Step 2b – Features of Cognate Accusatives: Verbal Semantics

Posted July 3, 2013 by Peter Bekins
Categories: Syntax

Tags: , , , , ,

In my reading on the cognate accusative in Biblical Hebrew, I’ve noticed that verbal semantics have been somewhat neglected. There are two aspects of verbal semantics that interest me based on the theoretical framework I developed for my dissertation:

First, to what degree does the event fit the transitive prototype? The canonical transitive clause has both grammatical and semantic aspects that line up as follows:

Grammatical Role Subject Object
Semantic Role Agent Patient [1]

Second, to what degree are the verb and object inherently related? Put differently, to what degree does the object add semantic information about the nature of event and to what degree can it exist independently of the event?

The cases that are most interesting linguistically are those in which the event deviates significantly from the transitive prototype and there is a strong inherence between object and verb.

First, consider a phrase like וַיַּעֲלוּ עֹלֹת “They offered burnt offerings” (Exod 32:6). This would seem to fit the semantic prototype since the עֹלֹת are affected/effected patients that undergo a change-of-state. Further, עֹלֹת seems to be semantically meaningful since you can העלה a wide variety of things besides עֹלֹת. Indeed, it is the presence of an object such as עלה, מִנְחָה, זֶבַח, or פַּר that invokes the specific semantic frame of sacrifice in the context of העלה rather than the more generalized meaning “to bring up”. Frankly, there is nothing particularly interesting here about the cognate nature of the accusative.

A phrase such as נִלְבְּנָה לְבֵנִים “Let’s ‘brick’ bricks” (Gen 11:3) would seem to be a middle case. On the one hand, it clearly patterns with verbs of creation where לְבֵנִים fills the role of effected patient; therefore, it fits the transitive prototype semantically. Of course, can you לבן anything else besides לבנים? This makes לבנים redundant, and liable to indefinite object deletion.[2] For instance, מַדּוּעַ לֹא כִלִּיתֶם חָקְכֶם לִלְבֹּן “Why have you not finished your order to ‘brick’?” (Exod 5:14).

Finally, we have cases such as חֲטָאתֶם חֲטָאָה גְדֹלָה “You have sinned a great sin” (Exod 32:30) and ‏וַיֶּחֱרַד יִצְחָק חֲרָדָה גְּדֹלָה “Isaac trembled a great tremble” (Gen 27:33). Neither חטא nor חדר fit the semantic prototype. For instance, חטא (Qal) returns 181 hits in the MT, but I only count two cases in which it governs an object (Exod 32:30 and 32:31) and both of these are cognate accusatives and qualified. Here, an otherwise intransitive verb is massaged into the transitive prototype by treating the phrase חֲטָאָה גְדֹלָה as an effected object which was created by the action of sinning. These are what I would consider proper cognate accusatives (or at least interesting cognate accusatives). [3]

In summary, the basic description of a cognate accusative is an accusative phrase that shares the same root (or a closely related root; note we haven’t really looked into these) as the verb. To analyze the significance of this construction I would divide the data based on noun phrase type and verbal semantics. First, is the accusative a bare indefinite noun phrase or qualified in some manner. Second, to what degree does the clause fit the semantic prototype and what is the inherence relationship between accusative and verb.

[1] An effected patient is created by the action of the agent, while an affected patient undergoes a resulting change of state due to the action of the agent.
[2] Indefinite object deletion is a phenomenon in which the object in a semantically transitive clause may be omitted if it is low in referentiality (indefinite) and can be inferred from context. Verbs of eating and drinking, for instance, often allow indefinite object deletion: ויאכלו וישתוּ “They ate and they drank” (Gen 24:54).
[3] One caveat—be careful about making assumptions concerning whether a verb is “transitive” or “intransitive.” Verbs on the boundaries are often treated differently by different languages.

Step 2a – Features of Cognate Accusatives: Noun Phrase Type

Posted July 1, 2013 by Peter Bekins
Categories: Syntax

Tags: , , ,

Now that we have a list of cognate accusatives (two lists actually) we can begin to develop some measures to further analyze the construction. The analytics window in Accordance shows a variety of patterns.

For instance, in our group of complements you’ll notice 13 particles. Those are the instances where the verb עלה is governing a prepositional phrase headed by על. The group of adjuncts has some similar cases that should be omitted.Image

You will also see that the group of complements can be divided between nouns in the absolute state and those in the construct:


At this point I have maximized what I can do with Accordance, so we will need to start tallying things by hand.[1] My first intuition is that we should look at two things: what types of noun phrases are showing up as cognate accusatives and with what semantic classes of verbs. In this post we will look at the former, the latter in the next.

Based on the analytics, the most obvious division is between bare indefinite noun phrases and noun phrases that are qualified in some manner (adjective, pronominal suffix, or in a construct relation).[2] For our complement cognate accusatives, there are about 80 bare indefinites and 90 qualified nouns. For the adjunct cognate accusatives I count roughly 25 bare indefinites and 15 qualified nouns. Overall it seems to be split rather evenly.

It is often assumed that the cognate accusative construction is used for manner modification. Biblical Hebrew (a root-pattern language) does not have a large set of derivational morphemes; therefore, to an English speaker it may seem like Biblical Hebrew has a shortage of adjectives and adverbs (I believe one of the teaching grammars uses this label).[3] More properly, the language simply expresses such concepts differently. For instance:

‏ וַיֶּחֱרַד יִצְחָק חֲרָדָה גְּדֹלָה

“Isaac trembled a great tremble” ≅ “Isaac trembled greatly”  (Gen 27:33)

In many cases, however, the qualifier does not add any adverbial nuance. Consider the following pair of sentences:

‏ וַיַּחֲלֹם יוֹסֵף חֲלוֹם

“Joseph dreamed a dream” (Gen 37:5)

‏וַיַּחֲלֹם עוֹד חֲלוֹם אַחֵר

“Joseph again dreamed another dream” (Gen 37:9)

Note how there is already an explicit adverbialעודin the second verse. The adjective אחר seems to simply be an adjective here, specifying that it was a second dream.

Further, if manner modification is the function of cognate accusatives, what is the point of all those bare indefinites? It is often suggested that the bare indefinite cognate accusatives also have an “emphasizing” function. Thus:

 שָׁם פָּחֲדוּ פָחַד

“There they shall be in great terror” (Ps 14:5 NRSV, emphasis mine)

In general, I find this highly suspect. Consider these counter-examples in Genesis:

‏ וַיִּדַּר יַעֲקֹב נֶדֶר

“Jacob really (?) made a vow” (Gen 28:20)

‏ וַיִּזְבַּח יַעֲקֹב זֶבַח

“Jacob really (?) made a sacrifice” (Gen 31:54)

‏ וַיַּצֵּב יַעֲקֹב מַצֵּבָה

“Jacob really (?) erected a sacred stone” (Gen 35:14)

While cognate accusatives can leverage the relationship between verb and object in order to introduce manner modification, this does not seem to be a good explanation for the bulk of the examples. In the next post we will consider whether verbal semantics can help us.


[1] In a perfect world, I would like to be able to dump the cognate accusatives from the previous search into a database and then tag each of them with several user defined parameters in order to cross-reference them, but Accordance is not that type of analytical tool. In a slightly less perfect world, I would like to run some sub-searches on the result set we produced with our previous search. Normally we could do this with the [CONTENTS ? ] command, but I realized there is a problem. CONTENTS returns a list of all verses that produced a hit, but the scope for a syntax search must be a chapter or book since it is possible for a syntactic unit to span two or more verses. We can use the analytics to get some broad analysis on the distribution.
[2] Conspicuously, the search returned no hits on nouns determined by the definite article. This seems to be a by-product of the way the [AGREE] command interacts with the syntax tags, namely it is not searching the entire complement phrase for roots that agree, but only the very first element in the phrase. For instance, the following verse did not produce a hit:

לִלְבֹּן הַלְּבֵנִים

“to ‘brick’ the bricks” (Exod 5:7)

Though it is syntactically analogous to this indefinite example:

נִלְבְּנָה לְבֵנִים

“Let’s ‘brick’ bricks” Gen 11:3

The grammars suggest that the typical “internal object” (look it up) is an indefinite cognate accusative. Were we doing a formal research project, we would obviously want to double check these definite examples to see if there is anything interesting going on.
[3] Of course, this line of thinking can lead to bad philosophical conclusions about differences between the Hebrew mind (i.e. they are concrete thinkers) and the Greek mind (capable of abstract thought). For instance, I recently had a man who had some seminary training (a while ago of course) try to argue with me that the “Hebrews” had no “concept of time”. I believe my face went into severe contortions and I began chanting the mantra James Barr… James Barr… James Barr…

Step 1b – Adverbial Cognate Accusatives?

Posted June 29, 2013 by Peter Bekins
Categories: Syntax

Tags: , , , ,

In the last post we did a search for complements which shared the same root as the verb. Before we analyze these results, we should do a second search to check for a similar situation with adjuncts.

If you saved the previous Hebrew Construct, simply exchange the complement phrase with an adjunct phrase.

Screen shot 2013-06-29 at 8.13.29 AM

Indeed, this search gives us another 159 hits.

Screen shot 2013-06-29 at 8.14.59 AM

If you browse these results, you’ll notice that many of the hits are infinitive absolutes. The other category of examples seem to be cases where a cognate accusative co-occurs with a direct object as in Exodus 12:14:

‏וְחַגֹּתֶם אֹתוֹ חַג לַיהוָה

“You shall celebrate it (as a) feast to YHWH”

Why are these considered adjuncts instead of complements, or why are the previous examples considered complements instead of adjuncts? Would the similar cognate accusative in this hypothetical sentence be analyzed as a complement:

וְחַגֹּתֶם חַג לַיהוָה

“You shall celebrate a feast to YHWH”

I am not precisely sure, but my intuition is that this question may need to be revisited. Cognate accusatives, of course, are one of the categories that present problems for the traditional label object, but we will worry about that later.

Our next step will be to develop some measures and criteria to analyze the distribution of this construction as we attempt to draw some conclusions about its function.

Step 1 – Searching for Cognate Accusatives

Posted June 28, 2013 by Peter Bekins
Categories: Syntax

Tags: , , , ,

Our first step will be to learn a bit about the cognate accusative construction. Rather than start with the grammars, let’s try this inductively by creating an Accordance search. Begin by opening the HMT-W4 text (not BHS-W4) and link a Hebrew Construct. So, where is the tag for cognate accusative?

When the Syntax project was introduced, one of the loudest complaints was the omission of traditional grammatical labels such as subject, direct object, indirect object, and adverb. Long story short, these labels may be convenient, but they are pre-theoretical and not well-defined for a rigorous syntax. The goal of the syntax DB, therefore, was to be as theory neutral as possible with as few labels as possible.

There are two major distinctions:

At the clause level the subject is distinguished from the predicate:

‏אַתָּה תְּשׁוּפֶנּוּ עָקֵב
[S You ] [P will bruise him (on the) heel ]

At the phrase level complements are distinguished from adjuncts:

תְּשׁוּפֶנּוּ עָקֵב
[P will bruise [C him ] [A heel]]

In theory, this allows you to define your own grammatical relations by combining the syntax data with the morphological database. The complement of an active verb is roughly equivalent to the traditional object, the adjunct of a verb is adverbial, the adjunct of a noun is adjectival, etc.

So, our first search will look for all complements of a verb in which both verb and complement share the same root.

Screen shot 2013-06-28 at 3.32.47 PM

Remember that the Syntax DB is hierarchical in nature; therefore, it is a good practice to always build your searches top-down: clause > phrase > etc. You will also want to check “search both directions” for this example.

I have combined the syntax and morphology tags in order to specify that the head of the predicate phrase is a verb. I am not sure if this is the best practice, but the search did not work as expected otherwise.

I have specified that I only want the first—highest level—complement. This is also due to the hierarchical nature of the DB. I am only interested in the complement of the verb, but other elements (like prepositions) take complements which would produce hits if this box was left unchecked.

This search produces 390 hits. Here is a screenshot from the first page:

Screen shot 2013-06-28 at 3.33.29 PM

A quick browse suggests that everything is in order, but we will have to look at the data more closely later. In the next post we will expand our search to consider the cases in which an adjunct shares the same root with the verb. Should these also be grouped with our cognate accusatives? Should we distinguish a subset of cognate objects?

Is there a cognate nominative?

Posted June 28, 2013 by Peter Bekins
Categories: Syntax

Tags: , , ,

If you have studied Biblical Hebrew, then I am sure you are well aware of the cognate accusative construction. On the simple definition, a cognate accusative is an object phrase that shares the same (or a closely related) root as the verb. The phenomenon is much more complex and interesting, of course, but I did not spend much time thinking about cognate accusatives for my dissertation since they are typically indefinite and, therefore, not object marked. 

The other day I found myself wondering whether there was also a cognate nominative and what this may tell us about the nature of the grammatical relations subject and object and/or the accusative case. A quick search of the standard BH grammars came up empty, but then I thought that this may be an interesting research question for a test of the Hebrew Syntax module in Accordance

While I participated in tagging several of the books in the module, to this point I haven’t really tried my hand at any complicated searches. Indeed, while I am intimately familiar with the theoretical framework and terminology of the DB, I must confess that I have not found earlier versions of the search interface to be particularly intuitive.  

In this series of posts we’ll do a little summer research project and see where it leads.  

A Midrash on את

Posted March 1, 2013 by Peter Bekins
Categories: Uncategorized

I presented my dissertation lecture yesterday afternoon (in lieu of a formal defense we present a public lecture and take questions), and I thought it appropriate to open with a midrash on את. The human mind is endlessly searching for patterns and meaning. Randomness is anathema. Potato chips look like Abraham Lincoln. The virgin Mary appears on grilled cheese. And so, both grammarians and interpreters have often felt that the use of את to mark an object phrase must have some meaning. The following is a midrash on את which I came across somewhat by random (HT: Davar Akher). 

Sefer HaBahir 1:32

דרש ר’ ישמעאל לר”ע מ”ד את השמים ואת הארץ אלמלא לא נאמר את היינו אומרים שמים וארץ אלהות הן א”ל העבודה נגעת אבל לא בררת כן דברת אבל את לרבות חמה ולבנה כוכבים ומזלות ואת לרבות אילנות ודשאים וגן עדן

Rabbi Ishmael expounded to Rabbi Akiva, “Why does it say את השמים ואת הארץ? Had it not said את, they would think that השמים and הארץ were gods!”

He replied, “Good Lord! You have laboured, but you have not sifted, and so you have spoken; however, את includes sun and moon, stars and constellations; את includes trees and vegetation and the Garden of Eden.”

The midrash concerns the use of את for the phrase את השמים ואת הארץ in Genesis 1:1. Rabbi Ishmael gives a reasonable grammatical argument: had השמים and הארץ not been overtly marked as direct objects, it would have been grammatically possible to read the phrase in apposition to אלהים as the subject. This is a common explanation for את; when it is used it must be necessary to distinguish subject from object. Its grammatical function is indeed to indicate the object, but when one considers the distribution of את broadly, this is not a likely explanation for the variation in its use. There are numerous examples where את is used when ambiguity between subject and object is low or where it is absent in cases where ambiguity may be high. 

Rabbi Akiva scolds Rabbi Ishmael for his interpretation, but not on grammatical grounds. Akiva’s midrash (see also Baba Qama 41b) plays on the fact that the object preposition את is a homonym of the comitative preposition את ‘with’. The idea is that wherever an object is marked by את, this also implies that there is something unsaid that should go along ‘with’ it. Here, את is alerting us that את השמים ואת הארץ should be taken as a hendiadys that includes all of the subsequent “structural” creation up to Eden (I assume his omission of birds, fish, animals, and humans was intentional). This is a good midrash, but a midrash nonetheless. 

I have argued in my dissertation that the distribution of את is not random, but I wouldn’t say that its presence is particularly meaningful in the normal sense of the word. Fundamentally, את is a grammatical marker of the direct object, and the variability in its use is largely a by-product of the way object marking systems develop. 

Dead Space

Posted January 14, 2013 by Peter Bekins
Categories: Uncategorized

The last couple years have been extremely busy. My dissertation took a bit of a rabbit trail, but I am very pleased with the results of the extra year of research. The dissertation is now complete and has been accepted, so I will formally become Dr. Bekins this June. I have posted a link to the dissertation under the pages sidebar if you are interested.

In the meantime, I have also been participating in the Accordance Syntax database project, teaching 2-3 courses per semester at Wright State University, working up several papers for submission for publication, and doing all my duties as dad and husband. So…. the blog has been the one to suffer. I hope to begin posting periodically again, but don’t keep your fingers crossed. Until then, you can read about all the wonders of ʾet.


Get every new post delivered to your Inbox.

Join 158 other followers