Language and thought in user-centered design

Deutscher, G. (2010). “Does language shape the way you think?”  New York Times Online Edition. Retrieved on October 4, 2010.

This article summarizes the history and the current status of the linguistic relativity hypothesis or the idea that “language shapes thought.” This idea comes in several different formulations. The version on which Deutscher’s review focuses is a cross-linguistic claim that speakers of different languages use different mental representations or processes as a result of having learned different languages as children.

A particularly strong version of the cross-linguistic hypothesis suggests that our native language provides us with a basic toolbox of conceptual representation, so if a concept or a mode of thought is not encoded in a given language, its monolingual speakers would be incapable of thinking about such a concept or thinking in such a mode.  For instance, linguistic relativity led to the prediction that people’s perception of differences in color should reflect the way their language encodes color.  This prediction was spectacularly disproven by the discovery that people whose language only has two color terms in its vocabulary perceive and categorize colors similarly to speakers of English, which has 11 basic color terms, plus many non-basic ones (Berlin & Kay, 1969; Rosch, 1972). Another empirical prediction from linguistic relativity, namely that speakers of languages that lack a specific grammatical form to mark counterfactual assertions (i.e. subjunctive, as in “If I were a rabbit) should have difficulty with counterfactual reasoning in general, also turns out to have little evidence to support it (Bloom, 1981; Au, 1983). In both of these very different domains, as well as in others, it seems that major differences among languages do not translate to major differences in their speakers’ cognition.

More recent cross-linguistic research on cognition uses sensitive measures to study much subtler mental habits that might be attributable to language differences. The theory behind this approach is that learning a language necessarily involves learning to “think for speaking” in that language (Slobin, 1996). Thinking for speaking involves attending to aspects of the world that a language we are learning designates as important. For instance, the rules of English grammar require that a sentence describing an event classifies it as a past, present or future event through obligatory tense marking, whereas many other languages, such as Mandarin Chinese, allow speakers to omit this information. By contrast, a language like Turkish requires any sentence about an event to mark the speakers’ source of information about the event (i.e. whether the speaker witnessed the event or heard it described by someone else), which is optional in English. Languages also differ in what they treat as required or optional information about relations and movement in space. Such differences should result in people who pay relatively more or less attention to various aspects of their experience. There are several kinds of evidence for this sort of cross-linguistic variation, in such domains as object categorization and grammatical genders, as well as spatial orientation and spatial vocabulary, among others.

I would argue, however, that the influence of language upon thought expresses itself not only in cross-linguistic differences, but also at the different ways of talking about one’s experience, even within the constraints of a single language. This is especially true when dealing with highly abstract domains of experience, which includes using complex interfaces. Much of our discourse on abstract topics is only possible through extensive use of metaphors, which involve systematic analogical mappings between abstract and concrete domainst (Lakoff & Johnson, 1983; 1999). By virtue of metaphor our understanding of abstract notions, such as data, is aided by concrete imagery of familiar physical objects, e.g. written or printed text.  In other words, when we talk about reading, writing, cutting, pasting, erasing or archiving digital data, we draw metaphorical inference from what we know about the concrete domain of writing on paper and apply it to the far less tangible domain of disembodied, computer-operated information. Although we are, for the most part, fully aware of the metaphorical nature of such talk, we are often at a loss for any other way of talking about abstract data manipulation. Metaphors are an important part of computer users’ experience part of computer-based product design.

Because the position and movement of objects in space is by far the most common source domain for conceptual metaphors, metaphorical language very often invokes a spatial frame of reference, an imaginary spatial scheme or a layout.  For instance, when we talk about looking ahead into the future, looking back into the past, getting sidetracked or stalled, and so forth, we invoke a spatial scheme in which we (or a company, a country, or the whole world) are moving forward in time, from the past into the future, so that covering a great distance means making great progress, being lost or disoriented means being confused about one’s goals and plans, going uphill means putting extra effort in, etc.  In the context of a computer interface, such mental imagery is buttressed by real graphical displays with which users interact in through visual input and motor actions.

I believe that the most basic challenge of effective metaphor use in interface design is achieving a degree of internal coherence between the graphical and the linguistic aspects of the interface. For example, describing file directories as “folders” leads a user to expect that they can navigate them by “opening/closing them” or “being inside/outside them”. If such terminology is combined with a graphical interface that does not allow for such navigation (e.g. a hierarchical graph where directories are represented as branches), it may give rise to confusion and/or increased information-processing load.  Beyond avoiding inconsistencies, a further challenge is to choose metaphors that are compatible with existing conventional metaphors of a given language. For example, in a language like English, where the forward and backward direction in space is used as a metaphor for the future and the past, respectively, email inbox navigation could be made easier by describing the earlier-date section of the mailbox as “previous” than by naming it “previous (as Yahoo mail does). Analogously, in Mandarin Chinese it would more felicitous to have newer emails to be ‘below’, rather than ‘above’ the older ones, for the same reason. But respecting language-specific and culture-specific metaphors ought to be merely the beginning of successful linguistic interface design. Languages allow multiple metaphors for the same abstract “target” domains. For example, the state of the national economy can alternatively be visualized as a ship (barely staying afloat), physical health of a person (on the verge of a relapse or still suffering from complications), or other relatively more concrete domains. Choosing the metaphor that is optimal from the users’ perspective, that enhances learnability and makes learning more intuitive can make a major difference in the overall success of a product.  Even more interesting are cases in which inventing novel metaphors is a part of product design.  This is probably a greater demonstration of the influence of language upon thought than anything produced in a psychological laboratory.