Thursday, March 22, 2012

Jennifer Spenader: "Variation in Demonstrative choice in Swedish" (2004)

This really thoughtful and interesting paper by Jennifer Spenader tries find out when Swedish speakers prefer references of the form det här huset ("this here house") over reference of the form detta hus ("this house").

She sums up her findings in the maxim "avoid marking that which can be determined from the context" (p. 246). More specifically, this means that the compound demonstrative is inappropriate for a recently mentioned referents. It is also less appropriate for animate referents, allegedly because they are more "accessible" and may thus be recovered from context.

Methodology
In order to test what kind of referents her subjects preferred in different situations, she asked them to fill in forms of the following type (cf. p. 249):
Many research studies show an increase in diabetes or "sugar sickness" among women working in the government sector.
Passive work combined with lack of physical activity leads to an unhealthy lifestyle.
The research also shows that the social part of the women's work often involves chatting over an unhealthy snack.
A. ____________ doesn't don't occur among women in the private sector.
C. ____________ate during certain periods great amounts of candy.
The two last sentences are alternatives that suggest either a more abstract (A) or a more concrete (C) referent for the empty slot. The subjects were only presented with one of them.

Results
Using logistic regression, Spenader found that the most predictive factor for the choice of form was sentence recency: The more time that had passed since a referent was mentioned last, the more likely the subject was to use a a compound form (den här ...). The dependency between the discourse distance and choice of form can be seen in this chart:


The second-most predictive factor was animacy/concreteness. This variable could take the three values Animate, Inanimate-abstract, and Inanimate-concrete.

An Alternative (Stochastic) Model
In section 5 of the paper, Spenader describes an alternative model she has devised within the paradigm of stochastic optimality theory. I didn't know what this involved, but I must say that I find it quite odd.

The idea in stochastic optimality theory is not, as one might think, one of putting weights on the different constraints so that the parsing of an input becomes a matter of maximization of a real function. No, the idea is rather to let the order of the different conflicting constraints be stochastic. It thus introduces noise into the application of the rules in order to reproduce rather than explain empirical variations.

In any case, Spenader sets up a system with seven constraints and trains a system to selecting the best ordering. This turns out—again—to select recency as the most decisive factor, followed by animacy. She tests the trained grammar not by a statistical method but by generating a single random observation and comparing that to the empirical data (using chi-square tests).

This seems like a methodologically bad choice. One could instead pick a test statistic and find the probability that a new data set generated from the trained model would be more extreme than the empirically observed data points. If need be, this could be done with Monte Carlo methods, but I'm not sure that it would even be necessary.

Sociolinguistic Explanations
I find it extremely odd that Spenader doesn't even mention the most obviously relevant dimension of explanation, namely that of high vs. low style.

There is something distinctly informal about constructions of the form det här huset, even though slightly less in Swedish than in the (African-American) English this here house. The closest we get to this is the initial discussion of written vs. spoken language (p. 229).

This formal/informal dimension should be seen in conjunction with, for instance, the social aversion against pointing directly at someone, especially if you're not talking directly to them. This tendency may in some cases have a mirror image in the Japanese tendency to suppress certain nominal phrases and find it coarse or impolite to fail to do so.

A striking English example of this tendency is John McCain's infamous reference to Barack Obama as "that one" during a presidential debate in 2008. He also, by the way, pointed at Obama (without looking at him) while making the reference. The results was that he emitted an air of superiority, as if Obama was an inanimate thing.

Wednesday, March 21, 2012

Prince and Smolensky: Optimality theory (2004)

This is essentially a book on phonology, but also a showcase of optimality theory. The main focus of the book is phonetic parsing, i.e., the problem of specifying how to split a word (say, giddyup) into syllables (gid.dy.up). It's written by Alan Prince of Rugter's university and Paul Smolensky of John Hopkins.

Prioritized Constraints and Optimal Candidates
The answer given by optimality theory is that the parse is produced by iteratively throwing away possible parses that violate important constraints. More specifically, the idea is to start with the whole set of possible parses and then go through an ordered list of constraints, progressively rejecting more and more candidates.

The list of constraints may for instance be something like the following (cf. p. 20):
  1. A syllable must have an onset
  2. The sonority of the syllable nucleus must be higher than or equal to the sonority of the nucleus of any competing syllable.
Applying these two constraints then gives the following procedure for identifying syllables:
  1. Start with the complete list of all available candidates.
  2. If some candidates have onsets and others don't, throw away those that don't.
  3. Among the candidates that remain, throw away any candidate whose nucleus is not of the highest possible sonority.
After having extracted one syllable in this way, one may repeat the procedure to find extract a second syllable from the string phonemes that are not yet parsed.

Optimality As Maximality
The meat of such a theory of course lies in the specific set of constraints that are identified, and the order they are put in. Price and Smolensky seem to believe that both the constraints and the application procedure is universal. All linguistic difference thus stems from different ordering of constraints in their view.

As they note in chapter 5.2.2, the selection procedure they represent indirectly defines a partial order on the candidate parses of a word. One decides whether p is more "harmonious" than q by first comparing their performance on criterion 1, and if that is undecided, their performance on criterion 2, etc. A correct parse is thus one that is maximal with respect to this ordering.

Although it seems somewhat ridiculous that everything in their system is clear-cut and categorical, there is some plausibility to the claim that people actually use algorithms of this sort. The serial procedure employed in optimality theory is very similar to the take-the-best algorithm introduced by Gerd Gigerenzer and Dan Goldstein.

It's interesting that their morphological system has a lot of equivalents to that of Pānini, to whom they explicitly refer.

Wednesday, March 14, 2012

Lewis: Convention (1969)

Lewis' definition of "convention" (p. 78) crucially relies on the concept of common knowledge as well as rationality: For a equilibrium R to be a convention, it must be common knowledge to everyone that the majority conforms to R.

So mere behavioral adaption without a mutual ascription of rationality doesn't count as "convention" (cf. p. 59). Every player has to think that every other player is rational and has the same knowledge as him- or herself. You can't think that you're the only one who's not a robot and still call it a "convention" according to Lewis.

He also excludes solutions to trivial coordination games (p. 78, item 5). There has to be at least two distinct equilibria available so that the fixation in one or the other becomes truly contingent. He doesn't seem to entertain the possibility that there could be something like degrees of contingency.

Tuesday, March 13, 2012

Lewis on bias and analogy in Convention (1969)

David K. Lewis' short book Convention: A Philosophical Study (1969) is an attempt to explain meaning in terms of equilibria in coordination games. The concepts that do the most of the heavy lifting are precedent, analogy, and salience.

The Ambiguity of Precedent
Lewis is aware that no two situations are ever alike, and that agent thus have to engage in some kind of extrapolation:
Suppose not that we are given the original problem again, but rather that we are given a new coordination problem analogous somehow to the original one. Guided by whatever analogy we notice, we tend to follow precedent by trying for a coordination equilibrium in the new problem which uniquely corresponds to the one we reached before. (p. 37)
A consequence of this is ambiguity. He immediately continues:
There might be alternative analogies. If  so, there is room for ambiguity about what would be following precedent and doing what we did before. Suppose that yesterday I called you on the telephone and I called back when we were cut off. We have a precedent in which I called back and a precedent—the same one—in which the original caller called back. But this time you are the original caller. No matter what I do this time, I do something analogous to what we did before. Our ambiguous precedent does not help us. (p. 37)
Or, that is exactly the big question: Whether and when precedent decides or even strictly determines what a construction means in some new situation.

Finding A Relevant Precedent
We then have a two-dimensional similarity space (me/you differences and caller/receiver differences). If these dimensions have the same weight, then both analogies yield the same average fit:


The big question is then how convention is possible at all, if everything is similar to everything else in some respects. Some sense of immediate similarity must be picked up along the way or have been there all along.

Lewis doesn't seem to have any psychological theory of salience or of similarity between new and old situations. However, one could construct a similarity measure based on possible courses of action, so that situations are similar when their elements can be handled in a similar way.

For instance, electricity can be much like water because many of our intuitions about how it behaves are reliable. Thus source elements such as "source," "direction," "pressure," etc. can be paired with certain target elements without needing much behavioral adjustment.

Similarly, source elements as "front" and "back" can be paired with the screen-side and the wall-side of a television in either of two ways. These pairings will on average require less behavioral adjustment than, say, pairing "head" and "feet" with screen-side and wall-side. A few candidate analogies are thus consistent with our prior knowledge, although no single analogy takes absolute priority.

Friday, March 2, 2012

Gibbs: The Poetics of Mind (1994), comments on ch. 3

In the discussion of metaphors in the context of his chapter 3, Gibbs tries, somewhat confusedly, to situate himself within the psycholinguistic field that he surveys.

Handy References
One of the more interesting texts he mentions (p. 104) is an article by Dawn Blasko and Cynthia Connine (1993) which shows that familiarity decreases reading times for idioms. He also mentions (p. 103) a study by Richard J. Gerring and Alice F. Healy (1983) which shows that topic-vehicle ordering can affects reading times for at least some idioms.

Both of these findings are consistent with a keyword theory of idiom comprehension proposed by Cristina Cacciari and Patrizia Tabossi (1988). They are also consistent with accounts based gradual fossilization of conventional metaphors.

The Gibbsean Dichotomy
In the section entitled "Are literal and figurative language processing identical?", Gibbs attempts to introduce a distinction between a Gricean model and his own conceptual model. However, in the process, he ends up drawing a pretty crude caricature of the Gricean view while at the same time virtually turning his own theory into a notational variant of Gricean pragmatism.

The pragmatic theory that Gibbs places at the opposing end of the theoretical spectrum is one that postulates the following list of steps during comprehension, or something like it (p. 111):
Recover literal meaning
Recover metaphorical meaning
Recover idiomatic meaning
Recover ironic meaning
Recover indirect meaning
In contrast with this (ridiculous) theory, he places his own short list (p. 112):
Understand with respect to conceptual knowledge
Understand with respect to "common ground"
This might seem a little odd, especially because of his invocation of "common knowledge." It seems more than usually difficult to construct a theory having "common ground" as one of its legs without ending up in a copy of Gricean pragmatics.

Gibbs is Grice
This suspicion is confirmed when he later gives examples of how the two legs of his theory merge to form a particular act of comprehension:
Suppose Mary and David have an agreement in which Mary walks the dog on sunny days. By uttering The sky is blue, Mary would be communicating "It's your turn to walk the dog."
Suppose David is a kite flyer but has a great fear of being struck by lightning in the process. By uttering The sky is blue, Mary would be implicating "It's safe to fly your kite."
Suppose that David and Mary have in common the knowledge that their friend Betty plays golf on every sunny day. Mary's answer The sky is blue to David's question Where do you suppose Betty is? would imply "Betty is at the golf course."
In all three examples, utterance of The sky is blue prompts the listener to infer a further message that is recoverable only with reference to common ground. (p. 114)
So after all, the theory still just boils down to a reference to implicature, recovery, and common ground. There is no discussion of how Gibbs expects his new formulation of this old proposal to avoid the problems he have just gone through pages and pages to point out.

Clark: "Responding to Indirect Speech Acts" (1979)

I haven't read all of this (quite long) paper, but it's intriguing because it employs a quite unconventional methodology: Herbert Clark makes inferences about people's processing of indirect speech acts by looking at how they respond to them verbally, in particular whether they respond to the literal meaning, the conveyed meaning, or both.

The method is this: You decide on a stimulus questions, for instance Could you tell what time it is?; then you look up a dozen shops in the phone directory and call them; you ask them the question, and you record the exact wording of their response. Nice and simple.

The result that comes out of this exercise is that people frequently respond to both the literal and the derived meaning of a query, in that order. So for instance, you might ask someone Could you tell me what time it is? and that person might respond Yes, it's four o'clock.

Clark's conclusion from this data is that both the literal and the derived meaning of the question must enter the hearer's mind. I'm not sure whether that actually follows from the data (there are other competing explanations), but the observation that people actually say Yes is indeed worth taking seriously for a psycholinguistic theory.

Parrot Responses

One of the reasons that Clark has to doubt his own conclusion is that people in fact also respond Yes when the response to the literal request is in fact No. For instance, you may ask me Would you mind telling me what time it is?, and I may respond Yes, it's four o'clock. Only a very small minority responds with a No (p. 447-48).

This of course undermines Clark's use of the data slightly, since it may imply that people only respond Yes as a matter of verbal habit or perhaps to convey a more general sense of acceptance or affirmation—and not because they actually process the literal meaning of the request.

Clark himself explains the problem away by expanding the Gricean two-stage theory into a three-stage theory: He supposes that the question is gradually broken down according to the following progression:
  1. Would you mind telling me what time it is?
  2. Will you tell me what time it is?
  3. Tell me what time it is!
In this way, he can explain the Yes as a response to an intermediate byproduct of the sentence processing (stage 2), while the other part of the answer is a response to the final product (stage 3).

That doesn't seem quite right—but OK, it's a theory.

A Reanalysis By Gibbs

Raymond Gibbs explains the same data by assuming that people include the Yes because it "is conventionally thought of as being polite" rather because they interpret the question literally as well as figuratively (Poetics of Mind, p. 89).

His support for this claim comes from an experiment in which he forced subjects into a literal reading of a question of the form Can't you ... ? This turns out to be difficult, and people take longer time to do this than to read the same question when it functions as an indirect request.

This data is quite dubious, both because the sentences are quite odd (Can't you be friendly?) and because the stories are quite badly written and don't unequivocally exclude an indirect request reading of the question in the so-called "literal" context.

However, Gibbs follows up with the following comment:
In general, people are biased toward the conventional interpretations of sentences even when these conventional meanings are nonliteral or figurative. Certain sentence forms, such as Can you … ? and May I … ?, conventionally seem to be used as indirect requests. Listeners' familiarity with these sentence forms, along with the context, helps them immediately comprehend the indirect meaning of these indirect requests. People may not automatically compute both the literal and indirect meanings of indirect speech acts. (Poetics of Mind, p. 91)
This seems more reasonable and in fact brings him much closer to the keyword theory of Cacciari and Tabossi (cf. Idioms (1995), chapters 2 and 11).

Conventionality and Frequency

The intuition that Gibbs has about the frequencies is not entirely unreasonable, although the story becomes a little bit more complex when we look at actual empirical frequencies. I've done a quick count based on the MICASE corpus and found the following estimates:

Sentence form   Literal   Indirect    Unclear   Other
Can you …? 17 17 12 9
Could you …? 12 23 13 18
May I …? 0 13 0 6
Would you mind …? 0 9 0 0

The numbers in the two first rows are based on a search in the "highly interactive" section of the corpus, and the numbers in the two last rows are based on a search in the entire corpus.

The phrase can you is so common in the corpus that I just picked 55 occurrences at random and categorized those. All other numbers are based on exhaustive search within the ranges specified above.

The category "Unclear" covers cases where both a question reading and a request reading are compatible with the phrase, for instance:
  • can you remember that?
  • can you cut it up so that everybody gets a piece?
  • can you predict you know what it's gonna be?
The category "Other" covers less interesting search noise like i wanna finish this in May. i wanna finish in the set of May I …? sentences.

It's interesting that there are in fact a relatively large overlap in the direct and the indirect function of the sentences. These are the cases where there is a actual way out for the hearer of the request, such as Could you say anything about that? The existence of such real ambiguities are of course what motivates the use of indirect speech acts as politeness strategies in the first place.

Gibbs: The Poetics of Mind (1994), overview of ch. 3

Chapter 3 (pp. 80-119) is one of the more interesting ones in Gibbs' book. It seems to defend the theory that figurative meanings are (almost always) accessed faster and more easily than literal meanings.

If this is a correct representation of his views, they are indeed consistent with the fact that Gibbs wants the data to eventually point towards cognitive metaphor theory. He thus wants to make plausible the claims that figurative meanings to be connected to the literal meanings by hard-wired neural connections.

What's On the Table

Gibbs' theory of ambiguity resolution should be seen in contrast with two distinct alternatives:
  1. Gibbs' own theory: People usually access the figurative meaning first, and only later and optionally the literal meaning. Possibly, the effort necessary to access a literal meaning "partly depends on the frequency and familiarity of these phrases" (p. 96).
  2. The strictly discrete Gricean theory: Comprehension runs through a series of steps; one looks for a literal reading, evaluates its fitness, rejects this literal reading, looks for a different reading, evaluates this second reading, etc.
  3. The lexical representation hypothesis: A number of mutually inhibitory readings are simultaneously activated, and one of them wins out (p. 93).
Probably no one has believed the strictly stepwise Gricean theory since the mid-1970s. The really decisive difference in this context is thus between a "horse race" theory (p. 93) and Gibbs' own "automatic" theory (cf. pp. 95-96).

Some Problems For Gibbs

Gibbs' theory is problematic mainly for two reasons:

First, it is conceptually confused, because the very idea of an idiom is a phrase is stable and frequently used. Gibbs has thus failed to give a definition of figurative language that would allow us to separate the effect of frequency imbalances from the effect of figurative thought.

Second, as Matthew McGlone has pointed out, there is something theoretically confused about the idea that source and target domains are tightly neurally connected. If there was really an automatic transfer of activation from temperature to emotions, it is unclear how we separate the two domains or arrive at different readings in different contexts. Nothing in Gibbs' theory explains how that is possible.

The Structure of the Chapter

The chapter contains the following sections:
  1. Untitled introduction (p. 80)
  2. The traditional view of figurative language understanding (pp. 81-84)
  3. Psycholinguistic research (pp. 84-109)
    1. Indirect speech acts (pp. 85-91)
    2. Idioms (pp. 91-97)
    3. Slang (pp. 97-98)
    4. Proverbs (pp. 98-99)
    5. Metaphor (pp. 99-106)
    6. Metonomy (pp. 106-107)
    7. Irony (pp. 107-109)
  4. Are literal and figurative language processing identical? (pp. 109-115)
  5. The processes and products of understanding (pp. 115-119)
  6. Conclusion (p. 119)
The subsections on indirect speech acts, idioms, and metaphors contain references and discussions of much of the same evidence as Gibbs recycles in chapters 4 and 6.

For instance, the (very interesting) experiment by Cristina Cacciari and Patrizia Tabossi that he discusses on page 95 reappears on page 287 almost as if he was not aware that he had already presented it.

(By the way, based on the prose style, it seems probable that the discussion in chapter 6 was written before that in chapter 3).

    Thursday, March 1, 2012

    Gibbs: The Poetics of Mind, comments on ch. 5 (1994)

    Most of chapter 5 of Gibbs' book is spent criticizing some rather uninteresting claims about metaphor, namely, the claim that they rely on preexisting similarity and the claim that comprehension moves through a process of finding and rejecting a literal reading before considering a metaphorical alternative. The chapter plods along without much reflection or analytical distance.


    Gibbs vs. Sperber and Wilson
    During a discussion of Searle's Speech Acts, Gibbs uses the opportunity to bring out his old point about the separation of process from product in theories of comprehension. He claims that pragmatic theories focus exclusively on products while making claims about processes as well (p. 228).

    He also briefly considers relevance theory as a possible alternative in the section named "The metaphor-as-loose-talk view." "Loose talk" is a relevance-theoretic name for vague assertions, especially vague assertions that are just precise enough given the context.

    Gibbs rejects that metaphors can be regarded as "loose talk" in this pragmatic sense because it
    incorrectly assumes that that metaphors [...] obligatorily demand additional cognitive effort to be understood. Furthermore, the very notion of metaphor as loose talk presupposes that metaphorical language only resembles speaker's thoughts rather than being a direct reflection of ideas or concepts that are actually constituted by the speaker. (p. 232)
    The second sentences obviously only constitutes an argument against relevance theory if one agrees that metaphors "reflect ideas." In any case, trying to explain meaning as relation between a thought—be it in terms of resemblance or identity—seems like a non-starter to me.

    Bidirectional Metaphor
    In the chapter on Max Black's theory ("The interactive view"), Gibbs briefly mentions that the metaphor PEOPLE ARE MACHINES is bidirectional or reversible (p. 238). This is followed by a generic reference to More Than Cool Reason, without any page number.

    He explains this by saying that "different things get mapped" without getting any more into the issue.

    Gibbs vs. Gentner
    Gibbs sees the theories of Ortony and of Dedre Gentner as elaborations of Black's interaction theory. In particular, he suggests that they may be compatible (p. 245), unlike what Gentner seems to think. However, he rejects both on the grounds that none of them
    is especially informative about metaphor understanding as an early on-line process, because they primarily focus on explaining the results, or late products, of that understanding. (p. 246)
    He doesn't give any further evidence for this claim.

    After thus having rejected the salience imbalance model and the structure-mapping model, he goes onto present Glucksberg's class inclusion model as a patch on this "problem" (p. 246-48).

    He finds this model "appealing" for its focus on process (p. 247) but rejects it because "it might very well be the case that such metaphorical groupings [= Glucksberg's ad hoc concepts] already exist as part of our everyday conceptual structures" (p. 247). And of course, if one believes that assertion, one should go for cognitive metaphor theory rather than Glucksberg's picture.

    Gibbs On Poetry and Freshness
    In his presentation of the cognitive theory of metaphor, Gibbs finds himself in a strange two-way situation with respect to the freshness of metaphor:
    The conceptual view of metaphor also explains why people find such great beauty and power in poetry and literary prose. Verse embellishes the more mundane ways of thinking about our worldly experiences. (p. 249-250)
    In fact, cognitive metaphor does exactly the opposite: It explains the intelligibility of metaphor by assuming that there is nothing new in a metaphor. As Lakoff and Turner claims, poetry uses the same metaphors as everyday discourse.

    If this were really true, it seems very difficult to explain why they would sometimes seem fresh, beautiful, and thought-provoking, while at other times mundane, flat, and uninteresting. Almost as an aside, Gibbs tries to do so by driving wedge in between the mapping and its instantiations:
    The view of metaphor as conceptual structure is particularly valuable as a linguistic theory of metaphor because it suggests a difference between having a metaphorical mapping of two disparate domains already existing as a unit in one's conceptual system and the mental act of putting together the same metaphor for the first time. (p. 251; my graying out)
    That's quite confusing given that the whole theory relies on metaphors to be unsurprising and effortless. In particular, does Gibbs mean "the same [conceptual] metaphor" or "the same metaphor[ical mapping]" in the last sentence? That makes quite a large difference for the theory.

    This is partly, but only partly, cleared up later on the same page:
    Of course, many linguistic metaphors specify new twists on some conventional conceptual metaphor and may require additional inferential processing to be properly understood. This additional inferential processing often operates on the products of earlier comprehension processes (i.e., is part of metaphor interpretation rather than comprehension per se). (p. 251)
    So yet another distinction has to be introduced to save the theory, this time between "interpretation" and "comprehension per se." Again, this pushes to conceptual mapping to some background role where it virtually has no observable qualities.

    Gibbs: The Poetics of Mind, overview of ch. 5 (1994)

    Chapter 5 of Gibbs' book, "Understanding metaphorical expressions,"  surveys various theories of meaning which apply directly to metaphor.

    This involves the usual list of authors that cognitive metaphor theorists refer to when they need to stick a name to a straw man position: Aristotle, Max Black, Richard Rorty, John Searle, etc.

    The chapter contains the following sections:
    1. An untitled introduction (pp. 208-210)
    2. Aristotle on metaphor (pp. 210-211)
    3. I. A. Richard's contribution (pp. 211-212)
    4. The metaphor-as-comparison view (pp. 212-218)
    5. The metaphors-without-meaning view (pp. 218-222)
    6. Metaphor as anomaly (pp. 222-225)
    7. Metaphor and speech acts (pp. 225-230)
    8. The metaphor-as-loose-talk view (pp. 230-232)
    9. The interactive view (pp. 232-239)
    10. Psycholinguistics and the interactive view (pp. 239-248)
    11. Metaphor as conceptual structure (pp. 248-260)
    12. Metaphor and literature (pp. 260-261)
    13. Conclusion (pp. 261-264)
    Note that sections 10 and 11 are longer than the rest. Together, these two sections constitute almost two fifths of the chapter.