I am a data scientist who has a question about collocations based on a book I am reading. The book is "Feature Engineering for Machine Learning: Principles and Techniques for Data Scientists" by Alice Zheng, Amanda Casari.

In chapter 3, regarding working with text data and natural language processing, the authors state:

"Collocations are more meaningful than the sum of their parts. For instance, “strong tea” has a different meaning beyond “great physical strength” and “tea”; therefore, it is considered a collocation. The phrase “cute puppy,” on the other hand, means exactly the sum of its parts: “cute” and “puppy.” Thus, it is not considered a collocation."

I'm struggling to understand the difference between the two examples they've given. Why exactly is it "cute puppy" is not a collocation whereas "strong tea" is? I've asked a friend who is an English language teacher, and he seems to think both are collocations. Are the authors incorrect in this instance?

The Book can be found here (links to relevant section).

  • 2
    These differences of opinion often point to reliance on different definitions. Here are two: ODO (habitual juxtaposition) and Cambridge (emergent meaning). I can see merit in both, but I'll leave it to the professional linguists to answer definitively.
    – Lawrence
    Sep 20, 2018 at 15:20
  • 1
    wikipedia gives the meaning as habitual juxtaposition. When a collection of words has a meaning different from the sum of its parts, that's an idiom. in "strong tea", "strong" isn't being use to mean "physical prowess", but a word having different meanings, and a particular one being understood in a particular context, is not a remarkable phenomenon. "strong tea" has exactly the meaning of its constituent words, when those words are understood in context. Sep 20, 2018 at 15:34
  • The example in the Cambridge definition @Lawrence linked might be instructive: By that definition, we can't freely substitute synonyms into collocations the way we can into regular phrases. So "adorable puppy" and "sweet puppy" sound fine and almost synonymous with "cute puppy", but the phrases "powerful tea" and "bold tea" don't sound quite natural and wouldn't easily substitute for "strong tea".
    – 1006a
    Sep 20, 2018 at 18:18
  • Actually, bold tea sounds like advertising. meleztea.com/products/bold-tea But you can't mix adjectives with more than one main meaning compared to adjectives that have basically one meaning.
    – Lambie
    Sep 20, 2018 at 18:59

2 Answers 2


Things either collocate or don't but it is not "scientific" at all. It is about semantics and discourse (a probable arrangement or existence of terms in a particular field or context). It is the expectation of finding one word with another in a particular context.

So (to give a very easy example) in a mathematical treatise, "shit" would not collocate with the discourse or the context of formal mathematics. However, change that context to mathematicians in a chat room, and one might say to another: "That's just a shit argument, mate."

The example above is collocation at the level of discourse. But anyone analyzing a text or speech has to define their context: at the lexical level, at the sentence level, in speech, in writing, in a particular field, etc. It's all up to the person analyzing some linguistic form or phenomenon. But, the definition(s) should cohere with a basic understanding of what a "word" (as found in a dictionary) actually is. Please read on...

Collocate just means "to be located in the same place as" or "to share a location".

That said, the book you are reading seems to have a somewhat different take on the matter, and, is describing what I would call a "set phrase", or a "cliché".

And the example of "strong tea" is not the best because the meaning "great physical strength" is not the meaning of "strong" in strong tea. The pertinent meaning here is: extreme,intense, rich in some active agent (Merriam Webster). This is borne out by considering: "weak tea", its "natural opposite" or adjectival antonym, if you will. Those two authors are not linguists per se. There is no "rule" that only first meanings of words can be used to compare "side-by-side collocations" of meanings of words in cliché phrases such as "strong tea" and "cute puppy".

A collocation linguistically speaking is most definitely not limited to an adjective + a noun as in "cute puppy" or "strong tea". One has to define one's one's playground. Those authors have limited their definitions to a very circumscribed usage of the term collocation: an adjective modifying (or describing) noun or one "set" next to one. And seem also to have based their analysis of "first meanings". That's the stumbling block for me here.

They are using the meaning of collocation as "side by side" as in to set or arrange in a place or position, especially: to set side by side (Merriam Webster). And the broader meaning in that same dictionary is: intransitive verb: to occur in conjunction with something. Ergo, not necessarily side by side. But in any case, I believe they have made a semantic error by assuming that only first meanings of adjectives count. They are actually comparing non-comparable items.

I think one might say that in general parlance "cute puppy" and "strong tea" are indeed collocations in general speech in English (even if you aren't Australian [joke]). One would very much expect to find those adjectives side by side with those nouns. What would, however, not collocate in that sense might be: "strong puppy" and "cute tea".

[Please note: my bias here,if I have one, is that I am a translator, and deal with this collocation issue in my daily grind all the time.]


There are two issues here:

  1. What do we mean by collocation? And,
  2. Regardless of what we ultimately decide about 1., is there some relevant difference between cute puppy and strong tea?

Let me start with 2.

Cute puppy vs. strong tea

Yes, there is a difference, but I agree with Lambie (see his or her answer on this page) that the book you are using has misidentified it. Contrary to what your book says, and as many people have already pointed out, the meaning of strong tea is exactly the sum of its parts, just like in the case of cute puppy. That's not what's different about them. Rather, what's different about them is the following.

In the case of cute puppy, many different words aside from cute could be used to express an idea similar to cute puppy: adorable puppy, likeable puppy, delightful puppy, precious puppy, ... Cute may be the most popular word nowadays to use in this context, but there is nothing unnatural or strange about the other ones.

In contrast, in the case of strong tea, all plausible alternatives end up sounding awkward and unnatural: ?powerful tea, ?concentrated tea, ?intense tea, ... Mind you, if someone uses these alternatives, they will be understood. But then they are likely to be asked, 'You mean, strong tea?' Because when we want to say that a tea is powerful/concentrated/intense, the word we are supposed to use is strong. (See e.g. here.)

What is going on here is something similar to various other linguistic phenomena, such as set phrases, clichés (as Lambie mentioned in his or her answer), fixed expressions, and idioms, but it is not exactly any of those. Strong tea is not an idiom, because an idiom should have a figurative meaning, and strong tea does not. It is not a fixed expression, because a fixed expression should have 'a more specific meaning than the expression itself', whereas strong tea means exactly what it seems to mean, no more and no less. It is not a cliché, because a cliché is 'overused to the point of losing its original meaning or effect, even to the point of being trite or irritating', none of which applies to strong tea. And it is not a set phrase, because a set phrase 'is a phrase whose parts are fixed in a certain order, even if the phrase could be changed without harming the literal meaning.' After all, there is nothing strange about saying That tea is strong; in other words, strong can function both attributively and predicatively with tea. But that is not so with actual set phrases. We routinely say bad luck, but ?my lack is bad would be acceptable only under special circumstances.

It is only to be expected that linguists would seek out a name that would capture the special relationship between words like strong and tea. The name they have settled on is collocation.

Which brings us to question 1:

What do we mean by collocation?

I have found several scholarly sources that discuss the history and multiple senses of this term. Below I will use Seretan's Syntax-Based Collocation Extraction.

In short (pp. 13-14),

2.2.3 Collocation vs. Co-occurrence

As indicated in Section 2.2.1, the term collocation has originally been used in a broad sense, for describing the general event of recurrent word co-occurrence. This purely statistical view was later contrasted by a more restricted, linguistically motivated view, which explicitly states that the items in a collocation are syntactically related. The second view has recently gained in popularity, and some authors have suggested to use distinct terms to distinguish between the two understandings. More precisely, it has been proposed to use the term association or co-occurrence for the general statistical understanding, and to reserve the term collocation for the restricted understanding corresponding to the linguistically-motivated approach. For example, Manning and Schütze (1999) and Evert (2004b) state:

It is probably best to restrict the collocations to the narrower sense of grammatically bound elements and use the term association and co-occurrence for the more general phenomenon of words that are likely to be used in the same context (Manning and Schütze, 1999, 185)
In order to make a clear distinction between the two approaches to collocations,3 I refer to the distributional notion as cooccurrences (. . .) I reserve the term collocation for an intensionally defined concept (Evert, 2004b, 17).

The distinction between co-occurrences and collocations seems to be nowadays unanimously accepted (Bartsch, 2004), and will also be adopted in our work.

The above is preceded by five pages describing the history of the various meanings of collocation, and is followed by four pages explaining in more detail the meaning used in this work. Most of it can be read on google books, see the link I provided above.

  • "I reserve the term collocation for an intensionally defined concept." One reaches a threshold of tolerance about translation errors. Anyway, the point is not that "cute" has synonyms. The point is that the authors cited by the OP makes a semantic judgment that is not unjustified to define their idea of collocation.
    – Lambie
    Sep 21, 2018 at 13:27
  • @Lambie I believe intensionally has to do with intension vs extension, but I would have to look at the original work to be sure. Sep 21, 2018 at 13:53
  • [from that link you give] Maybe: “intension” indicates the internal content of a term or concept that constitutes its formal definition; and “extension” indicates its range of applicability by naming the particular objects that it denotes. In any event, this is not really relevant here.
    – Lambie
    Sep 21, 2018 at 13:58
  • @Lambie OK, you were right: it rather has to do with distributional vs intensional. Evert's work is free to download here. Also in Seretan: 'A statistical approach, called distributional, and a phraseological (linguistic) approach, called intensional' (p. 14). Sep 21, 2018 at 14:07
  • Right, thanks. I just do not think it matters here.
    – Lambie
    Sep 21, 2018 at 14:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.