This is Chomsky’s formulation of the Strong Minimalist Thesis, i.e. the working hypothesis underlying the Minimalist Program.
Language = the narrow syntax, i.e. the computational system that builds structure using items from the lexicon. Structure is built using simple, yet powerful, operations.
The narrow syntax builds the structure and then the structure goes off to other systems with which the narrow syntax shares an interface. These other systems (of which Chomsky assumes there are two, one semantic, the other to do with production (speech or signs etc.)) are systems in their own right. This means they can only ‘see’ and deal with certain things. Therefore, the narrow syntax must produce something that an interface system can ‘read’. Since there are two interface systems, the narrow syntax must produce structures which are legible to both interface systems. The interface systems are different and so require different ‘legibility conditions’ to be met. The narrow syntax thus faces a problem of how to satisfy these conditions simultaneously. The hypothesis being followed in the Minimalist Program is that the narrow syntax that we have is an optimal solution to this problem – it meets the conditions of the interface systems in (one of) the best possible ways.
The Strong Minimalist Thesis is not a doctrine – it is a working hypothesis. It’s a bit like the assumption that natural phenomena can be modelled by mathematics – you assume an ideal, see how far the natural phenomenon matches the ideal, identify the areas where it does and does not, then return for more hypothesising. By using the Strong Minimalist Thesis as a working hypothesis linguists (of course this only applies to linguists who make the same assumptions as Chomsky) can try to establish:
(1) The ‘legibility conditions’ of the interface systems.
(2) The extent to which the narrow syntax does meet these conditions in some ‘optimal’ way.
(3) The extent to which the narrow syntax does NOT meet these conditions in some ‘optimal’ way.
(4) Reasons for why language may be optimal/sub-optimal.
Hopefully that has shed some light on what is at first glance…and second, third, fourth glances etc…a pretty obscure little sentence.
“John is trapped in a room made of cheese on top of a mountain, and can signal for help only by eating the walls of the room such that a message is formed that can be seen from afar.”
The Performance Grammar Correspondence Hypothesis (PGCH) was put forward by John Hawkins (2004) as an explanation for why grammatical patterns and the frequencies of those patterns cross-linguistically are the way they are.
In essence, it says that linguistic constructions which are easier to process are more likely to be grammaticalised. Conversely, those which are harder to process are less likely to be grammaticalised. Furthermore, processing ease is hypothesised to underlie our preferences for certain constructions over others (where there is competition between constructions) in usage. Linguistic performance thus shapes the grammar.
Hawkins suggests that there are three principles behind the hypothesis. Simplifying horrifically:
Minimise Domains: this basically means make the distance between elements which go together syntactically and semantically as small as possible, e.g. if an adjective goes with a particular noun, put them as close together as possible.
Minimise Forms: this basically means make those elements mentioned above as small and as meaningful as possible, e.g. consider spoken English “I’mma be there” where “I am going to be there” has very much had its form minimised.
Maximise Online Processing: this basically means arrange those elements in such a way that a listener will be able to process the structure of what you’re saying in the most efficient way possible. This involves making structures easier to recognise but also avoiding potential misinterpretations of structure, e.g. “I looked the number up” – consider where you place the “up” as the object gets longer. “I looked the number of my friend who just moved in next door up” vs. “I looked up the number of my friend who just moved in next door”. If the object is going to be very long, it is better to put “up” straight after the verb so that the verb (and its idiomatic meaning) can be recognised sooner. When the object isn’t so long, as in “I looked the number up,” efficiency isn’t greatly affected.
Note that language users flout these principles all the time, e.g. for stylistic effect, and are not consciously aware of them.
Using these three principles, Hawkins’ theory makes some very strong and interesting predictions about the types of patterns found in the languages of the world, and about which patterns are more likely or unlikely to be found.
Reference
Hawkins, J. (2004). Efficiency and Complexity in Grammars. Oxford: Oxford University Press.
Scotttrembls raised an interesting point: “Do you know if there’s any evolutionary relationships between SVO, SOV and VSO languages? The evolutionary explanation never seems to come up- has this already been disporved or do we not understand enough about language evolution?”
There’s no evolutionary relationship in the sense that all SVO languages are genetically related and separate from all SOV languages etc. SOV, SVO and VSO languages are distributed throughout the world and are found in many different language families. But we know that languages can change types over a period of time so, in this sense, there are evolutionary paths from one type to another. For example, Old English and Latin are considered to be canonically SOV languages but their descendants (English and the modern Romance languages) are SVO languages. You might wonder when an SOV language stops being an SOV languages and becomes an SVO language. You have to bear in mind that these types refer to canonical structures, languages may use other structures at the same time but their use will be more restricted (although there are languages which many would characterise as being ‘free word order’ in which case they would not fall into any of these categories). For example, English is canonically SVO, but English uses other word orders for questions, focus structures etc. So the relative frequencies of particular structures within a language may change over time resulting in what appears to be a single type-switch.
Work on implicational universals (universals of the form which says if a language has structure X then it will have structure Y) initiated by Joseph Greenberg and taken further by John Hawkins makes some interesting predictions for language change. Greenberg’s formulations were for the most part tendencies, i.e. if X then Y significantly more often than not, but Hawkins aimed to identify exceptionless universals which often involved adding extra conditions, i.e. if X then, if Y then Z. This places more constraints on the forms languages can take but it also makes strong predictions about evolutionary paths of language change. The reasoning is roughly: if these formulations hold for the present situation and if there is no reason to assume things were any different in the past then languages can only move through allowed ‘states’ as determined by the strong implicational universals.
We understand enough about the evolution of some language families to be able to test these predictions and the predictions have been largely correct so far. However, many would not take this evolutionary picture to be an ‘explanation’, rather it is seen as a ‘description’ of the facts which allows us to characterise possible evolutionary paths of change and distinguish them from impossible ones. Given that each stage of a language is a present-day language in its time, it is still ultimately up to the explanations offered by formal and functional approaches to account for the form a language takes at any particular point in its evolutionary history.
When people study language typology they study the ways in which languages vary. However, it’s more than just saying different languages use different words or these languages use very similar sounds. We study the ways in which structural features of languages differ (or are similar) and many go further asking questions about what the limits of linguistic structural variation are.
English speakers will know that in a simple transitive clause we start with the subject followed by the verb followed by the object, e.g. ‘Bob (S = subject) likes (V = verb) pizza (O = object)’, i.e. English has typically SVO word order. But are there other ways of arranging such a structure? Logically there are six ways: SVO, SOV, VSO, VOS, OSV, OVS. The next question that a typologist will ask is how are languages distributed across these possibilities. As a null hypothesis we might think that we would expect to find roughly equal numbers of languages in each group, but this is not what we find at all. SVO and SOV account for around 85% of all languages (with SOV being a bit more frequent than SVO). Adding VSO languages brings the total to around 95% of all languages. The question is: why is the distribution of languages so skewed?
Three broad types of answers suggest themselves as candidates (at least to my mind):
1) It could be down to chance – the distribution of languages today may represent a highly skewed sample. If we came back in 1,000 years we might see a completely different distribution. This approach is obviously not taken by language typologists. There is certainly something interesting about the distribution which demands an explanation. To write the pattern off as due to chance would be to miss potentially significant insights into the ways languages are structured and shaped.
2) The formal aspects of human language (perhaps as encoded by Universal Grammar) constrain the surface forms that human languages can inevitably take, i.e. variation is not limitless though it may be apparently vast.
3) The functional pressures that act on speakers and hearers every time they use language will affect which forms languages will prefer to take, i.e. structures that are easier to say and to comprehend will be preferred and so will come to dominate amongst the languages of the world.
Given the great success of generative linguistics in the past few decades, (2) is a very popular approach to take. However, many intuitively feel that the approach in (3) is ultimately more satisfactory as an explanation. Personally I’m inclined to think that if we can explain surface variation in terms of performance preferences, this is a good thing because it means there is less for the formal approach to account for. Furthermore formal aspects of language are most often thought to be all-or-nothing affairs. If a grammar rules out a particular structure, that structure cannot exist, whereas if performance factors disfavour a particular structure, that structure will be either non-existent or rare.
But are (2) and (3) incompatible? You might think so given the distinction that’s often made between competence and performance. Many would not consider performance factors as relating to language proper – it is extra-linguistic and not something the linguist should be looking at. But the fact is that all the (overt) language that we use to construct theories of both competence and performance is being ‘performed’ in some way (either spoken or written or signed). I think there may well be limits on variation set by formal properties of human languages (which will account for some of the totally unattested structures) but others will be set by performance. And then maybe others that are to do with physics and biology more generally (here I’m thinking more of phonological typological patterns).
For now then it may be useful to adopt either (2) or (3) as an approach to language typology with the aim of seeing how far they can go, but always with the ultimate aim of putting the two together in the end for a more comprehensive account of why languages are the way they are.
The words for ‘one hundred’ in Indo-European languages exemplify an ancient sound change – the centum/satem split (the Latin and Avestan words for ‘one hundred’ respectively).
Proto-Indo-European has been reconstructed as having three ‘series’ of velar consonants – palatal velars, plain velars and labiovelars (*kj, *k and *kw respectively). However, in nearly all daughter languages, these three series collapsed into two. Languages on the centum-side of the split merged the palatal and plain series to be left with *k and *kw and those on the satem-side merged the labiovelars and plain series to give *kj and *k.
It was thought to be the case that the centum/satem split represented an ancient dialect division of Indo-European languages. Most centum-languages are found in the west whilst most satem-languages are found in the east. However, a number of problems with this view exist. Tocharian is a centum-language but is (or was) the furthest east of any Indo-European language. There is also evidence that some languages kept the three series distinct in certain environments longer than others, e.g. Luvian (an extinct IE language spoken in Anatolia).
This, plus other evidence, suggests that the centum/satem labels are better viewed as descriptive shorthands which are used to label mergers which occurred independently in various Indo-European daughter languages (although this view raises problems of its own as well!).
Hauser, Chomsky and Fitch (HCF) wrote a paper in 2002 entitled The faculty of language: what is it, who has it, and how did it evolve? It explicitly introduced an important distinction between the Faculty of Language in the Broad Sense (FLB) and the Faculty of Language in the Narrow Sense (FLN).
FLB is made up of all the components used in language, e.g. ability for vocal learning, ability to conceptualise, a sensori-motor system, a computational system (syntax) capable of recursion etc. FLN is a subset of FLB, i.e. everything in FLN is also in FLB but not vice versa. More specifically, FLN contains only those components of FLB which are uniquely human. HCF hypothesise that FLN only contains recursion and that all other aspects of language (i.e. what constitutes the rest of FLB) can be found in non-human species.
You might not agree with HCF in terms of what to assign to FLN, but the general distinction is both useful and perhaps necessary for tackling the issue of language evolution. The first question of HCF’s title (what is it?) involves filling in the FLB category – and obviously we need to know what it is we are trying to explain the evolution of! The second question (who has it?) involves deciding whether something belongs to FLN as well. If we put something in the FLN box, we are saying it is uniquely human. This can be tested (in principle – how easy it is to test is another matter!) and this is one place where interdisciplinary research between biology and linguistics can be very fruitful and informative. Basically, from an evolution of language perspective, we want to know what exactly (if anything) makes human language special. This then leads to the third question – we can ask how the uniquely human part of language evolved, i.e. we can ask how did FLN evolve rather than the much larger question of how did FLB evolve.
Of course, having done all that, FLN might turn out to be empty! Perhaps human language is not qualitatively different, only quantitatively different. But let’s wait to see what we find!
Grimm’s Law was highly successful at predicting the forms of Germanic words but there were many exceptions. However, the discovery of Verner’s Law showed that exceptions might just be apparent; sound change is still ‘regular and exceptionless’, you just have to look a bit closer for the regularities.
An example of one such ‘exception’ is father, from above.
Note how Latin pater (which retains the /p/ and /t/ from PIE) shows up as father in English. IE /p/ > Gmc /f/ as predicted by Grimm’s Law, but IE /t/ has not come out as /θ/ instead we find /ð/. More telling are examples of related words which have the predicted sound in some cases but not in others! For example, English birthandburden are both related but show different outcomes of what was historically the same consonant.
Karl Verner noticed, however, that the unpredictable instances correlated with the position of accent in PIE. Sanskrit retains much of the earlier accent system which Germanic has subsequently changed. Sanskrit pater retains the accent following the /t/. Verner noticed that Germanic results from Law A were voiced unless they were immediately preceded by an accented syllable (in which case they would be voiceless) – this is Verner’s Law. Subsequently many of these Germanic voiced fricatives became voiced stops (thus leading to birthandburden). Germanic also underwent an Accent Shift whereby the position of accent changed. This annihilated the conditions for Verner’s Law but left the results of it unchanged, i.e. the results went from being conditioned and predictable (phonetic) to unconditioned and unpredictable (phonemic).
Verner’s Law also helped to explain cases of /s/-/r/ alternations, so called rhotacism. That is /s/ was pronounced as [z] by Verner’s Law unless preceded by accent. This [z] sound then underwent rhotacism to become /r/. Old Latin shows flos-floris‘flower’, English shows was-were etc. Many of the results of Verner’s Law have, however, been lost through analogical levelling. Latin underwent levelling to yield flor-floris‘flower’ and many English dialects have levelled the was-wereparadigm (as has Modern German), i.e. you might hear people saying ‘we was, you was, they was’.
Verner’s Law was and is a fantastic example of how powerful the comparative method is when applied carefully and rigorously. It also gave a great confidence boost to the Neogrammarian Hypothesis which says that sound change is regular and exceptionless. But that is not the end of Verner’s Law…it’s still around in places. When you next come across execute and executor/executive, think carefully about where the stress falls and how you are pronouncing the <x> in those cases – you might just see Verner’s Law in action!
Grimm’s Law (also called the First Germanic Sound Shift) refers to changes which affected the stop consonants in what became the Germanic subgroup of the Indo-European language family (Proto-Germanic being the ancestor of all Germanic languages, i.e. Gothic, German, Yiddish, Swedish, Icelandic, Dutch, Afrikaans, Old English, English etc.). There are in fact three series of changes which changed some aspect of the articulation of the IE stop consonants whilst retaining the same number of distinctions (number of phonemes).
Law A: IE /p t k/ > Gmc /f θ x/
Law B: IE /b d g/ > Gmc /p t k/
Law C: IE /bh dh gh/ > Gmc /β ð γ/ (which later became /b d g/)
Exactly when this happened is not known but we can at least work when the Laws may have taken effect relative to each other, e.g. Law A cannot have happened after Law B because otherwise we would expect IE /b d g/ to show up as /f θ x/ in Germanic.
For example:
Latinpater > Englishfather, German Vater(German orthographic <v> is pronounced /f/)
Greektri > English three
Latincord- > English heart (English /h/ descends from earlier /x/)
Sanskritbhratar > English brother, German Bruder
These are standard but selective examples. Standard in the sense that you’ll find them in text books; selective in that we cannot simply look at one language and expect it to faithfully represent changes which happened hundreds of years ago. Latin, Greek and Sanskrit have undergone changes since Proto-Indo-European and English and German have undergone changes since Proto-Germanic. Modern German shows evidence of a Second Germanic Sound Shift which changed the Germanic stop consonants again! English did not undergo this change as it had already separated from the language that was to become German (compare threeanddrei,daughterandTochteretc.).
In the most recent versions of Chomskyan theory, Merge plays a central (if not the central) role. It is the only structure building operation available in the language faculty. This differs from earlier versions where Move was considered to be a separate structure building operation but Move has since been reconceived as a different type of Merge.
The Minimalist Program has reduced the architecture of the language faculty to the bare essentials (referred to as the ‘(virtually) conceptually necessary’ components). This means that there is a lexicon, a structure building computational system and (at least) two ‘interfaces’ with other cognitive systems (one semantic, the other phonological, broadly speaking). Items are selected from the lexicon and copied into the Numeration if they are to be used to construct a sentence. The Numeration is like a holding bay.
Merge, the structure building operation, takes two items and forms a set, i.e. X and Y merge to form {X,Y} (the theory also involves labelling the set but I’ll ignore that bit). Now, when I said above ‘a different type of Merge’ I did not mean that the operation itself varies, rather the difference between the types of Merge lies in where X and Y come from. There are three possibilities.
1) X and Y both come directly from the Numeration.
2) Either X or Y but not both comes directly from the Numeration.
3) Neither X nor Y come directly from the Numeration.
Option (1) is the type of Merge that gets structure building started. Without (1) there would be no structure.
Option (2) is the type of Merge called External Merge (EM) because one of the merging items is from the Numeration, i.e. comes from somewhere external to the structure that has already been built. Option (2) allows the structure built by option (1) to be extended by merging further items to already existing structure.
Option (3) is the type of Merge called Internal Merge (IM) and this is the current conception of movement. When an item moves, it is going from one place in the structure to another so the items that are merging both come from somewhere internal to the structure that has already been built.
Note that this assumes there is only one monolithic Numeration. If we wanted to merge two existing structures, we would have to add to the options above or modify our assumptions about the nature of the Numeration.
my prof uses two question marks in her slides if she’s going to ask us to give the answer ?? but it just ends up looking like she’s talking like this ?????
fun fact about languages: a linguist who was studying aboriginal languages of Australia finally managed to track down a native speaker of the Mbabaram language in the 60s for his research. they talked a bit and he started by asking for the Mbabaram word for basic nouns. They went back and forth before he asked for the word for “dog” The man replied “dog” They had a bit of a “who’s on first” moment before realizing that, by complete coincidence, Mbabaram and English both have the exact same word for dog.
on a similar note, a traditional Ojibwe greeting is “Nanaboozhoo”so when the French first landed in southern Canada they thought that they were saying “Bonjour!” Which is fucking wild to think about. Imagine crossing the ocean and the first people you meet in months somehow speak French.
One of my university linguistic textbooks had a list of identical or near identical words in Ancient Greek and Hawaiian. They were all coincidences of course, but it was wild.
The whole point of the list was showing how there’s a finite number of sounds human can produce, and it’s possible that multiple languages will end up assembling similar sounding words meaning roughly the same thing, and that therefore assuming that similar sounding words MUST be related is hubrys.
My favourite one of these is swedish and japanese KojaandKoya (小屋). Both are pronounced basically the same (ko-ya), and both mean “hut/small hosue”, but they are not related to eachother in the slightest. Japanese koya comes from Ko- (prefix meaning “Small”) and Yameaning “dwelling”. SwedishKojaon the other hand comes from Koj(a sleeping place on a ship), in turn from latin Cavea(enclosure/cage/cavity), probably via Dutch Kooi.
“Look at all these similar words! These languages must be related!” is one of the banes of linguistics,stg.
If you find one, or several, apparently similar words in two languages, there’s three reasons why that might be: they’re related; the word was borrowed (from one to the other, or both from a common source); or it’s random chance. And we’re generally really bad at appreciating just how many look-alikes can be down to random chance.
In case you were wondering how linguists actually prove the relationships between languages:
large numbers of look-alikes that have systematic sound correspondences, such that you can at least partly reconstruct what their ancestor word would’ve looked like. These sound correspondences have to be highly consistent, and they might reveal that words that don’t look much alike actually share an ancestor. (Example: Spanish hojaand French feuilleboth come from Latin folium,but sound completely different.)
correspondences that apply to “core vocabulary” words, such as numbers, names of body parts, and question words. Almost all the Indo-European langauges, for instance, have recognizeable words for “father” and “mother”: Latin pater/mater,Greekpater/meter,Sanskritpitr/matr,Old Irish athir/mathir,Avestanpitar/matar,and so on.
similarities in the grammatical structures; the Bantu languages share an unusual noun-class prefixing system, for instance, and the Semitic languages have a distinctive root-and-pattern morphology.
This also includes shared irregularities. Most Indo-European languages, for instance, have a very irregular verb to be; in other languages, to be is perfectly regular.
Sketchy comparisons include:
languages that are far apart in space, in dissimilar cultures, with no account of how they could have a common origin. (Migrations of the Bantus or the Indo-Europeans are well known, even if we can’t pinpoint all the details; contact between Ancient Greece and Hawaii would’ve required time travel.)
mass comparisons without identifying any sound correspondences (lookin’ at you, Joseph Greenberg)
comparison between modern languages that ignore well-known ancestor languages (don’t compare Spanish to Hindi, compare Latin to Sanskrit).
Youcancompare a reconstructed language to another reconstructed language - say, Proto-Indo-European to Proto-Semitic – but it’s a lot like what happens when you keep resaving an image with progressively worse compression. Reconstructed proto-languages, even ones with lots of supporting evidence, can still have gaps, mistakes, or plain old unknowns; if you pixellate an image badly enough, it could look like anything. So anything claiming to be a reconstruction of “Proto-World” or the “first human language” is almost assuredly some kind of bullshit, it’s just a question of what kind.
Wish you were enrolled in an intro linguistics class this semester? Starting a linguistics major and looking for extra help? Trying to figure out whether you should study linguistics and what comes after? Whether you’re just trying to grasp the basics of linguistics or you’re trying to construct a full online linguistics course, here’s a comprehensive list of free linguistics websites, podcasts, videos, blogs, and other resources from around the internet:
So I’ve been learning Korean and I am crazy upset with the romanizations I’ve encountered on vocab lists. They do not reflect the sound alternations that Korean characters have in different environments making them misleading.
For example:
Dog - 개 is romanized as “gae” and that’s all well and fine because ㄱ can sound like “g” and ㅐcan sound like “ae” but 개 is pronounced /ke/ because ㄱ is only voiced intervocalically. So why romanize it as g? Now in instances where it is pronounced /g/ it should absolutely be romanized as g because that’s the sounds you hear and should make.
This is just an obstacle I have run in to many times and I don’t see why we can’t just represent the sounds as they are said.
