2.1 Psychophysics, Sensory Thresholds, and Signal Detection Theory

2.1 Psychophysics, Sensory Thresholds, and Signal Detection Theory

Psychophysics represents the oldest subdiscipline of experimental psychology, formally inaugurated by Gustav Theodor Fechner in his 1860 treatise Elemente der Psychophysik. The core objective of psychophysics is to mathematically quantify the functional relationship between physical stimulus magnitude (objective environmental energy) and subjective psychological sensation (conscious internal experience).

---

1. Classical Sensory Thresholds and Experimental Methodologies

Classical psychophysics established two primary forms of sensory thresholds:

1. Absolute Threshold (stimulus limen): The minimum physical energy of a stimulus required for an observer to detect its presence above baseline noise. Because human sensory systems fluctuate, classical psychophysicists operationally defined the absolute threshold as the stimulus intensity detected on 50% of trials.
2. Difference Threshold (just noticeable difference, JND; difference limen): The minimum change in physical stimulus intensity required for an observer to distinguish between two stimuli 50% of the time.

To determine these thresholds empirically, Fechner developed three classical psychophysical methods:

---

2. Classical Psychophysical Laws: Weber, Fechner, and Stevens

Stimulus Scaling Evolution:
Weber's Law (ΔI / I = k) ──> Fechner's Law (S = k log I) ──> Stevens' Power Law (S = k · I^a)
[Proportional JND]          [Logarithmic Sensation]          [Power Function Scaling]

Weber's Law

Ernst Heinrich Weber (1795–1878) discovered that the just noticeable difference is not an absolute constant quantity, but rather a constant proportion of the initial stimulus magnitude. Expressed mathematically:

$\frac{\Delta I}{I} = k$

Where:

• $\Delta I$ = the difference threshold (JND).
• $I$ = the initial stimulus intensity (standard or baseline).
• $k$ = the Weber fraction (a constant specific to each sensory modality).

Fechner's Law

Fechner extended Weber's empirical finding by introducing a key theoretical assumption: all just noticeable differences are subjectively equal in psychological magnitude. By mathematically integrating Weber's formula, Fechner derived a logarithmic relationship between physical intensity and psychological sensation magnitude:

$S = k \log I$

Where $S$ is subjective sensation magnitude, $I$ is physical stimulus intensity, and $k$ is a scaling constant incorporating the Weber fraction. Fechner's Law describes logarithmic response compression: as physical intensity grows geometrically (multiplicatively), subjective sensation increases only arithmetically (additively). Under Fechner's law, every tenfold step in intensity adds the same fixed increment of sensation; Stevens later showed that for brightness (exponent about 0.33) a tenfold increase roughly doubles perceived magnitude.

Stevens' Power Law

In the 1950s, Stanley Smith Stevens challenged Fechner's assumption that JNDs represent equal psychological steps. Using the method of magnitude estimation—in which observers assign direct numerical values to perceived sensations—Stevens demonstrated that sensory scaling obeys a power function rather than a universal logarithmic curve:

$S = k I^a$

Where:

• $S$ = perceived psychological sensation magnitude.
• $I$ = physical stimulus intensity.
• $k$ = scaling constant depending on units of measurement.
• $a$ = modality-specific exponent determining the shape of the psychophysical curve.

Perceived Sensation (S)
    ▲
    │          / Response Expansion (a > 1, e.g., Electric Shock a ≈ 3.5)
    │         /
    │        /   / Apparent Line Length (a = 1.0, Linear Scaling)
    │       /   /
    │      /   /   ─── Response Compression (a < 1, e.g., Brightness a ≈ 0.33)
    │     /   /  ─
    │    /   / ─
    │   /  ──
    └──┴────────────────► Physical Intensity (I)

The exponent $a$ defines three fundamental sensory response profiles:

1. Response Compression ($a < 1$): Perceived sensation increases more slowly than physical intensity. Examples include visual brightness ($a \approx 0.33$) and auditory loudness ($a \approx 0.60$). This protects sensory systems from saturation over massive dynamic ranges.
2. Linear Scaling ($a = 1$): Perceived sensation increases in direct, 1:1 proportion to physical intensity. The quintessential example is perceived line length ($a \approx 1.0$).
3. Response Expansion ($a > 1$): Perceived sensation increases dramatically faster than physical intensity. The classic example is electric shock applied to fingertips ($a \approx 3.5$). A minor physical increment in shock voltage produces an agonizing escalation in perceived pain, serving an urgent evolutionary defense mechanism.

---

3. Signal Detection Theory (SDT)

Classical threshold methods suffered from a fatal theoretical flaw: they assumed a static physical boundary below which a stimulus is imperceptible and above which it is instantly detected. In reality, detection performance is heavily contaminated by internal neural noise, ambient environmental noise, motivation, fatigue, and decision criterion (response bias).

Developed by engineers and psychophysicists in the 1950s (e.g., Peterson, Birdsall, and Fox; Green and Swets), Signal Detection Theory (SDT) abandons the concept of a rigid absolute threshold. SDT asserts that sensory observation is an exercise in statistical decision-making under uncertainty.

The 2x2 Decision Matrix

In a standard SDT detection task, trials consist of either a Noise alone ($N$) event or a Signal-plus-Noise ($S+N$) event. The participant renders a binary choice ("Yes, signal present" or "No, signal absent"). This yields four distinct empirical outcomes:

Because the two response alternatives are mutually exclusive and exhaustive within each trial type, the probabilities are complementary:

$P(\text{Hit}) + P(\text{Miss}) = 1.0$

$P(\text{False Alarm}) + P(\text{Correct Rejection}) = 1.0$

Therefore, an observer's complete detection performance can be fully characterized using just two numbers: the Hit Rate ($H$) and the False Alarm Rate ($FA$).

---

4. Disentangling Sensitivity ($d'$) and Response Criterion ($c$, $\beta$)

SDT models neural activity along a continuous sensory evidence axis. Because baseline neural firing fluctuates continuously, presentation of Noise alone produces a normal distribution of sensory activity with mean $\mu_N$. When a faint signal is added, it shifts the distribution upward to a new normal distribution with mean $\mu_{S+N}$.

Neural Evidence Distributions and Decision Criterion:

      Noise Alone (N)         Signal + Noise (S+N)
         Distribution             Distribution
            ┌───┐                    ┌───┐
           ╱     ╲                  ╱     ╲
          ╱       ╲                ╱       ╲
         ╱         ╲   Criterion  ╱         ╲
        ╱           ╲     │(c)   ╱           ╲
       ╱             ╲    │     ╱             ╲
      ╱  Correct      ╲   │    ╱               ╲
     ╱  Rejection      ╲  │   ╱       Hit       ╲
    ╱                   ╲ │  ╱                   ╲
   ───────────────────────┼────────────────────────► Internal Sensory Evidence (x)
                          │
              False Alarm │ Miss
              ◄───────────┼──────────►
                  "No"    │   "Yes"
                          │
              ◄───────────d'──────────►

Sensitivity ($d'$, d-prime)

Sensitivity ($d'$) measures the observer's physiological capacity to distinguish signal from noise. Geometrically, $d'$ is the distance between the center of the Noise distribution and the center of the Signal-plus-Noise distribution, normalized by the standard deviation ($\sigma$):

$d' = \frac{\mu_{S+N} - \mu_N}{\sigma} = z(\text{Hit}) - z(\text{False Alarm})$

Where $z$ denotes the standard normal deviate ($z$-score corresponding to the cumulative probability).

• If $d' = 0$, the observer cannot discriminate signal from noise at all (performance is at pure chance level; Hit Rate equals False Alarm Rate).
• A larger $d'$ (e.g., $d' = 2.0$ or $3.0$) indicates excellent sensory acuity and wide distribution separation.
• Crucial Exam Rule: $d'$ is an intrinsic physiological/sensory parameter. It is determined solely by stimulus intensity, sensory apparatus integrity, and neural noise; it does not change when the observer shifts their willingness to guess.

Decision Criterion ($c$ and $\beta$)

The response criterion reflects the observer's cognitive decision rule—their threshold for responding "Yes":

1. Criterion Location ($c$): The distance of the decision boundary from the intersection of the two distributions, measured in standard deviations:

$c = -0.5 \left[ z(\text{Hit}) + z(\text{False Alarm}) \right]$

1. Likelihood Ratio ($\beta$, beta): The ratio of the height of the $S+N$ distribution to the height of the $N$ distribution at the criterion point:

$\beta = \frac{f(x \mid S+N)}{f(x \mid N)}$

Criterion strategies fall into three categories:

• Neutral Criterion ($c = 0$, $\beta = 1$): The decision cutoff sits exactly at the midpoint between the two distribution peaks. The observer is unbiased.
• Liberal Criterion ($c < 0$, $\beta < 1$): The cutoff shifts to the left. The observer requires minimal sensory evidence to say "Yes". This maximizes Hits, but causes a high False Alarm rate. This strategy is adaptive when the cost of a Miss is catastrophic (e.g., a radiologist screening for malignant tumors, or military radar monitoring for inbound missiles).
• Conservative Criterion ($c > 0$, $\beta > 1$): The cutoff shifts to the right. The observer requires overwhelming sensory evidence to say "Yes". This suppresses False Alarms, but causes a surge in Misses. This strategy is adaptive when false alarms carry punitive costs (e.g., a legal trial operating under "proof beyond a reasonable doubt").

---

5. Receiver Operating Characteristic (ROC) Curves

A Receiver Operating Characteristic (ROC) curve plots an observer's Hit Rate ($y$-axis) as a function of their False Alarm Rate ($x$-axis) across varying criterion settings.

Hit Rate (True Positives)
1.0 ┌───────────────────────┐
    │               . · ─── │  High Sensitivity (d' = 3)
    │           . ·         │
    │       . ·             │  Moderate Sensitivity (d' = 1.5)
    │    . ·                │
    │  .                    │
    │ ╱                     │  Chance Line (d' = 0, Hit = FA)
    │╱                      │
0.0 └───────────────────────┘
   0.0                     1.0  False Alarm Rate (False Positives)

Key Interpretive Rules for the GRE Psychology Subject Test

1. The Chance Diagonal: The diagonal line connecting $(0,0)$ to $(1,1)$ represents $d' = 0$ (zero discriminability). Along this line, $P(\text{Hit}) = P(\text{FA})$.
2. Moving ALONG a Single ROC Curve: Represents a change in response criterion ($c$ or $\beta$) while sensory sensitivity ($d'$) remains strictly constant. Moving toward the upper-right corner reflects a liberal criterion shift; moving toward the lower-left reflects a conservative criterion shift.
3. Moving to a HIGHER, Outward-Bowing Curve: Represents a true increase in sensory sensitivity ($d'$). The closer the apex of the curve approaches the top-left corner $(0, 1.0)$—where Hit Rate = 100% and False Alarm Rate = 0%—the greater the observer's discriminative ability.