Unnaturalness in experimental design

I read a language related article today where the researchers presented their subjects sentences with the words making up the sentences being presented serially on a screen. Their goal was to determine if brain responses correlated with the presentation of words in a sentence carrying more information according to several probabilistic models of language. The material that they presented to subjects seemed basically like a slowed down version of this:

The authors did find the effect they expected – words bearing more information lead to a certain kind of increased brain activity, but they also made some more far reaching suggestions based on the probabilistic models they used.

Two of the probabilistic models the researchers employed didn’t make use of hierarchical structure. The predictions made by these models correlated better with the brain responses they found. Based on this finding, they state in the abstract: “These findings suggest that different information measures quantify cognitively different processes and that readers do not make use of a sentence’s hierarchical structure for generating expectations about the upcoming word.”

I understand that the researchers needed to present words serially in order to isolate the variable they were considering, namely higher information bearing words. But, how natural of a task is this really to make assumptions about something beyond the word like the role of hierarchical structure in language? In reading, our eyes don’t typically take in just one word at a time. At least, I know I don’t read like that.

For me, the brain responses in and of themselves are interesting. For instance, some corpus approaches to linguistics look at co-occurrence to see which words occur more often together than they should based on chance alone. This is referred to by some as mutual information. But, what is really interesting in these brain responses is that the words in the sentence that don’t normally cluster with the others receive the most significant brain responses. In terms of information theory, these words carry more information in the sentence because they are less predictable. So, perhaps it’s not just correlations, but also negative correlations that we need to pay attention to in probabilistic approaches to language when we are working at the level of the sentence. As interesting as that is, I’m not sure it’s necessary to reach farther than that to talk about hierarchical structure when it seems like the task the subjects performed was pretty unnatural.

In which Morrissey responds to Kris’ question on the utility of linguistics

Kris asks: what is linguistics good for? Here’s what Morrissey has to say:

I’m not quite as pessimistic as Morrissey. I Kris’ question has layers to it. I believe those who engage in exegesis (a group to which I no longer belong) benefit from using resources informed by modern linguistic theory. I believe they would benefit from even more resources of this kind being available.

Moving to the next level, though, do I think those who engage in exegesis would benefit from studying the linguistic theory that may be behind the resources themselves? I think some knowledge is necessary for the sheer reason that it helps to understand one’s tools. But just how much knowledge of linguistics a exegete should seek, I’m not sure.

At any rate, if you have thoughts on the subject, be sure to visit Kris’ post and chat with him over there.

Is Cognitive Linguistics growing “dominant”?

Short answer: I don’t think so. Or at least, I’m not sure. For the backstory and my reasoning for saying so read on.

Today I read a review of Constantine Campbell’s Advances in the Studies of Greek by Nicholas Ellis and Mike Aubrey over at Themelios. Lest I give any of those who know me any misconceptions, I didn’t do so because I care at all about GrEEK. I did so because Mike is a friend and an outstanding linguist. Any time he writes or posts something about linguistics online I try to take a look. Sadly, I think I mostly comment when I disagree with something so I can engage in longer discussions about linguistics (it’s lonely being a linguist in South Louisiana sometimes).

With that said, I had a quibble with the review, really over mostly one word. Mike may disagree with me in the comments here or elsewhere. But, that’s okay because I promise in advance to buy him a beer next time I’m in Bellingham (Kris may disagree with me too because on a Facebook thread he expressed similar sentiments to those I’m about to quibble with concerning the overlooking of cognitive linguistics. Kris, however, doesn’t drink beer so I’m not sure how we’ll get past this). Nicholas and Mike state:

Campbell seems to have attributed to Chomsky the definition of deep structure as argued by one of Chomsky’s opponents, Charles Fillmore in the so-called “Linguistic Wars.” Fillmore (e.g. his article, “The Case for Case,” in Universals in Linguistic Theory, ed. E. Bach and R. Harms [New York: Holt, Rinehalt and Winston]) and those with him eventually developed their own ideas into what is cognitive linguistics today, a set of principles and frameworks that are growing more and more dominant in the field at large.

A few lines later they indicate that Campbell gives short shrift to cognitive linguistics in his book and that this “is disappointing.”

But, is cognitive linguistics really growing more “dominant” in the field of linguistics at large? Here is an excerpt from Vyvyan Evan’s The Language Myth (a book I don’t highly recommend, but here it is any way):

Alas, this is not the case. The views that I classify as myths are presented as established fact in many of the linguistics textbooks currently in use in many of the stellar universities throughout the English-speaking world. I was trained using these textbooks, and they are still compulsory reading for today’s undergraduate and graduate students – tomorrow’s researchers, educators and language professionals – even at the university where I teach and work. University students are regularly told that there is a Universal Grammar, that language is innate, that language is incommensurable with non-human communication systems, and that all languages are essentially English-like. For instance, the world’s best-selling university textbook on language is An Introduction to Language, written by Professor Victoria Fromkin and colleagues. This book, now in its tenth revised edition, proclaims the following in its very first chapter:

This business is just what the linguist attempts – to find out the laws of a language, and the laws of all languages. Those laws that pertain to all human languages, representing the universal properties of language, constitute a Universal Grammar … To discover the nature of this Universal Grammar whose principles characterize all human languages is a major aim of linguistic theory… the more we investigate this question, the more evidence accumulates to support Chomsky’s view that there is a universal grammar that is part of the human biologically endowed language faculty.

A recently published textbook introduction to the English language, The Structure of Modern English, by Professor Laurel Brinton, makes the following claims in its introductory chapter:

Language is rule-governed, creative, universal, innate and learned, all at the same time … A more general set of constraints on language is known as language universals. These are features of language that are not language specific … Inherent in the notion of universals is the belief that language is innate, that we are born with an inborn capacity for language acquisition.

As we shall see, the claims made in both these representative textbooks are wrong – they fly in the face of, now, several decades of evidence-based research …

He goes on to bemoan the fact that Steven Pinker is the most widely known popular level linguistics writer in the world.

If all of the talk of Universal Grammar has thrown anyone off the trail a bit, this is a concept associated with Chomskyan, or generative, linguistics. So, here one of the better known practitioners of cognitive linguistics appears to be asserting the dominance of generative linguistics in undergraduate and graduate level classrooms and in the public at large. What I gather from Evans is that he is arguing throughout his book: even though the preponderance of evidence argues against generative linguistics, generative linguistics is still the dominant linguistic framework in the field of linguistics at large and generative linguistics needs to be overthrown.

Evans could, of course, be wrong about this. In fact, I think he probably overstates his case in many parts of his book since much of what I see is collaboration in my own area of work. I’m thinking of cross-talk between WordNet, a project founded by George Miller who is often mentioned as one of the co-founders of the “cognitive revolution” with Chomsky and who offers gushing praise of Chomsky in places like the book Language and Thought, and FrameNet, a project founded by the same Charles Fillmore that Nicholas and Mike mention as an opponent of Chomsky. All that aside, I think Evans’ perspective does at least give me pause about claiming that cognitive linguistics is growing “dominant.” Maybe more “prominent”? But, I’m not sure even that cognitive linguistics is “dominant” or “prominent” such that I think someone approaching linguistics from the outside would necessarily know very much about it.

But again, this is mostly a matter of word choice. At any rate, I’d be interested to hear what Mike and other interested linguists might think. At least my question would be, if cognitive linguistics is becoming “dominant,” how is it asserting that dominance?

According to Mihaly Csikszentmihalyi's book Creativity, Albert Einstein wore the same sweater and trousers every day on his walks. If it takes an average of two minutes per day to choose which clothes to wear, that would have freed up 12 hours per year for Einstein to do more Einstein-y kinds of things.

Kris Lyle on qadosh in TDOT

My friend and recent co-author Kris has written a blog post on the Hebrew word קדש in the Theological Dictionary of the Old Testament in the light of its upcoming release within Logos Bible Software.

"Dave, stop. Stop, will you? Stop, Dave. Will you stop?" So the supercomputer HAL pleads with the implacable astronaut Dave Bowman in a famous and weirdly poignant scene toward the end of Stanley Kubrick's 2001: A Space Odyssey. Bowman, having nearly been sent to a deep-space death by the malfunctioning machine, is calmly, coldly disconnecting the memory circuits that control its artificial brain. "Dave, my mind is going," HAL says, forlornly. "I can feel it. I can feel it."

I can feel it too. Over the last few years I've had an uncomfortable sense that someone, or something, has been tinkering with my brain, remapping the neural circuitry, reprogramming the memory. My mind isn't going--so far as I can tell--but it's changing. I'm not thinking the way I used to think. I feel it most strongly when I'm reading. I used to find it easy to immerse myself in a book or a lengthy article. My mind would get caught up in the twists of the narrative or the turns of the argument, and I'd spend hours strolling through long stretches of prose. That's rarely the case anymore. Now my concentration starts to drift after a page or two. I get fidgety, lose the thread, being looking for something else to do. I feel like I'm always dragging my wayward brain back to the text. The deep reading that used to come naturally has become a struggle.

~ Nicholas Carr, The Shallows (pp. 1-2) ~

Searching Semantic Roles is Awesome in Logos 6 #Logos6

Last week I wrote a post about exploring meaning using case frames in Logos 6. I plan to come back to that topic with more practical examples in the near future, but I also wanted to introduce you to the semantic roles data on its own terms by taking a look at how these roles have enhanced clause search. By the end of this post, I think you will agree with me that this dataset is, quite frankly, awesome.

First, a brief aside: the case frame data that I talked about in the last post is made up of the semantic roles data. The important point here is that you can approach this data by looking at case frames as units of analysis or by individual semantic roles as units of analysis. If that doesn’t make sense at this stage, it’s not so important, but it may help to know that the case frames and semantic roles are interrelated.

Now I will inductively work through an example that I think will demonstrate just how exciting this new feature is.  Imagine for a moment that we want to find all of the places in the Bible that talk about Jesus being baptized. In the past, we might have tried to accomplish this by using clause search and the grammatical categories of ‘subject’ and ‘object.’

First, we might think: John the Baptist (subject) baptized Jesus (object). So, our first search might be something like “object:Jesus verb-lemma:βαπτίζω.”  Let’s check out the results of this search:

jesus object baptize


What’s the issue? … We forgot to take into account the passive voice. Maybe all of the cases of Jesus being baptized are more like: Jesus (subject) was baptized by John.  Our next search might end up as “subject:Jesus verb-lemma:βαπτίζω.” Here’s what we would find:
jesus subject baptize

This is obviously better; however, we’ve still got some noise in our search. We’re seeing instances that mention Jesus actually doing the baptizing. Is there a better way to do this?

If we think briefly about the semantic roles for βαπτιζω, it is possible to find a better way. A look at the semantic roles for βαπτιζω tells us that the person being baptized in each case is a “Theme.” With that little bit of knowledge, it is possible to search by meaning (i.e., semantics), as opposed to grammatical categories like subject and object that don’t always overlap completely with meaning.

If the previous paragraph was too much to parse, don’t worry about it because I think the next example search will make it easier to understand. What if we were to try “theme:Jesus verb-lemma:βαπτίζω”? Here’s what we would find:

jesus theme baptize

Jackpot! Take a look at the results. Now we’re seeing exactly what we want. But wait … there’s more. What if on the other hand, we wanted to find all of the places that mention Jesus doing the baptizing (for example, baptizing with the Holy Spirit)?  Another brief reflection on the semantic roles of βαπτιζω would tell us that the person doing the baptizing is an “Agent.” Maybe we could try “agent:Jesus verb-lemma:βαπτίζω”:

jesus agent baptize

Bingo! Our ability to search the Bible becomes so much more powerful when we are searching by meaning as opposed to searching by grammatical categories.  I hope that this example has demonstrated that it is worth the time to get to know the semantic roles data because it could save you a lot of time in the long run.

So, go buy Logos 6 if you haven’t already! Then feel free to ask any questions about semantic roles here or in the Logos forums. Or, drop a cool example of how you’re using this tool in the comments below.

Explore Meaning with Case-frames in Logos 6 (Or, what I’ve worked on for the last year) #Logos6

Rick has already posted some of his favorite features in Logos 6. So, I thought I’d take some time to post on my favorite feature in Logos 6 while also mimicking his post title. Incidentally, I’m biased because I worked on the Hebrew data for this project. Paul Danove (whose work really inspired this feature) provided initial Greek data, and Mike Aubrey continued that work.

Case-frames provide a new way of exploring meaning within Logos 6. It may not be apparent on first glance how they do this. Here I will work from an English example to an original language example to demonstrate how this works.

Consider an English verb like “return.” This verb can have several different meanings as in the following sentences:

  1. He returned home.
  2. He returned the donkey to its pen.

In the first case, we might paraphrase “return” as “go back”: “He went back home.” In the second, we might somewhat poorly paraphrase as “bring back” (perhaps this isn’t the only possible interpretation, but this is only an example): “He brought the donkey back to its pen.”

The difference in these two meanings of “return” is reflected in the number of “arguments” that the verb takes in each example. I don’t want to get bogged down what an argument is, but it is linguistic unit (word, phrase, clause) that is required to determine the meaning of a verb (or some other kind of predicate). In the first example above, the verb has two arguments: someone (an Agent) who is going back somewhere and the place (Goal) to which the person is returning. The second example has three arguments: the person (an Agent) returning something, the object (a Theme) that is being returned, and the place (a Goal) to which the object is being returned.  As hinted at above, these arguments are labeled with roles like “Agent,” “Theme,” and “Goal” in a case-frame analysis. A fuller discussion of each one of these roles can be found in the glossary of semantic roles in Logos 6.

An analysis of the arguments in the two examples above would look something like this:

  1. [Agent He] returned [Goal  home].
  2. [Agent He] returned [Theme the donkey] [Goal to its pen].

The meaning of the verb “return” in the first example is reflected in the pattern Agent – Goal and the second meaning is reflected in the pattern Agent– Theme– Goal. Thus, the meaning of “return” is related to the construction in which it is involved.

Of course, every language has important differences. But, these underlying patterns also occur in a language like Hebrew (though in Hebrew the differences between the two meanings discussed above are also somewhat reflected morphologically – qal vs. hiphil –  for those who know their Hebrew).

The Biblical Hebrew word most often rendered “return” by English translations is שׁוב (šwb). To see how we can explore the meanings of this verb by looking at its patterns like we did for the English examples the first thing we want to do is run a Bible Word Study. So open up the Bible Word Study guide:


Since this isn’t really a tutorial on how to work with Hebrew in Logos I’ll give you the easy version of how to search for our Hebrew word. In the top left search box we can enter “h:shwb” then select the appropriate Hebrew word from the dropdown:


Once the guide has finished gathering all of the necessary information make sure the “case-frames” section is expanded by clicking the arrow next to “case-frames”:


And now we can look for the patterns that we’ve discussed so far, which I’ve highlighted already. We can open up the search results for each one of these patterns by clicking on the appropriate section of the case-frame wheel.  For the Agent – Goal pattern we see the following:


We notice right off that case-frames can either be abstract (“the deeds of a man’s hands will return to him”) or concrete, and we see concrete examples of the Agent– Goal pattern, such as Gen 8:9 “… [Agent she] returned [Goal to him] … “ and Gen 15:16 “… [Agent they] will return [Goal here] …” Again, these examples approximate the English example of “He returned home” that we looked at above.

When we look closer at the Agent– Theme– Goal pattern we see the following:


One interesting matter to note here is that the translation generally doesn’t use the word “return,” so if we were just looking at an English translation we may have no idea that these usages were related to the same verb as the Agent – Goal pattern. Again, we have some abstract and some concrete examples. Concrete examples occur in places like Gen 29:3 “[Agent (they)] put the stone back in (i.e., return the stone to) its place on the mouth of the well” and Gen 28:15 “[Agent (I)] will bring you back (i.e., return you) to this land.” Again, these meanings of שׁוב (šwb) approximate the meaning we discussed in the second example above: “He returned the donkey to its pen.”

To summarize, we can see here how different meanings of verbs can involve different patterns of semantic roles. We particularly looked at the English example of “return” and how to look at similar examples in a language like Hebrew using the case-frames analysis in Logos 6. This was only an introduction using a common English example, but there is a vast amount of semantic information to explore here. And I’m pretty excited about that. So, go buy Logos 6! Then, explore the new case-frames feature and drop a comment below with a favorite example. And if you have any question about this feature or any other, please feel free to ask. If I know the answer, I’ll help the best way I can. And if I don’t know they answer, I’ll try to point you in the right direction.