Comparing Estimates of Vocabulary Acquisition

Many estimates of vocabulary size exist, with variability being their one constant.  The difficulties inherent in measuring vocabulary size have not stopped multiple researchers from coming up with their own numbers, some of which I’ve summarized below.

Despite the almost inevitable variation, the studies that support these estimates have told us some important things, such as…

• School age language acquisition occurs primarily through incidental experience more than formal teaching.
• Word learning shifts from concrete and functional to abstract and unusual.  This shift occurs gradually from third grade through the high school years.
• Environment matters.  Extreme environments extremely matter. 

There are many more of these studies than even what I’ve listed, and while I’m not saying I’ve seen it happen, it is possible that people could cherry pick ones that most support the point trying to be made.  Also, there is no consensus among anyone really, of what exactly counts for a word in these studies.  Does examine, examines, examined, etc. count as different words or variations of the same word?  Ultimately the numbers themselves aren’t as important as are examining what’s possible and what’s actually occurring.  We know from these studies that it is possible to learn many, many words – at rates of up to 14 words a day according to at least two sources.  Methods of teaching vocabulary – such as teaching categories, word webs, and using reading to facilitate vocabulary acquisition – can be helpful, but ultimately nothing works like an enriched experience.

Keep reading for more information about the sources of these studies.

Anglin (1993) from Wagovich and Newhoff; The Single Exposure:  Partial Word Growth Through Reading; The American Journal of Speech Language Pathology, November 2004.

 Beck and McKeown (1991) from Beck, I.L. and McKeown, M.G. (1991). Social studies texts are hard to understand: Mediating some of the difficulties. Language Arts, 68, 482-490. from http://balancedreading.com

 Goulden, Nation, and Read (1990) from this link:  http://www.fltr.ucl.ac.be/fltr/germ/etan/bibs/vocab/cup.html

 Lipsett/ Mehrabian and Owens numbers are from Language Development  – An Introduction; Robert E. Owens, Jr.; Allyn and Bacon; 1996

 Hart and Risely (1995) taken from American Educator, Spring 2003, excerpt from their 1995 book, Meaningful Differences in the Everyday Experiences of Young American Children; Brookes Publishing

 Miller and Gildea (1997) from Later Language Development; Marilyn A. Nippold; Pro-Ed; 1998

We often seem to teach something for a long time before progress is made.  Then, it all seems to click, and suddenly the target is achieved.  In language, after this “click,” there is usually no need to continue teaching the structure.  The click factor encompasses two frequently observed phenomena:  1)  a student will use a target structure at a low percentage for some time, then suddenly use it at a high percentage.  2)  a student will not use a target structure at all, until being taught, after which time the student will suddenly use it at a high percentage.  There seem to be two reasons that this occurs.  One is that children may go awhile without a real world need for a target structure.  For example, Joe has been exposed to the word “she” in speech therapy, but with no sisters, and inconsistent correction from his parent on other occasions when the word “she” has been needed, he has continued to use “he” time and time again.  One day he refers to his grandmother as “he,” and is corrected by his grandfather.  Suddenly, it all clicks!  He realizes the reason for previous frustration, he knows how to eliminate this frustration, and he begins using “she” correctly.  If we’re all lucky, there’s quick generalization to other structures, and the goal of pronoun usage can be soon crossed off the SLP’s list.

Another possible contributor to this phenomenon is the measurement of progress that omits imperceptible increments.  Initially Joe may produce pronouns with no cueing at 20%, and may require extensive cueing to do anything beyond this.  With work, he soon may require less cueing, although his number correct without cueing may still be at 20%.  This sort of thing happens all the time.  It’s not necessarily problematic, though it can be if Joe moves or his SLP forgets exactly what Joe was previously working on.  For this reason, the more specific information a data sheet can provide, the more it can help describe exactly where Joe is at.  Forming good habits in criterion testing, such as testing often (such as with weekly quizzes), using structured data sheets, and including all relevant information can assist us in providing the best language teaching possible.

Accurate and frequent criterion referenced assessment can be the language teacher’s friend.  To accurately measure language progress several things are needed:  the method of elicitation, actual data, an accurate description of what’s being measured and the type and amount of cueing required.  The method of elicitation is what type of language is required from the student.  Data is a description of what the student has done.  Cueing describes what the teacher did to facilitate production.  Different types of each of these are listed below.

Aspects of Criterion Testing                 Examples of Types

Method of elicitation                            identification, choice, label, use

Data type                                              5 out of 5 sentences, 70% of choices with 2 foils

Cueing modality                                    verbal, visual, tactile, multimodality

Amount of Cueing                                 minimal, moderate, extensive, none

Statements of progress which omit any of these areas can not reliably be representative of a student’s actual skills.  Descriptions of cueing are often the most difficult to phrase.  It should be noted that classroom teachers do not include cueing information on their tests because the hindrances of the classroom setting preclude this.  A teacher is simply not able to give cues to large groups of students taking a classroom test.  It’s often not as essential in areas not as dependent upon social interaction as is oral language.  Information regarding cueing can be critical however to the language teacher that is able to administer and manipulate specific test items to specific students.  Consider the following examples, where nos. 1 and 3 are describing the exact same Joe, and nos. 2 and 4 are describing the exact same Sarah:

1)  Joe labeled subject pronouns (with no cues) with 20% accuracy.

2)  Sarah labeled subject pronouns (with no cues) with 20% accuracy.

3)  Joe labeled subject pronouns (with extensive verbal and visual cues) with 20 % accuracy.

4)  Sarah labeled subject pronouns (with extensive cueing) with 20% accuracy.

Without information regarding cueing, our knowledge of students usually looks like the first two examples above.  In these examples, Joe and Sarah seem to be at similar levels, while the last two examples tell something completely different.  Often we may know or feel that there is progress being made, but when we don’t attend to how much help is required, we miss a critical element of this progress.  For this reason, progress notes without measurements of cueing tend to contain weeks upon weeks of statements similar to the first two with seemingly little progress being made over large periods of time.  Ignoring the help required to elicit language requires the social dependencies fostering language acquisition.  This also ignores what many therapists know as the “click factor.”

A Summary of Patricia Kuhl’s Work:

Patricia Kuhl is co-director of the University of Washington’s Institute for Learning and Brain Sciences.  Her ongoing work, which began in the 1970’s, has altered how modern language theorists view the predisposition to language learning that infants are born with.  Prior to her work most followed the Piagetian view that babies were social isolates not yet ready to learn.  Kuhl’s research has changed this prevailing perception in a number of ways: 

• Human infants, as well as the young of other species such as birds, monkeys, and chinchillas, are born with an ability to distinguish between all sounds that exist in the particular language of that species.
• Human infants lose the ability to distinguish between sounds not in their language at about the same time that they begin producing varied babbling.  Humans have evidently evolved a predisposition toward learning a specific language.
• Parents have evolved specific techniques for teaching language – most prominantly is the high pitched, simplified version of language called “motherese.”  Kuhl’s research has shown a strong positive correlation between a child’s early language acquisition and the amount of “motherese” heard (or “parentese” as Kuhl diplomatically has called it).
• Interaction is crucial.  Babies that are not interacting as much do not learn language nearly as effectively, even if they appear to be attending to caregivers.
• The explosion in language learning that takes place between six months and three years of age in typically developing children seems to be the result of a combination of a child’s innate ability to detect sound differences, a seemingly innate ability to apply computational strategies to make language learning more efficient (Kuhl calls this statistical learning), and a nurturing social setting.

A recent summary of Kuhl’s research backed opinions can be found here.  A very interesting Seattle Times article on Kuhl and her work can be accessed here.

I couldn’t decide on just one more language acquisition study, like I initially wanted to, so I’ll simply give out a few honorable mentions.

Eric Lenneberg and the critical period hypothesis - In 1967 Eric Lennneberg released a widely influential book based on his research popularizing the notion that if language is not learned before an early age – usually estimated at 4 to 6 years – a child’s ability to learn any language becomes greatly compromised, or disappears altogether.  Though this research has been advocated for and debated against by linguistic giants such as Noam Chomsky and Steven Pinker, the evidence from Lenneberg and others is flimsy, draws extensively from widely divergent examples of feral children, and is largely theoretical.

Research hasn't proven this chimp knows what's you're thinking

Theory of Mind – Theory of Mind describes the ability to infer the mental states of others.  D.G. Premack and G. Woodruff initally espoused Theory of Mind in their seminal 1978 paper, “Does the Chimpanzee have a theory of mind?”  Research by Wimmer and Perner in 1983 used a famous false belief task to test study participants’ abilities to put themselves in others’ shoes.  Simon Baron-Cohen, Alan M. Leslie, and Uta Frith published research in 1985 suggesting that children with autism have deficits in theory of mind.  Research in this area has been widespread, divergent, and often theoretically powerful. 

Jean Piaget – The study of his own three children formed the basis for much of Piaget’s work.  Not strictly language, per se, but his view of language acquisition was extremely influential, while the middle ground belief (in terms of nature versus nurture) of cognition’s intertwining with language is probably closer to the truth than anything else currently out there.  Good info exists here and here.  This information is especially informative.

I can’t not mention Broca, Vygotsky, or Kegl’s study of Nicaraguan sign language.

Although Everett has been studying the Pirahã (pronounce pee-da-ha) Amazonian tribe, and their unique language since the 1970’s, his work remained relatively obscure until 2005, when an article he’d published on his website was then published in Cultural Anthropology.  According to Everett’s studies, the Pirahã’s language lacks many aspects of language that linguists argue are basic necessities of a universal grammar, such as color concepts, perfect tense, quantity concepts, and numbers over two.  Why?  According to Everett, their hunter-gatherer lifestyles have such little use for these concepts, that words to convey them simply don’t exist.  This research, which overtly repudiates the Chomskyian theory that has dominated the study of language for decades, has been called by Steven Pinker, “A bomb thrown into the party.”

The coverage of this study has greatly expanded recently, exemplified by this New Yorker article.  Language Log has some excellent links concerning the Piraha.

Brown described five stages of language development based on a child’s mean length of utterance (MLU).  His research demonstrated that MLU was a better predictor of what linguistic structures a child was able to use than was chronological age.  This research, which examined three children whom Brown dubbed Adam, Eve, and Sarah, was the ultimate explanation of language acquisition for years.  The complexity of Brown’s description has also, unfortunately, painted language acquisition as a complicated morass of agent+actions, entities+locatives, recurrences, and nominatives that quite frankly, has turned off many students (especially speech-language pathology students) from this entire area.  The structural analysis of language samples based on Brown’s language description is a staple of the SLP college experience often remembered with revulsion.  Despite this, the influence of this study can not be denied.  Neither can it’s untouched accuracy in describing the process of language development.

The best online description of Brown’s work is at Caroline Bowen’s speech therapy site.

After decades of collaborating to increase child language vocabulary, Betty Hart and Todd Risley spent 2 1/2 years intensely observing the language of 42 families throughout Kansas City.  Specifically, they looked at household language use in three different settings:  1) professional families; 2) working class; 3) welfare families.  Hart and Risley gathered an enormous amount of data during the study and subsequent longitudinal follow-ups to come up with an often cited 30 million word gap between the vocabularies of welfare and professional families by age three.  This number came from the data that showed welfare children heard, on average, 616 words per hour, while children from professional families (essentially children with college educated parents) heard 2153 words per hour.  The longitudinal research in the following years demonstrated a high correlation between vocabulary size at age three and language test scores at ages nine and ten in areas of vocabulary, listening, syntax, and reading comprehension.  This study was subsequently used to fuel the fire of arguments for early childhood programs such as Head Start.

For an excellent summary of this study, read this.  A good comment on this study and poverty’s influence on education can be found here.

Over the next few days I will be describing some of what I feel are the language acquisition studies marked by their significance to both our current knowledge of language acquisition as well as historical impact upon subsequent research in the field.  Without any adieu, and with no particular order, here they are.

image from (http://jean.berko.gleason.googlepages.com/home)

Jean Berko Gleason’s “Wugs”  – 1958

Berko Gleason and colleagues presented pictures of imaginary creatures to children.  The pictures were given labels such as “wug,” made up by the researchers.  The children were then presented with varieties of the make believe creatures to test their ability to apply linguistic rules.  The famous example is “This is a wug.”  (1 wug)  “What are these?”  (More than one.)  Very young children had difficulty, but children by age 4 or 5 could usually label the plural “wugs,” and most importantly – could do it without ever having heard the word used before.  These sorts of pictures were also used to test other aspects of syntax acquisition, such as possessives and verbs.  The nativists have long used this as evidence that language is not memorized.  A shortened explanation of what I think is going on can be found here. 

In an interview found on her website, Berko Gleason gave an eloquent explanation for what she thinks is going on.

I am inclined toward a parsimonious view that sees linguistic behaviour as similar to other behaviour that relies in part on a general cognitive capacity to recognize patterns in input and to generalize. That doesn’t mean that a part of our brain isn’t, at some point, dedicated to language: Babies build their brains, strengthening some connections and pruning others out, based on their experience. The fact that babies can learn to talk and my cat Wolfie can’t means that experience isn’t everything—humans have some genetic capacity to acquire language that cats do not. This is also true of playing the piano and dancing the csardas. But when people talk about language ‘mechanisms’ or ‘hard wiring’ I think we need to remember that these are metaphors, and the human brain may not function like the machines they are based on.

Read more about wugs at this improbable research blog.

A child with a deficit in a skill typically has not discovered the power of that skill.  Thus remains the initial opening for novelty.  I believe that children are often more open to suggestion than we often give them credit for.  In other words, initially discussing the benefits of a skill can be an extremely effective introduction to the teaching of a skill.  However, because complex language is not yet a favored method of input for children in language therapy, these explanations can be brief.  Why are working on verbs?  Because every sentence has them, and with them you can talk about what anything does.  Why practice comparatives and superlatives?  Because with them we can greatly increase our powers to describe.  And it always helps to relate these introductions in personal ways.  Statements such as, “With superlatives you can tell me that you are a faster runner than your brother.” tend to work well.

Activities themselves provide ample opportunities to motivate, as does the presentation of the activities.  If you’re working on past tense you use stories, because stories involve a lot of past tense.  If you’re working on pronouns activities that frequently refer to other people should be used to achieve goals.   This is often easier than it sounds.  Young kids working on pronouns can get tons of exposure to pronouns in many simple group turn taking activities.  An example of how to structure this would be to give two boys and a girl different types of blocks.  Questions such as “Whose block is this?” and “Who gets the big block?” provide natural incentives to learn the exact pronouns absent with in any child’s repertoire.  There are plenty of activities that work equally well in this manner, including puzzles, sticker activities, dot to dots, and coloring activities.  Because nothing in language exists that does not assist, a demonstration of how these units assist can utilize their inherent motivational powers.

 A wide variety of activities assists us in providing the best presentation.  Worksheets, games, drill, and quizzes are all ways to help learning, but each type of learning loses potency quickly when done successfully.  In other words, if you do nothing but ________, or even do _________ a few times in a row, where any activity type can go in the blank, there is a natural tendency for children to lose learning interest.  Research is extremely robust that supports the contention that repeated and varied exposures greatly assists memory (Dempster and Farris, 1990).  This even applies to games.  Fun can be one of the most effective motivators, as young children especially learn primarily through play.  It is human nature to seek to expand, and any activity which only uses what a child already has, not only grows quickly boring, but also proves unsatisfying.  Games like Go Fish can effectively provide natural motivation, but unless the specific student needs specific help with questions or helping verbs, activities like this should usually be “thrown back.”  Preparation and notation are two key ways to prevent falling into the habit of consecutive presentations.

In this article from Nature magazine, psychologist Paul Bloom discusses research suggesting that infants are born with the ability to distinguish phonology and certain aspects of meaning.  He specifically discusses a study that suggested babies are born with the ability to distinguish between loose fits and tight fits.  In English we lose this distinction, while in the Korean language the distinction is maintained with two separate verbs.  This research (and research like it) involves showing 5 month olds things like a ring around a post that is initially tight, until they bore of it and look away.  The researchers then show them the varying instances of the same or the converse – like a ring fitting loosely around a post.  If the babies seem more interested in the contrast they are said to be innately predisposed, which is in fact what happens.  For the study’s authors, Paul Bloom, and even St. Augustine, this kind of thing is considered evidence that we are all born with cognitive precursors to language – a variation of the famed Sapir-Whorf Hypothesis.

Read on for my take… 

Could the explanations for the babies’ interest possibly be an innate predisposition toward novelty rather than the specific instances of novelty?  For one thing, by repeatedly exposing the babies to the stimulus (other studies use sounds) you are teaching them that the stimulus exists.  Their brains are being wired with something like this thought:  “Hey that sound exists!”  The innate part may be that once our brain is wired in such a manner – once it knows some bit of information – it is no longer as interested in it.  A new sound may pique interest because we are innately predisposed for novelty.

Could the explanations for the babies’ interest possibly be an innate predisposition toward novelty rather than the specific instances of novelty?  For one thing, by repeatedly exposing the babies to the stimulus (other studies use sounds) you are teaching them that the stimulus exists.  Their brains are being wired with something like this thought:  “Hey that sound exists!”  The innate part may be that once our brain is wired in such a manner – once it knows some bit of information – it is no longer as interested in it.  A new sound may pique interest because we are innately predisposed for novelty.

Could the explanations for the babies’ interest possibly be an innate predisposition toward novelty rather than the specific instances of novelty?  For one thing, by repeatedly exposing the babies to the stimulus (other studies use sounds) you are teaching them that the stimulus exists.  Their brains are being wired with something like this thought:  “Hey that sound exists!”  The innate part may be that once our brain is wired in such a manner – once it knows some bit of information – it is no longer as interested in it.  A new sound may pique interest because we are innately predisposed for novelty.  Continue reading for my take…

Could the explanations for the babies’ interest possibly be an innate predisposition toward novelty rather than the specific instances of novelty?  For one thing, by repeatedly exposing the babies to the stimulus (other studies use sounds) you are teaching them that the stimulus exists.  Their brains are being wired with something like this thought:  “Hey that sound exists!”  The innate part may be that once our brain is wired in such a manner – once it knows some bit of information – it is no longer as interested in it.  A new sound may pique interest because we are innately predisposed for novelty.