Sunday, February 8, 2015

Another preview: more on the Lexile Framework

Adapted from another passage I recently wrote for this book (...when really I should have been blogging):

One factor that educators consider in deciding which texts to assign to which students at which grade levels are the texts' Lexile ratings. These ratings are determined by a text-processing tool known as the Lexile Analyzer. The Lexile Analyzer measures two things: how challenging the individual words are (based on their frequency) and, as a proxy for syntactic complexity, how long the sentences are. The problem, however, is that sentence length correlates only weakly with the aspects of complexity that make sentence processing challenging. A relatively long sentence may be quite easy to process if it consists of a series of simple short sentences conjoined with a coordinating conjunction like “and”: 
 “I love you and you love me and we’re a happy family" 
while a relatively short “garden path” sentence like  
“The horse raced past the barn fell” 
can be quite difficult to process. Identifying the kinds of syntactic complexity that make reading challenging—particularly for children with Specific Language Impairment and autism—requires more sophisticated linguistic processing than that performed by the Lexile Analyzer.

One indicator of how complex a sentence is how embedded its subordinate clauses are. But where exactly these subordinate clauses are embedded also makes a difference. Generally, so-called "right-branching" sentences like: 
 “This is the cat that chased the rat that ate the cheese that lay in the house that Jack built”
no matter how much embedding they contain, are easier to process than corresponding "left-branching" sentences like:
"The cat’s prey’s cheese’s location’s builder was Jack."
Most challenging of all is what’s called “center embedding,” seen in sentences like:
 “The cheese that the rat that the cat chased ate lay in Jack’s house.”
Or sentences in which what looks like the most obvious structure turns out to be wrong, as in: “The horse raced past the barn fell.”

There are also factors beyond vocabulary level and syntactic complexity that figure in text difficulty. These include the challenges of deducing the antecedents for the various pronouns and other anaphoric devices, including underspecified noun phrases like “that idea” and “this strategy.” An automated text-rating system might search for pronouns and deictics like “this” and “that” and compare the ratio of such terms to the number of noun phrases. A high ratio would indicate a highly interconnected, internally-referential text that requires lots of inferences to determine antecedents and flesh out the content.

The amount of inferencing that readers must do, in fact, is a large part of how hard a text is. Beyond the inferences that determine antecedents for anaphors, there are a host of others, including the pragmatic inferences that make sense of dialogs, the social inferences that make sense of character interactions, and the perspective-taking inferences that make sense of actions in general. Such inferencing tasks are particularly challenging for children with autism. And, yet, they pervade most reading assignments, especially fiction. They are the reason why reading fiction is one of the most demanding school-based tasks that autistic children face. But nonfiction is also challenging. Many texts, fiction and nonfiction, require yet another sort of inference: inferences that draw on general background knowledge. General background knowledge is the type of knowledge that most children pick up incidentally from social interactions and overheard conversations. Autistic children, less attuned to these sources of information, commonly have knowledge deficits, and, therefore, further deficits in reading comprehension.

An automatic text analyzer might give rough estimates of social and emotional-based inferencing demands in the same way that the Lexile Analyzer gives rough estimates of syntactic complexity.  For example, it might search for social and emotional vocabulary terms and compute their density within the text. However imperfect a measure of this is of the social and emotional-based challenges for reading comprehension, it still would be a highly useful one, given that these challenges are a huge determiner of text difficulty for autistic readers, and given that they don’t figure at all in current text rating systems.


Anonymous said...

I don't think that I'm autistic, but I can't make any sense out of the sentence "The horse raced past the barn fell."
could you explain what the sentence means?

Katharine Beals said...

The sentence a famous example from linguistics: it involves a reduced relative clause in passive voice. The unreduced counterpart is "The horse that was raced past the barn fell."
The sentence is especially tricky because "race" is most frequently an intransitive verb; here it is transitive, but there's no passive "by clause" (as in "raced by the jockey") to give that away.

Auntie Ann said...

Wikipedia's pretty good on this:

Garden path sentences