LINGUIST List 7.1794

Thu Dec 19 1996

Disc: Analogy

Editor for this issue: T. Daniel Seely <>


  1. Geoffrey S. Nathan, Re: 7.1784, Disc: Analogy
  2. Stirling Newberry, Re: 7.1784, Disc: Analogy
  3. benji wald, Analogy: synchronic and historical

Message 1: Re: 7.1784, Disc: Analogy

Date: Wed, 18 Dec 1996 09:25:18 -0600
From: Geoffrey S. Nathan <>
Subject: Re: 7.1784, Disc: Analogy

	While it is interesting to see the old debates about the role
of analogy rehashed I think that one of the issues that is crucial has
only barely been touched on--the importance of form-meaning pairs as a
base for analogies. While a facetious example of "I see the barn red"
was just posted (in which rose-colored glasses figured) I would like
to return to the question of why such analogies do not occur. It
seems to me that those who argued for semantic anomalies have the
upper hand here, because we can easily create such analogies given
changes in world conditions.
	Consider, for example, a world similar to that proposed in The
Lathe of Heaven in which some person has the ability to change the
color of objects merely by choosing to see them that way. In such a

1)	I saw the barn red. 

becomes grammatical in just the same way as 

2)	I painted the barn red.

	A change in the MEANING of `see' (or perhaps in the state of
affairs in the world) renders the analogy fine. No special
grammatical mechanisms are required, and the sentence is perfectly
grammatical. The anomaly in the current reading of the analogized
sentence rests in the meaning of `see', not in a violation of some
structural conditions. So, while analogy may indeed depend on some
grammatical conditions (such as whether a word is a noun or a verb)
what rules out the analogy in the famous video case is semantic, not

Geoff Nathan
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Re: 7.1784, Disc: Analogy

Date: Wed, 18 Dec 1996 12:31:28 -0500
From: Stirling Newberry <>
Subject: Re: 7.1784, Disc: Analogy

>Date: Wed, 18 Dec 1996 15:12:40 +0200 (EET)
>From: Esa Itkonen <>
>Subject: analogy
>	If one is asked to predict or formalize WHICH analogies are formed
>in fact, diachrony and synchrony have to be treated separately.
>First, grammaticalization (for instance) is a diachronic process, and all
>linguistic changes are unpredictable. Therefore, if grammaticalization is
>viewed as analogical (with respect both to reanalysis and to extension),
>then any theory that would claim to be able to predict which analogies
>will be made would be falsified eo ipso.

Not so the theory would merely have to show constraints to the process
of gramaticalization. That is results that are impossible to get to no
matter how long it went about getting them.

Analogical does not imply unbounded.

>	Second, as far as synchrony is concerned, it IS possible to
>formalize analogy. To do so, one only needs to retain the traditional
>notion of analogy as a matter of form-AND-meaning (i.e. analogy-1) and
>reject the homonymous neologism that treats analogy as a matter of form
>(i.e. analogy-2). Only if one adopts analogy-2 (i.e. the aberrant, formalist
>notion of analogy) can one claim - and with perfect justification - that
>'analogy does not exist', and support this claim with the
>ungrammaticalness of 'I saw the barn red' and the like.

This, while symbolically true, is also unconnected, in a formal
logical sense, with the original question: is analogy a process used
in gramaticalization - or is there a process which is even analogous
to analogy used in gramaticalization. The formalization argument shows
that there is a process which produces results *similar* to analogy
used in gramaticalization - but this does not prove that there is a
process that is actually similar to analogy itself.

I am not a supporter of attempts to come up with funny sounding
analogies as disproofs - they don't disprove anything. The question
is: is there a process or set of processes at the deep structure level
which skelentonizes gramatically correct utterances and then does
symbolic substitution (that is the process of drawing an analogy) or
is this process or set of processes going on at some other level.

To date in this discussion I have not even seen the design of an
experiment that gets at this problem - whether using aphasia subjects
or by some attempt to require response at a rate below which there can
be a search of memory at a deep structure speed, or by use of some
direct observation means to see the difference between people drawing
different types of "analogies".

Absent a set of experiments which are well defined redefining the word
analogy endlessly will not get us anywhere. With all due respect to
everyone, which is frankly in some cases a great deal and in others
considerably less, two different topics are being thrown into the same
discussion. They are not connected to each other by a reasonable
number of logical steps, and arguing the two of them is not
particularly productive.

The problem, again, is a strong claim/weak claim confusion. It is
counterproductive for UG supporters to make the strong claim that
analogy is invalid simply because it does not seem to function at the
deep structure level. Similarly no amount of proving that analogy
works as the final product of a persons process of creating an
utterance, deciding that they want to utter it, and then finally
uttering it, is going to answer the question of whether analogy enters
in at the deep structure level or at some later point in the
process. To claim otherwise is utterly ridiculous.

Instead I would underline that the question is a weak claim one: that
the process known as analogy at the language-as-used level does not
hold at the language-as-universal-mechanism level. One bears on the
other not in terms of existance - nor even in the types of analogies
that will and won't be drawn, since if analogy is a higher order
memory search and symbolic substitution it is dependant on the models
present in long term memory, but at what point in the existance of a
thought that the process of analogy alters the shape of the thought

>	Thus, it is trivially true that analogy-2 does not exist. (This was
>pointed out also by Farrell and Israel.) Less trivially, analogy-1 both
>exists and can be formalized. More precisely, the formalization of
>analogy-1 is not just a possibility; rather, it is a FACT (cf. Esa Itkonen &
>Jussi Haukioja: 'A rehabilitation of analogy in syntax (and elsewhere)',
>in: A. Kertesz (ed.): Metalinguistik im Wandel: die kognitive Wende' in
>Wissenschaftstheorie und Linguistik. Frankfurt a/M: Peter Lang, 1997,
>pp. 131-177). So far, the formalization (carried out in PROLOG) covers a
>fragment of English syntax (including wellknown counter-examples which
>thus turn out to be fictitious), but it can be (analogically) extended

To be blunt - just because I can get a computer to do something one
way does not mean that a human being does it that way - in fact at a
deep level it is just about a certainty that people don't do it the
way the computer does.

>	Now that the traditional notion of analogy (i.e. analogy-1) has
>been first distinguished from the formal 'analogy' (i.e. analogy-2) and
>then formalized, one sincerely hopes that diffuse a priori claims to the
>effect that 'analogy does not exist' will quietly subside.
>Esa Itkonen

Actually no they won't becuase the question was not whether analogy
exists, but whether it exists with respect to Universal Grammar. A
large number of things which exist elsewhere don't exist with respect
to UG - which is why it was come up with - becuase it was noticed that
language often mirrors the set of things that can be gramaticalized
more than it resembles the real world.

Stirling Newberry
Boston, Massachusetts
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: Analogy: synchronic and historical

Date: Thu, 19 Dec 1996 21:24:55 -0800
From: benji wald <bwaldHUMnet.UCLA.EDU>
Subject: Analogy: synchronic and historical

To begin with, I have to say that to a large extent I saw the discussion of
"analogy", as the discussion of a word used in introductory linguistic
texts for an extremely general concept, and, indeed, as Esa Itkonen perhaps
implies (I'm getting back to him later), in the context of historical

It is, of course, true that in synchronic linguistics, this term, in all
its seeming generality, was used to comment on Chomsky's theories of the
"Syntactic Structures" type by some linguists, Bazell, I think, was one, to
"explain" why "colorless green ideas sleep furiously" is "grammatical".
Chomsky characteristically disagreed with such suggestions, instead of
saying, perhaps, that what Bazell et al were calling analogy he was calling
"surface structure" independent of "meaning" (at that point in the
development of his argument). Later he substantiated this by pointing out
that analogy was too general a concept, because (I'm not sure of this)
"Joan (I'm paraphrasing) is easy to please" and "...eager to please" have
the same (?) surface structure, but there is no "it's eager to please J"
parallel to "it's easy to please J" = "to please J is X".

So, where does "analogy" come in?

Is it a "surface" structure thing, or a "deep structure" thing? Obviously
not surface structure, so why use the term at all , if it needs reference
to other concepts which can do very nicely-thank-you without it, i.e.,
concepts which contain whatever insights "analogy" had, but are more
precise (or suggest something more precise) -- and fragment the basic
insight of "analogy" into various distinct things.

Of course, "deep structure" has gone through various metamorphoses and
processes of fragmentation since then, so that "analogy" can make a
come-back, according to some discussants, with constraints like:

"Avoid conceptual anomaly."

This constraint, in its wording and implicit references, plays on the
enormous discipilinary disagreement about whether some (or all?) "failures"
of analogy have an origin in "semantics" rather than "syntax", where the
latter retains its structuralness, "independent" of "meaning" (where
"meaning" = "semantics" is a separate component/ module from syntax =
abstract arrangement principles for lexical categories).

"He saw the barn red" was also dragged into this as an example. Is it THE
SAME as "to please Joan is eager"?

I don't know (or haven't given it enough thought), but I can see the same
kind of humor (at least) in:

Yeah, well maybe Joan painted the barn red, but before she was even born I
actually saw the barn red. (STRESS the VERBS to get some linguistic
unsophisticate to smile.)

as in,

the two-th time I told you that, I wasn't upset, but when we got to the
three-th time, I began to lose my patience. (well, wait for the four-th
time and then things will be all right, but watch out for the five-th time,
unless you proceed directly to the six-th time).

But I don't see anything humorous about:

Yeah well maybe it's easy please Joan, but it's certainly not eager.
 to please Joan is easy, but it (what? to please her?)
is certainly not eager

Instead I see dumbfounded expressions trying to figure out what the basis
for the intended humor or wisecrack is, and maybe even winces of the
blackboard-scraping reaction type.

Next, the HEDGE "reasonable" was drawn into discussion, I forget who
started that, maybe Chomsky himself. For example, Patrick Farrell

"More generally, the universal grammar stance, according to which
formal principles of linguistic structure constrain analogy,
could be said to be interesting if one could point to
semantically/conceptually REAONABLE analogies that are not drawn
AND corresponding explanatory principles of linguistic form that
are independently motivated, technically viable, and non-vacuous."

"Interesting" suggestion, but it is not clear whether "reasonable" is
invoked here for informal inter-subjective agreement among discussants, or
whether it is the basis for a new discipline, or for a revamping of the
basis of Linguistics (with a capital "L") itself, in which case "rational"
might be a NEAR-paraphrase (in context, "rational" would be offensive if
used in its informal sense, cf. "that's not a RATIONAL suggestion".) In
any case, "reasonable" is a contentious, rhetorical word unless its intent
is further specified. Would "principled" be a less inflammatory

So far, all I'm saying is that "she saw the barn red" and "to please her is
eager" seem to be weird in quite different ways, not surprisingly, since
they are based on quite different "analogies" (one with properties of
"easy" and one with those of "paint"). Quite a while ago I even claimed
that "she saw the barn red" was "grammatical" as far as I was concerned (I
wouldn't exclude it from the grammar of English, and can even interpret it,
as could quite a few other discussants at that time, including in ways I
grasped but hadn't thought of beforehand.) I wouldn't say the same for "to
please her is eager". It is not interpretable to me by any stretch of the
grammar -- which is not to say that I insist the violation is "syntactic"
rather than "semantic", I don't know (and I'm not sure I even care for
purposes of this discussion). I'll stand with, we can do better than just
say "analogy". To this I think everybody would agree, and that's the end
of our agreement. So how can we do better?

(NB, maybe even worm brains are capable of analogy at a certain level of
abstraction, and that human language descends from such capacities -- but
does our insight into language and what linguists mean/meant by "analogy"
end there? If so, we have a lot more important things to do than discuss

Next, Esa Itkonen's distinction between analogy-1 and analogy-2. This
really got my attention, because it's a metalinguistic comment, just like
the one I started out this message with ("it's about the WORD "analogy").
And I also enjoyed it because it injects historical linguistics into the
discussion, though I have less respect for the distinction between
synchronic and historical linguistics than he does, and that may be why I
am not sure I agree with his comments. Also, I'm not sure I agree because
I'm not sure I understand the full impact of what he is saying. What
particularly got me thinking was his statement:

"First, grammaticalization (for instance) is a diachronic process, and all
linguistic changes are unpredictable. Therefore, if grammaticalization is
viewed as analogical (with respect both to reanalysis and to extension),
then any theory that would claim to be able to predict which analogies
will be made would be falsified eo ipso."

The good point in this quote is that it acknowledges the historical fact
that all analogies, or whatever you want to call them, that could be made
in a language at a particular time are not (necessarily?) made. Later we
see that they are possible, to the extent that we are right that they
reanalyse and/or extend a certain pattern. Too bad he didn't give even one
example. Maybe we should use as simple an example as the English change
(ca. 1700- and-something) from "the bridge is building" to "the bridge is
being built", to avoid even greater complications with most
(non-morphological) examples. I guess in terms of analogy, it has to do
with things like:

 they *built* the bridge: the bridge *was built*
 = they*'re building* a bridge: x

i.e., extension of the pattern of passivisation in conjunction with the
(much) older
innovation "be + V-ing" as a progressive, at some point reanalysed as
something like an "aux(iliary)". Originally, "V-ing" was gerundive, "they
are on=in building a bridge" with no passive, simply "the bridge is on=in
building" (NB transitive V-ing takes subject or object relations to V with
equal equanimity, and has throughout the relevant period, cf. the famous
"shooting of the hunters", and even "the bread's still (a-)bakin'", the
last leads to complications digressive to the point, so don't think about
it too hard).

The example is good to bring home the point that somehow the analogy
escaped speakers earlier, so even though they said "the bridge was built
(at last)"and "they're building the bridge", they didn't think of saying
"the bridge is being built (at last)", but kept saying "the bridge is
building (at last)". Of course, one can claim, and probably has, that the
*reanalysis* of "they're building a bridge" did not come about until it
also included "the bridge is being built", and/or that that's the (only?)
way that we really know there WAS a reanalysis. In that case, and maybe in
any case, I fail to see the difference between analogy-1 (historical) and
analogy-2 (synchronic). So "analogy" seems to be as "weak" (in the usual
sense of being too "powerful", explains everything and therefore nothing)
an explanation in historical linguistics as in synchronic linguistics. And
no doubt historical linguists, esp of the generative persuasion, have made
this point. I doubt Esa simply lists historical innovations, shows their
"analogical source", and then moves on to another case and does the same
thing ad infinitum.

So I anticipated what I see as the "bad side" of Esa's comment. I don't
see any fundamental difference between "analogy" as used in historical
linguistics and as used in synchronic linguistics.

Furthermore, I think the statement that " all linguistic changes are
unpredictable" misses the point. The intent, I think, is that we cannot
predict when or even whether a certain change will happen, which would be
like doing rain dances (in my opinion -- and sometimes it WILL rain). But
I don't think that's the fundamental business of historical linguistics. I
think rather that the fundamental business is to discover the constraints
on change, which change are possible and which are not, and under what
conditions. That is, there are a limited number of ways that some piece of
language, call it X, can change from one state to the next; let's stick to
"syntactic constructions" for the sake of argument, though it is obviously
the case in phonology as well (I'm making the limits to avoid the larger
range of "metaphorical" changes possible for lexical items, perhaps
unmanageable for a field like linguistics or anything else, not that we
can't find certain patterns even there). And, as dialect variation and
historical trees etc show, there is more than one way to change, but that
does not mean change is unprincipled, or even unpredictable WITHIN the
limits of possible changes. (Either this or that CAN happen, but NOT

I would, then, strongly disagree with Esa's comment if it were meant to
adopt a "know-nothing" stance with regard to constraints on change, and I
would ask, what's the point of studying change? Is it simply an
antiquarian interest? (There is plenty of that in historical linguistics --
are all languages related? -- but that's not all there is to it, and that's
NOT what bringing in the concept of analogy was about. And furthermore,
you can't resolve the antiquarian interests without the principled
knowledge that this change can happen but not that one, so your
reconstruction is wrong.)

Answer next year? Hope it's a good one, year that is, well, the answer too.

- Benji
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue