Interesting facts?: Reflections on 'Lark in my heartbeat'
Just over seven years ago, Cambridge University Press published my book of reflections on psycholinguistics, titled (mundanely enough) 'Reflections on Psycholinguistic Theories: Raiding the Inarticulate', the faint nod towards T.S. Eliot’s Four Quartets being the most that my editor would allow as to its true nature. It is fair to say that the book was both a critical and commercial failure (though the same editor noted that, with over 4000 downloads, the work was no less successful than many academic texts). It was not so much panned as ignored. On reflection (!), I understand many of the reasons why, chief among these being that it is neither a textbook nor a monograph, nor yet a book for the lay reader ("...didn't make heaven...didn't make Harrisburg, [it] died in a hole in between." Josh Ritter, Harrisburg)
Still, what I regret most about the book is not that it was "neither fish nor flesh", but that I did not carry the Liberal Arts project far enough—to approach the scientific question of the relationship between linguistic theory and mental representation of language through the medium of poetry, song, popular music and comedy. This is, as far as I am aware, the first academic text to have had its own Spotify playlist. Had I stuck to my convictions, the book would have had the title Lark in my Heartbeat: A Conversation about Minds and Languages, and the text in the next section would have been the Prologue below.
(In terms of content, the only clear mistake I made concerns the chapter titled "P is for Poverty of (the) Stimulus", which I deliberately left blank—P is also for polemical). In 2014-2016, when I was writing the draft, I was much more skeptical of Poverty-of-Stimulus claims than I am now. But that was before the advent of the Large Language Models. OpenAI may have persuaded some of the redundancy of Innateness; in my case, just the opposite is true.)
Prologue: Interesting facts
Peter Cook: Did you know that you’ve got four miles of tubing in your stomach?
John Cleese: No, I didn’t know that.
Peter Cook: It’s a good thing I’m here, then. You have…you got four miles of tubing in your stomach, all coiled up very tightly…It has to be coiled up very tightly, otherwise the people in charge would never be able to cram it all in…’Course, it’s not any old tubing they use…it’s a specialised form of tubing they use: it’s called intestines…Aren’t you interested in your intestines?
John Cleese: Not particularly.
Peter Cook: Well, you should be. ‘Cos without your intestines, you’d be unable to digest…and then you’d look a bit of a fool…Would you like to see a diagram of your intestines? I’ve got a diagram of your intestines here…
From Peter Cook Interesting Facts [Secret Policeman’s Ball 1979 version]
This book is not as comprehensive an overview of psycholinguistics theory as some others. But neither is it as boring. Unless you are a Gradgrind (‘Stick to Facts, sir!’) or Peter Cook’s character in the Interesting Facts sketch, part of which is reproduced here, a handful of arcane facts presented without justification is anything but interesting, and a compendium of facts about psycholinguistics can be staggeringly dull. Even Peter Cook’s importunate character is quick to tell John Cleese why he should be interested in his internal organs, to try to keep his attention. There may be some people who would rather settle down on a rainy afternoon with a nice fat encyclopedia of language processing facts than with a novel, or even the instruction manual for a new piece of electronic equipment, but outside of university departments and research institutes such people are vanishingly few in number. Happily so.
Peter Cook’s comedy sketch also provides a heads-up, right at the outset, to another ‘interesting fact’ about psycholinguistics (and the sciences in general). This is that verified facts—as opposed to best guesses, speculations, and well-informed hunches—are rather thin on the ground. Sparse, practically endangered, one might say. This is in part a virtue of the Scientific Method, which never establishes the truth about anything, as we shall see later. It is not true, by the way, that you have four miles of tubing in your stomach. And though it may be the case, as Peter Cook asserts later in the sketch, that a giraffe would avoid a lot of trouble if it could jump ‘pound-for-pound as high as a grasshopper’, that is no fact either: it remains only a conjecture, since it can never be demonstrated empirically. Not just because real giraffes’ vertical leaping abilities to any height are severely limited—we could always run a computer simulation—but rather because we have no clear idea of how much trouble counts as a lot, or even what trouble might mean—to a giraffe; see What is a Language Experiment? in Part 2 below.
Rather than offer just another introductory survey, I present a three-part exposition of psycholinguistics, which deals in turn with three sets of basic questions: Why?, What? and How? Why is psycholinguistics interesting? Why do psycholinguists disagree on the answer to this question, and on other more substantive issues? (Why should I bother to read Parts 2 and 3?); What is claimed to be represented and processed in the mind of a language user? What is involved in spoken language comprehension and production?; What is the difference between first and second language acquisition, aside from their order of appearance?; How do we know (how likely) these things are (to be) true? How could we do more experiments to find out more? And so on. The attentive reader will have noticed already that some of the ‘wh-questions’ are mixed up, but the point should be clear. Part 1 (this volume) is largely devoted to the Why? questions: treatment of the What? and How? questions—Parts 2 and 3—are postponed to a separate volume.
Scope
In this book I’ll restrict attention to research questions at the heart of what might be called ‘classical’ psycholinguistics, which concerns itself with two historically separate lines of inquiry: EXPERIMENTAL PSYCHOLINGUISTICS, which has mainly been concerned with theories and models of adult language processing, and DEVELOPMENTAL PSYCHOLINGUISTICS, where researchers’ primary focus has been on how children come to know and use their first language(s). Previously, the line between these two areas was fairly clearly demarcated by differences in the technologies applicable to each, and their associated modes of analysis. Early experimental psycholinguistics was invariably laboratory-based, employing technologies—and research assistants—that could not be easily used with children: too many heavy monitors, too few social skills. By contrast, early developmental psycholinguistics tended to be based on longitudinal observations of children’s language development, archetypally in the form of diary studies of researchers’ children. In current psycholinguistics, the distinction between the two areas is much less robust than it once was: technological advances have allowed most experimental methodologies that were traditionally used with adults to be adapted for much younger participants; at the same time, experimentalists have begun to investigate language processing in adult second language learners and other groups of multilingual speakers, as well the abilities of atypical language users.
The book will be concerned with both of these major strands of psycholinguistic research. In terms of the experimental studies to be presented in Part 2, the focus is on psychological/behavioural measures of human language performance: language experiments that elicit some kind of overt behavioural response, such as a naming decision (‘that’s a frog’), lexical decision (‘that’s an English/German/Turkish word’), a truth-value judgment (‘Yes, that’s true’) or a rating of some kind (‘That sentence sounds OK to me’); alternatively, chronometric studies which measure the time taken to respond to a verbal stimulus in different item conditions, or which track overt eye-movements during reading, or listening (preferential-looking tasks, for example), or those which record gestures accompanying speech, and so forth. Part 2 discusses published experiments involving all of these measures, and others besides.
In recent years however, many psycholinguists have moved beyond these established paradigms to embrace more neurophysiological measures of brain activity associated with language processing, including brain localization research: the use of ERP and fMRI measures, for instance. Although I will summarize the main findings of some key studies in order to bring the reader up to date, I shall largely limit attention to more traditional kinds of experimental data. One practical reason for this restriction is precisely that most classical psycholinguistics experiments are not ‘rocket science’: they can be carried out by anyone equipped with a personal computer, a reasonable degree of motivation, and some basic instruction in experimental design (something that this book also aims to provide, especially in Part 3). Neuroscience, on the other hand, makes literal rocket science look like a trivial exercise in trial-and-error ballistics; its experimental paradigms are correspondingly complex and intricate. Neurolinguistic experiments currently require extremely expensive equipment and laboratory time, trained and skilled technicians to run the experiments and analyse the raw data, and—not infrequently—fairly extensive ethics procedures. It is also considerably harder to recruit participants for neurolinguistic studies if you don’t have access to a pre-registered pool of volunteers. For all of these reasons, unless you are a very fortunate student, it is unlikely that you will be able to run your own neurolinguistic experiment. But almost anyone (equipped with ethics clearance for such a task) can design and run one of the psycholinguistic experiments discussed in Part 2, detailed in Part 3, either to probe previous results (by trying to replicate an earlier experimental design with a different group of participants, or different set of materials); or by modifying a previous design in order to test some original experimental hypothesis. In short, thanks in large part to advances in software development, classical psycholinguistics wins hands-down over neurolinguistic research in any cost-benefit analysis of the best way to spend research time.
In addition to these practical considerations, I remain to be convinced that, even if all the necessary technical and human resources were available, it is worthwhile carrying out such experiments, given our current ignorance of applicable ‘bridging theories’ to connect neurolinguistic results to psycholinguistic theory. This point echoes remarks by the cognitive scientist Gary Marcus, in a recent New York Times article:
‘What we are really looking for is a bridge, some way of connecting two separate scientific languages — those of neuroscience and psychology. Such bridges don’t come easily or often, maybe once in a generation, but when they do arrive, they can change everything. An example is the discovery of DNA…Neuroscience awaits a similar breakthrough. We know that there must be some lawful relation between assemblies of neurons and the elements of thought, but we are currently at a loss to describe those laws. We don’t know, for example, whether our memories for individual words inhere in individual neurons or in sets of neurons, or in what way sets of neurons might underwrite our memories for words, if in fact they do.’
Other languages, other language learners
A distinctive feature of the book, aside from its tri-partite structure, is the focus on data from languages other than English. The main purpose of presenting non-English examples is to draw attention to the ways in which alternative forms of construal and different patterns of phonological, lexical and grammatical organization influence models of language processing, and force a reconsideration of too narrow constraints on theories of language acquisition. Our own language invariably appears to us to be the most natural, logical, reasonable, and economical way of verbalising our thoughts—of moving from ‘intention to articulation’, as Levelt (1989) expresses it—but that is only because it is our native language.
A discussion of one phenomenon, speech segmentation, will suffice to illustrate this point. Take the nonce word vaitch. To an English native-speaker, it seems self-evident that the word comprises three ‘speech sounds’ (phonemes) contained within one syllable [v-ai-tʃ]. On the other hand, to a native-speaker of Japanese it is just as obvious how the same string should be analyzed: as three ‘sounds’, three MORAE [ba-i-tʃi]; バイチ in katakana, the syllabary used by literate Japanese speakers for most non-native Japanese words, including nonce words. Two radically different analyses, then: what counts intuitively as a discrete phonetic constituent in one language has no readily accessible correlate in the other. It is certainly possible for a Japanese listener to analyze a word like tako (たこ‘octopus’) as containing four phonemes (t-a-k-o), but this analysis is as unintuitive for him or her as it is for an English listener to treat the Japanese word fukurō ‘owl’ as containing four morae (ふくろう, fu-ku-ro-u), as opposed to three syllables, which is the preferred English analysis.
It should be observed immediately that neither analysis is ‘out there’, in the acoustic signal: English and Japanese speakers assign their own internal analyses to the same continuous and un-modulated phonetic stimulus, visualized in Fig. 2 below. (Whatever corresponds in to) phonological segments, like almost all of the more interesting objects discussed in Part 2 of this book, pertain to a level of psychological—rather than external, acoustic reality.
Of course, this property is not exclusive to linguistic analysis. Devlin (1998: 96) makes a similar observation regarding (the) calculus:
‘[The] methods of the calculus say as much about ourselves as they do about the physical world to which they can be applied with such effect. The patterns of motion and change we capture using the calculus certainly correspond to the motion and change we observe in the world, but, as patterns of infinity, their existence is inside our minds. They are patterns we humans develop to help us comprehend our world.’
The distinction between syllables and morae (the Japanese term is haku 拍) is brought out particularly sharply when we consider the Japanese verse-form known as haiku. As a child, this verse-form was unknown to me; Wordsworth and Tennyson were de rigueur in school, with Hilaire Belloc or Ogden Nash thrown in for light relief. These days however, it seems that every primary school class involves at least one annual stab at haiku, as though brevity was the guarantor of poetic accomplishment. Or perhaps brevity is its own reward: it must be easier to mark 3rd-grade attempts at 17-syllable completeness than to trundle through re-hashes of The Charge of the Light Brigade.
The problem with the assignment is that traditional Japanese haiku doesn’t involve 17 syllables, as is commonly supposed, but instead calls for 17 morae, arranged in a 5-7-5 configuration. To appreciate the difference, have a look at the two haiku in (1) below—presented together with their transliterations and free English translations—and try to decide which best conforms to the traditional metrical scheme:
江戸の雨何石呑んだ時鳥
Edo no ame/Nan goku nonda/Hototogisu.
'Of Edo’s rain/How many gallons did you drink/Cuckoo? (Issa)
富士の風や扇にのせて江戸土産
Fuji no kaze ya /Ōgi ni nosete /Edo miyage.
The wind of Mt. Fuji/I’ve brought on my fan!/ A gift from Edo. (Bashō)
If you read the transliterations of these haiku as though they were English, you may conclude that neither of these poems is very well-behaved: the poem on the left appears to consist of only 15 syllables (5-5-5), while that on the right seems to contain the correct number of syllables, but in the wrong configuration (6-6-5). In fact, the left-hand poem conforms strictly to the traditional verse-scheme, while the right hand one breaks the classical rule by containing an extra mora in the first line, but is otherwise complete. Most significantly, the second line of both poems contains exactly seven morae: na1-n2 go3-ku4 no5-n6-da7, and o1-o2-gi3 ni4 no5-se6-te7, respectively. A ‘fact of analysis’ that is as transparent to a native speaker of Japanese as end-rhyme is to a four year-old English child; see Part 2.
This brief discussion of the syllable/mora distinction shows that by examining data from languages other than our own we discover that a lot of what seems to be reflexive cannot be innate: typically developing children may be born with the capacity to acquire and to process any language, but the particular systems of categorization and analysis they end up regarding as intuitive arise through rich experience and extensive interaction with other language users, as well with discourse and text (language in the environment).
By appreciating these cross-linguistic differences in representation and processing, we also gain greater insight into our own language. Goethe asserted that “Those who know nothing of foreign languages know nothing of their own.”—or better, “Wer fremde Sprachen nicht kennt, weiß nichts von seiner eigenen.” ‘Nichts’ is an exaggeration, but the aphorism holds an important truth.
The book also gives unusual priority to the acquisition and processing of non-native languages, and especially to second language acquisition in adults and school-age children. This is a reversal of usual practice in language acquisition research, which generally takes native language acquisition and processing as a base-line—the standard of comparison—and then examines the ways in which second language learners’ abilities relate to this standard. As a result, second language acquisition is often characterised in terms of deficiency-based models, where second language learners fall short of native-speaker norms. By altering the perspective in this way, I hope to be able to shed a different light on core issues in language representation and processing.
It may also have been noticed that the title of this section is ‘Other languages, other language learners’. The plural is significant: real children do not acquire ‘Language’, they acquire (varieties of) English, Hindi, Thai, Fijian, Malayalam, and so on. Nor do real speakers process ‘Language’, rather, they process (varieties of) these different languages. Though this may seem to be a subtle distinction, it involves a crucial shift of perspective from philosophical (Platonic) abstraction to empirical investigation. I’ll suggest that this shift is indispensible if we want to understand what is in the minds of language users. The reification of Language is anything but harmless.
Just where it now lies I can no longer say
I found it on a cold and November day
In the roots of a sycamore tree where it had hid so long
In a box made out of myrtle lay the bone of song
The bone of song was a jawbone old and bruised
And worn out in the service of the muse
And along its sides and teeth were written words
I ran my palm along them and I heard
Lucky are you who finds me in the wilderness
I am the only unquiet ghost that does not seek rest…
From Josh Ritter Bone of Song [first part]
A route-map
Pour l’enfant, amoureux de cartes et d’estampes,
L’univers est égal à son vaste appétit.
Ah! que le monde est grand à la clarté des lampes!
Aux yeux du souvenir que le monde est petit!...
As I had hoped was clear from the subtitle, this volume of the book is intended as a conversation rather than as a monograph (or manifesto). Yet I’d be lying if I claimed to have no higher agenda. I care deeply about the use of language and the intellectual value of linguistic analysis, and want to convey the passion I feel about issues in language and mind to as broad an audience as possible. Linguistics offers us a framework and a set of tools with which to explore one of the most inherently fascinating and complex aspects of shared human experience. It should be fun—and given the right approach—it should be accessible to any intelligent reader who is prepared to make some effort. Yet too often linguistics comes across as leaden, bombastic and irrelevant, and a good deal of this impression is due to the fact it is too abstract, and often unnecessarily technical. Of course, some abstraction is essential, otherwise we can’t say anything interesting; some technical terminology is unavoidable if we are going to draw useful distinctions. But it is easy to get carried away by jargon or theoretical aesthetics, to ‘lose the plot’: the search for the simplest and most elegant theory results in a dismissal of the very phenomena it was intended to account for.
Some of my professional colleagues display a zealous obsession with the first part of Einstein’s dictum reproduced below, while ignoring the clause that follows:
‘The supreme goal of all theory is to make the irreducible basic elements as simple and as few as possible…
…without having to surrender the adequate representation of a single datum of experience.’
Albert Einstein (1934)
Hence, to the extent that this book has a serious purpose, it is to try to get to grips with the stuff of languages (that plural –s, again), to remind the reader what it is exactly that must not be surrendered.
For all that, the book is mostly intended as the academic equivalent of taking the dog for a weekend ramble across the moors. We’ll get there eventually, but the value is in the scenery along the way, not in the shortest distance between two points. If what you were looking for was a quick and dirty guide to psycholinguistics, you’re looking in the wrong place. (Given that we’re on page 30—13, excluding the Preface—you’ve probably figured this out already.)
Nevertheless, there may be some readers who might be prepared to follow me up and down the many “rabbit-holes” in the text—to use a prospective publisher’s analogy—but who’d still like to know where the conversation is going and, approximately, how we are going to get there. If you are one of those people, here is a brief route-map of the next 350-odd pages.
Part 1 consists of five sections. Section I offers a brief introduction to the intellectually fragmented world of classical psycholinguistics, in which I touch on some of the key research questions in experimental and developmental linguistics. I begin with the fairly uncontentious claim that the answers given to such questions over the last half-century have been framed in reaction to the philosophical arguments of one man, Noam Chomsky, whose views on language and mind have galvanized supporters and detractors in almost equal measure. Following a short historical overview, I first consider how Chomsky’s framing of the big questions led to a major rift in the theory and practice of psycholinguistic research. A proper understanding of why this rift came about involves an understanding Chomsky’s ideas from a wider intellectual perspective. In pursuit of this latter question, I spend some time considering the relationship between elementary-school knowledge of “times-tables” (declarative knowledge) and the (more abstract) algorithmic properties of arithmetic, an analogy that I’ll return to several times in the course of the book. This is the first major rabbit-hole.
Which then leads on to a discussion of the interchange between the ideas of Chomsky and those of the neuroscientist and psychologist David Marr, a key proponent of what became known as the Computational Theory of Mind (CTM). Emerging from this—and so equipped with a slightly better knowledge of the algorithmic differences between humming birds and eagles—we return to the main path. And to an interim conclusion, namely, whatever position one takes on “Level 1 questions”, it is vital to be able to distinguish (at Level 2) between linguistic representation and process, between declarative and procedural knowledge.
Deciding this question in practice is a really difficult problem. In Section II, I consider six different grammatical phenomena, any of which, in principle, might be essentially declarative or procedural in nature. (Six Different Ways). In each case, there is presumably a fact of the matter, though with scope for both individual and cross-linguistic variation. The research evidence for and against in each instance is discussed in greater detail in Part 2. My purpose here is to use these test cases to explore the intricate nature, and near-infinite variability, of our knowledge of languages. Each case can be seen as a separate rabbit-hole, though warren might be a more appropriate allusion. Indeed, the whole of this section could be skipped by readers who don’t need to be persuaded of the difficulty of the task. But those same readers would be missing out on a more entertaining diversion than this summary implies.
Section III presents a glossary of idealizations. Psycholinguistics, like all scientific research, is chock-full of apriori assumptions and idealizations. Most of these appear innocuous—reasonable, almost—when considered in isolation. In interaction, however, they can produce significant distortion; in some cases, the idealizations can lead to absurd conclusions that threaten to undermine the value of the empirical research on which they are based. (Now there’s a nice Escherian sentence to be getting on with.) So, in this section, through a set of largely self-contained essays, I offer a critical examination of some of the key notions that have underpinned psycholinguistic research over the last fifty years. The general intention is not to reject these idealizations outright, but to give the reader a clearer appreciation of their immediate implications, as well as to point out some of their unintended consequences. An extended caveat lector to Part 2, and perhaps to research in linguistics more generally.
Section IV focuses attention on two case-studies in language acquisition, taking the second language first. The French Class sketch (by the English comedian Catherine Tate and Aschlin Ditta) is not only a brilliant example of comedic writing, it is also an object lesson in second language learning, with significant implications for psycholinguistic theories of acquisition and processing, in which Lauren’s use of English is at least as interesting as her command (or otherwise) of spoken French. Another rabbit-hole, based on a comic fiction—or worthwhile diversion, depending on your frame of mind. In addition to discussing Lauren, I offer a different case study in this section—less hilarious, perhaps, but no less captivating (to me at least): the English of my middle son, Julian. Julianish offers a different kind of lesson, namely, how perfect generalizations can lead to ‘imperfect competence.’
So, there you are, a three-page route-map. I make no guarantees (as) to its accuracy, but warrant that the journey is more interesting, and the terrain more challenging, than the map itself suggests. But as I observed at the outset to this section, you really shouldn’t need a map. Baudelaire said it better…
…Mais les vrais voyageurs sont ceux-là seuls qui partent
Pour partir; coeurs légers, semblables aux ballons,
De leur fatalité jamais ils ne s’écartent,
Et, sans savoir pourquoi, disent toujours: Allons!
Endnote
では 始めましょう!
Comments