10. Analogy_ The Warp and Woof of Cognition.pdf

10. Analogy: The Warp and Woof of Cognition : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:35 PM

10. Analogy: The Warp and Woof of Cognition

RAIMO ANTTILA

Subject

Linguistics » Historical Linguistics

Key-Topics

cognition

DOI:

10.1111/b.9781405127479.2004.00012.x

Greek science was based on an analogical grid of a contiguity axis (also known as causal, or indexical) and a

similarity axis. Thus Aristotle defined genera in the way shown in figure 10.1 (Hesse 1966: 61; this has often

been quoted, e.g., Anttila 1977: 18; Itkonen 1994: 44; Itkonen and Haukioja 1996: 137).

Lining up secure similarities gives an anchor for going into the uncertainties (the dots in figure 10.1 ),

especially if there is an imbalance ( figure 10.2 ). 1 This is still the essence of scientific analysis (and everyday

perception and understanding). Note how water waves led to sound waves to light waves, and so on (Hesse

1966: 11, 68, 93–6). There are positive and negative analogies that build up explanations, but particularly

useful in everyday life is persuasive analogy - for example, the state is to its member as a father is to his

child - and such analogies are the essence of cultural networks and mythologies (there is nothing else, in

fact). 2 Analogy mediates between actuality and potentiality.

Figure 10.1 The structure of analogy

Figure 10.2 Gap filling

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747912

Page 1 of 11

10. Analogy: The Warp and Woof of Cognition : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:35 PM

Figure 10.3 Meaning and form

The two axes in the analogical frame (reflecting a proportional relationship, an expression of similarity of the

sort A : B :: C : D) cover any kind of material where we have similarity and contiguity. In figure 10.2 , we have

on the left two axes which share the top left corner unit. There is a gap x that calls out to be filled by

analogy; this has happened on the right, with the box x . This situation is usually given with numbers: 4/2 =

10/ x ; x = 5, and no problems arise, since we get exact results (identical relations). But with most material

fed into such structures we have to be happy with vaguer similarities (in other words, the similarity stretch

between x and x [>] can be a long gradient scale = drift). Note that the left-right sequence in figure 10.2

succinctly summarizes analogy's two great theoretical powers. First, it shows that analogy is the agent that

dives into the hermeneutic gap, the átopon , the ‘out of place,’ the strange, the problem that asks to be

explained or solved; second, at the same time it is an impelling force of closure in gestalt terms. In such

structural asymmetry perception strives for wholeness. Thus, hermeneutics (pattern explanation) and gestalt

theory work under the same laws of human understanding. We also secure imposing metatheoretical glory

for analogy, although we just generally see its practical application value.

When it comes to linguistic signs - and let's just say words at this juncture - we have to remember that they

are combinations of form and meaning (again simplifying the situation to a Saussurean colligation).

Incredible mistakes are committed if only form is considered, and thus analogy seems to fail (but it is the

linguist who has failed; cf. Itkonen and Haukioja 1996: 135). Similarity relations exist both in meaning and

in form, and meaning and form are combined in the symbolic colligation. Observe the six such colligations in

figure 10.3 , say, where the squares represent words. The top part of the square represents meaning and the

bottom form. Words (2) and (5) share the same meaning, and (1) and (4) the same form. Various degrees of

similarity can also be perceived ( figure 10.4 ). Thus a figural set-up with identical form would work toward

changing meaning (1) toward meaning (4) or vice versa (the diagrams again emphasize identity), or with (2)

and (5), the forms could go either way. The actual forces depend on the centrality of each feature in context,

culture, grammar, and so on. Numbers (2) and (5) could also portray allomorphy, as could (y) and (z), since

the lexical meaning is identical, and in this situation contamination (y, z) is also normal.

Figure 10.4 Similarity relations

The force here has been described (see, e.g., Anttila 1972), in a way used ever since the Ancient Greeks, as

“one form-one meaning” (although this particular characterization is mine, as well as the notation below), an

ideal in sign formation that of course will never be achieved, but the ideal pushes constant change (cf.

Anttila 1977:55–8, 1989: 100–1, 107, 129–30, 143–6, 407; Itkonen and Haukioja 1996: 162). The main

force in such change is analogy, as rationality, of course. What this principle says is that the configurations V

(two meanings - one form) and Λ (one meaning-two forms) tend to be leveled out to I or split into I, I .

(Then of course metaphor, metonymy, loan translation, and folk etymology again create polysemy I > V .)

Allomorphic alternation, Λ , as in the original shade/shadow , or cow/ki-ne , tends either to split into

independent words I, I ( shade and shadow with different meanings) or to get leveled out into I (as in

cow/cow-s ). Extension of alternation from a more restricted environment to practically every word as in

Lapp/Saami consonant gradation, Λ , I > Λ , still represents unity for diversity. Leveling and extension remain

as the most prevalent analogical change concepts.

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747912

Page 2 of 11

10. Analogy: The Warp and Woof of Cognition : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:35 PM

The situation (1, 4) in figure 10.4 is so-called homonymic clash, and if change occurs, formal differentiation

is expected. Keller's treatment of German englisch ‘angelic’ and englisch ‘English’ is a good example (1994:

80–3, 93–5, 124, 132, 156). In a context like englische Mädchen the conflict was insidious, and the first one

was replaced by engelhãft , restoring one meaning-one form. The identical base morphemes need not be

perceived; the sign is normally taken as a whole. But any feature perceived and any interpretation

successfully forced on a percept is a potential anchoring for analogy. Thus French cerise > cheris was

interpreted in English to have the pl. -s , which then necessitated a new analogical sg. cherry . Similarly Arabic

kitabu ‘book’ in Swahili was interpreted to contain the native noun classifier ki -, whereby the plural had to

manifest as vitabu . Such examples are commonplace (latest treatment in Itkonen 1999: §III).

These figural form-meaning colligations appear everywhere in language structure and use. Consider

borrowing, perfectly analogical. For example, note the following situation between American English and

Finnish as pertains to certain “tools of smudging,” forming thus a general semantic similarity field of

something like this articulation:

This kind of different partition of semantic fields is typical between languages, and it does not matter that

lyijykynä ‘lead-pen/quill’ is motivated. The relation here between the two languages is roughly Λ /I (with the

slash indicating the formal similarity (the arrow) in pensseli/pencil ). In American Finnish however, where

English is an extremely strong social force (the necessary indexical anchoring for analogy), it exerts the one

meaning-one form pressure on Finnish. Since pensseli /pencil is a formal match, it is kept, but with English

semantics, whereupon harja takes on the whole range of English brush :

The result is greater one-to-one unity, both in form and in meaning, between English and American Finnish

(i.e., I I). This is quite common in (American) immigrant situations; for example English like (1)

‘similar/equal’ and (2) ‘to be fond of’ versus German gleich (-) (1) has yielded Pennsylvania Dutch (2) ich

gleiche dich ‘I like you’. Similar examples could be multiplied by the thousands.

1 Definition of Analogy

The above was of course quite general, but sufficient; largely the proportional aspect was treated, hinted at

with language material. Now when the basic structure has been laid out, we can add a basic definition:

analogy is a relation of similarity, that is, a diagram (in the sense of Peirce 1965: vol. II, with warp and

woof). In other words it is structural similarity (Itkonen and Haukioja 1996: 157; cf. Holyoak and Thagard

1995: 208). A diagram is the central icon, central in any science. But it is central in perception and cognition

also, because if we would just rely on images (i.e., mere pictures of feeling-similarity), we would not get

anywhere (not out of our own heads, although we would not even know it). A diagram gives us a reasonable

map of reality pointing toward further knowledge. All this is heightened with the higher-order diagram, the

metaphor (which I try to avoid here for reasons of space).

The proportion brings out the relation quite nicely and convincingly (for most linguists). It can be said that

the faculty to analogize is innate, and language faculty falls under this imperative. More generally one can

say that we have here a relation between a model and a copy, and the copy can be quite blurred (or in other

words: mapping knowledge from one domain into another: Itkonen and Haukioja 1996: 137). In language it

has been quite comfortable to espouse the proportion (Paul 1880), but one needs the other end also (from

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747912

Page 3 of 11

10. Analogy: The Warp and Woof of Cognition : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:35 PM

Hermann 1931 to Winter 1969; see Anttila 1977: 72–6). 3 The ability to copy is enormously powerful, as

seen in language learning, or any learning, in the social context (Short 1999). Thus it is no wonder that

linguistic signs can also be copied and modified, lifted out of their original contexts. Note that in science we

end up with theoretical terms that are stipulated (“not seen”), whereas in language the new items pop out

immediately for approval (whether they get approved is another matter). There is no difference in the

analogical structure.

2 Transposition and Analogy

All cognition is based on relation and order, that is, gestalts. Gestalt is ultimately based on relations,

because it is the total relation of relations (Weinhandl 1960: 132, 166). 4 The most crucial concept in all this

is that of transposition: gestalts are invariants of transpositions, similarities of correspondences. “For

whatever one would mean by gestalt, the transposability of gestalt has to be taken as its essential property,

as already von Ehrenfels tried to show” (Kaila 1945: 65; below Kaila's emphasis is eliminated). “We have

verified that one essential side in symbol function is connected with intermodal transposability [and add

analogical extension]. Here one sees the connection of symbol function with gestalt formation” (pp. 65–6).

One has to assume that “the organs forming the gestalts reach the invariants contained in the multiplicity of

receptor excitation. I call the principle in this assumption the principle of invariance of perception.

‘Invariants’ mean here the unifiedly recurring relations in the different areas with a multiplicity of excitation”

(p. 86). “Thus the process of consciousness is from beginning to end a search for invariants, finding them,

and partly also creating them” (p. 89; my translation). Transposition holds the key role (Weinhandl 1960:

406–12) in connection with invariance, isomorphy, language, natural law, and constancy. When a factor

(structure-point) varies, matching covariance of other factors produces invariance (p. 406). Transposition is

crucial for our experience, memory, and cognition, and it presupposes recognition (p. 407), since we have to

recognize a structure in other materials. In the symbolic mode (verbal, graphic, numerical) we get

categorization in that we assign facts to recognized concepts, thereby getting an isomorphic representation

for the object (one meaning-one form; Shapiro 1991). Transposition thus provides (in immediate experience)

an isomorphic replica (Kaila 1945: 407); similarity is again central (cf. pp. 206, 408). 5 If we could not

experience similar structures or figures or facts, we would really have nothing. It is constancy that gives

another match to invariance of objects (as experienced or perceived), and thus fills another aspect of

phenomenal representation (p. 411). This is how we get a constant external world and a chance for a fixed

starting-point (e.g., for analogy). This is, again, how we can further explain the human mechanism for

fiction and hypostatization. More particularly, we see the immediate reasons for the necessity of

epiphenomenal meanings (grammatical meaning, metaphor, riddles, and the like).

We have again reached the concept of analogy, although it might not be apparent to all linguists. The step

from transposition to analogy can be best exemplified by the fact that our perception grasps the world

through complex formation (perception of wholes) and abstraction (Dörner 1977). Our concepts are

relational stencils that classify incoming information. Gestalts and super-signs are characteristically of the

structural kind, since their composition can be transposed into other media or units (p. 74). A structure or

configuration of relations establishes a gestalt, and since the relations are not contained in the parts of the

whole, but obtain between them, a gestalt is indeed “more than the sum of its parts” (p. 75). As for

transposability, Dörner states that it is nothing but the possibility of interchanging the components of a

structure with others. The gestalt principle is simply a structure of empty slots for the components (pp. 75–

6). Finally, Dörner shows how argument from analogy consists in (i) matching a known domain of reality with

another structurally similar one, in (ii) abstracting the structure, the gestalt from the known, and (iii) putting

this structure over the unknown area. “An argument from analogy is an attempted transfer of a structure

from one domain of reality to another” (p. 81). This is critical analogy, but the same holds for what we know

from language, and this is what philosophers, psychologists, and scientists have come up with time and

again. Gestalt principles give a solid philosophical foundation for analogy and inference in general. Analogy,

as used in traditional linguistics, is perfectly valid. Whatever its limits are, they cannot be rectified or

eliminated by denying the notion altogether, since it is all we have (cf. Holyoak and Thagard 1995: 148,

262). Further, it is no use trying to formalize it for “proper” explanation, as linguists wanted to do during

and since the 1970s. Harald Höffding (1924: 26) already analyzed concepts like analogy and symbol as

correlative concepts expressing a mutual relation. Höffding took synthesis and relation as correlative

categories , exactly like continuity and discontinuity, resemblance and contrast (difference). These are

fundamental categories; analogy is formal (cf. Itkonen 1994: 52), and totality real .

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747912

Page 4 of 11

10. Analogy: The Warp and Woof of Cognition : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:35 PM

In short, similarity is the most important holistic process in mental life. It is the basic axiom for all cognition,

and since we are dealing with similarity we have here a continuity agent between percepts, parts, and even

sciences (Höffding 1924; Anttila 1977: 5). Models of formal logic fail, because analogy does not fit into their

either-or tallies. Note that Leibniz already pleaded for topology, analysis situs; such notions have been

rediscovered in cognitive linguistics (see Heine and Traugott, this volume), curiously tied with metaphor, not

analogy.

In fact, rationality is a process of becoming from indeterminate vagueness, and thus change is a primary

aspect of reality (Shapiro 1991: esp. §5). The use of symbols involves their further determination and this

necessarily leads to change. Language use is largely problem-solving in communication (including its many-

faceted context) and thereby falls under rationality, since one cannot solve problems without any reason.

Language use falls likewise under emergence phenomena in which structure and becoming cannot be

separated. And indeed, analogy is the main force in language structure, and it is an agent of change. What

has confused many is that similarity/analogy works both in structure, giving it cohesion, and as a process

for problem-solving (Itkonen 1991: 313–20, 1994: 44; Itkonen and Haukioja 1996: 136, 142). This is

traditionally well understood, although since the 1960s both aspects have been badly blurred, apparently

both on purpose and by accident. 6

3 Analogy and Metaphor

The two crucial factors in any relevant conception of cognition, namely similarity and contiguity, come out in

(cognitive or otherwise) linguistics as metaphor and metonymy. Of course, today the Peircean terms iconicity

and indexicality also abound, particularly the former (see Anttila and Embleton 1995: n.9). Now there is no

end to the literature on metaphor, and often no indication is given that the notion was quite well understood

before. 7 Something like this was bound to happen, since Chomsky's denial of metaphor as a relevant thing

and the rejection of analogy by the whole school was startling incompetence. Of course formalists and those

sympathetic to them say that explaining everything with metaphor does not explain anything. 8 In principle

there is no difference between metaphor and analogy for our purposes. 9 It was a mindless coup in linguistic

theory to abolish analogy in the face of its long tradition in linguistics and philology (although its

reintroduction in Optimality Theory has now made some linguists revolutionary).

The problem with analogy seems to be the following: against the nice dualities like similarity d contiguity,

metaphor d metonymy, icon d index, abduction d induction, which all match, analogy is a mixed bag; it

mixes the two columns, as it were. Since the context (the warp) is so crucial, I have called analogy an

indexical icon (Anttila and Embleton 1995: 98). The contiguity aspect of analogy (nearness) is also

emphasized by Coates (1987: 337). Locality is again central in cognitive psychology and linguistics, and this

is true of analogy also, since it requires orientation as a crucial anchoring factor (Vaught 1986: 324–5; Haley

1997). In current cognitive theory the prototype gives the orientation point, and then metaphor carries it

further. Note that this is exactly what the Ancient Greeks had in their paradigm (example) and analogy

(proportion). The paradigm is the indexical part.

Cognitive linguistics has got great mileage out of body metaphors, here too ignoring earlier work (Anttila

1992a: 66). The body is also central in the problem of foundational analogy, or incongruous counterparts, in

orientation (right and left) (Vaught 1986: 314–16, 325–6; see also Haley 1997 for foundational analogies).

Finally, whether we accept it or not, it is constructive to remember Vaught’ s plea for an analogical relation

between the immediate and the dynamic object and the two interpretants (1986: 321). Analogy joins them,

but also keeps them separate (distant). Meaning is thus perfectly analogical. 10

One defense of metaphor in cognitive linguistics is that semantic field theory would not be able to shift

between fields, whereas metaphor offers such a possibility (cf. Anttila 1992a: 66). This is strong camouflage,

since analogy was always taken as giving this ability (Höffding 1924: 72; and others). 11 There is no

difference between analogy and metaphor in this context, and we have seen how analogy performed exactly

this service (transposition; Dörner 1977).

4 Leaking Syllogisms

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747912

Page 5 of 11

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: