Sleep Like Death, Death Like Sleep

The boys, the wife and I watched the latest episode of the rebooted Cosmos last night. About 10 minutes in, Neil deGrasse Tyson began talking about the idea that life on Earth may have begun by arriving on meteorites. It’s known that rocks from Mars, for example, have ended up on Earth this way. It’s also known that some bacteria are able to survive in space, as proven by bacteria that survived a stint traveling on the outside of the International Space Station. Finally, it’s known that some bacteria can survive for a long time without a food source. On this point, Tyson talks about some recently revived bacteria found in Antarctic ice:

Even more amazing are these creatures, awakened from a death-like sleep of eight million years…

I was interested to hear Tyson put it that way, because I’ve also been hearing another person talking about death-like sleeps recently, but she phrases it differently:

Did you hear that? She said:

Before the sun sets on her sixteenth birthday, she will fall into a sleep like death!

Both phrases are talking about a sleep, not about death. We know this from the context, and from the fact that the verbs fall and awaken collocate more strongly with sleep than with death. But they’re phrased in completely opposite orders from each other! Furthermore, it’s syntactically possible for each phrase to be referring to death, not to a sleep. No, I haven’t actually found any examples of this, but it could happen, OK?

Here are the structural differences all sorted out. The diagrams on the left refer first to a death that is like sleep, and then to a sleep that is like death. In these parses, the adjective like is looking for a noun-phrase complement on its right to form an adjective phrase. The diagrams on the right refer to a sleep that is death-like, then to a death that is sleep-like. Here, the adjective like forms a compound adjective with the noun phrase on its left.

The situation reminds me of Shel Silverstein’s “snake eating cake”.

Let’s Diagram the Oath of Office!

Just in time for tomorrow’s inauguration ceremony, but a little bit late for the actual swearing in that took place today, here is the presidential oath of office, as written in the Constitution, put into a tree diagram just for you! Over the years, I’ve used the PHP Syntax Tree Drawer to make my diagrams, but a couple of years ago, Miles Shang’s Syntax Tree Generator came online, so now I have two phrase-diagramming apps to choose from. I couldn’t decide which one to use this time, so I chose both! Now you can decide which style you prefer. Here’s the diagram from the PHP Tree Drawer, with the familiar blue labels and red words you’ve come to love, and the top node of the tree centered horizontally. Click to embiggen.

Now here’s the diagram done with Shang’s Tree Generator, with blue node labels and green words. The top node of the diagram, like all the nodes in the tree, dominates branches of equal length, instead of making one branch reach much farther than the other, as you can see happens with the diagram above. Another nice thing about Shang’s Tree Generator is that it allows you to draw movement lines, so that if your theory of syntax has WH words actually moving from a place inside a clause to the front of the sentence (for a WH-fronting language like English), you can do that. On the other hand, the PHP Tree Drawer makes it easier to put subscripts on the labels. Look closely at my VP labels, and you’ll see that in the upper diagram, they’re subscripted to show whether they are nonfinite (headed by a verb’s base form in this sentence) or finite, but no such subscripts appear in the diagram below.

If you want to try out these apps yourself, here’s the string I used to generate the tree for both of them:

[Clause_fin [NP [Pron I]] [VP_fin [Aux do] [VP_base [Adv solemnly] [VP_base [V_base [V_base swear] [Conj or] [V_base affirm]] [Clause_that [Comp that] [Clause_fin [NP [Pron I]] [VP_fin [VP_fin [Aux will] [VP_base [Adv faithfully] [VP_base [V_base execute] [NP [Det the] [Nom [N Office] [PP [P of] [NP [N President] [PP [P of] [NP [Det the] [Nom [Adj United] [N States]]]]]]]]]]] [Conj and] [VP_fin [Aux will] [VP_base [PP [P to] [NP [Det the] [Nom best [PP [P of] [NP [Det my] [N ability]]]]]] [VP_base [V_base [V_base preserve] [V_base protect] [Conj and] [V_base defend]] [NP [Det the] [Nom [N Constitution] ][PP [P of] [NP [Det the] [Nom [Adj United] [N States]]]]]]]]]]]]]]]

In particular, you might not agree with how I’ve parsed the adverbs. For example, I’ve taken faithfully to attach to the entire VP execute the Office of President of the United States, but you could also make a case that it attaches just to the verb execute, and that this string then forms the VP with the Office of President of the United States. And as we were reminded in 2009, adverbs have some flexibility in where they can be placed in a sentence, so you could even experiment with diagramming faithfully execute the Office…; execute faithfully the Office…; and execute the Office … faithfully. Have fun!

Diagramming Interrogatives

A couple of months ago, Rentz and Lentz at the Bcomm Teacher Xchange blog were kind enough to include one of my posts in their list of resources for learning how to diagram sentences. However, they express their preference for Reed-Kellogg diagrams over tree diagrams:

This blog post illustrates the differences between the Reed-Kellogg diagram and tree diagram methods for diagramming sentences. I prefer the Reed-Kellogg method. I know linguists prefer tree diagrams for their precision and more nuanced representation of sentence structures, but I’m not a linguist. I just want a visually accessible way for students to look at sentences, and (at least for me) the left-right reading orientation of the Reed-Kellogg diagram presents sentence structures more clearly than the top-down reading orientation of the tree diagram.

I’ll respond to their two reasons. First, it’s true that if you’re a linguist, you probably prefer tree diagrams to Reed-Kellogg diagrams. It’s also true that if you’re a cat, you prefer meat to vegetables. But if you’re not a cat, that doesn’t mean you prefer vegetables to meat. Case in point: my son Doug, who is not a cat, yet still likes his pizza with pepperoni and bacon when he can get it, and will pick off any peppers or onions. Likewise, you don’t have to be a linguist to like tree diagrams (if you like diagrams at all). I respect Rentz and Lentz’s preference for Reed-Kellogg diagrams, but dispute their non-linguisthood as a valid reason for the preference.

As for the “left-right reading orientation of the Reed-Kellogg diagram”, I’m afraid I don’t understand. One of the advantages of tree diagrams is that they preserve the linear order of an utterance. Reed-Kellogg diagrams use a mixture of left-right and top-down orientations, and if you don’t know the original sentence that is being diagrammed, you can’t always get back to it by reading off a Reed-Kellogg diagram. If you don’t believe me, check out this Reed-Kellogg diagram of the opening sentence from the Declaration of Independence, and then compare it to this tree diagram of the same sentence (you’ll need to use the magnifying-glass icon). Both diagrams are big and unwieldy, but only the tree diagram lets you read back the original sentence in unwavering left-to-right order.

This willingness to undo a sentence’s linear order to get at its structure shows up especially in Reed-Kellogg diagrams of interrogatives. An interrogative like Do you like cats? in a Reed-Kellogg diagram is indistinguishable from the emphatic You do like cats!, because subject-auxiliary inversion (e.g. Do you) is ignored.

Also, wh elements are always left in situ in Reed-Kellogg diagrams. That is, a sentence like What did you see? is diagrammed as if it were the question Did you see what? — or more accurately, as if it were You did see what?, what with the undoing of the subject-auxiliary inversion in Did you. That would be the question you might ask someone if they said to you, “I did see it!” and you didn’t know what the heck they were talking about.

There are even diagrams in which the combination of undoing subject-auxiliary inversion and leaving all wh items in situ collapse even more sentences into one representation. Take the sentence

Brynn will say who stole the cookies.

In a Reed-Kellogg diagram, but it looks like this:

We already know this will be indistinguishable from Will Brynn say who stole the cookies?, but there’s more. This is also the diagram for the interrogative sentence

Who will Brynn say stole the cookies?

In English, the who in these sentences is placed at the front of whichever clause is a question. If the question is about who stole the cookies, the who stays at the front of stole the cookies. If the question is about whom Brynn will accuse, the who goes in front of will Brynn say. But when you diagram it in situ, you don’t know which sentence you’re dealing with, and the semantic difference is more than just whether you’re asking about or stating the same proposition. In a language like Chinese, where all wh items really are in situ, the ambiguity of this diagram would be excusable, because the actual sentence would be ambiguous, too — but we’re diagramming English, not Chinese.

In fact, the above diagram is even the same as the one for …who Brynn will say stole the cookies, but we can cut a little slack here, since this is a subordinate clause, not a complete sentence. A Reed-Kellogg diagram would have to connect situate this clause within a larger one; for example, Fenster knows who Brynn will say stole the cookies.

For comparison, here’s how Brynn will say who stole the cookies and Who will Brynn say stole the cookies? look in tree diagrams (click to embiggen):

It’s worth noting that only the first of these four English sentences can be read off the diagram left-to-right.

A couple of other reasons I prefer tree diagrams can be seen in the diagrams in this post. First, it’s easier to collapse tree diagrams into triangles to hide the details. In the Reed-Kellogg diagram, even though I wasn’t interested in the internal structure of the verb phrase stole the cookies, I had to diagram it out, right down to hanging the the underneath cookies. In the tree diagram, I just took it down to the level of VP and left that phrase in a triangle of its own. Second, tree diagrams let you diagram a phrase without insisting that you diagram the entire sentence it came from. If I wanted to diagram just the structure of the predicate stole the cookies, I could do that easily with a tree diagram, whereas a Reed-Kellogg diagram would look incomplete with a predicate on one side of the vertical bar and no subject on the other side.

I’m not saying that tree diagrams always have it over Reed-Kellogg ones. For some sentences, neither kind has an advantage, and for some, Reed-Kellogg might even have an advantage. For example, Reed-Kellogg diagrams do a better job than tree diagrams of showing the unity of phrasal verbs such as throw away when they wrap around a direct object. For many sentences, though, especially the kind that syntacticians think about and traditional grammarians tend to overlook, tree diagrams are the way to go.

Diagramming Adverbs

Having covered the basics of diagramming verb phrases (VPs) with intransitive, transitive, and linking verbs in my last post, I’ll turn now to adverbs and adverb phrases. I’ll use an intransitive verb to keep things simple: swim. The full sentence will be She swims very well, with the adverb phrase very well modifying swims. On the right is how it looks in a Reed-Kellogg diagram. As before, the subject is to the left of the vertical divider; the predicate is to the right. Well is attached to the horizontal bar underneath swims on a diagonal line to show that it modifies it. Since very modifies well, it is shown on a diagonal line, too, which turns 90 degrees to attach to the line for well. In general, modification is shown by having items on diagonal lines underneath the things they modify, with hooked diagonal lines to show submodification.

On the left is how we’d diagram this sentence using a tree diagram. Well, actually, it’s just a diagram of the VP swims very well, since I don’t care about teh subject right now. This is one advantage of tree diagrams over R-K diagrams, by the way. You can diagram whatever phrase you’re interested in, without committing yourself to diagramming a whole sentence, or adding material to turn it into a sentence. Anyway, let’s look at this VP from the bottom up. First off, swims the verb makes up a VP all by itself: She swims is a grammatical sentence. Moving on to the AdvP very well, notice that well is an adverb, and forms an AdvP all by itself. The adverb very is on a sister branch to the AdvP well, and together, very and well make … another AdvP. This is how modification is shown in a tree diagram: You have a node of category X, which branches into something of category M (the modifier) and something of category X again. In very well, X is AdvP, and the M category is Adv. Now look at the top node: It’s a VP. It branches into the AdvP very well, and another VP, swims. Once again, it’s the general pattern for showing modification.

So far, so good. R-K shows modification with slanted lines; tree diagrams show it with the [X [X M]] or [X [M X]] pattern. One problem with the R-K system, though, is that all adverbs are shown modifying not an entire VP, but strictly the verb. On the simple example we’ve done, it’s no big deal, but now consider the verb phrase write badly well. In the R-K diagram, the adverbs badly and well both appear on diagonal lines under the horizontal for write. Oh, and I’ll just put in an understood you for a subject, so we’ll have a whole sentence to work with. (This is how you represent subjectless imperatives in the R-K system.) The problem is that well doesn’t modify just write. It modifies the VP write badly, which means to do a good job of (deliberately) producing bad writing. But as represented in the R-K diagram, the phrase seems to refer to an impossible activity, of writing both well and badly at the same time.

Now here it is in a tree diagram. The entire VP branches into a smaller VP, write badly, which is modified by the AdvP well. The smaller VP write badly branches into a third VP, write, modified by the AdvP badly. This is the structure of a VP that makes sense, and as far as I know, there is no way of showing this in the R-K system. This deficiency is a reflection of the tendency of traditional grammars to focus on relations between individual parts of speech, instead of between larger chunks of syntax.

Last, let’s consider the VP behave badly. We’ll put it in a sentence so we can diagram it R-K style, and here it is. Nothing unusual here; the adverb badly goes on the diagonal line under behaved, right? The problem here is subtle. A modifier should be something that’s optional. Take away very well from She swims very well, and you’re left with She swims. We know she swims; we just don’t know if she does it well, poorly, fast, slow, or any other way. Take away badly from this example and you’re left with She behaved. Unlike the swim example, this doesn’t mean that she behaved in some manner or other. Used this way, behave has a more specific meaning of behaving well. It’s kind of like how He drinks doesn’t mean he drinks something or other and we don’t know what, but that he drinks alcoholic beverages. This is not the behavior of a modifier; it’s more like a verbal complement, like a direct object, or a predicate nominative, or a that clause after a verb like believe.

And why not? If NPs, AdjPs, and entire clauses can be complements to verbs, why can’t AdvPs? In fact, this could be represented pretty easily in the R-K system, by putting badly on the horizontal line after behaved, separated by a vertical line. It’s what we did with excites me in the last post, and it’s how other complements are represented, too (with the exception that the vertical line gets slanted backward for linking verbs). However, I’ve never seen this actually done in a R-K diagram, probably because R-K practitioners are used to thinking of an adverb exclusively as “a word that modifies a verb, an adjective, or another adverb.”

As a tree diagram, behaved badly would look as shown on the left. Notice that the VP node on top does not divide into an AdvP for badly and another VP for behaved. The verb behaved combines directly with badly to form a VP, just like the verb excites combined with me, and is combined with awesome in the last post. (More on the verb behave in this post.) This is how a word and its complement(s) are shown in these diagrams: They are sister nodes under a mother node that’s not the same category as either of the daughter nodes.

But wait? How do you know which daughter is the complement and which is the non-complement (i.e. “head”)? That’s for another post. The main thing here is that modification is shown by a mother and daughter node having the same label, while other relationships (such as complementation) are shown by having sister nodes under a mother node that has a different label from either daughter. And that modification is more accurately represented in tree diagrams than in R-K diagrams.

Diagramming Intransitive, Transitive, and Linking Verbs

Every now and then a reader will ask me why I don’t use Reed-Kellogg sentence diagrams, or what I think of them, or recommend them over the tree diagrams I use. So I’ve decided it’s time to begin a series of posts comparing the two ways of diagramming sentences — or in the case of tree diagrams, phrases. They’re not limited to just sentences, which is one reason I prefer them.

The most recent person to write to me about R-K diagrams is Martha Kolln, a retired English professor at Penn State, and author of several grammar books. She argued that R-K diagrams “clearly [show] the differences among transitive, intransitive, and linking verbs.” This seems like as good and limited a topic as any to start with, so here goes. I’ll start with intransitive verbs, with the sentence Syntax rocks. The R-K diagram of this sentence, shown to the right, has the subject to the left of the vertical line that crosses the main horizontal line. The verb rocks is to the right. That’s all there is to it, because rocks is intransitive, and doesn’t require any direct objects, predicate nominals, or any other kind of complement to complete the verb phrase.

Now let’s see how this looks using a tree diagram. Since the whole thing is a sentence, the root node of the (upside-down) tree is labeled S. To show that this S consists of a noun phrase (NP) subject and a verb phrase (VP), this S node divides into two branches. On the left is the NP node. This NP consists of just the noun syntax, so it has only one branch coming off of it, leading to the N node. The “leaf” of this branch is the noun syntax. The VP node also consists of just one word, the verb rocks. So a single branch comes off the VP node, and ends in a V node, just above the rocks leaf.

So far, the two systems are pretty similar, although the tree diagram explicitly gives information that the R-K diagram assumes you will have; i.e., the category and part-of-speech labels. This will be useful for when we diagram isolated phrases that aren’t part of a sentence.

Next up, a transitive verb, illustrated in the sentence Syntax excites me. As before, the subject is to the left of the vertical line that crosses the horizontal, and the predicate is to the right. This time the predicate consists of the verb excites, and its direct object, me. The direct object appears to the right of another vertical bar, which does not cross the main horizontal. This is the hallmark of complements of verbs other than linking verbs in the R-K system.

Now for Syntax excites me in a tree diagram. The NP subject is as before. This time, the VP node branches into the V node for excites, and another NP node for its direct object. Thus, both systems distinguish between intransitive and transitive verbs. The direct object NP consists of the pronoun me, so just a single branch ends in a Pronoun node followed by the leaf me.

Finally, let’s look at linking verbs, with the sentence Syntax is awesome, which contains the linking verb is and the predicate adjective awesome. The R-K diagram on the right is like the other two as far as separating the subject syntax from the predicate. R-K diagrams distinguish linking verbs from transitive verbs by having a slanted line instead of a straight line between the linking verb and its complement. The slanting line is supposed to indicate that the subject has the property named by the predicate adjective (or that the subject and predicate nominative refer to the same thing).

Represented as a tree diagram, Syntax is awesome has precisely the same branch structure as Syntax excites me. The differences are that the leaf labels excites and me are now is and awesome, respectively; and that the NP and Pronoun nodes under the VP node are now Adj(ective)P(hrase) and Adj, respectively. The identicalness of these trees means that this kind of diagram does not distinguish between transitive verbs and linking verbs: Both direct objects and predicate adjectives (and predicate nominals and predicative prepositional phrases) are represented as complements of a V — that is, as a sister node to a V under a VP mother node.

How can a tree diagram represent the information of the slanting line in a R-K diagram? From a syntactician’s point of view, that’s not the diagram’s job. That information isn’t about syntax, but semantics. For that matter, neither R-K diagrams nor tree diagrams show that (at least in more formal varieties of English) pronominal complements of transitive verbs are in accusative case (e.g., They criticized her), while pronominal complements of linking verbs are in nominative case (e.g. It was she). To do that, you need enriched diagrams that can show features like these, but that’s a whole nother story.

As a final note, I should note that both kinds of diagram will let you parse an ungrammatical sentence. You could easily diagram *Syntax excites awesome in either system. To rule it out, you need a set of rules for what kinds of diagrams you can make. In R-K, you need to have a rule that only linking verbs get to have the slanting line after them, and can be followed by an AdjP, and indeed, such rules are taught in the grammar books. For tree diagrams, you need something similar. But with tree diagrams, you have the option of putting some of that enrichment I mentioned above in the category labels, so that you can verify each branching is legit. Below I’ve re-presented all three sentences in this kind of diagram. Here, the labels S and NP stay the same, and Pronoun is simply NP (because pronouns act as NPs). VP has been replaced by NP\S; that is, a VP is something that looks to its left for an NP, and results in an S. Instead of a simple V category for any kind of verb, we have labels showing what kind of complement the verb has to take before it can form an NP\S. Intransitive verbs are NP\S already! Transitive verbs are (NP\S)/NP; that is, they look for an NP on the right before they become an NP\S looking for a subject. Linking verbs are (NP\S)/Pr, with Pr standing for Predicative. This category covers predicative NPs, AdjPs and P(repositional)Ps. The adjective awesome is given the category Pr, since it can act as a predicative adjective phrase. (A noun such as former, though, which can’t be predicative, wouldn’t have this category. Instead, it would be N/N: something that looks for a noun on its right, and results in a bigger noun.) Syntax, by the way, is an NP only because it’s a mass noun, which can act as an NP. A count noun such as sentence would have category N, and would have to combine with something with category NP/N, such as the, in order to form an NP.

The nice thing about this kind of presentation is that at each branching node, you can see the label in the denominator of one daughter node get “canceled out” by the label of its sister node. For example, at the S node, you can see the NP in the denominator of the daughter NP\S node get canceled out by the NP of the other daughter node.

Let’s Diagram a Sentence, Part 2

In my last post, I took this reader-suggested sentence:

It was neither more nor less than a pig, and she felt that it would be quite absurd for her to carry it any further.

and got as far as diagramming each of its component sentences as far as shown below:

Today I’ll continue with that first sentence.

Let’s Diagram a Sentence, Part 1

Michael Foy asked in a recent comment:

So how would you analyze these sentences in tree diagrams?

1. Alice was just beginning to think to herself, ‘Now, what am I to do with this creature when I get it home?’ when it grunted again, so violently, that she looked down into its face in some alarm.

2. This time there could be no mistake about it: it was neither more nor less than a pig, and she felt that it would be quite absurd for her to carry it any further.

Let’s give it a try. I’ll tackle It was neither more nor less than a pig, and she felt that it would be quite absurd for her to carry it any further. The first thing we can do is notice that it’s a compound sentence, composed of two independent clauses joined by an and. This is typical of phrases joined by and: If the component phrases are of category X, the phrase as a whole will have category X, too. (And if the component phrases have different categories, which does happen? I’m not even going there today.) There are a couple of ways we could diagram this, shown below, and linguists have yet to agree on which approach reflects better what’s going on in human language.

The diagram on the left says that a coordinate structure has three parts: the two things that you’re coordinating, plus the conjunction. The diagram on the right says that a coordinate structure naturally falls into two chunks, not three: The first thing you’re coordinating, and a chunk consisting of the conjunction and the other thing you’re coordinating. Where I’ve put a ? label, there is further contention over what kind of phrase this is. To simplify things, I’ll go with the approach on the left.

The next easiest thing to do is to divide each sentence into subject and predicate, or I’m labeling them here, noun phrase (NP) and verb phrase (VP). I’ll present these as two diagrams, and trust that you can graft them onto the one covering the whole sentence at your leisure. The first sentence:

And now the second one:

Notice that I’ve labeled the pronouns Pro for short. (Actually, this is a bad choice on my part, since syntacticians often use Pro to refer to missing pronouns, but I’m not going to redraw the diagrams now. I’ll just make it each have only one branch coming down from them; that means that these NPs consist of only one thing, namely a pronoun.

Things are going to get more complicated now, as we move down into the sentences. What are we going to do with the correlative conjunctions neither … nor? What about the “dummy it” in it would be quite absurd to carry it further? In the Reed-Kellogg system of diagramming sentences, the answer to that last question would be to recast the sentence as to carry it further would be quite absurd, and then use the infinitive as the subject. But in the tree-diagramming system I’m using here, we’d like to show the structure of the sentence without destroying the information about how it’s actually spoken or written out.

So stay tuned for the next post, where we’ll kick it down to the next level, so to speak!