1. "CSI" = the entire Contextual Semantic Investigation curriculum, which should spell out everything that the teacher and students should do.
  2. "CVA" = the specific algorithm that is based on Cassie's program.
  3. X = the unknown/unfamiliar/hard word
  4. S(X) = the sentence containing X
  5. C(X) = the larger context surrounding X (of indeterminate size); the "current" context.
  6. H = a current hypothesis about a meaning for X.

Overall curriculum outline:

  1. Teacher models CSI
  2. Teacher models CSI with student participation
  3. Student(s) model(s) CSI with teacher assistance
  4. Students do CSI in small groups, working jointly
  5. Student does CSI on own.

The "basic algorithm":

  1. Become aware of X and of the need to understand X.
  2. Repeat the following until you have a plausible meaning for X in C(X):
    1. Generate H
    2. Test H

Elaboration of how to test a hypothesis:

At any time, if you can make a hypothesis, test it:
  1. Replace all occurrences of X by H (your definition).
    • S(X) → S(H).

  2. Does the sentence S(H) make sense?
    • If so, proceed with your reading,
    • else, generate a new hypothesis.

Elaboration of how to generate a meaning hypothesis:

  1. Make an "intuitive" guess, or "hunch".
    • You may have an inexpressible "feeling" about what the word might mean;
    • this step is intended to "allow" you to test it out.

  2. If you can't do that, or if your hunch fails the test, then:

    • NB: The following 3 steps (2a, 2b, 2c) can be done in any order

    1. Have you read or heard X before?
      • if so, can you remember its context, or what you thought it meant?
      • if so, test that

    2. Does X's morphological structure help?
      • i.e., use your PK about the meanings of prefixes, suffixes, roots, etc., to generate & test a hypothesis

    3. Create an environment in which you might be able to make an "educated" guess:

      1. (re-)read the S(X) slowly & actively (i.e., think about it as you read it)
      2. determine X's part of speech (N? V? Adj? Adv?...)
      3. try to summarize what the entire text says so far.
      4. activate (i.e., think about) your PK about the topic
      5. "put 2 and 2 together":
        • what can you infer from combining:
          • the info in the text so far
          • what you already know about the topic?

  3. If all of the above fails, then do CVA (note: these steps should be done in the order given):
    1. "Solve for X"
    2. Search C(X) for clues
    3. Create H

    In more detail:

    1. "Solve for X":

      • syntactically manipulate S(X) so that X is the focus/topic/subject (see examples)

      • NB: This can be your first hypothesis if you haven't yet been successful in generating one.
      • This hypothesis, however, does not have to be tested, except possibly as a test of your ability to solve for X.

        • Explanation:

          • Let X = "brachet"
          • let S('brachet') = "The brachet bit the hart".
          • "Solving" S('brachet') for X gives us:

              A brachet is something that bit the hart.

          • Testing this by replacing this hypothesized definition in the sentence gives us:

              The thing that bit the hart bit the hart.

            • This is correct and makes sense (well, sort of) but isn't very informative.
            • If we didn't know what a brachet was before, we still won't know.
            • But at least it tells us that we "solved for X" in the correct fashion, because we have a true, if tautological ("empty of meaning"), sentence.

      • Once you have "solved for X", think about possible synonyms;
        • these can be "hypotheses-in-waiting": i.e., hypotheses to test later if you find further clues.
        • E.g., let X = "grudgingly", as in:

            Sandra had won the dance contest and the audience's cheers brought her to the stage for an encore. "Every step she takes is so perfect and graceful," Ginny said grudgingly, as she watched Sandra dance.

        • Solving for X gives us:

            Grudgingly is the way that Ginny said "Every..."
            or: Grudgingly is a way to say "Every..."

        • Now, generate a list of ways to say this:


        • Later, if you find out more about Ginny and Sandra, you can either add to, or eliminate from, this list, thus generating one or more hypotheses that can be tested.
        • It gives you something to look for later in the story (i.e., it gives you something to think about as you continue reading).

    2. Search C(X) for clues:

      • Now try to figure out more about X by looking at C(X)—i.e., the textual context—(and invoking more of your PK!).

        • The operative metaphor here is this:

          • "Solving for X" is like looking at all the ideas in your mind (your "mental semantic network") from the point of view of your idea of X (the "node" labeled "X") and describing your immediate surroundings, the closest ideas (or "nodes") that you can see.
          • "Figuring out more" is like looking beyond those closest ideas (or nodes) to *their* neighbors, and so on.

        • A geographical metaphor might be this:

            Where is UB?
            In Amherst (that's "solving for X", maybe).
            Where is Amherst?
            In Erie County.
            Where is Erie County?
            In western NYS.
            Where is NYS?
            In the US.

            • You know the rest of this story.
            • You also know that by the time you've said "in the universe", you no longer have useful information.
            • But neither is "in Amherst" useful to someone who either already knows that or doesn't know where Amherst is.
            • You need to give information (i.e., a definition) that's somewhere in the middle—not too little, not too much.

      • How do you "figure out more"? I.e., how do you gather clues?
        Recall that earlier we determined X's part of speech. Well,

        1. If X is a NOUN, then search C(X) for information (clues) about X's:

          • class membership:
            • what kind of thing is an X?
            • what kinds of things are Xs?
          • properties (size, color, etc.)
          • structure
            • what parts do Xs have?
            • what wholes are Xs part of?
            • what is X's physical structure?
          • acts:
            • that Xs can do
            • that can be done to or with Xs
          • agents:
            • who can do things to or with Xs
            • to whom things can be done with Xs
            • ownership information
          • comparisons
            • look for/think of possible synonyms
          • contrasts
            • look for/think of possible antonyms

        2. If X is a VERB, then search C(X) for information (clues) about X's:

          • class membership:
            • what kind of act is Xing?
                (e.g., if X = "walk", then walking is a kind of moving)
            • what kinds of acts are Xings?
                (e.g., if X = "walk", then sauntering is a kind of walking)

          • properties of Xing (manner: how is it done?)
              (e.g., walking is a kind of moving by foot;
              sauntering is a kind of slow walking)

          • transitivity
              intransitive: you can X;
              transitive: you can X something
              ditransitive: you can X something to Y
              • here, look for agents and objects of Xing

          • comparisons & contrasts (synonyms & antonyms)

        3. If X is a MODIFIER (i.e., an ADJECTIVE or an ADVERB), then search C(X) for information (clues) about X's:

          • class membership
              (e.g., color, size, shape, etc.)
          • contrasts
              (e.g., if you read, "he did it Xly instead of Yly", where you know what Y means,
              then you can hypothesize that X might be an opposite or complement of Y)
          • parallels
              (e.g., if you read, "he did it Xly, Yly, and Zly"
              and if you know what Y and Z are,
              and—if you're lucky—it turns out that Y and Z are near-synonyms,
              then you can hypothesize that X means something very similar)

          Note that it doesn't matter if you're not sure if some clue that you find should be listed, say, as a "property" or "structural" information. What matters is finding the clues, not what to call them!

    3. Create H:

      • What do you do with all these clues?
        You need to create H:
        • a tentative definition
        • your current hypothesis about what X means (or might mean, at least in the current context).
        • It's the analogue of a detective's hypothesis about "who done it".

      • So, how do you create a definition?

        • Classically (going back to Aristotle), a definition consists of a "genus" and a "difference":
          • You say what kind of thing X is (its "genus")
          • and how it differs from other things of that kind.

        • Schwartz & Raphael (1985) suggest creating a definition "map" that looks like this:

          or Google schwartz raphael "concept of definition"

        • Essentially, you should include answers to the following questions:
          • What is it? (this gives you its genus or superclass)
          • What is it like? (this also gives you its differences from other things in that genus)
          • What are some examples?

        • Then try to create a single sentence (like a Collins COBUILD definition) that tells you what X is,
        • and then replace X by its meaning in the original sentence, as described above.

    Copyright © 2006 by William J. Rapaport (rapaport@cse.buffalo.edu)
    file: curriculum-outline-20060801.html