O n f e r e n c e

Input Processing: redundancy and meaningfulness 
VanPatten also proposes principles of input processing that consider redundancy and 
meaningfulness, given in (5)-(7) below. 
Redundancy
refers to the expression of the same 
information in multiple places, either lexically or inflectionally. For example, the sentence 
“He told me about it yesterday” encodes the past tense twice, lexically by “yesterday” and 
inflectionally by “told”. 
Meaningfulness
refers to whether a given form contributes to the 
communicative goal of the sentence, as opposed to merely being a requirement of the 
grammatical system, such as grammatical gender or case agreement (e.g. Lee 1987). 
5.
The Meaning-Before-Nonmeaning Principle
: Learners are more likely to process 
meaningful grammatical forms before nonmeaningful forms irrespective of redundancy. 
6.
The Lexical Preference Principle
: Learners will tend to rely on lexical items as opposed 
to grammatical form to get meaning when both encode the same semantic 
information. 
 
7.
The Preference for Nonredundancy Principle
: Learners are more likely to process 
nonredundant meaningful grammatical form before they process redundant meaningful 
forms. 
Principle (5) is usually irrelevant for the purposes of OTI, since token sentences containing 
a given target grammatical form are identical with regard to the meaningfulness of that 
form.
8
Automatic analysis of principles (6) and (7) requires specific considerations of each 
particular target grammatical structure and language. In other words, assessing 
grammatical redundancy cannot be done with any general formula or algorithm. However, 
one general approach applies to most cases.
9
For each target grammatical structure, a list 
of potentially redundant lexemes and part-of-speech tags is compiled.
10
The occurrence of 
these items in the same clause as the target grammatical structure is then tagged as 
redundant. Returning to the earlier example targeting past tense “He told me about it 
yesterday”, the ICALL system simply needs a precompiled list of words and phrases 
encoding the past tens
e. Upon finding “yesterday” in that list, it can tag “yesterday” as 
being grammatically redundant with respect to the target grammatical structure (past 
tense “told”).
It should be noted that SLA research of input processing generally casts lexical redundancy 
of grammatical meaning as a negative thing, since it distracts learners from target 
grammatical structures (VanPatten 2004). This is under the assumption that students are 
in “consumption mode”, i.e. trying to decipher the meaning of a text or utteran
ce. 
However, in the context of form-based exercises, redundancy can be leveraged in ways 
that open new possibilities.
7
This definition assumes that the target language structure can be assigned a single position. For multiword 
constructions, an average position can be used, or the position of one particular piece of the construction can be 
treated as representing the construction as a whole. 
8
There are some cases in which a single surface form may or may not be communicatively meaningful (e.g. 
gender agreement can be grammatical or referential). However, since identifying such cases requires higher-level 
analysis (e.g. discourse analysis, anaphor resolution, etc.), I leave this problem for future work.
9
Since authentic text ICALL applications will virtually always be based on some kind of automatic morphological 
analysis, I assume the availability of part-of-speech tags for redundancy analysis. 
10
If suitable corpora are available, then these lexical redundancies can be extracted automatically, based on 
collocational probabilities.

---

-296- 
2014 CALL Conference 
LINGUAPOLIS
www.antwerpcall.be 
For example, tokens that contain multiple expressions of the same information are 
excellent candidates for automatically-generated exerci
ses. For example, the token “He 
told me about it yesterday” can automatically be transformed into “He _____ me about it 
yesterday”, with either a multiple
-choice or fill-in-the-blank response method. 
The primary advantage of this kind of cloze exercise is 
that the context itself (“yesterday”) 
indicates that the correct form should be in the past tense 
–
as opposed to an explicit 
instruction such as “Fill in the following blanks with the correct past tense form.” This 
approach is generally highly recommended by SLA researchers, since it reinforces form-
meaning connections, while at the same time requiring comprehension of the entire phrase 
(e.g. Farley 2005). Furthermore, by eliminating the need for explicit instructions, this 
approach makes it possible to address multiple target language structures at the same 
time. For example, if a learner has already mastered several target language structures, 
an authentic text ICALL system could provide review activities in which generated 
exercises target more than one structure.
The ability to target multiple language structures has the added benefit that it expands the 
set of documents that can be felicitously implemented in authentic text ICALL. Although 
the primary motivation of the present study is an overabundance of tokens of the target 
language structure in a text, the opposite problem is also common. A large number of 
texts do not contain enough tokens of the target language structure to be useful. An 
ATICALL system that can only work with one target language structure at a time cannot 
make much use of a document that has only a few tokens of one target language 
structure, and a few tokens of another. However, the ability to create exercises for 
multiple target structures 
–
facilitated by the detection of redundancy 
–
can make some 
previously unusable documents more useable.