(Image Source: University of Oxford)
Abstraction, Sound, and the Mind: Phonology, Processing, and Language Change
(Inspired by the work and ideas of Professor Aditi Lahiri)
Introduction
This post is written at the intersection of phonological theory, psycholinguistics, historical linguistics, and cognitive neuroscience. It is inspired by decades of scholarship exemplified by Professor Aditi Lahiri, whose work demonstrates that linguistic abstraction is not an intellectual luxury but a cognitive necessity. Across languages, across time, and across brains, language operates through representations that are simultaneously economical, flexible, and deeply structured.
Based on an interview transcript and the research work of the professor, this post develops a coherent research narrative: how sound systems are represented, how they are processed in real time, how they change historically, and how experimental and computational tools can illuminate, but never replace, linguistic theory.
The intended audience includes advanced undergraduate and graduate students, researchers across linguistic subfields, and scholars in cognitive science seeking a principled account of phonological abstraction.
1: Why Phonology Must Be Abstract
Phonology has long occupied an uneasy position within linguistics. Unlike syntax, it lacks overt symbolic markers. Unlike phonetics, it cannot be directly observed. Yet phonology is indispensable: without abstract representations, the variability of speech would overwhelm comprehension.
This section argues that abstraction is not an analytical imposition but a property of the language faculty itself. Speakers effortlessly normalize across accents, speaking rates, noise, and individual variation. Such normalization presupposes representations that transcend surface detail.
We distinguish three levels:
The acoustic signal
Phonetic implementation
Phonological representation
Conflating these levels leads to theoretical confusion and experimental misinterpretation. Phonology provides the mapping principles that allow unstable signals to yield stable meaning.
2: The Featurely Underspecified Lexicon (FUL)
2.1 Origins and Motivation
The Featurely Underspecified Lexicon (FUL) emerged from dissatisfaction with binary feature systems and hierarchical feature geometries. Earlier models assumed that lexical representations were maximally specified and that redundancy rules derived surface variation. Empirical evidence from perception and processing contradicted this assumption.
2.2 Core Principles of FUL
FUL rests on four foundational claims:
Features are monovalent, not binary.
Lexical representations are underspecified.
Underspecification is strong, not weak.
Consonants and vowels share feature inventories.
These principles eliminate arbitrary dependencies (e.g., coronal subfeatures) and align phonological theory with psycholinguistic evidence.
2.3 Strong Underspecification
In FUL, absence of a feature is informative. An underspecified lexical entry can match multiple surface realizations, whereas a specified feature can mismatch and block recognition. This asymmetry predicts experimentally observed patterns in lexical decision, mismatch negativity (MMN), and reaction time studies.
3: Mapping Signal to Lexicon
Speech perception is not pattern matching. It is inference under constraint.
FUL explicitly models the interaction between:
Continuous acoustic input
Discrete lexical representations
Language-specific feature inventories
Cross-linguistic evidence shows that listeners attend only to contrastive features in their language. Aspiration may be decisive in one system and irrelevant in another. Yet comprehension remains robust because abstraction filters the signal.
Multilingual speakers provide especially revealing evidence: the same acoustic cue can be interpreted differently depending on the activated phonological grammar.
4: Universals, Particulars, and Coronality
Some features recur across languages with remarkable stability. Coronality is one such case. While languages vary in how they deploy voicing, aspiration, or tone, coronal place consistently exhibits underspecification.
This section argues that phonological universals are best understood not as surface regularities but as constraints on representation. Universality emerges from cognitive economy rather than typological frequency alone.
5: Language Change and the Phonological Lexicon
Historical linguistics offers a natural laboratory for testing representational theories.
The persistence of irregular plurals such as mouse–mice and goose–geese is not accidental. These forms reflect earlier Germanic noun classes marked by vowel alternation rather than affixation. Their survival demonstrates that lexical representations can remain stable even as surface phonetics undergo radical change.
The Great Vowel Shift illustrates how phonetic drift can obscure morphological relations without eliminating them. Orthography fossilizes earlier stages, misleading modern intuitions about sound–meaning correspondences.
6: Studying Languages One Does Not Speak
Most linguistic knowledge depends on the systematic study of non-native languages. This chapter dismantles the myth that fluency is a prerequisite for insight.
Key methodological principles include:
Structural focus over intuition
IPA-based transcription
Awareness of perceptual bias
Triangulation through multiple speakers
Fieldwork on endangered languages underscores the ethical and scientific urgency of documentation. Linguistic diversity is not noise. It is evidence.
7: Speech Production and Phonological Planning
Speech is planned, not improvised.
Production data show that speakers organize utterances into phonological phrases that do not necessarily align with syntactic constituents. Function words cluster prosodically with content words, forming planning units that shape rhythm, stress, and reduction.
Second-language speakers often master syntax before mastering planning. The result is grammatically correct but prosodically unnatural speech. This gap reveals how deeply phonology structures fluency.
Loanword adaptation further demonstrates that phonological planning operates within native constraints, reshaping foreign material to fit familiar patterns.
8: Experimental Windows into the Mind
Psycholinguistics advances through controlled manipulation, not observation alone.
This section surveys key experimental paradigms:
Reaction time
MMN and EEG
Infant perception methods
Production latency
Experiments do not replace theory; they test it. The challenge lies in translating abstract linguistic hypotheses into measurable contrasts without trivializing them.
9: Phonology in the Brain
Neural evidence confirms what theory predicts: phonological processing unfolds rapidly and incrementally. EEG studies reveal sensitivity to feature mismatches within milliseconds, long before conscious awareness.
Crucially, the brain responds to abstract features, not raw acoustics. This supports the claim that phonology is neurally real.
10: Computational Models and Their Limits
Deep neural networks excel at pattern extraction but lack explanatory transparency. When used uncritically, they risk collapsing competence into performance.
This section argues for theory-guided computation. Models should test representational hypotheses, not replace them. Without linguistic abstraction, computational success remains descriptively impressive but scientifically shallow.
11: Integration Across Linguistic Domains
The boundaries between phonology, phonetics, psycholinguistics, and historical linguistics are methodological conveniences, not cognitive divisions.
A unified account of language must explain:
Stability amid variation
Change without collapse
Diversity without chaos
Abstraction is the connective tissue across these domains.
12: Language Diversity and the Future
Global media and educational homogenization threaten linguistic diversity. Yet diversity is not a peripheral concern; it is the empirical foundation of linguistic theory.
Future research must engage with under-documented languages, prosodic systems, and sociophonetic variation. The goal is not preservation alone, but understanding.
13: Academic Life and Research Craft
Successful research depends less on brilliance than on design.
This section offers practical guidance:
Formulating answerable questions
Reading historically, not just recently
Designing modular projects
Writing incrementally
Treating grants as research architectures, not lotteries
Luck matters but preparation determines whether luck can be used.
Conclusion: The Necessity of Phonological Thought
Phonology teaches a broader lesson about human cognition: complex systems require abstraction to function. Sound systems are not accidents of articulation; they are structured solutions to communicative and cognitive constraints.
Understanding language means understanding how the mind extracts order from variability. Phonology, in this sense, is not merely a subfield; it is a theory of intelligibility itself.
Interview with Professor Aditi Lahiri - Oxford University Linguistics Society
