Friday, March 1, 2013

'Futureless' languages?

[If you've come to this post because you're wondering if a particular language is 'futureless' or not, skip down right to the bottom for a summary of the various points I make in this fairly lengthy post.]

This post follows on from my previous one about the work of Keith Chen, a behavioural economist at Yale. To recap, Chen's central hypothesis is that the language you speak may affect your savings behaviour, depending on how your language grammatically encodes statements about the future. My point in my first post was that he was drawing on research that suggests a link between language and thought, the hypothesis being that speakers of a certain language must pay attention to particular features of the world around them on a habitual basis because the language they speak makes it obligatory for them to mention such features.

My concerns about his work however, start when I look at Chen's application of these ideas to the tense systems of the languages in his study, and how these languages have been analysed for tense.

'Futureless' languages?

The term 'futureless' languages comes up in Chen's work (although he actually rejects it in favour of a different term) and deserves some clarification. Contrary to what the TEDx talk description says of 'languages without a concept for the future', in Chen's work, he cites Östen Dahl's definition of 'futureless languages":

Dahl defines “futureless” languages as those which do not require “the obligatory use [of grammaticalized future-time reference] in (main clause) prediction-based contexts”. In this framework, a prediction is a statement about the future that has no intentional component. (footnote 3, p. 1)
I'll explain why Chen needs to add the part about 'no intentional component' a bit later on. What is important to note here is that the term 'futureless' language does not refer to 'a language without a concept for the future', but rather to a language that does not obligatorily force its speakers to use some sort of grammatical marking e.g. a future tense inflection on a verb, when describing an event situated in the future. For this reason, Chen actually uses the more neutral term 'weak future-time reference (FTR) language' instead.

Mandarin Chinese, the main inspiration behind Chen's work, is a prime example of a 'futureless' language in this discussion, because it does not typically oblige its speakers to mark for futurity (most obviously because verbs are not obligatorily marked for future tense). However, this is not the same as saying that Mandarin does not have a concept of the future, or that it prevents its speakers from talking about events in the future. Mandarin speakers know that they can always use adverbs of time to specify if an event is going to take place in the future, e.g. 明天 míngtiān 'tomorrow'. It is simply not obligatory that speakers use such adverbs in order to construct grammatical sentences in Mandarin.

But the world isn't simply made up of 'futureless' and 'futured' languages, with nothing in between. And this analysis of Mandarin may not be entirely correct either, as we shall see later. If we look at the chapter on 'Future Tense' on the World Atlas of Language Structures Atlas, written by Viveka Velupillai and Östen Dahl himself, we find this observation:

"It is relatively rare for a language to totally lack any grammatical means for marking the future. Most languages have at least one or more weakly grammaticalized devices for doing so."

In fact, rather than thinking of two discrete categories, it might help (at this stage at least) to think of it more as a cline between a strong tendency to mark for 'futurity' and a weak tendency. In a study like Chen's, how does one draw the line between what is a 'futureless' and 'futured' language, which you may agree is a fairly subjective decision to make?

A European framework for a global study

Chen's solution to this question is to rely on a framework set out by the European Science Foundation’s Typology of Languages in Europe (EUROTYP) project which has criteria to determine what is a 'strong' future-time reference (FTR) language and what is a 'weak' FTR language. The guidelines state that the data collected for the project came from primary texts and responses to questionnaires developed by Dahl (see here for details), though I do not know how percentage of the data was textual and what was elicited through the questionnaire. Descriptive linguists will also be quick to point out the problems of relying too much on data collected using a questionnaire, since these often do not reflect actual language use.

Of greater concern is the inherently Eurocentric basis for comparing languages of the world in Chen's study. Chen states that, "to [his] knowledge, the EUROTYP project is the most extensive typological research program to study the cross-linguistic grammaticalization of FTR" (p. 9), but it does not change the fact that all the languages looked at are European. The survey itself cannot give a good idea of the kinds of grammatical categories one finds in languages all over the world. For instance, Kayardild, an indigenous language of Australia, can mark 'tense' on nouns - this may seem like a rare and extreme example, but it demonstrates that a set of criteria developed for only European languages will necessarily omit categories that may be found in other languages of the world.

To use an analogy from an article about Joe Heinreich's work on human behaviour by only looking at Western, Educated, Industrialized, Rich, and Democratic (WEIRD) people: it is like studying penguins while believing that one is learning insights applicable to all birds.

Wherefore the weather forecast?

This brings me to the analysis of English as a 'strong' FTR language. A number of people have raised similar examples to the one I'm about to use here to suggest that English is not a 'strong' FTR language:

(1) I am going to the shops now. (I'm on my way there as I say this.)
(2) I am going to the shops tomorrow. (I intend to go tomorrow.)

The main observation here is that in both sentences, the speaker is using a form that is described in English grammar as the 'present continuous' tense. However, while (1) refers to an event taking place at the time the speaker is uttering the sentence, i.e. 'the present' for the speaker; (2) refers to an event that might take place in the future.

[Note: Both (1) and (2) are perfectly grammatical and acceptable to native English speakers. One should not start thinking that (2) is wrong because it uses the 'present tense' to refer to the future! Think about the sentence I go to the shops every Friday. - you are using what is called the 'simple present' tense in English to describe an activity that you habitually do, not something that is happening at the very moment you utter the sentence.]

In both sentences, I could omit now and tomorrow and still have a grammatical sentence referring to an event in either the present or in the future. This shows English can actually be similar to Mandarin in not grammatically distinguishing between the present and the future on the verb, but through the use of an adverb of time.

So how does Chen get around this problem? He draws on an analysis by Bridget Copley, published in her 2009 book.

"Copley demonstrates that in English, “futurates” (sentences about future events with no FTR) can only be used to convey information about planned / scheduled / habitual events, or events which arise from law-like properties of the world." (footnote 9, p 4)

Using this argument, he then omits examples like the one I gave in (2), which describes a planned event. This is the main reason that has led Chen's study to focus on the language used in weather forecasts, since they do not have an intentional component to them. His own corpus study involved getting students to scour the internet for examples of weather forecasts in various languages and coding the verbs for FTR making.

This is one of my biggest problems with the research. The reason given for omitting the kinds of sentences that Copley lists is just not good enough. How can you just omit such a large chunk of data showcasing the way English speakers use English? What it looks like here is Chen trying to make English fit into the 'right' category so that it can then match his hypothesis / results.

On top of that, I've seen comments about how weather reports are not examples of typical or 'natural' speech, or how they may not even require the use of verbs, e.g. 'Tomorrow. Cloudy. Maximum temperature, 25 degrees." Chen notes that his study was confined to only languages that are widespread on the internet, but there is an underlying assumption that the 'weather report' as a genre exists for all language communities in his study. What about places that don't have a dependable meteorological service to announce the weather? Maybe a language might have a difference when talking about the weather that's likely later in the day and the weather that's likely in a few days' time?

Finally, even if Chen could convince me that weather reports are a reliable source of data for this cross-linguistic study, there is still one fundamental problem. If we consider the data that he has already collected (see Appendix B of his working paper), we find a set of 'verb ratios', referring to the percentage of verbs in a weather forecast about future weather which are grammatically marked as  future-referring. Some of these ratios are not 100% - this means that not all the verbs about future weather are grammatically marked as future-referring. Consequently, what we are looking at isn't obligatory future marking, but rather a tendency to do grammatical future marking. (Interestingly, Hungarian only has a 25% verb ratio, but it is still coded as being a strong FTR language.)

This goes against the basis for the hypothesis that an obligatory aspect of a language will make its speakers attend to it habitually every time they speak. Current research looking at the potential effects of language on thought is still in its nascent stage and it has necessarily been limited to testing hypotheses about obligatory features of a language. Once we start including a more general tendency to use a particular feature, the argument that it is the linguistic feature shaping how the speaker thinks becomes far less tenable: can we even tell if such a tendency is the result of the structure of a language affecting ways of thinking, or if it is something like a cultural norm that results in this particular use of language?

And if you're still in the mood...

As a final point, I should also point out that linguists rarely talk about 'tense' by itself, but 'tense, aspect and modality / mood' (TAM).  This is because languages rarely have markers just for 'tense', without conveying information about the other two. Simply put, 'tense' refers to the location of an event in time, 'aspect' refers to how the event relates to the flow of time, and 'modality' / 'mood' to the attitude of a speaker towards an event. (Click here for more info on the WALS site.)

For instance, the use of 'will' in English statements like 'It will rain' does not just convey information about when the event will take place (tense), but also the speaker's level of certainty that it will take place (modality/mood). Some Mandarin speakers might criticise Chen's study by saying that Mandarin too can convey future meaning on the verb, using the auxiliary verb 会 huì, which marks both futurity and certainty (much like English 'will') and is necessarily used when one wishes to convey both futurity and certainty.

In general, describing the TAM system of any language is very tricky. Two analyses of the same language by two different linguists may look very different, depending on such factors as the linguist's native language(s); the linguist's own theoretical orientation; the methodology used to collect the data; and the linguistic intuitions of the language consultant providing the data. At the cross-linguistic level, it is also very difficult to compare the TAM systems of different languages - it is not as simple as translating a sentence from one language to another and looking for a one-to-one correspondence. For instance, consider:

(3) Je suis à Lyon. 'I am in Lyon.' (I'm in Lyon now!)
(4) Je suis à Lyon pour 7 jours. 'I am in Lyon for 7 days.' (I intend to stay in Lyon for 7 seven days.)
(5) Je suis à Lyon depuis 7 jours. 'I have been in Lyon for 7 days.'

In French, all three sentences use the same present tense form suis (from the verb être 'to be'), while in English we see the present tense form am (from the verb to be) in only (3) and (4), not (5). Can we then say that the 'present tense' in French is the same as the 'present tense' in English?

As I mentioned in my last post, Chen's study is informed by more recent work on the relationship language and thought. However, there is one fundamental flaw in the application of this hypothesis to the data - namely, it starts to look at tendencies within a language rather than obligatoriness. The nature of the data collected in his study is also problematic, as is the reliance on what is a rather subjective way to code languages for tense. Admittedly, this lack of agreement when describing a language's TAM system and when comparing the TAM systems of different languages is something for linguists to work out, assuming consensus can ever be reached.

Given all that I've written, I actually do wonder what correlations Chen has found, since he seems to be convinced of many of the correlations he's found. I'd actually like to see more work on the potential impacts of one's language on behaviour, though perhaps not at the scale Chen has worked at, which is something I might write about soon. I might also spend one more post discussing some of the issues I have with the survey / census data Chen has used for his study.

[ADDENDUM (12/03/2013): If you've come here looking to see if a particular language is 'futureless' or not, there is no easy answer to this. Languages are not either 'futureless' or 'futured', and where you draw this subjective line depends on the criteria you choose. As in Chen's study, such criteria may also lead you to allow or ignore particular types of sentences from the language you are considering. Finally, you also need to consider the analysis of the language you are using, since grammatical analyses of languages can and do differ from scholar to scholar.]

1 comment:

  1. Hi Amos, I'm a recent graduate in language studies and remember this video being shown in my masters-level class. Shocking. This guy is an economist and obviously knows nothing about the grammar of English or Chinese at all. As you pointed out, future marking is obligatory in neither English nor Mandarin. Linguistics has no clout in the real world and that's why economists and psychologists can say whatever the hell they want and people will believe them even if there's scant proof for it.

    If you are interested in reading some scholarly articles about the experimental evidence relating to the Sapir-Whorf hypothesis and TMA in European languages, I can send you some from my previously completed course. I've also done some work on the differences in the use of the modal "will" in Inner-circle and Singapore English and presented at the ISB9. Email me at