Tuesday, April 30, 2013

Gibbs: Embodiment and Cognitive Science, ch. 4 (2005)

It's a bit surprising that someone as keen on criticizing introspective psychology as Ray Gibbs should find himself doing it extensively when the lights are dimmed. Pages 90–114 of his most recent book almost reads as an illustration of the kind of speculative neuroscience that he warns others against.

The Defective Mouth

For example, in a discussion of the idioms chew the fat and chew the rag — based on discussions in Goossens et al. (1995) — he motivates the meaning of the phrase as follows:
Given that both fat and rags can be chewed a long time, with little nutritional value coming from these activities, these idiomatic phrases express the idea of talking about something a long time with little new information to be gained from this experience. (p. 100)
Even if we accept this link, when did fat ever become low on nutrition?

More factual weirdness crops up as Gibbs discusses how people experience various bodily processes:
The metaphor "spit out" reflect the idea that has something of value in the body, which through effort he or she is able to gather up (like spit or phlegm) and say (or expectorate). (p. 100)
That's a bit gross, but it's also inaccurate: When was spit ever "valuable," and when did we ever have a reoccurring experience of desiring other people's spit? Even if the connection was unproblematic, the imagery wouldn't work.

Similarly, without any open reservations, Gibbs links the experience of coughing something up to a state of well-being:
"Getting something off one's chest," just like "blowing off steam" and "coughing something up," restores a sense of balance or well-being to an individual. (p. 103)
Even more oddly, the idiom lie through one's teeth is explained by means of an image of the truth being located inside the mouth of the speaker:
When people "lie through their teeth," the [mouth which is perceived as a] container is perceived as a hiding place where true information resides, but the container is somewhat defective, and we can see through the speaker's shameless attempt to lie about something when the truth can partly be seen. (p. 104)
It wouldn't matter that this image strikes me as completely bizarre if we only got some evidence that it is psychologically plausible. The problem is that, of course, we don't.

Murphy–Gibbs Revisited

Gibbs still has an old axe to grind with Gregory Murphy, who pointed out some serious gaps in his theory back in 1996. In the book, Gibbs essentially reiterates his defense of cognitive metaphor theory — constrain the theory so everything it claims is prefaced by "optionally."

So let's take the conclusions first: In the last section of the chapter on concepts, Gibbs repeats that "conceptual simulations surely … create imaginative understandings of events" (p. 122), and that
… human conceptual processing is deeply grounded in embodied metaphor, especially in regard to abstract understandings of experience. (p. 122)
On the previous page, he also talks about "embodied simulation in the creation of concepts in context" and "the sensorimotor nature of conceptual processing" (p. 121).

All of this talk of online simulation and temporary constructs seems quite consistent with the idea that comprehension isn't a matter of word-for-word recall, but rather of constructing something like a little film clip in the head which represents the meaning of a sentence. In the process of creating such a film, common bodily experiences may take precedence over other contextual cues.

I take this to be the idea behind his talk of emergence and self-organization, although the whole thing is a bit vague:
My suggestion is that image schemas are attractors within human self-organizing systems. The important point here is that attractors are not localized representations, but emerging patterns of entire systems in action (i.e., interplay of brain, body, and world). In this way, the stable properties of image schemas (e.g., the topographic structure of something like SOURCE-PATH-GOAL) are not separate from sensorimotor activity. Image schemas should not be reduced to sensorimotor activity, but it is a mistake to view image schemas as mental representations that are abstracted away from experience. (p. 115)
The problem is that this seems to contradict his claim that certain "attractors" can be ignored when we feel like it:
Each metaphoric construal of a concept [e.g., LOVE] in some context results in a concept that is independent as a temporary representation apart from embodied source domain information in long-term memory. My suggestion, then, is that conceptual metaphors may not preexist in the sense of continually structuring specific conceptual domains. But conceptual metaphors may be used to access different knowledge on different occasions as people immediately conceptualize some abstract target domain given a particular task. Conceptual metaphors may also simply emerge as the product of conceptualizing processes, rather than serve as the underlying cause of these processes. (p. 121)
As Murphy pointed out in the original debate, this leaves wide open the question of how people select the appropriate metaphor, and after the metaphor is chosen, how to select the correct properties from the source domain. Whenever someone postulates a context-sensitive selection, they also implicitly postulate a mechanism that can perform this selection.

What Does This Leaves Us With?

It's a part of this picture that Gibbs wholeheartedly supports Joe Grady's corrective to cognitive metaphor theory (1997), i.e., that non-primary metaphors are constructed out of several primary metaphors glued together. This means that the theory essentially only applies to primary metaphors.

So in sum, the picture that seems to emerge is the following: Whenever you understand affection, normality, power, quantity, etc., you do so by means of mental simulations of a bodily process. Comprehension is a matter of choosing he right analogy between a bodily process and an abstract concept. Because some experiences are more salient, they are also more likely to be used.

So that's at least a theory. It's probably wrong, though, since many sentences are understood perfectly fine without the use of any bodily simulation — in fact, bodily simulation can even be an obstacle to correct understanding in many cases.

Monday, April 15, 2013

Talmy: Toward a Cognitive Semantics (2000)

It appears that Leonard Talmy has put the whole manuscript for his book Toward a Cognitive Semantics (2000) online. That's nice of him.

I've just had a brief look at chapter 5, which is about figure/ground relations. It contains some pretty funny and telling linguistic examples. Unfortunately, it doesn't seem to include any psychological evidence, though.

Even so, I'd still like to show a few of his examples because they bring out his point so neatly.

I Put My Mouth Around the Food

He argues that some things are inherently more likely to be used as grounds. This is illustrated with contrasts like
    1. The bike is near the house.
    2. *The house is near the bike. (p. 314)
    1. The TV antenna was above the house.
    2. ?The house was below the TV antenna. (p. 317)
    1. My sister resembles Madonna.
    2. ?Madonna resembles my sister. (p. 318)
    1. Clark Kent is Superman.
    2. ?Superman is Clark Kent. (p. 318)
    1. y = 3x2 + 1
    2. 3x2 + 1 = y (p. 320)
    1. He dreamt while he slept.
    2. *He slept while he dreamt. (p. 324)
Being an old-school linguist, Talmy of course sums up the lesson learned from these examples with a story about universal principles of cognition and the like. That's all fine and good, but you can probably already predict what that story is going to look like based on these examples.

Before the Bomb Exploded, I Pushed the Button

One thing I found really thought-provoking was Talmy's discussion of how before and after are expressed in the native American language Atsugewi. According to his gloss, these relations are expressed in a way that literally translates as the following (p. 323, my italics):
  • Having-eaten, we left. (= We left after we ate.)
  • Still not having-left, we ate. (= We ate before we left.)
Notice that all four sentences actually mean the same thing (in a strictly logical sense).

But I for one needed some time to even see that. Time is difficult.

Sunday, April 14, 2013

Steen: Finding Metaphor in Grammar and Usage (2007)

Gerard Steen's 2007 book on metaphor is very long, and it contains both chapters that are interesting for me and chapter that aren't. Steen has chosen to read the chapters so as to cover all the possible combinations of the following pairs of dichotomies (cf. ch. 1.4):
  • grammar — usage
  • language — thought
  • symbols — behavior
So that's eight topics. Not surprisingly, this structuring a book this way produces a fair bit of repetition, but it also makes it easy to skip things that are less important to you.

From my perspective, the two important chapters the ones that touch on psycholinguistic aspects of the theory. That's chapters 9 and 12.

Curriculum

There's a list of books and papers that seem to have been more informative of Steen's position than others. I would say his core reading list consists of the following:
He also refers a number of times to Lynne Cameron's Metaphor in Educational Discourse (2003), although that seems to be less central to his argument.

With respect to the debates about the psychological reality of conceptual mappings, he cites the following works in particular a lot:
There are other references as well, but these are the ones that receive the most attention.

Conclusions

Much of the point of Steen extensive methodological discussion is to weed out some of the more crazy claims made by cognitive metaphor theorists. In the process of doing so, he occasionally his own opinions on these matters.

Somewhat surprisingly, he is quite pessimistic about the cognitive validity of cognitive metaphor theory. In his discussion of Perfetti and Giora, he poses the following three questions (pp. 351–52):
  1. Does the activation of polysemous words with metaphorical senses, in comprehension contexts requiring the metaphorical sense, always involve the activation of both the basic nonmetaphorical sense as well as the indirect metaphorical sense (processing of linguistic form)?
  2. Does the activation of polysemous words with metaphorical senses, in comprehension contexts requiring the metaphorical sense, always involve such a degree of activation of the basic nonmetaphorical sense that it is sufficiently rich to function as a conceptual source domain for mapping the indirect metaphorical sense (processing of conceptual structure domains)?
  3. Does the activation of polysemous words with metaphorical senses, in comprehension contexts requiring the metaphorical sense, always display some manifestation of an obligatory mapping from the basic nonmetaphorical sense to the indirect metaphorical sense (processing of conceptual structure of mapping)?
(Incidentally, this quote also illustrates Steen's almost meditatively repetitive style.) Anyway, his own answer to these questions is negative:
Salient word meanings [in the sense of Giora] are word meaning that are frequent, familiar, conventional, and prototypical, and it is these senses which are quickest to retrieve. Since metaphorical word senses may also be highly frequent, familiar, conventional, and prototypical, they may also be activated faster than their nonmetaphorical counterparts. It follows that questions 3, 2 and even 1 have to receive negative answers, as may be inferred from Giora's detailed review of her own and other researcher's work. (p. 352)
It follows that the symbolic analyses we know from cognitive metaphor theory don't really have much cognitive validity:
The fact that symbolic synchronic or historical analysis may helpfully privilege a particular class of senses as direct or nonmetaphorical does not directly map onto individuals' cognitive representation of such senses as prior or even necessary in the online comprehension of linguistic forms when these are to be processed in the metaphorical ways. (p. 352)
That's pretty radical, but it's perhaps less surprising if we see Steen as the strict methodology policeman of cognitive metaphor theory — in that case, we would expect him to have the most cautious conclusions in the field.

Saturday, April 13, 2013

Gibbs: Embodiment and Cognitive Science, contents (2005)

Unlike The Poetics of Mind (1995), Gibbs' new book not only discusses psycholinguistics, but also perception and other linguistic phenomena. However, chapters 4 and 6 still focus on cognitive metaphor theory as such, and they largely reiterate the claims he has made elsewhere.

Unfortunately, the new book, like the old one, still has a mostly flat structure, more in the style of a laundry list than a narrative. Since I find this a little difficult to find my way around in, I'll just quickly give a table of contents for the two chapters.

Chapter 4: Concepts

The contents of chapter 2 (pp. 79–122) are as follows:
  1. Untitled introduction (pp. 79–80)
  2. Traditional Views of Concepts (pp. 80–81)
  3. Problems with the Traditional View (pp. 81–86)
  4. Perceptual Symbols (pp. 86–90)
  5. Image Schemas and the Metaphorical Nature of Abstract Concepts (pp. 90–96)
  6. Thinking (pp. 96–99)
  7. Linguistic Action (pp. 99–104)
  8. Grammar and Spatial Concepts (pp. 104–107)
  9. Political Ideas (pp. 107–111)
  10. Mathematical Concepts (pp. 111–114)
  11. Questions about Image Schemas (pp. 114–115)
  12. Questions about Conceptual Metaphors (pp. 115–116)
  13. A New View of Embodied Metaphor (pp. 116–118)
  14. Is Cognitive Linguistic Evidence Relevant to the Study of Cognition? (pp. 118–121)
  15. Conclusion (pp. 121–122)
Sections 6 through 10 (pp. 96–114) contain various linguistic illustrations of the notion of an schema. The headings "Thinking" and "Linguistic Action" could thus also have been named "English words words for thought-related concepts" and "English words for speech-related concepts."

All three chapters are extremely speculative, and they rely almost solely on introspection and post-hoc rationalization of linguistic expressions.

Chapter 6: Language and Communication

Chapter 6 (pp. 158–207) contains the following subheadings:
  1. Untitled introduction (pp. 158–159)
  2. Time Course of Linguistic Communication (pp. 159–160)
  3. Language Change (pp. 160–161)
  4. Speech Perception (pp. 161–165)
  5. Gesture and Speech (pp. 165–170)
  6. Body Movement and Discourse (pp. 170–174)
  7. Word Meaning (pp. 174–180)
  8. Image Schemas and Utterance Interpretation (pp. 180–183)
  9. Embodied action in Metaphor Processing (pp. 183–184)
  10. Desire as Hunger: A Case Study in Embodied Metaphor (pp. 184–187)
  11. Understanding Time Expressions (pp. 187–190)
  12. Embodied Metaphors in American Sign Language (pp. 190–194)
  13. Neural Theory of Language (pp. 194–198)
  14. Embodied Construction Grammar (pp. 198–199)
  15. Embodied Text Understanding (pp. 199–205)
  16. A Case Study: Indexical Hypotheticals (pp. 205–207)
  17. Conclusion (p. 207)
Section 2 is a warning not to confuse diachronic and synchronic issues. Section 3 is a discussion of etymology with a reference to Eve Sweetser.

Section 10 is about the desire/hunger study that I have read, but been unable to obtain the data from. Section 13 is a discussion of Narayanan's seriously inadequate computation model of verb comprehension.

Monday, April 8, 2013

Kövecses: Metaphor in Culture (2005)

Let's have a little look at the fine print in Zoltán Kövecses book Metaphor in Culture. After trawling through about 300 pages of positive evidence, Kövecses gets around to considering some problems. A short passage within the last five pages of the book is dedicated partly to a discussion of incoherence in metaphor (pp. 289–292).

Kövecses looks at two examples. First, he considers the conflict between fear-is-cold metaphors (cold feet etc.) and the sentence Our fears are fueled by acts of terrorism (p. 289). Then he gives a metaphorical explanation for why Clinton couldn't have an affair, while Mitterrand could (p. 291). Let's look at those two examples one by one.

Fear and the "'Cognition over Embodiment' override"

About the fear that seems to be both cold and burning at the same time, Kövecses writes:
What's going on here then? Does metaphorical thought conflict with embodiment, thus contradicting one of the major claims of the cognitive linguistic view of (metaphorical) understanding? I believe this is exactly what is happening. But I do not think that this conflict creates a major difficulty for the experientialist view of the embodiment of metaphors. All that needs to be done is to make a slight change in the strong version of the theory. I suggest that we conceive of the simple but highly generic metaphors that are based on tight correlations in experience (such as INTENSITY IS HEAT) as powerful conceptual devices that can override local embodiment in other parts of the conceptual system (in FEAR IS COLD). In other words, once we have a highly [p. 290:] entrenched and generic simple metaphor such as INTENSITY IS HEAT, this metaphor can be applied even to cases in which the metaphor does not fit local embodiment (as in the case of FEAR IS COLD). I do not know how common similar cases of "embodiment override" are, but I guess they are common enough that they call for some explanation within the theory of embodiment of meaning. The notion of simple, generic, correlation-based metaphors that are powerful enough to override local embodiment would be my best best shot at solving the issue. (pp. 289-90)
The biggest problem with this suggestion might not be its theoretical shortcomings, but rather its blatantly ad hoc character. It is indeed a "good shot" at evading a problem but has little psychological motivation. With such a patch-on-patch methodology, it is difficult to see what would count as a "major difficulty."

More specifically, it is also difficult to see why INTENSITY IS HEAT should be more "tight," "simple," and "entrenched" than FEAR IS COLD. If a "tight correlation" is not a matter of embodiment anymore, how then are we supposed to understand this concept? All this seems to pull the carpet from under several fundamental aspects of the theory, in an obscure corner of the book.

Sex in France and the "'Social-Cultural Experience' Override"

And now for the sex. Kövecses opens his discussion of the Clinton/Mitterrand comparison with a quick brush-up of George Lakoff's theory of political emotions, in particular, the alleged SOCIETY IS A FAMILY metaphor. Then the same puzzle as above:
Given that the same metaphor and metonymy in the French conceptual system [as in the American], how come marital infidelity never became a political issue in France? I believe that thte answer is that another part of the conceptual system can override some of the mappings of the SOCIETY IS A FAMILY metaphor. If there is a culture, such as France, in which sexual freedom (of even family members) is an important value, then the metaphorical connection "family issue political issue" is not made (i.e., activated) because it would be inconsistent with the part of the conceptual system that maintains that sexual freedom is important. In other words, a part of the broader cultural context (in this case, the value of sexual freedom in France) can override the particular mappings that a culture sets up between the FAMILY source and the SOCIETY target. (p. 291)
So you see, your values are founded in metaphorical thought, except when they are founded in your values.

Who's On Top, Anyway?

The notion of "override," here as in Lakoff's version, implicitly requires one mechanism to have a higher priority than another.

This assumption is usually not spelled out when the concept is invoked, since it is usually called upon to explain inconvenient data while still maintaining the strong view that "metaphor governs thought." This is of course impossible if one simultaneously says that thought governs metaphor.

Naturally, Kövecses also faces this problem, which Michael Kimmel apparently pointed out to him (p. 292). Consequently, he goes on to disavow the rhetoric of the preceding pages:
The talk about "overrides" may seem to suggest some kind of temporal and causal progression from a "universal base" to a "cultural overlay" and an ontologically "most basic" part from which other things emerge, or develop. I do not intend any such interpretation. I do not claim that there is first a universally embodied shared metaphor in different cultures, just as I do not claim that the three systems that interact in metaphorical conceptualization [bodily, cultural, and cognitive] can be meaningfully separated from each other for other than heuristic purposes (see chapter 10). As I have argued elsewhere (Kövecses, 2000a) and in chapter 10, I view the emergence of metaphors as being simultaneously shaped by both embodiment and culture (and most likely also by communicative context). I simply use "override" as a convenient way of talking about certain "incoherences" and "conflicts" among the heuristically postulated systems. In reality, all that we see is the differences in metaphorical conceptualization across and within cultures. Thus the term override should be taken as representing a convenient fiction to talk about such differences as we move from culture to culture, from subculture to subculture, and so forth. (p. 292)
Falling back on "words don't matter" is a bizarre argument for somebody working in cognitive metaphor theory; but rhetoric aside, what does this mean in terms of the two examples above?

Specifically, have a second look at the following two snippets:
… generic metaphors that are based on tight correlations in experience … can override local embodiment …
… a part of the broader cultural context … can override the particular mappings …
What happens if we replace these sentences with the pedestrian observation that cultures differ?

There might be a reading under which this states more than the tautology that things differ when they differ, but I don't feel I'm getting much to work with here.

Wednesday, April 3, 2013

Cho and Kreps: "Signaling games and stable equilibria" (1987)

Because games often have several equilibria and no obvious way of choosing between them, it is always good sport to try to come up with some new, stronger refinement of the concept of the Nash equilibrium. This paper investigates a number of ways of doing so, mainly motivated by a single type of equilibrium that the authors find unintuitive.

The game that Cho and Kreps investigates is given by the following tree:


They motivate the game by the following annoyingly ridiculous story:
  • A can have two types, "wimp" (with 10% probability) and "surly" (with 90%).
  • A prefers quiche for breakfast when he's "wimp" and beer when he's "surly." He gets 1 point for having the breakfast he prefers, and 0 otherwise.
  • In addition, A prefers not to duel B, and he gets 2 points for avoiding a duel.
  • B prefers to duel A iff A is a "wimp" and not to duel him iff he is "surly." B gets 1 point for making the right decision and 0 otherwise.
Under these assumptions, the game has the following two equilibria:
  1. A has beer for breakfast regardless of his type; B duels A iff he has quiche for breakfast.
  2. A has quiche for breakfast regardless of his type; B duels A iff he has beer for breakfast.
So in the  first equilibrium, the game ends in one of the two lower branches of the subtrees on the left, with payoffs (2, 0) or (3, 1). In the other equilibrium, the game ends in one of the two lower branches of the subtrees on the right, with payoffs (3, 0) or (2, 1).

It is equilibrium no. 2 that Cho and Kreps find unintuitive and spend the most of the paper combatting. Their main focus in this effort is the notion of "stable equilibria" as defined by Elon Kohlberg and Jean-Francois Mertens in a 1986 paper—a concept that Cho and Kreps state that they have "mixed feelings" about (p. 181).

Johan van Benthem: "Games that make sense" (2008)

This is a chatty note on the various uses of game theory in semantics and pragmatics. It makes two points that I find worth mentioning.

First, van Benthem correctly points out that there are two different notions of "game" in play in semantics, and that these are sometimes confused. One is Hintikka-style verification games, and the other is Parikh-style signaling games. Although the verification games may in some sense be taken as idealized roadmap for a conversation, this fact is not completely obvious and cannot be taken for granted.

Second, he notes that signaling games have thrown a lot of the syntactic and semantic structure from logic overboard in its attempts to model the emergence of meaning. Since logic usually models hard, conventional facts about a language, this means that game-theoretic approaches to pragmatics have a hard time getting off the ground, because they take everything to be up to debate and revision in the online conversation situation. This is a false assumption in many cases.

Van Benthem writes:
Finally, from the viewpoint of natural language, we have not even reached the complete picture of what goes on in ordinary conversation. There may be games that fix meanings for lexical items and for truth or falsity of expressions whose meaning is understood. But having achieved all that, the ‘game of conversation’ only starts, since we must now convey information, try to persuade others, and generally, further our goals – and maybe a bit of the others’ as well. (p. 7)
He gives a tip of the hat to a number of people in dynamic epistemic logic and then continues:
But conversation and communication is also an arena where game theorists have entered independently, witness the earlier references in Van Rooij [42], and the recent signaling games for conversation proposed in Feinberg [22]. Again, there is an interface between logic and game theory to be developed here, and it has not happened yet. (p. 8)
But certainly a number of people are currently trying to smuggle more logical assumptions into the games, with various levels of success.