Rules vs. Analogy in Spanish Irregular Verbs

PIs	Nora Presson, Brian MacWhinney, Nuria Sagarra
Faculty	MacWhinney
Postdocs	n/a
Others with > 160 hours	n/a
Study Start Date	04/01/10
Study End Date	06/01/10
Learnlab	n/a
Number of participants (total)	105
Number of participants (treatment)
Total Participant Hours	~350
Datashop?	Not yet

Abstract

The goal of this experiment was to test explicit instruction of when irregular verbs take a change in the stem, and when they use a regular affixation pattern. We gave beginning Spanish students practice conjugating verbs that have irregularity in some inflected forms. Two irregularity types were compared: stem-change verbs, where the pattern of transformations is predictable based on a large gang of similar irregular verbs, and spelling-change verbs, where the pattern of transformations is predictable based on the spelling and the phonology of the inflected form and the infinitive. We compared explicit formulations of "when to change" the stem of a verb with an analogical comparison condition, where instead of a rule formulation, participants saw a similarly irregular verb they knew (judged by the instructor) conjugated appropriately as a model.

Glossary

• explicit instruction
• cue validity
• cue reliability
• cue availability

Research question

• What is the baseline accuracy (latency) for producing inflected verb forms for irregular verbs (i.e., verbs where at least one form requires an idiosyncratic transformation)?
• Do forms that require a transformation (irregular forms) show lower accuracy (higher latency) than forms that do not, even when the verb itself is irregular?
• Do spelling change verbs, where the change is predictable from the spelling of the infinitive, show different response patterns than stem change verbs, which are not visible from the infinitive?
• Does providing a rule about when change of the stem is necessary improve performance compared to providing a known form that is transformed in the same way as the practice verb?
• How does the amount of improvement after practice depend on:
  • Regularity of the form
  • Instruction type (rule vs. exemplar)
  • Change type (spelling vs. stem)

Background and significance

Learners have difficulty processing inflected verb forms (Clahsen & Felser, 2006; Johnson & Newport, 1989). Some models of language processing postulate that this difficulty stems from learners' lack of ability to compose regular inflected forms as native speakers can (e.g., the Declarative/Procedural model of Ullman, 1997; 2001). This prediction stems on the assumptions that:

• native speakers store irregular forms but compose (some but not necessarily all) regular ones (see for example Pinker, 1997)
• learners are unable to process (decompose) morphologically complex forms in a nativelike way (possibly because of a Critical Period effect)

The evidence that learners treat regulars and irregulars similarly (from which most infer that both are stored) comes from priming studies, where native speakers show priming of the base stem for regular but not irregular inflections, and also from production data showing frequency effects on latency for irregular but not regular inflected forms.

However, most previous studies of the difference between regular and irregular forms do not distinguish between levels or types of irregularity. There are irregular forms that have predictable changes, and inflected forms that do not show any idiosyncratic transformation, though other forms of the same verb do.

One level of distinction comes from a recent study by Bowden and colleagues (2010) that compared Spanish -ar verbs, which are the most frequent type of verbs, are most often regular, and encompass most novel words, to -er and -ir verbs, which have fewer types and are more likely to be irregular. They showed that native speakers only failed to show frequency effects when producing -ar but not -er/-ir regular inflected forms in both present (irregularities possible) and imperfect (almost no possible irregularities) tenses. Also, learners showed frequency effects (which they interpret to mean forms are stored and not composed) for all types of verbs and tenses.

Most of the "irregular" verbs used in that study, however, do show predictable patterns that can explain the stem transformations. We are interested in whether these regularities matter; that is, whether a wholly idiosyncratic transformation is really processed the same way as an irregular that follows a predictable pattern.

We compare two types of irregularity: first, spelling change verbs, where a change is needed to preserve the phonological form of the verb (e.g., sacar "to take out" -> saqué "I took out" because the regular transformation would lead to sacé, with a soft 'c'.). We compared those to stem change verbs, where the change is not meant to preserve phonology but does follow a similar pattern across a large number of verbs.

Learner performance is full of errors, so we are interested in how to improve performance with practice and explicit feedback. In this study, we used two feedback types: an explicit explanation of when to transform the verb stem (i.e., which forms are irregular), and presenting a correctly inflected form of a highly frequent verb their teacher judged that they would know.

Dependent variables

• Response accuracy (% inflected forms typed correctly at first attempt)
• Response latency (seconds to type inflected form)

Independent variables

• Spelling change verbs compared to stem change verbs
• Practice with a rule (e.g., for arrancar, a spelling change verb, "c --> qu when ending begins with e or i"; for volver, a stem change verb, "o --> ue when syllable is stressed") vs. practice with an analogy (a verb following the same pattern of irregularity that their teacher judged would be familiar)
• Forms that do require a transformation to the stem (irregular) vs. forms that do not (regular)
• Test time (pre-test, post-test, 1-week follow-up)

Hypotheses

1. Performance will improve after 3 sessions of training for ~15 minutes. Post-test accuracy immediately after training as well as after a one-week delay will be higher than pre-test accuracy.
2. Irregular forms will show lower accuracy than regular forms, even though all forms come from "irregular" verbs.
3. Training will be more helpful for irregular than regular forms.
4. Feedback with an analogy will be more helpful with stem change verbs (unpredictable from infinitive form), whereas an explicit rule will be more helpful with spelling change verbs (predictable from the infinitive).