O n f e r e n c e

Input Processing 
Input processing, as defined by VanPatten (2004: 5), is the “initial p
rocess by which 
learners connect grammatical forms with their meanings as well as how they interpret 
roles of nouns in relationship to verbs”. VanPatten proposes ten principles for predicting 
the likelihood that a learner will make form-meaning connections from a given input 
string.
6
It should be noted that Vajjala & Meurers (2013) used insights from ESL learner corpora (Lu 2010; Lu 2011a; 
Lu 2011b) to expand their feature-sets for machine learning readability analysis, although their training and 
testing was performed on L1 corpora. 

---

-294- 
2014 CALL Conference 
LINGUAPOLIS
www.antwerpcall.be 
Informally, these principles can be seen as a set of limitations and biases that naturally 
confront second language learners as they try to understand utterances or texts in their 
target language. 
In the following sections, I 
discuss VanPatten’s principles of input processing with an eye 
toward automatic analysis. Those principles that cannot be implemented reliably with 
current language technology are discussed only briefly. 
Input Processing: cognitive load 
Not surprisingly, some of VanPatten's principles of input processing are concerned with 
cognitive load. The result of learners’ “push to get meaning”, combined with limited 
cognitive resources, is that some forms may not be processed. In a loose sense, these 
principles get at the same issue addressed by readability analysis, i.e. whether a text is 
unattainable or makes too many demands on cognitive resources. In fact, operationalizing 
these principles would yield features that are very similar to features already widely used 
in readability research, and so they will not be discussed extensively. Two of these 
principles are listed below: 
1.
The Primacy of Content Words Principle:
Learners process content words in the input 
before anything else. 
2.
The Availability of Resources Principle
: For learners to process [grammatical forms], 
the processing of overall sentential meaning must not drain available processing 
resources
. 
Determining the cognitive load of a sentence would be facilitated greatly by the availability 
of a student model specifying known vocabulary and grammar knowledge. This is 
especially true of early and intermediate learners, whose repertoire is limited. Currently, 
no authentic text ICALL systems make use of such student models. 
3.
The Sentence Location Principle
: Learners tend to process items in sentence initial 
position before those in final position and these latter in turn before those in medial 
position (Klein 1986; Barcroft & VanPatten 1997; Rosa & O’neill 1998).
A common sense explanation for principle (3) is that the first few words of an utterance 
consume available cognitive resources, leaving fewer resources for remaining parts of the 
utterance. In addition, words in utterance-final position are more easily processed than 
those in the medial position because the listener has the benefit of reflection. Although this 
principle has only been established empirically with auditory input, it is at least possible 
that an analogous effect is present with textual input, as well. Textual input is primarily 
processed linearly 
–
like an auditory signal 
–
but the reader’s ability to control pace and 
repetition may diminish or negate the sentence location effect. Without empirical evidence 
for a sentence location effect in textual input, operationalizing principle (3) is not 
straightforward. I suggest the following revision, but stress that it is untested. 
4.
The Sentence Location Principle
(for textual input): Learners tend to process items in 
sentence initial position before those in later positions. 
Since both auditory input and textual input are processed in a linear fashion, I assume that 
the first few words of a sentence do have a processing advantage. However, since the 
reader controls pace and can pause to reflect at any time, I assume that there is no 
processing advantage for words near sentence-final position. Given these assumptions, 
operationalizing (4) is very straightforward: promote token sentences in which the target 
grammatical form is among the first 
x
words of the sentence, where 
x
represents a 

---

-295- 
2014 CALL Conference 
LINGUAPOLIS
www.antwerpcall.be 
hypothesized cognitive processing threshold.
7
Although an arbitrary general-purpose 
number could be used for 
x
, Hulme et al. (1995) showed that the limits of working 
memory are dependent on the lexical status of the words being processed, i.e. whether 
they are stored in long-term memory. This means that the threshold 
x
given above should 
ideally change according to the learner’s knowledge. Once again, a student model 
specifying known words would aid dramatically in automatic assessment of principle (4).