Once again, perhaps we are too used to the convenient abbreviations that our written dictionaries provide us with; just because written dictionaries include the syntactic categories corresponding to the alternative meanings of each word does not mean that the mental lexicon does likewise. After all, the OED also includes the approximate date at which each word entered the language-but just because the OED includes this information, and can search for words on the basis of this information, does not mean that the mental lexicon does likewise. So perhaps the simplest interpretation of Tanenhaus's `watch' experiment is to suppose that syntactic categories are not listed separately as in a dictionary. Why should they be? If the syntactic category of a word is nothing more than a reflection of its meaning (and there will be more about this in Chapters 9 and 13), they will not be listed separately. The meaning would have to be activated before the syntactic inappropriateness could be judged. And when we hear an ambiguous word like `bank', how could we judge which meanings were inappropriate if we did not activate them all?
It looks, again, as if all that psycholinguists have done is come up with the obvious (although it is probably true to say that hindsight makes the results of any scientific endeavour seem obvious). But, however obvious it may be, the view that has developed of how we access the mental lexicon is substantially different from the view we might originally have had on the basis of how we access a dictionary like the OED. Accessing the mental lexicon is far less restrained. We activate all the lexical entries compatible with the developing sequence of sounds entering the ear. If the same sequence of sounds has more than one meaning, we activate all the meanings compatible with that sequence and only subsequently are the contextually inappropriate meanings suppressed. Of course, an inevitable by-product of all this is that we must activate all manner of spurious, unintended, meanings. But given that the mental lexicon must reside within the neural circuitry of the brain, this is in fact a natural (if initially counter-intuitive) way for the system to work-the neural circuits are like so many combination locks, and as the speech input unfolds through time, so do the tumblers of the different combination locks move around, until eventually, just those combination locks whose sequences are completed spring open.
Unlocking the combination
How far should the combination lock analogy be pushed? To return to the opening theme of this chapter, might this not be just another example of an analogy that is simply inappropriate, simply wrong, or simply confusing? Maybe. But the continued use of the analogy helps explain one further fact concerning the manner in which we access the mental lexicon. The fact itself concerns a prediction that can be made on the basis of the way a combination lock operates, although when the prediction was originally tested, combination locks were probably the furthest things from the mind of the tester. The prediction, quite simply, is that wherever in the speech input a sequence is found that could correspond to a word, the lexical entry for that word should be activated. After all, rotate the dial of an old-fashioned mechanical combination lock and so long as the sequence of rotations contains, somewhere within it, the correct sequence for that lock, the lock will open.
In the late 1980s, Richard Shillcock carried out an experiment to determine whether, for example, listeners activate the lexical entry corresponding to `bone' if they hear the sequence `He carefully placed the trombone on the table'. He used the cross-modal priming paradigm described earlier in which a word related to `bone' (e.g. `rib') would be flashed up on a screen at the offset of `trombone'. Crucially, he also included sentences like `He carefully placed the bone on the table'. Shillcock found that the amount of priming he got from `trombone' to `rib' was the same as that from `bone' to `rib'. In other words, the lexical entry corresponding to the word `bone' is activated even when `bone' is heard simply as part of the word `trombone'.
Subsequent studies (in Italian) have shown similar effects with sentences equivalent to `He broke all records for the new distance'. Here, the lexical entry corresponding to the word `nudist' is activated (embedded in the sequence `new distance'). And although no one has tested for which lexical entries are activated on hearing `rampart', it would be surprising if the lexical entries for `ram', `am', `amp', `ramp', `part', and `art' were not activated in addition to that for `rampart'.
So it looks increasingly as if the process of lexical access is in fact rather simple: we consider all possible hypotheses about what could be present at any point in the incoming speech signal. But with this simplicity comes an obvious worry; surely there would be an explosion of possibilities? How do we determine which are the right ones? How do we know that we should be reading about `a new discovery' and that nudists have nothing to do with it (even though the meaning of `nudist' is activated)? Presumably, the answers are to be found in the manner by which we string the meanings of words together to give a coherent meaning to the entire sentence. Why we are not even conscious of all these spurious words is another matter. But, perhaps unfortunately, consciousness is beyond the remit of our ascent of Babel.
Psycholinguists have only brushed the tip of a theoretical iceberg. Many questions remain, and controversies abound. For instance, we are still unsure as to the nature of the bilingual lexicon. Only now are we developing theories concerning the processing of morphologically complex words-recall that many of the questions asked in that section of this chapter were never answered. And only now are we developing a better understanding of the tools we use, and the effects that the tools themselves can have on the phenomena we try to observe. But one thing is for certain: whatever the limits of our understanding, we now know not to trust whatever intuitions we may have had on the basis of the dictionaries on our bookshelves.
Time flies like an arrow
Without words there could be no language. Finding them, hidden away within the neural structures of our brains, is half the battle, but only half. Words are more than simple gestures, like pointing, or raising the eyebrows, or shaking the head. We use sequences of words to convey ideas, and by combining the words in different ways we change the idea being conveyed. Gestures and other signs do not have this property. Although they too convey ideas, the idea they convey is not altered simply by altering the sequence of signs. So the words themselves are only one ingredient-recognizing the way in which they are combined is another. Discovering that `Sue would like to marry you' may be the most important thing in the world to you, but only if it can be distinguished from `You would like to marry Sue'. Although this might also be true, it would say nothing about what Sue felt about it. So which way the words are combined is important. But it is important, and has meaning, only insofar as both the speaker and the hearer (or the writer and the reader) share some common knowledge regarding the significance of one combination or another. This shared knowledge is grammar.
But what has this to do with the title of this chapter? In fact, the title illustrates one of the most remarkable things about the way we interpret words and their combination. `Time flies like an arrow' is a classic example used in textbooks on psycholinguistics. It is an important example because, time and time again, the textbooks get it wrong. They remark how, in principle, 'Time flies like an arrow' could be interpreted not as a proverb concerning the passage of time and the flight of an arrow, but as something similar to the sentence `Fruit flies like a banana'. Well, that is certainly true, but hardly riveting. Presumably our lack of experience of any insect called a 'time fly' means that the only sensible interpretation we can come up with is the one in which time is doing the flying (not that you have ever seen it fly). But what really makes this sentence interesting is the way in which the textbooks get it wrong. There are not just two possible interpretations of the sentence, but at least 50. And if, instead of reading this sentence, we heard it, there would be well over 100.3 _link_ And this is the puzzle-our knowledge of the meanings of the different words, and our knowledge about what is allowed by the grammar of the English language, permits each and every one of these different interpretatio
ns. So why are we not aware of them? Why are we not aware of the problems involved in eating pizza with a friend, or with a nice bottle of wine, or with extra anchovies, or even with a knife and fork? What prevents our brains from clogging up with the multitude of possibilities? Why do ambiguities pass us by without even the merest flicker of recognition?
The answer is not as straightforward as one would hope, and has to do, in part, with the way we use our knowledge of grammar.
The conventional aspects of language
The grammar of a language is in fact nothing more than a set of conventions, passed on from generation to generation. If we think of a sentence as conveying information about who-did-what-to-whom (or what), then there is a convention, in English, for the subject to come before the verb (who before did-what), and for the verb to come before the object (did-what before whom). But this is not true of all languages. It is simply a convention; it is true of languages like English and Italian, but it is not true of Turkish (object-subject-verb), Japanese (subjectobject-verb), or either classical Hebrew or Welsh (verb-subject-object). Similarly, there is a convention in English for adjectives to precede the nouns they modify ('the green spaghetti'), but in Italian and Spanish (and other languages too), they tend to follow the nouns they modify (`la pasta verde').
These conventions are no different really from the convention we have adopted to interpret `512' as `five hundred and twelve' and not ,two hundred and fifteen', or from the convention to drive in one country on the left and in another country on the right. If you fail to observe the convention, you do so with potentially disastrous consequences. And just as which side of the road you drive on determines on which side of the car the foot pedals are found, so linguists have found that certain grammatical conventions also tend to go together-if you have one, you also have the other. For instance, if the prepositions in the language come at the beginning of each prepositional phrase ('the man with the freckles'), then the verbs will generally come before their objects ('she liked the man'). But if the verbs come after their objects, as happens in Japanese for example, then the prepositions generally come at the end of each prepositional phrase. In contemporary linguistic theory, the preposition /verb case arises because prepositions are assumed to have a grammatical function that is very similar to verbsthey both specify relationships between things: between `the man' and ,the freckles' and between `she' and `the man'.
Whatever the conventions in any one language, their purpose is to signal specific patterns of meaning. The following two sentences, like the example given in the first paragraph, contain exactly the same words, and yet their meanings could not be more different:
Politicians think that the public don't know. The public know that politicians don't think.
But so what? The same cooking ingredients, mixed and cooked in a different order, can lead to radically different dishes. What is so special about grammar? The words `lips' and `slip' contain exactly the same letters but in a different order, and they too mean different things-so why can we not simply say that each word in a sentence functions in exactly the same way as each phoneme or letter in a word?
One reason is that whereas we know the meanings of only a finite number of words, we have the potential to understand an infinite variety of sentences. We can store in the mental lexicon the meanings of each word we know, but we cannot store in some kind of super-lexicon all the meanings of all the possible sentences that we could ever understand. Instead, we have to create the meaning out of the parts. But what is it, exactly, that we create?
Within an individual word, each phoneme fulfils the same function: to narrow down the search through the mental lexicon. In this respect, the role of any one phoneme (or letter, if written) is the same as that of any other. But the same is not true of the words in a sentence. To return to an example used in Chapter 4: the function of `the girl' in `The girl thinks the language is beautiful' is different from the function of `the language'; `the girl' is the subject of the verb `thinks'-she is the one doing the thinking while the language is, according to her, the thing that is being beautiful. So part of the process of creating, or more properly deriving, the meaning of a sentence involves figuring out the roles that each of the participants in the sentence plays. The term participant here refers simply to the different things that are themselves referred to in the sentence. So the thought `the language is beautiful' participates in the sentence, and has a role to play as whatever was being thought. A verb like `thinks' signals two roles: one for whoever is doing the thinking, and one for whatever is being thought. A verb like `put' signals three roles: one for whoever is doing the putting, one for what is being put, and another for where it is being put. The grammar is simply the knowledge that tells us where in the sentence to find the things that fill these different roles. If we cannot find those things, we cannot understand the sentence.
This brings us back to an earlier theme-different languages employ different conventions for telling us where, within the sentence, to look for the recipients of the various roles that need to be assigned. In a language like English, we rely predominantly on word order. There is nothing to distinguish the different participants in `The boy gave the letter to the teacher' except for their relative ordering within the sentence; the boy gave, the teacher received, and the letter was given. But languages like German, for instance, do not rely solely on word order:
derJunge gab den Brief dem Lehrer (the boygave the letter (to) the teacher)
Here, different versions of the word `the' (in this case, `der', `den', and `dem') are used to reflect the role that the following noun plays in the sentence, and notice that as a result, the German version requires no preposition corresponding to the English `to'. `Der', `den', and `dem' are the same word, just inflected differently (see Chapter 6 for more on inflection and morphology). This kind of explicit marking of roles is called case-marking. And languages which explicitly mark their roles in this way do not require such a rigid ordering of the words as a language like English. Indeed, some languages allow almost (but not quite) unrestricted word order, relying instead on case-marking. So the German sentence given above could also appear as:
den Briefgab derJunge dem Lehrer dem Lehrer gab der lunge den Brief derJungegab dem Lehrer den Brief
So different languages employ different conventions for marking the roles of the different participants in a sentence (or, more correctly, in the event described by the sentence). It all sounds quite simple, really. The grammar of the language is like a recipe; follow the recipe, mix the ingredients in the right order, and you end up creating exactly what the author of the recipe intended you to end up with. In language, you end up reconstructing the idea that the speaker or writer intended to convey. So what is the problem?
Coping with ambiguity
Often, the grammar-as-recipe provides alternative ways to mix the ingredients, and this is why `time flies' could be a term to describe zippers that operate like a time lock on a safe, only opening at certain times. One of the jobs of the reader/listener is to work out which is the right mixture-that is, to work out which grammatical convention is intended by the writer/speaker. It is true that some of the blame for having to do this rests with the fact that the individual words within a sentence can often have more than one interpretation ('flies' as a noun or as a verb, for instance). But there are many cases where the words can be totally unambiguous, and yet we still have to figure out which of several grammatical interpretations is the right one. Eating pizza, with one's family or with one's fingers, is a case in point. The grammar of English permits whatever is in the position after `with' to be assigned a variety of roles, but which role is actually assigned appears to depend on the meaning of whatever will fill the role. But does this mean that we recognize the meaning of something, and try and fit to it all possible roles that are permitted by the grammar? In effect, do we access all possible interpretations before picking the most plausible? And if not, how else could it possibly work?
Fi
guring out what to do with whatever, or whoever, you happen to be eating your pizza with is relatively easy-as long as you can figure out the alternatives, you can settle on the most plausible one. But if someone left the book they were reading on the bus, was it the leaving or the reading that happened on the bus? And if someone was supposed to remind you that you were going to read to them tomorrow, was it the reading or the reminding that would happen tomorrow? Yet again, unless explicitly alerted to the ambiguity, we probably would not notice it. But unlike the pizza case, these cases cannot be resolved on the basis that, of all the alternatives, some are plausible and some are not. One can just as plausibly leave something on the bus as read it on the bus, or read to someone tomorrow as be reminded tomorrow. So it is not just a matter of rejecting implausible alternatives. Something else is going on. But what?
All the examples discussed so far have involved ambiguities that go unnoticed. Unless our attention is explicitly drawn to them, we tend not to notice the ambiguities we encounter in our everyday language. Sometimes, however, an ambiguity is noticed, or rather, the consequence of not noticing the ambiguity is noticed, and with surprising effect. A photograph in the Sunday Times of Sam Shepard, the actor, and Chuck Yeager, the first pilot to fly faster than the speed of sound, was accompanied by the following caption:
The Ascent of Babel: An Exploration of Language, Mind, and Understanding Page 11