The Mind and the Strangest of Languages

Leonard Talmy

Department of Linguistics and Center for Cognitive Science
University at Buffalo, State University of New York

Written interview conducted by Konrad Szczesniak
Appearing as: Mente e A Mais Estranha das Linguas
In: Ciencia Hoje (Science Today), vol. 46, n. . 275, pp. 88-10.
Oct. 2010. Brazil.

1. Your doctoral research on Native American languages led to some of the most significant ideas that you are famous for. When you chose your topic, did you get an impression you were striking gold?

My dissertation contrasted the ways that three languages represent concepts -- especially how they represent events of motion differently. The languages were English, Spanish, and Atsugewi, an American Indian language that was spoken near Mount Lassen in California. As a graduate student, I had spent a total of about a year there, mainly working with one of the last fully fluent speakers of the language. That dissertation and the papers I wrote based on it laid out a framework of semantic typology, especially motion typology, that became the first area I was well known for. The three languages I compared turned out to represent the three main patterns for representing events of motion seen in languages around the world. But it was sheer luck that Atsugewi -- which exhibits the rarest of the three patterns -- was the language I had found to work on, since many other California Indian languages have other patterns. But with Atsugewi as the language I did encounter, I was able to work out the three-way categorization that makes up the core of the motion typology. You ask if I knew I had struck gold. Well, if you’re referring to the language itself, yes, I knew it was one of the amazing languages of the world (see my answer to the next question). But if you’re referring to the typological framework I worked out, I’d have to say that I’ve never known how an analysis of mine might strike others.

2. What was the strangest language you’ve ever come across?

Of course, strangeness is in the eye of the beholder. To a speaker of Ojibwa, English must seem a stranger language than Atsugewi would seem. But I’ll assume you mean from the perspective of someone speaking a language like English. And from that perspective, Atsugewi is a good candidate. To be sure, most field linguists feel that the language they are working on is the most amazing of the world’s languages. I certainly feel that way about Atsugewi, but what I can do here is describe some of its characteristics and let you judge.

First of all, it is a "polysynthetic" language, where the verb consists of a root in the middle that is typically surrounded by a large number of prefixes and suffixes. One of these giant verbal words can be equivalent to the whole of a medium-sized sentence in a European language. Polysynthetic languages as a group are strange from an English perspective, but they can differ in their degree of polysynthesis -- and hence strangeness -- and Victor Golla, an accomplished Americanist, rates Atsugewi as a "10" on a scale of 10.

Apart from the sheer number of affixes around the verb root, Atsugewi is noteworthy for the meanings of these elements. To explain first with a comparison, in a sentence expressing a motion event, the verb in English typically refers to the manner of the motion, as in The bottle floated into the cave. In Spanish, the verb typically refers to the path of the motion, as in the corresponding sentence: La botella entro’ a la cueva flotando (The bottle entered to the cave floating). But in Atsugewi, the verb typically refers to the object in motion. In effect, the corresponding Atsugewi giant verb would express this event as: It-bottled into-the-cave due-to-the-current. Atsugewi has hundreds of verb roots that refer to one or another kind of object or material as moving or located. For example, -st’aq’- means `for runny icky material (e.g, mud, rotten tomatoes, guts) to move or be located’. And -swal- means `for a linear limp object susspended by one end (e.g., a sock hung to dry, a hunted rabbit suspended from one’s belt, a flaccid penis) to move or be located’.

At the same time, the slot just following the verb root must be filled by any one of some fifty suffixes that express the moving object’s path relative to some reference object. Thus, in the example above, Atsugewi would not actually use a phrase equivalent to "into the cave" as I first indicated, but would use a single suffix meaning `into a volumetric enclosure’. The language does have a noun for `cave’, but it would typically not include this noun in the sentence since the suffix is so specific. Among its fifty or so directional suffixes, some twelve subdivide the English preposition into, together with the noun it precedes, into an equal number of more finely distinguished geometric arrangements. For example, in addition to the suffix -ipsn for a path into a volumetric enclosure, the suffix -waw means `into a gravitic container’, as when acorns are put into a basket. The suffix -wamm means `into an areal enclosure’, as when a horse goes into a corral. The suffix -isp means `into an aggregate’, as when something is thrown into some bushes or a crowd of people. The suffix -ic’t means `into liquid’, as when a rock falls into a creek. And -ik’s means `horizontally into a solid substance’, as when an axe chops into a tree trunk.

And at the same time as all this, the slot just preceding the verb root must be filled by one of some two dozen prefixes that indicate the type of event that caused the motion of the object expressed by the verb root. Several of these cause prefixes express natural forces. For example, the prefix ca- means `as a result of the wind blowing on the object’. And the prefix cu- means `as a result of liquid current flowing against the object’. This is the prefix that would be used to express the phrase "due to the current" that I used above in the Atsugewi rendition of the bottle-floating example. Another group of cause prefixes distinguishes different ways that an instrument can be used to cause an object to move. One such prefix is for a linear instrument acting axially on the object, as in poking or prodding it. another prefix is for a linear instrument acting "circumpivotally" on an object, as in batting it. And one large group of cause prefixes distinguishes different body parts acting in particular ways on an object to cause it to move. For example, ma- is for one’s foot acting on an object, as in kicking it. The prefix phu- is for one’s mouth acting egressively on an object, as in spitting it out, while the prefix pri- is for one’s mouth to act ingressively on an object, as in sucking it in.

I’ll finish this sketch with examples of how Atsugewi picks a single element from each of the three categories I just discussed and puts them together in a single verb. Take a verb with the following prefix, root, and suffix, all of them cited above -- ca-st’aq’-ic’t -- which would in turn have to be surrounded by more prefixes and suffixes indicating tense, aspect, person, mood, voice, and the like. Literally, this verb would mean: `due-to-the-wind-blowing-on-it, runny-ikcy-material moved into-liquid’. It could be said, for example, if some deer’s intestines that had been lying on a river bank blew into the water as a result of a gust of wind blowing on it. For comparison, consider another verb that keeps the same root, but has a different cause prefix and directional suffix, and that adds a further suffix meaning `hither’: ma-st’aq’-ipsn-ukk. Literally, this verb means: `due-to-(your)-foot-acting-on-it, runny-icky-material-moved hither into-a-volumetric-enclosure’. And this verb could be said where an English speaker might say: "You’ve tracked mud into my house". As you can see, the two languages represent the same event by abstracting from it different components in different relationships.

3. Many scholars have used your work in support of the idea that language influences thought. Do you think we think through language?

Let me begin my reply by noting some distinctions. The word "thought" can be taken to refer to any form of structured ideational content -- or, more generally, experiential content -- that occurs over some portion of time in consciousness. By this definition, the word can apply to the semantic structure and content of any particular language. This includes the meanings of all the language’s basic linguistic elements -- the term for these elements is "morphemes" -- and how these meanings are put together in words and sentences. The word "thought" also applies to the structure and content of ideas that occur through a linguistic discourse such as a conversation, which, when present in the speaker at any point in the conversation, lead to what he says, and which in turn are evoked in the hearer by what the speaker says. "Thought", especially in its more generic sense of experiential structure and content, can also refer to what appears in an individual’s consciousness apart from linguistic discourse -- for example, while walking down a street experiencing the sights and sounds and reflecting on some incident. In an extension of this usage that now drops the requirement of occurrence through time, the word "thought" can further be applied to the structure and content of one’s beliefs and presuppositions, especially culturally based ones, pertaining to any domain, such as the social or the natural.

Your question about the relation between language and thought is of course in the tradition of the issues raised long ago by Benjamin Whorf. In this tradition, the emphasis is on structural correlations, rather than on ones of content. In terms of the framework just outlined, your question is to be interpreted as the relation between the structure of ideational content in a language and that in any of the other forms of thought in a speaker of the language.

In my reply to your second question above, the last example I gave shows that, at least at the level of a language’s basic elements of meaning -- its morphemes -- we humans can think very differently about phenomena. The question, though, is whether the differences we find at this morphemic level persist into higher levels of cognition. My impression is that, to a large extent, concepts represented below the sentence level can differ greatly across languages, but that the further you move toward the "paragraph level" of discourse and beyond, the more similar that the overall structure of conceptualizations across languages tends to look. And the structure of conceptualizations in speakers of different languages when in non-linguistic contexts -- say, as they walk along experiencing sights and sounds and reflecting on some incident -- may be quite comparable.

If so, this would mean that our cognition is structured in a way that permits relatively distinct organizations of conceptual content for different forms of thought. Of specific relevance here is that it permits different conceptual structures at the morphemic level and at the discourse level.

I suspect that some analysts’ motivation to adopt the Whorfian view of a strong connection between the generally narrower granularity of concepts in linguistic elements and the generally broader granularity of concepts in our ordinary mentations is in part due to a reluctance to believe that two such different conceptual structures could be maintained side-by-side in a single cognitive system. Even worse is that, in a conversation, these two levels with putatively distinct conceptual organizations would be in constant interaction.

My response to any such view is that, on the contrary, the norm in cognition is for differently organized systems to co-exist and interact. This is seen across the perceptual and motor systems, across the different sensory systems, and across the different subsystems within any given sensory system -- for example in vision, the distinct subsystems for processing form, color, and motion. Within the cognitive system of language, it is seen across the organizationally distinct subsystems of phonology, syntax, and semantics. It is therefore no stretch to posit that the organization of conceptual content at the morphemic level can be distinct from that at the discourse level.

Now, while the overall tendency may be for differences in conceptual structuring at the morphemic level across languages to fade away over larger swaths of discourse, some differences may persist. Dan Slobin has observed one such difference -- the best case I have seen. In my typology for a motion event, as I noted under question 2 above, the verb in a language like English typically expresses the manner of the motion, while the path is expressed (to simplify) in the preposition -- as was seen in The bottle floated into the cave. But in a language like Spanish, the verb typically expresses the path, while any manner -- if it is included at all -- is expressed in a gerund, such as flotando. But in Spanish-type languages, such a manner-expressing gerund is not grammatically required, and it is often omitted, in fact sometimes being stylistically awkward. But the manner in English-type languages is expressed by the verb, which cannot be omitted. So speakers of such languages regularly hear motion sentences in which the manner is necessarily expressed. This requirement for expressing manner seems also to have led to a great expansion in the sheer number of semantically distinct manner verbs in such languages -- greatly surpassing the number of semantically comparable verbs in Spanish-type languages. This recurrence and range of manner expression in English-type languages may attune its speakers to an awareness of the component of manner in motion events that persists to the discourse-level organization of thought.

This same pattern of an occasional correspondence amidst a majority of non-correspondence seems to hold as well between the conceptual structure of a language and the structure of beliefs in the cultural system of its speakers. To illustrate, David Wilkins worked on the language and culture of the Mparntwe Arrernte, an Australian Aboriginal group. He gathered together all that he could discern of the grammatical forms in the language that seem to reflect aspects of cultural structure. There are indeed several such forms, almost all of them involving kinship relations and totemic affiliations for both people and places -- intense cultural preoccupations in Aboriginal Australia. Nevertheless, the number and extent of these forms is minute compared to the entire grammatical system of the language. And even here, several of these cases involve no novel grammatical categories but only certain special applications of otherwise familiar categories. (The details are spelled out in my chapter "The cognitive culture system").

4. Apart from your career in linguistics and speaking a number of languages, you are also interested in music. How much do we know about the connection between language and music?

In my paper on the evolution of language (Talmy 2007), I compared the cognitive systems present in humans as against other primates. In the evolution toward humans, some cognitive systems, such as vision and motor control, seem to have remained virtually the same. Other cognitive systems, apparently present in a simpler form in other primates, seem to have become greatly elaborated in humans. Possible examples are the cognitive systems underlying culture, gesture, advanced thought, and perhaps even affect. But still other cognitive systems seem to have appeared essentially for the first time in humans. Notably among these are language and music.

The main argument in that paper is that the evolutionary breakthrough enabling language was the appearance of a cognitive capacity for combinatoriality. To describe this capacity in its general form, there first has to be present in cognition a relatively closed inventory of discrete units of a particular type. Then some of these units can be selected out and placed in one or another arrangement. Each such arrangement then constitutes a new higher-level unit with its own distinct identity. For example, each language has a relatively closed inventory of sounds -- typically its consonants and vowels. The linguistic term for these is "phonemes". These are discrete units, wholly distinct from each other, with separate identities. From the English inventory of phonemes, one might select a particular three: "k", "t", and "ae" (the sound in "bad"). These three could then be arranged in three different sequences: "kaet", "taek", and "aekt". Each of these sequences would then constitute a wholly distinct higher-level unit, a morpheme -- here, a word. The three words are spelled: kat, tack, act. Moreover, each of these higher-level units has its own new identity, an identity unrelated to those of the phonemes making it up.

This combinatorial pattern of phonemes forming morphemes may be the most notable and well-known example in language. But, by my count, language has some eight such cases of combinatoriality. The cognitive systems in common across primates seem to have nothing like this density of combinatoriality, and possibly even no full-fledge form of it at all. Thus, it either appears first in the evolution of humans or gets greatly elaborated in that lineage.

What is notable, then, is that combinatoriality appears not only in language, but also in music -- the other main cognitive system that first appears in the human line. In music’s main form of combinatoriality, the relatively closed cognitive inventory of discrete units is that of the notes of the scale. Certain notes can be selected from this inventory, and arranged in different sequences. Each such sequence then constitutes a new higher-level entity with its own distinct identity, a melody.

It is not clear what selective advantage music might have had that may have abetted its evolution. But music would seem to be connected with language -- and may have evolved together with it -- in view of their shared dependence on the otherwise modest occurrence or outright absence of the capacity for combinatoriality in cognition.

5. One of the most debated issues in modern linguistic research is the question of how independent our language is from other cognitive structures. Where do you stand on this question?

The Fodor-Chomsky modularity model largely holds that our cognitive faculties exist as impermeable compartments, each with its own mode of processing incommensurate with that of the other compartments. But I suspect, on the contrary, that little in cognition is unconnected with the rest of cognition. I use the term "systems" in referring to cognitive systems, from major to minor, to avoid the term "module" and so to suggest their interconnectedness. This interconnectedness here refers not only to their cooperative interactions, but also to commonalities of their organization and the kinds of processing they do.

I call this view the "overlapping systems model of cognitive organization". In this view, the major cognitive systems each have certain properties of organization. Many of these properties are comparable across systems -- an "overlap" of the systems. Some properties are shown by only one system, some by several, and some by all.

The major cognitive systems include motor control, perception (in general or in its several modalities: vision, hearing, kinesthesia, etc.), affect, a reasoning/inferencing system, a planning/simulation system, language, and a cognitive system for cultural structure.

Some 17 categories of organizing properties that can be found across all or most of these major cognitive systems are listed in a talk handout (available on my website) called "Relating language to other cognitive systems". To consider just one of these here, one category of organizing properties is schematic structure: the abstracted schematic delineations and partitionings that structure a system. This category includes any distinction between substantive content and schematic delineations that structure it; any differentiation of component "units" within the system; any structural relations between part and whole; and hierarchical embedding.

To illustrate this last type of schematic structuring, hierarchical embedding can be seen in a range of cognitive systems. First consider visual perception and how a viewer might parse a scene -- say, the inside of a restaurant. The perception might include three hierarchical levels. At the top level is a containing framework consisting of the ceiling, walls, floor, and the edges between them. EMbedded within this is the pattern of the people and the tables. And embedded within the outline of the individual persons is the details of their clothing. This hierarchical aspect of the parsing sorts out elements of the overall scene that, without it, might be scrambled.

Hierarchical embedding might also be present in motor control -- say, in someone’s act of moving toward a two-handled bucket to pick it up. Here, the containing framework might be the overall bodily movement toward the bucket. Embedded within this would then be the movements of the body bending at the waist and the arms stretching forward. And embedded within the arm motion might be the fine-level control over the fingers to grip the handles.

Last, consider language, where (among other aspects) the syntactic structure of a sentence has long been seen to include hierarchical embedding. Consider, for example, the sentence A jogger rushed past the houses. The whole sentence is at the top level of the hierarchy. Embedded within it are the subject phrase the jogger, and the predicate phrase rushed past the houses. Embedded within the predicate phrase are the verb rushed and the prepositional phrase past the houses. This last in turn has embedded within it a preposition and the noun phrase the houses, which in its turn consists of the determiner the and the noun houses. A still further level of hierarchical embedding is found even within three of the words of the sentence -- jogger, rushed, and houses -- each of which consists of a root and a suffix.

Even just this one set of examples shows that -- contrary to the Fodorian notion of cognitive modules with incommensurate forms of processing -- visual perception, motor control, and language share in common at least one type of schematic structuring. And schematic structuring is just one among a number of categories of organizing properties that they also share.

To amplify this argument, let me take an evolutionary perspective and consider a cognitive system that appeared early, say, visual perception, and a cognitive system that appeared after that, say, language. Any commonality in properties of organization that language has with vision -- say, the presence of hierarchical embedding -- can have three explanations. One is that some cognitive subsystem outside both vision and language is responsible for hierarchical embedding. When vision first evolved, it formed connections to this subsystem, and when language first evolved, it too developed such connections. Another explanation is that hierarchical embedding first arose as a subsystem within the visual cognitive system, and so appeared when vision first evolved. Then when language evolved, it developed connections to the visual system, in particular to its subsystem for hierarchical embedding. The third explanation is that the visual system and the language system each developed its own independent subsystem for hierarchical embedding when they evolved, perhaps under similar selective pressures. All of these explanations -- like the commonalities of organization in the first place -- argue for closer interconnections between language and the other cognitive systems than held in the modular view.

6. You are the brother of Shel Talmy, the producer of great rock songs by The Who and The Kinks during the explosion of the counterculture and pop. Did the cultural climate influence your work? Did you interact with your brother in the music area?

My brother Shel and I are fond of each other, and each of us appreciates the other for his work. But I don’t think much influence went in either direction. The most I interacted with Shel in the music area was to sit in on a session he was recording once and add to the background clapping (he relied on my sense of rhythm). Just possibly my love of the original musics of cultures around the world, untouched by fusion from the Western tradition, influenced him to record an Irish master of the ilian pipes. But it doesn’t seem to me that Shel’s work in music or the cultural climate of that time has colored my work much.

7. What do you find to be the most rewarding part of your work in linguistics?

Since you ask this question in personal terms, I’ll respond in that vein. What follows could sound a bit grandiose, but it does actually represent the way I feel. I don’t really think of myself as a linguist but as someone who wants to think about the nature of the mind. I see linguistics, the area I’ve gained expertise in, as an avenue to how the mind works. To the extent that I’ve gotten insights into our cognition and shared them, that’s what gratifies me. Sometimes doing this kind of work -- delving into the mind and seeing something of its phenomena -- feels like unveiling some of the mysteries of the cosmos.

References

Talmy, Leonard. 2000. Toward a Cognitive Semantics. volume I: Concept structuring systems. i-viii, 1-565. volume II: Typology and process in concept structuring. i-viii, 1-495. Cambridge: MIT Press.

Talmy, Leonard. 2007. Recombinance in the evolution of language. In Jonathon E. Cihlar, David Kaiser, Irene Kimbara & Amy Franklin (eds.), Proceedings of the 39th Annual Meeting of the Chicago Linguistic Society: The Panels, vol. 39-2, 26-60 Chicago: Chicago Linguistic Society.