, it moved from the direct object’s position, where it leaves a trace/gap, simplified here as follows: “whati [one] answers [what]i (to) the messengers.” The verb יענה (“answer,” 3.m.sg yiqtol) is singular, thus the following input מלאכי גוי (“messengers of a nation”) must be interpreted as the second object (dative) in an impersonal subjectless clause. 38 Parry (2020: 130) asserts that יענה in MT and יענו (“answer,” 3.m.pl yiqtol) in the scroll denote an impersonal subject, namely the two versions have the same impersonal interpretation. Contra Parry, as noted above, I propose that מלאכי גוי (מלכי in the scroll) was assigned with the overt subject role in its expected canonical position and that the verb יענה was changed to plural to agree with it. The LXX also translates מלכי as the subject of the passive ἀποκριθήσονται (employed in its active sense, i.e., “shall answer”; see Penner, 2020: 114).
4.1.6
Isa. 45.24: אַ֧ךְ בַּיהוָ֛ה לִ֥י אָמַ֖ר צְדָק֣וֹת וָעֹ֑ז עָדָיו֙ יָב֣וֹא וְיֵבֹ֔שׁוּ כֹּ֖ל הַנֶּחֱרִ֥ים בּֽוֹ [‘Only in the L 1QIsaa: אך ביהוה ליא יאמר צדקות ועוז ‘Only in the L LXX: λέγων Δικαιοσύνη καὶ δόξα πρὸς αὐτὸν ἥξουσιν, καὶ αἰσχυνθήσονται πάντες οἱ ἀφορίζοντες ἑαυτούς
The syntactic analysis of this verse in MT is difficult. Clause a may be a verbless clause containing a parenthetical impersonal string לי אמר (“[one] said to me”; e.g., Ibn Ezra, see also Parry, 2020: 324). The next input, יבוא (“come,” 3.m.sg yiqtol), does not agree in number with its following verb יבשו (“be ashamed,” 3.m.pl yiqtol), 41 indicating that עדיו יבוא may be interpreted as an impersonal scrambled clause (“to him [one] will come”; e.g., Kara, see also Notarius, 2021: 10). Some scholars suggest that the change of יבוא to יבואו results from the waw preceding ויבשו, 42 which is indeed absent in 1QIsaa. Although this is possible, I propose that יבואו points to one of three possible subject interpretations in 1QIsaa: (a) צדקות ועוז (“righteousness and strength”) from the preceding clause was interpreted as the preposed conjoined subject of יבואו, as translated in LXX (see Penner, 2020: 248, 249) 43 and as predicted from a processing perspective; 44 (b) יבואו and יבושו (“be ashamed,” 3.m.pl yiqtol) share the same subject כול הנחרים בו (“all who have raged against him”), as proposed by Rashi on MT, and as indicated by the cantillation marks, despite the lack of agreement in number between יבוא ויבשוin MT and the absent waw before יבושו in the scroll; or (c) it could be the case that the PP עדיו (“to him”) was interpreted as its homograph NP עדיו (“his witnesses”), thus interpreted as a preposed overt subject, semantically contrasting with הנחרים בו (“all who have raged against him”). None of these options is conclusive. From a processing perspective, however, option (a) is preferred, and the change of יבוא to יבואו is less likely to indicate an alternative form of an impersonal clause, as discussed above.
4.2 Non-scrambled subjectless clauses
4.2.1
Isa. 40.2:
45
דַּבְּר֞וּ עַל־לֵ֤ב יְרֽוּשָׁלִַ֙ם֙ וְקִרְא֣וּ אֵלֶ֔יהָ כִּ֤י מָֽלְאָה֙ צְבָאָ֔הּ ‘Speak tenderly to the heart of Jerusalemi and call heri because [she]i has served heri term’ 1QIsaa: דברו על לב ירושלים וקראו אליהא כיא ‘Speak tenderly to the heart of Jerusalem and call her, because LXX: ἱερεῖς, λαλήσατε εἰς τὴν καρδίαν Ιερουσαληµ, παρακαλέσατε αὐτήν· ὅτι ἐπλήσθη ἡ ταπείνωσις αὐτῆς
Isa. 40.2 presents the canonical SO linearization and an absent subject in the embedded clause כי מלאה צבאה (“because [she]i has served heri term”). The antecedent of the absent subject of מלאה in the embedded clause, ירושלם (“Jerusalem”), is not the subject of the preceding clauses but rather a part of the object of the imperative verbs דברו (“speak,” 2.m.pl; the object is a noun construction לב ירושלם, “the heart of Jerusalem”) and קראו (“call,” 2.m.pl). Because the object position is considered less prominent than the subject position, 46 an overt pronoun could have facilitated the identification of ירושלם as the antecedent for the inferred subject. 47 However, the change of the verb to the masculine form מלא in the scroll indicates that ירושלם was not interpreted as the antecedent for an inferred subject. Rather, צבאה (“her service,” m.sg), the object of מלאה in MT, was interpreted as the overt subject, 48 locally attached to the verb in its canonical position. The LXX also translates צבאה (ταπείνωσις, “humiliation”) as a subject of the passive ἐπλήσθη (“has been satisfied”; see Penner, 2020: 210, 211).
4.2.2
Isa. 41.25: הַעִיר֤וֹתִי מִצָּפוֹן֙ וַיַּ֔את מִמִּזְרַח־שֶׁ֖מֶשׁ יִקְרָ֣א בִשְׁמִ֑י וְיָבֹ֤א סְגָנִים֙ כְּמוֹ־חֹ֔מֶר וּכְמ֥וֹ יוֹצֵ֖ר יִרְמָס־טִֽיט ‘I stirred up [one]OBJi from the north, and [he]i has come from the rising of the sun [he]i called my name. And [he]i shall come [(upon) rulers]OBJ as (upon) mortar, and as a potter [he]i treads clay’ 1QIsaa (only the part of the verse following the ʾetnaḥta): ‘ LXX: ἐρχέσθωσαν ἄρχοντες, καὶ ὡς πηλὸς κεραµέως καὶ ὡς κεραµεὺς καταπατῶν τὸν πηλόν, οὕτως καταπατηθήσεσθε
Isa. 41.25 lacks a subject and has no overt antecedent in the preceding context; thus, it is probably an impersonal structure. According to this analysis, סגנים (“rulers”) is the internal argument of יבא (lit: “come,” 3.m.sg yiqtol), although in this meaning of בוא (“come upon,” “attack”), the patient role is usually realized syntactically by a PP headed by אל, על, or ל, that is, an indirect object. 50 By contrast, in 1QIsaa, סגנים―the first available candidate for subject of יבא―seems to have been interpreted as an overt subject (agent) in its canonical position, as Kutscher (1974: 403) argued, evidenced by the change of יבא into the plural יבואו to agree with it. The next verb ירמס( “tread,” yiqtol), however, remains singular in the scroll; thus, its subject is uncertain. The LXX also translates סגנים as a subject (see Penner, 2020: 222).
4.2.3
Isa. 33.23: נִטְּשׁ֖וּ חֲבָלָ֑יִךְ בַּל־יְחַזְּק֤וּ כֵן־תָּרְנָם֙ בַּל־פָּ֣רְשׂוּ נֵ֔ס ‘[Your ropes]SUBi are slack; [they]i cannot steady [the base of theiri mast]OBJ [they]i did not spread a sail’ 1QIsaa: נטשו חבליך בל יחזקו ‘[Your ropes]SUBi are slack, [they]i are not steady, LXX: ἐρράγησαν τὰ σχοινία σου, ὅτι οὐκ ἐνίσχυσεν· ὁ ἱστός σου ἔκλινεν, οὐ χαλάσει τὰ ἱστία· οὐκ ἀρεῖ σηµεῖον
The overt subject in MT is חבליך (“your ropes”), locally and canonically attached to its preceding verb נטשו (“slack”; lit. “abandoned”). The subject is not realized in the second and third clauses, which are also canonical SO clauses: “[they]i cannot steady [the base of theiri mast]OBJ; [they]i did not spread [a sail]OBJ.” The reading of the scroll indicates that the direct object of יחזקו (“steady,” 3.m.pl piʕel), כן תרנם (“the base of their mast”; noun construction) was not interpreted as a noun construction: כן (“base”) was changed into כי (“for”). This interpretation of כן generated what seems to be a scrambled clause: תרנם בל פרשו נס, “their mast (m.sg) did not spread (3.m.pl) a sail.” If this was the case, the first available phrase תרנם was interpreted as the overt preposed subject of בל פרשו (“not spread,” 3.pl qatal), that was changed to the singular פרש to agree with it. Thus, the scroll reads, “their mast did not spread a sail.” 51 The LXX also translates תרנם as a subject, but of an added verb ἔκλινεν (“has bent over”; see Penner, 2020: 184, 185).
4.2.4
Isa. 30.23: וְנָתַן֩ מְטַ֨ר זַרְעֲךָ֜ אֲשֶׁר־תִּזְרַ֣ע אֶת־הָאֲדָמָ֗ה וְלֶ֙חֶם֙ תְּבוּאַ֣ת הָֽאֲדָמָ֔ה וְהָיָ֥ה דָשֵׁ֖ן וְשָׁמֵ֑ן ‘And [he]? will give [rain (of) your seed that you sow ACC the ground and the bread of the land]OBJ and [it]SUB will be rich and plentiful’ 1QIsaa: ונתן מטר זרעך אשר תזרעך את האדמה ‘And [he]? will give [rain (of) your seed that you sow ACC the ground]OBJ and [ LXX: τότε ἔσται ὁ ὑετὸς τῷ σπέρµατι τῆς γῆς σου, καὶ ὁ ἄρτος τοῦ γενήµατος τῆς γῆς σου ἔσται πλησµονὴ καὶ λιπαρός
In Isa. 30.23, the inferred subject of נתן (‘give’, wəqatal) is probably God (e.g. Rashi, Kimhi), although it has no overt antecedent. However, the complexity in the first clause of this verse is not in connection to its subject, but rather to its object. נתן takes conjoined unparalleled direct objects, of which only one has a subordinate clause. We should note that the more structurally parallel conjuncts are, the easier they are to read and comprehend (e.g., Frazier et al., 1984). The fact that the conjuncts are not parallel may have impeded the processing of the conjoined objects in this case. The second clause lacks an overt subject, and its antecedent is uncertain; if the first clause contains conjoined objects (headed by “rain” and “bread”), referring to just one of them as an antecedent of an argument in the following clause may be costly. 52 We may recall that the object position is less prominent than the subject position; 53 this too may have impeded the inference of an absent subject in the second clause. Moreover, when the antecedent is an object, inferring a subsequent absent subject coreferential with it may not be straightforward, as mentioned in example 4.2.1 above. Thus, in the scroll we see that the second object לחם תבואת האדמה (“bread of the land,” m.sg, noun construction) was interpreted as the preposed overt subject of the second clause, evidenced by the change of והיה (“be,” wəqatal), which begins the second clause in MT, into יהיה (“be,” yiqtol) with no waw separating it from its overt subject. In the LXX translation according to most witnesses, לחם is a subject. 54
4.2.5
Isa. 10.15: אִם־יִתְגַּדֵּ֤ל הַמַּשּׂוֹר֙ עַל־מְנִיפ֔וֹ כְּהָנִ֥יף שֵׁ֙בֶט֙ וְאֶת־מְרִימָ֔יו כְּהָרִ֥ים מַטֶּ֖ה לֹא־עֵֽץ ‘Does [the saw]SUBi raise itself above (the one) who swings iti as if [one/iti] wields [a rodj and ACC (those) who lift itj up]OBJ or as if a staff should lift the one who is not wood’
55
1QIsaa: אמ יתגדל המשור על מניפיו ‘Does the saw raise itself above (the one) who swings it, as if [ LXX: ἢ ὑψωθήσεται πρίων ἄνευ τοῦ ἕλκοντος αὐτόν; ὡσαύτως ἐάν τις ἄρῃ ῥάβδον ἢ ξύλον
Isa. 10.15 has the infinitive verbal form הניף (“to wield”). Its subject is either a coreferential subject, controlled by the overt subject המשור (“the saw”) in the first clause; or it may be impersonal, that is, “as if one wields.” In MT, הניף takes a conjoined object: שבט ואת מרימיו (“a rod and ACC (those) who lift it up”), evidenced by the conjunctive waw preceding the accusative marker את. As discussed in example 4.1.4, from a processing perspective, an interpretation of syntactically similar phrases as conjoined phrases (as in MT) is to be expected, but in this case, it is at the expense of an overt subject. This may explain why some mediaeval commentators (e.g., Rashi, Kimhi, Kara) interpreted the second clause as reflected in the reading of 1QIsaa, namely, with an overt subject שבט and a single object את מרימיו. It seems that in this case, it was not only the preference for subject interpretation of the first available phrase in its canonical position which generated the interpretation in the scroll, but also semantic and pragmatic information; “the rod cannot wield those who lift it, as the saw cannot raise above those who swing it.” The LXX apparently translates this clause with a NS and שבט as an object (see Penner, 2020: 92).
In the next two examples, predicates seem to have been interpreted as overt subjects in 1QIsaa. This may suggest that the subject interpretation may not be limited to specific syntactic contexts, but rather reflects a general preference for the simplest available structure of the input, in which overt subjects are interpreted locally.
4.2.6
Isa. 66.18: וְאָנֹכִ֗י מַעֲשֵׂיהֶם֙ וּמַחְשְׁבֹ֣תֵיהֶ֔ם בָּאָ֕ה לְקַבֵּ֥ץ אֶת־כָּל־הַגּוֹיִ֖ם וְהַלְּשֹׁנ֑וֹת וּבָ֖אוּ וְרָא֥וּ אֶת־כְּבוֹדִֽי ‘And I their deeds and their thoughts [ ]? came to gather [all the nations and the languages]OBJi, and [they]i will come and [they]i will see my glory’ 1QIsaa: וׄאנוכי ‘And I [ LXX: κἀγὼ τὰ ἔργα αὐτῶν καὶ τὸν λογισµὸν αὐτῶν ἐπίσταµαι. ἔρχοµαι συναγαγεῖν πάντα τὰ ἔθνη καὶ τὰς γλώσσας, καὶ ἥξουσιν καὶ ὄψονται τὴν δόξαν µου
The first word string in this verse,ואנכי מעשיהם ומחשבתיהם (“and I their deeds and their thoughts”) is syntactically and semantically uncertain. Rashi does not account for the role of אנכי (“I”), and asserts that מעשיהם ומחשבתיהם is the conjoined subject of באה (“came,” 3.f.sg qatal) despite the lack of agreement in gender and number. By contrast, Kimhi considers ואנכי מעשיהם ומחשבתיהם a complete verbless clause, in which מעשיהם ומחשבתיהם is the predicate (“I am (with) 56 their deeds and their thoughts”). Accordingly, באה has no subject. Kimhi, following Ibn Ezra, says that the subject of “came” is עת (“time,” f.sg), i.e., “time has come,” that should be inferred from the context. Parry (2020: 439) summarizes as follows: “Inasmuch as the subject of באה is open to question, scholars propose a variety of theories. 1QIsaa’s variant of באו does not add understanding to the problem.” Contra Parry, I propose that the variant in 1QIsaa points to the reading of מעשיהמה ומחשבותיהמה as the overt conjoined subject of באה, that was changed to the plural באו to agree with it.
In the second clause, it seems that את כל הגוים והלשנות (“ACC all the nations and the languages”), which is the conjoined object of the infinitive לקבץ (“to gather”), was disjoined at some point, probably upon encountering the next input ובאו וראו (“come and see,” 3.pl wəqatal). Accordingly, הלשנות was interpreted as the overt subject of ובאו וראו, similarly to the assumed process in 4.1.4 above. This process/interpretation is evidenced by the superlinear conjunctive waw added to באו, which may indicate a revision of the text after production.
57
In other words, the scribe probably interpreted this verse as a coordination of two scrambled clauses with overt subjects, and then revised it with an added waw:
b[והלשוׄנות באו וראו את כבודׄי] a[מעשיהמה ומחשבותיהמה באו לקבץ את כׄול הגואים] ואנוכי ‘And I [[their deeds and their thoughts]SUB came to gather all the nations]a [[and the languages]SUB will come and see my glory]b’
Such a reading reflects the simplest structure of the input, and it is predicted in light of economic processing, given both scrambled clauses and absent subjects in MT. However, the final reading has the superlinear waw, and both initial and final readings of 1QIsaa do not account for the role of אנכי in this verse. The LXX translates אנכי as a subject and מעשיהם ומחשבתיהם as a conjoined object of an added verb, ἐπίσταµαι (“know”). It is also the subject of the next verb, ἔρχοµαι (“come”; see Penner, 2020: 336, 337).
4.2.7
Isa. 66.12: … הִנְנִ֣י נֹטֶֽה־אֵ֠לֶיהָ כְּנָהָ֨ר שָׁל֜וֹם וּכְנַ֧חַל שׁוֹטֵ֛ף כְּב֥וֹד גּוֹיִ֖ם וִֽינַקְתֶּ֑ם עַל־צַד֙ תִּנָּשֵׂ֔אוּ ‘[Here I am extending peace to her like a river]a [and the wealth of nations (I am extending to her) like a flooding stream]b [and youSUBi will nurse]c [on arm (lit: on side) youSUBi will be carried]d…’ 1QIsaa: הנ֯ני נוטה אליהא כנהר שלום וכנחל שוטפ כבוד גואים ‘[Here I am extending peace to her like a river]a [and the wealth of nations (I am extending to her) like a flooding stream]b [ LXX: Ἰδοὺ ἐγὼ ἐκκλίνω εἰς αὐτοὺς ὡς ποταµὸς εἰρήνης καὶ ὡς χειµάρρους ἐπικλύζων δόξαν ἐθνῶν· τὰ παιδία αὐτῶν ἐπ’ ὤµων ἀρθήσονται
Clauses a, b, c in MT are not scrambled, only clause d. The NS of וינקתם (“you nurse,” 2.m.pl wəqatal) in clause c should theoretically be identified with no substantial effort, as the second person is morphologically marked on the verb. However, the first two clauses are syntactically complex; the string נוטה אליה (“extending to her”) is overt only once and has to be inferred after וכנחל שוטף (“and like a flooding stream”). 58 The null subject “you” of the subsequent וינקתם is a new agent of a new event that follows the previous introduced referents “I” and “her” and their related events in the preceding input. Introducing “you” with the new event of nursing may be costly at this point, given the other referents and events in this verse.
Clause d of the verse is scrambled, with a preposed adjunct על צד (lit: “on side”). The following predicate תנשאו (“be carried,” 2.m.pl yiqtol) shares the coreferential NS of וינקתם. In 1QIsaa, however, a different interpretation is evident: וינקתם was interpreted as an NP, ויונק]ותיהמה] (probably “and their nursing children”; it must be an NP, as it bears the genitive suffix המה), and it was assigned with the overt subject role of תנשאו, which in turn was changed to תנשינה (3.f.pl yiqtol) to agree with it. Once again, economic processing may account for this variant: תנשאו has an overt subject interpreted close to it, and in the scroll, there is one new event fewer (nursing) compared to MT. The LXX also translates וינקתם as an NP in the subject position of the passive ἀρθήσονται (“will be carried”; see Penner, 2020: 336, 337).
4.2.8
Isa. 45.8: הַרְעִ֤יפוּ שָׁמַ֙יִם֙ מִמַּ֔עַל וּשְׁחָקִ֖ים יִזְּלוּ־צֶ֑דֶק ‘Shower, O heavens, from above, and [skies]SUB will pour down [righteousness]OBJ’ 1QIsaa:
59
הריעו שמים ממעלה ושחקים ‘Cheer, O heavens from above and skies, LXX: εὐφρανθήτω ὁ οὐρανὸς ἄνωθεν, καὶ αἱ νεφέλαι ῥανάτωσαν δικαιοσύνην
Isa 45.8 is an example of an apparent subjectless clause. The overt subject in the second colon is שחקים, (“skies,” m.pl), which stands in synonymous parallelism with שמים (“heavens,” m.pl) in the first colon. It seems that שמים and שחקים were interpreted as conjoined phrases (vocative) of the first imperative verb הרעיפו (“shower” 2.pl), which was changed into perfect/imperative הריעו (“cheer”). The interpretation of conjoined phrases, which is predicted by processing strategies, is evidenced by the addition of conjunctive waw before יזל (“pour,” 3.m.sg yiqtol), ruling out שחקים as its subject. 60 Accordingly, יזלו (“pour,” 3.m.pl yiqtol) no longer has a subject; as predicted, the clause was not interpreted as a subjectless clause with an overt direct object. Rather, the object צדק was assigned the subject role, evidenced by the change of יזלו into singular.
Kutscher (1974: 400) explained that “the entire verse was corrupted because the scribe did not understand the word הרעיפו […] he took צדק to be the subject of the second half of the verse, and hence misconstrued the entire verse.” I do not agree that the misinterpretation of הרעיפו motivated the syntactic analysis of this verse in the scroll, but rather propose that the (predicted) coordination of שמים and שחקים, resulting in an absent subject of יזלו, underlies this variant. The LXX translates צדק as an object, as in MT (see Penner, 2020: 244).
The examples discussed thus far all indicate the preference for an overt subject to be interpreted locally according to the canonical SO linearization. The following two variants are counterexamples, namely, the subject in two clauses in MT was interpreted as having a different syntactic role, at odds with MT and with the preference for interpreting the first available phrase as a subject.
4.3 Non scrambled (+subject)
4.3.1
Isa. 54.3: כִּי־יָמִ֥ין וּשְׂמֹ֖אול תִּפְרֹ֑צִי וְזַרְעֵךְ֙ גּוֹיִ֣ם יִירָ֔שׁ וְעָרִ֥ים נְשַׁמּ֖וֹת יוֹשִֽׁיבוּ ‘[For (to the) right and left you will spread out]a [and your seedSUBi nationsOBJ will dispossess (3.m.sg)]b [and desolate cities [they]i will settle (3.m.pl)]c’ 1QIsaa: כיא ימין ושמואל תפרוׄצי ‘[For (to the) right and left you will spread out]a [and LXX: ἔτι εἰς τὰ δεξιὰ καὶ εἰς τὰ ἀριστερὰ ἐκπέτασον, καὶ τὸ σπέρµα σου ἔθνη κληρονοµήσει, καὶ πόλεις ἠρηµωµένας κατοικιεῖς
Clauses a and c in MT are scrambled OS with null subjects. Clause b is SOV with an overt subject: זרעך (Subject) גוים (Object) יירש (Verb); “your seed will dispossess nations.” MT’s reading assigns זרעך (“your seed,” m.sg) with the subject role. The change of the verb יירש (“dispossess,” 3.m.sg yiqtol) to plural in 1QIsaa (and in 1QIsab) indicates that גוים (“nations,” m.pl), which immediately precedes the verb, was interpreted as its subject (and the inferred subject of clause c), leaving זרעך with the object role in clause b. Accordingly, the meaning of this verse in 1QIsaa is completely opposite to its meaning in MT: “the nations will dispossess your seed and will settle desolate cities.” The LXX translates זרעך as the subject of יירש (see Penner, 2020: 286).
The motivation for the change attested in 1QIsaa may be semantic, namely, the animacy of גוים versus the inanimacy of זרעך; animate entities are syntactically “privileged,” i.e., they tend to be realized both as subjects and in early word order positions. 61 Alternatively, this change may be structurally motivated; in clauses a and c, a locative argument is preposed, but in clause b, the subject is preposed. 62 Moreover, the verb in clause c is the plural יושיבו (“settle,” 3.m yiqtol); therefore, its inferred subject זרעך (m.sg) in clause b agrees with it only if denoting a collective noun. In clause b, however, זרעך agrees with a singular verb, יירש. The lack of agreement in number may have interfered with the interpretation of זרעך as the subject of clause b in the scroll. In summary, the variant in the scroll may be semantically or structurally motivated. Note that the resulting aberrant reading of the scroll(s) indicates this variant too cannot be intentional.
4.3.2
Isa. 6.10: פֶּן־יִרְאֶ֨ה בְעֵינָ֜יו וּבְאָזְנָ֣יו יִשְׁמָ֗ע וּלְבָב֥וֹ יָבִ֛ין וָשָׁ֖ב וְרָ֥פָא לֽוֹ ‘[Lest [he]i see with hisi eyes]a [and with hisi ears [he]i hear (3.m.sg)]b [and [hisi heart]SUB will understand]c [and [he]i will turn and [he]i will be healed]d’ 1QIsaa: ‘[Lest [he]i see with hisi eyes]a [and with hisi ears [he]i hear (3.m.pl)]b [
(The overt coreferential antecedent of [he]i is העם הזה, ‘these people’, m.sg, in the preceding verse. Ø represents the omitted waw in 1QIsaa.)
LXX: µήποτε ἴδωσιν τοῖς ὀφθαλµοῖς καὶ τοῖς ὠσὶν ἀκούσωσιν καὶ τῇ καρδίᾳ συνῶσιν καὶ ἐπιστρέψωσιν καὶ ἰάσοµαι αὐτούς
This verse presents two sentence-level variants. The first one is the change of ולבבו יבין (“and his heart will understand”) in MT, wherein לבבו is the preposed overt subject of יבין, to בלבבו יבין (“with hisi heart [he]i will understand”) in 1QIsaa, 63 in which the subject is absent and the preposition ב (“with”; lit. “in”) is added. In this case, clauses a and b, יראה בעיניו ובאזניו ישמע (“see with his eyes, and with his ears hear”), are semantically similar and structurally parallel, although chiastic. Syntactically, this structure may have primed 64 the interpretation of the next input לבבו as a PP. Semantically, לבבו is a body part, as are the preceding PPs בעיניו and באזניו. Thus, the syntactic priming of the PPs in the first bicolon, and their semantic role (instrument), may have facilitated the interpretation of לבבו as an adjunct of יבין rather than its subject (the LXX also translates לבבו as an adjunct of יבין; see Penner, 2020: 72). The other variant is somewhat hybrid: ובאזניו ישמע in MT has become ובאוזניו ישמעו in the scroll. The change of the verb to the plural ישמעו may point to the (predicted) interpretation of the preposed אזניו (f.pl) as the overt subject of ישמעו, 65 although the preposition ב has remained in 1QIsaa, ruling out a conclusive subject interpretation.
5 Conclusions
Sporadic textual variants in 1QIsaa exhibit a subject interpretation at the expense of other syntactic phrases in MT, mainly objects. These variants seem to reflect a general bias towards the canonical base-generated structure in SO languages, i.e. the simplest structuring of the input with no moved elements, as argued by Gorrell (2000) and De Vincenzi (1991). Because scrambled clauses are harder to process than canonical clauses, the preference for subject interpretation is explained by economic processing, that is, the preference for the easiest structure to process in terms of required cognitive resources.
In this article, it has been shown that the interpretation of the first fitting phrase in the input as an overt subject occurs in 1QIsaa in cases where MT has scrambled clauses, as may be predicted. Similarly, as proposed, it is also found in subjectless third-person masculine verb clauses. In §4.1, it has been shown that these two syntactic contexts are sometimes combined. In these contexts, the subject is not realized in its expected linear position. In line with Gibson’s Dependency Locality Theory (2000), I argue that the integration cost of an absent subject in the structure, whether a distant overt subject or an obligatory null impersonal subject, may account for this preference in subjectless clauses, despite the cost of integrating a new discourse referent.
Once again, it must be stressed that the variants discussed throughout are not assumed to be intentional, i.e., consciously created by the scribes. The findings from the literature on Noisy Channel processing show that comprehenders may unconsciously perceive the linguistic input as different from its actual form, to conform to what is probable in their language. Nevertheless, such Noisy Channel inferences do not occur consistently: in some instances, readers adopt a “close” alternative, whereas in other instances, they are faithful to the linguistic input.
In addition to philological and text-critical analyses, changes at the clause level in 1QIsaa and other copied texts should also be evaluated from a sentence-processing perspective, which accounts for the cognitive aspects involved in the copying task. In 1QIsaa, such a perspective may provide another explanation for differences in meaning of some verses when compared with MT and other textual witnesses of Isaiah.
Footnotes
Acknowledgements
This paper is part of my doctoral dissertation, “Aspects of Language Processing in the Text of the Great Isaiah Scroll (1QIsaa),” written under the supervision of Prof. Aya Meltzer-Asscher and Prof. Noam Mizrahi. The research was supported by the Israel Science Foundation through Prof. Mizrahi’s grant, “Revealing the Sealed Document: Revisiting the Qumran Isaiah Scrolls” (ISF 1000/20). I wish to thank two JSOT reviewers for their helpful comments. I owe special thanks to Dr. Simon Pulleyn, who kindly checked the citations from the Septuagint for me.
1
Throughout the paper, the following abbreviations are used:
2
The most recent and comprehensive work about variants in all textual witnesses of Isaiah is Parry’s valuable Exploring the Isaiah Scrolls and their Textual Variants (2020).
3
On syntactic variants in 1QIsaa see, e.g., Abegg (2011); Fassberg (2000); Fleck (2022, 2023); Muraoka (2000); Pulikottil (2001); 
: 369–446).
4
Some changes in number, including the variant in Isa. 58.11, are explained by Kutscher as errors of the scribe regarding the subject of the verb. In Isa. 58.11, this error is attributed to the influence of the plural noun עצמותיך, “your bones.” I follow Kutscher’s observation that עצמותיך was misinterpreted as the subject of the clause, but I argue for a different motivation.
5
According to , “the community that transmitted M[T] has left the biblical text virtually unchanged for some two thousand years since the time of the proto-Masoretic texts from the Judean Desert […] whereas many Qumran scrolls […] changed the orthography, morphology, and content of the text. Thus, 1QIsaa, dating to the 1st century
6
7
According to Bornkessel, Schlesewsky, and Frederici (2002: B22 n.1), the word order of a sentence is canonical if this sentence may be felicitously uttered with no constraining context. Because the context in BH is limited to a written corpus, observations about canonical word order are mainly frequency-based.
8
The canonical position of the verb in BH is not addressed in the current study, only the linearization of subjects and objects which is not debatable. I therefore present the views on BH as VSO/SVO but follow in using SO to denote a language in which the subject precedes the object with no regard to the position of the verb.
9
10
11
See, e.g., Hornkohl (2018); Moshavi (2000); Robar (2018); .
12
Impersonal structures in the passive voice are syntactically and semantically different and are therefore not discussed here.
13
14
The use of NS throughout this paper is only descriptive and does not entail the postulation of an empty syntactic category.
15
Coordinate structures with coreferential subjects as in example [1] are a context that favors null subjects even in non-NS languages (e.g., Almor et al., 2017: 101), but they may be subject to language-specific constraints. Consider an example from English, a non-NS language: They ate the food and drank the beer and danced and sang songs and partied (van Oirsouw, 2019: 104, example 6). The exact derivation of such structures is debatable; see e.g., for a review of this debate.
16
Traditionally, BH is considered an NS language and Modern Hebrew a partial NS language (see Holmstedt, 2013: 265–267 for an overview of the topic). Naudé (1994) regarded Qumran Hebrew as an NS language, but this view was emphatically rejected by Muraoka (2020: 4, n.2). Some scholars, however, point to a mixed behavior of the NS in BH as well. For example, Vainikka and Levy (1999: 643, n.21) argue that wayyiqtol is the only context that allows an absent referential third-person subject in main clauses: “Except for this special construction, third person subjects cannot be omitted in Biblical Hebrew.” According to Barbosa (2019), consistent NS languages, e.g., Italian, and partial NS languages, e.g., Brazilian Portuguese, exhibit distinctive behavior. One defining distinction between consistent NS and partial NS languages is the impersonal construction with a third-person singular verb. In contrast to partial NS languages, “A consistent NS language cannot use a plain null subject to convey the meaning of a generic (inclusive) subject and must resort to some overt strategy” (Barbosa, 2019: 490; see other distinctive features, a comprehensive discussion and references therein). In line with this observation, if BH is indeed a consistent NS language, then Isa. 7.24: בַּחִצִּ֥ים וּבַקֶּ֖שֶׁת יָ֣בוֹא שָׁ֑מָּה (example [4]; see also Notarius, 2021: 10, example 7b) does not convey the impersonal generic reading “one will go there,” but rather a referential reading, “he will go there.” As this study only identifies an absent subject of third-person masculine verbs in MT Isaiah as a relevant context for subject interpretation in 1QIsaa, namely, it focuses on surface structures, the questions of why and how an absent subject is licensed in BH, what type of absent subject may appear in different contexts, or whether BH is a consistent or partial NS language, are not expounded upon here.
17
See Popović, Dhali, and Schomaker (2021: 2). The authors have demonstrated that columns 1–27 in 1QIsaa were copied by one scribe and columns 28–54 by another scribe. Their findings support the view of previous scholars who argued that the scroll was the work of two scribes.
18
Kutscher (1974: 2–3) concluded that 1QIsaa reflects a later textual type than the Masoretic Text. According to : 239), Kutscher’s work and his conclusion account for most scholars’ view of 1QIsaa as secondary to the text of MT. See also Tov’s view (2012: 274) in n.5.
20
E.g., Basque: Erdocia et al. (2009); Dutch: Frazier and Flores d’Arcais (1989; among others); Finnish: Hyönä and Hujanen (1997); German: Schriefers, Friederici, and Kühn (1995; among others); Italian: De Vincenzi (1991); Japanese: Miyamoto and Takahashi (2002; among others); Serbo-Croatian: Stojanovic (1997); Russian: .
21
Bader, Meng, and Bayer (1999), cited in : 531).
22
Krems (1984), cited in Fanselow, Kliegl, and Schlesewsky (2002: 176).
23
E.g., the preferred subject-extracted relative clause: the man
24
The preference for the simplest structure is shown across a wide range of ambiguities, construction types, and languages. See, among many others, Carlson (2001); Carlson, Dickey and Kennedy (2005); Frazier and Fodor (1978); De Vincenzi (1991); .
25
First termed “Syntactic Position Locality Theory” (Gibson, 1998), later revised to the “Dependency Locality Theory” (DLT; Gibson, 2000).
26
Other approaches explaining the preference for the canonical word order include interference-based and expectation-based accounts. See, e.g., Gordon, Hendrick, and Johnson (2001; interference-based account); ; expectation-based account).
27
On Noisy Channel processing see, e.g., Huang and Staub (2021); .
28
31
LXX translations are given for the sake of comparison. It should be noted, however, that translating and copying are two different tasks.
32
33
BDB, 322.
34
According to British Library Ms Or 1474.
35
Engelhardt and Ferreira (2010: 505); : 27).
36
MT active singular > 1QIsaa active plural: Isa. 21.9; 30.14; 53.9. MT active plural > 1QIsaa active singular: Isa. 13.2; 30.6; 33.23 (1974: 403).
37
MT מלאכי (“messengers”) > 1QIsaa מלכי (“kings”/“messengers” with no א).
39
The clause division accounts for the lack of agreement between יבוא and יבשו.
40
The change of אמר to יאמר is irrelevant for the current discussion.
41
The accent on יבוא in MT connects it to יבשו despite the lack of agreement.
43
The verse division in LXX is quite different from MT and 1QIsaa. However, צדקות ועז were translated as the conjoined subject of יבואו.
44
The Late Closure processing strategy (Garden Path model; Frazier, 1987) suggests that the parser attempts to attach new items of the linguistic input into the clause or phrase postulated most recently. Applying this (unintentional) strategy in Isa. 45.24, the preferred attachment of עדיו יבואו is to its preceding input צדקות ועז, as suggested in option (a) and as LXX translates.
45
Moshavi notes, “According to de Regt [1991], the percentage of verb-first subordinate clauses is even higher than verb-first nonsubordinate clauses” (2010: 10, n.8). In line with this observation, Isa. 40.2 is classified as a canonical (-scrambled) clause.
46
See, e.g., Carminati (2002); .
47
48
According to Parry (2020: 277), “The noun צבא takes either a m.sg or f.sg verb (see BDB, 838), thus, either מלאה or מלא is grammatically acceptable.” However, BDB bases the uncertain feminine gender of צבא on Isa. 40.2 and Dan. 8.12, וְצָבָ֛א תִּנָּתֵ֥ן עַל־הַתָּמִ֖יד, a difficult verse in which תנתן (“will be given”) may be 3.f.sg, making צבא a subject agreeing with a feminine verb, or 2.m.sg, making צבא the object in this verse. : 392) nonetheless asserts that “according to the Scr[oll] the subject is צבא.”
49
The waw addition before the verb ירמס points to an interpretation of כמו חמר and כמו יוצר as conjoined phrases, as predicted from a processing perspective and discussed in example 4.1.4.
50
E.g., BDB, 98, III.1.b; cf. אתי על (“attack”) in Jewish Palestinian Aramaic in Sokoloff (2017: 59, meaning 3). : 293) presents BHS’ view that יבא was originally יבס, “with the sāmek dropping out via haplography ויב
52
53
See n.46.
54
55
The last clause of the verse, כְּהָרִ֥ים מַטֶּ֖ה לֹא־עֵֽץ, is syntactically and semantically difficult. As no change is attested in this clause in the scroll, I present only its NRSV translation with no analysis.
56
Kimhi adds “with.”
59
ויזל/ייזל; reading is uncertain. Kutscher (1974) and read ויזל.
61
See, e.g., Branigan, Pickering, and Tanaka (2008) and referenced works therein.
62
63
4QIsaf: אזנו ישמע ובלבבו יבין[וב].
64
On syntactic priming effects, i.e., the tendency for speakers or writers to reuse syntactic forms they have recently produced, see e.g., Sturt, Keller, and Dubey (2010) and referenced works therein.
65
Although עם (“people,” m.sg), which is the inferred subject of Isa. 6.10, may be collective and agree with a plural verb.