Implicit Agency in Time-Space Metaphors

As I've said before, the primary (if not the only) real experimental evidence for conceptual metaphor theory -- the theory that abstract concepts are structured by more concrete (i.e., closer to sensory/perceptual) concepts -- comes from one domain: time. Time, according to conceptual metaphor theory, is structured primarily through spatial metaphors. In English, there are two dominant spatial metaphors: the ego-moving metaphor, in which we are moving toward the future (e.g., "I'm coming up on my 30th birthday"), and the time-moving metaphor, in which we are stationary while time moves towards us (e.g., "My 30th birthday is fast approaching"). Most of the experiments on time-space metaphors follow the same basic desgin: either the ego-moving or time-moving metaphor is primed (e.g., by showing people a picture depicting motion consistent with one or the other), and then given an ambiguous temporal statement like, "Next Wednesday's meeting has been moved forward two days." When asked what day the meeting is now on, people tend to answer in a way that is consistent with the primed spatial metaphor. Thus, if the ego-moving metaphor was primed, people will tend to answer that the meeting has been moved to Friday, and they tend to answer Monday if the time-moving metaphor is primed. Here's a great illustration of this from a poster by Dennis and Markman1:

i-599c8e922197205693e5baefdf21c224-Time-Space Metaphors.JPG

In the past (including in the linked post), I've maintained that it was possible to explain the results of such experiments as simple priming effects. Here's how I explained it before:

But the priming procedures that all of these experiments use may provide evidence of nothing more than, well, priming. We would also predict these priming effects just by noting that temporal and spatial language use the same words. Sure, at some point in the distant past these words may have been used metaphorically, but now the metaphors are "lexicalized," or put differently, "dead." The words used to speak about time are now only about time, but are homonymous (or polysemous, which amounts to about the same thing6) with the words used to talk about space. We would predict the priming effect because, as research on suppression in language comprehension has shown, when we read/hear a word, we activate all of its common meanings, and then suppress those that are irrelevant in the present context7. However, simply activating the irrelevant meanings can lead to priming effects. Therefore, after viewing a spatial prime, reading a word that can have a spatial or temporal meaning will lead to priming effects even in temporal domains.

In other words, I have always argued that when spatial terms are used to talk about time, the spatial meanings are active long enough for the spatial primes in these experiments to influence their interpretation, even if time is not conceptualized through metaphorical mappings to space.

It turns out there may be another explanation for these results that also has nothing to do with whether time is conceptualized through metaphorical mappings to space. This explanation was proposed by Dennis and Markman, and has to do with the grammatical structure of the ambiguous temporal sentences. Again, the sentence that is generally used is, "Next Wednesday's meeting has been moved forward two days." Dennis and Markman note that a difference between the ego-moving and time-moving interpretations of the sentence has to do with whether the "ego" is either active or passive. In the ego-moving interpretation, the "ego" is an active agent, while in the time-moving interpretation, the "ego" is a passive patient. If the spatial primes in previous experiments prime a particular representation of agency (either active or passive), it can affect the interpretation of the spatially ambiguous sentence even if our concept of time is not metaphorically structured by our concept of space.

In order to test this interpretation, Dennis and Markman had participants unscramble two types of sentences. Half of the participants unscrambled sentences using the first-person subject pronoun "I." The other half unscrambled sentences using the first-person object pronoun "me." Here are their examples of each:

  • Subject pronoun: Mary I bridge under kissed the ("I kissed Mary under the bridge")
  • Object pronoun: Mary me kissed the bridge under ("Mary kissed me under the bridge")

The sentences using the first-person subject pronoun sentences prime the ego-as-active representation of agency, and the first-person subject pronoun sentences prime the ego-as-passive representation. Thus, Dennis and Markman predict that when participants unscrambed the subject-pronoun sentences, they would answer that the meeting had been moved to Friday, consistent with the ego-moving interpretation, and when they unscrambled the object-pronoun sentences, they would answer that the meeting had been moved to Monday, consistent with the time-moving interpretation. This is in fact what they found: 71% of participants answered "Friday" after unscrambling the subject-pronoun sentence, and 60% answered "Monday" after unscrambling the object-pronoun sentence (proportions consistent with past studies using spatial primes). Thus, they argue that the results of previous studies using spatial primes can be explained as resulting from people's "unconscious representations of agency/passivity."

Now, this research is ongoing (it was presented at a conference, but hasn't been published in a peer reviewed journal yet), but I'm not sure how you would interpret these results from a conceptual metaphor theory perspective, and even if you could, they show that the evidence for the metaphorical structuring of time in terms of space has plausible alternative interpretations. I'm sure the debate about how to interpret previous results will continue, and that more really interesting work will be done, but for now, I think it's safe to say that the evidence forthe metaphorical structuring of time is on very shaky ground. Without that evidence, it's not clear that conceptual metaphor theory has an empirical leg to stand on.


1Dennis, J.L.M., & Markman, A.B. (2005). Are abstract concepts structured via more concrete concepts? Proceedings of the 27th Annual Conference of the Cognitive Science Society, 2467.

More like this

I've said it before, and I'll say it again: conceptual metaphor theory sucks. Why does it suck? Well, because there's no experimental evidence for it (and plenty of evidence against it). Except, that is, in one domain: time. Specifically, the work of Lera Boroditsky, along with Dedre Gentner and…
Last week's article on the Aymara language and metaphorical depictions of time generated a lot of discussion. I think part of the confusion there had to do less with the specific example and more with basic questions about metaphorical representations of time, so today I'm going to cover some of…
The last post on time-space metaphor research has sparked a really interesting discussion in the comments (go check it out), so I thought I'd talk about some more research to see if we can't get even more people talking. If you've been following that discussion, this is the research by Daniel…
In his comment to my post on conceptual metaphor theory (CMT), reposted here, Dr. Gibbs writes: The topic of why conceptual metaphor theory arouses such vehemence is one that greatly interests me and is again the subject of my in progress book. My own opinion is that conceptual metaphor theory, and…

Hi Chris,
Alex Kranjec claims to have found a new type of temporal ambiguity (I forget what it was exactly; I talked with his research supervisor at the APS in May) but there is some stuff by him out there on the web. There is also this emobodiment study which is pretty cool...(page 16 of the .pdf)
http://www.cogling.org.uk/mambo/images/stories/PGCCL/bookofabstracts120…
Anyway, my question is do these manipulations (as in the .pdf above) get rid of the agency confound you write about?

Chris, I want to make sure we're not throwing out the baby with the bathwater as we pick apart conceptual metaphor theory.

If you interpret the theory to be saying that one domain is understood completely in terms of another domain, by appropriating the same words and concepts in a mindlessly reductive fashion, then yes, I think it's fair to say that such a theory can't possibly be true. (I'll leave the question of whether even Lakoff believes it in this extreme form aside for now.)

But if you take this straw man of a theory and leaven it with a little sanity and critical thinking, there is unquestionably a fundamental insight being breached here, and that is of the ubiquity of metaphor and analogy in our cognition. Almost certainly spatial terms cannot be borrowing the exact same mental real estate when they are used in a temporal context. But it seems to me that spatial terms themselves point at deeper, almost non-modal concepts, especially once these spatially grounded concepts prove through experience to be usefully applicable to other domains, a process which inflects the underlying concepts themselves, while permitting multiple "views" onto the concept for different purposes, some extremely common (eventually leading to a sort of conceptual mitosis), and some more ad hoc.

In general, I find extreme claims about the polysemy of words as they are used in different domains both a bit perverse and overly logocentric. Perverse because it seems like a terribly obtuse and inflexible way to design a brain; logocentric because the very invocation of a term like "polysemy" shifts focus unduly toward surface linguistic forms, away from the creative application, misapplication, elaboration, and abstraction of concepts which is the basis for all our cognitive abilities.

Looking forward to your upcoming posts ...

Matt, oh, I agree that it's improtant to understand just how ubiquitous metaphor and analogy are in cognition. I always recommend the book, The Analogical Mind, but I'm not sure conceptual metaphor theory really offers us anything.

I think Boroditsky's work can be debunked for a different reason. The supposed temporal judgments people make in the "abstract" domain of the time don't really involve temporal processing. They are just about days of the week. Days of the week adhere to very strong ordinal sequence. To just illustrate this point, the SNARC effect (an effect used to study number processing) has been shown to occur with linear orders, like the alphabet, months of the year, and yes, days of the week. Boroditsky's task just involves processing of a linear order.

If you want to show there is no evidence for empirical evidence, you would need to address some strong work by Casasanto.

There is also some empirical evidence for the GOOD-IS-UP conceptual metaphor, e.g. Meier & Robinson (2004).

"Therefore, after viewing a spatial prime, reading a word that can have a spatial or temporal meaning will lead to priming effects even in temporal domains."

As long as the schema structuring the temporal concept has a (predictable) spatial structure that is. The polysymous word will only have a priming effect if the word primes meaning in the context of a spatial structured schema. So I don't see why you consider this a criticism necessarily.

"I'm not sure how you would interpret these results from a conceptual metaphor theory perspective."

Linguistic representation necessarily involves spatial imagery or "simulation"? It can also involve representation of the ego (or not). Interpreted as such, the concept of agency/passivity is not a confound but an intrinsic component or primitive of the conceptual representation.

Able, not true. For the type of priming I described, all you'd need is for the same words to be used, because we know that when a word is used, all of its most frequent meanings are activated, and the contextually irrelevant meanings are subsequently suppressed. My point is that simply activating the spatial meaning for a moment is enough to lead to priming.

Shane, I hadn't thought of that, but it's an interesting possibility.

Michael, I read the abstract, and I have to admit I'm confused by the data. The "novice" participants showed no effect, and the aware participants only showed the effect on a few trials. I'm not sure what the hell that means, but I don't think it's obvious that it means they're using space to structure their time concepts.

"My point is that simply activating the spatial meaning for a moment is enough to lead to priming."

But this doesn't contradict the argument. The word "forward" has spatial and, its argued, a distinct temporal meaning. The meaning of forward is different depending on the domain. Moving something forward in space has a different meaning than moving something forward in time. Right?

Now we prime "forward" (spatial) motion in the context consistent with "ME" or "MT" temporal models.

Then we ask the ambiguous temporal question.

Now if forward motion is priming the spatial meaning of "the word forward" without priming something like a schematic representation of a temporal model, but rather the temporal meaning of the same, but primarily spatial "word forward" (linked by a conventional semantic network) why would participants give different answers to the ambiguous question consistent with the primed spatial model, unless *the spatial model is itself* somehow relevant to the strucuture of the temporal concept. That is, why should the primed word forward, accessing a secondary temporal meaning, lead to predictably consistent but different responses to the ambiguous question? What/where are the different semantic "links" you are appealing to explain the difference is response patterns? Anyway, its not very clear what you mean.

The "links" are evident in the research on polysemy in the other recent post, for example. The fact that there is some semantic overlap between the two meanings of the terms "forward" can affect temporal reasoning without the metaphorical structuring.

Chris, the polysemy experiments, and certainly your response don't address why some participants are being primed to associate "forward 2 days" with "Monday", and some with "Friday" and, more importantly, why those associations are consistent with the structure of the nonlinguistic spatial prime (that itslef is being argued to reflect the schematic structure of the temporal concept.) What are the "evident" semantic links that account for response patterns in experiments that use the ambiguous question about Wednesday's meeting? And if they're evident why not simply state what they are?

You're going to have to be much more explicit about what your calling "semantic overlap" in this specific case. Right now the pattern of responding is consistent with the structure of the theorized spatial schema. Do you have a more parsimonious semantic model that you are proposing to account for the specific pattern of responses one observes with the ambiguous question? If so, it shouldn't be difficult to explain.

Able, you're saying that in order for the spatial primes to result in forward meaning Friday or Monday, they have to be using one of the metaphors to structure their interpretation of the temporal relation in the sentence, but the Dennis and Markman study itself shows that's not the case.

Furthermore, I'm not arguing that we don't have an ego-moving and time-moving metaphor, simply that the meaning of "moving," "forward," "before," "after," etc. have specifically temporal senses that are largely independent of their spatial counterparts (i.e., they don't need an on-line mapping to the spatial meaning of the terms to get their content). If that's the case, then priming the spatial analog of a particular sense of the term "forward" would easily affect the corresponding temporal sense, without the metaphorical structuring. That's not more parsimonious in that it requires that you have two senses for "foward," as opposed to one used in two different contexts, but I don't think that's a problem.

Chris,
Do you have any thoughts of how you might test my explanation that the effects are down to processing of linear order and not really about temporal processing? I have a few ideas, but was wondering what you think.

Obviously, you would need a context that doesn't involve linear order, but that's tough with time. I proposed an experiment a while back (see here) using the DRM false memory paradigm, but the design needs some work. I think if the kinks could be worked out, it might provide a way around some of these problems.

Chris,

Able, you're saying that in order for the spatial primes to result in forward meaning Friday or Monday, they have to be using one of the metaphors to structure their interpretation of the temporal relation in the sentence, but the Dennis and Markman study itself shows that's not the case

I don't think D and M think their results show that people are not necessarily representing the problem spatially. Certainly representing "agency" (or the ego) is a vital component of the ME schema. One could very easily spin their results in a way to support the metaphorical view.

I'm not arguing that we don't have an ego-moving and time-moving metaphor, simply that the meaning of "moving," "forward," "before," "after," etc. have specifically temporal senses that are largely independent of their spatial counterparts (i.e., they don't need an on-line mapping to the spatial meaning of the terms to get their content).

In Boroditsky 2000 this issue is addressed at length. I don't think anyone claims temporal senses of words are no represented independently (on some level). Although, you're right I think, in that it isn't very clear what the relations may be.

"If that's the case, then priming the spatial analog of a particular sense of the term "forward" would easily affect the corresponding temporal sense, without the metaphorical structuring."

This is where you lose me. What is the "corresponding" temporal sense, independent of a spatial schema, that leads to the pattern of responses after spatial prming seen in Boroditsky 2000, Boroditsky and Ramscar 2002 (for example). How do you describe the pattern of results independent of the spatial model. I simply can't.

In one case "forward" is associated with "movement" to earlier times. In another case, to later times. According to your criticism, something about the spatial meaning of "forward motion" corresponds with earlier or later times independently of some critical spatial strucuture. The "correspondence" *is* the evidence of schematic spatial structuring. There is simply no way to explain the response pattern without the spatial model. There is no reason why forward motion should sometimes mean "movement to earlier times" and sometimes "movement to later times" independent of a spatial model. That so many studies have shown that spatial motion primes responses to questions about time with patterns of results predicted by the spatial models, is strong evidence for correspondence. You still haven't said what the alternate correspondence might be.

Abel, Dennis is lurking around here somewhere, so maybe we can get him to comment.

I will try to answer you more thoroughly after I get a post out on Gernsbacher's suppression mechanism. For now, I'll say that my position is that the temporal concepts were borrowed, metaphorically (analogically) from space, but at some point in the linguistic past, and that by corresponding senses, I mean senses derived from particular spatial senses.

For further support for Chris's, point have a look at Kemmerer (2004) -

The spatial and temporal meanings of English prepositions can be independently impaired

He found a double dissociation of brain damaged subjects who understood the spatial meaning of prepositions or but not the temporal meanings , and vice versa,

Chris, my call for evidence was to show directly that the results are due to linear order processing not to temporal processing per se, not to find a non linear order task. I just found a recent Cogsci proceedings paper by Matlock, Ramsar & Srinivasan whose results (but not their explanation) strongly supports my interpretation. If people are primed by counting from G to P they are more likely to say Friday. If they count from P to G, they are more likely to say Monday. The title of the paper (and their explanation) is :
Even the Most Abstract Motion Influences Temporal Understanding
So, everybody (e.g. Abe), do you think the results are best explained by some kind of abstract motion or spatial processing, or processing of linear order?

For a non linear order task, with non linguistic temporal processing, which does show evidence that the linguistic mappings from space to time influences our understanding of time I would look at Casasanto's work, which I keep mentioning...

Aside from purely linguistic issues, our conceptual structuring of time does involve spatial models. For example, analogue clock faces, calendars, and timetables. It would not be surprising to find that when we process temporal concepts (hours, days of week, months) we sometimes rely on spatial processing, using those spatial models.

Shane,

"The spatial and temporal meanings of English prepositions can be independently impaired."

Even in Boroditsky 2000 she argues that the time and space domains can be represented separately (she gives a rather complicated account of this partially to explain the theoretically predicted and observed asymmetry in inter-domain priming effects) so I don't think this general line of argument discredits the theory that spatial schemas structure temporal understanding.

"Aside from purely linguistic issues, our conceptual structuring of time does involve spatial models. For example..."

I agree completely. That's why I find the above criticism that spatial priming doesn't influence temporal understanding somewhat bewildering. Why would one be so resistent to the idea of spatially structured cognitive models of time when we rely on them so much "outside" our heads.

"So, everybody (e.g. Abe), do you think the results are best explained by some kind of abstract motion or spatial processing, or processing of linear order?"

I don't see why these accounts are mutually exclusive. How does one represent "linear order" non-spatially?

Shane, I don't know if I mentioned it in any of the posts on the new blog, but have you seen my descriptions of the study in which people's motivations (approach or avoidance) affected whether they moved the meeting to Monday or Friday? If not, let me know. I think even Casasanto's work is, at the very least, limited by that finding, because it shows a distinct departure from both the ego-moving and time-moving metaphors.

I'm willing to accept that time is structured metaphorically, at least some of the time, though. I'm just not convinced by the data so far.

"I don't see why these accounts are mutually exclusive. How does one represent "linear order" non-spatially? "

I think there are many interesting inter-relationships and differences between how the brain processes linear order, numbers, temporality, and spatial and non-spatial magnitude, with some shared neural substrates, with work in this area starting to show how they are related.

So, for the question: Is our understanding of time partly based and structured by spatial processes (and linear order etc)? I think the answer is definitely yes.

For the question: Do we metaphorically structure abstract domains in terms of concrete domains? And does Boroditsky's work provide evidence for conceptual metaphor theory? The answers are much less clear. I wouldn't call how the brain processes time as abstract, nor would I call use of spatial mechanisms in temporal processing metaphoric.

I found the post about avoidance and approach, and remember reading it now. You don't cite who the work is by, and can't find it on the interweb - can you let me know?

But, Casasanto's work doesn't have anything to do with ego moving or time moving metaphors. Firstly, it shows on a non lingustic psychophysics task (with no primes) that spatial judgments affect temporal judgments, but not the other way round.

Secondly, it appears the conceptual metaphors different languages use about time affect temporal processing. Languages like English use time as distance (long night) metaphors whereas in Greek, they use time as volume (big night). In cross-linguistic work it shows that judgments of volume affect temporal judgments more for the time-as-volume metaphor languages than distance judgments, but not for languages like English. Plus, if you train English speakers to use time-as-volume linguistic metaphors, they show similar results.