Volume 7 [Spring 2000]
Home | Contents | Subscriptions | Archives


Words and Rules: The Ingredients of Language
by Steven Pinker

Review by Jonah M. Knobler


The verbs in English are a fright.
How can we learn to read and write?
Today we speak, but first we spoke;
some faucets leak, but never loke.
If I still do as once I did,
Then do cows moo, as they once mid?
-Richard Lederer

When Lederer wrote (not writed) this bit of verse, he hit the nail right on the head: English is a language of wacky and capricious inconsistency. Its irregular verbs have long been the subject of desperate groans from students, pedantic lecturing from teachers, and smug irony from humorists. "Hoe," "know," and "go" are phonetically indistinguishable from one another, save for their first sound. Why, then, are their respective past tenses "hoed," "knew," and "went," for heaven’s sake? What twisted process could have engineered such an apparently arbitrary hodgepodge of supposed rules ("just add ‘-ed’") and glaring inconsistencies ("delete everything and replace with ‘went’")?

Enter Steven Pinker, psycholinguist extraordinaire and author of the newly released Words and Rules: The Ingredients of Language. Pinker, one of today’s foremost figures in cognitive science, has spent years studying the notorious system of English past-tense verb conjugation, in hopes of elucidating some of the inner workings of the human mind. Although he has already achieved widespread public renown and resounding critical success with his first two books (The Language Instinct, an engaging linguistic treatise, and How the Mind Works, a tour-de-force survey of evolutionary psychology), it is with Words and Rules that Pinker finally introduces lay readers to his own personal research.

Words and Rules: A Dynamic Duo

Specifically, Pinker has developed a "words-and-rules" theory of language which states that everything linguistic is fundamentally grounded in two complementary phenomena, "words" and "rules." "Words" are defined as the memorized chunks of linguistic data stored in the human mind - the entries in our mental dictionary - while "rules" are the active procedures performed in real-time to generate novel linguistic forms. English past-tense verb inflection requires both phenomena, words and rules, in concert (2). Without either one, fluent speech is impossible.

Pinker uses the term "words" in a somewhat non-standard sense. In common parlance, "walk," "walked," "walks," etc. are all words. But Pinker uses "word" to mean a lexical entry, a memorized unit of verbal information physically stored in the mental dictionary (3). According to Pinker, we don’t independently memorize all of the forms of each regular verb (or noun, or adjective). We just memorize the root - it’s vastly more efficient that way. (Nor do conventional dictionaries have separate entries for "walk," "walks," "walked," "walking"¼ Imagine how bloated such a tome would be!) With respect to regular verbs like "walk," the root form ("walk") is the only Pinkerian word (or lexical entry) of the bunch - the only one with a listing in the brain’s dictionary. The other forms are generated on the fly by rules, cutting down on the amount of brain space and memorization energy required for language.

Regular Verb Inflection: Rules in Action

Just as a computer program calls functions, passing data to those functions as input and returning a desired value as output, the mind invokes rules, passing words as input and generating correctly-inflected language as output. If Pinker is correct about rules, then somewhere in the mind lies something resembling the following function. Note: text in Courier New font represents "mentalese," or a pseudocode of what goes on in the mind.

function past(input: some_word)

begin

return(some_word + "-ed");

end.

This simple past() rule is enough to generate the past tense of each of the countless regular verbs in the English language. It’s clear that past(walk) is walked, and that past(talk) is talked, and that past(overreach) is overreached. With this simple bit of mentalese, we have rendered it unnecessary to memorize separately the past tense form of every regular verb we learn. Because the formation of the present tense third-person form ("walks") and the present participle ("walking") is also regular and controlled by rules, our mental storage needs for verbs are reduced to a quarter of what they would be if we had to memorize each form individually (18).

Of course, memorizing only the root forms as Pinkerian "words" and having a rule in mentalese to generate the others is the efficient way to go. But how do we know that our brain actually takes this efficient route? Can we prove that a past() rule exists? Indeed we can - new words come into the lexicon every day, and as soon as we hear a new verb ("to videoconference," say), we instantly have access to its past tense form ("I videoconferenced with my boss"), even if we have never heard it used (150). This is strong evidence for the presence of a rule in the mind that can take as input any verb, no matter how new or unfamiliar. Furthermore, if we assume that "to Sally Ride" means "to achieve great publicity as a woman astronaut," most speakers would say "Christa McAuliffe out-Sally Rided Sally Ride" and not "out-Sally Rode Sally Ride" (158). This shows that we don’t automatically inflect new verbs by analogy to inflections of similar sounding verbs. There’s actually a rule in there somewhere, operating independently of any previously memorized knowledge. Words and rules really are separate! As further evidence, Pinker goes on to show that word-based and rule-based linguistic processes actually use different physical areas of the brain (252).

So we have a rule at work churning out the past tenses of regular verbs as we speak. (The same is true for their present participles, for the plural forms of nouns, etc.) Why, then, do we say "guideguided" but "riderode" (not "riderided"), and "hidehid" (not hidehided)? Why doesn’t the past() rule work here?

Irregular Verbs: Where Words Rule Supreme

The answer, according to Pinker, is that the operation of the past() rule is inhibited by the presence of an irregular form in the mental dictionary. Evidently, when we want to conjugate a verb in the past tense, both pathways - the words pathway and the rules pathway - are simultaneously invoked. The events in the words pathway (specifically, the successful location of an irregular form) block the rules pathway from ever reaching completion (130).

Let’s take a closer look at how the words pathway operates. First of all, irregular verbs are stored in the brain in data structures somewhat like this. (Square brackets denote phonetic sequences written in the International Phonetic Alphabet. Just as a dictionary spells words phonetically to aid pronunciation, your mental dictionary contains phonetic information to tell you how to pronounce what you’re thinking.)

word:"RIDE"

sound =

part_of_speech = verb

meaning = "to use as transportation"

past_tense = "rode"

past_participle = "ridden"

Whereas regular verbs are stored in the brain somewhat like this:

word:"GUIDE"

sound =

part_of_speech = verb

meaning = "to direct or offer assistance"

When you are speaking and want to say "Yesterday I [ride, past tense] my bicycle to the Steven Pinker lecture," your mind simultaneously fires up the rules pathway and the words pathway. The latter calls up the data record for "ride," and scanning down the fields, notices that there is an entry for past_tense. Effortlessly, you retrieve "rode" and plug it into your sentence, blocking off the result of the rules pathway (the incorrect form "rided").

But if you want to say "Yesterday I [guide, past tense] a tour group around Harvard Yard," your brain will not find an entry for past_tense under its record for "guide." The words pathway reaches a dead-end here and gracefully bows out. Meanwhile, the rules pathway has already been activated. It invokes the rule past(guide), which, as we saw earlier, yields as output "guided." Thus, the rules faculty of our brain appears to be the default method of inflection that applies when we lack any previously specified information in our words faculty to preempt it.

Words-and-Rules Theory: What It Explains

This model of interaction between words and rules works beautifully to show why young children make those adorable errors like "Mommy, he hitted me!" Since the toddlers have seldom (or never) heard the irregular forms, they are absent from the mental dictionary, and the rules pathway runs to completion with no interference from the words pathway (198). The model also shows why people have such a hard time with uncommon irregular past tense forms like "strive-strove." When we want to use the past tense of "strive," we often waffle between "strived" and "strove," unsure of which one is really correct. This is because the data record for "strive" in most people’s minds looks something like this:

word:STRIVE

sound =

part_of_speech = verb;

meaning = "to make a great effort"

past_tense (=)"strove"

past_participle (=)"striven"

Here, the (=) signs in the past_tense and past_participle fields represent tenuous, weakly reinforced neural connections, the result of only a few instances of hearing "strove" and "striven" and associating them with "strive." As the mind scans unconsciously down the mental dictionary entry for "strive," these tenuous connections are not sufficient to preempt the rules pathway, which, by default, runs to completion and provides "strived," as if for a regular verb. The connections to "strove" are still there, however, as weak as they may be, so we always have a sense of uncertainty whenever we say "strived." The less frequently we hear a given irregular form, the more tenuous that form’s connection to its data record, and the more indecisive we are when we have to come up with it (131).

Often, an irregular past tense form will become so rarely used that a new generation of children never hears it; if this is the case, the verb is eventually regularized by subsequent generations. English was once far richer in irregular verbs than it is now, but children seldom heard the inflected forms of the rare verbs. The linkages in their mental dictionaries from the verb roots to their irregular past forms eventually became too weak to block off the rules pathway, which then could proceed uninhibited, as it does for regular verbs which have no such linkages at all. As a result, we now say "He has often chided her" instead of "He has often chidden her" (125). Words-and-rules theory, we now see, has done much to shape the development of the English language!

Phonology: From the Mind to the Mouth

There is actually one more process at work in the conjugation of verbs. After the words and rules pathways have been traversed, the resultant snippet of mentalese (that hopefully represents the correct conjugated form) must be translated from thought-language into pronounceable English (or Spanish, Xhosa, Cantonese¼). This is the responsibility of phonology, the faculty in the brain that converts the underlying proto-words into the surface forms in which they are actually pronounced. For example, although the past() rule affixes the mentalese suffix "-ed" to the end of a verb stem, we depend on phonology to tell us how to pronounce that suffix. In the word "bugged," the suffix is pronounced as a "d." In the word "bucked," it’s pronounced as a "t." And in the word "budded," it’s pronounced "id." (Say these three words out loud if the difference in pronunciation isn’t apparent.)

Phonology doesn’t just work at random. It actually operates according to numerous rules similar to the rules we’ve already discussed. However, phonological rules have nothing to do with tense, person, number, or other "grammatical" concepts, as rules like past() do. While inflection rules like past() convert raw mentalese into grammatical mentalese, phonological rules convert grammatical mentalese into speech. The phonological rule that accounts for the variation in pronunciation of the suffix "-ed" appears below. Assume that we have already defined a function final_sound() which returns the final phonetic sound of a given string, and a function voiced() which returns true if a given phonetic sound is voiced (one that your vocal cords vibrate while making, like vowels or the consonants "b," "g," "m," etc. - put your hand on your voicebox and see if you can feel it!) and false if it is not.

procedure pronounce_with_past_suffix(input: some_verb)

begin

if final_sound(some_verb) = "t" or "d" then say(some_verb + "-id");

else

if voiced(final_sound(some_verb)) = true then

say(some_verb + "-d");

else

say(some_verb + "-t");

end.

That is, when the "-ed" suffix follows a "t" or "d" sound, it’s pronounced "-id." If it follows a voiced sound, it’s pronounced "-d"; if it follows an unvoiced sound (one that your vocal cords don’t vibrate while making, like the consonants "p," "k," "f," etc.), it’s pronounced "-t." Try saying "butted," but pronounce the suffix "-d" instead of "-id." Phonological rules prevent incomprehensible tongue twisters like this one, making words easier for the speaker to pronounce and easier for the listener to understand.

Conclusion: Putting It All Together

To recapitulate: the process of verb conjugation depends on two pathways in the brain - the path of words and the path of rules, both of which feed into the system of phonology. Words are the memorized data records in the brain that contain fields for concepts like part of speech, meaning, and irregular forms. Rules are the active processes that occur automatically, algorithms coded in mentalese that accept input and return output. As we have seen, irregular verbs are inflected using only the words faculty, while regular verbs depend on the rules faculty for inflection.

While words and rules are both learned during early childhood (as they must be, since they vary from one language to another), the capacity and inclination toward learning words and rules are innate (208). From birth, babies are actively scanning an overwhelming stream of raw phonetic data for the patterns that signal the presence of words and rules. From one part of the world to another, the words and rules themselves differ, but their existence is universal, and the drive to find them is present in the mind of every human being at birth.

The critical point, as Pinker deftly points out, is that the words and rules theory explains more than just verb inflection. It sheds light on many long-standing linguistic conundrums, ranging from why individuals suffering from a certain brain disorder can conjugate irregular verbs but not regular ones, to why we say "the batter ‘flied out’ to left field" (not "flew out") (149). In Words and Rules, Pinker shows how his theory goes beyond the realm of verb inflection, tracing its ramifications through all of language, cognition, and consciousness. Illustrated with fascinating case studies and supported with the newest evidence, Words and Rules is truly an enthralling glimpse into our inner mental workings. Although Pinker focuses in detail on one particular microcosm of human intelligence - irregular verb conjugation - the result is a thrilling journey to the frontier of our understanding of the human mind.