Minimal pairs

As any well educated GGer knows, there is a big and important difference between grammaticality and acceptability (see here and here) (don’t be confused by the incessant attempts by many (especially psycho types) to confuse these very separate notions (some still call judgment tasks ‘grammaticality judgments’ (sheesh!!))). The latter pertains to native speaker intuitions, the former to GGers theoretical proposals. It is a surprising and very useful fact that native speaker’s have relatively stable converging judgments about the acceptability (under an interpretation) of linguistic forms over a pretty wide domain of linguistic stimuli. This need not have been the case, but it is. Moreover, this capacity to discriminate among different linguistic examples and to comparatively rate them consistently over a large domain has proven to be a very good probe into the (invisible underlying) G structure that GGers have postulated is involved in linguistic competence. So for lots of GG research (the bulk of it I would estimate) the road to grammaticality has been paved by acceptability. As I’ve mentioned before (and will do so again here), we should be quite surprised that a crude question like “how does this sound (with this meaning)?” has been able to yield so much. IMO, it strongly suggests that FL is a (relatively) modular system (and hence immune to standard kinds of interference effects) and FL is a central cognitive component of human mental life (which is why its outputs have robust behavioral effects).  At any rate, acceptability’s nice properties makes life relatively easy for GGers like me as it allow me/us to wallow in experimental crudity without paying too high an empirical price.[1]

That is the good news. Now for some bad. The fact that acceptability judgments are fast and easy does not mean that they can be treated cavalierly. Not all acceptability judgments are equally useful. The good ones control for the non-grammatical factors that we all know affect acceptability. The good ones general exploit minimal pairs to control for these distorting non-grammatical factors. Sadly, one problem with lots of work in syntax is its lack of fastidiousness concerning minimal pairs. Let’s consider for a moment why this is a problem.

If acceptability is our main empirical probe into grammaticality and it is understood that acceptability is multivariate with grammaticality being but one factor among many contributing to acceptability, then to isolate what the grammar contributes to an acceptability judgment requires controlling for all acceptability effects that are not grammatically induced. So, the key factor behind the acceptability judgment methodology is to bend over backwards to segregate those factors that we all know can affect acceptability but cannot be traced to grammaticality. And it is the practicing GGer that needs to worry about the controls because speakers cannot be trusted to do so as they have no special conscious insight into their grammatical knowledge (they cannot tell us reliably why something sounds unacceptable and whether that is because their G treats it as ungrammatical).[2] And that is where minimal pairs come in. They efficiently function to control for non-grammatical factors like length, lexical frequency, pragmatic appropriateness, semantic coherence, etc.  Or, to put this another way: to the degree that I can use largely the same lexical items, in largely the same order to that degree I can control for features other than structural difference and thereby focus on G distinctions as the source for whatever acceptability differences I observe. This is what good minimal pairs do and so this is what makes minimal pairs the required currency of grammatical commerce. Thus, when they are absent suspicion is warranted, and best practice would encourage their constant use.  In what follows I would like to illustrate what I have in mind by considering a relatively hot issue nowadays; the grammatical status of Island Effects (IE) and how minimal pairs correctly deployed, render a lot of the argument against the grammatical nature of island effects largely irrelevant. I will return to this theme at the end.

To get started, let’s consider an early example from Chomsky (1964: Current Issues). He observes that (1) is three ways ambiguous. It has the three paraphrases in (2).

1.     John watched a woman walking to Grand Central Station (GCS)

2.     a. John watched a woman while he was walking to GCS

b. John watched a woman that was walking to GCS

c. John watched a woman walk to GCS

The ambiguities reflect structural differences that the same sequence of words can have. In (2a), walking to GCS is a gerundive adjunct and John is the controller of the subject PRO.[3] In (2b) a woman walking to GCS is a reduced relative clause with walking to GCS an adjunct modifying the head woman. In contrast to the first reading, a woman walking to GCS forms a nominal constituent. In the third reading a woman walking to GCS is a gerundive clausal complement of watch depicting an event. It is thematically similar to, but aspectually different from, the naked infinitive small clause provided in (2c). Thus, the three way ambiguity witnessed in (1) is the product of three different syntactic configurations that this string of words can realize and that is made evident in the paraphrases in (2).

Chomsky further notes that if we WH move the object of to (optionally pied piping the preposition) all but the third reading disappears:

3.     a. Which train station did John watch a woman walking to

b. To which train station did John watch a woman walking

Given what we know about islands and movement, this should not be surprising. Temporal adjuncts resist WH extraction (CED effects), as do relative clauses (CNPC). Clausal complements do not. Thus, we predict that movement of (to)which train station from (1) with structures analogous to (2a,b) should be illicit, while movement from (1) with a complement structure like (2c) should be fine. Thus, we expect the movement to factor out all but one of the readings we find with (1). And this is what occurs.

Note that this explanation of the loss of all but one reading coincides with the fact that all but the third paraphrase in (2) resists WH extraction:

4.     a. *(To) which train station did John watch a woman while he was walking (to)

b. *(To) which train station did John watch a woman who was walking (to)

c.  (To) which train station did John watch a woman walk (to)

Thus the reason that (3) becomes monoguous under WH movement is the same reason that (4a,b) are far more unacceptable than (4c).  This argues for the fact that unacceptability wrt these sentences ((un)acceptability under an interpretation for (1) and tout court with (4)) implicates a syntactic source precisely because other plausible factors are controlled for, and they are controlled for because we have used the same words, in the same order thereby varying only the grammatical structures that they realize.[4] 

We can go a little further, IMO. Note the dependent measure in (4) is relative acceptability with (4c) as baseline. But, note that in this case the items compared are not identical. The fact that we get the same effects in (1)/(3) as we do in (2)/(4) argues that the data in (4) reflects structural differences and not the extraneous vocabulary items that differ among the examples.  Furthermore, the absence of the two illicit readings in (3) is quite clear. It is often asserted that acceptability judgments are murky and can be trivially enhanced/degraded by changing the WHs moved or the intervening lexical items. Perhaps. Here we have a case where the facts strike me as particularly clear. Only the event reading survives the extraction. The other ones disappear, which is exactly what a standard theory of islands would predict. This, I believe, is typical for well constructed minimal pair cases: the dependent measure will often be the availability of a reading and, interestingly, the presence/absence of a reading is often more perspicuous for native speakers than is a more direct relatively acceptability judgment.

I would like to consider one more case for illustration. This involves near minimal pairs rather than identical strings. What the above Chomsky case provides evidence for (rather clear evidence IMO) is that G structure matters for extraction. It shows this by factoring out everything but such structure as the relevant variable. However, it does not factor out one important variable: meaning. Sentence (1) has three readings in virtue of having three different syntactic structures. So, the argument cannot single out whether the relevant factor is syntactic or semantic. Does the difference under WH movement reflect the effects of formal grammatical structure (syntax) or of meaning (semantics)? As the two vary together in these cases, it is impossible to pull them apart. What we need to focus in on this are structures that are semantically and formally the same. And this is very hard to do. However, not quite impossible. Let me discuss a (near) minimal pair involving event complements.[5]

Consider the following two sets of sentences:

5.     a. Mary heard the sneaky burglar clumsily attempt to open the door

b. Mary heard the sneaky burglar’s clumsy attempt to open the door

c. What₁ did Mary hear the sneaky burglar clumsily attempt to open t₁

d. What₁ did Mary hear the sneaky burglar’s clumsy attempt to open t₁

6.     a. Mary heard someone clumsily attempt to open the door

b. Mary heard a clumsy attempt to open the door

c. What₁ did Mary hear someone clumsily attempt to open t₁

d. What₁ did Mary hear a clumsy attempt to open t₁

The main difference between (5) and (6) is that the latter tries to control for definiteness effects in nominals. What is relevant here is that both sets of cases distinguish the acceptability of the the c from the d cases with the former being judged better than the latter using standard Sprouse like techniques (i.e. we find a super additivity effect for (5c)/(6c)). Why is this interesting?

Well note that the near minimal pairs have a common semantics. Perception verbs take eventive internal arguments. These can come in either a clausal ((5a,c)/(6a,c)) or a nominal ((5b,d)/(6b,d)) flavor. The latter should show island effects under movement given standard subjacency reasoning. In sum, these examples control for semantic effects by identifying them across the two syntactic structures yet we still find the super-additivity signature characteristic of islands. This argues for a syntactic (rather than a semantic) conception of islands for this is the one factor we varied in these near minimal pairs, the meaning having been held constant across the a/b and c/d examples.

Howard Lasnik is constantly reminding those around him how important minimal pairs are in constructing a decent grammatical argument. He notes this because it is not yet second nature for GGers to employ them. And he is right to insist that we do so for the reasons outlined above. It allows us to make our arguments cleaner and to control for plausible interfering factors. Minimal pairs is the nod we give to the fact that acceptability judgments are little experiments with all the confounds that experiments bring with them. Minimal pairs is the price we pay for using acceptability judgments to probe grammatical structure. As Chomsky noted long ago in Syntactic Structures these sorts of judgments can really get you deep into a G structure very efficiently. They are an indispensible part of linguistic theorizing. However, to do their job well, we must understand their logic. We must understand that theories of grammar are not theories of acceptability and that there is a gap between acceptability (a term of art for describing data) and grammaticality (a term of art for describing the products of generative procedures). Happily the gap can be bridged and acceptability can be fruitfully used. But jumping that gap means controlling for extraneous factors that impact acceptability. And that is how minimal pairs are critical. Deployed well they allow us to control the hell out of the data and zero in the grammatical factors of linguistic interest. So, let’s hear it for minimal pairs and let’s all promise to use them in all of our papers and presentation from now on. Pledges to do so can be sent to me written on a five dollar bill c/o the ling dept at UMD.

---

[1] Jon Sprouse and friends have shown roughly this: that crude methods are fine as they converge with more careful ones.

[2] If undergrads are to be believed virtually all unacceptability stems from semantic ill-formedness. If asked why some form sounds off you can bet dollars to doughnuts that an undergrad will insist that it doesn’t mean anything, even when telling you what it in fact means.

[3] Which, you all know, does not exist but is actually a copy/occurrence of John due to sidewards internal merge. And yes, this is an unpaid political announcement.

[4] Note the use of ‘grammatical’ rather than ‘syntactic.’ These cases implicate structure but as syntactic structure and semantic interpretation co-vary we cannot isolate one or the other as the relevant causal element. We return to this with the second example of a minimal pair below.

[5] This is joint work that I did with Brian Dillon. He did most of the heavy lifting and deserves the lion’s share of the credit. It is published here.

Islands are not parametric

Every now and then the world works exactly the way our best reasoning (and theory) says it is supposed to work. Pauli (whose proposal was en-theoried by Fermi) postulates the neutrino to save the laws of conservation and momentum in the face of troubling data from beta decay (see here) and 20 years later the little rascalino is detected and 40 years later Nobel prizes are collected. Ditto for the Higgs field, but this time the lag was over 40 years from theory to detection, and again a Nobel for the effort. These are considered some of the great moments in science for they are times when theory insisted that the world was a certain way despite apparent counter-evidence, and patience plus experimental ingenuity proved theory right. In other words, we love these cases for they comes as close as we can hope to get to having proof that we understand something about the world.

Now linguistics (I am almost embarrassed to say this, but I really don’t want to be painted as hyperbolic) is not quantum mechanics. It’s not even Newtonian mechanics (if only!). But every now and then we can snag a glimpse of FL by considering how its theories, and the logic that allows us to develop these, yield unexpected validation of its core tenets. The logic I refer to in this case is the PoS. Many denigrate its value. Many are suspicious of its claims and dispirited by its crude reasoning. They are wrong. Today I want to point to one of its big success stories.  And I want to luxuriate in the details so that the wonders of PoS thinking shine clearly through.

The Athens participants, to a person, touted islands effects as one of GG’s great discoveries. It is such for several reasons.

First, they are non-obvious in the simple sense that the fact that islands exist is not cullable from inspection of live text. Listen all you want to the speech around you and you will nary spot an island. It is the classic example of a dog that doesn’t bark. It is only when you ask informants about extractions that the distinctive properties of structures as islands shines through.

Second, it takes quite a bit of technical apparatus to even describe an island. No conception of phrasal structure, no islands. No understanding of movement as a transformation of a certain sort, no islands. So, the very fact that islands are linguistic objects with their own distinctive properties only becomes evident when the methods and tools of GG become available.

Third, and IMO the most important point, islands are perfect probes into the structure of FL and were among the first linguistic objects to tell us anything about its structure. Why are they perfect probes? Because if islands exist (and they do, they do) then their existence must be grounded in the structure of FL. Let me say this very important point another way: the fact that there are islands cannot be something that is learned (in the simple sense learning as induction from the ambient linguistic data). That islands exist and constrain movement operations cannot possibly be learned because there is no data to learn this from in the PLD. Indeed, absent inquiries by pesky linguists interested in islands, there would be virtually no data at all concerning their islandish properties, neither in the PLD nor the positive LD.  Island data must be manufactured by hard working linguists to be seeable at all. Thus, island phenomena are the quintessential examples of linguists acting like real scientists: they are unobvious, based on factitious data, only describable against a pretty sophisticated technical background and pregnant with implication for the fine structure of the principle object of inquiry, FL.  Wow!! So islands are a really, really, really big deal.

And this sets up why this poster by Dave Kush, Terje Lohndal and Jon Sprouse (KLS) is so exciting. As the third point above notes, what makes islands so exciting is that they are the perfect probes into the structure of FL precisely because whatever properties they have cannot possibly be learned and so must reflect the native structure of FL. To repeat, there exists no data relevant to identifying islands and their properties in either the PLD or the positive LD. But this absence of data implies that island effects should not vary across Gs. Why? Because variation is a function of FL’s response to differential data input and if there is no plausible data input relevant to islands then there cannot be variation. However, for many years now, it has been argued that different Gs invidiously distinguish among the islands. In particular, since the early 1980s several have argued that the Scandinavian languages don’t have islands. KLS shows that this is simply incorrect. Here’s how it shows this.

The argument that Scandinavian Gs are not subject to islands rests on the claim that they do not display island effects. What are island effects? These are the native speaker judgments which categorize extractions out of islands as unacceptable. Thus, a native speaker of English will typically judge (1) as garbage, this being typically annotated with a “*”.

(1)  *What did the nominee hear the rumor that Jon got

In contrast to this judgment concerning (1) in English, the claim has been that speakers of Swedish and Norwegian accept such sentences and do not give them *s. Maybe they give them at most a ?, but never a *. So the basis for the claim that Scandinavian Gs are exempt from islands is that native speakers do not find them unacceptable. The conclusion that has been drawn is that Gs vary wrt to islands. But this is impossible given PoS reasoning that implies that Gs could never so vary, ever. As you can see, there is an impasse here: theory says no variation, empirics say there is. Not surprisingly, given the low value much of the field assigns to PoS style theoretical “speculation” (aka logic), much of the field has concluded that there is something deeply wrong about our theory of islands. [1]  

Now for a long time this is how things sat. The obvious answer to this empirical challenge is to deny that Scandinavian acceptability judgments are good indicators of grammaticality wrt Scandinavian islands. In other words, the fact that Scandinavian speakers accept movement out of Scandinavian islands does not imply that the structures so derived are grammatical. More pithily, in this particular case (un)acceptability poorly tracks (un)grammaticality.

Let me quickly say two things about this general point before proceeding lest too big a moral gets drawn. First, it is important to remember that within GG, conceptually speaking, “ungrammatical” is not a synonym for “unacceptable,” despite the practice within linguistics of confusing them (especially terminologically).  “Acceptable” is a descriptive predicate of data points, a report of speaker judgments. “Grammatical” is a theoretical predicate applied to G constructs. GGers have been lucky in that over a large domain of data the two notions coincide. In other words, often, (un)acceptability is a good indicator of (un)grammaticality. However, one obvious way of dissolving the Scandinavian counter-examples above is to suggest that the link between the two is looser in this particular case. In other words, that the English data is more revealing of the underlying G facts than is the Scandinavian data in this case.

Second, the fact that this may be so in this case does NOT mean that acceptability is always or generally or usually a bad indicator of grammaticality. It isn’t. First, as Sprouse and Almeida and Schutze have demonstrated in their various papers, over a large and impressive range of data, the quick and dirty acceptability judgment data is a very stable and reliable kind of data.[2] Second, cross-linguistic research over the last 60 years has shown that the MLGs discovered in one language using acceptability data in that language generally map quite well onto the acceptability data gathered in other languages for the same constructions. In fact, what made the Scandinavian data intriguing is that it was somewhat of an outlier. Many many languages (most?) exhibit English style island effects. So, even if the assumption that acceptability tracks grammaticality might not be perfectly correct, it is roughly so and thus it is prima facie reasonable to take it to be a faithful indicator of grammaticality ceteris paribus.[3]

Ok, back to KLS. How does it redeem the PoS view of islands? Well, it provides a method for detecting island effects independently of binary (i.e. ok vs *) acceptability judgments. The probe comes from a battery of relative acceptability data gathered using the now well-known techniques of Experimental Syntax (ES). ES does acceptability judgment gathering more carefully than we tend to do. Factors are separated out (distance vs island) and their interaction (more) carefully compared. Using this technique one can gather relative acceptability data even among sentences all of which are judged quite acceptable. Using this method, the empirical signature of ungrammaticality is a super additivity (SA) profile apparent when you cross distance and structure.[4]

With this machinery in place, we are ready for KLS’s big find. KLS applies this SA probe to English and Scandinavian and shows that both languages display a SA profile for sentences involving extraction from islands.[5] Where the languages differ then is not in being responsive to islands but in how speakers map an island violation into a binary good/bad judgment. English speakers judge island violations as bad and Scandinavian speakers often judge them as ok.[6]

Conclusion: Scandinavian obeys islands restrictions and this is visible when we use a more sensitive measure of G structure than ok vs *. In other words we clearly see the effects of islands in Scandinavian when we look at their SA profiles. 

We should all rejoice here. This is great. The PoS reasoning we outlined above is vindicated. Gs do not differ wrt their obeisance to island restrictions and these are still excellent probes into the structure of FL. I, of course, am not at all surprised. The logic behind the PoS is impeccable. I have the irrational belief that if something is logically impossible then it is also metaphysically impossible. Thus, if some G difference cannot be  learned due to an absence of any possibly relevant data, all Gs must be the same. This is as close to apodictic reasoning as we are likely to find in the non-mathematical sciences, so I have always assumed that the KLS results (or some other indication that Scandinavian obeys islands) must exist. That said, who can’t delight when logic proves efficacious? I know I can’t!

However, KLS is actually even more interesting than this. It not only vindicates the logic of the GG program against a long-standing apparent problem, but it also reshapes the domain of inquiry. How? Well, note that English and Scandinavian still differ despite the fact that both show island effects. After all, the former assign *s to sentences that the latter assign at most ?s to. Why? What’s going on? KLS offers some speculations worth investigating regarding how non-syntactic conditions might affect overall (i.e. binary) acceptability judgments.[7] I personally suspect that this overall measure is affected by many different factors including intonation (and hence old/new info structure) lexical differentiation and a host of other things that I really can’t imagine. Sorting this out will be hard if for no other reason that we have not really concentrated much on how these myriad effects interact to provide an overall judgment. What KLS shows is that if we are interested in how speakers construct an overall judgment (and, off hand, it is not clear to me that we should be interested in this but I am happy to hear arguments for why we would be), then this is where you need to look for “variation.” Why? Because they have provided very good evidence that islands are NOT parameterized, which is the conclusion that elementary PoS reasoning leads to.

Diogo Almeida, has a nice paper (here, and here)[8] that does similar things along these ES lines for extraction out of Wh-islands in Brazilian Portuguese (BP). The paper uses ES methods to probe not only sensitivity to islands but also to compare how sensitive  different constructions are to island restrictions. The paper compares Topicalization and Left Dislocation showing that both induce island effects (as On WH Movement would lead us to expect). It also shows that BP here differs in part from English, raising further interesting research issues. I for one would love to see ES applied in comparing Topicalization, Left Dislocation and Hanging Topics in languages where one finds case connectivity effects. How do these do wrt islands? I can imagine a simple predication wherein case connectivity being a diagnostic of movement implies that Hanging Topics do not exhibit SA effects in island contexts. Is this so? I have no idea.

Diogo’s paper provides one further service. It provides a nice name for SA-without-unacceptability effects. Diogo dubs them “subliminal islands.” Diogo argues that the simple existence of subliminal effects has interesting implications for how we understand SA effects. He argues, convincingly IMO, that we expect subliminal effects if grammaticality is one component of acceptability, not so much if we take a performance view of islands. The paper also has a nice discussion of the conceptual relation between acceptability and grammaticality that I recommend highly.

Ok, time to end. I have two concluding points.

First, IMO, this kind of work demonstrates that ES can tell us something that we didn’t know before. Heretofore, ES work has aimed to either re-establish prior results (Jon’s work was aimed at showing that island effects are real) or to defend the homeland against the barbarians (Jon & Co arguing that informal methods are more than good enough much of the time). Here, KLS and Diogo’s paper show that we can resolve old problems and learn new things using these methods. In particular, these papers show that ES methods can open up new questions even in well-understood domains. In particular, I believe that both papers show that ES has the potential to reinvigorate research into the grammatical structure of islands. I mention this because many syntacticians have been wary of ES, thinking that it would just make everyone’s cushy life harder. And though I generally agree, that ES methods do not displace the more informal ones we use, these papers demonstrate that they can have real value and we should not be afraid (or reluctant) of using them.

Second, go out and celebrate. Tell friends about this stuff. It’s science at its best. Yipee!

---

[1] The impression I have (even after Athens) is that such speculation is considered to be just this side of BS.  At the very least it really plays (and can play) no serious role in our practice. I may be a tad over-sensitive here.

[2] The fact that such a crude method of data collection has proven to be so reliable and useful raises an interesting question IMO: why? Why should such a dumb method of gathering data work so well? I believe that this is a function of the modularity of FL and the fact that Gs play a prominent role in all aspects of linguistic performance. More exactly, how Gs are used affects performance but performance does not affect how Gs are structured. Thus Gs make an invariant and important contribution to each performance. This is why their effects can be so easily detected, even using crude methods. At any rate, whether this diagnosis is correct, the fact that such a crude probe has been so successful is worth trying to understand.

[3] Note that this implies that the challenges to the universality of islands based on Scandinavian data were reasonable, even if ultimately wrong.

[4] See Jon’s thesis here, or many of his other writings for a simple explanation of these effects.

[5] KLS discusses Wh islands, Noun complement islands, subject islands and adjunct islands. It does not address relative clauses. However, Dave Kush has told me that they have looked at these too and the results are as expected: Scandinavian shows the same SA profile for RCs as for all the others.

[6] There is a kind of judgment one often hears from linguists which this methodology questions. It’s “the sentence is not perfect but it is grammatical.” The ES methodology suggests that one treat this kind of judgment very gingerly for it might indicate the underlying effects of the relevant G distinction, not its absence, as is typically concluded. Note that the quoted judgment above runs together “acceptable” and “grammatical” in a not wholly coherent way. It relies on the assumption that informal judgments concerning degree of unacceptability is a reliable probe of the binary grammatical/ungrammatical distinction. In other words, it assumes that grammaticality is binary (which it may be, but who knows) and that the severity of an acceptability judgment can reliably indicate the grammatical status of a structure. This assumption is challenged in the KLS paper.

[7] Sprouse’s thesis already demonstrated how e.g. D-linking could affect overall acceptability (i.e. the binary judgment) without inducing an SA signature.

[8] Same paper but for formatting. Latter is published version for citation purposes.