Research Statement

I am interested in understanding how languages are learned, represented, and processed. Using a combination of experimental and theoretical methods, my research explores constraints on morphology, phonology, and speech perception. My goal is use the patterns that are found among languages of the world as a basis to form hypotheses about how the mind must work in order to produce these patterns.

Phonological Learning

My experimental work has explored the psychological nature of theoretical phonological constructs such as features (Finley and Badecker, 2009), phonetic grounding (Finley, 2011-b; Finley, 2012; Finley, in preparation; Finley and Badecker, 2012), and markedness constraints (Finley, in progress). In my research, I train naïve adult participants on phonological patterns to address several questions pertaining to linguistic representations. For example, I have explored the nature of hierarchical representations in learning phonological systems (Finley, submitted). In order to understand how the mind represents complex, hierarchically structured linguistic patterns, I compared how adults learn two phonological patterns that on the surface are very similar, but require very different linguistic representations. One type of pattern (deemed opaque vowels) requires only adjacent, flat-structured representations (e.g., aXb, where X and b are required to share a feature. A similar type of pattern (deemed transparent vowels) is represented with non-adjacent dependencies (e.g., aXb where a and b are required to share a feature). Adult learners were exposed to a vowel harmony pattern that required either adjacent representation (opaque vowels) or non-adjacent representations (transparent vowels) that required nonadjacent vowels to conform to the harmony pattern. Participants in both conditions were able to learn the basic harmony pattern, but participants in the non-adjacent (transparent) condition failed to learn the non-adjacent aspect of the pattern. A series of follow-up experiments demonstrated that increasing the training in terms of types (number of words containing the relevant non-adjacent sequences) as well as tokens (number of times each type was heard) is sufficient for participants to learn the non-adjacent pattern. These results suggest that the information required for learning a language is dependent on the complexity of the representation, even when the surface structure of the words may appear similar.

Another example of how seemingly trivial surface differences can result in large differences in representations comes from consonant harmony. Languages with consonant harmony allow an unbounded number of intervening segments between the source and the target for harmony (e.g., agreement of /s/ vs. /sh/ in the word form /VsVCVCVCsV/). My research (Finley, 2011, 2012) has demonstrated that it is possible to learn consonant harmony patterns across large distances (shown for up to four intervening syllables), but adult learners are nonetheless biased towards adjacent patterns. Learners generalize from non-adjacent patterns to adjacent patterns, but not vice versa. These results demonstrate that learners form representations for novel patterns based on their biases towards adjacent elements.

In my labs at Elmhurst College and Waldorf College, I have continued to explore complex representations in phonology. I am in the process of completing a series of experiments that supports the notion of markedness as motivation for phonological processes. Participants were trained on a metathesis pattern in which consonants switched places when two morphemes were combined (e.g., /pel + kob à peklob/). Participants were more likely to extend the metathesis pattern to novel forms when the pattern improved syllable structure, suggesting that the representation for the pattern was not an arbitrary switching of sounds, but a pattern grounded in the markedness of syllable structure.

Theoretical Phonology

I also make use of theoretical phonology (Optimality Theory) in order to explore the hypothesis that a better understanding of phonological representations will lead to a more typologically plausible theory phonology. My approach to vowel harmony builds on the insights of Turbidity Theory (Goldrick, 2001), which allows for a third, abstract level of representation. This abstract level can be used to modulate spreading, allowing transparent and opaque interactions in vowel harmony to be captured without post-hoc mechanisms. Enriched representations allow the analyst to regulate exactly which structures are attested and which are not (Finley, in prep). To ensure that my theory of representations does not predict any unexpected pathologies, I make use of finite-state machinery to evaluate possible outputs (‘contenders’) using methodology developed by Riggle (2004) (Finley, 2010). Further, I have analyzed the interaction of epenthesis and vowel harmony, specifically the nature of non-participating epenthetic vowels (Finley, 2009).

Morphological Learning

During my postdoctoral research at the University of Rochester, Dr. Elissa Newport and I developed a paradigm to understand how language learners determine which words are morphologically related. While previous research has focused on word meaning as the driving force in learning to segment morphemes, we hypothesize that learners make use of a variety of distributional properties, such as frequency and similarity. These distributional properties of words – the parts that recur across a number of words in the language – allow the learner to determine the stems and the affixes from which words are constructed.

In our first set of experiments, we demonstrated that adults (Finley & Newport, 2010; in preparation) and school-aged children (Finley & Newport, 2011) use distributional information to parse complex words into stems and suffixes. I am currently testing children ages 6-12 to determine whether the ability to parse stems and suffixes is as strong for stems and prefixes. Cross-linguistically, there is a bias for suffixing languages compared to prefixing languages. While our adult data did not show any differences in learning suffixes, our preliminary data with children suggests a bias towards suffixes over prefixes.

We also addressed how adult and child learners use distributional information to parse non-concatenative morphology. Learners were exposed to a simulated version of this type of pattern in which the consonants (ptk, mbg) and the vowels (ae, ou) formed separable units (e.g., patek, potuk, mabeg, mobug). Our results demonstrate that English-speaking adult learners are biased to parse the words in terms of syllables (e.g., patek as pa + tek), and did not show evidence of parsing words into consonant patterns. This bias was overcome when the distributional information highlighted the consistencies of the consonants and the vowels. First, when words had prefixed forms in addition to the non-concatenative forms (e.g. patek, mu-patek), learners were able to parse consonants as separable units. Second, learners were able to parse the consonants when the variability of the vowels exceeded the variability of the consonants (e.g., 12 consonant patterns, and 24 vowel patterns). Third, learners were able to parse consonants and vowels when additional consonant-vowel structures were added in addition to the CVCVC shape used in previous studies (e.g., CCVC, CVCCV, etc.). These results suggest that learners are highly sensitive to the distributional properties of morphological systems, even when the morphological patterns are quite different from those of their native language. These results were recently replicated with children (Finley, talk to be presented at the 2014 LSA).

More recently, I have worked with undergraduate Elizabeth Wiemers at Elmhurst College to develop a paradigm to study how complex patterns of morphology are learned (Finley & Wiemers, 2013). For example, a language that marks number (singular, dual and plural) as well as gender (masculine, feminine, neuter), will have potentially nine difference markers (e.g., one for singular-masculine, one for feminine-neuter, etc.). In order to understand how languages with complex markings are able learned, we have experimented with different cues to learning. For example, learners exposed to a language that marked all nine potential markings were more likely to learn these markings when the language contained phonological cues (e.g., every singular form starts with /p/). This result supports the hypothesis that the patterns found in language are not coincidental, but help the learner to find consistencies between sound and meaning (e.g., German’s determiners der, dem and das all start with /d/).

Future Work

I am fully committed to an interdisciplinary research program that incorporates multiple methodologies in order to better understand spoken language and the relationship between the structure of language and the structure of the mind. I will continue to develop innovative experimental paradigms that address the nature of morphological and phonological representations, and attack the difficult questions in phonological research. My research will continue along three distinct lines.

First, I plan to continue my research on the nature of linguistic representations in adults and children. I will make use of a variety of experimental approaches, including artificial grammar learning, speech perception and perceptual adaptation. I plan to continue to compare children and adults (from infants to teenagers) in order to understand the developmental trajectory of language learning mechanisms. By understanding the similarities and differences in what children and adults are able to learn, we can better understand why children appear to be better language learners than adults.

Second, I will continue to develop computational methods as tools for connecting experimental and theoretical results. This type of work involves developing formal, axiomatic approaches to hypothesis testing, allowing for clear, concise links between data and theory. The use of computational tools to model and predict experimental data can provide further insights into the representations and mechanisms that underlie the learning process. I have begun work to use a computational model to explain recently collected data that shows that learners are better at parsing stems when the stems contain a frequent suffix.

Third, I will continue to use my experimental results to inform linguistic theory, and vice versa. I strongly believe that a research paradigm in linguistics is incomplete without a theoretical backbone. Because my experimental research is strongly grounded in linguistic theory, I am committed to providing major contributions to theoretical linguistics. My goal is to develop a theoretical paradigm that allows for the integration of experimental results within a formal, theoretical framework.

Through these lines of research, I have built a research program in which clear formal hypotheses about the representations of morphological and phonological processes can be tested using experimental, computational and theoretical methods. I believe that my research will lead to a better understanding of language and language development. To this end, I would be very interested in collaborating with various researchers on issues of language development and psycholinguistics.