Diagramming sentences

« previous post | next post »

Today's Frazz:

In fact, there are people whose (paid) job it is to "diagram" sentences — some examples of their output are here, here, here, here, and here. These results have engineering as well as scholarly and scientific applications.

But the Reed-Kellogg method of "diagramming" sentences has been intellectually obsolete for a hundred years. It's a shame and a scandal that no other mode of syntactic analysis has any grip on the popular imagination today.

In particular, there's an alternative standard which has been around, with minor modifications and additions, for more than 20 years; which has been applied on a large scale (millions of words) in published "treebanks" of languages as diverse as English, Greek, Chinese, and Arabic; and which is increasingly used for engineering, scientific, and humanistic applications, as documented in thousands of publications. Excellent tutorial materials have been developed for training people to use this standard.

And yet, if you were to ask faculty members in English departments and Schools of Education, I'd be surprised if one person in a hundred has ever even heard of this analytic standard — and I doubt that one in ten thousand would actually know anything substantive about it. So who is doing all the research, and writing all those thousands of scholarly, scientific and technical papers? Computational linguists and computer scientists.

There's something deeply wrong here, and plenty of blame to go around. The people who know about this stuff have done a dreadful job of public relations. The people who don't know, and should, are intellectually irresponsible in this as in many other ways.

Update — I should make it clear that the advantage I'm claiming for the "treebank" style is NOT that its analyses are superior to the long list of alternatives, starting with Reed-Kellogg diagrams. Most of the substance of the analyses would have been familiar to Otto Jespersen, for example. And the now-available explanatory materials are certainly not suitable for use by high-school or even university students.

My point is that this framework is stable and descriptively well-established, and has become a widely-accepted basis for computational and historical work. There are translations back and forth with a number of alternative representational formats, including dependency grammar, tree-adjoining grammar, categorial grammar, etc.; there could perfectly well in principle be translations to and from Reed-Kellogg diagrams, if those were somewhat better formalized. There are relatively good parsers — and better ones every year.

So if we accept the premise that some fraction of educated people ought to learn "grammar", in some sense of that word, then some version of the treebank framework is the obvious candidate for the kind of grammar that they should learn.



  1. Liz Weishaar said,

    April 14, 2013 @ 9:19 am

    I was very curious to find out why there's such a need for diagrammed sentences, but your blog doesn't seem to address this issue. And I clicked the "here" links in the first para, but then saw cover pages, if you will, of info that I couldn't access.

    It's Sunday morning and I'm only halfway through my tea, so maybe I'm being obtuse, but can you please clarify or amplify a bit?

    [(myl) If your question is "Why are engineers at places like IBM and Google interested in having access to large collections of accurately-parsed sentences?", the answer is that they believe that technologies from web search to information extraction to machine translation can be improved by accurate parsers, and that the way to get accurate parsers is to apply machine-learning methods to large treebanks.

    If your question is "Why do people who study the history of English (or French, or Icelandic, or ...) want large historical collections of parsed text?", I hope that the answer is obvious.

    If your question is "Why should intellectuals want to know anything about grammar?", there are two kinds of answer. The first type of answer is simple: for the same reason that they might want to know something about cosmology, or biology, or statistics, or history. The second type of answer is based on the traditional claim that an analytic understanding of how language works has practical value in areas like composition and language learning. This claim might be false -- certainly most of today's instructional practice assumes that it is -- but there's little real evidence either way, as far as I know.]

  2. Liz Weishaar said,

    April 14, 2013 @ 9:20 am

    You know, I misused "obtuse" in that post. I'm hiding in shame now.

  3. mike said,

    April 14, 2013 @ 9:25 am

    Are you aware of anyone who's attempted to adapt the Penn treebank markup for use in the curriculum for, say, middle schoolers or whoever the audience used to be for traditional diagramming? (Do you particularly recommend any of the tutorials you mention?)

    [(myl) I'm not aware of any such efforts. In my opinion, the right place to start would be in undergraduate-level linguistics courses, especially those with a computational flavor. ]

  4. John Roth said,

    April 14, 2013 @ 9:34 am

    I did a quick read of your reference, and I beg to demur. I don't think that representation would be at all satisfactory for a middle or high school approach. The (almost) binary tree representation in "A Student's Introduction to English Grammar," which I've seen in other contexts, including some K-12 materials, seems to be more suitable. Although it has defects, at least from my point of view, it has the benefit of a graphic diagrammatic representation that works for most cases (stranded PPs and various forms of internal linkage are exceptions) – but then, Reed-Kellog doesn't work for them either.

    [(myl) At the level that I'm talking about, there's not a great deal of difference. Both representations basically present surface constituent structures of a common-sensical kind, and most of the differences involve simple and systematic re-naming or local reconfigurations.

    The advantages of the (extended) treebank approach include its application to a dozen or so languages besides English, the existence of a very large body of annotated material, and connections to software like NLTK, as a result of which there are thousands of people already using these representations for all sorts of pedagogical and non-pedagogical purposes.

    But if the approach of "A Student's Introduction to English Grammar" were in widespread use, that would be terrific, and easily intertranslated.]

  5. Stephen Downes said,

    April 14, 2013 @ 9:58 am

    I think part of the objection is that the treebanks aren't really diagrams, properly so-called.

  6. Brian Hagerty said,

    April 14, 2013 @ 10:32 am

    Of course Reed-Kellogg diagrams stink, but it is simply not true that there are readily available tutorials for alternative methods. I have spent some time looking for a straightforward tutorial about how to do the standard linguistics-style tree diagrams (NP – VP etc.) and have failed. (Perhaps there's a book with a tutorial, but I haven't found one on the web.) I certainly haven't found any straightforward tutorials for the kind of analysis described in the "Penn Treebank" poster you linked to.

    The linguistics community shares responsibility for failing to promote an alternative to Reed-Kellogg diagrams. It's not fair to simply bash the method when there is no readily available, linguist-approved alternative system.

  7. Lucy Kemnitzer said,

    April 14, 2013 @ 10:37 am

    I would welcome an actually visual representation of the parts of a sentence that made sense — which the thing we do in schools doesn't, much — but the links you provided don't seem to lead to that kind of thing? I just saw large blocks of text?

    The point of diagramming sentences is to have a visual representation of the relationships among the words and phrases. It's supposed to make it easier for students to understand the roles that the words and phrases have in the whole, and it's supposed to make agreement and precedent easier to understand and use. But neither the old system nor blocks of text do that.

  8. Jan Freeman said,

    April 14, 2013 @ 10:45 am

    @Liz W: You didn't misuse "obtuse," but lots of people are now using it to mean "obscure" or "abstruse," so your fit of self-doubt is understandable.

  9. Jussi Piitulainen said,

    April 14, 2013 @ 10:46 am

    Liz, re why there's such a need for diagrammed sentences, assuming you refer to treebanks, the annotations make explicit and searchable some of the structure that people perceive in those sentences. A computer program can be written to count, for example, which verbs have their subjects and objects in what grammatical case how often when there is annotation that marks verbs as verbs and links the subjects and objects to the verb. Without such explicit annotation, the search is much harder to program.

    There are computer programs that produce such structure, but (1) the same question can be asked about the need for such programs, and (2) treebanks, in turn, help in the creation of such programs, so I think that the more fundamental answer is that the annotations make interesting aspects of the data explicit.

  10. joshua bowles said,

    April 14, 2013 @ 10:46 am

    @Liz Weishaar
    These treebanks were instrumental (and still are) in developing statistical models for languages. Such models can be (and probably are) used by many companies who need to process text.
    Even if a company or individual doesn't use formal language models built with treebank data the models and strategies built from having access to reliable treebank data (i.e., reliable diagrammed sentences) have un-arguably had an influence on the kinds of formal models used.

    Additionally, by "companies" I'm including the likes of Google, Microsoft, Apple, IBM… many of the products that have had a profound impact on our society include natural language processing (NLP)… and NLP is the field that employs formal/computational models of languages … and all such models have various ways of defining a formal standard for diagramming sentences.

  11. Nic Subtirelu said,

    April 14, 2013 @ 10:49 am

    It seems like the purposes for teaching syntactic analysis need to be better theorized, researched, and explicated in an accessible manner before anyone is going to adopt any system be it Penn Treebank or otherwise. At this point, most people arguing for grammar teaching in schools rely on unsubstantiated claims about how it will improve student writing. There are some elements of grammar (particularly those related to punctuation) where I think the ability to analyze syntax is useful (e.g. determining what's a dependent vs. independent clause or just a clause at all for that matter and thus punctuating accordingly). However, I think the English and Education departments referenced in this post rightly feel that these elements of orthographic accuracy are mere minutia next to the other priorities they have for students: helping students develop effective rhetorical strategies and writing processes as well as habits of critical reading and writing. If one could show how analysis of grammar could tie into awareness of how, on a micro-level, language is mobilized for specific purposes, then I think you'd have a stronger argument for incorporating grammar teaching into general education (and to be fair some scholars have started to do this particularly systemic-functional linguists and critical discourse analysts). Of course corpus and computational linguists have reasons for using Penn Treebank, but, useful as their work is, it's hardly a tragedy that every student is not educated in their coding systems.

    [(myl) The "grammar is irrelevant to writing" argument would be more convincing if current grammarless methods were doing a better job of teaching writing. As things stand, there isn't much real evidence for either position in this debate.

    My own feeling is that educated people ought to know some grammar for the same reasons that they ought to know some cosmology and some history and some biology. If grammar (and logic and rhetoric) help them to be better writers or orators, so much the better; but we don't teach cosmology because it will make people better at aiming spacecraft.]

  12. Steve said,

    April 14, 2013 @ 4:11 pm

    I agree with myl, understanding passive voice is at least as important as understanding cellular meiosis.

    More importantly, I think it's important to understand grammar and syntax are objects of scientific study. A lot of people don't know that, and it leaves them open to accepting unscientific arguments not based on evidence. It can also lead to making unscientific arguments not based on evidence, but that has lots of other causes as well.

    I think that if we wanted to encourage teachers to teach grammar and syntax according to current scholarly understanding of the subject (i.e., not teaching outdated or incorrect knowledge) then we'd have to do four things (at least in California):

    Have a linguistics CSET (standardized tests required to be taken and passed by prospective teachers) that reflects current scholarly understanding of grammar and syntax.
    Revise the current english standard to reflect grammar and syntax as it's currently understood, with periodic revisions and new knowledge becomes accepted by scholars on the subject.
    Adopt textbooks that reflect the new standard.
    Test the subject on the CSTs (tests taken by California students for the purposes of evaluating schools and teachers).

    This would require prospective teachers to learn the subject, teach it, make it easy for them to teach, and reward them for teaching it.

  13. AntC said,

    April 14, 2013 @ 5:35 pm

    As a Brit, I didn't come across the R-K diagrams until recent years, thanks to Language Log/A&L Daily. At UK "Grammar School" we used a Parts of Speech approach, with some parenthesising of sentences. Learning Latin was the main way I picked up syntax. My first experience of diagramming sentences was UnderGraduate Introductory Linguistics (Syntactic Structures/Aspects-influenced).
    I have to agree with @Stephen Downes that treebanks aren't really diagrams , going by the examples myl points to. OK, I can see that LISP-style expressions are isomorphic to trees, but they don't look like trees. Are there utilities that represent those expressions as trees, for school-level use?
    Brainerd seems too inviting for Caulfield's sense of mischief. Is Kellogg anything to do with the breakfast food company? Could Brainerd Kellogg be a Heavy Metal band?

  14. Keith M Ellis said,

    April 14, 2013 @ 6:52 pm

    "More importantly, I think it's important to understand grammar and syntax are objects of scientific study. A lot of people don't know that…"

    This is true, but I keep finding it startling to be reminded that many people think such things. For example, recently a governor said something to the effect that we need more scientists, not more anthropologists. This stopped me in my tracks; I wondered exactly what he thinks science is?

    The physical sciences, I'm sure. But speaking as someone whose primary cultural science affiliation has been with physics since childhood, I never thought of "science" as being limited to physics and the like. It's a rigorous, institutionalized, empirically-based study of the cosmos. Whether or not various social sciences are as mature as the natural sciences is beside the point — these disciplines, existing within the structures of contemporary science, are certainly "science".

    I'm somewhat at a loss to fully comprehend why people would be inclined to believe that linguistics or anthropology or whatever are prima facie not science because (as the reasoning goes) their objects aren't scientific. This isn't the prejudice between the "hard" and "soft" sciences; this is something else, something deeper.

  15. Carl said,

    April 14, 2013 @ 6:59 pm

    I really dislike this cartoon. The diagrams themselves aren't that important, but you'll never get good at a foreign language if you can't identify the parts of speech and their relationships in a sentence. The diagramming part of it is just an activity designed to help you internalize the structure.

    I teach a logic class with college students, and the lack of grammar knowledge students have today is a big hinderance for me. I tell them that categorical sentences have to be "All/some/no NP are (not) NP," but they don't know what noun phrases are, so they don't why "All my friends are out shopping" or "No one is eating cake" don't work. I wish they had been grilled on sentence diagrams as children!

    I can only imagine what pain and suffering foreign language teachers must feel.

  16. Nic Subtirelu said,

    April 14, 2013 @ 7:29 pm

    Well I think I've already made it quite clear that I'm not advocating a "grammar is irrelevant to writing" position. I concur with Steve that passive voice is a potentially useful concept for any English (or insert any other relevant language)-speaking citizen of the world to know about; indeed, it's mobilized for a number of textual or rhetorical purposes that can shed light on the text producer's organizational or ideological intentions (e.g. passive voice may be employed for purposes of new/old information management or for purposes of evading responsibility). Knowing what passive voice is and what its functions in a text might be are, in theory, useful learning objectives (for the reasons I just mentioned). However, convincing educators of this fact requires us to go beyond arguments of either (a) an 'educated person' should know it or (b) it's no more irrelevant than other things people are required to learn in school. Providing a solid rationale for incorporating grammar into school curricula requires us to consider carefully what a person might gain from this knowledge and having a clear answer to the inevitable "so what?" question. I acknowledge that there are engineering applications of syntax that have obvious value to us particularly in the realm of computers. I can definitely see an argument for making systems of grammatical analysis (like Penn Treebank) a basic requirement for students say of computer science (linguistics students learning syntactic analysis, I think, should go without saying, of course). However, when I consider the aims of general education courses, say of a basic writing class, I don't see a system of grammatical analysis that is silent about function as being useful to the purposes of those courses. Labeling parts of speech and parsing structure is perfectly useful for the computational purposes already mentioned by numerous comments here (and indeed many students, perhaps more than currently are being taught it, will want and need to learn). However, if we want to incorporate grammatical analysis into, for example, a 10th grade writing class or a first year composition course, then I think we would be best served with a grammatical analysis that pays close attention to the function of grammatical choices and the meaning-making capacities of text producers, e.g. systemic functional linguistics.

  17. Elijah Mayfield said,

    April 14, 2013 @ 8:32 pm

    I agree with everything you've said here, now that I've spent four years at CMU's Language Technologies Institute studying computational linguistics.

    However, I went to an undergraduate-only liberal arts college for my B.A., where there was no department of Linguistics and no computer scientist studying language technology. There was, however, a wonderful woman in the English department who taught a grammar course using sentence diagramming.

    As an independent study with her in my senior year cross-disciplinary thesis, I wrote a one-way, rule-based system which translated from Stanford dependency parses to Reed-Kellogg diagrams. The system bears no resemblance to anything I would do now (after many courses and thousands of hours studying NLP and machine learning), but it was good enough to be accepted to the ACL '09 Student Research Workshop:


    Overall, I enjoyed the experience of figuring out what the various formalisms meant, and needing to pick up enough grammar along the way to build a reasonable mapping.

  18. Mark F. said,

    April 14, 2013 @ 9:22 pm

    When I was a kid in the 70's, a few teachers taught (presumably R-K) diagramming, but I got the sense at the time that it was going out of fashion. I had it in 5th and maybe 6th grade, and I think that's it. Is it still commonly taught? I'm pretty sure my son has never seen a sentence diagram.

    My 8th grade English teacher mentioned at the time having gone to a continuing education seminar that was advocating a linguistics perspective and, in particular, the teaching of tree diagrams. Pretty much what MYL is advocating. Her response was something like "They say it's more scientific, but really it's just diagramming sentences." I didn't know enough to say "but tree diagrams are clearer and do a better job of expressing the relationships." So attempts to bring serious linguistic analysis into the teaching of grammar were being made by some people as long as 30+ years ago, but to some degree it was probably swimming against the tide of declining interest in teaching grammar at all.

    [(myl) With any system of grammatical analysis, in my opinion, the first crucial problem is a descriptive one: If you pick up a random text and go through it sentence by sentence, is there a well-defined answer to the question "What the analysis of this specific sentence?" This is quite different from the provision of analyses for a few hundred grammar-book examples. And I don't think that anyone has ever worked this out for the Reed-Kellogg diagrams. Certainly no one has ever worked it out for (for example) "construction grammar".

    In most practical contexts, including pedagogical ones, the next crucial question is stability. How often does the answer to the first question change? If the answer is that everything has to be re-done every decade or so, in ways that are not algorithmically specifiable, then the practical effort to develop instructional materials and train teachers becomes worthless before it's even finished. This is what happened to the considerable work done in the 1960s to use "transformational grammar" in the schools.

    The treebank representations (there are several essentially equivalent ways to present the same information) win on both of these dimensions. By design, they are inter-subjectively consistent ways to analyze arbitrary texts; and by virtue of the effort sunk in creating them, they change only gradually, and in carefully-specified ways (for instance by filling in properties or relationships that were previously left vague).]

  19. Keith M Ellis said,

    April 14, 2013 @ 9:55 pm

    "I teach a logic class with college students, and the lack of grammar knowledge students have today is a big hinderance for me. I tell them that categorical sentences have to be "All/some/no NP are (not) NP," but they don't know what noun phrases are, so they don't why "All my friends are out shopping" or "No one is eating cake" don't work. I wish they had been grilled on sentence diagrams as children!"


    Those examples are easily converted to standard categorical statements — e.g., All people who are my friends are people who are shopping, No person is a person who is eating cake.

    I've never been in a classroom where the grammar of categorical statements was discussed explicitly because it's never been necessary. You don't need students to have a formal grasp of what a noun phrase is, and certainly not diagramming, for them to understand why you need to convert "All my friends are out shopping" to "All [people who are my friends] are [people who are shopping]", with perhaps the brackets I included for clarification. It becomes obvious both from examples of such restatements and from attempts to talk about the subjects and predicates.

    What's not clear to me from your comment is whether you're arguing that your students represent an opposing experience than my own (they don't understand the restatement or why it is necessary for clarification and for identifying statements which aren't implicitly categorical) or that you're expecting them to avoid such informal assertions and generate categorical statements from scratch.

    I'd be very surprised at the former. That's why, when reading your comment, I assumed the latter — which I think is a pedagogical mistake given the context of teaching categorical logic as compared to teaching formal logic. The former is all about everyday language and the implicit reasoning within it. And arguably teaching categorical logic is a mistake as it can hinder students with formal logic — therefore, whatever utility teaching categorical logic has, it almost certainly lies within its relative informality. If you're going to insist on formalism, you might better be teaching actual formal logic.

  20. Dominik Lukes said,

    April 15, 2013 @ 12:23 am

    As someone who was taught a diagramming system designed by a Prague School trained structuralist, I have the experience of thinking 'what's the big deal' when introduced to Syntactic Structures. So perhaps the American system is due an upgrade, if you need one.

    [(myl) Thanks to Jan Hajič and others, the Prague style of grammatical analysis has been formalized and extended, so that it works as a description of English as well as of Czech -- see e.g. the Prague Dependency Treebank, which is built on top of the Penn Treebank. As you can learn from the documentation of that work, it was not a trivial enterprise to specify the system fully enough to make it possible to apply it in an inter-subjectively consistent way.]

    But I seriously doubt that it is doing anybody any good. It certainly had zero impact on most of the Czech cohort and Czechs are far from the World's best stylists (if only because such imaginary league table is non-sensical). You might argue that Czechs produce among the world's best linguists (per capita) but there are probably several dozen of those so its school-based diagramming obsession seems a rather wasteful enterprise.

    Why not show kids how to use a corpus, help understand collocability and principles of construction grammar?

    [(myl) Training in the use of a corpus would be at the core of the kind of instruction that I'm suggesting. But "construction grammar" would be an extremely poor choice for the formalism used. There's no agreement about what a "construction" is, or how to decide what the "constructions" in a given passage actually are; and no significant corpus of any language has ever been described in a "construction grammar" framework. So as an approach to teaching basic grammatical analysis in a corpus-based context, "construction grammar" would at present be useless.]

    Something that actually describes language at work, rather than language stripped of all its complexity to fit into an engineering model.

    [(myl) This is the opposite of the true situation, since there is no "construction grammar" description that can be applied to "language at work". There's value in the basic construction-grammar idea that some complex structures have properties that are not entirely predictable from the composition of their parts; but no one has ever cashed that idea out in a way that can actually be used to describe arbitrary real-world texts in English or any other language.]

    Let them teach trees in maths and science – they seem to tell only a small part of the utility story, anyway, how about stochastic processing – should Markov models be taught in English classes?

    [(myl) This is just confused, I'm afraid. Trees and labelled bracketings are just alternative notations for the same thing; dependency structures and other more general graph structures allow other non-hierarchical relationships to be represented, which is sometimes necessary, especially in languages with free word order; "stochastic processing" (whether generative or discriminative) can be applied to any sort of grammatical analysis, markovian or otherwise; ... ]

  21. Jason said,

    April 15, 2013 @ 1:02 am

    The diagrams themselves aren't that important, but you'll never get good at a foreign language if you can't identify the parts of speech and their relationships in a sentence

    I agree, but unfortunately the current school of thought is that you don't try to teach or use grammar in explicating a foreign language. I've studied French, Japanese and Melanesian Pidjin, and I've never taken a single course where there was anything more than a cursory emphasis on grammar, let alone the old-school course where you diagrammed sentences and identified the partitive article or the passé composé. It's all instrumental, hear and repeat, follow the pattern without really understanding it type stuff.

  22. Steve said,

    April 15, 2013 @ 1:08 am

    I'm somewhat at a loss to fully comprehend why people would be inclined to believe that linguistics or anthropology or whatever are prima facie not science because (as the reasoning goes) their objects aren't scientific. This isn't the prejudice between the "hard" and "soft" sciences; this is something else, something deeper.

    With regards to linguistics specifically, I think a lot of the problem is that it's in a transition period similar to what physics went through. Aristotle was once an authority on physics, and it took a lot of people doing a lot of scientific study to overturn him. Now, no one would ever think to cite him as an authority on motion. But centuries ago? Going against him would be political suicide.

    I think the same thing is going on with linguistics. E.B. White and Orwell are cited as authorities on grammar, despite the mountains of evidence contradicting them. Over time, their hold on the subject will fade. That probably can't some soon enough for Professor Pullum but, until then people will continue to site them.

    One reason why people may be slow to discard their advice, wrong as it is, is because language is a human construct. To scientifically study it, you need to study how it is used. Saying anything useful about language probably means that it has to be used by a lot of people.

    So linguists comes to the conclusion that "decimate" means, in many contexts, "total devastation" because that's how a lot of people use it.

    But then you have some people (peevers) who say "decimate" can't mean that because back in some language that isn't spoken anymore "decimate" mean to reduce by one tenth. That argument that, scientifically speaking, that's not the only meaning of "decimate" because lots of people use it to mean "total devastation" doesn't hold water for them. Facts are facts, just because people use a word to mean one thing doesn't make them right. To peevers, saying that is like saying evolution isn't real. Denying a fact doesn't make it not a fact.

    So elitism is a big barrier to change.

  23. Carl said,

    April 15, 2013 @ 1:46 am

    "Those examples are easily converted to standard categorical statements — e.g., All people who are my friends are people who are shopping, No person is a person who is eating cake."

    Yes, they are. My students don't understand how to do that though. :-( I show them the non-NP versions of the sentences and the NP versions, and they don't know why or how I've changed the sentences.

    Part of the issue is that I am in Hawaii, where students tend to have a weaker grasp of English. But that's compounded by the fact that grammar education here is especially weak. I really wish these students had been made to diagram sentences approximately ten years before I have them.

  24. Carl said,

    April 15, 2013 @ 1:55 am

    "It's all instrumental, hear and repeat, follow the pattern without really understanding it type stuff."

    I suppose the current approach makes sense if you're trying to train people to *speak* a foreign language. For that, since there is no source text to translate, you just need to output words in the right order. The further away they can get you from thinking in your native language, the better. But for translating things, a deep knowledge of grammar is essential. You have to be able to parse the sentence in both languages to get a good match. But nowadays there's less emphasis on written translation skills for whatever reason.

    [(myl) For an increasingly large (or at least increasingly relevant) segment of the population, formal standard English is essentially a foreign language, and learning to write in formal standard English is essentially an exercise in written translation.]

  25. Jason said,

    April 15, 2013 @ 1:57 am


    I really wish these students had been made to diagram sentences approximately ten years before I have them.

    I think though this might just be because it would be easier to teach them the formal language of logic if they've already had experience with formal language analysis of any kind, rather than that Understanding Grammar per se helps with understanding formal logic. It's the practice with formalism, rather than the unlikely linkages between the two formalisms, that is the transferrable skill here.

  26. Arekahanara said,

    April 15, 2013 @ 2:50 am

    I agree with myl, understanding passive voice is at least as important as understanding cellular meiosis.

    Tena tatou,
    I'm not sure what Steve is implying here. If he is minimising the importance of the passive he is perhaps displaying a monolingual english language view of the world. Te Reo Maori (New Zealand Maori) uses the passive voice in 80 % of spoken and language. It is not used to conceal the actor as has been suggested, but rather it focuses on the action rather the actor. As a fluent second language speaker of Maori for the past 30 years, I ffind it fascinating how grammar and voice are intricately interwoven with culture. I say that understanding cellular meiosis is very important.

  27. Keith M Ellis said,

    April 15, 2013 @ 3:32 am

    "Yes, they are. My students don't understand how to do that though. :-( I show them the non-NP versions of the sentences and the NP versions, and they don't know why or how I've changed the sentences."

    Huh. This just isn't my own experience, and mine was with a variety of students in varying educational contexts, from a large university to a community college to a liberal arts college. What you're describing was a non-problem; while, in contrast, a number of other things in categorical logic presented serious difficulties to a large portion of students.

    Anyway, to the larger discussion I'll add one data point: I had one English class in high school in the SW US around 1981 where we did a lot of sentence diagramming. It may have even then been largely falling by the wayside in secondary education, but it was still present.

    I absolutely hated it. Hated it.

    Per the discussion about linguistics and science and all that — I'm certain that I would have been deeply fascinated if I'd been taught grammar as a means of understanding how language works. As, indeed, I later was fascinated with grammar when I learned classic greek. But the way it was taught in that high school English class seemed to epitomize everything I hate most about a certain kind of pedagogy. It wasn't in the least about writing better. It wasn't about understanding language. It was taught as an almost arbitrary, meaningless formalism, lacking all context.

  28. peter said,

    April 15, 2013 @ 3:28 pm

    Sentence diagramming strikes me as a very odd activity to be doing at all, if you accept the theory that brain functions in most people are apportioned by hemisphere. Processing text and processing images are, for most pople, functions apportioned to different hemispheres of the brain. Thus, those students whose text-processing hemisphere (typically, the left hemisphere in right-handed people, and the right hemisphere in left-handed people) is dominant will find the task of diagramming text difficult, while those students whose image-processing hemisphere (typically, the right hemisphere in right-handed people) is dominant will find the task of diagramming text difficult. Is this why most students find the activity hard to do?

  29. Chris C. said,

    April 15, 2013 @ 6:07 pm

    @Peter — I think the brain hemisphere thing is rather overstated. Students have a hard time diagramming sentences to the same degree that they have a hard time understanding grammar itself, I think. We parse natural language without thinking about it. Having to explicitly parse a sentence is, for many students, akin to the centipede thinking about how he coordinates all his legs.

  30. MikeE said,

    April 16, 2013 @ 7:22 am

    Lucy Kemnitzer was asking about a visual representation, and AntC mentioned what he learned as a Brit.

    Whatever happened to "box analysis", which is what I learned in England at the age of about 11. This was pre-Chomsky, but – from memory – the diagrams looked generally very much like the tree diagrams in "Aspects …" when it came out . . . except that there were boxes.

    I wonder if it is the idea of _any_ (hierarchical) analysis of sentences (or anything else) that is really important as a child. It seems so obvious now, but I'm not sure it was at the age of 11.

  31. ThomasH said,

    April 16, 2013 @ 11:24 am

    Can the alternative be taught to seventh graders? What I was taught was fun.

  32. Warren Oliver said,

    April 16, 2013 @ 11:53 am

    For a web-based attempt to teach English grammar in a modern way at undergraduate level, see
    Internet Grammar of English from University College London,

    There are also Apple and Android apps available. The sections on phrases, clauses and sentences do cover constituent structure, although they don't provide explicit visual representations of trees.

  33. Ray Dillinger said,

    April 18, 2013 @ 3:59 pm

    FWIW, I remember a really excellent honors English class in high school (1980s, Midwest USA) where we did various different *kinds* of diagramming. We did the Reed-Kellogg classic that we had done in junior high, then we did clause-based tree diagramming, then we did reduction tables with a Chomsky grammar, then we did clause labeling with parens and tags, then we did element linkages (the formalism that was later embodied/automated in Link Grammars). Each time we covered something new, we compared and contrasted it to the other methods we'd covered.

    Turned out most of them were effectively equivalent; the Reed-Kellogg technique didn't preserve sentence word ordering and the linkage technique left some relationships implicit that were explicit in other models; but the rest of them were effectively ways of representing a tree structure. We spent 18 weeks on that, then for the second half of the school year the class was reading short stories and poetry and doing analysis of the author's technique (including grammar but not limited to it) and how the techniques used contributed to communicating the ideas the stories were built around.

    It is a trip to study grammar hard for 18 weeks and then get ee cummings and Ken Kesey thrown at you – figure out what the relationship of their work to grammar is and whether it worked or not.

RSS feed for comments on this post · TrackBack URI

Leave a Comment